Prediction of Clinical Response of Transcranial Magnetic Stimulation Treatment for Major Depressive Disorder Using...

Area of Science:

• Neuroscience
• Computational Psychiatry
• Artificial Intelligence

Background:

• Cognitive control deficits are common in major depressive disorder (MDD).
• The precise mechanisms of transcranial magnetic stimulation (TMS) and its impact on cognitive control in MDD remain unclear.
• Understanding TMS efficacy is crucial for improving treatment outcomes.

Purpose of the Study:

• To investigate the predictive capability of cognitive variables for clinical response to TMS in MDD patients.
• To evaluate the effectiveness of hyperdimensional computing (HDC) in classifying TMS treatment outcomes.
• To compare HDC with Support Vector Machines (SVM) for predicting treatment response.

Main Methods:

• Utilized 34 cognitive variables from 22 MDD subjects over an eight-week TMS treatment period.
• Employed hyperdimensional computing (HDC), a novel brain-inspired computing paradigm, for classification.
• Implemented leave-one-subject-out cross-validation (LOSOCV) and feature selection using SelectKBest.
• Assessed performance using accuracy, sensitivity, specificity, and Area Under the Curve (AUC).

Main Results:

• HDC achieved an AUC of 0.82, outperforming SVM by 0.07, despite SVM having higher accuracy.
• Optimal classification performance for both HDC and SVM was achieved with SelectKBest feature selection.
• The 'ws_MedRT' (median rate for the Websurf task) feature demonstrated a clear distinction between responders and non-responders.

Conclusions:

• Hyperdimensional computing (HDC) shows significant potential for predicting clinical responses to TMS in MDD.
• Feature selection is critical for enhancing the performance of classification models in predicting TMS efficacy.
• The findings suggest HDC as a valuable tool for personalized treatment strategies in psychiatric disorders.