OBJECTIVES : This review aimed to examine the literature, in particular, focusing on the extent to which the AI and ML approaches have been employed in clinical research in oncology.

METHODS : A targeted literature review was conducted from database inception through February 12, 2020. Embase^® and PubMed were searched to identify studies that (1) contained a comprehensive description of any AI and ML functionality in healthcare; (2) focused on any cancer indication; and (3) were conducted in a randomized controlled (RCT) setting.

RESULTS : A total of 1,979 citations were screened. We identified 74 RCTs evaluating different applications of AI and ML in oncology. Prostate cancer was the most frequently explored (n=14 studies), followed by colorectal cancer (n=13) and lung cancer (n=8). More rarely explored were cancers of the breast (n=5 studies); esophagus and lymphocytes (n=4 each); nasopharynx and kidney (n=3 each); stomach, cervix, skin, pancreas, head & neck (n=2 each); and brain, bladder, and liver (n=1 each). We identified three major categories where algorithm tools were tested, i.e., prediction (n=37 studies), diagnosis (n=29 studies), other models (n=8 studies). Most of the studies were primarily aimed at the identification of biomarkers, risk stratification, disease severity, staging, and associated complications. Artificial neural network was the most commonly utilized algorithm (n=22 studies), followed by support vector machines (n=16), convolutional neural network (n=11), random forest (n=7), elastic net (n=5), regression (n=4), decision tree (n=2), and LASSO (n=1). Also, the number of published studies evaluating AI and ML increased by 150% between 2014 and 2020.

CONCLUSIONS : ML models and AI are promising in the field of oncology. However, its application is currently limited and still grounding on diagnostic and prognostic services. The review highlights that more efforts are needed to facilitate the implementation of relevant ML studies for testing new biomedical treatments for efficacy and safety and expanding their access.