Athanasios Tsanas (‘Thanasis’)





































My research is multi-disciplinary, encompassing fields such as biomedical engineering, signal processing, machine learning, applied mathematics, and communications. Some of the projects I am actively involved include:

Longitudinal monitoring of mental health using self-assessments and objective sensor-based data

We monitor longitudinally (>1 year) patient groups and healthy control participants who self-assess their mood states on a range of validated clinical questionnaires. Moreover, we record a range of objective data including activity, sleep, relative geolocation, communication practices, and heart rate. The aim is to develop robust biomarkers which may be used as clinical precursors for detecting clinical symptoms.


Sleep structure architecture and characterisation of mental and neurological disorders

We automatically assess the sleep structure (stages, arousals, sleep spindles) using a range of collected raw signals such as electroencephalogram (EEG), and electrocardiogram (ECG). The aim is to determine changes in sleep patterns of mental and neurodegenerative subjects compared to age- and gender-matched healthy controls, and also to determine sleep patterns which might be linked to symptom severity. Moreover, we aim to differentiate neurological disorders using patterns obtained from signals recorded during sleep.


Telemonitoring of Parkinson's disease using nonlinear speech signal processing

We have developed a novel way of estimating the average progression of Parkinson's disease by a useful clinical metric, called UPDRS, using solely speech signals. Speech is ideal for telemonitoring systems, which require simple, self-administered tests. Moreover, we have reported about 98.5% accuracy in differentiating people with Parkinson’s disease from healthy controls. We are now expanding these findings aiming to link them to the underlying physiological understanding of voice production mechanisms.


Feature selection

In many applications we can collect a large number of characteristics (features) which we believe might be predictive of the investigated system. One practical example is using a large number of genes which we believe are somehow related to an outcome, e.g. whether someone has cancer. My work tries to identify the feature set which may be most predictive of the outcome of interest.


Cardiovascular system modelling and analysis with a view to mechanical circulatory support devices

We develop a model based on physiological principles which explains the changes brought to the circulatory system with patient specific parameters (age, gender etc.), as well as in cases of exercise, disease, cardiac arrest etc. Ultimately, I would like my research to help in the development of algorithms for monitoring patients and for mechanical circulatory support devices (of which the artificial heart is the most well-known, but there is a plethora of other systems).


© Athanasios Tsanas

Last updated: 23 November 2016