RT info:eu-repo/semantics/doctoralThesis
T1 Design and reliability validation of diffusion MRI biomarkers for the brain's white matter
A1 París Bandrés, Guillem
A2 Universidad de Valladolid. Escuela de Doctorado
K1 Resonancia magnética
K1 MRI
K1 RM
K1 Diffusion
K1 Difusión
K1 Reliability
K1 Fiabilidad
K1 Noise
K1 Ruido
K1 33 Ciencias Tecnológicas
AB Diffusion Magnetic Resonance Imaging (dMRI) provides a unique, non-invasive window into the microstructural organization of the human brain by capturing the diffusion of water molecules. However, a fundamental limitation remains: the physical diffusion process occurs at microscopic scales, while the actual imaging resolution is millimetric. This mismatch introduces significant challenges in linking what is measured to the true underlying tissue properties. Various modeling approaches have been developed to bridge this scale gap and extract quantitative biomarkers that describe relevant features of the tissue. Despite this progress, there is still a notable divide between what cutting-edge research protocols can achieve and what is feasible in standard clinical practice or more general research contexts. This thesis addresses this gap by focusing on making advanced microstructural biomarkers more practical, reliable, and robust for wider use.The work is structured around three main objectives. First, it introduces and implements new biomarkers specifically designed to describe local white matter diffusion properties. These markers aim to be informative for clinical and neuroscientific applications, while avoiding over-reliance on high-end equipment or impractically complex acquisition protocols. Second, it examines the role of thermal noise (a fundamental and unavoidable feature of MR data acquisition) highlighting its impact on the stability and interpretability of model-based estimates. Despite its influence, thermal noise is often treated implicitly or neglected altogether. This thesis explicitly quantifies its effects and develops strategies to mitigate noise-induced bias and instability, using robust estimation and informed regularization. Third, it extends current approaches to assessing reliability in dMRI by clearly distinguishing two fundamental aspects: repeatability (the stability of a metric under the same conditions) and separability (the capacity to differentiate between different conditions or subjects). By separating these dimensions, the thesis provides a more detailed perspective on what “reliability” truly means in the context of diffusion metrics, which is particularly important for studies with limited sample sizes or variable conditions.To support these objectives, the research combines synthetic data and repeated in vivo scans, providing a strong basis for evaluating how different estimation strategies perform under realistic conditions. The results demonstrate that commonly used test-retest measures alone are insufficient to fully describe the reliability of microstructural biomarkers. Instead, a multidimensional approach that includes both repeatability and separability yields a clearer picture of a metric’s practical utility. Additionally, the findings confirm that thermal noise can systematically distort parameter estimates, especially in models with degenerate or ill-posed solution spaces: a typical challenge in biophysical diffusion modelling. The thesis proposes and validates robust constrained estimators and noise-aware corrections that improve estimation quality under realistic acquisition conditions.In conclusion, this work delivers both practical tools and conceptual advances for the field of diffusion MRI. It advocates for a more nuanced approach to reliability, moving beyond simple test-retest metrics, and emphasizes the need to explicitly address thermal noise in the modelling process. Overall, the methods and insights presented here are intended to help close the gap between advanced diffusion MRI research and its translation to broader clinical and research applications, ultimately making sophisticated biomarkers more accessible, interpretable, and dependable in everyday practice.
YR 2025
FD 2025
LK https://uvadoc.uva.es/handle/10324/81931
UL https://uvadoc.uva.es/handle/10324/81931
LA eng
NO Escuela de Doctorado
DS UVaDOC
RD 21-ene-2026