A co-evolving agentic AI system for medical imaging analysis

2025-09-24

AI Summary

Current AI systems for medical imaging suffer from fragmented ecosystems, limited tooling, and absence of real-time expert feedback. To address these challenges, we propose TissueLab—a co-evolving multi-agent system enabling natural-language querying, automated generation of interpretable analytical workflows, and seamless integration of multimodal tools from pathology, radiology, and spatial omics. TissueLab introduces a domain-agnostic “tool factory” mechanism with standardized interfaces, vision-language-model-driven workflow synthesis, and an active learning loop incorporating real-time expert validation. This design enables rapid, zero-shot adaptation to novel clinical scenarios without large-scale retraining. Experiments demonstrate state-of-the-art performance across multiple quantitative benchmarks; notably, TissueLab generalizes to unseen disease conditions within minutes. The framework—including code, tools, and ecosystem—is fully open-sourced.

Technology Category

• Computer Vision: Medical and Biological Imaging
• Machine Learning: Multimodal Learning
• Multiagent Systems: Teamwork

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Agentic AI is rapidly advancing in healthcare and biomedical research. However, in medical image analysis, their performance and adoption remain limited due to the lack of a robust ecosystem, insufficient toolsets, and the absence of real-time interactive expert feedback. Here we present "TissueLab", a co-evolving agentic AI system that allows researchers to ask direct questions, automatically plan and generate explainable workflows, and conduct real-time analyses where experts can visualize intermediate results and refine them. TissueLab integrates tool factories across pathology, radiology, and spatial omics domains. By standardizing inputs, outputs, and capabilities of diverse tools, the system determines when and how to invoke them to address research and clinical questions. Across diverse tasks with clinically meaningful quantifications that inform staging, prognosis, and treatment planning, TissueLab achieves state-of-the-art performance compared with end-to-end vision-language models (VLMs) and other agentic AI systems such as GPT-5. Moreover, TissueLab continuously learns from clinicians, evolving toward improved classifiers and more effective decision strategies. With active learning, it delivers accurate results in unseen disease contexts within minutes, without requiring massive datasets or prolonged retraining. Released as a sustainable open-source ecosystem, TissueLab aims to accelerate computational research and translational adoption in medical imaging while establishing a foundation for the next generation of medical AI.

Problem

• Limited performance and adoption of agentic AI in medical image analysis
• Lack of robust ecosystem and real-time interactive expert feedback
• Insufficient toolsets for explainable workflows across medical imaging domains

Innovation

• Co-evolving AI system integrating pathology, radiology, omics tools
• Automatically plans explainable workflows with real-time expert feedback
• Uses active learning for rapid adaptation without retraining