The GASPAR trial is a pragmatic, parallel-arms, single-center, non-blinded, superiority randomized control trial in neurorehabilitation. The main objective is to test whether a 4 weeks gait rehabilitation program that uses augmented reality is superior to a conventional treadmill training program of equivalent intensity. Baseline assessments precede allocation, which consists in blocking randomization (2:1 ratio) with stratification according to the disease etiology. Post-intervention assessments serve to compare the short-term efficacy of the intervention between the two groups. Three months after discharge, follow-up assessments take place to detect potential long-term effects.

After a lesion to the central nervous system, many patients suffer from a diminished walking capability. During the first rehabilitation phase, specific cares help the recovery of motor function to offer the best chances of returning to normal walk. For example, the repeated practice of walking exercises facilitates the gain of muscular strength and stimulates motor relearning. However, in patients, who have finished the first rehabilitation phase, it is often observed strong limping, unsteady gait, and a bad management of obstacle clearance. Thus, these individuals fall frequently, with risk of severe injury. The aim of the study is to assess the efficacy of gait training on a treadmill equipped with an augmented reality system. Shapes are displayed onto the treadmill belt with a projector (beamer): the patient adapts his/her gait to the drawing that unfolds in front of him/her. It is possible to train gait symmetry for coordination enhancement or to exercise obstacle clearance and gait agility. Another advantage of this type of treadmill is the capability to continuously analyze gait with integrated sensors and, hence, to give real-time feedback to the patient. Whereas the method seems promising, thorough studies that would confirm its efficacy are lacking. Therefore, the investigators seek to conduct a randomized controlled trial to compare visually-guided gait training with a more classical training method. During four weeks, participants will train, according to their abilities, three to five times a week during 30 minutes on the augmented-reality treadmill. A group of control patients will train at comparable intensity but without the visual guidance of gait. Using the data collected during the training sessions, the investigators will study whether the evolution of locomotor function differs between both groups. Furthermore, several questionnaires will be filled in by the participants to document the self-evaluation of their progress. In the long term, the investigators expect to first improve the efficacy of gait rehabilitation and second to better understand the recovery process of locomotor function during the first months after a stroke.

Participants fulfilling all of the following inclusion criteria are eligible for the study:

• Informed Consent as documented by signature
• Stroke patients, patients with traumatic brain injury, or patients with spinal cord injury in the acute to sub-acute phase (maximum 40 days after the incident)
• Need for gait rehabilitation in reason of at least one of the following conditions: Paresis of the lower extremities. Severe balance disorders. No walking at entry due to the neurological injury
• Ability to walk 2 minutes without or with minimal aid, namely: With the help of one person, or: With walking aid (cane, walker)

The presence of any one of the following exclusion criteria will lead to exclusion of the participant:

• Age < 18 years
• Other clinically significant concomitant disease states (e.g., renal failure, hepatic dysfunction, cardiovascular disease, etc.) that may threaten the health in case of sustained exercise.
• Concomitant gait disorders induced by acute to sub-acute musculoskeletal injuries, e.g. fracture of the lower extremities.
• Severe pre-existent gait disorders that deeply affect walking abilities and gait pattern, either due to musculoskeletal (e.g. severe osteoarthritis) or neurological (e.g. Parkinson's disease) etiologies.
• Severe non-corrected visual impairment
• Inability to follow the procedures of the study, e.g. due to language problems, psychological disorders, dementia, etc. of the participant.

At the end of the study, final anonymized database will be made available through a specialized online repository