Machine Learning for Melanoma Detection
Authors: Trevor Cabreros, Mee Her, Jaskiran Kalsi
Advisor: Dr. Matthew Leineweber
Melanocytes are melanin-producing cells responsible for pigmentation. Deep in the skin, melanocytes can deteriorate from UV exposure which could result in lesions. These damaged tissues may develop into melanoma, a dangerous form of skin cancer, and are identifiable based on their visually-distinct features. A doctor’s judgement is typically used to detect melanoma, however, medical opinions can vary due to individual bias and potentially lead to misdiagnoses. To mitigate such occurrences, we developed machine learning algorithms to remove ambiguity by using vast amounts of data to systematically reach a diagnosis. The k-Nearest Neighbor and Support Vector Machine classifiers are implemented along with MATLAB’s Imaging Toolbox to evaluate abnormal melanocytic tissue. The software differentiates between benign and malignant lesions and may allow for optimal treatments. Thus, a supplemental approach is created that provides a second medical opinion. Additionally, this method opens doors for diagnosing other cutaneous-related diseases.