Data-driven assessment of cardiovascular ageing through multisite photoplethysmography and electrocardiography.

The cardiovascular system is designed to distribute a steady flow through its elastic properties. With ageing, fatigue and fracture of elastin lamellae cause a loss of elasticity defined arterial stiffness. Arterial stiffness causes changes of the pulse wave propagation through the arterial tree, which volumetric counterpart can be assessed non-invasively through photoplethysmography (PPG). PPG may be employed in combination with electrocardiography (ECG). It is here reported an implementation of analysis of multisite PPG and single lead ECG relying on Deep Convolutional Neural Networks (DCNNs). DCNNs generate peculiar filters allowing for data-driven automated selection of the features of interest. The ability of a DCNN to predict subject's age from PPG (left and right brachial, radial and tibial arteries plus fingers) and ECG (Lead I) in a healthy male population (age range: 20-70 years) was investigated. A performance in age prediction of 7 years of root mean square error was obtained, which was superior to other feature-based procedures. The accuracy in age prediction of the machinery in the healthy population may serve for the generation of age-matched normal ranges for the identification of outliers suggesting cardiovascular diseases manifesting as fastened cardiovascular ageing which is recognized as a risk factor for ischemic diseases.