Globally, liver cancer is considered one of the most prevalent and significant causes of cancer-related mortality. Techniques for automatically segmenting and classifying liver tumors are crucial for supporting physicians during the tumor diagnosis procedure. Strong classification algorithms can alter a wide range of real-world applications that are becoming available due to the development of artificial intelligence. Classifying liver tumors is a challenging task because of the significant appearance, diversity, noise, and non-homogeneity observed in cancer tissue. Computed tomography (CT) examinations can be used to guide biopsies and other easily determined procedures, as well as to plan and manage tumor treatments properly. For many CT images, manual segmentation and classification are laborious and time-consuming procedures. Computer-aided diagnosis (CAD) systems play a critical role in the early detection of liver disease, which lowers the mortality rate of liver cancer. In recent decades, a wide range of advanced techniques for autonomous liver segmentation have been developed by researchers. Yet, the most difficult procedure is segmenting a liver tumor due to the wide variety in the tumor’s size, boundary depth, and position with other organs around the liver. This survey aims to analyze the different kinds of techniques utilized for the segmentation and classification process of liver tumors. The performance measures of those techniques, such as mean square distance, dice coefficient, volume overlap error (VOE), and accuracy, are calculated. The input images considered in each work, like CT and magnetic resonance imaging (MRI) scans, are analyzed. The datasets, methods, and algorithms used by each study for the segmentation, as well as the classification process, are categorized. Furthermore, the experimental findings, advantages, and drawbacks of each liver tumor detection research are analyzed and classified accordingly. Finally, the model helps the researchers to get precise outcomes in the field of liver tumor segmentation and classification; various challenges in the existing works are discussed.
Dharaneswar et al. (Thu,) studied this question.