Multi-scale Multi-task Distillation for Incremental 3D Medical Image Segmentation

This work studies incremental 3D medical image segmentation, where a segmentation model must adapt to new tasks while preserving performance on previously learned structures.

The core idea is to use multi-scale, multi-task knowledge distillation to transfer information from prior models into the updated model. Instead of treating each new segmentation target as an isolated training problem, the framework encourages the model to retain useful representations across multiple levels and tasks, reducing forgetting during incremental updates.

This project reflects my broader interest in annotation-efficient and continually adaptable medical AI systems, especially for deployment settings where full retraining is costly and new anatomical targets or datasets may arrive over time.

Recommended citation: Tian, M., Yang, Q., & Gao, Y. (2023). "Multi-scale Multi-task Distillation for Incremental 3D Medical Image Segmentation." Computer Vision – ECCV 2022 Workshops, Lecture Notes in Computer Science, 13803, 369–384.
Download Paper