The continuous monitoring of the breathing is essential and fundamental in many cases, especially in pathological subjects, as patients affected by the Duchenne Muscular Dystrophy (DMD), where respiratory insufficiency occurs at 16-18 years of age and the respiratory and cardiac failure are the main causes of death. Currently, there are several techniques to monitor the breathing, but all of them have different disadvantages, because they are expensive, cumbersome or they can be used only in clinics. Recently, a particular attention was given to accelerometers and inertial sensors to monitor the movement of the thorax and the abdomen and to extract breathing parameters, such as the breathing rate, the inspiratory and expiratory time. Different approaches were proposed in literature, using one accelerometer or one Inertial Measurement Unit (IMU), to register the movement of the thorax or the abdomen due to the breathing. In particular a recent work proposed a system composed by three peripheral units placed on the abdomen, on the thorax and in a reference position (not subject to respiratory movements) able to communicate via the Bluetooth Low Energy (BLE) technology to a central unit connected to a PC. In each unit there are a MARG (Magnetic, Angular Rate and Gravity) sensor array, a microcontroller to implement a sensor fusion algorithm, that obtains the quaternion components relative to the sensor orientation and a BLE module. The tests made in static condition (supine position) produced good results in the estimate of the breathing rate, the inspiration and expiration times, compared to the results obtained with the Optoelectronic Plethysmography (OEP). Starting from the prototype version realized in this study, the aims of this thesis work are: • Realize a better and improved version of the device proposed in the previous study. In particular, each one of the three peripheral units is designed and printed on a single, compact Printed Circuit Board (PCB), contained in a plastic housing. The firmware and the software are also improved to have a respiratory signal with a better quality. Moreover, a new implementation of the device is proposed, composed by two peripheral units (on the thorax and on the abdomen) and one reference central unit, able to receive and save the data from the other two units on an SD card. This reference unit is also able to communicate and interface with a smartphone/tablet/PC using the BLE technology. A prototype version of this implementation is realized and a PCB project is also designed; • Test the device designed and evaluate the accuracy of the parameters estimated (breathing rate, inspiration and expiration times) in different condition: in static conditions, considering two positions (seated and supine), in semi-static conditions, using a wheelchair, pushed by another person and in dynamic conditions, during the gait. For the static acquisitions, the results are compared with the results obtained using contemporaneously the OEP and the trials for each subject involves variation of breathing rate and tidal volume; • Propose a new data processing, that fuses the four quaternion components with the Principal Component Analysis (PCA) and implements an adaptive filter with different analysis methods according to the type of acquisition made (static, semi-static or dynamic). The parameters obtained are averaged over each 3-minute trial performed. The device designed in this thesis is able to work for about 22-23 hours, it is compact , wearable and wireless, so it is suitable for a continuous monitoring of the breathing during all the day. The performances in terms of stability and convergence of the sensor fusion algorithm are sharply improved by the introduction of the sensors calibration and also the data loss during the Bluetooth communication is reduced thanks to the hardware and firmware optimizations. The results obtained for the estimate of the breathing parameters in static conditions show an optimal correlation with the measures of the OEP. In fact, the values of the R2 found in the regression analysis are very high for the breathing rate (R2>0,99) and good for the inspiration and expiration time (R2>0,83). Regarding the errors between the two measures, the best performances are reached during the breathing in normal conditions using the abdomen signal, probably because in these conditions the subjects tested tend to use more the abdomen compartment. For the acquisition in wheelchair, the results are quite good, especially for the parameters estimated using the abdomen signal. This demonstrates that it is possible to extract the respiratory signal also during the movement, partially removing the oscillations introduced by the wheelchair motion. For the dynamic test, the results shows that the gait introduces oscillations in the signal recorded by the units at very specific frequencies, related to the steps cadence; accounting for this effect, it is possible to implement a notch filter to remove these oscillations and to achieve a good estimate of the breathing parameters. However, the limitation of this approach is evident when the breathing rate and the gait have a frequency content very close each other: in this case the notch filter eliminates also the respiratory signal. In conclusion, an improved version of the device proposed in the previous study is presented. It can be used for the long term monitoring of the breathing not only in static conditions, but also in semi-static and dynamic conditions, with a good estimate of the temporal breathing parameters. Moreover, a new version of the device is also proposed, able to save the data on an SD card and to interface with a smartphone/tablet/PC.
Il monitoraggio continuo delle funzioni respiratorie è essenziale e fondamentale in molti casi, specialmente in soggetti patologici, come i pazienti affetti dalla distrofia muscolare di Duchenne (DMD), dove l’insufficienza respiratoria avviene tra i 16 e i 18 anni e l’ arresto cardiaco e respiratorio sono le principali cause di morte. Attualmente, esistono diverse tecniche di monitoraggio del respiro, ma presentano tutte degli svantaggi, poiché sono costose ,ingombranti e spesso possono essere usate solo in ospedali o cliniche. Recentemente, una particolare attenzione è stata data a accelerometri e sensori inerziali in grado di monitorare il movimento dell’addome e del torace dovuto alla respirazione e di ottenere dei parametri respiratori, come la frequenza respiratoria, il tempo di inspirazione e di espirazione. Diversi approcci sono stati proposti in letteratura, usando un accelerometro triassiale o una IMU (Inertial Measurements Unit), per registrare il movimento del torace o dell’addome dovuto alla respirazione. In particolare, un recente studio ha proposto un sistema composto di tre unità periferiche posizionate su addome, torace e in una posizione di riferimento (non soggetta a movimenti respiratori) in grado di comunicare attraverso la tecnologia Bluetooth Low Energy (BLE) a un’ unità centrale collegata ad un computer. In ogni unità sono presenti un array di sensori MARG (Magnetic, Angular Rate and Gravity), un microcontrollore che implementa un algoritmo di “sensor fusion”, ottenendo le componenti del quaternione relativo all’orientazione del sensore e un modulo BLE. I test condotti in situazioni statiche (posizione supina) hanno prodotti buoni risultati per la stima della frequenza respiratoria, del tempo di inspirazione e di espirazione, paragonati ai risultati ottenuti con la pletismografia optoelettronica (OEP). Partendo dalla versione prototipo sviluppata in questo lavoro, gli obiettivi di questa tesi sono: • Realizzare una versione migliorata del dispositivo proposto nello studio precedente. In particolare, ognuna delle tre unità periferiche viene stampata su un singolo compatto circuito stampato (PCB), contenuto in un contenitore di plastica. Anche il firmware delle unità e il software di acquisizione sono migliorati per ottenere un segnale respiratorio con una qualità superiore. Inoltre, viene proposta una nuova implementazione del dispositivo, composta da due unità periferiche (su addome e torace) e un’ unità centrale di riferimento, in grado di ricevere i dati dalle altre due unità, salvarli su una scheda SD e comunicare via Bluetooth con uno smartphone/tablet/PC. Vien anche realizzata una versione prototipo di questa implementazione e un progetto per la PCB viene proposto. • Testare il funzionamento del dispositivo e l’accuratezza della stima dei parametri respiratori (frequenza, tempo di inspirazione e espirazione) in diverse condizioni: statiche, considerando due posizioni (supina e seduta), semistatiche, usando una sedia a rotella spinta da una persona e dinamiche, durante il cammino. Nelle acquisizioni statiche i risultati sono confrontati con quelli ottenuti contemporaneamente dall’OEP e le prove consistono in variazioni di frequenza respiratoria e volume corrente. • Proporre una nuova elaborazione dei dati, che fonde l’informazione delle quattro componenti del quaternione usando la Principal Component Analysis (PCA) e implementa un filtro adattivo con diverse metodologie di analisi in base al tipo di acquisizione (statica, semistatica e dinamica). I parametri ottenuti respiro per respiro vengono mediati su ognuna delle prove, che durano 3 minuti. Il dispositivo progettato in questa tesi è in grado di funzionare per 22-23 ore, è compatto, indossabile e wireless, per cui è adatto per un monitoraggio continuo del respiro durante tutta la giornata. Le prestazioni in termini di stabilità e convergenza dell’algoritmo di “sensor fusion” sono state nettamente migliorate grazie all’introduzione della calibrazione dei 3 sensori e anche la perdita di dati durante la trasmissione Bluetooth è stata migliorata grazie alle ottimizzazioni hardware e firmware. I risultati ottenuti nelle acquisizioni statiche mostrano per i parametri respiratori una correlazione ottimale con i risultati dell’OEP. Infatti, i valori del coefficiente di determinazione R2 calcolati nell’analisi di regressione lineare sono molto alti perla frequenza respiratoria (R2>0,99) e buoni per i tempi di espirazione e inspirazione (R2>0,83). Per quanto riguarda gli errori calcolati tra le due misure, le prestazioni migliori sono raggiunte durante le prove di respiro normale usando il segnale dell’addome, probabilmente perché i soggetti testati usavano maggiormente il compartimento addominale nelle condizioni testate. Anche per le acquisizioni in sedia a rotelle, i risulati sono abbastanza buoni, specialmente quelli ottenuti con il segnale dell’addome. Ciò dimostra che è possibile estrarre il segnale respiratorio anche durante il movimento, riuscendo a rimuovere parzialmente le oscillazioni introdotte dal movimento della sedia a rotelle. Nei test dinamici, i risultati mostrano che il cammino introduce nel segnale registrato dalle unità delle oscillazioni a frequenza specifiche, legate alla cadenza del passo; tenendo conto di questo effetto, è possibile implementare un filtro notch per rimuovere queste oscillazioni dal segnale e raggiungere una buona stima dei parametri respiratori. Comunque, la limitazione di questo approccio è evidente quando la frequnza del respiro e quella legata al cammino sono molto simili: in questo caso un filtro notch elimina entrambi i contributi. In conclusione, una versione migliorata del dispositivo proposto nello studio precedente viene presentata. Può essere utilizzato per un monitoraggio a lungo termine del respiro, non solo in condizioni statiche, ma anche semistatiche o dinamiche. Viene inoltre proposta una versione prototipo del dispositivo in grado di salvare i dati su una scheda SD e di comunicare con uno smartphone/tablet/PC.
A wearable device based on multiple MARG sensors for the respiratory rate monitoring
GANDOLFI, STEFANO
2016/2017
Abstract
The continuous monitoring of the breathing is essential and fundamental in many cases, especially in pathological subjects, as patients affected by the Duchenne Muscular Dystrophy (DMD), where respiratory insufficiency occurs at 16-18 years of age and the respiratory and cardiac failure are the main causes of death. Currently, there are several techniques to monitor the breathing, but all of them have different disadvantages, because they are expensive, cumbersome or they can be used only in clinics. Recently, a particular attention was given to accelerometers and inertial sensors to monitor the movement of the thorax and the abdomen and to extract breathing parameters, such as the breathing rate, the inspiratory and expiratory time. Different approaches were proposed in literature, using one accelerometer or one Inertial Measurement Unit (IMU), to register the movement of the thorax or the abdomen due to the breathing. In particular a recent work proposed a system composed by three peripheral units placed on the abdomen, on the thorax and in a reference position (not subject to respiratory movements) able to communicate via the Bluetooth Low Energy (BLE) technology to a central unit connected to a PC. In each unit there are a MARG (Magnetic, Angular Rate and Gravity) sensor array, a microcontroller to implement a sensor fusion algorithm, that obtains the quaternion components relative to the sensor orientation and a BLE module. The tests made in static condition (supine position) produced good results in the estimate of the breathing rate, the inspiration and expiration times, compared to the results obtained with the Optoelectronic Plethysmography (OEP). Starting from the prototype version realized in this study, the aims of this thesis work are: • Realize a better and improved version of the device proposed in the previous study. In particular, each one of the three peripheral units is designed and printed on a single, compact Printed Circuit Board (PCB), contained in a plastic housing. The firmware and the software are also improved to have a respiratory signal with a better quality. Moreover, a new implementation of the device is proposed, composed by two peripheral units (on the thorax and on the abdomen) and one reference central unit, able to receive and save the data from the other two units on an SD card. This reference unit is also able to communicate and interface with a smartphone/tablet/PC using the BLE technology. A prototype version of this implementation is realized and a PCB project is also designed; • Test the device designed and evaluate the accuracy of the parameters estimated (breathing rate, inspiration and expiration times) in different condition: in static conditions, considering two positions (seated and supine), in semi-static conditions, using a wheelchair, pushed by another person and in dynamic conditions, during the gait. For the static acquisitions, the results are compared with the results obtained using contemporaneously the OEP and the trials for each subject involves variation of breathing rate and tidal volume; • Propose a new data processing, that fuses the four quaternion components with the Principal Component Analysis (PCA) and implements an adaptive filter with different analysis methods according to the type of acquisition made (static, semi-static or dynamic). The parameters obtained are averaged over each 3-minute trial performed. The device designed in this thesis is able to work for about 22-23 hours, it is compact , wearable and wireless, so it is suitable for a continuous monitoring of the breathing during all the day. The performances in terms of stability and convergence of the sensor fusion algorithm are sharply improved by the introduction of the sensors calibration and also the data loss during the Bluetooth communication is reduced thanks to the hardware and firmware optimizations. The results obtained for the estimate of the breathing parameters in static conditions show an optimal correlation with the measures of the OEP. In fact, the values of the R2 found in the regression analysis are very high for the breathing rate (R2>0,99) and good for the inspiration and expiration time (R2>0,83). Regarding the errors between the two measures, the best performances are reached during the breathing in normal conditions using the abdomen signal, probably because in these conditions the subjects tested tend to use more the abdomen compartment. For the acquisition in wheelchair, the results are quite good, especially for the parameters estimated using the abdomen signal. This demonstrates that it is possible to extract the respiratory signal also during the movement, partially removing the oscillations introduced by the wheelchair motion. For the dynamic test, the results shows that the gait introduces oscillations in the signal recorded by the units at very specific frequencies, related to the steps cadence; accounting for this effect, it is possible to implement a notch filter to remove these oscillations and to achieve a good estimate of the breathing parameters. However, the limitation of this approach is evident when the breathing rate and the gait have a frequency content very close each other: in this case the notch filter eliminates also the respiratory signal. In conclusion, an improved version of the device proposed in the previous study is presented. It can be used for the long term monitoring of the breathing not only in static conditions, but also in semi-static and dynamic conditions, with a good estimate of the temporal breathing parameters. Moreover, a new version of the device is also proposed, able to save the data on an SD card and to interface with a smartphone/tablet/PC.File | Dimensione | Formato | |
---|---|---|---|
2017_07_Gandolfi.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Testo della tesi
Dimensione
4.94 MB
Formato
Adobe PDF
|
4.94 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/10589/135452