Deep Learning Method for Melanoma Discrimination Using Blood Flow Distribution Images

We have developed a multipoint laser Doppler velocimeter (MLDV) that can measure blood flow velocity non-invasively. The device can acquire blood flow velocity in absolute value and image the blood flow distribution. Absolute values of blood flow velocity indicated a follow-up on the affected area. Therefore, we have performed a follow-up for melanoma and breast cancer. However, this device does not have the ability to determine whether a measurement site is cancerous or not. Thus, in this study, we built a deep learning system with blood flow distribution images as input and tested whether it can discriminate melanoma or not. The results showed that this technique was particularly effective in the early stages of the disease when no abnormalities were found on the skin surface. (c) 2021 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.