JMIR Publications

ABSTRACT

Background:

Bipolar disorder (BD) is the tenth common cause of frailty in young individuals and has triggered morbidity and mortality worldwide. BD patients have 9–17 years lower lifetime as compared to the normal population. It is a predominant mental disorder but misdiagnosed as depressive disorder that leads to difficulties in the treatment of affected patients. 60% of patients with bipolar disorder are looking for the treatment of depression. However, machine learning provides advanced skills and techniques for the better diagnosis of bipolar disorder.

Objective:

This review aims to explore the machine learning algorithms for the detection and diagnosis of bipolar disorder and its subtypes.

Methods:

The study protocol adapts PRISMA extension guidelines. It explores three databases, which were Google scholar, ScienceDirect, and PubMed. To enhance the search, we performed backward screening of all the references of the included studies. Based on the predefined selection criteria, two levels of screening were carried out: the title and abstract review and the full review of the articles that met the inclusion criteria. Data extraction was performed independently by all investigators. To synthesize the extracted data, a narrative synthesis approach was followed.

Results:

573 potential articles were retrieved from three databases. After pre-processing and screening, only 33 articles were identified, which met our inclusion criteria. The most commonly utilized data belonged to the clinical category (n=19, 57.57%). We identified different machine learning models used in the selected studies, classifications models (n=18, 55%), regression models (n=5, 16%), model based clustering methods (n=2, 6.06%), natural language processing (n=1, 3.03%), clustering algorithms (n=1,3.03%), as well as deep learning based models (n=3, 9.09%). The most common data utilized was magnetic resonance imaging (MRI) for classifying bipolar patients compared to other groups (n=11, 34%) while the least common utilized data was microarray expression dataset and genomic data. The maximum ratio of accuracy was 98% while the minimum accuracy range was 64%.

Conclusions:

This scoping review provides an overview of recent studies based on machine learning models used to diagnose bipolar disorder patients regardless of their demographics or if they were assessed compared to patients with psychiatric diagnoses. Further research can be conducted for clinical decision support in the health industry.