How AI is Empowering Researchers in Movement Disorder Science

Key points

• DeepLabCut, an open-source deep-learning tool, achieved markerless tracking comparable to human accuracy after labeling only around 200 video frames.
• Models trained on ordinary single-camera clinical video predicted walking speed, cadence, knee flexion and the Gait Deviation Index with correlations of roughly 0.73 to 0.83 against laboratory measurements.
• A Movement Disorder Society task force cautions that more data has not yet translated into proportionally better clinical understanding, so each new model still needs validation.

Why manual analysis limited the field

The study of movement disorders has long been constrained by the limitations of human observation. Researchers analyzing gait patterns or tremor characteristics relied on frame-by-frame video review or on marker-based motion capture that, as Kidziński and colleagues (2020) put it, requires expensive systems and highly trained personnel — constraints that limited study sizes and, for visual review, introduced observer variability. Artificial intelligence is changing what's possible.

What machine learning adds

Machine learning can now process movement data at a scale that manual review never could. DeepLabCut, an open-source deep-learning tool, achieved markerless tracking comparable to human accuracy after labeling only around 200 video frames (Mathis et al., 2018). In clinical gait, models trained on ordinary single-camera video predicted walking speed, cadence, knee flexion, and the Gait Deviation Index with correlations of roughly 0.73 to 0.83 against laboratory measurements (Kidziński et al., 2020). Automating that analysis doesn't just save time — it opens new avenues of investigation. (For how camera-based tracking works, see what markerless motion capture is.)

AI excels at pattern recognition across large datasets, identifying subtle correlations between movement characteristics and clinical outcomes that might escape even experienced researchers. These approaches are being investigated as ways to identify early markers of disease, tailor treatment, and better target patient subgroups for clinical trials — goals a Movement Disorder Society task force on technology has highlighted, alongside the caution that more data has not yet translated into proportionally better clinical understanding (Espay et al., 2016). In clinical care the same tension applies to tremor, where higher-resolution measures still have to show change beyond natural variability; see quantifying tremor beyond visual rating.

Longitudinal analysis at scale

Perhaps most significantly, AI enables longitudinal analysis at population scale. Researchers can track movement changes across diverse patient groups over extended periods, building evidence bases that inform clinical guidelines and therapeutic development. And because markerless methods work from stored video, existing recordings can be re-analyzed as pose-estimation algorithms improve (Wade et al., 2022) — though each new model still has to be validated before its outputs are trusted. That validation requirement is what separates research-grade output from a measurement clinicians can act on, as we discuss in what makes movement measurement clinical-grade.

For the movement disorder research community, AI represents more than efficiency — it represents the possibility of discoveries that were simply inaccessible before. Institutions embracing these capabilities position themselves to lead the next generation of neurological research and translate findings into better patient care more quickly.

References

1. Mathis A, Mamidanna P, Cury KM, et al. DeepLabCut: markerless pose estimation of user-defined body parts with deep learning. Nature Neuroscience. 2018;21(9):1281-1289. doi:10.1038/s41593-018-0209-y
2. Kidziński Ł, Yang B, Hicks JL, Rajagopal A, Delp SL, Schwartz MH. Deep neural networks enable quantitative movement analysis using single-camera videos. Nature Communications. 2020;11:4054. doi:10.1038/s41467-020-17807-z
3. Espay AJ, Bonato P, Nahab FB, et al. Technology in Parkinson's disease: challenges and opportunities. Movement Disorders. 2016;31(9):1272-1282. doi:10.1002/mds.26642
4. Wade L, Needham L, McGuigan P, Bilzon J. Applications and limitations of current markerless motion capture methods for clinical gait biomechanics. PeerJ. 2022;10:e12995. doi:10.7717/peerj.12995