Papers for

medical image analysts

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

AutoThreshold improves mask thresholding with joint confusion prediction

AutoThreshold: Joint Confusion-Profile Learning for Adaptive Mask Binarization

Abstract: A single validation-tuned threshold can become brittle when dense-score distributions shift. We introduce AutoThreshold, a post-hoc policy that predicts the joint mass of positive and negative pixels over score bins. This target directly parameterizes threshold-indexed confusion counts, avoiding products of independently estimated prevalence and class-conditional rates. Softmax normalization and cumulative decoding yield coherent, monotone confusion profiles; a plug-in decoder then selects a threshold for any $F_β$ objective without retraining the head. Across three independently trained Kvasir pipelines transferred to four polyp datasets, Joint exceeds continuous Otsu by $0.0080 \pm 0.0046$ external macro Dice, with positive gains in 3/3 runs. On the reference pipeline, three head seeds achieve $0.6324 \pm 0.0027$, 0.0085 above Otsu; direct-utility Ridge achieves 0.6342, with the paired difference interval spanning zero. The same Joint checkpoints outperform Otsu for $F_{0.5}$, Dice, and $F_2$. On developmental KITTI Road leave-one-corruption-family-out evaluation, Joint improves over scalar regression by 0.0082 [0.0064, 0.0101], while direct utility is 0.0022 higher. These results establish joint confusion prediction as a useful, metric-retargetable inductive bias, without claiming Bayes-optimal $F_β$ decisions.

Tue 10 NovComputer Vision and Pattern Recognition
The gist
Finding the right cutoff point to turn prediction scores into clear yes-or-no decisions can be tricky when the data changes. The authors introduce AutoThreshold, a method that predicts how many correct and incorrect pixels fall into score ranges together, instead of doing it separately. This approach makes it easier to choose a best threshold for different accuracy goals without retraining the model. Tests on medical and road datasets show AutoThreshold improves segmentation quality compared to common techniques.
Open → 2611.14279v1

Recurrent neural network improves 4D flow MRI spatial and temporal resolution

CURRENT: A Recurrent Neural Network for Concurrent Spatial and Temporal Super-Resolution of 4D Flow MRI

Abstract: Objective: To improve the spatial and temporal resolution of aortic 4D flow MRI through a unified deep learning framework for post-acquisition super-resolution (SR) of reconstructed velocity vector fields. Methods: We propose CURRENT, a recurrent neural network that performs spatial and temporal SR with a single network architecture. Supervised training was conducted using paired low- and high-resolution datasets generated by embedding computational fluid dynamics simulations of aortic flow into in vivo thoracic MRI backgrounds, followed by controlled spatial and temporal downsampling. CURRENT was evaluated for two-fold spatial and temporal SR against 4DFlowNet and for two-fold spatiotemporal SR against sequential spatial and temporal SR pipelines based on 4DFlowNet. Image-based and k-space-based quantitative metrics were considered and conventional interpolation were used as baselines. In addition, the impact of CURRENT's spatiotemporal SR on physics-informed relative pressure estimation was assessed. Finally, the ability of CURRENT to generalize from synthetic training data to in vivo acquisitions was investigated. Results: CURRENT achieved superior spatial SR performance and comparable temporal SR performance to dedicated 4DFlowNet models while providing a unified solution for both tasks. Simultaneous spatiotemporal SR using CURRENT yielded more accurate relative pressure estimates than either spatial or temporal SR alone and outperformed sequential 4DFlowNet-based spatial and temporal SR approaches. Furthermore, the proposed model demonstrated promising generalization to in vivo 4D flow MRI data despite being trained exclusively on synthetic datasets. Conclusion: CURRENT provides a unified framework for spatial and temporal SR of 4D flow MRI. The proposed approach has the potential to enhance physics-based analysis and broaden the applicability of accelerated 4D flow MRI.

Tue 10 NovComputer Vision and Pattern Recognition
The gist
4D flow MRI is a way to see how blood moves through the aorta, but the images can be blurry or slow. The authors developed CURRENT, a deep learning model that sharpens these images both in space and time using just one system. They trained CURRENT on computer-simulated data mixed with real MRI backgrounds, then tested it against other methods and showed it gives clearer images and more accurate pressure estimates in the heart. Importantly, even though it was trained on synthetic data, it works well on real patient scans.
Open → 2611.14058v1

Hierarchical method compresses brain scans while keeping detail

A Hierarchical Tokenization Framework for Voxel-Level fMRI Representation Learning

Abstract: We introduce HBAR (Hierarchical Brain Activity Representation), a hierarchical tokenization framework for learning compact, multi-scale representations of neuroimaging data. HBAR organizes voxel-level resting-state fMRI into a hierarchy spanning sub-parcel voxel groups, parcels, and large-scale functional networks, enabling up to 415x compression while preserving fine-grained spatial information beyond conventional parcellation. The framework supports fixed atlas-based, fully learned, and softly prior-guided hierarchies, allowing us to systematically study the role of structured priors in brain representation learning. On HCP-YA, at approximately 415x compression, HBAR achieves higher parcel and network correlations than rate-matched flat tokenizers, while at a lower 52x compression rate, shuffling the anatomical grouping reduces parcel correlation from 0.920 to 0.886. Demographic readouts with autoregressive training also reveal potential gains from the hierarchical structure. By combining sub-parcel inputs, compact discrete tokens, and multiscale brain structure, HBAR offers a practical framework for representation learning in high-dimensional neuroimaging data.

Sun 8 NovMachine Learning
The gist
Brain scans like fMRI generate enormous amounts of data, making it hard to analyze in detail. The authors developed a new method called HBAR that organizes brain data into layers from small groups of brain cells up to large brain networks. This approach compresses the data a lot—over 400 times smaller—without losing important detail. It also lets computers learn better brain patterns by using the brain’s natural structure. Their tests show HBAR works better than simpler compression methods and helps reveal information about people’s demographics.
Open → 2611.11081v1

Multi-modal image registration improved with shared structural feature learning

RegNC-Mamba3: Shared Structural Feature Learning Using Non-Causal State Space Principles for Multi-Modal Image Registration

Abstract: The main challenge of multi-modal image registration is shared structural feature extraction with higher accuracy and lower computational complexity. Compared with a State Space Model (SSM), the State Space Principles (SSP) formulation of Mamba-3 uses a more accurate discretization with lower truncation error while maintaining linear complexity. However, vanilla SSP is designed for natural language processing (NLP) and is not directly suitable for images. To address this issue, we propose a novel non-causal SSP (NC-SSP) with shared structure awareness and develop a coarse-to-fine registration method named RegNC-Mamba3. First, NC-SSP is designed to eliminate the dependence of the integral computation on the scanning path. By combining relative-weight normalization with aggregation over all circular scan paths, NC-SSP constructs a globally hidden state without long-range decay, which provides a wider receptive field and higher efficiency than the original Mamba-3. Second, NC-SSP is modified to strengthen the extraction of cross-modal structural features. When computing hidden states and feature interactions, we construct complex-valued sequences in NC-SSP by applying learnable Log-Gabor kernels, with different orientations to each head for frequency-domain modulation. This improves the perception of shared structural information across modalities and increases both registration accuracy and quality. Experiments on VIS-SAR and VIS-IR demonstrate that RegNC-Mamba3 outperforms the compared detector-based and semi-dense multi-modal registration methods.

Fri 6 NovComputer Vision and Pattern Recognition
The gist
Matching images from different sources, like visible light and infrared, is hard because their features look different. The authors created a new method that better picks out common shapes and structures in these images using a special mathematical approach called non-causal state space principles. This method looks at image data from many angles and frequencies to find shared details more accurately and quickly. Tests showed their approach registers images better than other methods.
Open → 2611.09336v1

Semantic visual memory improves few shot image segmentation accuracy

MESAM: Enhancing SAM2 for Few-Shot Segmentation through Semantic-Visual Memory Generation and Refinement

Abstract: This paper investigates how to leverage the powerful capabilities of the Segment Anything Model 2 (SAM2) to tackle few-shot segmentation (FSS) tasks. One major challenge lies in the substantial appearance differences between the support and query images, which may diminish the effectiveness of the memory-matching process in SAM2. Through systematic analysis, we find that the memory generated based on the categorical semantic information is robust to this appearance divergence and is more compatible with SAM2. Based on these observations, we propose a new Memory-Enhanced SAM2 (MESAM) method, which first introduces a Semantic-Visual Memory Generation (SVMG) module to separately construct a query semantic memory, a query foreground visual memory, and a support adaptive visual memory. Then, in our designed Interactive Semantic Memory Refinement (ISMR) module, we further enrich the obtained semantic memory with fine-grained visual details, and suppress its misatcivated background responses. Extensive experiments on the PASCAL-$5^i$, COCO-$20^i$, and LVIS-$92^i$ datasets demonstrate that MESAM consistently outperforms existing FSS methods, achieving new state-of-the-art segmentation performance. Our source code is available at: \href{https://github.com/tianyu-zou/MESAM}{https://github.com/tianyu-zou/MESAM}.

Fri 6 NovComputer Vision and Pattern Recognition
The gist
Segmenting objects in an image with only a few examples is hard because things can look very different between pictures. The authors studied how to use a tool called SAM2 better by creating a memory that combines both the meaning of an object and its visual details. They made a new system called MESAM that builds and refines this memory to help find objects more precisely, even with big appearance differences. Their experiments show MESAM works better than other methods on several standard image datasets.
Open → 2611.08943v1

Conditional mask diffusion tested for left atrial scar segmentation in MRI

CMDiff: Left Atrial Scar Segmentation from 3D LGE MRI Using Conditional Mask Diffusion Model: A Feasibility Study

Abstract: Late gadolinium enhancement (LGE) cardiac MRI enables non-invasive assessment of left atrial (LA) scar, an imaging marker associated with atrial fibrillation ablation outcome. However, automatic scar segmentation is hindered by low contrast, the thin atrial wall, scarce labels, and extreme class imbalance. Conditional diffusion offers iterative mask generation, but its suitability for sparse 3D LA scar segmentation remains unclear. To address this gap, we propose Conditional Mask Diffusion (CMDiff), which predicts clean masks using a dynamic-encoder-conditioned 3D sinusoidal UNet. We compared CMDiff with a deterministic baseline on identical cases in five-fold cross-validation of 60 scar-labelled public LASCARQS subjects, with 48 training and 12 held-out cases per fold. CMDiff achieved a scar DSC of 0.47 $\pm$ 0.08, HD95 of 12.89 $\pm$ 6.39 mm, and ASD of 1.91 $\pm$ 0.96 mm, compared with deterministic baseline values of 0.51 $\pm$ 0.08, 10.86 $\pm$ 5.39 mm, and 1.58 $\pm$ 1.03 mm, respectively. Under the same implementation, clean-mask prediction performed substantially better than the noise-prediction configuration evaluated in this study. Thus, CMDiff was technically feasible but did not improve segmentation performance: DSC was lower, and both HD95 and ASD were higher. Direct mask diffusion therefore provides an analytical alternative, but not a competitive performance advance for this task. The source code is publicly available at https://github.com/jizhang02/CMDiff.

Thu 5 NovComputer Vision and Pattern Recognition
The gist
Segmenting scars in the heart's left atrium using MRI images is hard because scars are small and hard to see clearly. The authors tested a new method called CMDiff that tries to improve this by generating clearer masks of scars in 3D images. They compared CMDiff to a traditional method and found it was possible to use CMDiff but it did not perform better. This means CMDiff is an alternative way to do the task, but it doesn't yet improve accuracy or reliability.
Open → 2611.07814v1

Framework improves pathology image quality filtering and segmentation accuracy

Agent-Driven Generative Priors for Pathology Image Synthesis and Segmentation

Abstract: Deep learning methods for pathology are constrained by the limited availability and high cost of large-scale pixel-level annotations. Synthetic data can alleviate this limitation, but existing pipelines often lack reliable quality control and remain weakly connected to downstream analysis. To address these issues, we propose a three-stage framework that integrates automated data screening, generative representation learning, and efficient adaptation to downstream tasks. First, we develop a pathology agent system to assess image quality and filter samples, resulting in TCGA-200K, a dataset containing 217,030 pathology patches selected from the public TCGA repository. Second, we train a generative model on the screened data. In addition to image synthesis, its backbone learns structural and semantic representations that can be transferred to downstream tasks. Third, we adapt the pretrained generative backbone to semantic segmentation using a lightweight task-specific head. We further introduce PathSeg-380K, a semantic segmentation dataset with expert annotations for approximately 380,000 cells across 12 cell types. Experiments on multiple pathology datasets show that the proposed framework achieves state-of-the-art segmentation performance. Our code is publicly available at https://github.com/script-Yang/Agent-GenPath.

Wed 4 NovComputer Vision and Pattern Recognition
The gist
Pathology images, which help doctors understand diseases, are hard to label in detail because it takes a lot of expert work. The authors built a system that first selects high-quality pathology image patches from a large public database, then trains a model to both generate similar images and learn useful features. Finally, they adapt this model to identify and label different types of cells in images, achieving very good results. They also created a large dataset with expert-marked cell types for this purpose.
Open → 2611.06701v1

Captor grounds text in images without needing text encoders

CAPTOR: text-encoder-free grounding by counting caption co-occurrence over the visual words of a frozen encoder

Abstract: Grounding a text query in an image needs a matrix scoring every term against every patch. Vision-language models compute it with an aligned text encoder, which many image encoders, such as histopathology-related models trained on slides alone, lack. The Caption-Anchored Patch-Term Occurrence Representation (CAPTOR) estimates the matrix without a text encoder. It clusters a frozen image encoder's patch tokens into visual words and describes each captioned image by its share of patches in each word. A term signature is the standardized difference of those shares between images whose captions mention the term and the rest. The matrix factorizes into a term-by-word table and a query-independent image factor, so an image is encoded once and a new term adds one table row. Against every training-free baseline, CAPTOR leads on Common Objects in Context (COCO), Pascal Visual Object Classes (VOC) and Cityscapes. On COCO, scored on the cells one object owns, it reaches 0.565 mean intersection over union (mIoU) against 0.502. On all three, it is also level with a text alignment trained on the same backbone, which needs gradients, a text encoder and several times the captions. On held-out breast-cancer slides, it exceeds the strongest baseline scoring patches with a pathology text encoder by 0.160 mIoU. Read through CAPTOR, UNI, an encoder trained on slides alone, exceeds such baselines by 0.132 on a common grid. Counting can therefore give a text interface to any image encoder trained without language, wherever captioned images exist.

Tue 3 NovComputer Vision and Pattern Recognition
The gist
Connecting words in a sentence to parts of an image usually needs both an image reader and a text reader trained to work together. The authors created CAPTOR, which allows finding where words belong in an image using only an image reader. It does this by grouping image parts into visual categories and counting how often words appear with these categories in image captions. This method works well even when standard text readers are not available or hard to train.
Open → 2611.05599v1

Compact network improves joint nuclei segmentation and classification accuracy

TME-ContextNet: Class-Conditioned Spatial Supervision and Grouped Decoding for Joint Nuclei Instance Segmentation and Classification

Abstract: Joint nuclei instance segmentation and classification requires precise boundaries and neighborhood information that helps distinguish morphologically similar cell types. We present TME-ContextNet, a multi-task framework that learns class-conditioned spatial context from existing nuclear annotations and organizes decoding into geometry-related and semantic task groups. Class-conditioned spatial supervision extends nuclear labels into a dense target using Euclidean distance transforms, requiring no additional tissue-region annotation. We derive its nearest-region interpretation and characterize its stability to small annotation perturbations. A local--global context module combines dilated convolution and self-attention, while a grouped task decoder reduces redundant computation. The spatial head supplies training supervision and is omitted during inference. Experiments on PanMix, CoNSeP, and PanNuke yield Dice scores of 80.23%, 65.32%, and 79.13%, and mean multi-class panoptic quality (mPQ+) scores of 67.62%, 35.50%, and 43.74%, respectively. PanNuke results are averaged across three test folds. Target and decoder controls support the combination of distance-based supervision and grouped decoding while identifying their dependence on task configuration. The framework uses 36.40 million parameters, compared with 54.74 million for HoVer-Net. These findings support a compact, context-aware formulation for joint nuclei analysis in computational pathology.

Mon 2 NovComputer Vision and Pattern Recognition
The gist
Distinguishing and classifying different cell nuclei in microscope images is important but challenging because many cells look alike and are tightly packed. The authors present TME-ContextNet, a new model that uses detailed spatial information based on distance maps from existing labels and groups decoding tasks to better separate and identify nuclei. Their method includes modules that combine local and global image context and reduces extra computations, resulting in good accuracy on several pathology datasets while using fewer parameters than previous models. This means it can analyze cells more efficiently and accurately without needing extra annotation work.
Open → 2611.02671v1

Gpu method speeds up shape analysis on large images by up to 100 times

CuPH: Parallel Cubical Persistent Homology via Level-Synchronous Component Tracking

Abstract: Persistent homology (PH) is a mathematically principled multi-scale topological descriptor of shape and connectivity in data. It is widely applied in image classification, segmentation, and shape analysis across medical and aerial imaging, providing structural information that pixel-wise deep learning may overlook. Standard cubical PH decomposes an image into cubical cells (vertices, edges, and squares) and reduces them sequentially in filtration order. This sequential reduction has $Θ(n)$ parallel depth for an $n$-pixel image, limiting within-image GPU parallelism and constraining its use in topology-aware deep learning. We introduce CuPH, a GPU-based parallel algorithm for computing cubical PH on 2D images that operates directly on pixels rather than cubical cells. CuPH buckets pixels by filtration values, activates them with a level-synchronous sweep, and tracks connected components with a lock-free union-find method. We prove that CuPH produces persistence diagrams identical to standard cubical PH on integer inputs, with a stability bound for floating-point inputs. CuPH reduces parallel depth from $Θ(n)$ to $O(L\log n)$ for an image with $L$ intensity levels. We implement the level sweep as a single CUDA kernel, eliminating cross-level kernel-launch overhead. Empirically, CuPH is $18.6$--$99.6\times$ faster than the existing GPU baseline (CMP). As a drop-in PH backend, CuPH accelerates topology-aware classification ($4.2$--$16.7\times$) and segmentation ($8.3$--$44\times$) while preserving accuracy. It also enables computing full-image PH losses on high-resolution images, improving segmentation Dice by $2.2$--$4.6$ points over patch-local PH baselines and accelerating training by $25$--$31\times$.

Sat 31 OctComputer Vision and Pattern RecognitionComputational GeometryDistributed, Parallel, and Cluster Computing
The gist
Analyzing the shapes and connections inside images helps computers better understand pictures, but traditional methods are slow and not very parallelizable on GPUs. The authors developed CuPH, a new way to do this shape analysis faster by working directly with pixel values and grouping pixels efficiently on GPU hardware. Their method produces the same results as the standard approach but runs much quicker, enabling better image classification and segmentation, especially on large images. This speed-up also helps train models faster while improving accuracy in medical and aerial image tasks.
Open → 2611.01078v1

Fisher contrastive learning improves leukemia diagnosis image analysis

Fisher Contrastive Learning to Mitigate Feature Suppression: A Case Study in Acute Promyelocytic Leukemia Diagnosis

Abstract: Self-supervised contrastive learning (SSCL) learns representations by contrasting augmented views of the same sample, but it can exhibit feature suppression, where dominant or easy-to-learn factors suppress other class-relevant signals, degrading the performance of representations, particularly in high-dimensional data with heterogeneous features. We introduce Fisher Contrastive Learning (FCL), which uses a regularized Fisher trace criterion to address feature suppression. FCL learns embeddings by maximizing the discriminative strength of Fisher discriminant functions, producing representations that preserve class-relevant information implied by the augmentation process. We prove that the representations learned by FCL could alleviate the feature suppression effects. We demonstrate FCL on simulated benchmarks designed to induce feature suppression and on a real-world biomedical application, pretraining on large white blood cell microscopy image dataset and transfer learning for diagnosis of acute promyelocytic leukemia in an external cohort. In all experiments, FCL outperforms the self-supervised baselines and yields more robust and discriminative representations.

Thu 29 OctMachine Learning
The gist
Machine learning can struggle when some features dominate and hide others important for identifying diseases. The authors introduce Fisher Contrastive Learning, a technique that helps the model focus on the right signals by balancing features better. They show this method works well on test data and improves diagnosis of acute promyelocytic leukemia from microscope images. This approach produces clearer and more useful representations of complex data.
Open → 2610.40162v1

Segmentation label downsampling keeps more regions with minimum loss

What Survives Label Downsampling? Exact Region Survival for Coarse Segmentation Targets

Abstract: Segmentation networks that predict at an output stride are trained against label maps downsampled by strided sampling, nearest-neighbor resampling or majority pooling. These methods label each coarse grid cell on its own, so a region that wins no cell is absent from the target and receives no positive label. We present MatchDown, which builds the coarse target from a minimum-cost matching between regions and the grid cells they occupy. It keeps as many regions as any one-region-per-cell target can, certifies how many losses are unavoidable, and with priority tiers (MatchDown-P) keeps every evaluated region (a region that a metric scores) of at least one cell's area. On COCO, Cityscapes and BBBC038 at factors 2 to 16, MatchDown-P loses exactly the minimum number of regions, whereas the downsampling used by StarDist, nnU-Net and connectomics pyramids loses four to five times as many on COCO at factor 8. Keeping a region in the target and reconstructing its shape are different problems: we show that no region lost by majority pooling can be reconstructed above IoU 0.5 by upsampling any coarse target, yet on BBBC038 a star-convex head trained on MatchDown targets segments 35% of the nuclei that majority pooling loses above IoU 0.5, against 15% when trained on majority targets and 24% on StarDist's strided target. On two nucleus datasets, with thresholds cross-validated for count error, this head has instance accuracy at IoU 0.5 (Acc50) 2.6 to 4.2 points above the strided target and at most 2.8 points below cell-center sampling, and a lower count error than the strided and cell-center targets on BBBC039. On BBBC039, an affinity head, which outputs one label per cell, lowers count error relative to majority targets by 9.3 nuclei per image at 7.2 points lower Acc50. MatchDown runs once, before training, in place of the label-downsampling step.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Training image segmentation models often requires reducing detailed labels to coarser versions, but this can lose important object regions. The authors introduce MatchDown, a method that pairs regions to coarse cells in a way that keeps as many regions as possible while minimizing unavoidable losses. Their approach improves how well segmentation models can identify and count objects, especially small or star-shaped ones, compared to standard downsampling. MatchDown performs this label downsampling before training and consistently preserves regions better across several datasets.
Open → 2610.39880v1

Cross-encoder method improves zero-shot detection of subtle image defects

Cross-Encoder Anchor Refinement for Zero-Shot Anomaly Detection

Abstract: Zero-shot anomaly detection aims to transfer anomaly discrimination from auxiliary training domains to unseen categories while localizing subtle defects. A key challenge is to construct discriminative references that combine contextual representations with complementary local evidence. We present Cross-Encoder Anchor Refinement (CEAR), which connects two visual encoding paths for the same image at the anchor stage, before normal and anomalous prototypes are constructed. An independent convolutional encoder pretrained for industrial product recognition builds an image-specific memory shared across selected vision transformer (ViT) layers. Spatial anchors read from ViT patch features query this memory through cross-attention and undergo residual refinement. Image-conditioned normal and anomalous queries then gate anchor aggregation to construct prototypes for anomaly scoring. Across 13 industrial and medical datasets, CEAR achieves the best performance in 57 of 68 evaluations spanning image-level detection and pixel-level localization. Notably, CEAR surpasses all other state-of-the-art methods in 41 of 44 pixel-level comparisons. Across 11 localization datasets, mean gains in pixel-level maximum F1 score and average precision are 4.45 and 3.79 percentage points, respectively. This study shows how cross-encoder interaction at the anchor stage can integrate complementary visual representations for fine-grained anomaly localization across domains.

Thu 29 OctComputer Vision and Pattern Recognition
The gist
Detecting unusual or defective parts in images without specific training on those defects is very hard. The authors propose a new approach called Cross-Encoder Anchor Refinement that combines two types of image understanding to better find and pinpoint tiny problems. They tested this method on many industrial and medical image sets and found it better than previous techniques, especially at precisely locating defects. This approach shows how linking different visual processes early on can help spot subtle anomalies in new situations.
Open → 2610.39727v1

Relative gradient operator improves edge detection under brightness changes

From the Superderivative to a Relative-Gradient Edge Operator

Abstract: First-derivative edge detectors such as Sobel, Prewitt, Roberts and Scharr measure absolute grey-level differences. We revisit the log-gradient magnitude, a relative-gradient edge operator, and motivate it by analogy with the superderivative, a multiplicative derivative in which the independent variable is perturbed multiplicatively rather than additively. In one and two dimensions the logarithm of the superderivative is a radial logarithmic derivative; its local, origin-free, isotropic analogue is the magnitude of the logarithmic gradient. The resulting operator measures relative rather than absolute change, giving an edge-strength map that is invariant under global intensity rescaling. We present the theory, show how the operator can be implemented in two steps---a classical gradient kernel plus a pointwise division---and compare it with standard first-derivative filters on synthetic images with ground-truth edges and on the official BSDS500 benchmark. We also study a parametric family that interpolates between the absolute gradient and the log-gradient; empirically, the best members are small intermediate exponents, which edge the classical kernels by a small but statistically detectable margin in the per-image metric and, with the same Gaussian smoothing as the single-$σ$ Canny, match it in ODS and lead it in OIS and AP, while the tuned Canny remains ahead in ODS. The single-threshold margin over the kernels does not replicate on the held-out test split, and the fully relative endpoint is never the best member, its ranking depending on the edge-map normalisation. The contribution is the construction and the invariance it buys, not a new state of the art.

Wed 28 OctComputer Vision and Pattern Recognition
The gist
Edge detectors usually look for absolute changes in brightness to find edges in images, but this can be affected if the whole image gets brighter or darker. The authors revisit an operator called the log-gradient which measures relative changes in brightness, making it more stable when the overall brightness changes. They explain the math behind it, show how to compute it efficiently, and compare its performance to classic edge detectors on benchmark images. Although the relative operator isn’t the absolute best in all tests, it offers an important property of brightness invariance and slightly improves detection in some settings.
Open → 2610.38530v1

Parallel algorithm speeds up contour extraction on multicore CPUs

TRUCO: A Scalable Lock-Free Algorithm for Parallel Contour Extraction

Abstract: Transforming binary masks into geometric contours remains a critical step in many computer vision pipelines, particularly for post-processing high-resolution segmentation outputs. Yet, standard algorithms such as the Suzuki-Abe method are limited by their sequential nature and the overhead of building topological hierarchies. As modern applications transition to high-resolution imaging, there is a critical need for contour extraction that leverages multi-core CPU architectures without the complexity of tile stitching or heavy synchronization. This paper presents TRUCO (Threaded Raster Unrestricted Contour Ownership), a lock-free parallel algorithm designed for high-performance geometric boundary extraction. TRUCO introduces a start-point ownership rule, combined with a speculative tracing model, in which threads trace contours downward into adjacent regions and use an upward-termination logic to ensure each contour is processed exactly once. Experimental results on synthetic and real-world datasets, conducted on both consumer and server-grade CPUs, demonstrate that TRUCO significantly outperforms existing baselines. The method achieves speedups of over an order of magnitude compared to the industry standard while maintaining robust scalability on high-resolution scenes across diverse hardware architectures.

Tue 27 OctComputer Vision and Pattern Recognition
The gist
Extracting outlines from images is an important step in understanding their content, but common methods work slowly because they process the image one step at a time. The authors created TRUCO, a new way to trace these outlines that lets many processors work together without getting in each other's way. This makes contour extraction much faster on modern computers with multiple cores, especially for very large images. The method ensures each outline is found exactly once by using a special ownership rule and smart thread coordination.
Open → 2610.37297v1

Semi supervised segmentation improves accuracy with dual branch voting

HDVS: semi-supervised semantic segmentation via heterogeneous dual-branch voting supervision

Abstract: Semi-supervised semantic segmentation, which aims to utilize large volumes of unlabeled images to achieve accurate segmentation results with fewer human annotations, has attracted increasing attention. Prior methods were primarily based on the assumption that the pseudo-labels generated by each branch were sufficiently reliable to supervise the remaining branches. However, there is a concern that the errors in pseudo-labels could accumulate during co-training, potentially resulting in suboptimal performance. To this end, we present a novel framework called heterogeneous dual-branch voting supervision (HDVS), which is designed to enhance the reliability of pseudo-labels and mitigate the issues arising from pseudo-labeling. Specifically, based on the cross-supervision framework, we introduce a voting mechanism that correlates pseudo-labels from heterogeneous branches to produce pseudo-labels with enhanced reliability. Concurrently, we employ a feature communication module to introduce perturbations at the feature level in each branch to maximize prediction diversity across the dual branches while maintaining convergence. Comprehensive evaluations of the proposed HDVS on two benchmark datasets (PASCAL VOC 2012 and Cityscapes) demonstrate its superiority to the state-of-the-art approaches.

Tue 27 OctComputer Vision and Pattern Recognition
The gist
Semantic segmentation means dividing images into meaningful parts, but it usually needs many labeled images which can be expensive. This paper presents a new way to do segmentation by using lots of unlabeled images plus a few labeled ones, improving the process with two different models that 'vote' to decide the best labels. This voting helps avoid mistakes that can happen when models learn from each other's errors. They also add a feature-level step to make the two models different yet consistent, helping the final result. Tests on standard image sets show this method works better than previous ones.
Open → 2610.37236v1

ReliMorph improves 3D deformable image registration with reliability guidance

Learning Where to Trust: Reliability-Guided Expert Routing and Velocity Gating for 3D Deformable Registration

Abstract: Deformable image registration relies on accurate anatomical correspondences. However, explicitly coupling local matching reliability with spatial context selection and deformation update strength remains insufficiently explored in learning-based registration. This challenge is particularly pronounced in homogeneous, weakly textured, and severely deformed regions, where ambiguous matches may induce erroneous deformation updates. We propose ReliMorph, a reliability-routed symmetric framework for 3D deformable registration that couples correspondence assessment with spatial context selection and update strength.Local correlation distributions characterize correspondence ambiguity using shared anatomical features with modality-specific adaptation. These reliability cues guide spatial expert fusion and gate residual velocities during coarse-to-fine refinement. Updated inverse-inconsistency feedback guides routing and gating at finer pyramid levels. To our knowledge, ReliMorph is the first 3D deformable registration framework to jointly condition spatial expert routing and residual velocity gating on correspondence reliability. Extensive experiments demonstrate that ReliMorph consistently outperforms competing methods in anatomical overlap, highlighting its effectiveness across unimodal and multimodal registration tasks.

Tue 27 OctArtificial IntelligenceComputer Vision and Pattern Recognition
The gist
Matching images that show soft tissues or organs in 3D is tricky because some parts don’t have clear details, making it hard to line them up correctly. The authors introduce ReliMorph, a system that assesses how reliable each local match is and uses that information to decide where and how to update the 3D alignment step-by-step. By focusing more on trustworthy areas and avoiding uncertain ones, ReliMorph reduces errors from ambiguous matches. Their tests show this approach works better than existing methods for various types of medical images.
Open → 2610.36672v1

Image based networks predict organ size and age from mri scans

Targeted Deep Regression - Initial Applications in Whole-Body Water-Fat MRI in UK Biobank

Abstract: UK Biobank is a British clinical study which, in addition to large amounts of medical data collected, plans to include detailed Magnetic Resonance Imaging (MRI) examina- tions from 100,000 participants. The image data can be analyzed for example by auto- mated analysis with neural networks for image segmentation or classification to expand the collection of image-derived phenotypes for medical research. In this work, we pro- pose, present and evaluate a methodology that uses image-based regression to estimate numerical properties from the collected neck-to-knee water-fat MRI scans. It applies a ResNet50-style network to 2D multi-channel collages that represent targeted and study- specific anatomical sub-regions from the MRI scan. The targeted sub-region sampling is achieved by propagating manually defined subregions from a reference scan to the other >30,000 analyzed scans using image registration. The resulting collages can be compact while retaining high-resolution, for example by providing multiple slices of kidney tissue for a kidney volume regression network. This approach also enables group-wise saliency analysis that highlight image and anatomical regions used by the network for the pre- dictions. The method was applied for the regression endpoints of kidney volume, heart volume, and age. Predictive performance was found between R2 = 0.731-0.984 and the group-wise saliency maps were found to highlight medically relevant regions. We con- clude that the method can be used to standardize and restrict the information presented to a deep regression analysis pipeline, increasing information density and opening up a wide range of future medical inference studies.

Mon 26 OctComputer Vision and Pattern RecognitionMachine Learning
The gist
Estimating specific body measurements from MRI scans can help in medical research and diagnosis. The authors developed a way to focus neural networks on small, important areas of MRI images to predict things like kidney size or heart volume. They use special image collages from targeted body parts which keep details but reduce data size. Their method showed accurate predictions and revealed which image areas mattered most. This approach could improve how computers analyze medical images and extract useful numbers from them.
Open → 2610.35339v1

Ki67 cell typing from images using pretrained embeddings and clustering

K-CLUE: A Segmentation-to-Clustering Framework for Ki67 Cell Typing

Abstract: Ki67 is a key marker of tumor proliferation in breast cancer, but its quantification from immunohistochemistry images normally depends on supervised classifiers trained per dataset. We examine the role of frozen pretrained visual representations in annotation-free Ki67 cell typing. To this end, we use K-CLUE, a framework that segments individual nuclei, represents each nucleus with a pretrained encoder, reduces the resulting embeddings with principal component analysis, and separates Ki67-positive from Ki67-negative nuclei using k-means clustering, without task-specific training or manual annotations. Clustering the raw components is highly sensitive to the choice of encoder: across seven encoders, the Ki67-positive F1-score on matched MIST nuclei varies by 44 points, although Ki67 status remains linearly decodable from every representation, with supervised probe AUCs above 0.98. Analysis of the embedding spaces shows that clustering succeeds when DAB-related variation occupies the leading directions of variance. Whitening the retained principal components removes this dependence, reducing the spread across encoders to 1.00 F1-score point. Handcrafted color representations are also competitive, confirming that Ki67 status is strongly reflected in chromatic information, but they require an explicit choice of color space. In end-to-end evaluation, K-CLUE reaches F1-scores of 85.00% and 81.91% for Ki67-positive and Ki67-negative nuclei on MIST, and 79.59% and 81.16% on BCData, outperforming DeepLIIF and IHCScoreGAN on both datasets. These results show that pretrained representations can support annotation-free Ki67 typing when their variance structure is accounted for.

Fri 23 OctComputer Vision and Pattern Recognition
The gist
Quantifying Ki67, a protein marker for tumor growth in breast cancer, usually requires training on labeled images for each dataset. The authors present K-CLUE, a method that identifies and classifies cells stained for Ki67 without needing any manual labeling or retraining. It first segments nuclei in images, extracts features using a pretrained model, then groups the cells using clustering to distinguish Ki67-positive from negative. Their study shows that by adjusting how features are processed, this clustering can be made reliable across different pretrained models, and their method performs well on real breast cancer image datasets.
Open → 2610.32493v1

Flow matching completes hidden curvy objects in images accurately

ROCOCO: Robust Outline Completion of Occluded Curvilinear Objects

Abstract: We propose a learning-based generative model for completing curvilinear features under occlusion by environmental obstacles. To this end, we adopt a flow matching-based approach that supports the inpainting of occluded regions with the appropriate curvilinear features. To eliminate the reliance on privileged information about the occlusion a priori, we also train a UNet-based mask prediction model, which can identify regions for inpainting given only the occluded observation. The components of our approach are validated via experiments using procedurally-generated test cases, and the complete pipeline is validated using large public datasets for curvilinear objects. Against off-the-shelf inpainting baselines, ROCOCO is the strongest method by a wide margin when no occlusion mask is given, and is competitive with mask-supervised inpainting when one is, achieving best on full-image error and second only to LaMa on the occluded region, while using a completer of $7.7$M parameters against LaMa's $51$M and Stable Diffusion's $860$M.

Wed 21 OctComputer Vision and Pattern RecognitionMachine Learning
The gist
Sometimes curved lines or shapes in pictures are partially blocked by other things, making it hard to see the full shape. The authors developed a way for computers to predict and fill in the missing parts of these curved shapes without needing prior knowledge of where the blockage is. Their method uses a special learning technique that looks at the visible parts and guesses the hidden sections accurately. Tests show their approach works better than many existing methods, especially when no information about the blocked areas is given.
Open → 2610.29177v1

Surface matching method separates shape shifts from rigid motions

Surface matching for image-derived geometries with a projection-based separation of rigid body modes

Abstract: The large diffeomorphic deformation metric mapping (LDDMM) framework offers numerous techniques for analyzing shapes and images. Surface matching computes a non-rigid, smooth, and invertible mapping that registers one surface to another. The computed deformation may contain rigid body motion, which frequently does not represent meaningful shape differences but is due to external factors like imaging modalities, such as the object's positioning during image acquisition. We present a simple extension to the LDDMM framework's surface matching that uniquely decomposes the computed deformation into rigid body modes (RBM) and RBM-free motion by introducing a suitable projection. This enables a focused analysis of each component individually and additionally helps to preserve the surface discretization quality. Three sets of examples illustrate our approach. A small-scale example demonstrates the proposed method's effects, followed by an example using a single pair of heart geometries, and finally an example involving a cohort of 98 heart shapes.

Mon 19 OctComputational Engineering, Finance, and Science
The gist
When comparing shapes from images, sometimes the shapes appear different because of how they were positioned or moved, not because their actual forms changed. The authors improve an existing mathematical method so it can tell the difference between true shape changes and simple movements like rotation or sliding. This helps in focusing analysis on meaningful shape differences and keeps the shape details clear. They show how this works using examples with heart shapes from medical images.
Open → 2610.26238v1

Glow builds brain regions linked to variables via improved clustering

GLOW: General Linear Models Optimized with Ward's Method

Abstract: Neuroimaging seeks brain regions whose imaging features correlate with variables of interest (e.g., diagnosis, age), usually via the General Linear Model (GLM). Region-Based Analysis dilutes or splits any effect that does not fill exactly one atlas region. Voxel-Based Analysis, with Cluster Extent Thresholding and Threshold-Free Cluster Enhancement built on it, fits a GLM per voxel and groups adjacent significant voxels. But neighboring voxels can share a statistic without sharing a GLM, which a per-voxel statistic cannot detect. GLOW (General Linear Models Optimized with Ward's method) instead builds a hierarchy of nested regions by greedily minimizing GLM error (provably a special case of Ward's clustering). Maximum-statistic (Westfall-Young) permutation controls family-wise error over all regions. Significance spreads to an effect's nested fragments and the regions containing it; a final pruning step seeks whole, pure effect extents. We plant synthetic effects of known extent and varying strength in Human Connectome Project (HCP) diffusion images and in white noise. Pruning delivers wholeness: GLOW returns a spatially consistent effect as one region on 57% of HCP trials, the best baseline on 35%; baselines recover comparable volume, but in many pure pieces. The cost is positive predictive value (PPV): GLOW is more sensitive (by up to 0.29), but its regions overrun the effect, mostly because no hierarchy region isolates it, not because of pruning. On HCP, GLOW's mean Dice gains 0.059 over the best baseline at weaker effects and gives back 0.005 at stronger ones. On white noise, GLOW trails by up to 0.61 Dice: its hierarchy holds the effect, but its permutation test does not certify it. Runtime grows roughly linearly with voxel count, about 4.5 times the voxel-wise methods' on the full HCP volume. The PPV loss may trace to the assumption of spatially independent residuals.

Fri 16 OctMachine Learning
The gist
Finding brain areas connected to specific traits often uses simple models per tiny spots in brain images, which can miss broader patterns. The authors propose GLOW, which groups nearby brain spots by how well linear models fit, making regions that better capture effects. Tested on brain scans and fake signals, GLOW finds more complete areas but may include extra nearby parts. It runs slower but shows promise in detecting linked brain regions more reliably than some existing methods.
Open → 2610.24085v1

Optimal schedules improve labeling decisions for machine learning tasks

Towards optimal selection schedules for online active learning

Abstract: We consider the problem of online linear classification, where a stream of unlabeled samples is available, and annotation is costly. We investigate the optimal query policy, that prescribes whether and when a sample should be labeled and used for a training step, or instead discarded, so as to minimize a given objective. We focus on the realizable case, under Gaussian design, in the high-dimensional regime where data dimension and number of unlabeled samples grow proportionally. We consider two objectives: 1) the generalization error under a labeling budget, and 2) a cumulative cost function where mistakes have unit cost and queries cost $Δ$. We first show that the myopic policies that greedily minimize at all times the increment in the objective correspond to either small or intermediate-margin policies, that prescribe the selection of samples whose margins fall within certain intervals that shrink with time. Within this class, we numerically ascertain the optimal schedules for the selection thresholds that minimize the final values of the objectives. Taking inspiration from the myopic policy for the cumulative cost, we design a very simple query policy schedule that we compare to other proposed heuristics, on synthetic data and real medical image datasets. Finally, we discuss myopic policies for two-layer neural networks, and exemplify how naive margin-based policies can cease to be optimal.

Fri 16 OctMachine Learning
The gist
Labeling data can be expensive and time-consuming, especially when machines learn from a continuous stream of information. The authors studied how to best decide which data points to label to improve learning accuracy while limiting costs. They found that choosing samples with certain confidence levels—called margins—in a particular way can optimize learning outcomes. They tested their method on synthetic and medical image data and also explored how it applies to neural networks, showing that simple approaches might not always work best.
Open → 2610.23477v1

Tissue graph encoder improves medical image analysis accuracy and efficiency

TAN: A Size- and Form-Adaptive Masked-Convolution Encoder for Tissue Graphs

Abstract: Tissue graphs represent histology images as graphs over tissue-conforming superpixels, which keeps homogeneous tissue regions intact instead of cutting them into rectangular patches. To embed superpixels, current encoders crop them to bounding boxes and then pad, resize, or mask them, so background or fill values can leak into the node features. This matters in practice, and especially in medical imaging, where a prediction should depend on the tissue and not on an arbitrary preprocessing choice: on BACH, changing only the fill of the masked region shifts the accuracy of frozen encoders by up to 6 percentage points, larger than the differences among the four best encoders under mean fill, and the preferred fill differs between datasets, so it cannot be fixed once and for all. We present TAN (Tissue-Aware Network), a compact convolutional encoder (0.41M parameters) that processes each superpixel at its native shape and extent and excludes background from every operation, which makes its embeddings exactly invariant to the fill by construction. Under identical masks and an identical graph classifier, TAN's accuracy on BACH is not significantly different from that of the 303M-parameter foundation model UNI (81.1% vs. 80.7-82.6%, depending on the fill), with ~740x fewer parameters, ~1,190x fewer FLOPs, and no pretraining. On BRACS, TAN is close to a ResNet18 trained under the same protocol and trails the histopathology foundation models. An ablation with TAN's architecture but standard processing of fill-padded crops shows that masking keeps accuracy similar (nominally higher on BACH, lower on BRACS, within noise) at less than half the FLOPs.

Fri 16 OctComputer Vision and Pattern Recognition
The gist
Analyzing tissue images is easier when the regions are kept whole, not chopped into squares. The authors show that current methods can accidentally include background information that changes results depending on arbitrary settings. They introduce TAN, a new small computer model that looks at each tissue region exactly as it is, ignoring surrounding space, which stops background from confusing the analysis. TAN works as well as much bigger models but uses far less computing power and no pretraining.
Open → 2610.23475v1

Hierarchical prompt learning improves zero-shot anomaly detection and localization

Knowledge-Guided Distributional Prompt Learning with Visual Hierarchy for Zero-Shot Anomaly Detection

Abstract: Zero-shot anomaly detection (ZSAD) aims to detect anomalies in previously unseen classes using auxiliary images, with applications such as industrial visual inspection and medical imaging. Recent vision-language-model-based approaches improve ZSAD by aligning textual and visual representations, while distributional prompting further models prompt variability for unseen categories. However, existing methods do not explicitly combine normality-referenced distributional modeling with hierarchy-specific prompt specialization, limiting their ability to model anomalous semantics relative to normality while addressing both image-level detection and fine-grained localization. We propose Hierarchical Knowledge-Guided Distributional Prompt Learning (HiDist), a normality-referenced hierarchical distributional prompt framework for ZSAD. HiDist models each prompt token with a diagonal Gaussian distribution and generates stochastic prompt variations through Monte Carlo sampling. Rather than learning normal and anomalous distributions independently, anomalous prompt distributions are parameterized as deviations from learned normal references, explicitly encoding their dependency. The resulting distributions are specialized into global and local representations for image-level detection and pixel-level localization, respectively, while an adapted global visual representation guides hierarchical alignment. Experiments on seven industrial and medical datasets demonstrate strong cross-dataset performance across anomaly detection and localization tasks. Extensive ablation and representation analyses further support the effectiveness of the proposed distributional modeling and hierarchy-specific representation learning.

Fri 16 OctComputer Vision and Pattern Recognition
The gist
Detecting unusual or faulty images in new categories without prior examples is hard. The authors developed a method that learns what "normal" looks like and then identifies anomalies by comparing differences, using a special way to handle prompts that guide the model. Their approach also looks at both the whole image and specific parts to find anomalies more precisely. Tests on various industrial and medical images show it works well across different cases.
Open → 2610.23172v1

Improving segmentation of thin shapes by adaptive connection fixes

TRACE: Topology-Aware Reconnection via Prediction-Adaptive Connectivity Enforcement for Thin-Structure Segmentation

Abstract: Thin, elongated structures are particularly sensitive to connectivity errors, as even a few local segmentation mistakes can disrupt otherwise continuous paths. Existing skeleton-based objectives typically supervise fixed centerline locations and do not account for alternative local routes that may restore connectivity. We propose TRACE, a topology-aware, prediction-adaptive objective that uses a target skeleton to define required local connections. For each connection, TRACE identifies the easiest valid reconnection under the current prediction and jointly supervises its bottleneck with the corresponding skeleton point. The supervision thus adapts to the evolving prediction rather than remaining fixed to the centerline. We further introduce a topology-aware skeleton carrier that preserves both foreground and class-specific topology, enabling reliable definition of these connections. Extensive experiments across six 2D and 3D benchmarks show that TRACE improves connectivity and topology while maintaining strong segmentation accuracy with low training overhead. Code will be released upon acceptance.

Thu 15 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Thin and long structures in images are hard to correctly outline because small mistakes can break their shape's continuity. The authors developed a method called TRACE that looks for easier ways to reconnect broken parts based on what the computer currently predicts. Instead of focusing only on a fixed path, TRACE adapts to these changing predictions to keep the shapes connected. This approach leads to better consistency in segmenting thin structures without extra training cost.
Open → 2610.22210v1

Prostate MRI model improves tasks through continual pretraining

Continual pretraining Prostate MRI ViT Encoder for task alignment

Abstract: Building foundation models for MRI has emerged as a promising direction for advancing medical image analysis and improving accessibility of task specific models. Existing approaches primarily focus on effective pretraining strategies and evaluating their performance across diverse downstream tasks. While these methods show performance gains, limited attention has been given to understanding task alignment between pretraining and downstream objectives. Unlike language models, where the two stages are naturally aligned, such alignment is less evident in vision based applications. Motivated by this gap, we explore continual pretraining. Specifically, continually pretraining a pretrained ViT encoder using Low-Rank Adaptation (LoRA) on a prostate segmentation task to enhance task alignment and downstream performance. We further perform layer-wise representational analysis using Centered Kernel Alignment (CKA) to assess representational shifts and their correlation with task outcomes. Our experiments show improved performance in 3 of 6 downstream tasks and comparable results in the others. Our Feature analysis reveals that, for segmentation tasks, continually pretrained encoder features exhibit greater similarity to those of the finetuned encoder than the original pretrained encoder, correlating with observed performance gains.

Wed 14 OctComputer Vision and Pattern RecognitionMachine Learning
The gist
Medical images like prostate MRIs need computer programs that can understand them for different tasks, like finding organs or abnormalities. The paper’s authors discovered that continuing to train a vision model pretrained on general images specifically on prostate MRI data helped the model better prepare for certain tasks. This continual training aligned the model’s internal understanding with the needs of segmentation tasks, improving performance on some tests. They also analyzed how the model’s internal features changed during this process to explain why the improvements happened.
Open → 2610.20217v1

Robust image prototype discovery using student’s t mixture models

Pro-t-VAE: Closed-Form Student's t Mixtures for Robust Prototype Discovery

Abstract: We present a robust deep generative framework for unsupervised image prototype discovery, built on a heavy-tailed Student's t mixture prior, that bridges statistical robustness and mathematical tractability. Existing generative approaches typically rely on noise-prone Gaussian priors or computationally intensive diffusion models; neither simultaneously provides explicit latent structure nor exact analytic optimization. While diffusion models can yield strong priors, iterative sampling and intractable likelihoods limit scalability and interpretability for prototype discovery. We address these issues by constructing a Student's t latent space via a Bayesian scale-mixture formulation. This yields a fully closed-form augmented ELBO for unsupervised Student's t mixture variational autoencoders, which, to our knowledge, is the first such objective applied to robust image prototype discovery, avoiding biased variational approximations and high computational costs in prior heavy-tailed or diffusion-based frameworks. Treating precision as a latent variable automatically modulates each data point's influence, downweighting outliers to learn a data-dependent measure of trust. The result is improved robustness, accuracy, and efficiency, with well-separated latent representations and more stable manifold learning than state-of-the-art methods. We validate our model on standard vision benchmarks and challenging neuroimaging tasks. In brain MRI-driven neurodegenerative disease analysis, our model recovers clean, anatomically coherent prototypes where clinical noise often collapses rare subtypes. These results highlight the ability of our framework to preserve anatomy-associated patterns for precise discovery in clinical settings. Our code is available at: https://github.com/xuzhikangnba/torch-tvae.

Sat 10 OctMachine Learning
The gist
Images often have noisy or unusual parts that make it hard for computers to find common patterns. This paper presents a new method that uses a special mathematical approach involving Student’s t distributions to better handle these noisy parts. The authors show their method makes the computer find clearer, more accurate prototypes or typical examples in images without getting confused by noise or outliers. They tested it especially on brain MRI images, where it helped identify meaningful patterns in diseases more reliably than older methods.
Open → 2610.14894v1

Sciencebuddy-pv audits imaging model for vascular intervention research

ScienceBuddy-PV: A Verifier-Governed Recursive Improvement Protocol for Panvascular Intervention Research

Abstract: ScienceBuddy-PV adapts the ScienceBuddy recursive-in-recursive design to panvascular intervention research through a structured evidence ledger, device-vessel ontology, abstention contract and retained failure bank. Recursive harness optimization and outer-loop model learning, including reinforcement learning, are specified but not executed. We audit a fixed component, rather than the complete agent system, using separate ARCADE syntax and stenosis tasks (1,000 training, 200 validation and 300 test images per task). A shared task-masked convolutional model was trained once per representation setting; its parameters and the heuristic support gate were then held fixed, with the stenosis threshold selected on validation data. On the locked test sets, syntax Dice was 0.475 (95% CI 0.459-0.490) and stenosis Dice was 0.231 (95% CI 0.214-0.249). Stenosis pixel AUROC was 0.957 and AUPRC was 0.180 at prevalence 0.00975. The gate accepted 76.7% of stenosis-frame outputs versus 100% for a score-only comparator. In a constructed mask audit, it accepted 44.3% of reference candidates and rejected 55.7%, exposing the cost of abstention. Acceptance does not establish correctness or clinical safety. This public-data component audit makes support and abstention measurable and provides a reproducible protocol for future collaborative agent studies. It does not demonstrate recursive improvement, patient-level validity or clinical device recommendations.

Sat 10 OctArtificial IntelligenceComputer Vision and Pattern Recognition
The gist
Panvascular interventions are medical procedures on blood vessels. The authors adapted an AI system called ScienceBuddy to help analyze imaging data for these procedures by creating a special protocol to verify parts of the AI’s performance. They tested a specific model on images showing vessel narrowing and measured how well it marked problem areas, including its ability to say when it was unsure. This work provides a way to measure and reproduce support and hesitation in AI models, but does not claim it improves itself or directly helps patients yet.
Open → 2610.14523v1

Gpu-based method speeds up topological analysis of images dramatically

GPU-Accelerated Computation of Persistent Homology for Topological Analysis of Image Data

Abstract: In recent years, persistent homology has seen rapid adoption in deep learning, yet its computation remains a major bottleneck in network training. This paper introduces TopoGPU, a GPU streaming pipeline that computes persistence diagrams of cubical complexes induced by 2D and 3D images. TopoGPU streams the input image chunk by chunk, processing each chunk with massively parallel GPU kernels on a grid of GPU blocks; the resulting boundary relations are accumulated in host memory, where the CPU performs the boundary matrix reduction. TopoGPU introduces a stratification-aware discrete Morse matching that provably preserves persistent homology under streaming, together with a parallel topological sorting algorithm and a parallel V-path parity algorithm for deriving Morse boundaries on the GPU. TopoGPU outperforms Cubical Ripser, a state-of-the-art method for persistent homology computation, on every benchmark evaluated, achieving an average end-to-end speedup of 53.24x and a maximum of 198.01x. We further integrate TopoGPU into a topology-preserving deep network, demonstrating that it substantially reduces the cost of persistent homology computation during network training. TopoGPU is open source, with pre-built binaries, Google Colab notebooks, and Docker images available at the project's GitHub page: https://github.com/seravee08/GPU-Computation-of-Persistent-Homology-for-Image-Data.

Wed 7 OctComputer Vision and Pattern Recognition
The gist
Topological analysis helps computers understand the shape and structure of data, but it can be very slow, especially for image data used in deep learning. The authors present TopoGPU, a system that uses graphics processors (GPUs) to speed up these calculations by breaking images into chunks and processing them in parallel. Their approach carefully preserves the important topological features during streaming, resulting in much faster computation than previous tools. This makes it easier and faster to include topological analysis in training deep neural networks.
Open → 2610.10959v1

Efficient algorithms for watershed image segmentation on graphs

Playing with Kruskal: algorithms for flat and hierarchical watershed cuts

Abstract: In the framework of edge-weighted graphs, watersheds have proven to be linked to well-known optimization problems, as Minimum Spanning Tree, which allowed the design of efficient algorithms for computing (hierarchical) watershed segmentations. In the present article, after reviewing the literature related to watershed segmentation, we present a detailed end-to-end pipeline of algorithms to compute (hierarchical) watershed segmentations, starting from the computation of graph-based image representations, up to the computation of connected components of the final (hierarchical) segmentation. We consider the several variations of watersheds, including their supervised and unsupervised versions, and the various ways of computing seeds, to name a few. For the first time, we bring together all these watershed notions and algorithms in a compact and understandable way. We aim at providing a reference for those interested in employing and reimplementing the watershed segmentation framework for their task at hand.

Wed 7 OctComputer Vision and Pattern Recognition
The gist
Segmenting images into meaningful parts is important for many uses, but it can be complex. The authors explain how watershed methods, which are connected to well-known graph algorithms like minimum spanning trees, can help break images into segments reliably. They bring together various ways to compute these watershed segmentations, including hierarchical and supervised versions, in one clear guide. This makes it easier for people building image analysis tools to understand and use these techniques.
Open → 2610.10012v1

Medical image anomaly detection improved with quasi-binarized autoencoders

Quasi-Binarized Autoencoders: An Architecture-Independent Information Bottleneck for Medical Image Anomaly Detection

Abstract: Unsupervised anomaly detection, which learns only from normal images, is a central task in medical image analysis and remains an open problem. Reconstruction-based methods pass an image through an encoder-decoder network trained on normal data and detect anomalies from the residual between the image and its reconstruction. This works only if the information passed from the encoder to the decoder is limited; otherwise the network learns an identity mapping and reconstructs anomalies too. This limit is usually imposed through architectural choices, tuned per dataset, that cannot be stated in bits. We introduce the quasi-binarizing (QB) layer, which squashes each latent element into [0, 1] and adds Laplace noise of scale 1/epsilon. Each element is then epsilon-locally differentially private, and the mutual information between an image and its reconstruction is bounded by a quantity that depends only on epsilon and the number of QB elements, whatever the encoder and decoder. Placing a QB layer on every encoder-decoder path, including all skip connections, we build QBAE, a seven-level attention U-Net with 32,768 QB elements. On the seven datasets of the MedIAnomaly benchmark, QBAE with one architecture and one configuration reaches a mean image-level AUROC of 0.828, the highest among methods that do not adapt to each dataset, and the best reported results on BraTS2021 (AUROC 0.911, pixel-level AP 0.838). The noise is kept at test time, so that every reconstruction satisfies the bound. Without input corruption, the bottleneck alone prevents identity collapse (mean AUROC 0.805 vs. 0.590). Code is available at https://github.com/hanaokalog/MedIAnomalyQB.

Wed 7 OctComputer Vision and Pattern Recognition
The gist
Detecting unusual features in medical images is important but hard when only normal images are available for training. The authors introduce a method called quasi-binarized autoencoders that limits how much information passes through the network, preventing it from just copying the image and missing anomalies. Their approach uses a special layer that controls information flow mathematically and adds noise to improve detection. Tested on multiple medical datasets, this method performed better than others that use a fixed design for all data.
Open → 2610.09670v1

Brain MRI inpainting with zero-shot 2.5D flow models

Zero-Shot Brain MRI Inpainting with 2.5D Unconditional Flow Priors

Abstract: Generative inpainting of brain MRI volumes is essential for synthesizing healthy tissue in pathological regions, improving the accuracy and reliability of automated downstream brain analysis applications such as image registration, brain extraction, and segmentation. However, standard 3D approaches are computationally prohibitive, while efficient 2D slice-wise methods suffer from severe inter-slice discontinuities. Furthermore, traditional models rely on conditional training, requiring task-specific learning of masked inputs. We propose a zero-shot brain MRI inpainting framework utilizing 2.5D unconditional flow priors to capture spatial context along the superior-inferior axis without the overhead of full 3D convolutions. During training, our flow matching model learns the joint distribution of adjacent axial slice triplets, modeling the manifold of healthy brain anatomy while explicitly excluding pathological regions from the loss function. At inference, the model processes the input triplets autoregressively along the depth axis. We employ the Restora-Flow solver to constrain the unconditional prior using the input mask, achieving accurate zero-shot inpainting. Evaluations show our 2.5D strategy resolves the structural discontinuities of 2D baselines, synthesizing plausible healthy tissue while maintaining volumetric consistency across the axial, sagittal, and coronal planes. As a final step, we generate and average an ensemble of multiple stochastic reconstructions to form the final prediction. Quantitative results benchmarked on the official BraTS 2026 Inpainting Challenge validation set demonstrate the effectiveness of our proposed approach, yielding an SSIM of 0.816 $\pm$ 0.112, MSE of 0.007 $\pm$ 0.005, and PSNR of 22.923 $\pm$ 4.343. Code is available at https://github.com/imigraz/brats2026-inpainting.

Tue 6 OctComputer Vision and Pattern Recognition
The gist
The paper addresses the problem of filling in missing or damaged areas in brain MRI scans to help with analyzing brain images more accurately. The authors create a method that looks at slices of the brain in small groups (2.5D) instead of taking full 3D scans, making it faster and avoiding visual issues between slices. Their approach can fill in missing brain tissue without needing to be trained specifically for particular masks, working 'zero-shot.' This method also creates consistent results across different views of the brain and produces realistic healthy tissue in place of damaged areas.
Open → 2610.08983v1

Brain MRI anomaly detection needs clearer evaluation steps

MIRTO: a registration-gated, multiverse-tested evaluation protocol for unsupervised anomaly segmentation in brain MRI

Abstract: Unsupervised anomaly detection (UAD) methods for brain MRI are ranked by a single score, yet that score rests on choices that are rarely reported: how each anomaly map is aligned with the reference, how and on which data the threshold is set, and which false-positive budget, metric, aggregation and lesion definition are used. We present MIRTO, an evaluation protocol that makes these choices explicit and measures their effect. It gates the geometry of every comparison with a registration check and label-free diagnostics of known power, sets thresholds on validation data alone and reports the false-positive volume actually realised on test, repeats each comparison over 15,552 defensible evaluation pipelines, and attaches paired subject-bootstrap intervals with multiplicity control. Applied to four UAD methods trained on the same healthy data and tested on 312 BraTS 2020 subjects, MIRTO showed that an axis-order mismatch between stored maps and the reference lowered a diffusion model's voxel AUROC from 0.873 to 0.583 whilst barely moving its slice-level AUROC. Within each metric, the method explained at least 0.95 of the variance in voxel AUROC and AUPRC and 0.77 in Dice, but only 0.14 in lesion sensitivity, where the lesion definition and hit criterion dominated. A Dice advantage that was significant at validation thresholds vanished at equal realised false-positive burden, and an exact identity attributes it to threshold transfer. A training-free change to REFLECT's latent aggregation raised Dice at equal burden by 0.052. Nine hypotheses were tested against explicit criteria; because the same cohort served to develop the protocol, all inference is exploratory.

Thu 1 OctComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Detecting unusual regions in brain MRI scans without labels can be tricky because the way results are checked and measured varies and isn’t always clear. The authors created MIRTO, a method that carefully details and tests how each step in evaluating these detections affects the final scores. They showed that small changes in how maps are aligned or thresholds set can greatly impact the results. MIRTO tested many combinations and found which parts of the evaluation process matter most for fair comparison.
Open → 2610.02136v1

Sparse cubical complexes speed up topological analysis in images

Sparse cubical complexes for efficient topology-preservation in image data

Abstract: Persistent homology (PH) is a frequently used tool for extracting and preserving topological information from image data, particularly in image segmentation, where preservation of topological structures is important. However, despite its general applicability across dimensionality, domains, and target structures, the runtime cost of PH-based methods often makes their practical use infeasible. In this work, we argue that this runtime cost is largely driven by processing information that is unimportant for downstream application (e.g. as optimization objective). We propose sparse cubical filtrations as an alternative foundation for PH computation, reducing subsequent computational costs by factors of up to 100 on real datasets. We show close agreement with the optimization signal of the dense counterpart and empirically evaluate our solution's effectiveness as an optimization objective in realistic training regimes where other PH-based objectives can practically not operate (i.e., 3D data with large patch sizes). We show how our solution improves topological accuracy by up to 80\% across six diverse datasets while maintaining pixel- and region-based accuracy.

Tue 29 SeptComputer Vision and Pattern RecognitionComputational GeometryMachine Learning
The gist
Extracting shapes and structures from images is important for tasks like image segmentation, but doing this while preserving the shapes exactly often takes too much computer time. The authors found that a lot of this time is wasted on processing unimportant details. They introduced a method called sparse cubical filtrations that speeds up the process by up to 100 times while still keeping the important topological information intact. Their method works well on complex 3D image data and improves shape accuracy without losing detail in the images.
Open → 2609.37177v1

Preference guidance improves open-vocabulary image segmentation in new domains

Preference-Guided Adaptation for Open-Vocabulary Semantic Segmentation via Prompt Disagreement

Abstract: Open-vocabulary semantic segmentation (OVSS) enables pixel-level prediction over arbitrary text-specified vocabularies and has shown strong generalization on common benchmarks. However, OVSS performance often degrades in specialized domains such as medical imaging, remote sensing, and industrial inspection, where dense pixel-level masks for adaptation are costly to obtain and require domain-specific expertise. We propose a preference-guided adaptation framework that replaces dense mask supervision with binary preferences. We observe that different prompt templates produce systematically different segmentations for the same image, a phenomenon we call prompt disagreement, and we repurpose it as a built-in source of preference supervision. Building on this, we mine localized preference queries from regions of high cross-template uncertainty, and adapt the OVSS model with Region-Localized Preference Optimization (RLPO) together with consistency regularization that stabilizes updates outside the queried region. Across extensive experiments on the MESS benchmark, the proposed method achieves consistent gains across diverse OVSS backbones without any pixel-level annotation, and remains effective under noisy preferences. Our code is available at https://github.com/blue-531/pref-ovss.

Mon 28 SeptComputer Vision and Pattern Recognition
The gist
Semantic segmentation assigns labels to every pixel in an image, but models often struggle when applied to specialized areas like medical images or satellite photos. This paper shows how asking simple yes-or-no questions about uncertain parts of an image can help improve these models without needing detailed, hard-to-get labels. The authors use disagreements between different text prompts as clues to figure out where to ask these questions, leading to better segmentation results. Their method works well on various existing models and remains effective even if some answers are noisy.
Open → 2609.34528v1

Model uncertainty does not match human ambiguity in vision tasks

Does Model Uncertainty Track Human Ambiguity? Evidence from Multi-Annotator Vision Benchmarks

Abstract: Human-model alignment is critical for trustworthy AI-assisted decision-making systems. Yet, most work evaluates model predictions against single ground-truth labels, overlooking that humans themselves often disagree on labels, a signal of genuine ambiguity. We investigate whether models struggle on the same instances that humans find difficult. We measure this on two vision datasets (FER+ and CIFAR-10H) where multiple human annotations per image capture human disagreement patterns. We evaluate eight pretrained models across three architectures (ResNet, EfficientNet, MobileNetV3) in two parts: first, whether model uncertainty (softmax confidence, entropy) correlates with human disagreement, and second, whether predictive multiplicity measures (inter-model disagreement, Jensen-Shannon divergence) do. We find that it does not: alignment is weak in both dimensions. At the discrete label level, 50.4% of CIFAR-10H images and 33.5% of FER+ images receive multiple valid classifications from humans, while the models converge on only one. These instances represent a critical failure case where humans perceive ambiguity and would request expert review, yet models decide confidently. At the continuous score level, single-model uncertainty correlates weakly with human disagreement ($ρ= 0.24--0.55$), and predictive multiplicity provides only modest improvement. Widely-used uncertainty quantification methods do not reliably identify instances humans find ambiguous. Model uncertainty should not be treated as a trustworthy signal by default for decision-making in high-stakes scenarios.

Mon 28 SeptArtificial Intelligence
The gist
Humans often disagree when labeling images because some pictures are truly unclear or ambiguous. This paper shows that popular AI models, even when uncertain, do not flag the same confusing images as humans do. The researchers tested many AI models on datasets with multiple human labels and found that models tend to be confidently wrong on images where humans see ambiguity. Therefore, relying on AI model uncertainty alone might be risky in situations needing human-like judgment.
Open → 2609.34506v1

Whole-slide image analysis models tumor microenvironment interactions dynamically

Modeling Whole-Slide Images as Dynamic Tumor Microenvironment Fields

Abstract: Due to the gigapixel-scale nature of whole-slide images (WSIs), weakly supervised WSI analysis is commonly formulated as a multiple instance learning (MIL) problem, where patch-level features are aggregated into slide-level representations. However, diagnostic and prognostic evidence often arises from spatially coherent tumor microenvironment regions and their interactions, rather than isolated patches alone. Existing patch-level or static region-based methods usually overlook how tissue regions should be adaptively formed and subsequently evolved through microenvironment interactions across heterogeneous boundaries. In this paper, we propose Concept-Guided Tumor Microenvironment Evolution (TMEvolve), a reaction-diffusion-inspired framework that models WSIs as latent tumor microenvironment fields over discrete patch graphs. TMEvolve instantiates this view as a learnable graph-discretized evolution process over patch neighborhoods. It first forms adaptive soft tissue regions as coherent microenvironment units, then performs pseudo-time evolution through two complementary local dynamics: intra-region diffusion, which stabilizes latent states within coherent tissue compartments, and concept-guided boundary flux, which propagates visual feature signals and language-derived concept signals across heterogeneous region interfaces. The evolved microenvironment regions are finally aggregated for slide-level prediction. We evaluate TMEvolve on six datasets across three weakly supervised WSI tasks: survival prediction, gene expression prediction, and histological subtype classification. TMEvolve consistently improves over representative MIL methods, pathology foundation models, and concept-guided baselines. Ablation studies and visualizations further support the effectiveness and interpretability of TMEvolve, highlighting the value of dynamic region modeling and boundary interaction.

Mon 28 SeptComputer Vision and Pattern RecognitionArtificial Intelligence
The gist
Whole-slide images of tissue samples are huge and hard to analyze all at once. The authors designed a new method, TMEvolve, that groups nearby tissue patches into regions and models how these regions interact and change over time, like a dynamic environment. This helps capture important spatial relationships that simpler methods miss. Their method improves predictions about cancer survival, gene activity, and tissue types compared to older techniques.
Open → 2609.34451v1

Quantum method improves medical image retrieval with fewer parameters

QuPID: Quantum Parameter-Efficient Input-Dependent Retrieval Adaptation for Medical RAG

Abstract: Fidelity-based quantum retrieval ranks candidates by the fidelity between query and archive states. Applying a shared input-independent unitary after fixed state encoding leaves that fidelity unchanged, so training the circuit cannot alter the ranking. Quantum parameter-efficient input-dependent retrieval adaptation (QuPID) repairs this by making the circuit input-dependent through data re-uploading and by comparing measurement readouts, vectors of local Pauli expectations, rather than states. The result is a small readout for adapting frozen image features to a local archive with limited data: training simulates the circuit classically, and inference runs on a GPU with fixed learned parameters. We characterize the class as a structured factorization of input-modulated quadratic feature maps, bound the frequency support of its re-uploading channel, and give a parameter-count generalization bound that motivates its small budget. Under a shared frozen backbone and a label-free protocol, QuPID's 60 parameters give higher precision-at-5 (P@5) on ChestX-ray14 and MURA than frozen medical encoders, and than adapters and low-rank adaptation (LoRA) with up to 5.25 million trainable parameters. On ChestX-ray14, the P@5 gain over the frozen encoder is +0.116, the lead over retuned adapters is widest at 512 adaptation examples (+0.040), and the full-budget margin over an equally compact classical rotation-plane head is +0.023 with a 95% interval excluding zero. Medical imaging is the primary testbed; the pattern recurs on two non-medical benchmarks, in report generation, and under simulated gate noise and finite-shot readout.

Sun 27 SeptArtificial IntelligenceMachine Learning
The gist
Finding relevant medical images quickly can help doctors diagnose patients better. The authors introduce a quantum-inspired method called QuPID that adjusts how images are matched using fewer adjustable parts, making it work well even with little training data. QuPID outperforms other methods on medical chest X-rays and bone images, showing that small, smart quantum approaches can improve accuracy. It also works well on some non-medical tasks and handles noisy conditions better than alternatives.
Open → 2609.33351v1

Efficient image restoration with flexible learned degradation operators

Perturb-and-Solve: Efficient Learned-Operator Conditioning for Latent Diffusion Inverse Problems

Abstract: Latent diffusion models serve as powerful priors for solving inverse problems in image restoration, such as deblurring, inpainting, and super-resolution. Current methods have a trade-off between generality and efficiency. Solvers that are restricted to a fixed set of degradation operators are fast and efficient. Methods that support arbitrary degradation operators are slow and require gradients through the diffusion network. To break this bottleneck, we introduce PASEO (Perturb-And-Solve for Efficient Operator conditioning), a method that uses a small (1M parameters) learned network to degrade diffusion model predictions in latent space. PASEO supports learned degradation operators without back-propagating through the diffusion network. We efficiently sample reconstructions from an approximate posterior by combining the diffusion model's prediction with the observed image. We do this by adding noise and solving linear equations based on a local linear approximation of the learned network, without building or inverting large covariance matrices. Across super-resolution, deblurring, and inpainting on FFHQ and COCO, PASEO achieves strong perceptual quality while running up to 9x faster and using up to 34% less peak memory than the tested baselines, with the same or fewer model evaluations.

Sun 27 SeptMachine LearningComputer Vision and Pattern Recognition
The gist
Restoring images that are blurry, missing parts, or low resolution is a common problem. The authors present a new way to do this faster and with less computing power by teaching a small network to mimic how images get degraded, without needing to retrace through a large image generator. Their approach combines this learned degradation with a powerful image generator, leading to faster and better image repairs for various tasks like sharpening and filling in missing parts.
Open → 2609.33171v1

Simple methods improve gene expression prediction from H&E images

Is H&E Image-to-Spatial Transcriptomics Simpler Than It Looks?

Abstract: Predicting spatial gene expression from routine H&E histology offers a scalable route toward spatial molecular profiling. Recent work has pursued increasingly sophisticated architectures to capture spatial context and richer expression structure. At the same time, simple estimators have shown strong performance in several studies, but what they already solve and where additional complexity is needed remain unclear. We study this behavior through the structure of prediction error under the mean-squared error (MSE) objective. Differences in average expression across genes can account for a substantial part of aggregate prediction performance, while a key unresolved error lies in recovering variation within each slide. Decomposing MSE into slide-level and within-slide components, we find that the within-slide component has lower residual-normalized parameter sensitivity in controlled neural experiments. This motivates Component-Guided Loss (CGL), which increases supervision of the within-slide component. CGL-Linear is a closed-form affine instantiation that achieves overall state-of-the-art performance across HEST-1k cohorts and gene-panel sizes. The same within-slide supervision improves existing neural models. These results suggest that substantial gains can come from aligning the training objective with prediction-error structure rather than increasing model complexity.

Sat 26 SeptComputer Vision and Pattern RecognitionMachine Learning
The gist
Predicting gene activity in tissue samples usually requires complex AI models, but the authors found that simpler methods can perform just as well by focusing on specific errors within each sample. They show that understanding the types of mistakes made during prediction leads to better training methods. This means that models can be improved not by being more complicated but by using better ways to learn from the data.
Open → 2609.32857v1

Ultrasound image clarity improved by new self-supervised despeckling method

Mask2Restore: Self-Supervised Ultrasound Despeckling via Inpainting

Abstract: Medical ultrasound (US) is inherently degraded by speckle, a granular interference pattern that is often treated as a complex form of noise in image restoration. However, unlike random noise, US speckle originates from coherent scattering within tissue and is therefore highly spatially dependent and deterministic under fixed acquisition conditions, making US speckle suppression fundamentally different from natural image denoising. Because speckle-free US targets are unavailable in practice, self-supervised denoising is necessary. Blind-spot networks (BSN) are the dominant self-supervised paradigm for natural images, but their pixel-wise masking strategy assumes spatially independent noise, an assumption poorly matched to US speckle, which is spatially correlated over multiple pixels rather than pixel-wise independent. To address this mismatch, we propose Mask2Restore, a self-supervised US despeckling framework that reformulates despeckling as contextual inpainting with block-wise masking on single noisy images. Unlike pixel-wise BSN masking, block-wise masking addresses this multi-pixel speckle correlation by removing locally correlated speckle neighborhoods and shifting the reconstruction cues used by the network from adjacent speckle correlations to broader anatomical context. We further introduce cross-resolution context regularization (CRCR), which suppresses residual speckle bias by enforcing consistency across multi-resolution predictions. Experiments on simulated and in vivo carotid US, unseen fine-structure cases, and downstream cardiac segmentation demonstrate improved speckle-detail trade-offs, better preservation of fine anatomical structures, and practical value for subsequent image analysis.

Sat 26 SeptComputer Vision and Pattern Recognition
The gist
Ultrasound images used in medical scans have a grainy pattern called speckle that makes the pictures hard to read. This speckle is different from normal random noise because it depends on how tissues scatter sound waves. The authors created a new method called Mask2Restore that removes this grainy pattern without needing perfect clean images to learn from. Their approach masks blocks of pixels instead of single pixels to better handle the speckle’s pattern, and uses multiple image sizes to improve accuracy. Tests show this method keeps more fine details and helps with medical tasks like heart image segmentation.
Open → 2609.32844v1

Data scale factors affect brain image model training outcomes

Samples, Sources, Space: Decomposing Data Scale in Spatially Structured Representation Learning of Human Brain Microarchitecture

Abstract: Scaling studies typically represent training data by a single count of samples. For hierarchically and spatially structured data, however, the same number of samples can be drawn from few or many sources and distributed differently across the underlying domain. We therefore study data scaling as an allocation problem, separating unique sample count, source diversity, and spatial coverage. We study this decomposition in microscopic whole-brain histology, where a source is an individual brain, and a sample is an image patch at a specific spatial location. Across 93 controlled pretraining runs of a contrastive model that uses spatial proximity for supervision, we vary data allocation, compute, and model capacity over 11.6 million spatially anchored image patches from 21 human brains. Performance improves with more unique samples, broader spatial coverage, additional compute, and larger model capacity. At fixed sample count, distributing samples across one to 18 subjects produces no detectable improvement, even though representations generalize substantially better to subjects encountered during pretraining. Inter-subject variation therefore strongly affects generalization, but additional subjects provide no benefit when a fixed sample budget is distributed across more sources. These results establish sample count, source diversity, and spatial coverage as distinct axes of data scaling in spatially structured representation learning.

Fri 25 SeptArtificial Intelligence
The gist
When training computer models to understand detailed images of the human brain, how the training data is spread out matters. The authors found that increasing the total number of image samples, covering more brain areas, and using bigger models all help performance. But simply using images from more people does not improve the model if the total number of images is fixed. This means that where and how many images are taken can be more important than just the number of people they come from.
Open → 2609.31201v1

Image recovery improves using neural priors with Fourier phase data

Image Reconstruction from Phase with Untrained Neural Priors

Abstract: Fourier phase encodes important spatial image structure, but recovering an image without measured spectral magnitude requires additional constraints and leaves absolute intensity ambiguous. We propose a projection-based two-stage framework that combines Fourier-phase and spatial-support constraints with an image-specific neural prior. The first stage alternates constraint enforcement with regularized neural-prior updates, while the second performs phase/support refinement alone with guaranteed convergence. We evaluate two neural-prior implementations on the same 77 microscopy images and compare them with a constraint-only baseline. After 500 final refinement passes, the best-performing variant achieves 31.41 dB pooled PSNR, 35.75 dB mean PSNR, and 0.9531 mean SSIM, improving pooled PSNR by 1.51~dB and reducing pooled MSE by 29.3% relative to the baseline. The results demonstrate the benefit of combining neural guidance with explicit constraint refinement at the evaluated iteration budget, while showing that lower phase residual alone does not guarantee greater reconstruction accuracy.

Fri 25 SeptComputer Vision and Pattern Recognition
The gist
Recovering an image when you only have its phase information is tricky because some details and brightness can't be determined directly. The paper’s authors created a two-step way to reconstruct images by combining known image shapes and a special, flexible neural network that adapts to each image. They tested their method on microscopy pictures and found it produced clearer and more accurate images than simple methods. However, they also found that just reducing phase errors doesn't always mean better final images.
Open → 2609.30659v1

Physics informed method reduces shadows in fetal ultrasound imaging

Shadow Reduction in Ultrasound Imaging Using Differentiable Simulation and Radiance Field Decomposition

Abstract: Acoustic shadows from bone and other highly attenuating tissues obscure clinically important structures in ultrasound. In fetal brain imaging, skull-induced artefacts disproportionately degrade the hemisphere closer to the transducer (proximal), limiting symmetric assessment of the two hemispheres. Existing correction methods require raw scanner data, impose restrictive assumptions on tissue properties, or rely on generative models that may hallucinate anatomy. We present RFlash, a physics-informed post-processing method that decomposes beamformed ultrasound images into explicit attenuation and scatter-intensity maps using a differentiable radiance-field formulation of image formation. Attenuation-adaptive re-rendering then removes the dependence of the signal at each depth on the intervening tissue, equivalent to virtually advancing the transducer into the tissue. Across 1,261 3D fetal brain volumes, 143 real 2D curvilinear abdominal scans, and 1,200 simulated 2D linear-probe liver scans, RFlash reduces shadow-related intensity differences more effectively than classical Hughes-Duck attenuation correction. For a gestational-age model trained on the distal hemisphere (further from the transducer) and applied to the proximal hemisphere, prediction error decreases by 5.1 days (40%) relative to the original images. The estimated attenuation maps also yield shadow-confidence maps that improve random-forest bone-shadow segmentation over the image alone and receive greater SHAP importance than an existing neural confidence-map baseline, suggesting greater physical consistency. RFlash requires neither hardware modification nor access to raw scanner data and supports 2D and 3D acquisitions with linear and curvilinear probes, making it widely applicable allowing clinicians to use our method on their already acquired scanners and images.

Thu 24 SeptComputer Vision and Pattern Recognition
The gist
Ultrasound images of fetal brains can be blurry or blocked by shadows from bones, making it hard to see both sides equally. The authors developed a method called RFlash that uses physics and image simulation to separate shadow effects from real signals, allowing clearer views by correcting for these shadows after the image is taken. This helps doctors better measure fetal brain development without needing special ultrasound machines or raw data. Their tests show that RFlash improves shadow removal more than older methods and makes it easier to identify bone shadows.
Open → 2609.29373v1

Polyp image synthesis improves colonoscopy data with adaptive mucosal context

Beyond the Foreground: FOV-Aware Polyp Image Synthesis via Lesion-Guided Adaptive Mucosal Context Propagation

Abstract: Synthetic image and mask pairs can alleviate scarce colonoscopy annotations, but realistic synthesis requires preserving the supplied lesion while generating compatible mucosa. Existing foreground-guided methods treat all non-foreground pixels as background and rely mainly on local integration. Directly applying them to colonoscopy causes two problems: non-mucosal black regions contaminate generated tissue, and local reasoning produces inconsistent mucosal texture and illumination. We propose LAMP, the first foreground-guided framework for polyp image synthesis based on lesion-guided adaptive mucosal context propagation. LAMP explicitly separates the lesion, valid mucosa, and camera exterior using a field-of-view (FOV) mask. Lesion-to-Mucosa cross-attention extracts lesion appearance conditions for valid-mucosa locations, while FOV-constrained multidirectional Vision Receptance Weighted Key Value propagates them over legal tissue support. An adaptive gate then controls their residual fusion into the diffusion U-Net. Extensive experiments on five polyp datasets demonstrate that LAMP substantially outperforms existing methods in overall generation quality and consistently improves five downstream segmentation models. Our code will be released at https://github.com/wangtong627/LAMP.

Thu 17 SeptComputer Vision and Pattern Recognition
The gist
Finding enough annotated colonoscopy images with polyps is hard, which slows progress in medical imaging. The authors present a new method called LAMP that creates realistic synthetic colonoscopy images with polyps by carefully copying details from the lesion and surrounding mucosal tissue. This technique avoids mistakes caused by black non-tissue areas and produces better textures and lighting. Their approach helps train segmentation models that can better detect polyps, potentially improving colonoscopy image analysis.
Open → 2609.19966v1

Normal kidney images help detect rare glomerulus abnormalities accurately

Seeing Abnormal from Normal: Glomerular Abnormality in Representations of Normal Renal Morphology

Abstract: Fine-grained evaluation of glomerular pathology must distinguish normal glomeruli from abnormalities such as global and segmental glomerulosclerosis, obsolescent, ischemic, solidified, disappearing, and atubular glomeruli. Supervised classification requires labeled examples of every category, which is impractical when subtypes are rare or absent from the training cohort. One-class anomaly detection offers an alternative by modeling normal data and scoring deviations, allowing previously unseen abnormalities to be detected. We use the frozen residual U-Net backbone of Omni-Seg, pretrained to segment structurally normal renal primitives without abnormal-subtype labels. We propose NoRDeC (Normal-Reference Detection and Characterization), a framework combining Mahalanobis normal-reference scoring with layer-wise representation analysis to determine whether and where glomerular pathology is encoded, how spatial aggregation affects detection, and whether abnormalities alter inter-layer relationships differently. Using glomerular images from two institutions, we evaluate backbone layers and aggregation strategies, compare NoRDeC with PaDiM and PatchCore, and analyze representations using centered kernel alignment (CKA). Layer 4 with Center-70 aggregation achieved a pooled AUROC of $0.926\pm0.013$. NoRDeC achieved the highest AUROC in six of seven abnormality categories and in the pooled analysis, while CKA suggested subtype-dependent changes in inter-layer relationships not captured by anomaly scores alone. The normal-reference model is fitted using only normal glomeruli; abnormality labels are used for configuration selection, evaluation, and grouping in the representation analysis. These results show that a frozen renal feature extractor can support both detection and representation-level characterization of glomerular abnormalities without using abnormal examples to fit the detector.

Wed 16 SeptComputer Vision and Pattern Recognition
The gist
Detecting kidney problems in tiny parts called glomeruli is tricky because some abnormalities are rare or missing in training data. The authors use a model trained only on normal kidney images to spot unusual ones by measuring how different they look from normal patterns. Their technique, called NoRDeC, analyzes features at different model layers to find and understand abnormalities without needing examples of each problem type. Tests showed it performs better than some existing methods in identifying various glomerular abnormalities. This approach helps detect kidney disease signs even when rare abnormal cases are not available for training.
Open → 2609.19444v1

Synthetic data adaptation improves few-shot cryo-ET classification accuracy

Bridging the Synthetic-to-Real Gap for Few-Shot Cryo-ET Classification

Abstract: Subtomogram classification in cryo-electron tomography (cryo-ET) is a challenging problem due to the scarcity of labeled examples. While cryo-ET simulators can be adopted to generate unlimited synthetic data, the substantial domain gap between synthetic and real subtomograms hinders its practical utilization. In this work, we propose a novel synthetic-to-real adaptation framework with a learnable transformation module, bridging this gap at both the input and feature levels. Extensive experiments demonstrate that our method consistently outperforms existing transfer learning baselines in few-shot settings.

Sat 12 SeptComputer Vision and Pattern RecognitionMachine Learning
The gist
Classifying tiny 3D images from cryo-electron tomography is hard because there aren’t many labeled examples to learn from. The authors show that while computer-generated synthetic images could help, they usually look different from real ones, causing mistakes. They created a new method that transforms synthetic data to look more like real data and adjusts features to close this gap. This approach helps computers learn better from very few real examples and improves classification performance.
Open → 2609.14097v1

Deep learning improves brain surface labeling with limited expert data

Geometric-to-Semantic Spherical Transfer Learning for Cortical Sulci Labeling

Abstract: Deep learning on cortical surfaces faces a dilemma: capturing the complex topology of over 60 nomenclature-dependent sulci per hemisphere requires high-capacity models, yet the extreme scarcity of expert annotations ($N=62$ subjects) inevitably causes overfitting. Standard supervised approaches fail to generalize in this data-scarce regime, particularly for variable and small sulci where topological ambiguity is high. To overcome this limitation, we introduce a Geometric-to-Semantic Spherical Transfer Learning framework. First, we leverage massive unlabeled data (UK Biobank, $\approx$30,000 subjects) to pre-train a spherical encoder using a locally-optimized strategy. By relying solely on continuous surface features (curvature and depth), the relevance of this pre-training is confirmed by the model's ability to detect localized and rare topological traits, such as sulcal interruptions. The downstream labeling task, however, introduces extracted sulcal fundi (lines) as an explicit semantic input. To bridge this dimensional domain gap (from purely geometric to semantic) without causing catastrophic forgetting, these anatomical lines are integrated into the pre-trained backbone via a soft-initialized Topological Prior Injector. Our experiments demonstrate that this approach outperforms fully supervised baselines trained from scratch, achieving a mean Dice of 0.77. Crucially, a local analysis reveals that the self-supervised geometric priors yield the largest performance gains on variable and tertiary sulci (up to 14.8%), confirming that learning the cortex shape is highly beneficial for identifying its rarest parts.

Fri 11 SeptMachine LearningComputer Vision and Pattern Recognition
The gist
Labeling the grooves on the brain’s surface is hard because there are many small, complex features and only a few expert examples to learn from. The authors used a huge set of unlabeled brain scans to train a model to understand brain shape before teaching it to recognize specific labeled grooves. This two-step approach helps the model identify rare and tricky brain features better than training with limited labeled data alone. Their method increased accuracy notably, especially on the most variable and uncommon brain grooves.
Open → 2609.12627v1

SkNeXt reduces cost of brain cell mapping from huge microscopy data

SkNeXt enables topology-guided neuronal reconstruction from petabyte-scale microscopy data

Abstract: Recent advances in high-resolution fluorescence and electron microscopy have enabled nanoscale imaging across increasingly large brain volumes, but the resulting terabyte- to petabyte-scale datasets make complete neuronal reconstruction prohibitively expensive in computation, data movement, and manual proofreading. Here, we present SkNeXt, a topology-first framework for scalable neuronal reconstruction from large volumetric microscopy datasets. Instead of densely processing entire image volumes, SkNeXt first converts neuronal morphology into compact SWC skeletons that preserve long-range connectivity. Proofreading is therefore focused on sparse neuronal trees, allowing branch, continuity, and connectivity errors to be corrected before high-resolution reconstruction. The corrected skeletons then serve as persistent structural priors for recovering detailed morphology while preserving neuronal identity and topology. Crucially, SkNeXt also uses neuronal skeletons as spatial indices for selective data access, retrieving high-resolution image regions only along reconstructed trajectories and bypassing most background and signal-free volumes. This substantially reduces I/O and computational overhead, allowing reconstruction cost to scale with neuronal morphology rather than total dataset size. Using SkNeXt, we reconstructed neurons from a petabyte-scale super-resolution fluorescence dataset of the mouse brain on a single GPU within one week, without requiring exhaustive dense inference across the complete imaging volume.

Wed 9 SeptComputer Vision and Pattern Recognition
The gist
Mapping the shape and connections of brain cells from large, detailed microscope images is usually very slow and expensive. The authors developed SkNeXt, a new method that first summarizes brain cell shapes in simple outline forms to focus on the important parts. This lets them fix mistakes easily and only look at detailed images where needed, saving lots of time and computer work. Using SkNeXt, they mapped neurons from a very large mouse brain dataset using a single graphics processor in just one week.
Open → 2609.09832v1

Attention guidance improves reliability in multiple instance learning tasks

CAR-MIL: Counterfactual Attention Regularization for Multiple Instance Learning

Abstract: Multiple Instance Learning (MIL) is widely used for weakly supervised learning, particularly in digital pathology, where fine-grained annotations are costly. Most MIL methods aggregate instance features via attention mechanisms. However, attention weights do not always faithfully reflect instance importance and may focus on spuriously correlated regions. In this work, we propose CAR-MIL, a framework that explicitly guides attention learning through a counterfactual attention regularization objective inspired by counterfactual explanations. Built on a standard attention-based MIL architecture, our approach introduces a lightweight counterfactual attention branch trained to produce an alternative prediction while remaining close to the factual attention distribution. This encourages prediction changes to arise from minimal, structured redistributions of attention, leading to more informative evidence allocation. The resulting factual and counterfactual attention maps capture complementary evidence: the former highlights regions supporting the prediction, while the latter reveals regions whose reweighting would challenge it. We evaluate our method on synthetic MIL benchmarks with instance-level ground truth enabling controlled analysis of attention behavior and on five digital pathology datasets across four tasks. CAR-MIL maintains competitive classification performance, with the largest gains observed on more challenging tasks, while improving attention reliability, demonstrating the benefits of integrating counterfactual explainability reasoning into attention learning. Code is available at: https://github.com/ImaneCR/CAR-MIL/.

Tue 8 SeptComputer Vision and Pattern Recognition
The gist
When computers are trained to recognize details in images without exact labels, they can get confused about what parts of an image really matter. The authors designed a new way to teach the computer to pay attention to important areas more accurately by showing it what would happen if it shifted its focus slightly. This approach leads to better detection and explanation of relevant image parts, especially in complex medical images. Their method performs well on test data and makes the computer's focus more trustworthy.
Open → 2609.08419v1