Novel non-invasive heart monitoring
Cardiovascular disease (CVD) is a leading cause of mortality in developed countries. It has been estimated that up to 90% of CVD could be prevented by early prevention and diagnosis of the disease.
Using accelerometers for acquiring information on subject’s heart signal has been known for a long time, but only recently been spotted as having a high potential for diagnostic value in cardiac settings, due to arise of new miniaturized MEMS devices. Since ECG (which is the gold standard for non-invasive diagnosis of heart diseases) measures the electrical commands received by the heart, the SCG reveals the mechanical response of the heart beating, which is complementary to ECG. The utilization of gyroscope (for example within smartphone IMU, Inertial Measurement Unit) for the analysis of heart signal has been proposed by our research team. As gyroscope measures tiny rotational vibrations of the body (we denote this approach as gyrocardiography, GCG), it can be used as a complementary cue to the SCG for heart analysis. Joint SCG / GCG analysis can be used to give a better signal representation (than SCG only) in smartphone cardiography. Ballistocardiography (BCG), which is a well-known method, measures the pressure changes induced by blood flow throughout the body with sensors installed, for example, under the bed. The above mentioned non-invasive heart monitoring methods can currently be seen as promising ways for early prevention of heart diseases.
Key words:
Non-invasive heart monitoring, Cardiovascular disease, Signal processing, AI, MEMS, Wearable electronics, Digital biomarkers, Diagnostic innovation
Utilization of MEMS accelerometers and gyroscopes and BCG in measuring the mechanical activity of the heart has been and is currently investigated in several of our research projects. During the past 10 years main application areas have been:
- Atrial fibrillation (including randomized studies for detection and follow-up)
- Heart failure (including also CRT investigations)
- ST-elevation infarction
- Aortic stenosis (including randomized studies)
Internet of things monitoring of health in pregnancy
Overweight and obesity in women of reproductive age is a global health problem. Obesity and excessive gestational weight gain (GWG) have several negative consequences for both mother and child. Pregnant women with obesity are at increased risk of antenatal and postnatal depression, gestational diabetes mellitus, and hypertensive disorders. Obesity and excessive GWG are also significant risks for the child. Both independently increase the risk of infant obesity, macrosomia, birth complications and glucose, insulin and cardiometabolic dysregulation in the offspring.
The research group has developed an everyday IoT monitoring system that can detect and predict physiological and psychological health problems in obese women during the perinatal period (13 weeks of gestation to 12 weeks postpartum). The system follows a cybernetic approach to health, integrating lifestyle and environmental data and bio-signals with medical knowledge to develop a closed-loop health system that optimizes maternity care.
Using this system, the group has collected continuous 9-month data on overweight or obese peripartum women (n=54), e.g. on depressive symptoms, stress, pregnancy-related anxiety, weight gain, blood pressure, heart rate, heart rate variability, physical activity, sleep and dietary intake. The research will be centered on ubiquitous pregnancy monitoring services, enabling related pregnancy complications detection, prediction, and assessment through the development of data analytics and machine learning algorithms.
The candidate can work with data processing algorithms such as data fusion and data abstraction methods at the sensor node to exploit higher-level information from collected PPG data (optical measurement of arterial volume) and activity data. For example, employing bio-signal processing techniques to extract stress levels, subjectively estimated via fusing PPG-driven respiration rate and heart rate variability values. Moreover, personalized multi-modal machine learning-based models will be built using daily behavior, lifestyle, and special conditions of the mothers extracted from sensory data and the person’s history for cardiometabolic complications detection. Finally, exploit time-series-based explainable artificial intelligence (AI) approaches to investigate the association between observations (e.g., physiology, behavior, etc.) and the predictions in a person-centric way.