Automated AI Skin Medical Analysis using Deep Learning

Computer Engineering

Skin medical lesions are various abnormalities on the skin that appear on a patient due to many different reasons. These can be due to the exposure of skin to harmful UV rays or they can be mere birthmarks or an uncontrolled growth in the skin tissue, defined as cancer. Not all skin lesions are cancerous and therefore it is important to identify them at an early stage which can potentially improve the survival rate in case the pigment on skin turns out to be cancerous. The proposed system helps the user to identify if the spots on the skin are cancerous or benign by using their mobile devices. The model was trained on HAM10000 ("Human Against Machine with 10000 training images") dataset using convolutional neural network. The dataset comprised of seven different classes of skin lesions and augmentation was used to increase the number of images in the dataset. Due to limited number of images in the dataset, to achieve better results, the model was trained using transfer learning with MobileNet adapted for mobile use. The model was then integrated into the android application which could be used to detect whether the skin lesion is cancerous or benign using the camera of the mobile device. The proposed system can be used to detect if the spots on skin are cancerous or not by using the mobile device of the user.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.