A high-speed C++/MEX solution for long-duration arterial blood pressure characteristic locations detection

The major concentration of this study is to describe the structure of a C++/MEX solution for robust detection and delineation of arterial blood pressure (ABP) signal events. Toward this objective, the original ABP signal was pre-processed by application of à trous discrete wavelet transform (DWT) to extract several dyadic scales. Then, a sliding window with fixed length was moved on the appropriately selected scale. In each slid, mean, variance, Skewness and Kurtosis values of the excerpted segment were superimposed to generate a newly defined multiple higher order moments (MHOM) metric to be used as the detection decision statistic (DS). Then, after application of an adaptive-nonlinear transformation for making the DS baseline static, the histogram parameters of the enhanced DS were used for regulation of the α-level Neyman–Pearson classifier aimed for false alarm probability (FAP)-bounded delineation of the ABP events. The proposed method was applied to all 18 subjects of the MIT-BIH Polysomnographic Database (359,000 beats). The end-systolic and end-diastolic locations of the ABP signal as well as the dicrotic notch pressure were extracted and values of sensitivity Se = 99.86% and positive predictivity P+ = 99.95% were obtained for the detection of all ABP events. This paper proves the proposed MHOM-based ABP events detection–delineation algorithm as an improvement because of its merits such as: high robustness against measurement noises, acceptable detection–delineation accuracy of the ABP events in the presence of severe heart valvular, arrhythmic dysfunctions within a tolerable computational burden (processing time) and having no parameters dependency on the acquisition sampling frequency.

► The proposed method was applied to 359,000 beats of the PhysioNet Database. ► All characteristic locations of the ABP signal and dicrotic notch pressure were detected. ► Values of sensitivity Se = 99.86% and positive predictivity P+ = 99.95% were achieved. ► The proposed ABP events detection algorithm is robust against measurement noises. ► Also, accuracy of method is acceptable within a tolerable processing time with no dependency to acquisition system sampling frequency.