CASC-AI: Consensus-aware Self-corrective AI Agents for Noise Cell Segmentation

2025-02-11

AI Summary

Existing methods for multi-class cell segmentation in high-resolution whole-slide images (WSIs) struggle to handle pixel-level annotation noise introduced by non-expert annotators, leading to substantial increases in false positives (FP) and false negatives (FN). To address this, we propose a consensus-aware self-correcting AI agent framework. Our method introduces a novel dynamic supervision mechanism based on a consensus matrix, integrating feature-similarity-driven adaptive weighting and pixel-wise contrastive learning to decouple representations of noisy versus reliable regions. Supervisory signal quality is further enhanced via multi-stage iterative label purification. Evaluated on real-world non-expert-annotated WSIs and two types of synthetically corrupted datasets, our approach achieves significant improvements in segmentation accuracy while effectively suppressing FP and FN errors. The source code and annotations are publicly available.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Multi-class cell segmentation in high-resolution gigapixel whole slide images (WSI) is crucial for various clinical applications. However, training such models typically requires labor-intensive, pixel-wise annotations by domain experts. Recent efforts have democratized this process by involving lay annotators without medical expertise. However, conventional non-agent-based approaches struggle to handle annotation noise adaptively, as they lack mechanisms to mitigate false positives (FP) and false negatives (FN) at both the image-feature and pixel levels. In this paper, we propose a consensus-aware self-corrective AI agent that leverages the Consensus Matrix to guide its learning process. The Consensus Matrix defines regions where both the AI and annotators agree on cell and non-cell annotations, which are prioritized with stronger supervision. Conversely, areas of disagreement are adaptively weighted based on their feature similarity to high-confidence agreement regions, with more similar regions receiving greater attention. Additionally, contrastive learning is employed to separate features of noisy regions from those of reliable agreement regions by maximizing their dissimilarity. This paradigm enables the AI to iteratively refine noisy labels, enhancing its robustness. Validated on one real-world lay-annotated cell dataset and two simulated noisy datasets, our method demonstrates improved segmentation performance, effectively correcting FP and FN errors and showcasing its potential for training robust models on noisy datasets. The official implementation and cell annotations are publicly available at https://github.com/ddrrnn123/CASC-AI.

Problem

• Handles annotation noise adaptively
• Improves multi-class cell segmentation
• Corrects false positives and negatives

Innovation

• Consensus Matrix guides AI learning
• Contrastive learning separates noisy features
• Iterative refinement enhances label robustness