Revolutionizing Cancer Detection Diagonalizing with Texture Features Using the TexDiagNet Model

Cancer diagnosis using medical imaging has witnessed significant advancements with the integration of deep learning. This study proposes TexDiagNet, a novel deep learning model optimized for texture feature extraction and classification in medical images, specifically chest CT scans. Using the publicly available Chest CT-Scan Images Dataset from Kaggle, the proposed method applies advanced preprocessing techniques to enhance image quality and emphasize regions of interest. TexDiagNet achieved an impressive accuracy of 97.10%, outperforming nine existing models, including NASNet (95.40%) and EfficientNet (93.80%). The model's training and validation losses were notably lower (0.10 and 0.12, respectively), demonstrating superior generalization capabilities. In addition to accuracy, the model excelled in other performance metrics such as precision (96.90%), recall (96.50%), and F1-score (96.70%). TexDiagNet's computational efficiency was also highlighted, processing images at 100 frames per second. These results suggest that TexDiagNet is a highly effective tool for diagnostic applications in cancer detection.