| Popis: |
Non-invasive pulse oximeters are used by nurses to monitor oxygenation in $>$600 NICU's in the USA for $>$80% of mechanically ventilated infants. Motion causes false alarms up to 29% of the time, resulting in poor specificity. I developed a fuzzy logic computer prototype alarm system using fuzzy sets of desaturation, within limits, and artifact. The purpose of this study was to compare the new (NEW) vs. conventional (OLD) alarm systems to test the hypothesis that a neonatal pulse oximeter alarm system based on fuzzy logic will have equivalent sensitivity and improved specificity vs. the OLD alarm system. Thirty-eight infants were enrolled with a mean (range) study weight of 1495g (470-3390). Oxygen saturation signals (Nellcor N200) were collected for 1 hour per infant on a strip chart recorder and saved on a computer. Reference (REF) signals were simultaneously recorded from a second pulse oximeter and transcutaneous O$\sb2$ monitor. If an alarm persisted, each 30 second interval was considered a separate event. The NEW vs. OLD alarms were compared to the REF for 919 alarm events. There were 451 new alarm events, with a mean of 12.1 (range 1-36) new events per subject, and the others were persistent alarms. Oxygen desaturation was defined as transcutaneous $\rm PO\sb2 |
