SPART-real time spatial analysis of ECG signal

Abstract

MethodsPresented SPART method was developed for real time computation the preferential plane of VCG spatial loops. The space loop transformation is done using modifed Karhunen-Loeve Transform (KLT). The major loop (ML) are result of such transform. Algorithm execute morphology analysis and adaptive clustering using Fourier Descriptors (FDs) of ML. Respiration signal was derived based on the calculations of spatial electrical axes of ML. SPART method estimate the spatial beat-tobent parameters variability and decrease the respirationinduced injluence.Preliminaly testing confrm utility of SPART method to real-time analysis of VCG signal during electrophysiology investigations.The low noise signal from three bipolar orthogonal X Y Z leads is digitized at 1000 samples/sec with 12-bit resolution. The adaptive QRS detection and delineation of the reference detection point are performed similar to method proposed by Wolf [lo]. We use detection algorithm based on spatial velocity computed from the three filtered lead signals. Detection point estimate the R wave peak.For each QRS complex we extract baseline-corrected 200 samples (200 ms) data segment, with detection point at the 80th sample. Let hT = [x, y, z] be a beat segment data with mean mh . Covariance matrix was computed as IntroductionElectrical activity of the heart has a spatial character. The vectorcardiogram (VCG) is graphical presentation of spatial vector rotation [5]. Mechanical action of the respiration cause the rotation of cardiac vector and respiratory sinus arrhythmia [6]. This phenomenon may introduce artefacts during ECG parameters measurement 1131.The main target of spatial VCG loops analysis is to found their own preferential plane [ 5 ] . Previous procedures of VCG spatial rotation has a manual character [SI. To algorithmic solution one can use a Karhunen-Loeve Transform (KLT) 181. Many KLT based procedures of signal compression and morphology representation lost a part of input signal [SI. Real-time ECG analysis is highly desirable in some applications such monitoring during exercise tests or HRV analysis during tilt tests [lo, 131.In this study we present an automatic method SPART to estimate the spatial beat-to-beat variability with QRS loops clustering [2]. SPART method provide algorithm computation of X Y Z space to preferential UVW space transformation. 0276-6547195 $4.00 0 1995 IEEE 357Let bT = [U, v, w] be a beat segment data after transformation. The KLT transformation is defined aswhere A is a transformation matrix whose rows are the normalized eigenvectors 4; of matrix C, for k = 1, 2, 3, corresponding to the eigenvalues hk. Eigenvectors derived from loop signals are the basis vectors, i.e. these vectors point in the direction of maximum variance of the data. It is possible to rotate the loop so that it is aligned to some standard or invariant direction. KLT transformation result may be reflected about the origin in the any axis [4].We developed spatial loops orientation matching method for sorting and arranged Q so there was direct rel...

Automatic rhythm analysis using SPART method

“…The results estimated for X Z and W planes were comparable. All three methods mentioned in introduction: Young and Huggins [4], CAVIAR [20] and SPART [3] use mapping spatial loops into preferential space. However these methods use different techniques for minimizing the norm.…”

“…In the first stage SPART method provides transformation algorithm from standard X Y Z space to preferential UVW space [3]. This transformation is linear and losses.…”

“…Spatial transformation for template beat loop is done by previously reported method [3]. For beat-to-beat matching of the next loops we developed the new modified SLOM method.…”

“…This phenomenon may introduce artifacts during measurements of ECG parameters [3]. On the other hand, the cardiac activity can be adequately represented by planar vector loop in the preferential plane [17].…”

“…In this study we present a modified SPART method for automatic estimation of the spatial beat-to-beat variability [3]. The SPART method provides algorithm computation of XYZ space to the preferential UVW space transformation and QRS loops clustering based on FD's.…”