Name
Abstract | ||
|---|---|---|
High rates of false monitoring alarms in intensive care can desensitize staff and therefore pose a significant risk to patient safety. Like other critical arrhythmia alarms, asystole alarms require immediate attention by the care providers as a true asystole event can be acutely life threatening. Here, it is illustrated that most false asystole alarms can be attributed to poor signal quality, and we propose and evaluate an algorithm to identify data windows of poor signal quality and thereby help suppress false asystole alarms. The algorithm combines intuitive signal-quality features (degree of signal saturation and baseline wander) and information from other physiological signals that might be available. Algorithm training and testing was performed on the MIMIC II and 2015 PhysioNet/Computing in Cardiology Challenge databases, respectively. The algorithm achieved an alarm specificity of 81.0% and sensitivity of 95.4%, missing only one out of 22 true asystole alarms. On a separate neonatal data set, the algorithm was able to reject 89.7% (890 out of 992) of false asystole alarms while keeping all 22 true events. The results show that the false asystole alarm rate can be significantly reduced through basic signal quality evaluation. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1109/EMBC.2017.8037313 | 2017 39TH ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY (EMBC) |
Field | DocType | Volume |
Signal quality,ALARM,Computer science,Asystole,Medical emergency,Intensive care | Conference | 2017 |
ISSN | Citations | PageRank |
1094-687X | 0 | 0.34 |
References | Authors | |
2 | 2 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Remi Dekimpe | 1 | 0 | 0.34 |
| Thomas Heldt | 2 | 2 | 9.54 |