Nowadays, the amount of heterogeneous biomedical data is increasing more and more thanks to novel sensing techniques and high-throughput technologies. In reference to biomedical image analysis, the advances in image acquisition modalities and high-throughput imaging experiments are creating new challenges. This huge information ensemble could overwhelm the analytic capabilities needed by physicians in their daily decision-making tasks as well as by biologists investigating complex biochemical systems. In particular, quantitative imaging methods convey scientifically and clinically relevant information in prediction, prognosis or treatment response assessment, by also considering radiomics approaches. Therefore, the computational analysis of medical and biological images plays a key role in radiology and laboratory applications. In this regard, frameworks based on advanced Machine Learning and Computational Intelligence can significantly improve traditional Image Processing and Pattern Recognition approaches. However, conventional Artificial Intelligence techniques must be tailored to address the unique challenges concerning biomedical imaging data. This thesis aims at proposing novel and advanced computer-assisted methods for biomedical image analysis, also as an instrument in the development of Clinical Decision Support Systems, by always keeping in mind the clinical feasibility of the developed solutions. The devised classical Image Processing algorithms, with particular interest to region-based and morphological approaches in biomedical image segmentation, are first described. Afterwards, Pattern Recognition techniques are introduced, applying unsupervised fuzzy clustering and graph-based models (i.e., Random Walker and Cellular Automata) to multispectral and multimodal medical imaging data processing. Taking into account Computational Intelligence, an evolutionary framework based on Genetic Algorithms for medical image enhancement and segmentation is presented. Moreover, multimodal image co-registration using Particle Swarm Optimization is discussed. Finally, Deep Neural Networks are investigated: (i) the generalization abilities of Convolutional Neural Networks in medical image segmentation for multi-institutional datasets are addressed by conceiving an architecture that integrates adaptive feature recalibration blocks, and (ii) the generation of realistic medical images based on Generative Adversarial Networks is applied to data augmentation purposes. In conclusion, the ultimate goal of these research studies is to gain clinically and biologically useful insights that can guide differential diagnosis and therapies, leading towards biomedical data integration for personalized medicine. As a matter of fact, the proposed computer-assisted bioimage analysis methods can be beneficial for the definition of imaging biomarkers, as well as for quantitative medicine and biology.
Oggigiorno, la mole di dati biomedicali eterogenei è in continua crescita grazie alle nuove tecniche di sensing e alle tecnologie ad high-throughput. Relativamente all'analisi di immagini biomedicali, i progressi relativi alle modalità di acquisizione di immagini agli esperimenti di imaging ad high-throughput stanno creando nuove sfide. Questo ingente complesso di informazioni può spesso sopraffare le capacità analitiche sia dei medici nei loro processi decisionali sia dei biologi nell'investigazione di sistemi biochimici complessi. In particolare, i metodi di imaging quantitativo forniscono informazioni scientificamente rilevanti per la predizione, la prognosi o la valutazione della risposta al trattamento, prendendo in considerazione anche approcci di radiomica. Pertanto, l'analisi computazionale di immagini medicali e biologiche svolge un ruolo chiave in applicazioni di radiologia e di laboratorio. A tal proposito, framework basati su tecniche avanzate di Machine Learning e Computational Intelligence permettono di migliorare significativamente i tradizionali approcci tradizionali di Image Processing e Pattern Recognition. Tuttavia, le tecniche convenzionali di Intelligenza Artificiale devono essere propriamente adattate alle sfide uniche imposte dai dati di imaging biomedicale. La presente tesi mira a proporre innovativi metodi assistiti da calcolatore per l'analisi di immagini biomedicali, da utilizzare anche come strumento per lo sviluppo di Sistemi di Supporto alle Decisioni Cliniche, tenendo sempre in considerazione la fattibilità delle soluzioni sviluppate. In primo luogo, sono descritti gli algoritmi classici di Image Processing realizzati, focalizzandosi sugli approcci basati su regioni e sulla morfologia matematica. Dopodiché, si introducono le tecniche di Pattern Recognition, applicando il clustering fuzzy non supervisionato e i modelli basati su grafi (i.e., Random Walker e Automi Cellulari) per l'elaborazione di dati multispettrali e multimodali di imaging medicale. In riferimento ai metodi di Computational Intelligence, viene presentato un innovativo framework evolutivo basato sugli Algoritmi Genetici per il miglioramento e la segmentazione di immagini medicali. Inoltre, è discussa la co-registrazione di immagini multimodali utilizzando Particle Swarm Optimization. Infine, si investigano le Deep Neural Network: (i) le capacità di generalizzazione delle Convolutional Neural Network nell'ambito della segmentazione di immagini medicali provenienti da studi multi-istituzionali vengono affrontate mediante la progettazione di un'architettura che integra blocchi di ricalibrazione delle feature, e (ii) la generazione di immagini medicali realistiche basata sulle Generative Adversarial Network è applicata per scopi di data augmentation. In conclusione, il fine ultimo di tali studi è quello di ottenere conoscenza clinicamente e biologicamente utile che possa guidare le diagnosi e le terapie differenziali, conducendo verso l'integrazione di dati biomedicali per la medicina personalizzata. Difatti, i metodi assistiti da calcolatore per l'analisi delle immagini biomedicali sono vantaggiosi sia per la definizione di biomarcatori basati sull'imaging sia per la medicina e biologia quantitativa.
(2019). Computer-Assisted Analysis of Biomedical Images. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2019).
Computer-Assisted Analysis of Biomedical Images
RUNDO, LEONARDO
2019
Abstract
Nowadays, the amount of heterogeneous biomedical data is increasing more and more thanks to novel sensing techniques and high-throughput technologies. In reference to biomedical image analysis, the advances in image acquisition modalities and high-throughput imaging experiments are creating new challenges. This huge information ensemble could overwhelm the analytic capabilities needed by physicians in their daily decision-making tasks as well as by biologists investigating complex biochemical systems. In particular, quantitative imaging methods convey scientifically and clinically relevant information in prediction, prognosis or treatment response assessment, by also considering radiomics approaches. Therefore, the computational analysis of medical and biological images plays a key role in radiology and laboratory applications. In this regard, frameworks based on advanced Machine Learning and Computational Intelligence can significantly improve traditional Image Processing and Pattern Recognition approaches. However, conventional Artificial Intelligence techniques must be tailored to address the unique challenges concerning biomedical imaging data. This thesis aims at proposing novel and advanced computer-assisted methods for biomedical image analysis, also as an instrument in the development of Clinical Decision Support Systems, by always keeping in mind the clinical feasibility of the developed solutions. The devised classical Image Processing algorithms, with particular interest to region-based and morphological approaches in biomedical image segmentation, are first described. Afterwards, Pattern Recognition techniques are introduced, applying unsupervised fuzzy clustering and graph-based models (i.e., Random Walker and Cellular Automata) to multispectral and multimodal medical imaging data processing. Taking into account Computational Intelligence, an evolutionary framework based on Genetic Algorithms for medical image enhancement and segmentation is presented. Moreover, multimodal image co-registration using Particle Swarm Optimization is discussed. Finally, Deep Neural Networks are investigated: (i) the generalization abilities of Convolutional Neural Networks in medical image segmentation for multi-institutional datasets are addressed by conceiving an architecture that integrates adaptive feature recalibration blocks, and (ii) the generation of realistic medical images based on Generative Adversarial Networks is applied to data augmentation purposes. In conclusion, the ultimate goal of these research studies is to gain clinically and biologically useful insights that can guide differential diagnosis and therapies, leading towards biomedical data integration for personalized medicine. As a matter of fact, the proposed computer-assisted bioimage analysis methods can be beneficial for the definition of imaging biomarkers, as well as for quantitative medicine and biology.File | Dimensione | Formato | |
---|---|---|---|
phd_unimib_802621.pdf
accesso aperto
Descrizione: tesi di dottorato
Dimensione
12.83 MB
Formato
Adobe PDF
|
12.83 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.