Xception Spiking Fractional Neural Network for Oral Squamous Cell Carcinoma Classification Based on Histopathological Images

Oral Squamous Cell Carcinoma (OSCC) is the leading cancer caused due to the immense growth of malignant cells in different regions of the oral cavity particularly in the area of the neck and head. The habits of smoking, betel nuts, and chewing tobacco that mainly affect the pharynx, nasopharynx, viral infection, and oral cavity regions are the major causes of OSCC. OSCC has to be detected at the initial stages to provide accurate treatment and to assess the severity. Thus, various screening and detection models are used to accurately detect OSCC with great significance for reducing morbidity and mortality. In this research, a novel deep learning model Xception Spiking Fractional Neural Network (XSFN-Net) is introduced to classify OSCC from the histopathology images. Here, the histopathology images are enhanced initially using the Medav filter. The pixel-wise image segmentation is performed using Parallel Reverse Attention Network (PraNet) along with the designed loss function from the enhanced image. Later, the extraction of relevant features from the segmented output and the OSCC is finally classified using XSFN-Net from the resultant extracted features. The XSFN-Net attained high performance with True Positive Rate (TPR) of 96.604%, accuracy of 97.538%, and True Negative Rate (TNR) of 95.545%.