Task-driven framework using large models for digital pathology
Jiahui Yu, Tianyu Ma, Feng Chen, Jing Zhang, Yingke Xu · Communications Biology · 2024
Microscopy is an indispensable tool for collecting biomedical information in pathological diagnosis, but manual annotation, measurement and interpretation are labor-intensive and costly. Here, we propose a task-driven framework powered by large models that excel in visual analysis and real-time control, paving the way for the next generation of microscopes. We achieve proof-of-concept success on clinical tasks, specifically in adaptive analysis of H&E-stained liver tissue slides. This work demonstrates the advanced capabilities for future digital pathology, setting a new standard for intelligent, efficient, and real-time analysis in clinical applications. A large model-powered smart microscope framework is developed to achieve adaptive decision-making and automated analysis by responding to the pathological features, accelerating the diagnostic paradigm of future digital pathology.