ABSTRACT
Melanoma is a skin cancer that causes the highest mortality among different types of cancers. Its early-stage detection and intervention increase the survival rate. The dermatologist often fails to detect/classify rashes, blisters, nevus or seborrheic keratosis, and melanoma by visualising the dermoscopy images. Different Artificial Neural Network models combined with image processing techniques were recently used to classify benign and malignant. However, its limitation is that the model’s accuracy depends on the extracted features from the image. Further, recently, Conventional Neural Networks (CNN) are being used in melanoma detection. It has the advantages of high classification accuracy and does not need features to be extracted. However, CNN’s highly data-driven property limits its application/deployment in a lite weight embedded device platform. Hence, developing a liter machine learning algorithm with expected accuracy is essential to deploy such automatic classifiers in an embedded platform. This chapter aims to design, develop a deployable melanoma detection model using a CNN. The model is easily accessible and tunable to assist patients, doctors, and the medical community. In this proposed work, the complexity of the network is reduced with 95.25.
