Individualized identification of first-episode bipolar disorder using machine learning and cognitive tests

Area of Science:

• Neuroscience
• Psychiatry
• Computational Psychiatry

Background:

• Cognitive dysfunction is a hallmark of Bipolar Disorder (BD), often correlating with illness severity and episode frequency.
• Early identification of cognitive impairments in BD is crucial for timely therapeutic interventions.
• Previous research highlights the link between cognitive deficits and the progression of Bipolar Disorder.

Purpose of the Study:

• To employ machine learning (ML) techniques to identify first-episode Bipolar Disorder (FE-BD) patients based on cognitive test performance.
• To assess the utility of ML models in distinguishing individuals with early-stage BD from healthy controls.
• To investigate whether cognitive deficits observed in chronic BD patients are also present in the early stages of the disorder.

Main Methods:

• Utilized two cohorts: Cohort 1 (74 chronic BD patients, 53 healthy controls) and Cohort 2 (37 FE-BD patients, 18 healthy controls).
• Administered the Cambridge Neuropsychological Test Automated Battery (CANTAB) to assess visual processing, spatial memory, attention, and executive functions.
• Trained a linear Support Vector Machine (SVM) model to differentiate between chronic BD patients and healthy controls, achieving 77% accuracy.

Main Results:

• The trained ML model successfully identified chronic BD patients with 77% accuracy.
• Applying the model to the FE-BD cohort yielded a classification accuracy of 76% (AUC = 0.77).
• These findings indicate significant cognitive impairments in FE-BD patients, comparable to those in chronic BD.

Conclusions:

• Cognitive deficits are present in the early stages of Bipolar Disorder and may persist throughout the illness course.
• These early cognitive impairments can serve as potential biomarkers for the detection of BD.
• The developed ML tool demonstrates potential for early BD detection by identifying cognitive markers.