Machine learning model based on spontaneous brain activity detected by functional MRI for distinguishing unipolar...

The therapeutic strategies for bipolar disorder (BD) and unipolar depression (UD) are quite different. However, the majority of patients with BD often present with a depressive episode as their initial symptom and are misdiagnosed as UD. To date, no reliable tool has been able to accurately differentiate BD patients from UD patients.

The spontaneous brain activity derived from functional MRI of 79 BD patients and 79 matched UD patients was used to establish machine learning (ML) models for distinguishing BD patients from UD patients. Furthermore, the imaging signatures obtained from the optimal model and statistically significant clinical characteristics were incorporated into the predictive nomogram. The performance of the nomogram was evaluated by calibration curve and decision curve analysis (DCA).

The ML model based on spontaneous brain activity of 10 brain regions with significant differences between BD patients and UD patients achieved optimal diagnostic performance, with an AUC of 0.894 in the validation dataset. Disease duration was identified as an independent clinical predictor. A nomogram integrating disease duration with the imaging signatures derived from the optimal model demonstrated good discriminative efficacy, with a C-index of 0.926. The calibration curve and DCA indicated excellent reliability and significant net clinical benefit.

Our study provides a preliminary proof-of-concept that a nomogram integrating spontaneous brain activity with clinical information may serve as a potential diagnostic tool for differentiating BD patients from UD patients.