Abstract:
Early diagnosis and accurate identification of skin cancer together with dermatological diseases represent major worldwide health issues that require precise medical assessment for proper therapeutic interventions. A diagnostic model development project focuses on skin cancer identification by examining skin lesion images with high precision. The research uses MobileNetV2 as an advanced deep learning architecture to analyse image-based diagnostic models that detect different types of skin cancer. The study maintains uniqueness through its unification of machine learning techniques for skin cancer recognition with Ayurvedic remedy guidance which stems from Sri Lankan traditional medical practices. The research leverages a dermatology dataset which includes diverse 27200 labelled images of skin lesions under 10 skin diseases to support training data representation. The classification performance evaluation of MobileNetV2 outperforms ResNet50 and other models by achieving better accuracy and speed in model convergence. Experimental findings indicate that MobileNetV2 delivers 64.54% accuracy performance beyond CNN but demonstrates similar results to ResNet50 at 62.97% accuracy. The study connects the developed model with Ayurveda treatment suggestions using a rule-based system to develop an expanded method for complete skin cancer healthcare.
