Title :
The next-generation holter revolution: from analyse-edit-print to analyse-print
Author :
Pardey, J. ; Jouravleva, S.
Author_Institution :
Oxford Instrum. Med. Ltd., UK
Abstract :
A new three-channel Holter analysis algorithm called Medilog ADAPT is presented and its modular design is described. The use of three-channel analysis improves the accuracy and robustness of the algorithm beyond that of two-channel algorithms and supports the direct printing of reports from unedited analysis results. It also enables Medilog ADAPT to continue working even during periods when two channels are unusable due to signal drop-out or noise corruption. Beat classification uses neural network estimators of Bayesian a posteriori probabilities. A new Holter system running Medilog ADAPT has been performance tested according to ANSI/AAMI standards EC38.1998 and EC57:1998. Gross statistics for QRS Se, QRS +P, VEB Se and VEB +P are 99.9%, 99.9%, 98.4% and 97.6% respectively for the 78 nonpaced ECGs in the AHA database and 99.9%, 99.9%, 97.5% and 93.6% respectively for the 44 nonpaced ECGs in the MIT-BIH database.
Keywords :
Bayes methods; electrocardiography; medical computing; neural nets; signal classification; AHA database; Bayesian a posteriori probability; ECG; MIT-BIH database; Medilog ADAPT; QRS detection module; VEB detection module; beat classification; gross statistics; neural network estimators; noise corruption; robustness; signal drop-out; three-channel Holter analysis algorithm; ANSI standards; Algorithm design and analysis; Bayesian methods; Databases; Electrocardiography; Neural networks; Noise robustness; Printing; Probability; System testing;
Conference_Titel :
Computers in Cardiology, 2004
Print_ISBN :
0-7803-8927-1
DOI :
10.1109/CIC.2004.1442950