Graph-guided Sequential Transfer for Medical Image Segmentation

Illustration of the optimal sequential transfer path for a specific target task, 16-T2-NCR.

The medical image processing field often encounters the critical issue of scarce annotated data. Transfer learning has emerged as a solution, yet how to select an adequate source task and effectively transfer the knowledge to the target task remains challenging. To address this, we propose a novel source selection framework designed to identify the landmark source with an effective sequential transfer path for the given target task. Specifically, we first assess the relatedness among source tasks, estimated by our task affinity metric.

Considering the characteristics of medical image segmentation tasks, we first analyze the image and label similarity between tasks and compute the task affinity score. Based on the analysis, we then construct a comprehensive source graph and combine the informativeness and representativeness of each node to identify the landmark source for the target. To ensure a positive transfer, our proposed method finds a sequential transfer path to the target by minimizing both transfer and search costs. by incorporating intermediate source tasks, it gradually narrows the domain discrepancy and consequently improve the transfer performance on the target task.

We conducted extensive experiments on three brain MRI medical datasets to demonstrate the efficacy of the proposed framework in finding the best source sequence. The results show that our method outperforms other transfer learning approaches by a considerable margin, improving state-of-the-art performance by 6.61% for FeTS 2022, 0.66% for iSeg-2019, and 1.70% for WMH in terms of segmentation Dice score.

Publication

@inproceedings{yang2024graph,
  title={Graph-guided Sequential Transfer for Medical Image Segmentation},
  author={Yang, Jingyun and Wang, Jingge and Zhang, Guoqing and Li, Yang},
  booktitle={2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM'24)},
  year={2024},
  organization={IEEE},
  address={Lisbon, Portugal},
  month={December}
}

Project list