Can Arterial Blood Pressure Predict Age? A ConvNet Classification Task

Abdalrhman Mostafa, Nour-Mounira Z. Bakkar, Mohamed Abdelhack, Ahmed F. El-Yazbi

Keywords: ConvNet, Spectrogram analysis, Blood Pressure Variability, Time series analysis

TL;DR: We use a ConvNet to predict age from high resolution blood pressure data.

Abstract: Blood pressure (BP) increases throughout life, and is controlled by several feedback mechanisms in mammals. Therefore, high resolution BP data may contain information related to the health and functionality of those systems, and the organism as a whole. We used beat-to-beat BP data at different sampling rates collected from a heterogeneous population of rats to predict their age, and achieved 90%+ test accuracy, AUROC, and AUPRC across all sampling rates. Higher frequency components appeared to contribute less to 100 Hz sample rate BP data classification, but nonetheless could point to previously unknown Spectro-temporal patterns that the model identified.