Stroke produces widespread disturbances in whole-brain organisation, yet the mechanisms through which focal lesions alter functional connectivity and how structurally intact regions reorganise remain incompletely understood. Connectivity gradients provide a framework for characterising macroscale functional architecture, capturing continuous axes of organisation that extend beyond discrete network boundaries. This thesis examines how stroke affects these gradients and evaluates whether gradient changes explain variation in post-stroke cognition. The thesis also explores the use of structural imaging for modelling post-stroke functional reorganisation, assessing its value in contexts where resting-state fMRI is not routinely obtained in clinical settings. This work draws on two chronic stroke cohorts: a group of individuals with semantic aphasia and a subset of the PLORAS dataset. Resting-state fMRI from both cohorts (mean post-stroke interval > 4 years) was used to characterise long-term functional changes. First, in semantic aphasia, lesion information from structural MRI was used to model the direct effects of stroke in connectivity space and to test whether simulated post-stroke gradients predicted semantic control impairments. Second, connectivity gradient reorganisation was characterised in the PLORAS cohort to determine whether large-scale organisational changes in the unlesioned cortex were linked to language outcomes. Third, an iterative propagation simulation model was evaluated, capturing direct and indirect effects of stroke across the cortex, and tested to determine whether gradient–behaviour associations observed in the simulations were reflected in empirical stroke gradients. Across the three empirical studies, the thesis demonstrates that stroke consistently attenuates the brain's principal connectivity gradient axes, reducing differentiation between transmodal and unimodal systems, visual and motor networks, and control-default mode regions. In semantic aphasia, lesion-based simulations reproduced the empirically observed reorganisation moderately well, particularly along the transmodal–unimodal axis. These simulations revealed that shifts of the left inferior frontal gyrus toward a more unimodal connectivity profile were associated with poorer semantic control, highlighting how altered embedding within large-scale hierarchies contributes to semantic deficits. In the larger and more diverse PLORAS cohort, distributed reorganisation of connectivity gradients in the unlesioned cortex explained language outcomes beyond the effects of focal lesions. Shifts along the control-default mode axis, particularly within the left inferior frontal gyrus, were associated with better speech but poorer writing performance. The final study showed that iterative propagation simulations improved correspondence with empirical gradients, especially for the visual-motor axis. One replicated finding was that a shift of the right inferior frontal gyrus toward the motor end of the visual–motor axis was associated with better phonology and working memory. Together, these findings show that connectivity gradients provide a low-dimensional, clinically informative framework for understanding how focal lesions and distributed reorganisation shape post-stroke cognition. They also highlight the current potential and limitations of structural imaging-based simulations for modelling post-stroke functional reorganisation.
