Objective: To compare the predictive value for motor outcomes of multimodal MRI to clinical predictors alone in STN DBS for PD.

Background: STN DBS is an effective treatment for the motor symptoms of PD, yet outcomes remain highly variable. Currently, pre-surgical levodopa response provides the most established prediction of clinical outcome, although its predictive power varies greatly across studies. MRI-based methods including resting state functional connectivity (FC), subcortical and cortical volumetrics, and DTI-based structural connectivity (SC) independently predict DBS outcomes, while a comprehensive model integrating these methods has not yet been developed.

Method: We included 61 participants who had multimodal MRI prior to STN DBS for PD and who had adequate clinical data before and after DBS, as well as good quality volumetric, DTI, and rs-fcMRI data. We applied clinical and MRI-based predictors from published studies including thalamic and ventricular volumes, FC and SC connectivity between STN and GPI, VL thalamus, and motor cortex. Regularized linear regression using the least absolute shrinkage and selection operator (LASSO) and leave-one-out cross-validation was used for model construction and factor selection, optimized to RMSE to avoid overfitting with addition of more variables. The primary outcome was percent change in UPDRS-III from the initial preoperative OFF-medication examination to the average OFF-medication, ON-stimulation score over the first year after DBS.

Results: The “traditional” model, constrained to only clinical predictors (including levodopa response, age, sex, handedness, LEDD, preoperative UPDRS-III) was modestly predictive of motor improvement (R² = 0.29, RMSE = 14.9. In comparison, the total model, which included FC, SC, volumetric, and clinical predictors more strongly predicted motor improvement (R² = 0.55, RMSE = 13.6). Factors included in the optimized model included STN-GPI FC, GPI-VL thalamus SC, VL thalamus-motor cortex SC, ventricular volume, age, preoperative UPDRS-III, and levodopa responsiveness.

Conclusion: Multimodal MRI greatly outperforms levodopa responsiveness alone in prediction of motor outcomes for STN DBS in PD. Addition of cognitive and psychiatric features both as predictors and as outcomes may further enhance the predictive power and clinical utility of this approach.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/quantifying-the-value-of-multimodal-mri-in-outcomes-prediction-for-stn-dbs-in-pd/