Attention-Based Prototype Calibration for Multi-Rater Few-Shot Medical Image Segmentation

2026-06-15

AI Summary

This work addresses the often-overlooked systematic discrepancies among multiple expert annotations in few-shot medical image segmentation by introducing an attention-based prototype calibration mechanism that requires no modification to the backbone network. The method explicitly models the deviation between individual annotator-specific prototypes and the consensus representation in the prototype space, employing a lightweight attention module to dynamically calibrate annotator-specific prototypes. This enables personalized segmentation while preserving semantic consistency. Compatible with existing prototypical networks, the proposed approach effectively captures annotation variability and demonstrates significant performance gains over baseline methods on multi-annotator medical imaging datasets, thereby validating its efficacy and practicality.

Abstract

Few-shot medical image segmentation methods typically assume a single ground-truth annotation, overlooking systematic variability across expert raters commonly observed in clinical datasets. We propose an attention-based prototype calibration framework for few-shot multi-rater segmentation that models rater-specific deviations from a consensus representation in prototype space. A lightweight yet principled attention operator directly refines rater prototypes without modifying the backbone feature extractor, making the approach fully compatible with existing prototype-based few-shot segmentation methods. This design preserves semantic consistency while enabling personalized segmentation outputs with minimal computational overhead. Experiments on multi-rater medical imaging datasets demonstrate consistent improvements over baseline prototype approaches, highlighting the effectiveness of structured prototype calibration for modeling annotation variability. Our code is available at https://github.com/truong2710-cyber/JAPC.

Problem

• few-shot segmentation
• multi-rater
• annotation variability
• medical image segmentation
• prototype calibration

Innovation

• attention-based calibration
• multi-rater segmentation
• prototype refinement
• few-shot learning
• medical image segmentation