Heart rate variability (HRV) is commonly used as a clinical measure to assess Autonomic Nervous System function and overall health. Various factors, including age, gender, physical fitness, and physiological conditions, can influence HRV. This study focuses on evaluating electrocardiogram (ECG) features during a complex postural control task in a Virtual Reality (VR) environment to determine their significance in classifying subjects who experienced Motion Sickness (MS) symptoms. The study utilized the BioVRSea setup, which combines VR with a platform that simulates waves to induce MS in subjects. HR, along with other biosignals, was measured using advanced ECG sensors. A binary index extracted from a questionnaire was used to evaluate and differentiate individuals based on symptom changes. Statistical analysis and Machine Learning (ML) models were employed to determine the most significant HRV features in classifying subjects with MS symptoms during the postural control task. Seventy healthy volunteers participated in the experiment, and a total of 100 HRV features were obtained from the ECG signals considering all the different phases of the experiment. The statistical analysis revealed six features that showed statistically significant differences between subjects with and without MS symptoms. ML models, including Random Forest, Gradient Boosting and Logistic Regression algorithms, were trained using different wrapper feature selection techniques. The best-performing model achieved an accuracy of 77.1%, precision of 62.1%, recall of 67.4%, and F score of 85.4%. This study highlights the importance of ECG features in classifying MS symptoms during a complex postural control task in a VR environment. The findings contribute to understanding the autonomic responses and cardiac control mechanisms associated with MS. The results can have implications for future research on MS susceptibility and the development of personalized interventions to mitigate MS symptoms.
La variabilità della frequenza cardiaca è una misura clinica utilizzata per valutare la funzione del sistema nervoso autonomo e le condizioni di salute complessive dei pazienti. Diversi fattori come età, sesso, forma fisica e condizioni fisiologiche possono influenzare tale variabilità. Questo studio è incentrato sull'analisi delle features cardiache durante un compito di controllo posturale complesso in un ambiente di realtà virtuale (VR) con piattaforma mobile (BioVRSea) al fine di classificare dei soggetti affetti da cinetosi (Motion Sickness). Per valutare e quantificare i sintomi del Motion Sickness, è stato utilizzato un questionario specifico sulla base del quale è stato calcolato un indice binario per differenziare gli individui in base ai sintomi. Sono stati impiegati modelli di machine learning e tecniche statistiche per determinare le features cardiache più significative nella classificazione dei soggetti con sintomi durante l'esperimento. Nello studio sono stati coinvolti 70 volontari sani, dai quali sono state ottenute un totale di 100 features cardiache dai segnali ECG, considerando le diverse fasi dell'esperimento. L'analisi statistica ha rivelato che 6 features mostravano differenze statisticamente significative tra i soggetti con e senza sintomi del Motion Sickness. Sono stati addestrati numerosi modelli di machine learning, tra cui Gradient Boosting, Random Forest e Logistic Regression, utilizzando diverse tecniche di selezione delle features. Il modello più performante ha raggiunto un'accuratezza del 77,1%, una precisione del 62,1%, una sensibilità del 67.4% e un punteggio complessivo dell'85,4%. I risultati di questo studio contribuiscono alla comprensione delle risposte autonomiche e dei meccanismi di controllo cardiaco associati al Motion Sickness. Inoltre, tali risultati possono avere implicazioni per la futura ricerca sulla suscettibilità al Motion Sickness e lo sviluppo di terapie paziente-specifiche per mitigarne i sintomi.
ECG features analysis using machine learning for motion sickness prediction during a complex postural control task
LINDEMANN, ALESSIA
2022/2023
Abstract
Heart rate variability (HRV) is commonly used as a clinical measure to assess Autonomic Nervous System function and overall health. Various factors, including age, gender, physical fitness, and physiological conditions, can influence HRV. This study focuses on evaluating electrocardiogram (ECG) features during a complex postural control task in a Virtual Reality (VR) environment to determine their significance in classifying subjects who experienced Motion Sickness (MS) symptoms. The study utilized the BioVRSea setup, which combines VR with a platform that simulates waves to induce MS in subjects. HR, along with other biosignals, was measured using advanced ECG sensors. A binary index extracted from a questionnaire was used to evaluate and differentiate individuals based on symptom changes. Statistical analysis and Machine Learning (ML) models were employed to determine the most significant HRV features in classifying subjects with MS symptoms during the postural control task. Seventy healthy volunteers participated in the experiment, and a total of 100 HRV features were obtained from the ECG signals considering all the different phases of the experiment. The statistical analysis revealed six features that showed statistically significant differences between subjects with and without MS symptoms. ML models, including Random Forest, Gradient Boosting and Logistic Regression algorithms, were trained using different wrapper feature selection techniques. The best-performing model achieved an accuracy of 77.1%, precision of 62.1%, recall of 67.4%, and F score of 85.4%. This study highlights the importance of ECG features in classifying MS symptoms during a complex postural control task in a VR environment. The findings contribute to understanding the autonomic responses and cardiac control mechanisms associated with MS. The results can have implications for future research on MS susceptibility and the development of personalized interventions to mitigate MS symptoms.File | Dimensione | Formato | |
---|---|---|---|
2023_12_Lindemann_Tesi_01.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Testo della tesi
Dimensione
22.65 MB
Formato
Adobe PDF
|
22.65 MB | Adobe PDF | Visualizza/Apri |
2023_12_Lindemann_Executive Summary_02.pdf
non accessibile
Descrizione: Executive Summary
Dimensione
954.48 kB
Formato
Adobe PDF
|
954.48 kB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/214382