Mobile Menu

Multimodality MRI and ML techniques in bipolar disorder identification

Researchers have combined structural and functional magnetic resonance imaging (MRI) with machine learning (ML) techniques to aid in accurately identifying bipolar disorder.

Bipolar disorder

Bipolar disorder is a chronic and debilitating mood disorder that affects up to 2.5% of the population. It is characterised by extreme fluctuations in mood, functionality and energy, alongside recurrent depressive and manic episodes. Early onset of disease results in high rates of self-inflicted injury and hospitalisation. Estimates suggest that the risk of suicide in patients with bipolar disorder is 20-times higher than the general population. Most importantly, the clinical symptoms of bipolar disorder overlap with other mood disorders, such as major depressive disorder and schizophrenia. As a result, bipolar often goes undiagnosed or misdiagnosed for extended periods of time. This prevents the effective treatment of bipolar disease and results in increased disease burden. Therefore, identification of new biomarkers is critical to assist in diagnosis and therapeutic monitoring.

In recent years, MRI has been utilised in neuroimaging studies as a potential biomarker. For structural MRI, voxel-based morphometry (VBM) is one of the most common techniques to analyse focal differences in brain anatomy. For functional MRI, regional homogeneity (ReHo) is an established approach to evaluate local activity in the brain while at rest. Study of both of these features has revealed that bipolar is a disorder with several morphological and functional brain abnormalities. Recent machine learning approaches have addressed some of the inconsistencies seen within these methods.

Multimodality MRI

In this study, published in BMC Psychiatry, researchers expanded on previous studies that have looked at just single-modality MRI. Instead, they constructed a support vector machine (SVM) model using a combination of structural and functional MRI to accurately identify patients. Specifically, they used VBM and ReHO measurements in grey matter volumes as features to differentiate between patients with bipolar disorder and healthy controls. In total, they recruited 44 patients with bipolar disorder and 36 healthy controls. Based on these features, the team were able to establish the SVM model and perform discriminant analysis.

The SVM classifier was able to effectively identify patients with bipolar disorder at an accuracy of 87.5%, sensitivity of 86.4% and specificity of 88.9%. These results highlight that the combination of structural and functional MRI could improve the recognition of bipolar disorder using an SVM classifier. In turn, this classifier could be used to identify patients for clinical trial studies or to stratify cohorts for further study.

Image credit: By kjpargeter –

More on these topics

Bipolar / Machine Learning / SVM

Share this article