Zebra Optimization Algorithm with Convolutional Neural Network for Detecting Subtle Changes in Skin Pigmentation for Melanoma Screening

Melanoma is easily detectable by visual examination since it occurs on the skin surface. Traditionally approach inefficiently to capture to diagnosis, it need to facilitate the diagnosis process while still yielding an accurate of diagnosis. This paper, Zebra optimization Algorithm with Convolutional Neural Network (ZOA-CNN) combined to performed in the classification skin pigmentation for melanoma screening achieve better accuracy. The ZOA is used to optimize the CNN technique enhance the classification accuracy. ZOA balances exploration as well as exploitation that supports the CNN avoid local minima and attain best solutions at the time of training. This flexibility enhances the robustness of the model in detecting melanoma. Initially data obtained from the ISIC 2019 dataset and pre-processing involved the Histogram Equalization and Convert RGB images to grayscale images, which is enhance the image quality. Segmentation utilized U-net architecture includes skip connections that allow high resolution feature from the encoder and decoder path. Feature extraction using MobileNet is designed for efficiency to extracted the high level feature and depthwise separable convolutions reduces the amount of parameters while maintaining the performance. The performance of proposed model evaluated the various metrics and it attains the better accuracy of 0.91 compared to GoogleNet.