Investigating usefulness and usability of medical content-based image retrieval tool for skin lesion images

Content-Based Image Retrieval (CBIR) is an application of computer vision techniques for searching an existing database for visually similar entries to a specific query image. One application of CBIR in the dermatology domain is displaying a set of visually similar images with a pathology-confirmed diagnosis for a given query image of a skin lesion. Recently, CBIR algorithms using machine learning techniques have gained more attention; however, we lack insights into how interactive CBIR decision support tools are actually perceived by end users. We present the design and evaluation of a CBIR user interface for dermoscopic skin images, and investigate users' classification accuracy, and users' confidence, trust, timing, and the system's educational value. Our study with 34 non-medical users indicates that CBIR enables users to make significantly more accurate and confident classifications on new skin lesion images from four categories commonly observed in clinical practice: nevus, seborrheic keratosis, basal cell carcinoma, and malignant melanoma.