Neuromodulation to Promote Upper Limb Recovery After Stroke: Clinical and Preclinical Insights Into Cerebellar Transcranial Direct Current Stimulation Combined With Robotic or Motor Training

Introducing and Refining Language in Vision Models

This thesis investigates how to extend the role of language within multimodal artificial intelligence, moving from its initial grounding in visual systems toward improved usability in representation, evaluation, and generative control. While vision-and-language models (VLMs) have achieved impressive capabilities, they still struggle with two key challenges: integrating language into domains that traditionally rely on purely visual or sensory information, and effectively leveraging the full ex...

From Words to Diagnosis: Exploring the Predictive Potential of Language Competence and Implicit Learning in Parkinson's Disease

Parkinson’s disease (PD) is traditionally considered a motor disorder, but current medical research increasingly adopts a systemic perspective on this condition, acknowledging that non-motor symptoms are also present. Albeit more subtle, they precede motor symptoms in the time course of disease progression. For this reason, focusing on non-motor symptoms is critical to accelerate disease detection and might represent a paradigm shift in PD diagnosis, which is currently clinical- and motor-bas...

Exploring the potential of e-Health-based multimodal interventions for the management of chronic non-cancer pain in older adults

Nanoparticle-based extracellular vesicles and stem cells: labelling, imaging and safety assessment

Nanoparticles (NPs) are transforming biomedical research thanks to their unique physicochemical properties, which enable precise interactions with biological systems and new tracking applications in regenerative therapy. This thesis investigates the use of NPs, specifically iron oxide and gold NPs, in combination with stem cells and extracellular vesicles (EVs) to optimize the non-invasive tracking of those therapeutic agents in preclinical models. Experiments in vitro and in vivo wer...

Un’idea di storia dell’arte. La metodologia storico-artistica di Max Dvořák (1874-1921) attraverso le lezioni di storia dell’arte barocca all’Università di Vienna

The Vienna School of Art History has played a pivotal role in shaping art history as a scientific discipline and in proposing new areas of analysis. One of its members, Max Dvořák (1874-1921), has contributed significantly to the development of a new scientific methodology, particularly in describing the metamorphosis this methodology underwent from the beginning of the 19th century to the 20th. This aspect of his thought has been overlooked by Italian historiography, which has concentrated m...

Motor resonance in functional movement disorders: unravelling repertoire-congruent mechanisms of action observation

Functional Movement Disorders (FMDs) are a subtype of Functional Neurological Disorders defined by abnormal involuntary movements such as weakness, tremor, dystonia, gait disturbances, myoclonus, or parkinsonism. These clinical features are typically inconsistent with recognized patterns of organic neurological disease and often fluctuate or change markedly under distraction. FMDs are increasingly conceptualized as disorders of network-level dysfunction, in which aberrant predictive coding, i...

Recreating biological barriers in a microfluidic system: applications of a new bioreactor

The need to develop increasingly faithful in vitro models capable of representing human organs and tissues under both physiological and pathological conditions is becoming pressing. Every year, countless topical and oral formulations are tested in the laboratory to evaluate their biodistribution, efficacy, safety, and potential toxicity. It is clear that the more accurately a laboratory model reflects the in vivo situation, the more informative and predictive the data obtained during this pre...

Advanced Neural Networks Verification for Safe and Explainable Intelligent Systems

Deep neural networks (DNNs) have revolutionized artificial intelligence (AI), enabling breakthroughs across domains ranging from medical robotics to sustainable energy management. Yet, their fragility to adversarial perturbations and their opacity as “black boxes” pose significant risks when deployed in safety-critical contexts. In this thesis, we address these challenges by advancing the field of neural network verification, with a particular focus on enhancing the safety and explainability ...