Artificial intelligence applications in the diagnosis of gallbladder neoplasms through ultrasound: A review

The occurrence of cancer in the gallbladder is infrequent. However, it is an aggressive disease that is often diagnosed at a late stage. Ultrasound imaging is a common first-line diagnostic tool for gallbladder diseases, but its accuracy relies on the expertise of the sonographer and the radiologist. The use of AI techniques, such as machine learning and deep learning, can enhance the diagnostic accuracy and efficiency from ultrasound imaging, serving as a supplementary diagnostic tool. This paper aims to address the existing gap in reviews on the application of AI in diagnosing gallbladder malignancies using ultrasound images. It provides insights into current trends in this field and suggests directions for future research. From the reviewed studies, it appears that despite the promising results, several challenges persist. These include the lack of large and comprehensive datasets, scarcity of publicly available datasets, and questions regarding the robustness, generality and reliability of AI models, which affect the models' practicality. In addition, the YOLOv8 model is evaluated as the object detector in the methodology pipeline of one of the reviewed papers. A fusion method that combines the bounding boxes of Faster R-CNN and YOLOv8, leveraging the benefits of both techniques, is also presented. By using the bounding boxes from the proposed fusion method, superior classification performance was obtained with an accuracy of 92.62%. This outperformed the individual use of Faster R-CNN and YOLOv8, which yielded accuracies of 90.16% and 82.79%, respectively.