Leveraging Transfer Learning with Class-Specific Decoders for Laparoscopic Segmentation

2026-07-31

AI Summary

This study addresses the challenges of multi-organ segmentation in laparoscopic surgery, particularly the anatomical complexity and severe class imbalance exacerbated by small or partially exposed organs. To tackle these issues, the authors propose the Class-specific Encoder-decoder with Multi-domain adaptation (CEMD), which uniquely integrates class-specific decoders with cross-surgical-domain transfer learning. By conducting cross-domain pretraining and full fine-tuning between rectal and cholecystectomy datasets, and systematically analyzing feature adaptation mechanisms in both encoder and decoder across training stages, the work demonstrates the decoder’s pivotal role in preserving and transferring anatomical knowledge. Experiments show that CEMD achieves a Dice score of 62.4% after cross-domain fine-tuning and converges significantly faster than training from scratch; however, performance gains for low-frequency organs remain limited, underscoring the persistent challenge of class imbalance even in transfer learning scenarios.

Abstract

Effective multi-organ segmentation in surgical data requires learning the intricate anatomical features and alleviating the challenge of class imbalance, which results from relatively lower proportions of small and limitedly exposed structures. Recent works on laparoscopic multi-organ segmentation focus on learning structure-specific features through class-specific decoder architectures and report favorable results. This work extends the decoder-focused architectures to investigate knowledge sharing in the cross-surgical domain. We utilize two datasets representing different surgical domains, rectal and cholecystectomy surgeries, to explore how surgical conceptual knowledge transfers under partially common anatomical representations. Additionally, we compare the feature adaptation for the encoder and decoder at different training stages to analyse the knowledge adaptation and retention in the network. Our results corroborate previous findings on decoder-specific architectures and demonstrate that the organ-specific decoder model (CEMD), fully fine-tuned after cross-domain pre-training, achieves the highest segmentation performance (62.4\% dice) while converging substantially faster than training from scratch. However, we also find that class imbalance in surgical data remains a persistent challenge that transfer learning does not fully resolve for underrepresented anatomical structures.

Problem

• laparoscopic segmentation
• class imbalance
• transfer learning
• multi-organ segmentation
• surgical domain adaptation

Innovation

• transfer learning
• class-specific decoders
• laparoscopic segmentation
• cross-surgical domain
• multi-organ segmentation