Objective DBS program selection using smartphone-based kinematics

Deep brain stimulation (DBS) has transformed the treatment of Parkinson's disease, yet programming remains a slow, subjective process guided by clinical intuition rather than quantitative evidence (1,2).Despite major advances in electrode design and imaging, clinicians still rely on trial-and-error adjustments and rater-dependent scores that scale poorly with modern multidimensional parameter spaces (3).Objective, clinician-facing, data-driven tools are needed to make programming faster, reproducible, and transparent during routine programming sessions.Recent progress in computer vision now enables precise quantification of motor performance directly from standard smartphone video, without wearables or motion-capture hardware (4-6).Building on our previously validated open-source platform (VisionMD) (4), we developed StimVision, a smartphone-based kinematic framework designed for in-clinic use that objectively ranks DBS programs within a single programming session and characterizes the quantitative motor signature of stimulation-induced improvement. J o u r n a l P r e -p r o o fprogramming to a reproducible, quantitative process, StimVision offers a practical route toward datadriven, standardized neuromodulation.The approach is device-agnostic and could readily extend to remote or home-based assessments, aligning with the field's shift toward tele-programming and digital biomarkers.While our sample size was modest, the convergence of quantitative precision and clinical feasibility positions this framework as a scalable proof-of-concept for objective, personalized DBS programming.Broader validation across centers and integration with automated parameter search algorithms could further accelerate the transition from expert-driven tuning to measurable, algorithm-assisted therapy optimization.All code and derived data used in this work are openly available at https://github.com/Flolan2/StimVision_data to encourage replication and clinical translation.