Convolutional Neural Network for Monkeypox Detection

Machine learning has been implemented in medical applications, especially in classification models to support diagnosis. In dermatology, it is of great relevance, due to the high difficulty in differentiating between pathologies that are similar, such is the case of its wide application in skin cancer. One of the diseases that has recently become relevant due to a recent outbreak is monkeypox, which is an exanthematic disease; these types of pathologies are very similar if you are not an expert, so diagnostic support would favor their identification, mainly for adequate epidemiological control. Therefore, the objective of this work is use a public database of monkeypox and control group images. These images were preprocessed, divided into 80/20 for training and testing set respectively. Implementing MiniGoggleNet, 6 experiments were carried out, with different number of epoch. The best model was the one of 50 epochs with accuracy of 0.9708, a loss function of 0.1442, an AUC for class 0 of 0.74, AUC for class 1 of 0.74, AUC for micro-average of 0.76 and AUC for macro-average of 0.74.