Prostate cancer (PCa) is the second most commonly occurring cancer in men worldwide. The current screening procedure, involving PSA (prostate-specific antigen) test, is characterized by low specificity and high false positives rate. This thesis has been carried out within DIAG-NOSE project, proposing an innovative tool for the early diagnosis of PCa based on urine analysis by electronic nose, with the aim to develop a decisional model solid and stable over time. Besides proposing some improvements to the experimental protocol, this thesis faced with drift, one of the main problems associated with the use of gas sensors for long-term applications, which has up to now limited their diffusion at an industrial scale. To do this, this thesis focused on the optimization of the data processing procedure by implementing specific pre-treatments and classification models and also by applying a drift correction method, based on Orthogonal Signal Correction (OSC) algorithm, on urine headspace data collected in about a year. The defined decisional model has proved to effectively compensate for drift. Even with a 1-year-old sensor array, a diagnostic accuracy of about 80% has been achieved. This result is comparable to the one achieved by new sensors and it is significantly higher than the one of the current diagnostic protocol based on PSA serum level (i.e. 58% of accuracy). Moreover, the decisional model has proven to be capable of staging PCa according to tumour aggressiveness, which has not achieved by any other diagnostic tool. With the purpose of validating the results, in the last months of this thesis work, specific blind tests, i.e. the analysis of totally unknown and independent samples, were carried out. These tests confirmed the performance of the diagnosis model by achieving an accuracy of 78% and proving the potentialities of the staging model to stratify prostate cancer patients based on tumour aggressiveness.
Il cancro alla prostata (PCa) è il secondo tumore più comune negli uomini a livello mondiale. L’attuale procedura diagnostica, effettuata tramite l’analisi del PSA (antigene prostatico specifico), è caratterizzata da bassa specificità e da un alto tasso di falsi positivi. Questa tesi è stata svolta all’interno del progetto DIAG-NOSE, che propone uno strumento innovativo per la diagnosi del PCa basata sull’analisi delle urine tramite naso elettronico, con lo scopo di sviluppare un modello decisionale solido e stabile su lungo periodo. Oltre a proporre alcuni miglioramenti del protocollo sperimentale, in questa tesi è stato affrontato il problema del drift associato all’utilizzo di sensori in applicazioni su lungo periodo e che, ad oggi, limita la loro diffusione su scala industriale. Per fare ciò, questa tesi si focalizza sull’ottimizzazione della procedura di elaborazione dati, implementando specifici pretrattamenti e modelli di classificazione, ed applicando un metodo per la correzione del drift, basato sulla tecnica Orthogonal Signal Correction (OSC), sulle analisi condotte nell’arco di circa un anno. Il modello decisionale sviluppato si è dimostrato capace di correggere il drift. Infatti, nonostante siano stati utilizzati sensori con più di un anno di età, l’accuratezza del modello di diagnosi si è attestata attorno all’80% circa. Tale risultato è comparabile con quello ottenuto utilizzando sensori nuovi e significativamente migliore della corrente procedura diagnostica basata sul PSA (58%). Inoltre, il modello decisionale si è dimostrato capace di stadiare il PCa rispetto all’aggressività del tumore, risultato fino ad ora non raggiunto da nessun altro strumento. Con lo scopo di validare i risultati ottenuti, negli ultimi mesi di questa tesi, sono state effettuate specifiche prove in cieco, condotte analizzando 41 campioni incogniti. Questi test hanno confermato la capacità del dispositivo diagnostico innovativo di rilevare e stadiare il cancro prostatico in modo non invasivo con un’accuratezza del 78%.
Early prostate cancer detection through the analysis of urine odour by electronic nose : development of specific drift correction models for extended datasets in time
GASPARI, GIULIA;PRUDENZA, STEFANO
2019/2020
Abstract
Prostate cancer (PCa) is the second most commonly occurring cancer in men worldwide. The current screening procedure, involving PSA (prostate-specific antigen) test, is characterized by low specificity and high false positives rate. This thesis has been carried out within DIAG-NOSE project, proposing an innovative tool for the early diagnosis of PCa based on urine analysis by electronic nose, with the aim to develop a decisional model solid and stable over time. Besides proposing some improvements to the experimental protocol, this thesis faced with drift, one of the main problems associated with the use of gas sensors for long-term applications, which has up to now limited their diffusion at an industrial scale. To do this, this thesis focused on the optimization of the data processing procedure by implementing specific pre-treatments and classification models and also by applying a drift correction method, based on Orthogonal Signal Correction (OSC) algorithm, on urine headspace data collected in about a year. The defined decisional model has proved to effectively compensate for drift. Even with a 1-year-old sensor array, a diagnostic accuracy of about 80% has been achieved. This result is comparable to the one achieved by new sensors and it is significantly higher than the one of the current diagnostic protocol based on PSA serum level (i.e. 58% of accuracy). Moreover, the decisional model has proven to be capable of staging PCa according to tumour aggressiveness, which has not achieved by any other diagnostic tool. With the purpose of validating the results, in the last months of this thesis work, specific blind tests, i.e. the analysis of totally unknown and independent samples, were carried out. These tests confirmed the performance of the diagnosis model by achieving an accuracy of 78% and proving the potentialities of the staging model to stratify prostate cancer patients based on tumour aggressiveness.File | Dimensione | Formato | |
---|---|---|---|
2020_12_Gaspari_Prudenza.pdf
non accessibile
Descrizione: testo tesi
Dimensione
4.19 MB
Formato
Adobe PDF
|
4.19 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/10589/170755