Optimized feature selection for the classification of uterine magnetomyography signals for the detection of term delivery. (April 2020)

• Record Type: Journal Article
• Authors: T., Ananda Babu; P., Rajesh Kumar
• Is Part Of: Biomedical signal processing and control. Volume 58(2020)
• Publication Date: 2020-04
• Subjects: Uterine magnetomyography -- Discrete wavelet transform -- Feature optimization -- Firefly algorithm -- Classification; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2020.101880
• Languages: English
• ISSNs: 1746-8094

Abstract

Graphical abstract: Highlights: Classification of MMG records of term patients (Physionet). DWT used with different mother wavelets to extract the features. GA, PSO and FFA techniques are used for feature optimization. Accuracy of 97.21% is obtained using FFA technique with SVM classifier. Abstract: The classification of uterine Magnetomyography (MMG) signals is attempted to characterize the uterine contractions that lead to true labor. This is crucial for providing proper care to the infants and mothers since most problems started at the onset of labor. Term records of the Physionet mmgdb database analyzed with the discrete wavelet transform (DWT) for this work. The DWT performed up to six levels to compute the features that are optimized using the firefly algorithm (FFA). The selected features then fed to four different classifiers for the labor assessment. The performance of each classifier is calculated for different mother wavelets. The experimental results compared with the genetic algorithm (GA) and particle swarm optimization (PSO) algorithms. The discrete wavelet features extracted using the coif5 wavelet and optimized using the firefly algorithm outperformed the other algorithms. The support vector machine (SVM) classifier produced the best results of 97.2105% accuracy, 97.3108% precision, and 5.3784% false-positive rate (FPrate). The results obtained from this research assist clinicians in the understanding of the parturition process and term labor assessment.