Active Learning Pipeline for Biomedical Image Instance Segmentation with Minimal Human Intervention

2025-11-06

Bildverarbeitung für die Medizin

AI Summary

High annotation costs and poor domain-specific generalization of foundation models hinder practical deployment of instance segmentation in biomedical imaging. Method: We propose an active learning framework that synergistically integrates large foundation models with conventional neural networks. Specifically, foundation models generate high-quality pseudo-labels; an uncertainty- and diversity-aware core-set selection strategy identifies the most informative samples; and minimal human annotations guide automatic configuration and fine-tuning of nnU-Net. Contribution/Results: Our approach jointly leverages the zero-shot capability of foundation models and the data efficiency of active learning, substantially reducing annotation dependency. On multiple biomedical datasets, it achieves near fully supervised segmentation performance using ≤5% manual annotations. This yields a highly automated, transferable solution for low-resource scenarios, advancing scalable and cost-effective biomedical image analysis.

Technology Category

• Machine Learning: Active Learning
• Computer Vision: Segmentation
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Biomedical image segmentation is critical for precise structure delineation and downstream analysis. Traditional methods often struggle with noisy data, while deep learning models such as U-Net have set new benchmarks in segmentation performance. nnU-Net further automates model configuration, making it adaptable across datasets without extensive tuning. However, it requires a substantial amount of annotated data for cross-validation, posing a challenge when only raw images but no labels are available. Large foundation models offer zero-shot generalizability, but may underperform on specific datasets with unique characteristics, limiting their direct use for analysis. This work addresses these bottlenecks by proposing a data-centric AI workflow that leverages active learning and pseudo-labeling to combine the strengths of traditional neural networks and large foundation models while minimizing human intervention. The pipeline starts by generating pseudo-labels from a foundation model, which are then used for nnU-Net's self-configuration. Subsequently, a representative core-set is selected for minimal manual annotation, enabling effective fine-tuning of the nnU-Net model. This approach significantly reduces the need for manual annotations while maintaining competitive performance, providing an accessible solution for biomedical researchers to apply state-of-the-art AI techniques in their segmentation tasks. The code is available at https://github.com/MMV-Lab/AL_BioMed_img_seg.

Problem

• Reducing manual annotation requirements for biomedical image segmentation
• Combining strengths of foundation models and neural networks effectively
• Addressing performance limitations on specialized biomedical datasets

Innovation

• Leverages active learning and pseudo-labeling for segmentation
• Combines foundation models with nnU-Net self-configuration
• Minimizes manual annotation via representative core-set selection