Abstract:
Functional Magnetic Resonance Imaging (fMRI), particularly at ultra-high-field strengths such as 7-Tesla (7T), has revolutionized neuroscience by enabling detailed exploration of the human brain’s functional architecture. This thesis leverages the advanced spatial resolution of 7T-fMRI to investigate the sensorimotor cortex, focusing on age-related changes and disease-specific alterations associated with Amyotrophic Lateral Sclerosis (ALS). While the unprecedented resolution of 7T-fMRI allows for the analysis of fine-grained features, such as cortical columns and laminar structures, the high dimensionality and complexity of the data necessitate sophisticated analytical methods.
This work employs 7 T task-based fMRI with advanced multivariate techniques, including Robust Shared Response Modeling (rSRM), Columnar Shared Response Modeling (C-SRM) (introduced as a novel method), and Partial Least Squares Regression (PLSR), to address the challenges posed by high-dimensional fMRI data. SRM and C-SRM are utilized to align functional data across participants and examine fine-scale neural organization, particularly at the columnar level, while PLSR links neural activity to clinical and behavioral outcomes. These approaches enable the identification of both shared and individual-specific neural patterns, offering a nuanced understanding of functional changes in the sensorimotor cortex.
In the context of aging, rSRM and C-SRM show that the hierarchical layout of Brodmann areas (BA) 3b → 1 → 2 is preserved in older adults, yet digit representations become less precise. In BA1, the optimal number of functional columns drops relative to young adults, indicating enlarged, less se- lective columns; while BA3b remains stable. These findings reveal a subtle dedifferentiation—blurred maps but an intact hierarchy—consistent with compensatory pooling of sensory inputs.
In ALS, rSRM combined with PLSR distinguishes patients from controls with high accuracy based on task-evoked BOLD patterns. Connectivity-derived latent variables outperform activation maps in clustering disease onset site and staging, and they exhibit an atopographic signature: foot and face regions of MI track progression regardless of the initial symptom locus. The data suggest an early, network-wide hyper-connective compensation that collapses as degeneration advances.
Together, these results validate advanced alignment and dimensionality-reduction strategies for extracting fine grained insights from 7 T data. They establish enlarged columns as a fingerprint of healthy ageing and identify network-level connectivity markers for ALS staging—outcomes that can guide larger multicentre, multimodal studies and inform therapeutic efforts aimed at preserving sensorimotor function across the lifespan and in neurodegenerative disease.
