In the clinical context the calculation of energy consumption during walking of a subject with movement abnormalities is very important because it allows to compare a normal locomotion to a pathological one, or to quantify the severity of a motor disorder, as also makes it possible to evaluate the effectiveness of a rehabilitation therapy or a surgical intervention. Currently indirect calorimetry is by far the most commonly used method for estimating energy expenditure during physical activity and is based on knowledge of oxygen consumption and carbon dioxide production, which occur when the muscles are working to produce energy during movement. In particular, the ergospirometer, a metabolic system that can be both mobile and stationary, currently constitutes the reference device for estimating energy consumption during activity-related energy expenditure, or AEE, both in terms of accuracy and practicality of utilization. But altermately other wearable devices can be used, such as inertial systems or other metabolic-motor sensors, much less expensive and manageable, which however need to be validated, comparing their measurements with those provided by a reference device. In this regard, in this thesis work, along with the energy expenditure measures provided by the ergospirometer, physiological and motor data were acquired during a walking test, through an inertial measurement unit, or IMU, and a metabolic-motor holter, in order to determine if these two sensors can act as a valid alternative to the ergospirometer (gold standard) for the estimation of the calories expended in a walk of 6 minutes. Specifically, as far as the IMU is concerned, we wanted to see if its acceleration data can be used to predict energy consumption, so various simple linear regression models have been obtained, in order to relate its accelerations (or rather their elaborations ) with the AEE supplied by the ergospirometer; subsequently the more representative acceleration variable was taken into consideration, together with the weight variable and other kinematic quantities (always estimated by the IMU) relating to the test, in order to develop multiple linear regression models that would explain the AEE variable even better. With regards to the metabolic-motor holter, which directly outputs the energy consumption, a correlation model was obtained between the AEE values supplied by it and those obtained from the ergospirometer, to determine whether the two devices provide a comparable measurement of the energy consumption during the walking test. In this study 6 healthy subjects (2 women and 4 men) were recruited, without particular inclusion criteria, with age = 30 ± 7.3 years, weight = 68.9 ± 16.3 kg and height = 1.72 ± 0.09 m, which repeated the trial twice, in order to have 12 experimental tests to analyze. The test consisted of walking as fast as possible for 6 minutes along a straight path of a length of 25 meters, with the aim of making as many meters as possible with respect to one's abilities, managing the effort independently. For the acquisitions the subjects were equipped with the IMU, positioned below the line connecting the two superior posterior iliac spines, the metabolic-motor holter, placed at the level of the triceps of the arm and the ergospirometer, which was worn at the level of the back thanks to a harness; all three devices were activated/deactivated on-board and the acquired data was stored on internal memories, the download of which on the PC was made from time to time for each subject immediately after the end of the test. To assess the predictive significance of the regression models obtained, statistical tests and the R2 goodness-of-fit measurement (called determination coefficient) were used, while to assess the reliability of the metabolic-motor holter with respect to the ergospirometer were used a graphical comparison and calculation of percentage differences and a correlation index. Concerning the IMU from the results it emerged that the acceleration variable IAAtot (sum of the integrals of the accelerations along the three axes) is the most explanatory predictor with respect to AEE (measured in kJ/kg), moreover if we combine with it predictors representative of the weight and average cadence of the subjects, the predictive capacity of the model towards AEE (measured now in kJ/min) grows further (R2 = 0.90). The metabolic-motor holter instead turned out to be an invalid device for the estimation of AEE: the percentage difference expressed in average and standard deviation was found to be equal to 22.80 ± 11.34 % and the Pearson correlation coefficient gave a value equal to 0.307. However, due to the low number of subjects considered, the results obtained are not very robust and for this reason they can only be considered as indicative. Surely it would have been possible to obtain better results by increasing the number of subjects of the analysis, so as to be able to discard any outliers (which in this work it was preferred not to eliminate, due to the already small number of samples). As far as the IMU is concerned, this would have allowed us to obtain a regression line that better describes the linearity of the experimental data but also that it is able to generalize with respect to new observations; as far as the metabolic-motor holter is concerned, this would have allowed us to understand if it tends to overestimate or underestimate the values measured by the ergospirometer, otherwise if a fluctuating trend of values emerged (as in this study). The present work therefore aims to be mainly of a methodological nature, namely it aims to provide an example of a rigorous and standard approach of validation and calibration of the devices used, so that in a future analysis with a higher number of samples, repeating this study, we obtain more accurate results that can be compared and exchanged without ambiguity between the various institutes and laboratories concerned.
In ambito clinico il calcolo del consumo energetico durante il cammino di un soggetto con alterazioni del movimento è molto importante perché permette di confrontare una locomozione normale da una patologica, oppure di quantificare la severità di un disordine motorio, come anche rende possibile la valutazione dell’efficacia di una terapia riabilitativa o di un intervento chirurgico. Attualmente la calorimetria indiretta è di gran lunga il metodo più comunemente usato per la stima del dispendio energetico durante un’attività fisica e si basa sulla conoscenza del consumo di ossigeno e sulla produzione di anidride carbonica, che hanno luogo quando i muscoli stanno lavorando per produrre energia durante il movimento. In particolare l’ergospirometro, un sistema metabolico che può essere sia mobile che stazionario, costituisce attualmente il dispositivo di riferimento per la stima del consumo energetico durante prove da sforzo (activity-related energy expenditure, o AEE) sia in termini di accuratezza che di praticità di utilizzo. Ma in alternativa possono essere utilizzati altri dispositivi indossabili, quali i sistemi inerziali o altri sensori metabolico-motori, molto meno costosi e maneggevoli, che necessitano però di essere validati, confrontando le loro misure con quelle fornite da un dispositivo di riferimento. A tal proposito in questo lavoro di tesi sono stati acquisiti, insieme alle misure di dispendio energetico fornite dall’ergospirometro, dati fisiologici e motori durante una prova di cammino, tramite un sistema di misurazione inerziale (noto anche come inertial measurement unit, o IMU) ed un holter metabolico-motorio, al fine di determinare se questi due sensori possano fungere da valida alternativa all’ergospirometro (strumento di riferimento) per la stima delle calorie spese in una camminata di 6 minuti. Nello specifico per quanto riguarda l’IMU si è voluto vedere se i suoi dati di accelerazione possono essere usati per predire il consumo energetico, quindi sono stati ricavati vari modelli di regressione lineare semplice, al fine di relazionare le sue accelerazioni (o meglio loro elaborazioni) con l’AEE fornito dall’ergospirometro; successivamente si è presa in considerazione la variabile di accelerazione più rappresentativa, insieme alla variabile peso e ad altre grandezze cinematiche (stimate sempre dall’IMU) relative alla prova, per elaborare modelli di regressione lineare multipla che spiegassero ancor meglio la variabile AEE. Per quanto riguarda invece l’holter metabolico-motorio, il quale fornisce in uscita direttamente il consumo energetico, è stato ricavato un modello di correlazione tra i valori di AEE da esso forniti e quelli ricavati dall’ergospirometro, per determinare se i due dispositivi forniscono una misura confrontabile del consumo energetico durante la prova di cammino. In questo studio sono stati reclutati 6 soggetti sani (2 donne e 4 uomini), senza particolari criteri di inclusione, con età = 30 ± 7.3 anni, peso = 68.9 ± 16.3 kg e altezza = 1.72 ± 0.09 m, i quali hanno ripetuto la prova due volte, in modo da avere 12 prove sperimentali da analizzare. Il test consisteva nel camminare il più velocemente possibile per 6 minuti lungo una traiettoria rettilinea di lunghezza pari a 25 metri, con l’obbiettivo di fare più metri possibile rispetto alle proprie capacità, gestendo autonomamente lo sforzo. Per le acquisizioni i soggetti sono stati equipaggiati con un IMU, posizionato al di sotto della linea che collega le due spine iliache posteriori superiori, con un holter metabolico-motorio, posto a livello del tricipite del braccio e con un ergospirometro, il quale è stato indossato a livello del dorso grazie ad una imbragatura; tutti e tre i dispositivi sono stati attivati/disattivati on-board e i dati acquisiti erano immagazzinati su memorie interne, il cui download sul PC era fatto di volta in volta per ciascun soggetto subito dopo il termine della prova. Per valutare la significatività predittiva dei modelli di regressione ottenuti sono stati utilizzati test statistici e la misura di bontà di adattamento R2 (detto coefficiente di determinazione), mentre per valutare l’attendibilità dell’holter metabolico-motorio rispetto all’ergospirometro si è ricorso ad un confronto grafico e al calcolo di differenze percentuali e di un indice di correlazione. Per quanto riguarda l’IMU dai risultati è emerso che la variabile di accelerazione IAAtot (somma degli integrali delle accelerazioni lungo i tre assi) è il predittore più esplicativo nei confronti di AEE (misurata in kJ/kg), per di più se ad essa uniamo i predittori rappresentativi del peso e della cadenza media dei soggetti, la capacità predittiva del modello nei confronti di AEE (misurata ora in kJ/min) cresce ulteriormente (R2 = 0.90). L’holter metabolico-motorio invece è risultato essere un dispositivo non valido per la stima di AEE: la differenza percentuale espressa in media e deviazione standard è risultata essere uguale a 22.80 ± 11.34 % e il coefficiente di correlazione di Pearson ha dato un valore pari a 0.307. Comunque a causa della bassa numerosità della popolazione di soggetti presa in esame, i risultati ottenuti sono poco robusti e per tale motivo possono solamente essere considerati indicativi. Sicuramente sarebbe stato possibile ottenere risultati migliori aumentando il numero di soggetti oggetto di analisi, così da poter scartare eventuali outliers (che in questo lavoro si è preferito non eliminare, per il già esiguo numero di campioni). Per quanto riguarda l’IMU ciò avrebbe consentito di ottenere una retta di regressione che descrivesse meglio la linearità dei dati sperimentali ma anche che fosse in grado generalizzare nei confronti di nuove osservazioni; per quanto riguarda l’holter metabolico-motorio invece, ciò avrebbe permesso di capire se esso tendeva a sovrastimare o sottostimare i valori misurati dall’ergospirometro, oppure se emergeva un andamento altalenante dei valori (come in questo studio). Il presente lavoro vuole quindi essere principalmente di carattere metodologico, cioè vuole fornire un esempio di approccio rigoroso e standard di validazione e di calibrazione dei dispositivi utilizzati, così che in un’analisi futura a più alta numerosità o svolta con altre tipologie di sensori indossabili e di test motori, ripetendo questo studio, si abbiano risultati più accurati che possano essere confrontati e scambiati senza ambiguità tra i vari istituti e laboratori interessati.
Validazione di due sistemi indossabili per la stima del dispendio energetico durante il cammino
ROTA SCALABRINI, MICHAEL
2018/2019
Abstract
In the clinical context the calculation of energy consumption during walking of a subject with movement abnormalities is very important because it allows to compare a normal locomotion to a pathological one, or to quantify the severity of a motor disorder, as also makes it possible to evaluate the effectiveness of a rehabilitation therapy or a surgical intervention. Currently indirect calorimetry is by far the most commonly used method for estimating energy expenditure during physical activity and is based on knowledge of oxygen consumption and carbon dioxide production, which occur when the muscles are working to produce energy during movement. In particular, the ergospirometer, a metabolic system that can be both mobile and stationary, currently constitutes the reference device for estimating energy consumption during activity-related energy expenditure, or AEE, both in terms of accuracy and practicality of utilization. But altermately other wearable devices can be used, such as inertial systems or other metabolic-motor sensors, much less expensive and manageable, which however need to be validated, comparing their measurements with those provided by a reference device. In this regard, in this thesis work, along with the energy expenditure measures provided by the ergospirometer, physiological and motor data were acquired during a walking test, through an inertial measurement unit, or IMU, and a metabolic-motor holter, in order to determine if these two sensors can act as a valid alternative to the ergospirometer (gold standard) for the estimation of the calories expended in a walk of 6 minutes. Specifically, as far as the IMU is concerned, we wanted to see if its acceleration data can be used to predict energy consumption, so various simple linear regression models have been obtained, in order to relate its accelerations (or rather their elaborations ) with the AEE supplied by the ergospirometer; subsequently the more representative acceleration variable was taken into consideration, together with the weight variable and other kinematic quantities (always estimated by the IMU) relating to the test, in order to develop multiple linear regression models that would explain the AEE variable even better. With regards to the metabolic-motor holter, which directly outputs the energy consumption, a correlation model was obtained between the AEE values supplied by it and those obtained from the ergospirometer, to determine whether the two devices provide a comparable measurement of the energy consumption during the walking test. In this study 6 healthy subjects (2 women and 4 men) were recruited, without particular inclusion criteria, with age = 30 ± 7.3 years, weight = 68.9 ± 16.3 kg and height = 1.72 ± 0.09 m, which repeated the trial twice, in order to have 12 experimental tests to analyze. The test consisted of walking as fast as possible for 6 minutes along a straight path of a length of 25 meters, with the aim of making as many meters as possible with respect to one's abilities, managing the effort independently. For the acquisitions the subjects were equipped with the IMU, positioned below the line connecting the two superior posterior iliac spines, the metabolic-motor holter, placed at the level of the triceps of the arm and the ergospirometer, which was worn at the level of the back thanks to a harness; all three devices were activated/deactivated on-board and the acquired data was stored on internal memories, the download of which on the PC was made from time to time for each subject immediately after the end of the test. To assess the predictive significance of the regression models obtained, statistical tests and the R2 goodness-of-fit measurement (called determination coefficient) were used, while to assess the reliability of the metabolic-motor holter with respect to the ergospirometer were used a graphical comparison and calculation of percentage differences and a correlation index. Concerning the IMU from the results it emerged that the acceleration variable IAAtot (sum of the integrals of the accelerations along the three axes) is the most explanatory predictor with respect to AEE (measured in kJ/kg), moreover if we combine with it predictors representative of the weight and average cadence of the subjects, the predictive capacity of the model towards AEE (measured now in kJ/min) grows further (R2 = 0.90). The metabolic-motor holter instead turned out to be an invalid device for the estimation of AEE: the percentage difference expressed in average and standard deviation was found to be equal to 22.80 ± 11.34 % and the Pearson correlation coefficient gave a value equal to 0.307. However, due to the low number of subjects considered, the results obtained are not very robust and for this reason they can only be considered as indicative. Surely it would have been possible to obtain better results by increasing the number of subjects of the analysis, so as to be able to discard any outliers (which in this work it was preferred not to eliminate, due to the already small number of samples). As far as the IMU is concerned, this would have allowed us to obtain a regression line that better describes the linearity of the experimental data but also that it is able to generalize with respect to new observations; as far as the metabolic-motor holter is concerned, this would have allowed us to understand if it tends to overestimate or underestimate the values measured by the ergospirometer, otherwise if a fluctuating trend of values emerged (as in this study). The present work therefore aims to be mainly of a methodological nature, namely it aims to provide an example of a rigorous and standard approach of validation and calibration of the devices used, so that in a future analysis with a higher number of samples, repeating this study, we obtain more accurate results that can be compared and exchanged without ambiguity between the various institutes and laboratories concerned.| File | Dimensione | Formato | |
|---|---|---|---|
|
tesi finale.pdf
accessibile in internet per tutti
Descrizione: elaborato tesi
Dimensione
1.23 MB
Formato
Adobe PDF
|
1.23 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/10589/146180