A Review - On Artificial Intelligence for Detecting and Preventing Non- Alcoholic Fatty Liver Disease

Abstract

Non-alcoholic fatty liver disease (NAFLD) is a growing global health problem. Liver biopsy is still considered the reference standard for diagnosis, but alternative non- invasive approaches are being proposed, including clinical scoring systems and conventional ultrasonography (US). Nevertheless, the utility of these non-invasive modalities, especially in the accurate characterization and quantitation of NAFLD, has been challenged. However, these limitations have recently made way for integrating artificial intelligence (AI) into Diagnosis Processes, successfully addressing some of the limitations. This study will analyze the performance and effectiveness of different AI techniques for the qualitative and quantitative evaluation of NAFLD, considering different algorithms' performance on ultrasound images. To that end, a systematic review was performed on this topic, narrowing its context to how artificial intelligence can improve the accuracy of non-invasive diagnostics. We conducted a systematic search of 5 major scientific bibliographic databases (PubMed, Association for Computing Machinery [ACM] Digital Library, IEEE Xplore, Scopus, and Google Scholar) to identify relevant studies. The results highlight the promise of AI to enhance the accuracy and consistency of NAFLD diagnosis. AI has proven to discover subtle features and measure hepatic fat content more sensitively and specifically than traditional goals using machine learning algorithms and deep learning models. Japan, and even making invasive procedures such as liver biopsy less necessary, thanks to the merging of artificial intelligence with classical diagnosis.