Structuro-functional surrogates of response to subcallosal cingulate deep brain stimulation for depression

Area of Science:

• Neuroscience
• Psychiatry
• Medical Imaging

Background:

• Subcallosal cingulate deep brain stimulation (scDBS) offers long-term improvement for severe treatment-resistant depression (TRD), but response rates vary.
• Identifying predictors of scDBS response is crucial for optimizing patient selection and treatment outcomes.

Purpose of the Study:

• To investigate whether baseline structural and functional brain attributes predict response to scDBS in patients with TRD.
• To develop machine learning models for classifying scDBS response status using neuroimaging data.

Main Methods:

• Retrospective analysis of MRI-derived brain volumes and 18F-fluorodeoxyglucose-PET glucose metabolism in a TRD cohort.
• Support vector machines (SVM) trained on preoperative imaging features (volume, metabolism) and electrode engagement for response prediction.
• Validation in an independent cohort and assessment of longitudinal changes.

Main Results:

• Machine learning models incorporating preoperative brain volumes achieved 83% accuracy in predicting binary response.
• Adding glucose metabolism data improved prediction accuracy to 86% and explained 67% of clinical variance.
• Electrode engagement with specific white matter tracts (uncinate fasciculus, cingulum) independently predicted response and significantly enhanced SVM model performance (up to 100% accuracy).

Conclusions:

• Responders and non-responders to scDBS exhibit distinct baseline and longitudinal differences in brain volume and metabolism.
• Preoperative imaging features, particularly when combined with tract engagement, are powerful predictors of scDBS response in TRD.
• These findings can inform patient selection and clinical decision-making for scDBS therapy.