Deep Learning Model to Predict Sentinel Lymph Node Status in Melanoma Patients

The study by T. Okamoto from the Department of Dermatology, Yamanashi Daigaku Igakubu Daigakuin Sogo Kenkyubu Igakuiki, Japan, aimed to develop an artificial intelligence (AI) model that can predict the prognosis of malignant melanoma patients, specifically their sentinel lymph nodes (SLNs) status, using histopathological images of melanoma. The study used around 400 skin cutaneous melanoma samples from The Cancer Genome Atlas database for training and validation. The model was trained using an imagenet pretrained convnet followed by an attention-pooling layer, and the accuracy was evaluated using the area under ROC curve (AUROC). The best prediction AUROC was 0.65. The model is currently being validated with a dataset from the University of Yamanashi to improve its accuracy. The results suggest that the histological features of the primary melanoma can somewhat predict the status of lymph node metastasis.