Deep Learning-Based Detection of Squamous Cell Carcinoma with Statistical Validation

Squamous cell carcinoma is a form of cancer that arises in multiple organs within the human body. The manual examination of histopathological images for the diagnosis of squamous cell carcinoma is a time-intensive and laborious task that often requires specialized expertise. Although several deep learning-based approaches have been introduced to classify these images as either cancerous or non-cancerous, there has been relatively little focus in existing research on differentiating between the specific subtypes of squamous cell carcinoma, particularly keratinized and non-keratinized forms. In this study, we present a custom-designed convolutional neural network model that classifies input images into three distinct categories: keratinized squamous cell carcinoma, non-keratinized squamous cell carcinoma, and normal tissue. To enhance the interpretability of the model's predictions, Gradient-weighted Class Activation Mapping is employed. Furthermore, the performance of the proposed model is thoroughly evaluated using statistical methods to ensure the reliability of the results.