Browse Prior Art Database

Detection and Structured Representation of Waveforms (Qrs Complexes) in Electrocardiogram Records

IP.com Disclosure Number: IPCOM000059715D
Original Publication Date: 1986-Jan-01
Included in the Prior Art Database: 2005-Mar-08
Document File: 2 page(s) / 14K

Publishing Venue

IBM

Related People

Miller, JM: AUTHOR

Abstract

This article describes a technique which solves the problem of rapidly detecting, locating and characterizing the shape of heartbeat waveforms (QRS complexes) in long electrocardiogram records. Long electrocardiogram (ECG) records contain large amounts of digitized data which must be processed either manually or by computer to detect heartbeat waveforms, and classify them as normal or abnormal. This article describes a method for detecting the waveforms (QRS complexes) and representing them in a way which permits rapid classification in subsequent processing. The structured representation of the waveform is obtained for each ECG channel by the following steps: 1. Any of several simple common methods of smoothing or filtering are used to remove some of the high-frequency components of the original data.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Detection and Structured Representation of Waveforms (Qrs Complexes) in Electrocardiogram Records

This article describes a technique which solves the problem of rapidly detecting, locating and characterizing the shape of heartbeat waveforms (QRS complexes) in long electrocardiogram records. Long electrocardiogram (ECG) records contain large amounts of digitized data which must be processed either manually or by computer to detect heartbeat waveforms, and classify them as normal or abnormal. This article describes a method for detecting the waveforms (QRS complexes) and representing them in a way which permits rapid classification in subsequent processing. The structured representation of the waveform is obtained for each ECG channel by the following steps: 1. Any of several simple common methods of smoothing or filtering are used to remove some of the high- frequency components of the original data. Typically, this is digitized data sampled at rates of 60 to 600 points/second. 2. First differences are obtained as follows: Data point i-x is subtracted from data point i, with i incremented by y, to obtain succeeding differences. The following conditions are imposed: y > x and x > = 1 Examples: point i-1 may be subtracted from point i, with i incremented by 2; point i-2 may be subtracted from point i, with i incremented by 3; point i-3 may be subtracted from point i, with i incremented by 4, etc. x and y should be chosen to maximize processing speed without degrading performance. The choice will depend, in part, on the sampling rate of the ECG data. 3. The preceding step results in runs of positive and runs of negative numbers. Only the sequence of extreme values (largest positive, smallest negative) above a threshold value is retained for further processing. The threshold values are KPOS/8 for positive numbers and KNEG/8 for negative numbers (/ = divided by). KPOS and KNEG are updated using a moving average of the elements associated with each QRS complex detected in step 4 below. Initial values are KNEG = KPOS = 0. 4. The sequence of first difference extreme values obtained from the preceding step is examined to find sets of 2 or 3 elements to represent the QRS complex. These are numbers which: . exceed the threshold test values KPOS/4 or KNEG/4, . have one of the sign patterns: -+, +-, -+-, +-+, and . are larger in absolute magnitude than other numbers of like sign within the interval s...