Please use this identifier to cite or link to this item: http://hdl.handle.net/2445/65343
Title: lnvestigation of quantitative imaging biomarkers for assessing perínatal outcomes
Author: Bonet Carné, Elisenda
Director: Gratacós Solsona, Eduard
Marqués, Ferran
Keywords: Neonatologia
Malalties neonatals
Diagnòstic per la imatge
Ecografia
Neonatology
Neonatal diseases
Diagnostic imaging
Ultrasonic imaging
Issue Date: 20-Nov-2014
Publisher: Universitat de Barcelona
Abstract: [spa] Esta tesis está compuesta por cuatro estudios para probar el uso de los métodos cuantitativos de análisis de texturas de imágenes para la predicción del riesgo en dos patologías fetales. En la mayor parte de la tesis (tres estudios) se utilizan imágenes de ultrasonido del tórax fetal para predecir la morbilidad respiratoria neonatal. En el primer estudio se relacionan las texturas obtenidas de las imágenes ecográficas del pulmón fetal con la edad gestacional, demostrando que se puede extraer información del tejido de forma no-invasiva que se correlaciona con el proceso fisiológico subyacente, la maduración normal del pulmón fetal. En el segundo estudio se correlacionan las texturas de las imágenes con el resultado del test TDx-FLM II, test que utiliza una muestra de líquido amniótico para predecir la madurez pulmonar feta, demostrando que el análisis de texturas de las imágenes pulmonares fetales contiene información sobre la madurez pulmonar fetal. En el tercer estudio, se desarrolla y evalúa un nuevo método de análisis de imagen del pulmón fetal para la predicción de la morbilidad respiratoria neonatal. Se describen los principios básicos de este nuevo método y se evalúa el funcionamiento del mismo con muestras ciegas. Los resultados obtenidos con el método no invasivo desarrollado para predecir la morbilidad respiratoria neonatal son similares a los reportados por las pruebas actuales, que requieren de líquido amniótico para el análisis y, por tanto, de una muestra obtenida de forma invasiva. Adicionalmente, en el último estudio se ha evaluado la capacidad del análisis cuantitativo de imagen en imágenes de resonancia magnética para detectar anomalías en distintas áreas del cerebro fetal que pueden estar asociadas con un neurocomportamiento neonatal anormal. De esta forma se prueba la transversalidad de las técnicas de análisis de texturas en distintos tipos de imágenes y patologías. Como conclusión final de los cuatro estudios, esta tesis aporta evidencias de que las técnicas no invasivas de análisis de imágenes médicas extraen información cuantitativa del tejido examinado que puede usarse para ayudar en el diagnóstico clínico.
[eng] This Thesis consists of different studies focused in advancing towards the development of non­invasive imaging biomarkers to predict perinatal clinical outcomes. The structure of the PhD Thesis is divided in four projects to explore the development of a series of new methods based on image texture analysis allowing the analysis of medical images (i.e. ultrasound or magnetic resonance imaging) in the field of fetal medicine applications -mainly fetal lung maturity and fetal brain assessment-, to test their reproducibility and to select the best performing approach to develop an imaging biomarker predicting a clinical outcome of interest. The majority of the work was focused on developing a quantitative imaging biomarker for neonatal respiratory morbidity. In order to achieve the objectives and to explore the development of a quantitative imaging biomarker fetal thorax ultrasound images were used for the studies 1, 2 and 3 to predict neonatal respiratory morbidity. To test the transversality of the quantitative texture analysis in other pathological models, fetal brain magnetic resonance images from Small-for-Gestational Age fetuses were used in study 4. First study demonstrates that quantitative image features extracted from fetal thorax ultrasound images correlate with gestational age. This study also demonstrated that it is posible to extract information from the tissue in a non-invasive manner that correlated with the underlying physiological process, regular fetal lung maturation. In the second study the correlation between texture analyses and the existing fetal lung maturity test was tested. Thus, the second study provided evidence that the image features from lung ultrasound images correlate with fetal lung maturity test assessed by a standard test as TDx-FLM II. These findings opened the possibility to explore the introduction of non-invasive techniques into clinical practice to test fetal lung maturity. In the third study, the basic principles of a novel method to predict neonatal respiratory morbidity risk (quantusFLM™) were described, and a validation was performed to assess the ability of the method to blindly predict the risk of neonatal respiratory morbidity. Remarkably, this study provides evidence that purpose­developed software based on quantitative texture analysis of fetal lung ultrasound images predicts neonatal respiratory morbidity with a similar performance to that reported for commercial fetal lung maturity tests in amniotic fluid. Additionally, in the last study the ability of image texture analysis to detect abnormalities in different fetal brain areas was evaluated, and their association with abnormal neonatal neurobehavior was tested. This study demonstrated the potential of quantitative imaging texture analysis for other image acquisition techniques and clinical outcomes. As a final conclusion, this Thesis provides evidence that the non-invasive quantitative imaging techniques based on texture analysis extract quantitative information related to the underlying tissue that could be used to assist in clinical diagnosis.
URI: http://hdl.handle.net/2445/65343
Appears in Collections:Tesis Doctorals - Facultat - Medicina

Files in This Item:
File Description SizeFormat 
EBC_PhD_THESIS.pdf6.05 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.