IoT based Mobile App for Skin Cancer Detection using Transfer Learning

Skin cancer is a significant health concern, approximately 125,000 fresh melanoma issues diagnosed each year. As a kind of cancer, skin cancer is extremely hazardous. Variations in the skin’s genetic material, generated by damaged and not repaired deoxyribonucleic acid (DNA), are the leading cause of skin cancer. Early detection is key since skin cancer has a tendency to migrate (transfer to other portions of the body over time) and is therefore less treatable in later stages. Dermatoscopy, clinical assessment, and other visual approaches were among the first used in the identification of skin lesions. It has been observed that the diagnosis accuracy of skin lesions can be negatively affected by the presence of novice physicians. The death rate from skin cancer can be reduced if it is detected early. Earlier research has shown that Deep Learning outperforms human specialists on a wide range of vision-based tasks. In this study, an IoT-based smartphone app is developed in collaboration with DL to enable the automatic identification of skin cancer. Transfer Learning DL models, including VGG-16 and AlexNet, are built with the HAM10k data. Techniques like reshaping and augmenting the data are used to eliminate the data imbalance. When comparing the two DL models for cancer prediction, the metrics show that AlexNet performs better. Finally, the AlexNet model is deployed in the development of the smartphone app, bringing the results of this study into the hands of the general public.