Abstract
The ARGUS/H system for Holter-tape analysis has been modified to reduce the time required for human editing and to improve the true-PVC detection rate without increasing the false-positive PVC rate. (PVC equals Premature Ventricular Contraction). The strategies employ increased use of context for more reliable waveform recognition by the computer and ″machine learning″ through editor interaction. The algorithm changes and additions have been made at or following the last processing stage (CYCLE) in order to capitalize on information already available from the proven system. Evaluation results show that editing times have been reduced by a factor of at least four, 99. 8% of all false-positive PVCs due to T-waves are automatically corrected, and the number of missed PVCs has been halved. Most (88. 2%) of the remaining missed PVCs are due to excessive signal noise. Further efforts to improve system performance are now addressed to two-channel processing.
Original language | English |
---|---|
Pages | 7-13 |
Number of pages | 7 |
State | Published - 1975 |
Event | Unknown conference - Rotterdam, Neth Duration: Oct 2 1975 → Oct 4 1975 |
Conference
Conference | Unknown conference |
---|---|
City | Rotterdam, Neth |
Period | 10/2/75 → 10/4/75 |