Cardiovascular diseases (CVDs) are the number one cause of death globally. An estimated 17.9 million people died from CVDs in 2016, representing 31% of all global deaths. Therefore, many products that monitors patient’s heart conditions have been introduced on the market such as: Apple Watch, QardioMD, and Astroskin. Most of these powerful devices can record patient’s biometric signals at rest. Devices such as Atroskin for example, can measure 3 lead electrocardiogram (ECG), breathing (upper and lower chest), body temperature, blood pressure, and acceleration. The main goal of this thesis is to develop a cardiac monitoring system that can detect heart anomalies continuously during normal activities such as resting, walking, cycling, running. This patient-specific monitoring systems starts by learning the patient’s ECG in normal cardiac state creating a patient specific neural model. As data get acquired and compensated from artifacts created by body motion, it can then detect in real-time departures form normality that are only due to cardiac anomalies. In order to do so, the system use a LSTM neural network that combines ECG with accelerometer data to detect departure from the normal patient spe-cific model. By fusing this information, one can reduce false positive created by motion artifacts and allow true continuous heart anomaly detection.

For more information on this project please communicate with Hong Zu Li