Systolic Blood Pressure Estimation from PPG Signal Using ANN

High blood pressure is one of the most important precursors for Cardiovascular Diseases (CVDs), common cause death in 2020, as reported by World Health Organization (WHO). Moreover, many patients affected neurodegenerative diseases (e.g., Parkinson’s Disease) exhibit impaired autonomic control, with inversion normal circadian arterial cycle, and consequent augmented cardiovascular fall risk. For all these reasons, a continuous monitoring could represent significant prognostic factor, help adjusting their therapy. However, existing cuff-based methods cannot provide readings. Our work inspired newest approaches based on photoplethysmographic (PPG) signal only, which has been used to continuously estimate systolic (SP), using artificial neural networks (ANN), order create more compact wearable devices. first database was derived from PhysioNet resource; we extracted PPG (ABP) signals, collected at sampling frequency 125 Hz, hospital environment. It consists 249,672 periods relative SP values. The second STMicroelectronics s.r.l., Agrate Brianza, MORFEA3 device digital sphygmomanometer, reference. pre-processing phase, remove noise motion artifacts segment into periods, carried out Matlab R2019b. removal challenging parts study because inaccuracy during everyday-life activity, this reason why dataset acquired controlled environment static position. Different solutions were implemented choose input features that best period morphology. train multilayer feed-forward network back-propagation model, whereas test it. results obtained project are promising match Association Advancement Medical Instruments (AAMI) British Hypertension Society (BHS) standards. They show Mean Absolute Error 3.85 mmHg Standard Deviation 4.29 mmHg, under AAMI standard, reach grade A BHS standard.