Customization of Entropy Estimation Measures for Human Arterial Hypertension Records Segmentation E.M. Cirugeda–Rold´ an 1 , A. Molina–Pic´ o 2 , D. Cuesta–Frau 3 , S. Oltra–Crespo 4 , P. Mir´ o–Mart´ ınez 5 , L. Vigil–Medina 6 , M.Varela–Entrecanales 7 Abstract— This paper describes a new application of the recently developed Coefficient of Sample Entropy (CosEn) measure. This entropy estimator is specially suited for cases where the length of the time series is extremely short. CosEn has already been used successfully to characterize and detect atrial fibrillation, using as few as 12 heartbeats. We have customized the methodology employed for heartbeat interval series to blood pressure hypertensive (BPHT) human records. Little can be found about BPHT records and its nonlinear regularity analysis. The method described in this paper provides a good segmentation between control and pathologic groups, based on the corresponding labeled BPHT records. The experimental dataset was drawn from the available records at the Hypertension Unit of the University Hospital of Mostoles, in Spain. The hypertension related variables studied were systolic blood pressure (SBP), diastolic blood pressure (DBP) and mean blood pressure (MBP). The hypothesis test yielded the following results in each case: acceptance probability of 0 for SBP, 0.005 for DBP and 0 for MBP. The confidence intervals for the three variables were nonoverlapping. I. INTRODUCTION Biological systems can be considered a manifestation of complex and nonlinear processes. Such systems exhibit not only the readily observable stationary or periodic behavior, but they also usually have a nonpredictable, chaotic, nonlin- ear or nonstationary behavior [1]. Nonlinear methods based on entropy computations or data complexity statistics have become very popular recently in applications related to the analysis of biological signals due to their good results. Their capability to unveil hidden nonlinear information embedded in records has proven very powerful in signal class segmenta- tion applications. Classical linear methods lack of robustness or characterization depth in most of these cases [2], [3]. 1 E.M. Cirugeda–Rold´ an and A. Molina–Pic´ o are PhD students of the Computer Science Department (DISCA) at Polytechnic University of Valencia, Alcoy Campus (EPSA-UPV), 03801 Alcoy, Alicante, Spain [ecirugeda,anmopi]@giica.com 3 D. Cuesta–Frau is with the Computer Science Department (DISCA) at Polytechnic University of Valencia, Alcoy Campus (EPSA-UPV) 03801 Alcoy, Alicante, Spain dcuesta@disca.upv.es 4 S. Oltra–Crespo is with the Mathematics Department at Polytechnic University of Valencia, Alcoy Campus (EPSA-UPV) 03801 Alcoy, Alicante, Spain soltra@mat.upv.es 5 P. Mir´ o–Mart´ ınez is with the Statistics Department at Polytechnic University of Valencia, Alcoy Campus (EPSA-UPV) 03801 Alcoy, Alicante, Spain pamimar@eio.upv.es 6 L. Vigil–Medina is with the Hypertension Unit of Internal Medicine Service at the University Hospital of M´ ostoles 28935 M´ ostoles, Madrid, Spain lvigil.hmtl@salud.madrid.org 7 M. Varela–Entrecanales is with the Internal Medicine Service at the University Hospital of M´ ostoles 28935 M´ ostoles, Madrid, Spain mvarela.hmtl@salud.madrid.org Blood pressure (BP) is considered to be a key parameter when evaluating the cardiovascular control system of a pa- tient since essential hypertension (HT) is considered to be a trigger of a variety of mayor cardiovascular diseases, such as cerebral stroke or myocardial infarct [4]. BP has been widely characterized by traditional, linear methods, which assume a certain degree of stationarity. On the contrary, little can be found about BP studies with nonlinear entropy methods [1], [5], [6]. Most of this few studies are based on animal blood pressure hypertensive (BPHT) records. Additionally, usual nonlinear methods employed in these cases are correlation dimension (CD) [1], Lempel–Ziv (LZ) [5] and detrended fluctuation analysis (DFA) [6]. However, due to the specific features of BP records, these metrics do not properly fit to this BP analysis task since a large number of samples are needed in order to obtain a good entropy estimation [1]. Most of them require a number of samples in the order of several hundreds or thousands, whereas a long–term BP record may contain some 120 samples at most. In this paper, our interest is focused on human BPHT records. These data series are usually noninvasively recorded by means of a digital sphygmomanometer. This technique termed sphygmomanometry is known to be the most accurate and noninvasive method for BP data acquisition, although it is quite uncomfortable for the patient. A cuff surrounding the arm should previously be inflated until its pressure is higher than the Systolic Blood Pressure (SBP), and then deinflated so as to take a measure. During the data acquisition it is convenient that the patient remains still in a steady state, such as sitting, relaxed, with the arm straight and immobile [4]. Owing to such uncomfortability and constraints, long or continuous BPHT records are not usually possible. There- fore, in order to enable a nonlinear analysis of such records, a more robust entropy measure is needed. An increased BP variability in the different ways it can be recorded (ambulatory or home) implies a worse prognosis in several studies. However, as far as we know, a complexity analysis of these arterial BP measures and its correlation with a clinical prognosis has not been carried out yet. Our previous research has proved that there is a progres- sive loss of complexity from a normality state to illness in thermo–regulation [7] and in glucoregulation [8], and such loss entails a worse prognosis. In other works, it has been observed that there is an inverse correlation between variability and complexity, and probably both phenomena are manifestations of the same deterioration process of the fine control physiological systems. However, a complexity 34th Annual International Conference of the IEEE EMBS San Diego, California USA, 28 August - 1 September, 2012 33 978-1-4577-1787-1/12/$26.00 ©2012 IEEE