OBJECTIVES : Machine learning (ML) consists of algorithms which offers improved detection, diagnosis, and therapeutic monitoring of disease by analyzing vast and complex data-sets. The aim of this study was to summarize the available literature on the use of ML techniques in early detection/diagnosis of diseases

METHODS : A literature search on PubMed was performed for identifying studies involving use of ML techniques for early detection/diagnosis of diseases. Studies were included if they were using ML techniques for early disease detection/diagnosis using clinical or real-world data-sets and were published within 5 years.

RESULTS

: 152 studies out of total 268 reviewed met the inclusion criteria. These studies were further examined for disease area covered, machine learning algorithm deployed, and data-sets used as part of the study.

Disease area covered:

• Neurological disorder studies have the highest use of ML techniques in early prediction (48 studies). This was followed by cancer prediction studies (39 studies) and cardiac event prediction studies (16 studies)
  • Other areas studied include arthritis, diabetes related complications, liver complications etc.

Algorithm Used:

• Support vector machine (SVM) is the top choice for prediction studies (57 studies) followed by Neural network-based algorithm (42 studies)
  • Other deployed methods include random forest, logistic regression etc.

Datasets Used:

Is primarily dependent on the disease area being studied.

• Neuroimaging data (MRI scans etc.) was the most common dataset for Neurological disorder studies
• Patient history, gene profile, and various lab test result datasets were used in combination for majority of oncology studies
• Electrocardiogram (ECG) datasets and patient history were used for cardiac event prediction studies

CONCLUSIONS : This review provides insights into rising role of ML techniques for early disease detection. These techniques have potential to harness the power of real-world datasets and through sophisticated algorithms convert them into life-saving action items