back to homepage


The Brain Computer Interface (BCI) is a direct communication pathway between the brain and the computer without stimulating any muscular activity. The applications of BCI technologies are in various fields, i.e., in sensorimotor rehabilitation for people with locked-in syndrome, in lie detection, in mode assessments for neuroergonomics perspectives or in brain fingerprinting etc. There are mainly two types of BCI according to the invasiveness of signal acquisition techniques from the brain. Invasive techniques include electrocorticography (ECoG), micro electrode array, etc., while noninvasive techniques include electroencephalography (EEG), magnetoencephalography (MEG), functional magnetic resonance imaging (fMRI), functional near infrared spectroscopy (fNIRS) and functional transcranial Doppler ultrasonography (fTCD) etc. In particularly, the EEG based BCIs are very popular due to its cost efficiency, mobility and noninvasive signal acquisition process. The spatio-temporal resolution of EEG carries useful information for decoding cortical events. But, there are some critical challenges like high computational latencies due to multichannel EEG recordings, inter-session and inter-subject variabilities in EEG dynamics, etc., oppose the enhancement of existing BCI systems. Our research goal is to investigate on these challenges and to figure out solutions.

Diabetes is a chronic disease and often causes cardiovascular abnormalities like Cardiac Autonomic Neuropathy (CAN) after prolonged existence. It is crucial to diagnose physiological changes due to diabetes, so to minimize unfavorable effects of diabetes. Heart Rate Variability (HRV) measurements from electrocardiogram (ECG) recordings sometimes provide meaningful biomarkers to predict potential cardiovascular risk factors. We aim to investigate on linear and nonlinear features extracted from consecutive R-R intervals of healthy and diabetes cohorts.

Major Depressive Disorder (MDD) is an emerging risk factor for human health by contributing to the diseases like cognitive impairments, cardiovascular risks etc. The primary challenge for treating depression is to diagnose the depression severity. Traditional methods to identify depression severity based on questionnaire based methods, i.e., Hamilton Depression Scale, Beck Depression Inventory, etc., are not always effective. However, based on the patients’ response, they are categorized in different risk levels (low, moderate or high). Now-a-days, different signal & image processing approaches are being used to establish suitable biomarker(s) from various biomedical signals for diagnosing MDD. In such cases, signal can be captured from the brain using electroencephalography (EEG) or using functional magnetic resonance imaging (fMRI), from the chest using electrocardiogram (ECG), from the skin using photoplethysmography (PPG) etc. In addition, different machine learning approaches are being applied in EEG signal to predict MDD patients’ response to selective serotonin reuptake inhibitors (SSRIs) treatment while the fMRI technique is being used to monitor brain functions during major depression. In the meantime, the ECG is being used as to identify the changes of arterial stiffness in MDD patients before/after treating with antidepressants. During diagnosing MDD using ECG and PPG, features like RR Intervals, Pulse Transit Time (PTT), Pulse Wave Velocity (PWV), Systolic Volume, Diastolic Volume, Pulse Wave Amplitude (PWA), etc., are measured as potential indicators of MDD patients’ states. In our study, we analyze simultaneously recorded ECG/PPG signals from both MDD subjects and healthy subjects. Additionally, we aim to predict suicidal ideation within MDD cohorts using ECG/PPG features.