Analysis of uterine contraction signals

In the United States there is a 27% of increase in premature birth over the past 20 years. The lack of an effective method for differentiating true labor from false labor causes unwanted hospital stays and treatment which could be avoided if there is a better understanding of the uterus dynamics. Studies show that labor progress correlates with electrical activities of the uterus. There must be a pattern in the contractions itself or in their coordination which can reveal the difference between uterine contractions that lead to delivery and the ones that do not. To understand the low level activities of uterus, first we analyzed magnetic recordings from the maternal abdomen which were acquired from Squid Array for Reproductive Assessment (SARA) system at the University of Arkansas for Medical Sciences. Neural networks were used for classifying those recordings into two groups. The first group of classes has signals that include contraction and the second group of signals has the ones that do not include contractions. Wavelet transforms were used for feature extraction of the signal.