Promptable cancer segmentation using minimal expert-curated data

2025-05-23

AI Summary

Medical image cancer segmentation heavily relies on large-scale, high-quality annotated data; however, expert annotation is costly and suffers from substantial inter-observer variability, hindering clinical deployment. Existing weakly supervised and promptable segmentation methods either require massive paired pathology–imaging datasets or fail to generalize robustly to lesion regions. To address this, we propose a dual-classifier collaborative prompting framework that achieves accurate single-point-prompt-driven segmentation using only 24 fully annotated and 8 weakly annotated images. By synergistically integrating fully supervised and weakly supervised signals, our method establishes a multi-level classifier collaboration mechanism for prompt-guided region search. In prostate cancer segmentation, it matches the performance of fully supervised models and significantly outperforms state-of-the-art promptable approaches (e.g., SAM), reducing annotation requirements to just 1% of those needed by comparable methods.

Technology Category

• Computer Vision: Segmentation
• Natural Language Processing: Prompt Engineering / Prompting
• Machine Learning: Semi-Supervised Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Automated segmentation of cancer on medical images can aid targeted diagnostic and therapeutic procedures. However, its adoption is limited by the high cost of expert annotations required for training and inter-observer variability in datasets. While weakly-supervised methods mitigate some challenges, using binary histology labels for training as opposed to requiring full segmentation, they require large paired datasets of histology and images, which are difficult to curate. Similarly, promptable segmentation aims to allow segmentation with no re-training for new tasks at inference, however, existing models perform poorly on pathological regions, again necessitating large datasets for training. In this work we propose a novel approach for promptable segmentation requiring only 24 fully-segmented images, supplemented by 8 weakly-labelled images, for training. Curating this minimal data to a high standard is relatively feasible and thus issues with the cost and variability of obtaining labels can be mitigated. By leveraging two classifiers, one weakly-supervised and one fully-supervised, our method refines segmentation through a guided search process initiated by a single-point prompt. Our approach outperforms existing promptable segmentation methods, and performs comparably with fully-supervised methods, for the task of prostate cancer segmentation, while using substantially less annotated data (up to 100X less). This enables promptable segmentation with very minimal labelled data, such that the labels can be curated to a very high standard.

Problem

• Automated cancer segmentation reduces costly expert annotations
• Weakly-supervised methods need large paired histology-image datasets
• Promptable segmentation struggles with pathological regions without big datasets

Innovation

• Promptable segmentation with minimal expert-curated data
• Combines weakly and fully-supervised classifiers
• Guided search process from single-point prompt