Objective: To develop and validate an artificial intelligence (AI) machine learning (ML) model for preoperative cognitive risk stratification for deep brain stimulation surgery (DBS).

Background: DBS is a well-established neurosurgical therapy for movement disorders with 12,000-15,000 new implantations annually worldwide and clear benefit in patients’ quality of life.¹ Evaluating DBS candidacy is a time intensive process best served by an interdisciplinary team. Recognizing barriers to access, particularly the time, cost, and availability of neuropsychologists in rural or global settings, we explored an ML-based algorithm using a routine clinical T1-weighted brain MRI as an objective, scalable surrogate for traditional neuropsychological testing.

Method: In a retrospective cohort of 390 patients evaluated for DBS at the Fixel Institute between 2020 and 2025, preoperative T1-weighted MRI scans were collected and processed using FreeSurfer’s SynthSeg tool to extract volumetric morphometry features. At the Fixel Institute, the results of our comprehensive neuropsychological evaluation is summarized in a DBS Cognitive Rating Scale (DBS CRS) which provides a 1–5 score (1 = least cognitive concern; 5 = most concern).² We converted these to a binary classification: CRS 1–2 as ‘Go’ (low cognitive risk) and ≥3 as ‘No Go’ (higher cognitive risk). We developed a neuroimaging ML pipeline using a leakage controlled nested cross-validation with Platt sigmoid calibrator. Our primary performance metric was balanced accuracy. Our secondary metrics were the area under the receiver operating curve (AUROC) and area under the precision recall curve (AUPRC).

Results: From 390 subjects, 332 met inclusion criteria for AI analyses, requiring sufficient MRI image quality and complete clinical documentation including DBS CRS scores. The final analysis cohort had a mean (SD) age of 64.8 (12.0) years and a mean (SD) educational attainment of 14.4 (3.3) years, with 64.7% of participants being male. Our model achieved a mean (SD) balanced accuracy of 0.66 (0.068), with a mean AUROC of 0.71 and a mean AUPRC of 0.61.

Conclusion: We present promising preliminary results of a machine learning framework that leverages routine T1-weighted MRI brain scans to provide presurgical cognitive risk stratification in DBS candidates. This approach may enhance DBS surgical candidacy assessment and expand access to advanced neurologic therapies in resource-limited clinical settings.

References: 1. Wong JK, Mayberg HS, Wang DD, Richardson RM, Halpern CH, Krinke L, Arlotti M, Rossi L, Priori A, Marceglia S, Gilron R, Cavanagh JF, Judy JW, Miocinovic S, Devergnas AD, Sillitoe RV, Cernera S, Oehrn CR, Gunduz A, Goodman WK, Petersen EA, Bronte-Stewart H, Raike RS, Malekmohammadi M, Greene D, Heiden P, Tan H, Volkmann J, Voon V, Li L, Sah P, Coyne T, Silburn PA, Kubu CS, Wexler A, Chandler J, Provenza NR, Heilbronner SR, Luciano MS, Rozell CJ, Fox MD, de Hemptinne C, Henderson JM, Sheth SA and Okun MS (2023) Proceedings of the 10th annual deep brain stimulation think tank: Advances in cutting edge technologies, artificial intelligence, neuromodulation, neuroethics, interventional psychiatry, and women in neuromodulation. Front. Hum. Neurosci. 16:1084782. doi: 10.3389/fnhum.2022.1084782

2. Kenney L, Rohl B, Lopez FV, Lafo JA, Jacobson C, Okun MS, Foote KD and Bowers D (2020) The UF Deep Brain Stimulation Cognitive Rating Scale (DBS-CRS): Clinical Decision Making, Validity, and Outcomes. Front. Hum. Neurosci. 14:578216. doi: 10.3389/fnhum.2020.578216

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/ai-guided-presurgical-risk-stratification-for-deep-brain-stimulation-surgery/