Research On Computer-aided Diagnosis Technology In Fetal Heart Rate Monitoring

With the publication of two-child policy in China,the number of elderly pregnant women has gradually increased.Correspondingly,the health status of pregnant women and fetuses during the perinatal period has also received more and more attention.Therefore,as a common method of monitoring fetal health in clinic,the accuracy and real-time nature of Fetal Heart Rate(FHR)monitoring is very important.Traditional FHR monitoring is usually interpreted by doctors,which lacks unity and real-time natrue.In recent years,with the maturity of medical-industrial technology,computeraided diagnosis technology has proven to be an effective solution.This paper mainly studies the automatic analysis and diagnosis of related indicators in FHR monitoring under computer-aided conditions.First,this paper introduces the detection process of the FHR signal and key indicators such as the baseline during monitoring,which provide a conceptual basis for subsequent analysis.The interpolation algorithm experiment is performed for the missing value problem caused by the detection process,thereby determining the use of the interpolation algorithm.The baseline is extracted through a low-pass filter and the specific factors affecting the baseline effect are analyzed to provide an optimization direction for subsequent baseline estimation.Secondly,in view of the baseline problem proposed in Chapter 2,this paper conducts research and analysis on the existing FHR baseline extraction algorithm.Based on this,a multiple filtering algorithm is designed in combination with a progressive correction strategy to improve the stability of baseline and reduce the impact of acceleration/deceleration.Then,this paper combines the real data set to model the Sinusoidal Fetal Heart Rate(SFHR)to realize the online detection of SFHR.In this paper,an Empirical Mode Decomposition(EMD)algorithm is used to reconstruct the baseline of the SFHR.The SFHR is modeled using the periodic standard deviation and fractal dimension,and the signal fluctuation limit is added to further modify the model.Experimental simulations verify that the model has good recognition effect.Finally,above algorithms are integrated into the software,which is written in C# and C ++.Test the algorithm in the real environment and analyze the cause of the deviation caused by the algorithm running in the real environment.