Walking Stability in Patients With Benign Paroxysmal Positional Vertigo: An Objective Assessment Using Wearable Accelerometers and Machine Learning

Yuqian Zhang, He Wang, Yifei Yao, Jian‐Ren Liu, Xuhong Sun, Dongyun Gu

Abstract

Background: Benign paroxysmal positional vertigo (BPPV) is one of the most common peripheral vestibular disorders leading to balance difficulties and increased fall risks. This study aims to investigate the walking stability of BPPV patients in clinical settings and propose a machine-learning-based classification method for determining the severity of gait disturbances of BPPV. Methods: Twenty-seven BPPV outpatients and twenty-seven healthy subjects completed level walking trials at self-preferred speed in clinical settings while wearing one accelerometer on the head and one on the lower trunk. Temporo-spatial variables and six walking stability related variables (root mean square (RMS), harmonic ratio (HR), gait variability, step/stride regularity, and gait symmetry) derived from the acceleration signals were analyzed. A support vector machine model (SVM) based on the gait variables of BPPV patients were developed to classify the BPPV severity of gait disturbances. Results: The results showed that BPPV patients employed a conservative gait and significantly reduced walking stability compared to the healthy controls. Significant different mediolateral HR at the lower trunk and anteroposterior step regularity at the head were found in BPPV patients among mild, moderate, and severe DHI (dizziness handicap inventory) subgroups. SVM classification achieved promising accuracies with area under the curve (AUC) = 0.87, 0.80, and 0.95 respectively for classifying the three stages of DHI subgroups. Conclusions: Results suggested that the proposed gait analysis that is based on the coupling of wearable accelerometers and machine learning provides an objective approach for assessing gait disturbances and handicapping effects of dizziness imposed by BBPV.

Introduction

Benign Paroxysmal Positional Vertigo (BPPV) is a prevalent inner-ear disorder characterized by brief episodes of intense dizziness or spinning sensations triggered by changes in head position. These symptoms often lead to significant balance impairments and an elevated risk of falls, particularly in older populations. Currently, clinical diagnosis relies heavily on the Dix-Hallpike maneuver, which has a sensitivity of approximately 79%. Furthermore, assessing the functional impact of BPPV typically depends on the Dizziness Handicap Inventory (DHI), a subjective self-report questionnaire.

There is a critical need for objective, quantifiable measures of gait instability that can be easily implemented in clinical environments. While laboratory-based 3D motion capture systems offer high precision, their cost and complexity limit their utility in routine clinical practice. This study explores the use of wearable accelerometers to provide a "digital signature" of BPPV-related instability, offering a portable and objective alternative for assessing disease severity.

Methods

The study included 27 patients diagnosed with idiopathic unilateral BPPV of the posterior semicircular canal and 27 age-, weight-, and height-matched healthy controls. Participants performed 20-meter walking trials at their self-preferred speed in a hospital corridor. Two wireless tri-axial accelerometers were used: one positioned on the head and another on the lower trunk (L3 vertebrae).

Data analysis focused on several key stability metrics derived from the acceleration signals:

  • Temporo-spatial variables: Walking speed, step length, and cadence.
  • Root Mean Square (RMS): Measuring the magnitude of acceleration in vertical (VT), mediolateral (ML), and anteroposterior (AP) directions.
  • Harmonic Ratio (HR): Assessing gait smoothness and rhythmicity.
  • Step/Stride Regularity and Gait Symmetry: Quantifying the consistency and balance of the gait cycle.

To classify the severity of gait disturbances, a Support Vector Machine (SVM) model with a linear kernel was developed. Patients were categorized into three subgroups based on their DHI scores: Mild (0–30), Moderate (31–60), and Severe (61–100). The model's performance was evaluated using 5-fold cross-validation.

Results

BPPV Patients vs. Healthy Controls

The analysis revealed that BPPV patients adopt a "conservative gait" strategy, characterized by significantly lower walking speeds, shorter step lengths, and reduced cadence compared to healthy controls. Despite this cautious approach, patients exhibited significantly impaired stability. Specifically, BPPV patients showed lower step regularity across all three axes at the head and reduced gait symmetry at the lower trunk.

Comparison of gait variables between BPPV patients and healthy controls.

As shown in the data, significant differences were observed in several RMS and HR metrics, particularly at the head and trunk, indicating that the vestibular dysfunction in BPPV directly impacts the body's ability to maintain a stable and rhythmic gait.

Severity Classification and Machine Learning

The SVM model demonstrated high accuracy in distinguishing between the three DHI severity levels. The Area Under the Curve (AUC) values were 0.87 for the mild stage, 0.80 for the moderate stage, and 0.95 for the severe stage.

Gait variables across different DHI severity levels in BPPV patients.

Key objective markers for disease progression were identified: mediolateral (ML) Harmonic Ratio at the trunk and anteroposterior (AP) step regularity at the head. These metrics decreased significantly as the self-reported DHI scores increased, providing a clear quantitative link between subjective handicap and objective gait instability.

Discussion and Conclusion

This study demonstrates that wearable accelerometers, combined with machine learning, can effectively quantify the gait disturbances associated with BPPV. The findings highlight that even when patients attempt to compensate with a slower, more cautious gait, objective stability metrics remain significantly compromised. The direct measurement of head dynamics proved particularly valuable, likely due to the proximity of the sensors to the vestibular system.

While the results are promising, the study is limited by a relatively small sample size, particularly in the severe subgroup, and a focus solely on posterior canal BPPV. Future research should expand to other BPPV types and explore more advanced deep learning architectures. Nonetheless, this approach offers a viable, objective tool for clinicians to assess the handicapping effects of dizziness and monitor patient progress in real-world clinical settings.

Summary updated

Abstract

<title>Abstract</title> <bold>Background:</bold> Benign paroxysmal positional vertigo (BPPV) is one of the most common peripheral vestibular disorders leading to balance difficulties and increased fall risks. This study aims to investigate the walking stability of BPPV patients in clinical settings and propose a machine-learning-based classification method for determining the severity of gait disturbances of BPPV. <bold>Methods:</bold> Twenty-seven BPPV outpatients and twenty-seven healthy subjects completed level walking trials at self-preferred speed in clinical settings while wearing one accelerometer on the head and one on the lower trunk. Temporo-spatial variables and six walking stability related variables (root mean square (RMS), harmonic ratio (HR), gait variability, step/stride regularity, and gait symmetry) derived from the acceleration signals were analyzed. A support vector machine model (SVM) based on the gait variables of BPPV patients were developed to classify the BPPV severity of gait disturbances. <bold>Results</bold>: The results showed that BPPV patients employed a conservative gait and significantly reduced walking stability compared to the healthy controls. Significant different mediolateral HR at the lower trunk and anteroposterior step regularity at the head were found in BPPV patients among mild, moderate, and severe DHI (dizziness handicap inventory) subgroups. SVM classification achieved promising accuracies with area under the curve (AUC) = 0.87, 0.80, and 0.95 respectively for classifying the three stages of DHI subgroups. <bold>Conclusions:</bold> Results suggested that the proposed gait analysis that is based on the coupling of wearable accelerometers and machine learning provides an objective approach for assessing gait disturbances and handicapping effects of dizziness imposed by BBPV.<bold>Trial registration: T</bold>he trial was registered in the Chinese Clinical Trial Registry (http://www.chictr.org.cn) on March 29, 2018. Registration number: ChiCTR1800015432 (http://www.chictr.org.cn/showproj.aspx?proj=25587).
BibTeX
@article{zhang2021walking,
  author = {Yuqian Zhang and He Wang and Yifei Yao and Jian‐Ren Liu and Xuhong Sun and Dongyun Gu},
  title = {{Walking Stability in Patients With Benign Paroxysmal Positional Vertigo: An Objective Assessment Using Wearable Accelerometers and Machine Learning}},
  journal = {Journal of NeuroEngineering and Rehabilitation},
  year = {2021},
  doi = {10.1186/s12984-021-00854-y},
}
Figure