Acoustic Analysis Of Infant’s Cry To Assess The Postoperative Pain

Background and ObjectiveThe assessment of pain represents a great challenge,especially for infants, because infants can not describe their pain, leading to a lack of simple, feasible, clinically effective pain assessment scale for infants. “Cry” was one of the most important parameters in pain assessment which was included in many pain scale tools for children. It’s reported that the acoustic characteristics of pain cry is differ from those of no pain cry. However, there was no report about whether these acoustic characteristics of pain cry can be used to assess postoperative pain in infants. In this study, we will establish a model of recognizer for postoperative pain cry, and we will assess the diagnostic performance of this model. We wish our study can provide research foundation to find a more convenient pain assessment scale for infants. Materials & Methods155 post-operative infants aged < 6 mouths were recruited in this prospective study(Chi CTR-OCH-14004648). Infant’s cry was recorded from they were finished the operation and came into the recovery unit until they were discharge from the recovery unit. FLACC(Face,Legs,Activity,Cry,Consolability) scales were used to evaluate postoperative pain intensity. Adobe audition 3.0 was used to preliminary treatment the cry audio files; the characteristic parameters of the cry audio were analyzed by Praat software. MFCC(Mel Frequency Cepstrum Coefficient) parameter and HMM(Hidden Markorv Model) were used to built a model of recognizer for postoperative pain cry.With FLACC as a standard, we used ROC(Receiver operating characteristic curve) curve to test the diagnostic performance of the model of recognizer for postoperative pain cry. The Kruskal-Wallis(K-W) test was used to determine whether continuous variables followed a normal distribution. Variables that were not normally distributed were log-transformed to approximate normal distribution for analysis. The results are expressed as the mean ± standard deviation or median, unless otherwise stated. The characteristics of the subjects according to relevant groups were assessed using one-way analysis of variance(ANOVA) for continuous variables and the χ2 test for categorical variables. Pearson and Spearman analytical methods were employed for correlation analysis of two variables. variance between the two groups using t-test, the non-homogeneity of variance between the two groups using wilcoxon test, Data were analyzed using SPSS19.0(USA) and R v 2.15.1 for the statistical mapping. Results1. The fundamental Frequency(F0) of the pain cry was 654.7±195.8Hz, it was significantly higher than the F0 of cry before operation(464.6±146.1Hz), p<0.01.2.The fundamental Frequency(F0), First formant(F1), RMS(Root mean square) of pain cry audio were significantly higher than no pain / mild pain cry audio(p<0.01), while the time of interval between Syllables(IS) decreased significantly(p <0.01).3. The best model to recognize the post-operative pain cry was HMM(Hidden Markorv Model) Model of 18 state together with 12 mixture. The AUC(area under curve) of this model to diagnose pain was 0.81±0.049,(95% CI: 0.713-0.906), the cut-off point was 0.558, the sensitivity was 80.0%, and the specificity was 77.1%. The AUC of this model to diagnose severe pain was 0.764 ± 0.059,(95% CI: 0.648-0.880), the sensitivity was 91.3%, and the specificity was 50.9%. ConclusionThe acoustic characteristics of post-operative pain cry are differing from those of no pain cry. These acoustic characteristics of pain cry can be detected by the model of recognizer for post-operative pain cry and be used to assess post-operative pain in infants.