An article recently published in the journal Maturitas explored the use of artificial intelligence (AI) and computer vision to assess gait and better understand fall risks in people with Parkinson’s disease (PD). The researchers aimed to overcome the limitations of current inertial-based gait assessments by integrating wearable video glasses with AI to provide a more complete analysis of fall risk factors.
Study: Better understanding fall risk: AI-based computer vision for contextual gait assessment. Image Credit: Microgen/Shutterstock.com
Background
PD is a neurodegenerative disorder characterized by symptoms like tremors, stiffness, and slow movements, which significantly affect quality of life. A major concern for people with PD is the risk of falling, which can lead to serious injuries or even death.
Gait issues are common in PD and serve as indicators of disease progression and fall risk. Traditional gait assessments, such as visual observation and instrumented walkways, have evolved to include wearable inertial measurement units (IMUs).
IMUs are portable, easy to use, and cost-effective, allowing gait tracking in real-world settings. While IMUs are valuable for monitoring daily activities, they do not capture visual information, which is important for understanding the context of gait problems.
About the Research
In this paper, the authors aimed to improve the understanding of fall risk in people with PD by combining IMU data with visual information from wearable video glasses. They used eye-tracking video glasses, along with AI-based computer vision, to gather data efficiently and ethically in real-world environments.
Three participants with PD were equipped with eye-tracking video glasses and an IMU worn around the waist. The devices were time-synchronized, and participants were instructed to do their daily activities for approximately three hours.
The IMU data provided key gait metrics, such as initial and final contact events, essential for assessing clinically relevant gait characteristics like step time and swing time asymmetry.
The researchers used the you-only-look-once version 8 (YOLOv8) computer vision model, trained with a dataset of annotated home environment images. This model was designed to anonymize and contextualize video data by detecting and blurring sensitive objects, such as people and written information. Additionally, it identified participants' walking paths and obstacles, providing valuable context for interpreting the IMU data.
Key Findings
The outcomes showed that integrating IMU data with AI-based computer vision provided a more complete understanding of fall risk in PD patients. The authors observed that while IMU data alone could identify abnormal gait patterns, adding contextual information from video glasses provided critical insights into the reasons behind these abnormalities.
By providing a more comprehensive overview of intrinsic and extrinsic factors influencing gait, their approach could lead to more accurate and personalized interventions to reduce fall risk.
For example, one participant displayed stable gait signals during continuous walking on a level terrain, as indicated by the IMU data. However, another participant showed unstable gait patterns at certain moments. The video revealed that these fluctuations happened when navigating stairs and obstacles, highlighting the importance of context in assessing fall risk.
The AI-based computer vision model also successfully anonymized video data by blurring sensitive objects, addressing privacy concerns in video research. The model also provided a detailed environmental context, which is important for accurate fall risk assessment.
Furthermore, the researchers used eye-tracking data to study gaze behavior and visual attention. They found that this data provided further insights into fall risk, especially when combined with IMU and computer vision data. For example, unusual gaze behavior while navigating a doorway and steps was detected by the computer vision model.
Applications
This research has promising applications in both clinical and research settings. It can help develop more precise and personalized fall risk assessments for individuals with PD and other neurodegenerative conditions.
By capturing internal and external factors, clinicians can better tailor interventions to reduce fall risk and improve outcomes. The technology can be used in home settings, providing a more natural and continuous way to assess gait compared to traditional methods.
This approach could also benefit other populations at risk of falls, such as older adults or those with mobility challenges. Its ability to monitor gait and environment in real-world conditions offers a more accurate and natural assessment than clinical methods.
Conclusion
In conclusion, this novel approach has proven effective in enhancing the understanding of fall risks in individuals with PD and has the potential to transform how healthcare professionals, such as doctors and nursing staff, assess fall-related factors. By addressing the limitations of traditional inertial-based assessments, it opens new avenues for future research and clinical application.
Further work should aim to refine these methods and validate the findings with larger groups. Integrating additional sensing technologies, such as light detection and ranging (LiDAR), could enhance environmental context accuracy, improving fall prevention strategies. Additionally, research on visual attention using eye-tracking videos may offer valuable insights into how individuals adjust their gait to navigate environmental challenges.
Journal Reference
Moore, J., & et, al. Better understanding fall risk: AI-based computer vision for contextual gait assessment. Maturitas, 2024, 189, 108116. DOI: 10.1016/j.maturitas.2024.108116, https://www.sciencedirect.com/science/article/pii/S0378512224002111
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.
Article Revisions
- Sep 19 2024 - Revised sentence structure, word choice, punctuation, and clarity to improve readability and coherence.