Inr models improve brain MRI segmentation with fewer parameters
How Far Can INRs Go? Cross-Domain Parameter-efficient INR-Based Semantic Segmentation for Brain MRI
Computer Vision and Pattern Recognition
Summary
Segmenting parts of brain scans is important for medical care but can be hard because of limited examples and changing scan styles. The authors studied a new type of lightweight model called Implicit Neural Representations (INRs) that can do this segmentation well with fewer parameters. They found INRs work best when limited in size and training data, unlike traditional methods that improve with scale. By combining clues from multiple internal layers, they built a better INR model, HierINRSeg, which performs more accurately on brain MRI scans even across different hospitals.
What this means in practice
- •For medical imaging engineers: Design MRI segmentation systems that run efficiently with limited hardware while maintaining accuracy across different hospital data.
- •For biomedical device developers: Implement compact segmentation models in portable brain imaging devices to support diagnosis in resource-constrained settings.$Commercial implications: This paper enables smaller, robust segmentation software suitable for medical imaging devices sold to healthcare providers.
Authors
Ziyao Shang, Pouya Sadeghi, Letian Jiang, Alexander Wong, Sirisha Rambhatla
Abstract
Biomedical image segmentation is central to medical image analysis, but practical deployment often faces limited annotations, memory constraints, and cross-site distribution shifts. Implicit Neural Representations (INRs) have recently emerged as a lightweight alternative for semantic segmentation, achieving competitive performance with substantially fewer parameters than conventional architectures. However, the mechanisms, scaling behavior, and domain generalization abilities of INR-based segmentation remain insufficiently understood. In this work, we study these questions in the context of cross-domain brain MRI segmentation. We analyze INR-based segmentation across low-parameter regimes, comparing it with conventional pipelines in both in-domain and out-of-domain settings. Surprisingly, we find that INR-based models do not simply improve with increasing parameter budget. Their advantage is most pronounced under low-parameter and limited-augmentation settings, while U-Net-based models benefit more from larger capacity and standard augmentation. We also investigate how INRs encode semantic information in their hidden features and show that complementary segmentation-relevant structure is distributed across multiple INR layers. Building on this insight, we introduce HierINRSeg, a hierarchical INR-based architecture that aggregates multi-layer representations for improved robustness and generalization. Extensive experiments show that HierINRSeg consistently outperforms MetaSeg, a strong recent INR-based segmentation baseline, with an average improvement of 5.6 percentage points in Dice for the in-domain test set and 8.2 percentage points out-of-domain. Overall, our analysis identifies the conditions under which INR-based segmentation is most effective, providing concrete guidance for model selection and future research.