Riprogrammare l'esplorazione visiva: Dall'apprendimento statistico alla riabilitazione post-ictus attraverso il biofeedback basato sullo sguardo

This thesis examines the nature and mechanisms of visuospatial attention, the impact of stroke on visual exploration, and a potential route toward effective rehabilitation. Chapter 1 sets the stage by tracing how attention evolved as a core mechanism for filtering relevant information from complex sensory inputs, highlighting its tight linkage with oculomotor control. It underscores that attention is not purely perceptual but fundamentally tied to motor actions that help organisms select and orient toward salient aspects of the environment. Chapter 2 shifts the focus to clinical scenarios, illustrating the wide variety of visual impairments experienced by stroke survivors. These may range from visual field defects and oculomotor disorders to attentional impairments, such as unilateral spatial neglect, in which individuals fail to notice stimuli on one side of space. Although several rehabilitation strategies have been proposed for these disorders, including restitution-based exercises, compensatory scanning protocols, and optical or sensory substitutions, robust empirical evidence supporting their long-term efficacy has remained elusive. Chapter 3 brings to light a novel experimental approach to modifying attentional biases through statistical learning, in which individuals search for invisible targets guided in real-time by auditory feedback which is tightly linked to the recorded eye movements. The demonstration that statistical learning of spatial regularities transfers to rest and free viewing of natural scenes suggests that prior experience, or “selection history”, can bias where the eyes move, even outside the original training context, thereby providing a novel perspective as well as a potential tool for ameliorating pathological spatial biases in patient populations. Chapter 4 channels these insights into a new rehabilitation protocol that combines real-time gaze monitoring and auditory feedback to drive the patient’s visual exploration of naturalistic scenes towards a normative pattern obtained from healthy observers. Building on the active inference framework, this protocol treats saccadic eye movements as an adaptive response to form accurate hypothesis about the external world, ensuring that neglected regions of the visual scene are sampled. Finally, Chapter 5 introduces preliminary clinical evidence indicating that participants who completed multiple sessions with this gaze-contingent biofeedback approach showed improved coverage of visual scenes and, in certain cases, better performance in daily tasks. Although the sample sizes were small and the clinical profiles highly heterogeneous, these early observations point to the protocol’s promise in helping2 patients to restore normal viewing patterns, raising hopes for a more robust, large-scale evaluation in future research. Chapter 6 summarizes the findings of the current thesis, weighs its scientific and clinical contributions, and charts future outlooks.