Accepting PhD Students

PhD projects

1. Brain-Machine Interfaces for Neural Rehabilitation after Stroke The brain-machine interface (BMI) technology aims to restore motor function disability in patients after stroke. This research work will contribute to the development and cross-validation of the BMI systems in Kazakhstan to improve the quality of life for disabled and motor-impaired people. It will focus on areas at the interface between engineering, robotics, neuroscience, and medicine. A physical human-robot interface will be designed and augmented with a non-invasive BMI to provide neural rehabilitation therapy for patients after stroke across a broad spectrum of impairment severity in the rehabilitation tasks. One of the key contributions of this proposed work will include: devising advanced electroencephalogram (EEG) interface methods (e.g. signal processing and machine learning algorithms) to stroke patients and developing a BMI for the control of the therapeutic robot that will improve upper and lower limb motor function, as well as to conduct clinical trials in collaboration with the local rehabilitation centers. Annually, over 49 thousand people suffer a stroke in Kazakhstan, 80% of them became wholly or partly disabled after the strokes (http://en.tengrinews.kz/health/-50-000-Kazakhstan-citizens-suffer-strokes-annually-9657/), and this research project will have a substantial impact in the engineering of novel tools for efficient stroke rehabilitation of patients locally. 2. Design and Optimization of a P300 Visual Brain-Computer Interface Speller for patients in Kazakhstan. Approximately half a million people living with Amyotrophic Lateral Sclerosis (ALS), a neurological disease that affects neurons responsible for voluntary movements. Patients rapidly lose the ability to move their arms, legs, and face muscles. Gradually, ALS patients become unable to communicate. However, the brain of an ALS patient is fully functional; this is an opportunity that provides a gateway to developing assistive devices for communication. The only means of communication for those patients are using a Brain-Computer Interface (BCI). The BCI speller is a system that enables persons unable to communicate naturally due to neuromuscular diseases to communicate with the external world. It acquires brain signals in response to visual stimuli the person is shown to on the screen and then analyzes in real-time to predict the desired symbol. To date, most BCI design paradigms have been focused on the development of a speller to communicate English or Latin-based languages. Due to a lack of BCI spellers for patients speaking the Kazakh language, this study will design and optimize a new BCI speller in collaboration with the local medical centers to enable thousands of ALS patients to communicate the Kazakh language. Besides, the BCI speller could be potentially useful for people with communication problems: persons afflicted with brain or spinal cord injuries, multiple sclerosis, muscular dystrophies, and cerebral palsy. 3. Feature Representation Learning of High-Frequency Oscillations strongly relevant to Ictogenesis Epilepsy affects approximately 1% of the population, and about 30~40% of them fail to control seizures with medications. For these patients with medically refractory epilepsy, epilepsy surgery is the best treatment option currently available. The success of epilepsy surgery largely depends on the precise localization of the epileptogenic region, brain areas that are capable of generating spontaneous seizures now or in the future, and that should be surgically removed or disconnected. The current clinical practice encompasses the identification of the seizure onset zone (SOZ) and extended epileptic area (or irritative zone) by multiple days of intracranial EEG recordings and the surgical resection of these regions. However, these invasive procedures do not always warrant favorable surgical outcome: for its worse, the seizure freedom rate after surgery declines with time. Among many candidates of an electrophysiological biomarker of an epileptic brain, high-frequency oscillations (HFO) are becoming the most promising ones. This study aims to investigate how different types of HFOs are related to Ictogenesis through developing a mathematical framework such as feature representation learning algorithms. And, to manifest the existence of novel predictive/distinct features that are hidden but essential biomarkers useful for accurate delineation of SOZ. The proposed machine learning-based extraction of HFO signatures will help physicians 1) to indicate accurate localization of epileptogenic zone in pre-surgical evaluations and, 2) to better understand which epileptic HFO discharges are highly relevant to ictal onset zone and lead to a better postsurgical outcome.