Multi-Scale Cascaded Spatial Segmentation Transformer for Liver Cancer Classification

Early and accurate detection is crucial for treating liver cancer, the main cause of cancer deaths. Despite its widespread use, Computed Tomography (CT) imaging generally struggles with liver tumors' low contrast, uneven borders, and overlapping features. The variety in tumor forms, sizes, and complicated anatomical aspects makes CT image liver cancer segmentation and categorization difficult. Variability in tumor size and shape, low contrast, overlapping structures, and complex liver anatomy are some of the difficulties that this method aims to address when using CT images to diagnose liver cancer. The Multi-Scale Cascaded Spatial Segmentation Transformer (M-SCSST) is an innovative approach developed for the Classification of Liver Cancer from CT Images that is introduced in this research. The M-SCSST uses a cascaded processing approach to include multi-scale spatial information into its transformer-based architecture. Accurate segmentation and classification of complicated and heterogeneous liver cancers in CT images are made possible by enhancing the detection of subtle features by utilizing advanced attention mechanisms (AAM) . Improved diagnostic accuracy is achieved by employing the suggested M-SCSST method on a large dataset of CT images of liver cancer. Its use helps radiologists identify cancerous from benign areas, which leads to earlier diagnosis of liver cancer and better treatment choices. The effectiveness of M-SCSST with CT scans is assessed through comprehensive simulation research. Research measures include precision, recall, computational efficiency, noise resilience, and classification accuracy. With improved accuracy and reliability, M-SCSST detects liver cancer from CT scans more effectively than conventional segmentation and classification approaches.