Post-stroke respiratory complications using machine learning with voice features from mobile devices

Abstract

Abnormal voice may identify those at risk of post-stroke aspiration. This study was aimed to determine whether machine learning algorithms with voice recorded via a mobile device can accurately classify those with dysphagia at risk of tube feeding and post-stroke aspiration pneumonia and be used as digital biomarkers. Voice samples from patients referred for swallowing disturbance in a university-affiliated hospital were collected prospectively using a mobile device. Subjects that required tube feeding were further classified to high risk of respiratory complication, based on the voluntary cough strength and abnormal chest x-ray images. A total of 449 samples were obtained, with 234 requiring tube feeding and 113 showing high risk of respiratory complications. The eXtreme gradient boosting multimodal models that included abnormal acoustic features and clinical variables showed high sensitivity levels of 88.7% (95% CI 82.6–94.7) and 84.5% (95% CI 76.9–92.1) in the classification of those at risk of tube feeding and at high risk of respiratory complications; respectively. In both cases, voice features proved to be the strongest contributing factors in these models. Voice features may be considered as viable digital biomarkers in those at risk of respiratory complications related to post-stroke dysphagia.

Prediction of dysphagia aspiration through machine learning-based analysis of patients’ postprandial voices

“…[38] Some studies have employed the Praat program to extract these sound parameters and analyze each indicator, either using voice-only or combining voice with clinical data indicators, trained with algorithms such as Logistic Regression, Decision Tree, Random Forest, SVM, GMM, and XGBoost. [12] Another study reported the results of dysphagia prediction using speci c phonation or articulation features trained using support vector machine (SVM), random forest, and other methods. [39] However, these studies have limitations in that they only analyzed speci c numerical indicators of voice and failed to analyze the overall voice itself.…”

“…[14][15][16] Previous studies analyzing the voice of patients with dysphagia have reported signi cant changes in parameters such as RAP, SHIM, and NHR due to aspiration into the airway. [11][12][13][14] However, these studies often extracted speci c vocal parameters rather than analyzing the patient's voice itself, which may limit their universal application in diagnosis and monitoring.…”

Computer-aided screening of aspiration risks in dysphagia with wearable technology: a Systematic Review and meta-analysis on test accuracy

“…Engineering and Computing (Sarraf Shirazi et al, 2012;Sarraf Shirazi et al, 2014), IEEE Transaction of Biomedical Engineering (Sejdic et al, 2013), IEEE Journal of Biomedical and Health Informatics (Shu et al, 2022), and Scientific Reports (Park et al, 2022) that spanned across the clinical, engineering, and interdisciplinary science fields.…”

“…The authors attempted to separate the sounds of breathing and swallowing through an aural and visual examination of the time-frequency signal spectrum. On the other hand, while Frakking et al (2022) made use of an omnidirectional condenser microphone (C417, AKG Acoustics, Vienna, Austria) in the form of a circular O-ring lateral to the cricoid cartilage, Park et al (2022) recorded the speaking sound using an iPad (with an embedded microphone) placed 20 cm in front of the participants' faces (Umayahara et al, 2018). The sampling frequency was 44.1 kHz and was band-passed between Regarding the protocol for measurement, most of them referred to the standard swallowing assessment procedures that fed food/ liquid with different consistencies or thicknesses (Lee et al, 2006;Lee et al, 2011;Merey et al, 2012;Frakking et al, 2022;Shu et al, 2022).…”

“…Nonetheless, classification performance for other GAN models seemed to be better than the proposed one. Park et al (2022) classified the acoustic signal using XGBoost, which produced an Frontiers in Bioengineering and Biotechnology frontiersin.org accuracy of 74.8%, yet the performance was better than the model using both the acoustic signal and clinical data.…”