The quantitative and qualitative assessment of superficial soft tissues, particularly muscle and subcutaneous fat, plays a crucial role in nutrition, geriatrics, and metabolic medicine for the diagnosis and monitoring of conditions such as metabolic syndrome and disease-related malnutrition. Although advanced techniques such as computed tomography and magnetic resonance imaging provide detailed data, they are highly complex to perform. Ultrasound, on the other hand, is practical, free from ionizing radiation, low cost, and repeatable, but suffers from interpretative variability and a lack of standardization. The aim of this work, carried out in collaboration with Strumedical s.r.l., is to develop and validate a prototype software for the automatic segmentation of ultrasound images, aimed at providing support to clinicians in measuring clinical parameters such as the thickness and echogenicity of the tissue of interest. To this end, several automatic segmentation algorithms based on Region Growing, Watershed, and Chan-Vese were tested, integrated with a standardized acquisition protocol and pre-processing procedures to improve image quality. The results showed that the Region Growing-based method was the most suitable for the purpose, ensuring segmentations and estimates of clinical indices most similar to the reference ones available in the analyzed database. In conclusion, the software prototype developed can be a valuable aid to clinicians in the ultrasound evaluation of soft tissues. This thesis work is a small step towards the standardization of ultrasound image analysis and, potentially, towards the integration of these technologies into portable devices and telemonitoring solutions, with potential clinical and research applications.
La valutazione quantitativa e qualitativa dei tessuti molli superficiali, in particolare muscolare e adiposo sottocutaneo, riveste un ruolo cruciale in nutrizione, geriatria e medicina metabolica per la diagnosi e il monitoraggio di condizioni quali sindrome metabolica e malnutrizione correlata alla malattia. Sebbene tecniche avanzate come tomografia computerizzata e risonanza magnetica forniscano dati dettagliati, presentano complessità sperimentali elevate. L’ecografia, al contrario, garantisce praticità, assenza di radiazioni ionizzanti, costi ridotti e ripetibilità, ma soffre di variabilità interpretativa e mancanza di standardizzazione. L’obiettivo di questo lavoro, realizzato in collaborazione con Strumedical s.r.l., è sviluppare e validare un prototipo di software per la segmentazione automatica delle immagini ecografiche, volto a fornire un supporto al clinico riguardo misurazioni di parametri clinici come dello spessore ed ecogenicità del tessuto di interesse. A tal fine, sono stati testati diversi algoritmi di segmentazione automatici, basati su Region Growing, Watershed e Chan-Vese, integrati con un protocollo di acquisizione standardizzato e procedure di pre-processing per il miglioramento della qualità d’immagine. I risultati hanno evidenziato il metodo basato su Region Growing come più adatto allo scopo, garantendo segmentazioni e stime degli indici clinici più simili a quelle di riferimento disponibili nel database analizzato. In conclusione, il prototipo di software sviluppato può rappresentare un valido supporto al clinico nella valutazione ecografica dei tessuti molli . Questo lavoro di tesi costituisce un piccolo passo verso la standardizzazione dell’analisi di immagini ecografiche e, potenzialmente, verso l’integrazione di tali tecnologie in dispositivi portatili e soluzioni di telemonitoraggio, con potenziali applicazioni cliniche e di ricerca.
SVILUPPO DI UN SOFTWARE PER IL RICONOSCIMENTO DI TESSUTI MOLLI SUPERFICIALI DA IMMAGINI ECOGRAFICHE
SABBATINELLI, MATTIA
2024/2025
Abstract
The quantitative and qualitative assessment of superficial soft tissues, particularly muscle and subcutaneous fat, plays a crucial role in nutrition, geriatrics, and metabolic medicine for the diagnosis and monitoring of conditions such as metabolic syndrome and disease-related malnutrition. Although advanced techniques such as computed tomography and magnetic resonance imaging provide detailed data, they are highly complex to perform. Ultrasound, on the other hand, is practical, free from ionizing radiation, low cost, and repeatable, but suffers from interpretative variability and a lack of standardization. The aim of this work, carried out in collaboration with Strumedical s.r.l., is to develop and validate a prototype software for the automatic segmentation of ultrasound images, aimed at providing support to clinicians in measuring clinical parameters such as the thickness and echogenicity of the tissue of interest. To this end, several automatic segmentation algorithms based on Region Growing, Watershed, and Chan-Vese were tested, integrated with a standardized acquisition protocol and pre-processing procedures to improve image quality. The results showed that the Region Growing-based method was the most suitable for the purpose, ensuring segmentations and estimates of clinical indices most similar to the reference ones available in the analyzed database. In conclusion, the software prototype developed can be a valuable aid to clinicians in the ultrasound evaluation of soft tissues. This thesis work is a small step towards the standardization of ultrasound image analysis and, potentially, towards the integration of these technologies into portable devices and telemonitoring solutions, with potential clinical and research applications.| File | Dimensione | Formato | |
|---|---|---|---|
|
Sabbatinelli_Mattia.pdf
Accesso riservato
Dimensione
1.35 MB
Formato
Adobe PDF
|
1.35 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/98773