Phybrata Sensors and Machine Learning for Enhanced Neurophysiological Diagnosis and Treatment

Abstract

Concussion injuries remain a significant public health challenge. A significant unmet clinical need remains for tools that allow related physiological impairments and longer-term health risks to be identified earlier, better quantified, and more easily monitored over time. We address this challenge by combining a head-mounted wearable inertial motion unit (IMU)-based physiological vibration acceleration ("phybrata") sensor and several candidate machine learning (ML) models. The performance of this solution is assessed for both binary classification of concussion patients and multiclass predictions of specific concussion-related neurophysiological impairments. Results are compared with previously reported approaches to ML-based concussion diagnostics. Using phybrata data from a previously reported concussion study population, four different machine learning models (Support Vector Machine, Random Forest Classifier, Extreme Gradient Boost, and Convolutional Neural Network) are first investigated for binary classification of the test population as healthy vs. concussion (Use Case 1). Results are compared for two different data preprocessing pipelines, Time-Series Averaging (TSA) and Non-Time-Series Feature Extraction (NTS). Next, the three best-performing NTS models are compared in terms of their multiclass prediction performance for specific concussion-related impairments: vestibular, neurological, both (Use Case 2). For Use Case 1, the NTS model approach outperformed the TSA approach, with the two best algorithms achieving an F1 score of 0.94. For Use Case 2, the NTS Random Forest model achieved the best performance in the testing set, with an F1 score of 0.90, and identified a wider range of relevant phybrata signal features that contributed to impairment classification compared with manual feature inspection and statistical data analysis. The overall classification performance achieved in the present work exceeds previously reported approaches to ML-based concussion diagnostics using other data sources and ML models. This study also demonstrates the first combination of a wearable IMU-based sensor and ML model that enables both binary classification of concussion patients and multiclass predictions of specific concussion-related neurophysiological impairments.

Keywords: concussion; machine learning; neurological; physiological impairment; vestibular; wearable sensor.

Figure 1

Classification use cases investigated in the present study.

Figure 2

The Random Forest SHAP values for all patients in the training set (n = 108). Features are ranked top to bottom (top being the largest contributor). A single point represents a patient‘s SHAP value for a given feature. Along the x-axis, a positive (negative) SHAP value indicates the features’ impact toward classifying a patient as concussed (healthy control). Color indicates the actual value of a subject’s feature value.

Figure 3

ROC curves for multiclass impairment prediction using the RF model and NTS preprocessing pipeline. Each curve represents the specific impairment vs. the rest of the classes (one vs. rest) and depicts the trade-off between True Positive Rates (TPR) and False Positive Rates (FPR) for each physiological impairment condition. The ROC performance is based upon the optimal threshold selected by the model for the testing set. The three colored curves correspond to vestibular impairment = green, neurological impairment = blue, both = yellow. The dotted black line represents random performance.

Figure 4

Testing set confusion matrix of RF model. Model predictions for each impairment (x-axis) are contrasted with actual impairment outcomes (y-axis) to categorize the correct and incorrect predictions made.

Figure 5

Mean SHAP value contribution (x-axis) of each feature (y-axis) for every impairment condition (blue = neurological, pink = vestibular, green = both), ranked from top to bottom in terms of importance. Taking each class together, the total mean SHAP value reflects each feature’s global impact on model classification.