Functional MRI Data Processing and Analysis: Novel Approaches with Application to Atypical Reading

Functional MRI (fMRI) has traditionally been used to localize brain areas engaged during tasks or at rest, and to map brain function. However, understanding complex cognitive processes requires more refined approaches to capture dynamic and spatially specific neural responses. Numerous methodological choices influence the outcome of fMRI studies, from acquisition—where single-echo and multi-echo echo-planar imaging (EPI) sequences affect signal quality and sensitivity—to preprocessing, where denoising, task regression, and filtering strategies shape the estimation of functional connectivity (FC). Each stage introduces variability but also opportunity: by systematically addressing them, we can improve the robustness and specificity of insights into brain function. In parallel, ultra-high-field (UHF) MRI and ultra-high-resolution acquisitions extend access to information such as layer-specific dynamics, broadening the scope of fMRI. This doctoral research develops and evaluates a framework that integrates methodological advancements in fMRI with cognitive neuroscience applications in atypical reading. The work has two aims: (1) to advance fMRI methods, including acquisition protocols, preprocessing pipelines, and modeling strategies; and (2) to apply these methods to investigate neural mechanisms supporting reading under increasing perceptual and cognitive demands. The thesis is organized into four main studies. The first examines FC patterns derived from task-based fMRI. While FC is usually computed from resting-state data, connectivity extracted from task-evoked signals is gaining interest. We assessed how preprocessing choices influence connectomes and their clinical utility, emphasizing the need for methodological rigor when using FC for trait-level inferences. The second study compares single-echo and multi-echo EPI sequences. Multi-echo protocols allow better separation of BOLD and non-BOLD fluctuations and support advanced denoising. We evaluated their impact on signal quality and model estimation at both individual and group levels, focusing on sensitivity to task-related activations during visually degraded word reading. This analysis clarifies methodological trade-offs and guides the choice of acquisition strategies for probing higher-order functions under demanding conditions. The third study introduces and validates an fMRI paradigm designed to probe reading under atypical conditions. Stimuli were manipulated along several visual dimensions with varying difficulty, enabling investigation of brain responses to increasing cognitive load. Behavioral measures were integrated with activation patterns to explore inter-individual differences in reading strategies and resilience to visual disruption. The final study employs UHF fMRI to investigate layer-specific responses. Laminar imaging distinguishes feedforward (bottom-up) from feedback (top-down) signals based on cortical layer profiles, offering unique insight into hierarchical processing. Using the degraded word reading task, this preliminary work sets the stage for examining how cortical layers contribute under varying perceptual demands. Together, these studies demonstrate how methodological decisions—from acquisition to preprocessing and modeling—profoundly affect the mapping of brain function. By integrating these findings into a unified framework, the thesis advances fMRI methodology and shows how careful design improves the accuracy, interpretability, and generalizability of results. At the same time, it contributes to understanding reading as a distributed cognitive process, with implications for both basic science and clinical interventions in individuals with reading difficulties.