Objective dyspnea measurement in real time using a forehead wearable and artificial intelligence

Area of Science:

• Physiology
• Biomedical Engineering
• Machine Learning

Background:

• Dyspnea is a complex symptom typically measured via subjective patient ratings.
• Continuous, automated dyspnea monitoring is crucial in critical care and trauma where patient communication may be impaired.

Purpose of the Study:

• To develop and validate an objective dyspnea score (ODS) using physiologic sensor data and machine learning.
• To assess the accuracy of the ODS in predicting patient-reported breathlessness and exertion.

Main Methods:

• Prospective enrollment of 54 pulmonary rehabilitation subjects undergoing treadmill trials.
• Automated collection of dyspnea measurements via physiologic sensors and patient-reported ratings (RPB, RPE) at one-minute intervals.
• Training machine learning models (7-feature and 19-feature) to generate ODS, with performance assessed on a held-out test set.

Main Results:

• The 7-feature machine learning model achieved the best performance.
• The ODS demonstrated strong correlation with patient-reported scores: 0.84 (78.7% accuracy) for RPE and 0.86 (83.6% accuracy) for RPB.
• The developed system accurately predicts subjective dyspnea levels.

Conclusions:

• The objective dyspnea score (ODS) system shows promise for automated, real-time dyspnea measurement.
• This technology can aid in monitoring patients, particularly those with communication impairments in critical settings.
• Further validation in diverse clinical populations is warranted.