Prostate cancer (KP) is the fifth most common cancer worldwide and its incidence is expected to considerably increase up to 2040. Current diagnostic protocols, based on prostate specific antigen (PSA) and prostate biopsy, are poorly accurate with an overall accuracy of about 58%, and have a low specificity (i.e. about 30%), which results in the over-diagnosis and over-treatment of patients. There is, thus, an urgent need for innovative and more accurate tools for early KP detection. Many researches proved the existence of a correlation between the alteration of urine and the KP presence, thereby suggesting to investigate urine as source of useful information for diagnostic purposes. The most promising results published up to now were obtained relying on trained dogs, who achieved a diagnostic accuracy above 97% in discriminating urine samples from controls and KP patients. However, trained dogs are not suitable for the development of a large-scale diagnostic tool, due for instance to dogs’ training costs and lack of compliance with hospital protocols. This PhD project, carried out in collaboration with the Humanitas Mater Domini Hospital in Castellanza (VA), aimed to transfer promising results achieved by Dott. Taverna with trained dogs to an instrumental method. The project focused on the development of an electronic nose for the analysis of urine odour aimed at the identification of a specific olfactory fingerprint of KP samples. 534 subjects (205 controls and 329 KP patients) were involved in the study to define the experimental protocol, train the eNose and validate the developed predictive model. The eNose proved capable to distinguish urine samples from KP patients from controls with a diagnostic accuracy above 80%, which is considerably higher than the one achieved by the current diagnostic procedure. A unique result achieved within this PhD project concerned the eNose capability to stage KP by differentiating KP patients suffering from high-aggressive cancers from subjects affected by low-aggressive tumours. This result, which – to the best of our knowledge – has never been reported before by any other diagnostic method, is extremely important, since the information about cancer aggressiveness is decisive for the definition of the prognostic pathway for patients. Indeed, according to current guidelines, low-aggressive KP patients undergo active surveillance, while in case of high-aggressive tumours, patients immediately undergo radiotherapy treatments or radical prostatectomy. One of the most innovative aspects of this research concerns the study and the development of specific methods for sensors’ drift correction, which is one of the most critical limiting factors to the scaling up of eNoses from research objects to effective diagnostic devices for large-scale use. The developed model proved capable to compensate the drift of a 1-year old sensor array, allowing achieving a classification performance comparable to the one achieved by means of new sensors not subjected to drift. Finally, the method was validated through the execution of double-blind tests, thereby confirming a diagnostic accuracy above 80% and eNose ability to stage the KP. It is worthy to underline that, despite the undoubtable importance of result validation by means of blind tests, it is the first time that the execution of such type of tests is reported in the scientific literature in the field of cancer diagnosis by means of eNoses. Results achieved within this PhD project showed the opportunity of developing a non-invasive, reliable, and cheap diagnostic tool for the early KP detection based on the analysis of urine odour, and led to the filing of a European patent request (EP 19160856.1) in March 2019, which has been extended to an International patent request (WO 2020/178284 A1) in March 2020.
Il cancro prostatico (KP) è il più diffuso tra gli uomini e la seconda causa di morte legata al cancro. La bassa specificità (30%) e l’elevato tasso di falsi positivi dell’attuale procedura diagnostica comportano un eccessivo trattamento dei pazienti e incentivano l’interesse nello sviluppo di strumenti diagnostici alternativi, non invasivi e più accurati. È ormai provata in letteratura l’esistenza di una correlazione tra l’alterazione delle urine e la presenza del KP, che suggerisce l’analisi delle urine come fonte di informazioni utili ai fini diagnostici. I risultati più promettenti finora pubblicati sono stati ottenuti affidandosi all’olfatto di cani addestrati, che hanno raggiunto un'accuratezza diagnostica superiore al 97% nel discriminare i campioni di urina da controlli e pazienti KP. Tuttavia, i cani addestrati non sono adatti per lo sviluppo di uno strumento diagnostico su larga scala, a causa, ad esempio, dei costi di addestramento dei cani e della mancata osservanza dei protocolli ospedalieri. Questo progetto di dottorato, svolto in collaborazione con l'Ospedale Humanitas Mater Domini di Castellanza (VA), si è proposto di trasferire i promettenti risultati raggiunti dal Dott. Taverna con cani addestrati ad un metodo strumentale. Il progetto si è concentrato sullo sviluppo di un naso elettronico (eNose) per l'analisi dell'odore delle urine. 534 soggetti (205 controlli e 329 pazienti KP) sono stati coinvolti nello studio per definire il protocollo sperimentale, addestrare l'eNose e validare il modello predittivo sviluppato. L'eNose si è dimostrato in grado di distinguere i campioni di urina di pazienti KP da controlli con un'accuratezza diagnostica superiore all'80%, che è notevolmente superiore a quella raggiunta dall'attuale procedura diagnostica. Un risultato unico raggiunto nell'ambito di questo progetto di dottorato ha riguardato la capacità del eNose di stadiare il KP, differenziando i pazienti affetti da KP affetti da tumori ad alta aggressività da soggetti affetti da tumori a bassa aggressività con un’accuratezza pari al 70%. È importante sottolineare che le informazioni sull'aggressività del cancro sono decisive per la definizione del percorso prognostico per i pazienti. Infatti, secondo le attuali linee guida, i pazienti con KP a bassa aggressività sono sottoposti a sorveglianza attiva, mentre in caso di tumori ad alta aggressività, i pazienti si sottopongono immediatamente a trattamenti radioterapici o prostatectomia radicale. Uno degli aspetti più innovativi di questa ricerca riguarda lo studio e lo sviluppo di metodi specifici per la correzione della deriva dei sensori, che ostacola lo scaling up dei eNose da oggetti di ricerca a dispositivi diagnostici efficaci su larga scala. Il modello sviluppato si è dimostrato in grado di compensare la deriva di un array di sensori di 1 anno, consentendo di ottenere una prestazione di classificazione paragonabile a quella ottenuta mediante nuovi sensori non soggetti a deriva. Infine, il metodo è stato validato attraverso l'esecuzione di test in doppio cieco, confermando un'accuratezza diagnostica superiore all'80% e la capacità del eNose di stadiare il KP. Nonostante l'indubbia importanza della validazione dei risultati mediante prove in cieco, è la prima volta che l'esecuzione di tali prove viene riportata nella letteratura scientifica nel campo della diagnosi dei tumori mediante eNoses. I risultati ottenuti hanno dimostrato l'opportunità di sviluppare uno strumento diagnostico non invasivo, affidabile ed economico per la rilevazione precoce di KP basato sull'analisi dell'odore delle urine, e hanno portato al deposito di una richiesta di brevetto europeo (EP 19160856.1) a marzo 2019, che è stata estesa a una richiesta di brevetto internazionale (WO 2020/178284 A1) a marzo 2020.
A novel approach for the non-invasive detection of prostate cancer based on urine odour analysis by an electronic nose
Carmen, Bax
2021
Abstract
Prostate cancer (KP) is the fifth most common cancer worldwide and its incidence is expected to considerably increase up to 2040. Current diagnostic protocols, based on prostate specific antigen (PSA) and prostate biopsy, are poorly accurate with an overall accuracy of about 58%, and have a low specificity (i.e. about 30%), which results in the over-diagnosis and over-treatment of patients. There is, thus, an urgent need for innovative and more accurate tools for early KP detection. Many researches proved the existence of a correlation between the alteration of urine and the KP presence, thereby suggesting to investigate urine as source of useful information for diagnostic purposes. The most promising results published up to now were obtained relying on trained dogs, who achieved a diagnostic accuracy above 97% in discriminating urine samples from controls and KP patients. However, trained dogs are not suitable for the development of a large-scale diagnostic tool, due for instance to dogs’ training costs and lack of compliance with hospital protocols. This PhD project, carried out in collaboration with the Humanitas Mater Domini Hospital in Castellanza (VA), aimed to transfer promising results achieved by Dott. Taverna with trained dogs to an instrumental method. The project focused on the development of an electronic nose for the analysis of urine odour aimed at the identification of a specific olfactory fingerprint of KP samples. 534 subjects (205 controls and 329 KP patients) were involved in the study to define the experimental protocol, train the eNose and validate the developed predictive model. The eNose proved capable to distinguish urine samples from KP patients from controls with a diagnostic accuracy above 80%, which is considerably higher than the one achieved by the current diagnostic procedure. A unique result achieved within this PhD project concerned the eNose capability to stage KP by differentiating KP patients suffering from high-aggressive cancers from subjects affected by low-aggressive tumours. This result, which – to the best of our knowledge – has never been reported before by any other diagnostic method, is extremely important, since the information about cancer aggressiveness is decisive for the definition of the prognostic pathway for patients. Indeed, according to current guidelines, low-aggressive KP patients undergo active surveillance, while in case of high-aggressive tumours, patients immediately undergo radiotherapy treatments or radical prostatectomy. One of the most innovative aspects of this research concerns the study and the development of specific methods for sensors’ drift correction, which is one of the most critical limiting factors to the scaling up of eNoses from research objects to effective diagnostic devices for large-scale use. The developed model proved capable to compensate the drift of a 1-year old sensor array, allowing achieving a classification performance comparable to the one achieved by means of new sensors not subjected to drift. Finally, the method was validated through the execution of double-blind tests, thereby confirming a diagnostic accuracy above 80% and eNose ability to stage the KP. It is worthy to underline that, despite the undoubtable importance of result validation by means of blind tests, it is the first time that the execution of such type of tests is reported in the scientific literature in the field of cancer diagnosis by means of eNoses. Results achieved within this PhD project showed the opportunity of developing a non-invasive, reliable, and cheap diagnostic tool for the early KP detection based on the analysis of urine odour, and led to the filing of a European patent request (EP 19160856.1) in March 2019, which has been extended to an International patent request (WO 2020/178284 A1) in March 2020.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi BAX - PhD_DEF.pdf
accesso solo da BNCF e BNCR
Licenza:
Tutti i diritti riservati
Dimensione
8.73 MB
Formato
Adobe PDF
|
8.73 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/204452
URN:NBN:IT:POLIMI-204452