How AI-Powered Gait Analysis Reveals Insights Into Your Health

**Yuval Naveh, Co-Founder and Chief Scientific Officer of OneStep**

Gait—how we walk—provides powerful, often overlooked insights into our physical, cognitive, mental, and social health. It’s more than just movement; it serves as a window into our overall well-being, with impairments in gait and mobility linked to falls, cognitive decline, chronic conditions, and even mortality. While healthcare professionals have long studied these connections, there’s something universally intuitive about how we assess a person’s health just by watching them walk.

Though step counters and fitness trackers have gained popularity for monitoring daily activity, these devices offer only a broad snapshot of wellness—distinguishing between the active and inactive. Gait speed is now recognized as the “sixth functional vital sign,” however, further mobility and gait analysis can provide much deeper, more detailed insights. In this article, Yuval Naveh, Co-Founder and Chief Scientific Officer of OneStep, will explore how the advancement in AI and ML is unlocking new possibilities for clinical care and patient monitoring through gait analysis, with the potential to improve outcomes across a wide range of conditions and patient populations.

Introduction

Gait, the way people walk, is a window to their physical, cognitive, mental, and social well-being. Mobility and gait impairment has been associated with morbidity, falls, cognitive decline, mortality, and symptoms of chronic disorders^1–5. Beyond clinical and academic conversation, on the more layman-intuitive level, we all have strong impressions of others’ physical-emotional condition only by observing how they walk. We can look at someone walking on the street and immediately sense if they are confused, tired, weak, or need help. We look at our parents as they age and move slower and more heavily. Some people say they can know what a person does for a living only by looking at the way they walk. While step counters and daily physical activity gadgets have become more popular, they can only roughly distinguish between active and healthy people and less active and less healthy people. Gait speed is now considered as the sixth functional vital sign⁶. Still, people who walk slowly and also people who walk faster can have so many different mobility dysfunctions derived from so many different impairments. We can instinctively draw much deeper conclusions about health simply by watching an individual’s movements, can we also capture such characteristics of mobility quality digitally? Could we use them to deliver better care?

Monitoring people’s real-world mobility allows for monitoring disease progression and recovery trends, evaluating treatment efficacy, and developing more precise intervention tools for healthcare. Just recently, Mobilise-D, a 5-year IMI-funded consortium involving dozens of international research institutions, pharma, and tech companies, has finished its mission to outline a roadmap of the effort needed to bring digital mobility outcomes (DMOs) and biomarkers from concept to approval. As part of this project, its participants have published technical validation studies of single wearable devices that estimate DMOs, worn on wrist⁷ or waist^8,9 in lab and real-world settings; published clinical validations of DMOs in various health conditions, such as COPD^10–12, PD^13–15, MS^16–18, PFF¹⁹; and described the regulatory qualifications of DMOs required and expected in order to be adopted as standards for use in clinical trials and care²⁰.

Use smartphones instead

While Mobilise-D examined wearable sensors, including an inertial measurement unit (IMU), similar solutions can be applied using smartphones where all smartphones—budget, mid-range, and flagship models—integrate IMUs as standard components. Mobile phones are usually placed in various positions, and while they are being carried over the body, they may be positioned differently in different types of pockets. This may require more advanced algorithms to handle these “degrees of freedom”, nevertheless, studies show^21–25 that some state-of-the-art machine learning (ML) algorithms can even outperform the results reported in Mobilise-D’s real-world technical validation²⁶ and provide clinically valid results compared to gold standard systems.

In addition to passive mobility measurement, smartphones and mobile apps enable the collection of mobility measures in controlled settings, including gait analysis, standard tests (such as timed up and go, sit to stand, 6-minute walk test), dual tasking, and collection of patient-reported outcomes (PROs). This complementary data is crucial as it allows the extraction of further mobility characteristics, subjective perception, and context, including environmental conditions. Moreover, it allows for an “apples to apples” comparison, which is essential to track changes and evaluate conditions vs. norms.

The opportunity of measuring DMOs with smartphones

The ability to measure valid mobility measures using any mobile device provides opportunities to conduct clinical studies that were inaccessible or too complicated otherwise. Since everyone has a smartphone, recruiting patients to participate in a study can be much easier. The following results are taken from an analysis performed to establish benchmarks for post-total hip arthroplasty (THA) recovery. The cohort includes hundreds (600 approximately) of patients who recorded measurements pre- and post- surgery, and compares In-app measurements vs. Background – unconsciously recorded measurements. (updated results following Dr. Teitz’s analysis).

In-app vs. background gait trends in THA. Nine different gait parameters are presented for THA patients post-surgery. In-app active walks in light green, background measurements in dark green. Error bars represent standard error.

These results show continuous normative trajectories of the recovery trend expressed in different gait parameters. Such parameters can also be assessed in motion labs even more accurately; however, gait and motion labs provide only a momentary snapshot and cannot provide a continuous representation of recovery progression. Where it might be already known in the literature that after six to eight weeks, THA patients return to their pre-op gait baseline values, such as in gait speed and stride length, the structure and pace of that trend were unknown ^27–29.

Moreover, these results compare controlled measurements when the patients were aware of the measurement and background passive measurements. The results imply that in-app walks demonstrate faster gait recovery than background walks across all measured parameters. This pattern could result from “Hawthorne effect”³⁰, which describes how people change their behavior when they know they are being observed or from the other side due to environmental conditions, such as being outdoors, walking on uneven surfaces, or in crowded areas. Anyway, background passive data and in-app controlled measurements provide a complementary representation of patients’ functional and mobility status.

Clinical applications

These kinds of analyses are the foundations of new tools for intervention, such as alerting caregivers when patients recover slower than expected. Furthermore, it enables the comparison of different surgery methods, variations in patient populations, implants, rehabilitation protocols, and other factors. Eventually, this analysis demonstrates the potential of monitoring DMOs with smartphones for tracking patients’ status and progression, assessing quality of care, and developing new digital care tools, not only for post-op and recovery patients but for many other clinical applications as well.

Smartphone-based tracking of DMOs has far-reaching potential. It can assess risks of coming adverse events, monitor disease progression, assess therapeutic efficacy and side effects, complement standard medical assessments, support clinical trials, and enhance patient awareness of their symptoms and overall health. Could smartphone-based DMO monitoring provide reliable indicators of the risk of falls, pain levels or frailty in chronic patients? Track the progression of neurodegenerative diseases like Multiple Sclerosis or Parkinson’s Disease? Detect complications in diabetes or acute adverse events? Maybe even assess non-mobility-related conditions such as visual impairment? And furthermore, predict longevity in young and healthy populations?

The clinical potential is vast—almost limitless. However, adoption within healthcare is not just a question of technological feasibility and clinical value; it must also align with the operational and economic realities of the medical ecosystem. Still, the opportunities to transform healthcare delivery are immense, and the technology is already here.

About Yuval Naveh, Co-Founder and Chief Scientific Officer of OneStep

Yuval Naveh is a co-founder and the Chief Scientific Officer of OneStep, a digital care platform powered by motion intelligence that provides comprehensive gait and mobility analysis directly from a smartphone. With nearly 20 years of experience in machine learning and over a decade leading ML and data science teams, Yuval brings a combination of scientific rigor and practical innovation to the company. He oversees OneStep’s technological research in motion analysis and leads the exploration of clinical insights derived from real-world patient mobility data.

Yuval holds an MSc in Computer Science from the Hebrew University of Jerusalem, where his thesis focused on machine learning and time series prediction. He is also an alumnus of the prestigious Talpiot program of the Israel Defense Forces, where he led dozens of R&D initiatives and data science teams, delivering large-scale, cross-unit technological solutions, one of which earned the Israel Defense Award.

Yuval and the OneStep team are committed to advancing motion technology to enable more accessible, personalized, and proactive care through continuous, real-world monitoring.

References

1. Jehu, D. A. et al. Risk factors for recurrent falls in older adults: A systematic review with meta-analysis. Maturitas 144, 23–28 (2021).

2. Walsh, J. A. et al. Gait speed and adverse outcomes following hospitalised exacerbation of COPD. Eur. Respir. J. 58, 2004047 (2021).

3. Del Din, S. et al. Analysis of free-living gait in older adults with and without Parkinson’s disease and with and without a history of falls: Identifying generic and disease-specific characteristics. J. Gerontol. A Biol. Sci. Med. Sci. 74, 500–506 (2019).

4. Cameron, M. H. & Nilsagard, Y. Balance, gait, and falls in multiple sclerosis. Handb. Clin. Neurol. 159, 237–250 (2018).

5. Dyer, S. M. et al. A critical review of the long-term disability outcomes following hip fracture. BMC Geriatr. 16, 158 (2016).

6. Middleton, A., Fritz, S. L. & Lusardi, M. Walking speed: the functional vital sign. J. Aging Phys. Act. 23, 314–322 (2015).

7. Kluge, F. et al. Real-world gait detection using a wrist-worn inertial sensor: Validation study. JMIR Form. Res. 8, e50035 (2024).

8. Micó-Amigo, M. E. et al. Assessing real-world gait with digital technology? Validation, insights and recommendations from the Mobilise-D consortium. J. Neuroeng. Rehabil. 20, 78 (2023).

9. Kirk, C. et al. Mobilise-D insights to estimate real-world walking speed in multiple conditions with a wearable device. Sci. Rep. 14, 1754 (2024).

10. Delgado-Ortiz, L. et al. Real-world walking cadence in people with COPD. ERJ Open Res. 10, 00673–02023 (2024).

11. Buekers, J. et al. Laboratory and free-living gait performance in adults with COPD and healthy controls. in M-Health/e-health (European Respiratory Society, 2023). doi:10.1183/13993003.congress-2023.oa3185 .