Advanced Liver Tumour Detection Using Optimized YOLOv8 Modules

Health fraternity is invariably challenged with early diagnosis, detection, identification, classification, treatment and convalescence of globally prevalent and life-threatening fatal diseases as liver cancer. The early detection of liver cancer through medical image processing technique is so challenging that an iota of deviation conspicuous among healthy tissues, benign tumour and malignant tumour tissues is a matter of wake up call. This work is entailed with introduction of a novel, optimized YOLOv8-based model for liver tumour detection, harnessing the strengths of transformer-based feature extraction, global attention mechanisms, and advanced feature aggregation techniques. The model was subjected to rigorous performance with relevant methods and messages as parameters time and again for repeated refinements. Eventually, it was concluded that the proposed model surpasses all the models in extant now in terms of precision, recall, and means average precision (mAP). This is ascertained by inference drawn from the model’s achievement of attaining 95.34% precision, 96.49% recall, and 97.31% [email protected]. In regard to tumour classification, the proposed model excels in differentiating normal cases, benign tumours, and malignant tumours. These innovations represent a significant step toward improving the accuracy of automated liver tumour diagnosis systems, with the potential to revolutionize clinical workflows and enhance patient outcomes.