Comparison of four methods to calibrate respiratory inductive plethysmograph in premature infants
- 646 Downloads
KeywordsPremature Infant High Weight Calibration Method Reference Period Face Mask
Respiratory inductive plethysmography (RIP) is a widely used method providing important information on the breathing pattern. This non-invasive technique is particularly interesting in infants and newborns. However, the different calibration methods developed in adults to allow the measurement of volume have not been validated in premature infants, even if they are frequently used. We developed a calibration method using the least square method to identify the coefficients (α ; β) to calculate the reconstructed volume from the RIP signals (V = α ribcage + β abdomen) with a pneumotachograph (PNT) signal as reference, and using an individually adapted filter to suppress heart artefacts.
To compare the accuracy of this calibration method with three frequently used calibration methods: a simple fixed-coefficients method with either equal weight attributed to the abdomen and rib cage (method 1:1) or with a higher weight attributed to the abdomen (i.e. β = 2α ; method 1:2); and an equivalent of the qualitative diagnostic calibration method, using coefficients proportional to the standard deviation of abdomen and ribcage signals (β/α = SD(abdomen)/SD(ribcage)) (method SD).
Twelve premature infants breathing spontaneously were studied. Birth weight was (mean ± SD) 1528 ± 340 g; gestational age 31.2 ± 0.6 weeks; age 6 ± 2 days. A specially adapted RIP jacket was installed 15 min before the study. RIP signals were recorded continuously. A face mask connected to a PNT was applied during two 30-s periods. The most regular 15 PNT cycles were identified and served as reference period, during which the coefficients (α ; β) were calculated with each method. The four reconstructed flows were then recalculated, and we estimated the distance R2 between reconstructed flows and the PNT reference signal for the entire period of PNT; with R2 = 1 – (mean square difference between the normalized flows / variance of PNT); that is, a R2 value closer to 1 represents a better reconstruction of the signal.
The mean distance R2 was 0.62 ± 0.22 with method 1:1, 0.63 ± 0.22 with method 1:2, 0.62 ± 0.23 with method SD, and 0.70 ± 0.17 with the least square method (P = 0.02; repeated-measure ANOVA with a significant difference between the last method and the three others). The ratios β/α were 2.1 ± 1.5 and 2.9 ± 4.3 with the two last methods, respectively.
This new RIP calibration method using the least square method with a short period of PNT as reference and an individually calculated filter gives a better reconstructed flow as compared with three other calibration methods. This method attributes a higher weight to the abdominal signal.