OBJECTIVES: Because of the complexity of different disease mechanisms, tere is a substantial need to diagnose various diseases effectively. Machine Learning – an area of artificial intelligence (AI)- has been increasingly used for patients, providers and payers to solve some of these issues. Paper will review how machine learning is being used to help in early identification of numerous diseases.

METHODS: Scopus and Web of Science databases between the years 2011-2022 are utilized to find original research publication. Both Machine Learning Algoritims and Deep Learnig were reviewed. Performance measures used in the literatures are identified. Several exclusion criterias are applied such as dublication, in accessibility of the entire text, non human studies and incomplete information related to test results. Publications are grouped by year, subject area, journal, countries as well as citations.

RESULTS: Number of publications related with Machine Learning based diagnosis went from 5 in 2011 to 535 in 2022. Multiple algorithms are used in machine learning based disease diagnosis modeling. Convolutional Neural Networks, Support Vector Machine and Logistic Regression were most common techniques that are used. There were several inconsistencies in terms of assessment measured published by the literatures. Model interpretation were absent in nearly all investigations. Most widely used disease diagnosis were heart disease, diabetics, kidney disease, Parkinson disease, cancer and covid-19. Most datasets were from clinical centers and health care claims data were rarely used.

CONCLUSIONS: Machine learning and deep learning literature in disease diagnosis have grown last decade. There were both data, algorithm and disease related challenges applying these techniques. The review identifies those challenges and suggest appropriate methods to mitigate those biases.