Application of deep learning in the prognosis of liver cancer patients: A systematic review

Background and aims: Prediction of survival after the treatment of liver cancer has been vastlyinvestigated, yet remains deficient. Deep learning has demonstrated its capability to recognizespecific features that can prognosis of liver cancer patients. Artificial intelligence is rapidlyemerging because of the ability to process large amounts of data and find hidden connectionsbetween variables. Artificial intelligence and deep learning are increasingly used in several topicsof liver cancer research, including diagnosis, pathology, and prognosis. The purpose of this articleis to assess the role of deep learning in the prediction of survival following liver cancer treatment.Method: A systematic review of the published literature focused on the prognostic impact of deeplearning in the management of liver cancer was undertaken. Databases PubMed and the Web ofScience and research books were systematically searched using the words “artificial intelligence”,“deep learning” and “liver cancer” (and synonyms). English language articles were retrieved,screened, and reviewed by the authors. The quality of the papers was assessed using the risk ofbias In the Non-randomized Studies of Interventions tool. Data were extracted and collected in adatabase.Results: Among the ۳۸۷ articles screened a total of ۱۲۷ studies reported on the use of deep learningin liver cancer. Among these articles, only ۹ (۷.۱%) studies referred to the employ of deeplearning in the prediction of survival among patients with liver cancer and were included in thisreview. Other studies using deep learning in liver cancer were excluded; specifically, these studiesreported on the employ of deep learning for the diagnosis of the tumor (n = ۷۶, ۵۹.۸%), identificationof specific genes or pathways (n = ۱۷, ۱۳.۴%), prediction of tumor response after therapy(n = ۱۶, ۱۲.۶%), and the prediction of pathological aspects (n = ۹, ۷.۱%). All studies included inthe analytic cohort were published in the last decade.Conclusion: Deep learning used for survival prediction after liver cancer treatment provided enhancedaccuracy compared with conventional linear systems of analysis. While a few limitationshave been identified in these studies, there was an optimal level of accuracy of the deep learningused in the prognosis of liver cancer patients. Improved transferability and reproducibility will facilitatethe widespread use of deep learning methodologies, so healthcare providers are suggestedto take advantage of deep learning capabilities to predict liver cancer patients’ survival.