Breathing frequency (fB) is one of the most important vital sign that can predict clinical adverse events, such as cardiopulmonary arrest and respiratory failure, thus the continuous monitoring of this parameter may avoid hospitalization or admission to intensive care unit by supporting prompt intervene. Recently, particular attention was given to a recent approach using inertial sensors to derive breathing signal by measuring the movement of the thorax and/or the abdomen. The use of these sensors would prompt the development of low-cost, wearable, non-intrusive, and easy to use breathing-monitoring systems. This paper proposes a new automatic algorithm to extract breathing temporal parameters from chest wall (abdomen+thorax) inclination change signals measured using Magnetic and Inertial Measurement Units (MIMUs). An important aspect of the processing algorithm is dimension reduction that allows the extraction of a single respiratory signal starting from 4 quaternion components. Three different methods are proposed and compared in terms of accuracy of breathing temporal parameter estimation, in a group of healthy subjects (n=8), considering different breathing patterns and different postures; Optoelectronic Plethysmography was used as reference system. In this preliminary study, we found that the fusion of the quaternion components by using principal component analysis provided the best fB estimation performance in terms of mean absolute errors (<2 breaths/minute), correlation (r>0.96) and Bland-Altman Analysis. The proposed algorithm and system were able to successfully reconstruct the breathing signal, and to accurately determine the respiratory rate in an automatic, position-independent manner so it has been exploited for a clinical investigation. The second part of the thesis presents the validation of the device in a target population made of neuromuscular patients (n=8 Duchenne, n=1 Limb-Girdle Muscular Dystrophy) in static condition, in supine and seated posture compared with Optoelectronic Plethysmography, as reference system. We obtained good correlation values (R2>0.92) and low relative errors (<6%) for fB. Then the system was used by patients for long-term breathing frequency monitoring. Beyond the utility, users well-accept the device during their daily life. To conclude, in this thesis, a real application of a wearable device based on MIMUs for everyday monitoring is presented with very good results.
La frequenza respiratoria (fB) è uno dei più importanti segni vitali che può predire eventi clinici avversi, come il collasso cardiaco e quello respiratorio, perciò il continuo monitoraggio di questo parametro potrebbe evitare l’ospedalizzazione o la terapia intensiva grazie a un pronto intervento. Recentemente, particolare attenzione è stata rivolta ad un approccio che prevede l’utilizzo di sensori inerziali che misurano i movimenti del torace e/o dell’addome per derivare il segnale respiratorio. L’utilizzo di questi sensori ha portato allo sviluppo di dispositivi economici, indossabili, non fastidiosi e facili da utilizzare per il monitoraggio respiratorio. Questa tesi propone un nuovo algoritmo automatico per estrarre i parametri respiratori temporali a partire dai cambiamenti di inclinazione della parete toracica (addome+torace) misurati con Unità di Misura Inerziali e Magnetiche (MIMU). Un aspetto importante dell’algoritmo per il processamento del segnale è la riduzione delle dimensioni che permette l’estrazione di un singolo segnale respiratorio a partire dalle 4 componenti del quaternione. Sono stati proposti tre metodi differenti e sono stati confrontati tra loro in termini di accuratezza della stima dei parametri temporali, in un gruppo di soggetti sani (n=8), considerando diversi pattern respiratori e differenti posture; la Pletismografia Optoelettronica è stata utilizzata come sistema di riferimento. Attraverso questo studio preliminare, abbiamo ricavato che la fusione delle componenti del quaternione usando l’analisi delle componenti principali ha prodotto la miglior stima di fB in termini di errore assoluto medio (<2 respiri/minuto), correlazione (R2>0.96) ed analisi di Bland-Altman. L’algoritmo ed il sistema proposti sono in grado di ricostruire il segnale respiratorio e di determinare accuratamente la frequenza respiratoria in modo automatico ed indipendente dalla posizione, per questo è stato sfruttato per uno studio clinico. La seconda parte della tesi presenta la validazione del dispositivo in una popolazione fatta da pazienti neuromuscolari (n=8 Duchenne, n=1 Distrofia dei Cingoli) in condizioni statiche, in supino ed in posizione seduta utilizzando come sistema di riferimento la Pletismografia Optoelettronica. Abbiamo ottenuto buoni valori di correlazione (R2>0.92) e bassi errori relativi (<6%) per fB. Il sistema è stato poi usato dai pazienti per il monitoraggio a lungo termine della frequenza respiratoria. Oltre all’utilità, gli utenti hanno accettato di buon grado l’utilizzo del dispositivo durante la loro vita quotidiana. Per concludere, in questa tesi, è stata presentata un’applicazione reale di un dispositivo indossabile basato su MIMU per il monitoraggio respiratorio giornaliero con risultati molto buoni.
Analysis of different algorithms for respiratory rate monitoring through inertial sensors in healthy subjects and patients with neuromuscular diseases
PREVITALI, YLENIA
2017/2018
Abstract
Breathing frequency (fB) is one of the most important vital sign that can predict clinical adverse events, such as cardiopulmonary arrest and respiratory failure, thus the continuous monitoring of this parameter may avoid hospitalization or admission to intensive care unit by supporting prompt intervene. Recently, particular attention was given to a recent approach using inertial sensors to derive breathing signal by measuring the movement of the thorax and/or the abdomen. The use of these sensors would prompt the development of low-cost, wearable, non-intrusive, and easy to use breathing-monitoring systems. This paper proposes a new automatic algorithm to extract breathing temporal parameters from chest wall (abdomen+thorax) inclination change signals measured using Magnetic and Inertial Measurement Units (MIMUs). An important aspect of the processing algorithm is dimension reduction that allows the extraction of a single respiratory signal starting from 4 quaternion components. Three different methods are proposed and compared in terms of accuracy of breathing temporal parameter estimation, in a group of healthy subjects (n=8), considering different breathing patterns and different postures; Optoelectronic Plethysmography was used as reference system. In this preliminary study, we found that the fusion of the quaternion components by using principal component analysis provided the best fB estimation performance in terms of mean absolute errors (<2 breaths/minute), correlation (r>0.96) and Bland-Altman Analysis. The proposed algorithm and system were able to successfully reconstruct the breathing signal, and to accurately determine the respiratory rate in an automatic, position-independent manner so it has been exploited for a clinical investigation. The second part of the thesis presents the validation of the device in a target population made of neuromuscular patients (n=8 Duchenne, n=1 Limb-Girdle Muscular Dystrophy) in static condition, in supine and seated posture compared with Optoelectronic Plethysmography, as reference system. We obtained good correlation values (R2>0.92) and low relative errors (<6%) for fB. Then the system was used by patients for long-term breathing frequency monitoring. Beyond the utility, users well-accept the device during their daily life. To conclude, in this thesis, a real application of a wearable device based on MIMUs for everyday monitoring is presented with very good results.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi finale con modifiche_Ylenia.pdf
solo utenti autorizzati dal 04/12/2021
Descrizione: Testo della tesi
Dimensione
8.19 MB
Formato
Adobe PDF
|
8.19 MB | 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/10589/145188