NIH Reporter #1P20GM161995-01

Explainable AI Platform for Skin Lesion Triage – Development, Data Collection, and Validation

This opportunity involves the National Institute of General Medical Sciences (NIGMS) funding the University of Hawaii at Manoa to develop an explainable AI platform for skin lesion triage. - Government Buyer: - National Institute of General Medical Sciences (NIGMS), National Institutes of Health - Project performed at University of Hawaii at Manoa, Honolulu, HI - OEMs and Vendors: - No specific OEMs are named - Prospective vendors include MedTech Innovations and SkinCare Solutions - Products/Services Requested: - Development of an explainable AI platform for skin lesion triage using full body 2D photography and AI - Data collection and annotation of skin lesion images (200 units) - Validation and testing of the AI platform (100 units) - Unique/Notable Requirements: - Platform must leverage computer vision and deep learning - Use of the ABCDE rule for lesion assessment - Incorporation of datasets from the International Skin Imaging Collaboration (ISIC) and new data from Hawaii - Adherence to NIH guidelines - Focus on improving diagnostic accuracy and explainability for melanoma and other skin cancers, especially in Hawaii's population

Description

This project aims to develop an explainable artificial intelligence (AI) platform to improve the diagnosis of skin cancer, particularly melanoma, by combining full body two-dimensional photography with AI techniques. The platform will leverage the ABCDE rule for lesion assessment and use datasets from the International Skin Imaging Collaboration and prospective data from Hawaii. The goal is to enhance diagnostic accuracy and address the lack of explainability in current AI models. The project includes development, data collection, annotation, validation, and testing phases.

View original listing