Mechanical ventilation is widely used in clinical applications for the treatment of patients with acute respiratory failure, by providing a support to the patient’s ventilation activity, based on the control of flow and volume or pressure. The assessment of the patient’s respiratory mechanics is crucial for the setting of the ventilation and usually it is realized through the monitoring of the parameters of resistance and compliance, following the linear single-compartment model. However, a continuous on-line estimation is not yet available in the clinical setting. This work proposes an algorithm based on the Recursive Least Squares (RLS) estimation to obtain the on-line assessment of the mechanical parameters, implemented in the MATLAB® environment to carry out tests on traces acquired from mechanically ventilated pigs. These data, after some pre-processing operations, are inputted in the estimation algorithm, while, in parallel, they are used to compute the same parameters from the signal obtained through the flow interruption maneuver, which is regarded as reference for the evaluation of the results of the estimation algorithm. The results of this work are based on the comparison of the outputs of the two computation techniques, comprehensive also of the data simulation in Simulink®. They revealed that the RLS algorithm tends to underestimate the resistance values when the reference values are higher (e.g., pathological cases, high ventilatory flow), but both techniques provide the same trends when considering the progression between different conditions, e.g., from healthy to pathological, different ventilation settings. Instead, the estimated compliance reflects both the reference ranges and trends. In conclusion, the two techniques provide comparable performances when considering the total pressure signal and the passage from different conditions, even if they do not necessarily share the same role in the characterization of the respiratory system. Furthermore, the estimation algorithm offers new insights into the study of the time progression of the parameters since their waveform is also dependent on the conditions in which the traces are acquired.
La ventilazione meccanica è una delle tecniche principali utilizzate per il trattamento di pazienti affetti da insufficienza respiratoria, fornendo loro un supporto nell’attività ventilatoria attraverso il controllo di flussi, volumi e pressione. La caratterizzazione meccanica del sistema respiratorio del paziente è fondamentale per la qualità del trattamento, sebbene ad oggi non sia ancora possibile ottenere un monitoraggio continuo nel tempo dei parametri principali, rappresentati da resistenza e compliance respiratorie. Questo progetto si propone di testare un algoritmo di stima basato sul metodo ricorsivo ai minimi quadrati (RLS) su un dataset composto da tracce acquisite da maiali ventilati meccanicamente in diverse condizioni. L’algoritmo implementato in questo progetto è in grado di processare le tracce di flusso, volume e pressione, per poi fornirle in input all’algoritmo di stima vero e proprio e, in parallelo, utilizzarle per calcolare gli stessi parametri, sfruttando le porzioni di segnale corrispondenti alla manovra di interruzione di flusso, tecnica di riferimento nel settore clinico. I risultati ottenuti dimostrano che l’algoritmo tende a sovrastimare il valore di resistenza rispetto al riferimento quando quest’ultimo ha valori maggiori, ad esempio nella condizione patologica o per alti flussi ventilatori. In ogni caso, l’andamento dei valori stimati risulta essere comparabile a quello dei valori di riferimento nelle varie fasi dello studio e, soprattutto, quando si considera il valore di pressione totale stimato. Per quanto riguarda la compliance, invece, la stima riesce a fornire valori che rispettano sia i range che l’andamento dei valori di riferimento. Si può quindi concludere che i parametri forniti dall’algoritmo di stima riescano a caratterizzare le componenti meccaniche del segnale pressorio, sebbene non siano perfettamente sovrapponibili ai valori di riferimento, come confermato anche dalla simulazione dei dati. Infine, la stima on-line potrebbe introdurre un nuovo tipo di analisi basata sullo studio delle forme d’onda dei parametri durante il ciclo respiratorio, siccome si sono dimostrate essere dipendenti dalle diverse condizioni considerate.
Estimation of the mechanical parameters of the respiratory system in mechanically ventilated pigs
MORELLI, MIRIAM
2022/2023
Abstract
Mechanical ventilation is widely used in clinical applications for the treatment of patients with acute respiratory failure, by providing a support to the patient’s ventilation activity, based on the control of flow and volume or pressure. The assessment of the patient’s respiratory mechanics is crucial for the setting of the ventilation and usually it is realized through the monitoring of the parameters of resistance and compliance, following the linear single-compartment model. However, a continuous on-line estimation is not yet available in the clinical setting. This work proposes an algorithm based on the Recursive Least Squares (RLS) estimation to obtain the on-line assessment of the mechanical parameters, implemented in the MATLAB® environment to carry out tests on traces acquired from mechanically ventilated pigs. These data, after some pre-processing operations, are inputted in the estimation algorithm, while, in parallel, they are used to compute the same parameters from the signal obtained through the flow interruption maneuver, which is regarded as reference for the evaluation of the results of the estimation algorithm. The results of this work are based on the comparison of the outputs of the two computation techniques, comprehensive also of the data simulation in Simulink®. They revealed that the RLS algorithm tends to underestimate the resistance values when the reference values are higher (e.g., pathological cases, high ventilatory flow), but both techniques provide the same trends when considering the progression between different conditions, e.g., from healthy to pathological, different ventilation settings. Instead, the estimated compliance reflects both the reference ranges and trends. In conclusion, the two techniques provide comparable performances when considering the total pressure signal and the passage from different conditions, even if they do not necessarily share the same role in the characterization of the respiratory system. Furthermore, the estimation algorithm offers new insights into the study of the time progression of the parameters since their waveform is also dependent on the conditions in which the traces are acquired.File | Dimensione | Formato | |
---|---|---|---|
2024_04_Morelli_Executive.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Executive Summary
Dimensione
829.41 kB
Formato
Adobe PDF
|
829.41 kB | Adobe PDF | Visualizza/Apri |
2024_04_Morelli_Tesi.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Testo Tesi
Dimensione
4.26 MB
Formato
Adobe PDF
|
4.26 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/10589/219438