Discrimination of vestibular function based on inertial sensors

Liu, Xinyu; Yu, Shudong; Zang, Xiaohan; Yu, Qianru; Yang, Licai

doi:10.1016/j.cmpb.2021.106554

Cited by 1 publication

(5 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ML models pair naturally as analytic tools, for the time series data generated by sensors. Recently, investigators have used sensors and ML to classify healthy patients versus those with unilateral vestibular hypofunction 8,26 . Using an accelerometer to assess gait, Zhang et al applied an SVM model to classify healthy controls versus patients with benign paroxysmal positional vertigo and achieved an AUC of 0.78 7 .…”

Section: Discussionmentioning

confidence: 99%

“…Further, the ideal placement of a sensor and which sensors are optimal for evaluating vestibulopathy remain unclear. Placement is often at the hip or trunk, as is performed here, whereas others have evaluated sensor placement on the head 7,26 . In regard to sensor type, there is some evidence that magnetometers may provide added value compared to accelerometers and gyroscopes 26 …”

Section: Discussionmentioning

confidence: 99%

“…Recently, investigators have used sensors and ML to classify healthy patients versus those with unilateral vestibular hypofunction. 8,26 Using an accelerometer to assess gait, Zhang et al applied an SVM model to classify healthy controls versus patients with benign paroxysmal positional vertigo and achieved an AUC of 0.78. 7 Ahmadi and colleagues paired body sway patterns and ML to diagnose a large cohort of 293 adults, including patients with peripheral disorders (unilateral vestibular hypofunction) and central disorders (anterior lobe cerebellar atrophy), among others who were healthy or had neuromotor disease, such as Parkinson's.…”

Section: Discussionmentioning

confidence: 99%

“…Placement is often at the hip or trunk, as is performed here, whereas others have evaluated sensor placement on the head. 7,26 In regard to sensor type, there is some evidence that magnetometers may provide added value compared to accelerometers and gyroscopes. 26 This study does not make use of pure physical activity metrics and includes participant age and gender.…”

Section: Discussionmentioning

confidence: 99%

“…7,26 In regard to sensor type, there is some evidence that magnetometers may provide added value compared to accelerometers and gyroscopes. 26 This study does not make use of pure physical activity metrics and includes participant age and gender. Age was found to be the most important feature for classification by the SVM model, which is consistent with the welldocumented observation that the odds of developing vestibular dysfunction significantly increase with aging.…”

Section: Discussionmentioning

confidence: 99%

See 4 more Smart Citations

Machine Learning Analysis of Physical Activity Data to Classify Postural Dysfunction

et al. 2023

View full text Add to dashboard Cite

BackgroundMachine learning (ML) analysis of biometric data in non‐controlled environments is underexplored.ObjectiveTo evaluate whether ML analysis of physical activity data can be employed to classify whether individuals have postural dysfunction in middle‐aged and older individuals.MethodsA 1 week period of physical activity was measured by a waist‐worn uni‐axial accelerometer during the 2003–2004 National Health and Nutrition Examination Survey sampling period. Features of physical activity along with basic demographic information (42 variables) were paired with ML models to predict the success or failure of a standard 30 s modified Romberg test during which participants had their eyes closed and stood upon a 3‐inch compliant surface. Model performance was evaluated by area under the receiver operating characteristic curve (AUC‐ROC), balanced accuracy, and F1‐score.ResultsThe cohort was comprised of 1625 participants ≥40 years (median age 61, IQR 51–71). Approximately half (47%) were diagnosed with postural dysfunction having failed the binarized (pass/fail) scoring mechanism of the modified Romberg exam. Five ML models were trained on the classification task, achieving AUC values ranging from 0.67 to 0.73. The support vector machine (SVM) and a gradient‐boosted model, XGBoost, achieved the highest AUC of 0.73 (SD 0.71–0.75). Age was the most important variable for SVM classification, followed by four features that evaluated accelerometer counts at various thresholds, including those delineating total, moderate, and moderate‐vigorous activity.ConclusionsML analysis of accelerometer‐derived physical activity data to classify postural dysfunction in middle‐aged and older individuals is feasible in real‐world environments such as the home.Level of Evidence3 Laryngoscope, 2023

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%