Applied AI in CDSS in Medicine: A Systematic Review

Objective: Clinical decision support systems (CDSS) are continuously developing to solve medical problems and try to improve healthcare management, which has shown a significant result in reducing medical errors and improving multiple healthcare processes. These days, artificial intelligence (AI) becomes more influential in healthcare supporting physicians to make a clinical decision. 

Materials and Methods: A systematic review was conducted to identify articles related to CDSS using AI algorithms. The original research was published between 2009 and 2019 in the English language. In a total of 3,687 identified articles, 1,112 articles were analyzed, and 199 articles are represented within this review.

Results: The top three clinical areas related to CDSS using AI algorithms in medicine were Neurology (N=78), Cardiology (N=73), and Oncology (N=48). After analyzing these three main specialties, we found that the tasks of CDSS could be grouped into diagnosis, prognosis, treatment optimization, clinical workflow, and screening. The top five common algorithms were support vector machine (SVM), neural network (NN), decision tree (DT), random forest (RF), and other classifiers. Interestingly, 79.4% of those algorithms were unexplainable AI.

Conclusion: More AI algorithms are applied in CDSS more each year and are important in improving clinical practice and healthcare management. Although deep learning has become more popular, especially in imaging informatics, machine learning is still a majority of prediction and classification models in the medical field. There are several limitations including lack of friendly user-interfaces for web-based tools, lack of software and mobile applications, and narrow scope of users.