Resumen

TAPIA C., Gonzalo  y  GLARIA B., Antonio. ARTIFICIAL NEURAL NETWORK DETECTS PHYSICAL STRESS FROM ARTERIAL PULSE WAVE. Rev. ing. biomed. [online]. 2015, vol.9, n.17, pp.21-34. ISSN 1909-9762.

The main goal of this work is to study the initial technical feasibility of detecting physical stress caused by exercise associated with episodes of rising Blood Pressure (BP) by means of analyzing Pulse Wave (PW), in order to reduce intrusiveness resulting from the current use of non-invasive BP Monitors. Lead I Electrocardiogram (EKG) and right index finger and main toe Arterial Pulse Waves (PW) were recorded on healthy volunteers, Before and After Exercise (BAE). Trained Artificial Neural Networks (ANNs) were used for stress detection. A common training set was used for different ANN. PW Phase Planes BAE, vectorized and heartbeat segmented, were used as input vectors, while rest or stress condition BAE were used as target vectors. Pan-Tomkins algorithm was applied to EKG for PW segmentation. A digital polygraph was used to register the signals. Thirteen university students, 2 females and 11 males (24.3 ± 2.83 years old), participated as healthy volunteers. They usually carried out recreational sports. Their BP raised (43.4 ± 18,6)/(12.7 ± 12,0) mmHg after physical exercises. Stress condition detection (n = 200) reached up to100% on True Positives and 12% in False Positives. Results are promising to continue improving the methodology. Its development should contribute to the detection, control and monitoring of Arterial Hypertension.

Palabras clave : Arterial Hypertension; Artificial Neural Networks; Pulse Waves; Minimally-Intrusive.

        · resumen en Español | Portugués     · texto en Español     · Español ( pdf )