The trial aims to test a novel rehabilitation device for subacute stroke hemiplegic upper limbs based on state-of-the-art non invasive Brain-Computer Interface (BCI) robotic rehabilitation in a clinical setting. The investigators aim to prove the clinical efficacy and safety of BCI therapy over traditional rehabilitation methods.

The proposed rehabilitation device is the first neuro-rehabilitation system which combines non-invasive BCI and robotic rehabilitation for the paralysed stroke upper extremity within 6 months of stroke.Spontaneous recovery after stroke takes place over the first 6-12 months after stroke. The first 3-6 months are the most crucial periods as this is the period of maximal neurological recovery and neuroplasticity. Differential rates of recovery occur for various types of impairments post-stroke. In general, motor functions (mobility, walking, upper limb function, activities of daily living (ADL)) recover faster than cognitive or language impairments which may recover over 12 months. A number of approaches to stroke injury rehabilitation have been introduced to facilitate intrinsic recovery or aid adaptive compensation for stroke-related impairments. Generally for rehabilitative training to be effective, it must be commenced as early as possible after stroke. Current research proves that rehabilitation using traditional neuro-facilitation approaches is effective in improving neurological and functional recovery and is superior to no treatment or nursing care alone. Rehabilitated patients have shorter total hospitalization stays, lower complication rates, earlier and higher rates of discharge home than patients who do not receive rehabilitation. In addition, rehabilitation involving a multidisciplinary team approach led by rehabilitation physician or specialist result in better functional outcomes compared to acute general ward-based therapies. In order for rehabilitation to be effective in modifying cortical neuroplasticity, it must be targeted at the specific stroke impairment, task specific, exercise must be repetitive and intensive, goal directed and command the attention of the stroke patient. Some of the components of rehabilitation include physical therapy, gait and balance training, aerobic conditioning, functional Activities of Daily Living (ADL) training, physical modalities to treat pain, Functional Electrical Stimulation (FES) or Neuro-Muscular Electrical Stimulation (NMES). Other methods include specific treatments to address complications of rehabilitation such as spasticity, ataxia, contractures and bladder or bowel incontinence. Often, one-to-one and highly labour-intensive and expensive therapies with close hand-over-hand treatments are required. Limitations of current physiotherapy and occupational therapy techniques include the following:(1) Difficulties in rehabilitation for the severely paralysed arm and hand which are often treated with passive modalities such as NMES, passive ROMs and other modalities. (2) Difficulties in achieving intensive rehabilitation and high repetitions in those with moderate to severe upper extremity paralysis either due to non participation or pain which is commoner in those with severe paralysis. (3)Problems in motivating and sustaining patient interest in repetitive exercises.(4)Therapy is often perceived to be boring and due to lack of immediate biofeedback. (I) Robot Aided Rehabilitation: MIT (USA) has developed a robot, named the MIT-MANUS, to aid therapy of stroke victims. Small clinical trials have reported that the robot significantly improved patients' recovery of arm motor movement and function with sustained gains several months after cessation of treatment. This system is being clinically used as a rehabilitation training tool in over 20 centres world-wide. Advantages of robot aided rehabilitation include the ability to document and store motion and force parameters, the ability to achieve thousands of repetitions per treatment session (100 times more than conventional treatment or FES) without causing tissue injury or pain, high intensity with low friction, attention training and increased biofeedback through the incorporation of interactive video games, which can simulate trajectories, mazes, ADL tasks such as preparing a meal and spatial task simulation such as going shopping. In addition, after the initial training period, supervision of the patient by the therapist can be reduced due to the sustainability of participation of the patient from the robot or BCI based Robotic Rehabilitation. Hence, productivity of the human therapist is increased by the robot. The robot thus acts as a high technology aid to the clinician and therapist. The system is also portable, giving rise to the possibility of tele-rehabilitation options with the performance and progress of the patient being monitored by the institution remotely. (II)BCI-based BCI based Robotic Rehabilitation:This non invasive device aims to use a novel approach in robotic training, which has not been employed in the therapeutic realm before.In the MIT-MANUS and related commercially available systems, there is no direct communication between the patient's mind or thinking processes or motor volitional thinking and the robotic system. Although some sensors are used to detect the patient's weak movement, it never knows when and how the patient wants to move. The robot arm to which the patient is tethered or constrained plans the trajectory of movement for the patient and reduces its active role as the patient recovers voluntary motion. In most times, the patient can only passively follow the predefined program, which may not fully explore the patient's motor initiatives and potential or attention processes.

Inclusion Criteria:

1. Demographics: 21 to 65 years, within 12 months of first, single clinical stroke (ischaemic or haemorrhagic).
2. Moderate to severe upper extremity (UE) weakness post stroke.
3. Fugly-Meyer motor score of the upper limb < 40.
4. Upper limb motor power MRC grade 3 or less /5 in at least 1 arm region.
5. Able to give own consent and understand simple instructions and learn through practice.
6. Resting brain states determined by FMRI criteria

Exclusion Criteria:

1. Recurrent stroke.
2. Previous brain surgery.
3. Spasticity of Modified Ashworth scale > 2.
4. Fixed contracture of any upper limb joint
5. Ataxia, dystonia or tremor of the involved upper limb or previous cervical myelopathy
6. Upper limb pain or painful joints in upper limb.
7. Severe cognitive impairment (Abbreviated Mental Test <7/10), or severe aphasia which may affect ability to participate in training.
8. . History of seizures in the past 12 months.
9. Severe left neglect