Gastroenterology Meets Machine Learning: Status Quo and Quo Vadis

Table 6

GE activities using ML.

Aim of study	Number of studies	Application/Description
Disease classification and discrimination	27	The usage of ML in disease classification is very frequent. Indeed, as ML systems are capable to analyze large volumes of patient data, they can, efficiently and accurately, correlate these features with some disease state. This is particularly useful for difficult-to-diagnose diseases, such as celiac disease which involves multiple clinical presentations and symptoms shared with other diseases. ML ability to accurately classify disease states (present/absent), etiology, and subtype allows subsequent investigations, treatments, and interventions to be delivered in an efficient and targeted manner.
Risk stratification	17	The accurate assessment of a patient’s risk of adverse events remains a mainstay of clinical care; MLTs form an attractive platform to build risk metrics because they can easily incorporate disparate pieces of data, yielding classifiers with improved performance.
Endoscopic imaging examination	16	Endoscopic procedures generate a large amount of images in one examination of a patient. It is hard for clinicians to leave continuous time to examine the full endoscopic images. Thus, the use of ML to assist in endoscopic imaging examination tasks represents a response to the urgent need for new technologies to supplement existing imaging techniques.
Early detection of cancer	7	Early identification of cancer is challenging because symptoms are non-specific (or absent) and compounded by overlap with symptoms of other diseases. That is why ML has emerged as a promising technique for handling complex interactions of high-dimensional medical data related to cancerology tasks.
Survival prediction	7	Survival probability prediction is one important problem encountered in medical studies when the primary endpoint of interest is time to an event. An accurate survival probability prediction can provide a useful tool for selecting prevention and treatment strategies. Thus, considerable studies in the reviewed literature have introduced MLTs as a rapid and reliable technique to predict survival.
Others tasks	14	Other applications of MLTs that have been studied in literature with promising results include drug development and treatment planning (6 studies), endoscopy or surgery candidate selection (4 studies), and surgical/clinical outcomes prediction (4 studies).