Alexander Jaus

Alexander Jaus

Postdoctoral Researcher in Computer Vision

ISBI 2026

Learning to Look Closer: A New Instance-Wise Loss for Small Cerebral Lesion Segmentation (Oral)

L. Bouteille, A. Jaus, J. Kleesiek, R. Stiefelhagen, L. Heine

IEEE International Symposium on Biomedical Imaging (ISBI), 2026

Paper
Medical Imaging Segmentation

Key contributions

  • CC-DiceCE: a connected-component-wise loss derived from the CC-Metrics framework, giving equal weight to every lesion regardless of size.
  • The first rigorous multi-dataset comparison of instance-wise losses (CC-DiceCE and blob loss) against a properly configured nnU-Net baseline across five heterogeneous brain MRI datasets.
  • Empirical evidence that CC-DiceCE improves lesion detection (recall) with minimal precision trade-offs, and generally outperforms blob loss.
Per-voxel gradient comparison: BlobDiceCE (left) vs. CC-DiceCE (right). CC-DiceCE assigns stronger, more spatially focused gradients to small lesions via Voronoi region partitioning, directly boosting their influence on training.
Per-voxel gradient comparison: BlobDiceCE (left) vs. CC-DiceCE (right). CC-DiceCE assigns stronger, more spatially focused gradients to small lesions via Voronoi region partitioning, directly boosting their influence on training.

How it works

Traditional loss functions in medical image segmentation, such as Dice, often under-segment small lesions because their small relative volume contributes negligibly to the overall loss. To address this, instance-wise loss functions and metrics have been proposed to evaluate segmentation quality on a per-lesion basis. We introduce CC-DiceCE, a loss function based on the CC-Metrics framework, and compare it with the existing blob loss.

Both are benchmarked against a DiceCE baseline within the nnU-Net framework, which provides a robust and standardized setup. We find that CC-DiceCE loss increases detection (recall) with minimal to no degradation in segmentation performance, though with dataset-dependent trade-offs in precision. Furthermore, our multi-dataset study — spanning five heterogeneous brain MRI datasets including lacunes, cerebral microbleeds, brain metastases, white matter hyperintensities, and gliomas — shows that CC-DiceCE generally outperforms blob loss.

Citation

Bouteille, L., Jaus, A., Kleesiek, J., Stiefelhagen, R., & Heine, L. (2026). Learning to Look Closer: A New Instance-Wise Loss for Small Cerebral Lesion Segmentation. In 2026 IEEE 23rd International Symposium on Biomedical Imaging (ISBI). IEEE. https://doi.org/10.1109/ISBI61048.2026.11515682