Multi-Input Melanoma Classification using MobileNet-V3-Large Architecture

Abstract

All over the globe there exists a serious problem with skin cancer, most especially melanoma; a malignancy which is known to behave aggressively and able to metastasize. Detecting this early is the key to saving life. This study introduces a new method of classifying melanoma using an advanced model known as MobileNet-V3-Large. This technique differs from others in that it considers both the images of the skin lesion and tabular data including factors consisting of patient’s approximate age, gender and the location of the lesion on the body of the patient is considered here. Such an approach empowers the predictions of whether skin lesion may be malignant or benign. Tested on huge collection consisting of skin images combined with the tabular data, it was established that the method outperforms others already existing. The results of this study showed that a high accuracy of 99.56% was achieved using the proposed model. The study indicates that utilizing multi-input method will substantially enhance diagnosis for melanoma hence reducing mortalities in the future.