Artificial Intelligence in the Diagnosis of Movement Disorders: A New Lens for Clinical Neurology

By Campion Quinn, MD

Diagnosing movement disorders has long relied on sharp clinical observation. The neurologist’s eye, trained to catch a tremor, judge gait rhythm, or note subtle asymmetries, has been central to this. But what if technology could enhance this faculty, allowing us to see what our eyes might miss? That’s what artificial intelligence (AI), specifically computer vision, is beginning to do in neurology.

AI isn’t replacing clinicians—it’s offering a second set of digital eyes—tools that see, measure, and track movement patterns with precision. This article explores how AI reshapes diagnosis and monitoring for conditions like Parkinson’s disease, dystonia, and tremor syndromes.

Movement disorders often involve overlapping symptoms that can create diagnostic ambiguity. For example, tremor is found in Parkinson’s disease, essential tremor, and dystonia.  Clinical diagnosis requires careful inspection, but standardized tools like the Unified Parkinson's Disease Rating Scale (UPDRS) suffer from subjectivity and interrater variability. They’re constrained when detecting early disease stages (1).

Computer vision solves these limitations. Using video, often captured with a smartphone, AI algorithms can now track limb movement, tremor amplitude, and even subtle changes in joint angles. One core technique, called pose estimation, identifies key body points like elbows, fingers, and eyes and then translates those movements into clinical metrics.

Pose estimation is a form of artificial intelligence that detects and tracks the position of a person’s body parts, like hands, arms, knees, or head, in a video. It doesn’t rely on wearable devices (e.g., accelerometers or motion sensors). Instead, it identifies key anatomical landmarks using just a camera and software. It works seamlessly and can run in real time on portable devices such as an iPhone, making it scalable for clinics, telemedicine, and even patient homes (1).

AI Spotlight: VisionMD

Developed by Dr. Diego Guarin at the University of Florida, VisionMD is an open-source AI tool that analyzes standard videos—captured via smartphone, laptop, or Zoom—to assess motor symptoms in Parkinson’s disease and other movement disorders. The software delivers objective, precise motion metrics without requiring cloud computing or technical expertise.

Already in use internationally, VisionMD helps clinicians identify optimal treatment settings, such as deep brain stimulation configurations, and reduces the subjectivity of traditional assessments.

In Parkinson’s disease, AI models track tremors and assess bradykinesia from finger-tapping videos. These tools can match accelerometer accuracy, enabling clinicians to monitor changes more frequently and objectively (2, 3).

Dystonia assessment has also evolved. For example, the severity of head tremors in cervical dystonia can be quantified using AI tools, helping physicians titrate botulinum toxin dosing or evaluate therapy outcomes (4). Similarly, facial tics, blepharospasm, and gait abnormalities are being detected with increasing precision.

Nystagmus, critical in stroke and vertigo diagnostics, is notoriously hard to quantify at the bedside. Convolutional neural networks can now analyze eye movements from smartphone videos with accuracy comparable to that of infrared eye trackers (5).

These tools are also empowering patients. Home-based video assessments allow for asynchronous care. A patient records a 30-second clip tapping their fingers or walking down a hallway. AI evaluates and returns results. Clinicians review this alongside history and imaging, improving access, especially for rural or mobility-impaired patients (6).

AI is equally transformative in clinical trials. Traditional scales often fail to capture small but clinically significant changes. AI-derived digital biomarkers detect subtle shifts over time, reducing trial sizes and durations. In one study, gait analysis via video captured a year of progression in spinocerebellar ataxia, outperforming existing scales (7).

Despite its promise, AI comes with caveats. Bias is a concern. If algorithms are trained on non-diverse datasets, they may underperform in underrepresented populations. Techniques like federated learning—training models without moving raw data—can reduce privacy risks and promote equity, but adoption remains limited due to logistical barriers.

Terms like domain generalization (where algorithms ignore demographic variation and focus on disease-specific features) are becoming essential safeguards against algorithmic bias.

Another hurdle is interpretability. Black-box tools can be challenging to trust. Clinicians require transparent AI systems that clearly demonstrate how and why a decision was made. Advances in explainable AI (XAI) are addressing this, but usability in real-world settings still lags.

Importantly, AI tools must integrate into clinical workflows. They must be plug-and-play and not reliant on elaborate setups. Cross-disciplinary collaboration between clinicians, engineers, and patients is essential for success.

As clinicians, we are not just observers—we are participants in the evolution of diagnostic medicine. AI provides a new lens. One that augments our capacity to detect, measure, and monitor disease with clarity and consistency.

Let’s not view AI as a replacement for clinical wisdom. Instead, see it as an upgrade to our stethoscope that helps us hear movement. With careful validation and ethical implementation, AI can help us deliver more precise, efficient, and equitable neurological care.

References

1. Friedrich, M. U., Relton, S., Wong, D., & Alty, J. (2025). Computer vision in clinical neurology: a review. JAMA Neurology, 82(4), 407–415. https://doi.org/10.1001/jamaneurol.2024.5326

2. Williams, S., Fang, H., Relton, S. D., Wong, D. C., Alam, T., & Alty, J. E. (2020). Accuracy of smartphone video for contactless measurement of hand tremor frequency. Movement Disorders Clinical Practice, 8(1), 69–75. https://doi.org/10.1002/mdc3.13119

3. Morinan, G., Dushin, Y., Sarapata, G., et al. (2023). Computer vision quantification of whole-body Parkinsonian bradykinesia using a large multi-site population. NPJ Parkinson’s Disease, 9(1), 10. https://doi.org/10.1038/s41531-023-00454-8

4. Peach, R., Friedrich, M., Fronemann, L., et al. (2024). Head movement dynamics in dystonia: a multicenter retrospective study using visual perceptive deep learning. NPJ Digital Medicine, 7(1), 160. https://doi.org/10.1038/s41746-024-01140-6

5. Friedrich, M. U., Schneider, E., Buerklein, M., et al. (2023). Smartphone video nystagmography using convolutional neural networks: ConVNG. Journal of Neurology, 270(5), 2518–2530. https://doi.org/10.1007/s00415-022-11493-1

6. Alty, J., Bai, Q., Li, R., et al. (2022). The TAS Test project: a prospective longitudinal validation of new online motor-cognitive tests to detect preclinical Alzheimer’s disease. BMC Neurology, 22(1), 266. https://doi.org/10.1186/s12883-022-02772-5

7. Ilg, W., Müller, B., Faber, J., et al. (2022). Digital gait biomarkers allow for the capture of 1-year longitudinal change in spinocerebellar ataxia type 3. Movement Disorders, 37(11), 2295–2301. https://doi.org/10.1002/mds.29206