Objective: To determine whether baseline motor phenotyping with the Integrated Motion Analysis Suite (IMAS), a multimodal sensor-based assessment platform, can identify clinically meaningful Parkinson’s disease (PD) subgroups with differential longitudinal response to electrosonic stimulation (ESStim).
Background: ESStim is a novel noninvasive brain stimulation approach that combines transcranial direct current stimulation (tDCS) and transcranial ultrasound (TUS). In a previously described single-center, randomized, double-blind, sham-controlled trial comparing ESStim, tDCS, TUS, and sham delivered for 10 days, 20 min/day (n=12/group, 48 total), ESStim showed greater improvement in ON-medication UPDRS-III than the other interventions (~4.7-point mean change from therapy completion to 6 weeks post-stimulation, p < 0.001, mean baseline score=22.5). IMAS provides high-dimensional quantitative motor metrics for data-driven stratification.
Method: Sixty-two baseline IMAS features were analyzed using a self-organizing map (SOM), followed by higher-level clustering of SOM neuron weights to define patient subgroups. Cluster-defining features were identified using random forest feature ranking and evaluated across subgroups using ANOVA with Bonferroni-adjusted post hoc tests. Subgroup membership was then carried forward to compare change from baseline in ON-medication UPDRS-III scores through treatment and across follow-ups up to 6 weeks post-treatment in the ESStim group.
Results: A 2-group solution identified two baseline phenotypes, postural-instability/fast and slow/tremor. The postural-instability/fast group showed greater improvement than the slow/tremor group (~20% vs 15% through 6 weeks, p<0.05). Stratification by baseline UPDRS-III severity or conventional tremor-dominant versus akinetic-rigid subtype did not separate responders. A 3-group solution further separated patients into fast/variable, postural-instability, and slow/tremor phenotypes; the postural-instability group showed the greatest improvement and differed from both other groups (both p<0.05).
Conclusion: Baseline sensor-derived motor phenotyping may identify PD subgroups with differential response to ESStim, and this stratification appears to outperform standard clinical subgrouping such as baseline UPDRS-III severity or tremor-dominant vs akinetic-rigid categorization.
References: Electrosonic stimulation for the treatment of motor symptoms in Parkinson’s disease: motion analysis assessments. Dipietro et al. Pan American Parkinson’s Disease and Movement Disorders Congress, Houston, USA 2026
Integrating Big Data, Artificial Intelligence, and motion analysis for emerging precision medicine applications in Parkinson’s Disease. Dipietro et al. J Big Data. 2024;11(1):155
Novel methods of transcranial stimulation: electrosonic stimulation. Wagner T, Dipietro L. In: Krames ES, Peckham PH, Rezai AR, editors. Neuromodulation. Cambridge: Academic Press. 2018
Funding: Work reported in this publication was supported in part by the National Institutes of Health NINDS (Award Number R44NS080632, R44NS110237, R43NS113737). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
To cite this abstract in AMA style:
L. Dipietro, U. Eden, S. Elkin-Frankston, M. El-Hagrassy, D. Doruk Camsari, C. Ramos-Estebanez, F. Fregni, T. Wagner. A Machine Learning-Based Approach to Identify Responders to Electrosonic Brain Stimulation in Parkinson’s Disease [abstract]. Mov Disord. 2026; 41 (suppl 1). https://www.mdsabstracts.org/abstract/a-machine-learning-based-approach-to-identify-responders-to-electrosonic-brain-stimulation-in-parkinsons-disease/. Accessed October 1, 2026.« Back to 2026 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/a-machine-learning-based-approach-to-identify-responders-to-electrosonic-brain-stimulation-in-parkinsons-disease/