Advances in Medical Image Processing for Liver Tumour Recognition: A Comprehensive Survey

The last several decades have seen a lot of activity in the medical image processing domain especially when it comes to the segmentation of liver and liver tumours. The survival rate of liver cancer is rather low when compared to some other types of cancer. Nevertheless, early detection greatly improves the prognosis of liver cancer patients. As a consequence, multiple researchers have developed automated deep-learning (DL) systems for forecasting the development of cancer cells using various medical imaging modalities. A summary of some previously published studies on liver cancer diagnosis is currently lacking in review publications. But new mechanisms and architecture in the prognosis of liver cancer were beyond the possibility of these investigations. This review emphases on the topologies of DL for liver cancer diagnosis. The survey that follows examines the datasets used, describes existing DL-based designs, assesses the benefits and drawbacks of earlier studies, and discusses image processing techniques. Furthermore, a thorough examination of several imaging modalities, performance metrics and results, challenges and future research goals are provided. The comparison study of different methods reveals that HFCNN has the highest accuracy, at 97.22%.