Carregant...
Fitxers
Tipus de document
TesiVersió
Versió publicadaData de publicació
Si us plau utilitzeu sempre aquest identificador per citar o enllaçar aquest document: https://hdl.handle.net/2445/217178
Application of Radiomics-based Machine Learning models on complex cardíac diseases
Títol de la revista
Autors
ISSN de la revista
Títol del volum
Resum
[eng] This doctoral thesis investigates the combination of radiomics and machine learning (ML) within the field of cardiology, emphasizing early detection and prognosis of complex cardiovascular diseases (CVDs). Radiomics converts standard medical images into detailed, high-dimensional data sets that expose subtle cardiac pathologies invisible to the naked eye. A specialized pipeline was developed to extract radiomic features and apply ML for effective classification and analysis. Each chapter modifies and optimizes this pipeline for different scenarios, culminating in a survival analysis ML pipeline in the final chapter. The thesis starts with a fundamental overview of cardiovascular diseases, radiomics, and ML, laying the groundwork for their use in cardiac imaging. The second chapter illustrates how cardiovascular magnetic resonance (CMR) radiomics and ML can differentiate left ventricular non-compaction cardiomyopathy (LVNC) from other types such as hypertrophic and dilated cardiomyopathies. In this chapter, we illustrated that radiomics is crucial in this differentiation process, showing that an automated ML-Radiomics pipeline can achieve state- of-the-art benchmarks from the clinical variables used in current medical practice, but without need for clinicians manual delineations. The subsequent chapter explores the benefits of integrating CMR radiomics with electrocardiogram (ECG) data to enhance atrial fibrillation (AF) detection. In this study of the UK Biobank participants we demonstrated that an ECG-based model had lower accuracy to detect AF in female subjects compared to males. The inclusion of CMR radiomics combined with ECG increased the model performance in women. The main universal implication is that a combined approach of ECG and atrial imaging might lead to better assessment of female participants suspected of AF. The fourth chapter evaluates the capacity of CMR radiomics and ML to forecast significant cardiovascular events like AF, heart failure (HF), myocardial infarction (MI), and stroke, using data from the UK Biobank. Incorporating radiomic features with vascular risk factors (VRFs) and CMR indices significantly enhances the performance of these predictive models. Radiomics features provided additional information over VRFs, although the improvement was only marginal compared to conventional CMR metrics. The improvement was most prominent in AF and HF prediction, which highlight that the performance of radiomics models is dependent on the disease aetiology and mechanism. The final chapter focuses on the use of radiomics and ML for identifying genetic cardiomyopathy in patients with excessive trabeculation. In a multicenter cohort study of individuals diagnosed with excessive trabeculation of the left ventricle radiomics analysis of standard, non-enhanced cine CMR images provided added value beyond left ventricular ejection fraction in the identification of a genetic or familial substrate and of adverse prognosis. Textural radiomics features were instrumental to recognize a genetic or familial substrate, while shape features dominated the identification of adverse prognosis. This thesis underscores the potential of radiomics and ML to advance cardiac diagnostic and prognostic capabilities, providing a more precise, personalized approach to managing CVDs.
Descripció
Matèries (anglès)
Citació
Citació
IZQUIERDO, Cristián. Application of Radiomics-based Machine Learning models on complex cardíac diseases. [consulta: 30 de novembre de 2025]. [Disponible a: https://hdl.handle.net/2445/217178]