Objective: To develop and evaluate a novel deep learning-based computer vision approach for tracking tongue movements in Parkinson’s disease (PD) patients with levodopa-induced dyskinesias to enable precise medication titration.

Background: Current dyskinesia assessment in PD relies primarily on clinician-administered rating scales such as the Unified Dyskinesia Rating Scale (UDysRS) and modified Abnormal Involuntary Movement Scale (mAIMS). While generally adequate, these scales often fail to capture the fine granularity of movement changes critical for personalized medication titration during hospitalization. Wearable sensor systems have emerged as complementary tools for tracking limb movements, yet they require specialized equipment, involve complex setup procedures, and rarely capture orofacial movements including those of the face and tongue.

Method: We recorded 30-second video segments of dyskinesias in a hospitalized PD patient, every day for 4 days, to capture reaction to medication cycle. Videos were processed through the DeepLabCut algorithm, which estimated and tracked tongue and lip landmarks. We then analysed several glossographic kinematic features: total distance travelled by markers, peak velocity, mean velocity, velocity variability, and discrete movement onsets per time unit using velocity thresholds. These quantitative metrics were used to titrate dopaminergic medication during the patient’s hospital stay.

Results: The system successfully quantified dyskinesia severity and temporal patterns. The features showed temporal change towards improvement throughout the hospital stay, which was corroborated by clinical findings and patient-reported outcomes. Details will be presented in person during the conference. The patient was discharged with symptoms at their neurological baseline.

Conclusion: Continuous tongue motility tracking can refine personalized dopaminergic therapy adjustments, particularly in advanced PD cases, where narrow therapeutic windows necessitate precision medication adjustments. We stipulate that glossography may be a useful metric complimentary to holistic clinical neurological assessment.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/glossography-a-novel-computer-vision-based-technique-for-assessing-and-treating-involuntary-tongue-movements-in-dyskinetic-episodes-in-parkinsons-disease/