Currently, the main tool used for acne diagnosis and treatment is human eye. The differential analysis of the pathology is based on the patient’s clinical history and on his or her physical exams. When dealing with physical exams, doctors focus their attention on skin topographic information, on lesions’ typology, distribution and seriousness and on the possible presence on abnormalities in blood pressure or in skin pigmentation. The seriousness of the pathology is estimated on the basis of a manual calculation of every lesion in those body region with most evident symptoms. Then, comparing calculation results with landmark values, the doctor is able to establish the pathology extension that will represent a leading value in the choice of the treatment to perform. This is non-standardized diagnostic approach and is therefore subject to intra and inter-subjective mistakes, thus resulting inaccurate. There are some more accurate methods based on the biopsy of some subcutaneous tissue’s samples or on photographic analysis. However, these methods result invasive and expensive. Aim of this thesis is to realize a biomedical device able to support and automatize acne diagnosis using low-cost hardware components and non-invasive methods for photographic analysis. A secondary aim is represented by the attempt to realize a tool that, thanks to its reliability and accuracy in the pathology analysis, may be used as a standard device in the medical environment. In order to produce the device, the project has been developed through a two-folded and parallel method. On the one hand there’s the characterization of those optical components necessary to capture superficial and sub-superficial images of the skin; on the other hand, there’s the settlement of the algorithms for image processing aimed to extract from the captured images clinical information important for the examined pathology. Starting from a detailed analysis of the state of art about the optical methods used in skin imaging diagnostic and on the state of art of indicators that characterize acne presence, in the first part of the project, we explain how it has been possible to highlight those methods which resulted more suitable to gain clinical information necessary in the pathology diagnosis. In particular, when the doctors classify lesions, they also evaluate the presence of indicators such as inflammation, pigmentation, pus, pain or lesions depth and size. These indicators may be measured through the evaluation of etiologic and biologic acne parameters. From an optical point of view, these parameters behave as either chromophores or fluorophores. This means that they react to a certain light source either absorbing or re-emanating the light at some specific wavelengths. The second part of the thesis illustrates in details the process used to choose the hardware components as well as the process used to implement the image processing algorithms. The choice of the hardware components has been led by the attempt to find the best trade-off both between measure accuracy and cost-saving opportunities and between medical needs and device encumbrances. In particular, the selected optical components allow the implementation of three different methodologies for image acquisition and skin samples analysis. It is known from the state of art, that the cited methods allow to gain those subcutaneous information necessary to achieve the proposed goals. The second pathway that has been followed in the project development concerns the definition and implementation of algorithms aimed to extract information from the acquired images and to use those images as a tool to support the acne diagnosis. Algorithms herein described refer both to image acquisition, pre-processing and segmentation and to the methods used to classify those images. Aside from describing in detail how these three different methods to scan the skin have been simultaneously implemented to improve the diagnosis accuracy, a general description of the entire infrastructure necessary to the system to operate is offered. This infrastructure ( in the thesis named ‘ecosystem’) is composed by physical and virtual, cloud-based elements whose architecture aimed to increase the device overall efficiency. The third part of the thesis introduces the results gained through the project development. Assessments on device functioning will be performed during a clinical trial that will take place in January 2018. Until that moment it will not be possible to have the consistent data set necessary to validate the project, the team has chosen to use a database of images acquired in a prototypical version of the device either with standard or in fluorescence illumination. These images have been thus used as a pre-validation set for the employment of those low-cost hardware components for the acne diagnosis performed through images in fluorescence. In particular, images acquired under standard illumination have been used to gain feedbacks from results obtained through processing images under fluorescence illumination. On the one hand, those results concern the measurement of the skin fluorophores involved in the acne generating process that has been then used in the pre and post monitoring phases of the pathology; on the other hand, they concern the segmentation of the images used to calculate and classify lesions. According to the results obtained in this process, it has been noticed that the chosen technology perform interesting results even though it still lacks any kind of assistance from additional components that in future will characterize the final version of the device. Hardware specifications, optical methods and algorithms proposed in this thesis, stand as a starting point for a wider project aimed to realize the first low-cost medical device for an effective acne diagnosis.
Lo strumento ad oggi più utilizzato per la diagnosi e il trattamento dell’acne è l’occhio umano. La diagnosi differenziale dell’acne è basata sulla storia clinica del paziente e su esami fisici. Nell’esame fisico il dermatologo focalizza l’attenzione su informazioni topografiche della pelle, su tipologia, distribuzione e gravità delle lesioni e su possibili anormalità nella pressione sanguigna e nella pigmentazione della cute. La gravità della patologia è valutata dal medico tramite conteggio manuale di ogni tipo di lesione, nelle aree del corpo in cui i sintomi risultano più evidenti. Confrontando poi i risultati con dei valori di riferimento, il medico stabilisce il livello di estensione della patologia, che andrà a guidarne la scelta di trattamento. Questo approccio diagnostico non è standardizzato, è soggetto ad errori intra ed intersoggettivi ed è poco accurato. Esistono dei metodi più accurati basati su biopsia di campioni sottocutanei o su analisi fotografiche che si presentano però invasivi o costosi. Obiettivo di questo lavoro di tesi è la realizzazione di un dispositivo biomedicale che automatizzi e sia di supporto alla diagnosi dermatologica dell’acne, tramite l’utilizzo di componenti hardware a basso costo e metodi non invasivi di analisi fotografica. L’obiettivo secondario è riuscire a realizzare uno strumento che, grazie all’affidabilità e all’accuratezza ottenibili sulla detezione delle lesioni della cute, possa proporsi come standard da adottare in ambito clinico. Il progetto è stato svolto seguendo parallelamente due filoni principali per lo sviluppo del dispositivo: la caratterizzazione delle componenti ottiche, necessarie per l’acquisizione di immagini superficiali e sub-superficiali della pelle, e la definizione degli algoritmi di image processing, per l’estrazione, da tali immagini, delle informazioni cliniche significative per la patologia in esame. Nella prima parte del testo si spiega come, partendo da una profonda analisi della letteratura sui metodi ottici ad oggi utilizzati per l’imaging diagnostico della pelle e sugli indicatori caratterizzanti la presenza di acne, è stato possibile evidenziare i metodi fotografici adatti ad acquisire le informazioni cliniche necessarie per la diagnosi della patologia. In particolare per la classificazione delle lesioni dell’acne il dermatologo valuta la presenza di alcuni indicatori come infiammazioni, pigmentazioni, pus, dolore e profondità delle lesioni. Tali indicatori sono misurabili attraverso la valutazione dei parametri eziologici e biologici dell’acne. Questi parametri dal punto di vista ottico si comportano da cromofori o fluorofori: rispondono cioè ad una illuminazione assorbendo o riemettendo la luce a specifiche lunghezze d’onda. La seconda parte del testo spiega nel dettaglio il processo di scelta delle componenti hardware e l’implementazione degli algoritmi di image processing. La scelta delle componenti è stata effettuata cercando il miglior trade-off tra accuratezza di misura e costi di realizzazione e tra esigenze mediche e ingombri dello strumento. In particolare le componenti ottiche selezionate permettono l’implementazione di tre diverse metodologie di acquisizione e analisi dei campioni di tessuto cutaneo: polarizzazione, spettroscopia multispettrale e fluorescenza. È noto infatti dalla letteratura che questi metodi permettono l’acquisizione di informazioni sottocutanee necessarie per il raggiungimento dell’obiettivo preposto. Il secondo filone principale di sviluppo del progetto riguarda la definizione e l’implementazione degli algoritmi per l’estrazione delle informazioni dalle immagini acquisite e il loro utilizzo per la diagnosi dell’acne. Gli algoritmi descritti in questo lavoro di tesi riguardano in particolare l’acquisizione, il pre-processing, la segmentazione delle immagini e i metodi di classificazione delle lesioni. Oltre alla descrizione dettagliata di come è implementato l’utilizzo simultaneo di tre diversi metodi di scansione della pelle, al fine di aumentare accuratezza nella diagnosi, è riportata anche una descrizione generale di tutta l’infrastruttura necessaria al funzionamento del sistema. Tale infrastruttura, nel testo indicata come “ecosistema”, si compone di elementi fisici e virtuali (basati su cloud) la cui architettura è finalizzata all’aumento di efficienza del dispositivo. La terza parte del testo introduce i risultati ottenuti durante lo svolgimento del progetto. I test sul funzionamento del dispositivo verranno effettuati durante un trial clinico che avrà inizio nel Gennaio 2018. Non avendo la possibilità fino a tale data di disporre di un set di dati consistente per la validazione del progetto, è stato scelto di utilizzare una banca dati di immagini acquisite con una versione prototipale del dispositivo in condizioni di illuminazione standard (luce bianca) e in fluorescenza. Queste immagini sono state quindi utilizzate come set di pre-validazione per l’utilizzo di componenti hardware a basso costo per la diagnosi dell’acne attraverso immagini in fluorescenza. In particolare le immagini acquisite in condizioni di illuminazione standard sono state utilizzate per avere un riscontro sui risultati ottenuti processando le immagini in fluorescenza. Tali risultati riguardano da una parte la misura della concentrazione dei fluorofori principali della pelle coinvolti nel processo di generazione dell’acne, utilizzata per il monitoraggio pre- e post-clinico della patologia, e dall’altra la segmentazione delle immagini utilizzata per il conteggio e la classificazione delle lesioni. Dai risultati ottenuti è emerso che la tecnologia scelta permette prestazioni interessanti nonostante ancora priva dell’ausilio delle componenti aggiuntive che caratterizzeranno la versione finale del dispositivo. Le specifiche hardware, i metodi ottici e gli algoritmi proposti in questo lavoro di tesi presentano il punto di partenza di un progetto più ampio volto alla realizzazione del primo device medico low cost per una diagnosi efficace dell’acne.
Low cost medical device for image processing-based acne diagnosis
BURGO, MAURO
2015/2016
Abstract
Currently, the main tool used for acne diagnosis and treatment is human eye. The differential analysis of the pathology is based on the patient’s clinical history and on his or her physical exams. When dealing with physical exams, doctors focus their attention on skin topographic information, on lesions’ typology, distribution and seriousness and on the possible presence on abnormalities in blood pressure or in skin pigmentation. The seriousness of the pathology is estimated on the basis of a manual calculation of every lesion in those body region with most evident symptoms. Then, comparing calculation results with landmark values, the doctor is able to establish the pathology extension that will represent a leading value in the choice of the treatment to perform. This is non-standardized diagnostic approach and is therefore subject to intra and inter-subjective mistakes, thus resulting inaccurate. There are some more accurate methods based on the biopsy of some subcutaneous tissue’s samples or on photographic analysis. However, these methods result invasive and expensive. Aim of this thesis is to realize a biomedical device able to support and automatize acne diagnosis using low-cost hardware components and non-invasive methods for photographic analysis. A secondary aim is represented by the attempt to realize a tool that, thanks to its reliability and accuracy in the pathology analysis, may be used as a standard device in the medical environment. In order to produce the device, the project has been developed through a two-folded and parallel method. On the one hand there’s the characterization of those optical components necessary to capture superficial and sub-superficial images of the skin; on the other hand, there’s the settlement of the algorithms for image processing aimed to extract from the captured images clinical information important for the examined pathology. Starting from a detailed analysis of the state of art about the optical methods used in skin imaging diagnostic and on the state of art of indicators that characterize acne presence, in the first part of the project, we explain how it has been possible to highlight those methods which resulted more suitable to gain clinical information necessary in the pathology diagnosis. In particular, when the doctors classify lesions, they also evaluate the presence of indicators such as inflammation, pigmentation, pus, pain or lesions depth and size. These indicators may be measured through the evaluation of etiologic and biologic acne parameters. From an optical point of view, these parameters behave as either chromophores or fluorophores. This means that they react to a certain light source either absorbing or re-emanating the light at some specific wavelengths. The second part of the thesis illustrates in details the process used to choose the hardware components as well as the process used to implement the image processing algorithms. The choice of the hardware components has been led by the attempt to find the best trade-off both between measure accuracy and cost-saving opportunities and between medical needs and device encumbrances. In particular, the selected optical components allow the implementation of three different methodologies for image acquisition and skin samples analysis. It is known from the state of art, that the cited methods allow to gain those subcutaneous information necessary to achieve the proposed goals. The second pathway that has been followed in the project development concerns the definition and implementation of algorithms aimed to extract information from the acquired images and to use those images as a tool to support the acne diagnosis. Algorithms herein described refer both to image acquisition, pre-processing and segmentation and to the methods used to classify those images. Aside from describing in detail how these three different methods to scan the skin have been simultaneously implemented to improve the diagnosis accuracy, a general description of the entire infrastructure necessary to the system to operate is offered. This infrastructure ( in the thesis named ‘ecosystem’) is composed by physical and virtual, cloud-based elements whose architecture aimed to increase the device overall efficiency. The third part of the thesis introduces the results gained through the project development. Assessments on device functioning will be performed during a clinical trial that will take place in January 2018. Until that moment it will not be possible to have the consistent data set necessary to validate the project, the team has chosen to use a database of images acquired in a prototypical version of the device either with standard or in fluorescence illumination. These images have been thus used as a pre-validation set for the employment of those low-cost hardware components for the acne diagnosis performed through images in fluorescence. In particular, images acquired under standard illumination have been used to gain feedbacks from results obtained through processing images under fluorescence illumination. On the one hand, those results concern the measurement of the skin fluorophores involved in the acne generating process that has been then used in the pre and post monitoring phases of the pathology; on the other hand, they concern the segmentation of the images used to calculate and classify lesions. According to the results obtained in this process, it has been noticed that the chosen technology perform interesting results even though it still lacks any kind of assistance from additional components that in future will characterize the final version of the device. Hardware specifications, optical methods and algorithms proposed in this thesis, stand as a starting point for a wider project aimed to realize the first low-cost medical device for an effective acne diagnosis.File | Dimensione | Formato | |
---|---|---|---|
Tesi.pdf
non accessibile
Descrizione: thesis text
Dimensione
2.51 MB
Formato
Adobe PDF
|
2.51 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/10589/133269