Respiratory failure is a severe condition affecting the respiratory system of a person. Different techniques have been introduced in the clinical practice to identify the cause of failure and guide the treatment. Most of them are time consuming, invasive or expose the patient to potential health risks. Novel, non-invasive, fast diagnostic methods capable to identify the causes of respiratory failure are continuously researched. In this context, this thesis, carried out in collaboration with ASST Papa Giovanni XXIII hospital in Bergamo (BG) and the TechRes Lab of the Department of Electronic, Information and Bioengineering (DEIB, Politecnico of Milano), proposes the use of an electronic nose (eNose) for the analysis of exhaled breath samples aimed at the identification of respiratory failure. Given the widespread diffusion of SARS-CoV-2, causing several cases of pneumonia and respiratory failure in the infected subjects, this disease was selected as case study. Population involved in the study comprised 53 subjects divided in three classes: 23 symptomatic SARS-CoV-2, 8 asymptomatic SARS-CoV-2 and 22 healthy controls. The experimental activity carried out within this project focused on two aspects: i) the definition of a specific protocol for exhaled breath sampling from respiratory failure patients and ii) the implementation of a specific eNose data processing procedure for signal analysis. The results achieved proved the potentialities of the eNose in the discrimination of symptomatic SARS-CoV-2 patients from healthy subjects and confirmed the efficacy of the experimental protocol developed in safely sampling respiratory failure patients. Indeed, a good classification performance was achieved by SVM linear and kNN based models, providing a clinical accuracy of about 82%. Furthermore, the analysis of the asymptomatic SARS-CoV-2 suggested that the proposed method can also be used to determine the severity of the patient status. Unfortunately, due to poor dimension of the population, this evidence represents only a preliminary result.
L’insufficienza respiratoria è una condizione critica riguardante il sistema respiratorio di una persona. Varie tecniche sono state introdotte nella pratica clinica per identificare le cause della malattia e guidare il suo trattamento. Molte di esse sono invasive, espongono il paziente a potenziali rischi per la salute oppure richiedono molto tempo. Metodi diagnostici innovativi, rapidi e non invasivi, sono continuamente ricercati per identificare le cause dell’insufficienza respiratoria. In questo contesto, questa tesi, svolta in collaborazione con l’ospedale ASST Papa Giovanni XXIII di Bergamo e con il Dipartimento di Elettronica, Informazione e Bioingegneria (DEIB, Politecnico di Milano), propone l’uso di un naso elettronico (eNose) per l’analisi di campioni di esalato con l’obiettivo di identificare l’insufficienza respiratoria. Data la diffusione su larga scala del virus SARS-CoV-2, causa in molti casi di polmoniti e insufficienza respiratoria in soggetti infetti, questa malattia è stata selezionata come caso studio. La popolazione coinvolta nello studio comprende 53 soggetti divisi in 3 classi: 23 SARS-CoV-2 sintomatici, 8 SARS-CoV-2 asintomatici e 22 controlli sani. L’attività sperimentale svolta in questo progetto si concentra su due aspetti: i) la definizione di un protocollo specifico per il campionamento dell’esalato da pazienti con insufficienza respiratoria e ii) l’implementazione di una specifica procedura di elaborazione dati per analizzare i segnali del naso elettronico. I risultati ottenuti dimostrano la potenzialità dell’eNose nella discriminazione dei soggetti SARS-CoV-2 sintomatici rispetto soggetti sani e conferma l’efficacia del protocollo sperimentale sviluppato per campionare in sicurezza i pazienti soggetti da insufficienza respiratoria. Infatti, una buona performance di classificazione è stata ottenuta tramite modelli SVM lineare e kNN, raggiungendo un’accuratezza di circa 82%. Inoltre, l’analisi degli asintomatici, suggerisce la possibilità di usare il metodo per determinare la severità della malattia nel paziente. Sfortunatamente, per via della bassa numerosità della popolazione, questa ipotesi può essere considerata solo come un risultato preliminare.
Novel approach for SARS-CoV-2 detection in respiratory failure patients based on exhaled breath analysis with an electronic nose
RATTI, CHRISTIAN
2020/2021
Abstract
Respiratory failure is a severe condition affecting the respiratory system of a person. Different techniques have been introduced in the clinical practice to identify the cause of failure and guide the treatment. Most of them are time consuming, invasive or expose the patient to potential health risks. Novel, non-invasive, fast diagnostic methods capable to identify the causes of respiratory failure are continuously researched. In this context, this thesis, carried out in collaboration with ASST Papa Giovanni XXIII hospital in Bergamo (BG) and the TechRes Lab of the Department of Electronic, Information and Bioengineering (DEIB, Politecnico of Milano), proposes the use of an electronic nose (eNose) for the analysis of exhaled breath samples aimed at the identification of respiratory failure. Given the widespread diffusion of SARS-CoV-2, causing several cases of pneumonia and respiratory failure in the infected subjects, this disease was selected as case study. Population involved in the study comprised 53 subjects divided in three classes: 23 symptomatic SARS-CoV-2, 8 asymptomatic SARS-CoV-2 and 22 healthy controls. The experimental activity carried out within this project focused on two aspects: i) the definition of a specific protocol for exhaled breath sampling from respiratory failure patients and ii) the implementation of a specific eNose data processing procedure for signal analysis. The results achieved proved the potentialities of the eNose in the discrimination of symptomatic SARS-CoV-2 patients from healthy subjects and confirmed the efficacy of the experimental protocol developed in safely sampling respiratory failure patients. Indeed, a good classification performance was achieved by SVM linear and kNN based models, providing a clinical accuracy of about 82%. Furthermore, the analysis of the asymptomatic SARS-CoV-2 suggested that the proposed method can also be used to determine the severity of the patient status. Unfortunately, due to poor dimension of the population, this evidence represents only a preliminary result.| File | Dimensione | Formato | |
|---|---|---|---|
|
2021_12_Ratti.pdf
non accessibile
Dimensione
4.9 MB
Formato
Adobe PDF
|
4.9 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/10589/183213