Model reduces manual effort in 3d medical image segmentation
Confidence-Aware Teacher-Student Distillation for 3D Medical Segmentation
Computer Vision and Pattern Recognition
Summary
Medical image segmentation usually needs detailed labels on many 3D medical scans, which takes a lot of time. The authors propose a method where only a few dots on a single 2D slice are needed as input. Their method uses a pretrained model to generate initial 3D labels with confidence scores, then trains a smaller model that focuses on the most reliable parts of those labels. This approach improves accuracy compared to other limited-label methods and greatly reduces how much manual labeling is needed.
What this means in practice
- •For medical imaging teams: Produce accurate 3D anatomical segmentations from minimal user input to reduce manual labeling time.
- •For clinical software developers: Integrate annotation-efficient segmentation models to enable faster and cheaper deployment of 3D medical image analysis tools.$Commercial implications: Enables development of segmentation features that require far less manual input, improving market competitiveness of imaging software.
Authors
Georgios Triantafyllou, Dimitris K. Iakovidis
Abstract
Medical image segmentation models typically rely on large amounts of densely annotated volumetric data, limiting their scalability across tasks and imaging modalities. This work addresses the challenge of predicting entire 3D anatomical structures from extreme annotation sparsity. An annotation-efficient student-teacher framework is proposed for automatic 3D medical segmentation that requires only a set of point prompts on a single 2D slice per volume, as input. A foundation model serves as an offline teacher, utilizing the provided point prompts from the selected slice to full-volume pseudo-annotations alongside their corresponding spatial confidence scores prior to student training. To mitigate the error propagation of noisy pseudo-annotations, a task-specific 3D student network is trained using a confidence-aware optimization strategy. By leveraging the teacher's pre-computed confidence scores, this strategy explicitly excludes statically uncertain regions of the pseudo-annotations from the loss calculation, while simultaneously emphasizing regions with higher confidence. Evaluated on 3D cardiac MRI datasets, our framework outperforms state-of-the-art semi-supervised methods, improving segmentation performance by up to 43.6%. Furthermore, it drastically reduces the manual annotation burden to just a few positive point prompts per volume, while improving surface boundary precision by up to 14.7% over the teacher and successfully recovering up to 34.1% of the performance gap toward the fully supervised upper bound.