An Attentive Representative Sample Selection Strategy Combined with Balanced Batch Training for Skin Lesion Segmentation

2025-03-21

AI Summary

To address data-efficient learning under low annotation budgets in skin lesion segmentation, this paper proposes a synergistic framework integrating representative sample selection with balanced batch training. To mitigate performance instability caused by random sampling in few-shot settings, the method introduces a clustering-driven representative sample selection mechanism and—novelly for medical image segmentation—an unsupervised balanced batch loading strategy. Technically, it unifies prototype-based contrastive learning, K-means clustering, attention-weighted sampling, and dynamic batch balancing. Evaluated on the ISIC 2018 dataset under extremely low annotation budgets, the approach achieves a 3.2% improvement in Dice coefficient over the state-of-the-art sampling methods. Moreover, it significantly enhances model generalizability and training stability.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Semi-Supervised Learning
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

An often overlooked problem in medical image segmentation research is the effective selection of training subsets to annotate from a complete set of unlabelled data. Many studies select their training sets at random, which may lead to suboptimal model performance, especially in the minimal supervision setting where each training image has a profound effect on performance outcomes. This work aims to address this issue. We use prototypical contrasting learning and clustering to extract representative and diverse samples for annotation. We improve upon prior works with a bespoke cluster-based image selection process. Additionally, we introduce the concept of unsupervised balanced batch dataloading to medical image segmentation, which aims to improve model learning with minimally annotated data. We evaluated our method on a public skin lesion dataset (ISIC 2018) and compared it to another state-of-the-art data sampling method. Our method achieved superior performance in a low annotation budget scenario.

Problem

• Effective training subset selection for medical image segmentation
• Improving model performance with minimal supervision setting
• Unsupervised balanced batch dataloading for medical image segmentation

Innovation

• Prototypical contrasting learning for sample selection
• Cluster-based image selection process
• Unsupervised balanced batch dataloading