A Novel Dicepn-Clun-Based Pitting/Crusting Ratio Estimation and Multiclass Skin Cancer Classification Using Dermoscopy Images

The patient’s lifetime is significantly increased by an early detection of Skin Cancer (SC). Nevertheless, none of the existing works concentrated on analyzing the pitting/crusting ratio. Therefore, this paper proposes an effective Dermoscopy, Deep Incremental Convolutional Elastic-tanh Pooling Neural- Cosinu-sigmoidal Linear Unit Network (DICEPN-CLUN)-based pitting/crusting ratio estimation and multiclass SC classification employing dermoscopy images. Primarily, the International Skin Imaging Collaboration-2019 (ISIC-2019) dataset is gathered and then pre-processed. Afterward, the hair removal process is done, followed by lesion segmentation. Likewise, from the segmented lesions, the 3D heat map is constructed. Similarly, from the dataset, the Metadata is extracted, followed by data pre-processing. Then, the features are extracted. In the meantime, the pitting/crusting region identification and pitting/crusting ratio estimation are carried out. An effective Fuzzy Rational Quadratic Weibull Inference System (FRQWIS) is established to identify the PCR. Lastly, the eight categories of SC are efficiently classified by the proposed DICEPN-CLUN. Hence, the proposed work obtained better outcomes with 99.9046% accuracy.