Statistical processing for gastric slow-wave identification

Successful identification of gastric slow waves in canine gastric electrical activity (GEA) data was achieved using a statistical data-processing procedure based on the multiple linear regression (MLR) curve fitting technique. Both distal and proximal waveforms were identified, first by construction...

Full description

Saved in:
Bibliographic Details
Published in:Medical & biological engineering & computing 2002-07, Vol.40 (4), p.432-438
Main Authors: GRANT, M. S, WILLIAMS, R. D
Format: Article
Language:English
Subjects:
Algorithms
Animals
Biological and medical sciences
Computation
Dogs
Electrodiagnosis
Gastrointestinal Motility
Linear Models
Mathematical models
Matrices
Medical
Medical sciences
Muscular system
Regression
Signal Processing, Computer-Assisted
Slow waves
Stomach - physiopathology
Stomach Diseases - diagnosis
Thresholds
Waveforms
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Successful identification of gastric slow waves in canine gastric electrical activity (GEA) data was achieved using a statistical data-processing procedure based on the multiple linear regression (MLR) curve fitting technique. Both distal and proximal waveforms were identified, first by construction of separate orthonormal bases from pre-selected sets of representative distal and proximal gastric slow waves (GSWs). Respective basis matrices were used to fit proximal and distal data to an MLR data model. Residual waveforms were computed from the original and 'fitted' waveforms and used in identifying GSWs in the data. Canine GEA data were split into 1,800-point blocks, and each 245-point data segment in a block was processed to identify the GSWs. Gastric slow waves were located in the data using a residual mean-squared error (MSE) threshold and, for distal GEA data, the minimum value of the main distal waveform peak. All threshold values were determined empirically and were set to detect GSWs while limiting false matches. Identification rates of 95% and 99% for proximal and distal GSWs, respectively, represent a significant improvement over those obtained in a previous study in which the same data were analysed using linear signal-processing methods. The use of the method presented in this paper for real-time identification of GSWs in conjunction with an implantable gastric pacer unit appears promising. Because the technique is inherently customisable, results obtained in this study should also be applicable to human subjects.
ISSN:0140-0118
1741-0444
DOI:10.1007/BF02345076