Federated learning with knowledge distillation for multi-organ segmentation with partially labeled datasets

Soopil Kim, Heejung Park, Myeongkyun Kang, Kyong Hwan Jin, Ehsan Adeli, Kilian M. Pohl, Sanghyun Park

Abstract: Highlights•We propose a knowledge distillation-based federated learning to address forgetting issue.•Local training is regularized with the knowledge of a global model and local organ-specific models.•We design a multi-head U-Net architecture that learns a shared embedding space for various organs.•We evaluate the proposed method using 8 publicly abdominal CT datasets for 7 different organs.•We achieve SOTA performance in accuracy, inference time, and the number of parameters.

External IDs:dblp:journals/mia/KimPKJAPP24