In the last years, the researchers have consolidated the feasibility of extracting physiological signals by the use of the smartphone camera and motion sensors such as the photo-plethysmography (PPG) and seismocardiography (SCG) signals, that could be useful to overcome the use of standard clinical instrumentations (e.g. the electrocardiogram (ECG)). In this work, the development of an Android 3-modular app for the separate and simultaneous acquisition of the PPG and SCG signals was proposed, aiming to capture the heart rate (HR) and pulse transit time (PTT) parameters of the users through two different experimental protocols. The first protocol was conducted on 7 healthy subjects (28 [27; 29] years old) which utilized autonomously the app in their familiar environment to acquire both PPG, at 30 fps, and SCG signals, separately, one time standing and one time in a supine position. At the end of the experiments, 3 questionnaires were administered to evaluate the app usability and perceived e-health engagement: the system usability scale (SUS); an ad-hoc questionnaire, and the “Twente engagement with eHealth technologies scale” (TWEETS). To extract the HR from the smartphone signals, the fast Fourier’s transform (FFT) algorithm was applied to the unfiltered PPG data and to the z component of the accelerometer signal, that was filtered by an 8th order Butterworth bandpass between [5, 25] Hz. The peak after the DC component was taken and multiplied by 60 to obtain the mean HR as beat-per-minute (bpm). The second protocol aimed to compare the mean HR and PTT parameters of 1 female user (22 years old) extracted by the simultaneous acquisition of the app PPG, at 30, 60 and 120 fps, and SCG to the mean HR and pulse arrival time (PAT) extracted by the synchronized Nexfin ECG and PPG (PLT) signals (“gold standard” measurements) with the collaboration of the Instituto Auxologico of San Luca (MI). The HR were extracted with the same procedure of the first protocol, while the PTT and PAT were estimated as the beat-to-beat time difference between the foot fiducial points of the PPG and SCG AO points, and between the R-peaks and the PLT foot points, respectively. For the first protocol, the t-paired test applied separately to the mean HR of the SCG (p < 0.02) and PPG (p < 0.003) acquired in the standing and supine positions demonstrated the app capacity to capture the participant change of postures. The total median scores of the SUS and the ad-hoc questionnaires of 70 [64;76] and 68 [56; 79], respectively, suggested a good usability of the app, while the total median scores of the first (items 1, 2, 3, for the behavioral engagement), second (items 4, 5, 6, for the cognitive engagement), and third (items 7, 8, 9, for the affective engagements) TWEETS compartments equal to 8 [7; 9], 8 [6; 10] and 6 [4; 8], respectively, underlined the app capacity to facilitate the e-health user engagement. For the second protocol, the mean HR of both PPG and SCG signals differed from the mean HR of the reference measurements with an error < 5%, confirming the feasibility of the app as instrument to retrieve the HR of the users. Also, as the SCG was always acquired at 250 Hz, the accuracy of the HR obtained from the PPG seemed independent from the frame rates (30, 60 or 120 fps) settings of the acquisition. The PTT measurements obtained a low correlation (r^2=0.013) with the PAT measurements, due to the Android programming tools that didn’t allow a built-in synchronization of the PPG and SCG signals, even if the Bland-Altman plot carried out to a CV = 100*SD/bias of around 10% (bias = 83 msec and SD = 8 mesc). This new app was proved to represent a great opportunity, opening the possibility to conduct different studies in as many contexts, spacing from research applications to, perhaps, commercial activities, as measurement tool as well as persuasive technology.
Negli ultimi anni, i ricercatori hanno consolidato la fattibilità di estrarre segnali fisiologici tramite l’utilizzo della fotocamera e dei sensori di movimento dello smartphone come ad esempio la foto-pletismografia (PPG) e la sismocardiografia (SCG), le quali potrebbero aiutare nel superamento dell’uso di strumentazioni cliniche standard (es. l’elettrocardiogramma (ECG)). In questo lavoro, lo sviluppo di una app Android 3-modulare per l’acquisizione separata e simultanea dei segnali PPG ed SCG è stato proposto, con l’obiettivo di acquisire la frequenza cardiaca (HR) ed il “pulse transit time” (PTT) degli utenti attraverso due differenti protocolli sperimentali. Il primo protocollo è stato condotto su 7 soggetti in salute (28 [27; 29] anni), i quali utilizzavano autonomamente la app nel loro ambiente casalingo per acquisire entrambi i segnali PPG, a 30 fps, ed SCG, separatamente, una volta in posizione eretta ed una volta supina. Alla fine degli esperimenti, 3 questionari sono stati somministrati per valutare l’usabilità della app e l’”eHealth engagement” percepito: il “system usability scale” (SUS); un questionario ad-hoc e il “Twente engagement with eHealth technologies scale” (TWEETS). Per estrarre l’HR dai segnali smartphone, l’algoritmo “fast Fourier’s transform” (FFT) è stato applicato ai dati PPG non filtrati e alla componente z del segnale SCG, che è stata filtrata con un filtro passabanda di Butterworth dell’ottavo ordine tra [5, 25] Hz. Il picco dopo la componente DC è stato rilevato e moltiplicato per 60, ottenendo così l’HR in battiti al minuto (bpm). Il secondo protocollo mirava al confronto tra l’HR medio e il PTT di un volontario femmina (22 anni), estratti dall’acquisizione simultanea del PPG, a 30, 60 e 120 fps, e SCG della app, e l’HR medio e il “pulse arrival time” (PAT), estratti dai segnali ECG e PPG (PLT) sincronizzati del monitor Nexfin con la collaborazione dell’Istituto Auxologico di San Luca (MI). L’HR medio è stato estratto con la stessa procedura utilizzata nel primo protocollo, mentre il PTT ed il PAT sono stati stimati tramite le differenze temporali battito-battito tra i punti caratteristici del PPG, chiamati “foot”, e del SCG chiamati “AO”, e quelle tra i picchi-R dell’ECG e i “foot” del PLT, rispettivamente. Per il primo protocollo, i “t-paired test” applicati separatamente agli HR medi del SCG (p < 0.02) e PPG (p < 0.003) acquisiti in posizione eretta e supina hanno dimostrato la capacità della app nel catturare i cambiamenti di postura dei soggetti. I punteggi mediani totali del SUS e del questionario ad-hoc di 70 [64;76] and 68 [56; 79], rispettivamente, hanno suggerito una buona usabilità della app, mentre i punteggi mediani totali del primo (domande 1, 2, 3, per l’engagement comportamentale), del secondo (domande 4, 5, 6, per l’engagement cognitivo) e del terzo (domande 7, 8, 9, per l’engagement emotivo) compartimento uguali a 8 [7; 9], 8 [6; 10] and 6 [4; 8], rispettivamente, sottolineavano la capacità della app nel facilitare l’eHealth engagement degli utenti. Per il secondo protocollo, l’HR medio di entrambi i segnali PPG ed SCG differivano dal HR medio di riferimento con un errore < 5%, confermando la fattibilità di usare l’app come strumento per rilevare l’HR medio degli utenti. Inoltre, come l’SCG era acquisito sempre a 250 Hz, l’accuratezza del HR ottenuta dal PPG sembrava essere indipendente dalla frequenza di cattura settata per l’acquisizione (30, 60 o 120 fps). Le misure del PPT hanno ottenuto una bassa correlazione (r^2=0.013) con le misure del PAT, dovuto alla mancanza degli strumenti di programmazione Android nel poter gestire la sincronizzazione dei segnali PPG e SCG, sebbene il grafico di Bland-Altman abbia portato ad un CV=100*SD/errore (errore = 83 msec e SD = 8 msec) di circa il 10%. Questa nuova app ha dimostrato di rappresentare una grande opportunità, con la possibilità di poter condurre diversi studi in altrettanti contesti, spaziando dalla ricerca a, magari, attività di tipo commerciale, sia come strumento di misurazione che tecnologia persuasiva.
Development and evaluation of an Android app to exploit smartphone-embedded camera and inertial sensors for monitoring of physiological parameters
PIGHINI, CLAUDIO
2018/2019
Abstract
In the last years, the researchers have consolidated the feasibility of extracting physiological signals by the use of the smartphone camera and motion sensors such as the photo-plethysmography (PPG) and seismocardiography (SCG) signals, that could be useful to overcome the use of standard clinical instrumentations (e.g. the electrocardiogram (ECG)). In this work, the development of an Android 3-modular app for the separate and simultaneous acquisition of the PPG and SCG signals was proposed, aiming to capture the heart rate (HR) and pulse transit time (PTT) parameters of the users through two different experimental protocols. The first protocol was conducted on 7 healthy subjects (28 [27; 29] years old) which utilized autonomously the app in their familiar environment to acquire both PPG, at 30 fps, and SCG signals, separately, one time standing and one time in a supine position. At the end of the experiments, 3 questionnaires were administered to evaluate the app usability and perceived e-health engagement: the system usability scale (SUS); an ad-hoc questionnaire, and the “Twente engagement with eHealth technologies scale” (TWEETS). To extract the HR from the smartphone signals, the fast Fourier’s transform (FFT) algorithm was applied to the unfiltered PPG data and to the z component of the accelerometer signal, that was filtered by an 8th order Butterworth bandpass between [5, 25] Hz. The peak after the DC component was taken and multiplied by 60 to obtain the mean HR as beat-per-minute (bpm). The second protocol aimed to compare the mean HR and PTT parameters of 1 female user (22 years old) extracted by the simultaneous acquisition of the app PPG, at 30, 60 and 120 fps, and SCG to the mean HR and pulse arrival time (PAT) extracted by the synchronized Nexfin ECG and PPG (PLT) signals (“gold standard” measurements) with the collaboration of the Instituto Auxologico of San Luca (MI). The HR were extracted with the same procedure of the first protocol, while the PTT and PAT were estimated as the beat-to-beat time difference between the foot fiducial points of the PPG and SCG AO points, and between the R-peaks and the PLT foot points, respectively. For the first protocol, the t-paired test applied separately to the mean HR of the SCG (p < 0.02) and PPG (p < 0.003) acquired in the standing and supine positions demonstrated the app capacity to capture the participant change of postures. The total median scores of the SUS and the ad-hoc questionnaires of 70 [64;76] and 68 [56; 79], respectively, suggested a good usability of the app, while the total median scores of the first (items 1, 2, 3, for the behavioral engagement), second (items 4, 5, 6, for the cognitive engagement), and third (items 7, 8, 9, for the affective engagements) TWEETS compartments equal to 8 [7; 9], 8 [6; 10] and 6 [4; 8], respectively, underlined the app capacity to facilitate the e-health user engagement. For the second protocol, the mean HR of both PPG and SCG signals differed from the mean HR of the reference measurements with an error < 5%, confirming the feasibility of the app as instrument to retrieve the HR of the users. Also, as the SCG was always acquired at 250 Hz, the accuracy of the HR obtained from the PPG seemed independent from the frame rates (30, 60 or 120 fps) settings of the acquisition. The PTT measurements obtained a low correlation (r^2=0.013) with the PAT measurements, due to the Android programming tools that didn’t allow a built-in synchronization of the PPG and SCG signals, even if the Bland-Altman plot carried out to a CV = 100*SD/bias of around 10% (bias = 83 msec and SD = 8 mesc). This new app was proved to represent a great opportunity, opening the possibility to conduct different studies in as many contexts, spacing from research applications to, perhaps, commercial activities, as measurement tool as well as persuasive technology.| File | Dimensione | Formato | |
|---|---|---|---|
|
2020_04_Pighini.pdf
accessibile in internet per tutti
Descrizione: Development and evaluation of an Android app to exploit smartphone-embedded camera and inertial sensors for monitoring of physiological parameters
Dimensione
8.29 MB
Formato
Adobe PDF
|
8.29 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/10589/164863