Mobile Application to Detect the Most Common Forms of Skin Cancer Through Computer Vision and Machine Learning

Skin cancer is the most common form of Cancer in the United States.  The deadliest form, melanoma, is curable if detected early. Squamous cell and basal cell carcinoma can cause death and disfigurement if left untreated. People living in rural areas depend on primary care physicians as the first step toward a health related diagnosis.  Physicians in rural areas are less likely to have specialized training and the equipment needed for the timely diagnosis of skin cancer.

SCIDS, the Skin Cancer Identification System, is a mobile application to assist with the early diagnosis of skin cancer by evaluating the characteristics of asymmetry, border irregularity, color, and size via captured images of lesions. These are commonly referred to as the ABCD’s of skin cancer. SCIDS combines the technologies of computer vision and machine learning in an Android mobile device utilizing the on-board camera and computational abilities to evaluate a skin lesion. The process involves 1) pre-assessment, 2) image acquisition, 3) image processing, and 4) classification and data analysis. The resulting analysis supplies information on the patient’s lifetime risk for skin cancer, the values of the ABCD’s and the results of the classification of the lesion as benign or malignant. We expect to have a working version that evaluates asymmetry, border irregularity and size by October 2012.

A hand held mobile device application that accurately evaluates data relevant to the diagnosis of skin cancer would be beneficial to people living in rural areas where accelerated diagnosis is directly related to improved outcome.