LISynSeg: Data-Centric Label-to-Image Synthesis for Cross-Modality Whole-Heart Segmentation
2026-08-31 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors looked at how changing the way training data is made and used can help improve heart image segmentation from CT and MRI scans without altering the model itself. They created a method called LISynSeg, which mixes real images with synthetic ones made from heart labels, making the synthetic images more realistic by adding chest context and simulating variations in heart structure. Their experiments showed that using both real and synthetic images together helps the model work better across different types of scans, especially MRI. This suggests that focusing on data preparation can improve heart segmentation performance despite differences in imaging modality and annotations.
whole-heart segmentationcomputed tomography (CT)magnetic resonance imaging (MRI)nnU-Netdata augmentationlabel-to-image synthesiscross-modality segmentationcardiac label variationtraining supervisionCARE Whole-Heart benchmark
Authors
Jiacheng Wang, Ivana Isgum, Ipek Oguz
Abstract
Whole-heart segmentation (WHS) in computed tomography (CT) and magnetic resonance imaging (MRI) is affected by acquisition shifts and heterogeneous cardiac annotations. Existing WHS systems combine architectural design, transfer learning, and generic spatial or intensity augmentation. We investigate whether changes to data augmentation and training supervision can improve cross-modality WHS while the segmentation architecture is held constant. We present LISynSeg, a data-centric approach that augments real-image nnU-Net training with label-to-image synthesis. Synthetic volumes are generated from cardiac label maps using contrast and acquisition perturbations calibrated to the training cohort, then mixed with real images to retain thoracic context absent from the labels (and thus the synthesized images). We model cardiac label variation through controlled changes in myocardial wall thickness and partial supervision of uncertain vessel endpoints. On the CARE Whole-Heart benchmark, synthetic-only training performs worse than the real-image nnU-Net baseline, whereas calibrated real-synthetic training improves cross-modality segmentation without changing the architecture; the improvement is larger for MRI than for CT. The results show that modifying the training data strategy can benefit model development for heterogeneous cardiac data. Code and trained weights will be released at https://github.com/MedICL-VU/Care26_LISynSeg.