Convolutional Neural Networks for Malignant and Benign Cell Classification using Dermatoscopic Images

In general, cancer is considered as the second most fatal disease across the globe. In the past decades, skin cancer is increasing at a high rate as it is a major type of cancer. Diagnosis of unknown pores and skin cancer is very significant for delivering proper treatment. Curable with early analysis, simplest extraordinarily trained dermatologists that capable of appropriately understand skin lesions as malignant or benign. Skin disease is analyzed by utilizing clinical screening, dermoscopic investigation, histopathological assessment and a biopsy. As number of biomedical images collected and stored are rapidly increasing so a need of a technology to meet the requirement. As understanding is in restrained supply, systems classify skin lesions automatically as of benign or malignant cancer are beneficial as preliminary screening gear. The proposed methodology presents a CNN model, trained on feature extracted from a highway CNN pretrained on images of dermoscopic of pores and skin lesions. Further, the model is optimized using two different optimizers i.e., RMSPROP and ADAM individually and then comparison of accuracies is done, and conclusion would be based on the good optimizer. Further, it doesn't price a good deal for computational energy in order to teach the model. It is a basic scratch model, in which this research work has worked only upon one DL algorithm and then in future, different DL algorithms will be analyzed to achieve a better accuracy.