Intelligent medical image grouping through interactive learning

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• Image grouping in specialized fields like dermatology is complex, requiring significant human expertise.
• Traditional machine learning struggles to integrate expert knowledge effectively.
• Manual annotation of medical images is time-consuming and often insufficient.

Purpose of the Study:

• To develop an interactive machine learning paradigm for automated, interpretable grouping of dermatological images.
• To integrate domain expertise directly into the machine learning model's training process.
• To enhance the accuracy and efficiency of medical image analysis.

Main Methods:

• An interactive machine learning framework was designed, enabling expert input.
• Dermatologists provided domain knowledge by grouping a small subset of images.
• A learning algorithm incorporated expert-defined groupings as constraints for dataset reorganization.

Main Results:

• The developed paradigm effectively improved image grouping based on expert input.
• The interactive loop of model computation, visualization, and expert feedback enhanced grouping quality.
• Demonstrated the feasibility of integrating human expertise into automated image analysis.

Conclusions:

• Interactive machine learning offers a powerful solution for knowledge-rich image grouping tasks.
• This approach bridges the gap between computational analysis and expert medical knowledge.
• The paradigm facilitates more accurate and interpretable dermatological image classification.