A Generalized Model for Skin Cancer Cell Detection Using Integrated Circuit and Decision Framework (ICDF) to Avoid Metastasis
Soumen Santra, Dipankar Majumdar, Surajit Mandal and Arpan Deyasi
A novel model using convolution neural network has been developed for near accurate detection of skin cancer by carefully preprocessing the data and addressing class imbalance. 500 JPG images in daylight or infrared mode were collected between December 2017 and March 2018 using a newly developed hardware model that has been integrated with a storage device. Of these, 380 displayed skin melanomas, and 120 displayed a carcinoma factor. The model’s accuracy in training, testing, and validation was 91.14%, 91.33%, and 91.12%, respectively. The fact that negligible losses are documented in every instance supports the quality of the separation between healthy and diseased skin cells. The current study clearly outperforms peer-reviewed work conducted using CNN, with 92.39% sensitivity, 96.63% specificity, and 97.03% accuracy. This supports the work’s uniqueness. An infrared camera that is integrated with the CNN analysis storage device is linked to the model.
Keywords: Integrated circuit and decision framework, skin melanomas, carcinoma, convolution neural network, metastasis