Investigating association relationship between fetal heart rate parameters from cardiotocography employing multi-objective evolutionary algorithms

• Published in: International journal of information technology (Singapore. Online) 2022, Vol.14 (4), p.1923-1935
• Main Authors: Piri, Jayashree, Mohapatra, Puspanjali, Dey, Raghunath
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Computer Imaging; Computer Science; Image Processing and Computer Vision; Machine Learning; Original Research; Pattern Recognition and Graphics; Software Engineering; Vision
• ISSN: 2511-2104; 2511-2112
• DOI: 10.1007/s41870-022-00909-w

Cardiotocography (CTG) is being used to track the heart rhythm of a baby and the contractions of a pregnant woman when the fetus is in the womb. In the case of acute or chronic issues, the CTG may be the fetal diagnostic test throughout pregnancy. However, the implementation of CTG in prenatal care did little to mitigate fetal mortality or morbidity due to the inability to agree on the understanding of features and the inter and intra-clinical differences. A comprehensive framework of clinical decision support is needed to assure that the symptoms of hypoxia are detected at the onset. One of the problems in data mining is the extraction of association rules, which focuses on discovering useful and interesting associations in massive data. Though association rule mining (ARM) is very common and helpful, there is no related analysis of CTG properties in accordance with our understanding. This research is aimed at the automated extraction of understandable, interesting, and reliable association rules (AR) from UCI CTG data. Traditional ARM methods cannot be effectively utilised to deal with this problem because of the numerical properties of CTG information. In this work, we hypothesise that multi-objective metaheuristic approaches can be employed efficiently for association analysis of CTG data automatically. Because of the quantitative variables in the CTG records, evolutionary intelligent MODENAR, MOGA, and QAR-CIP-NSGA-II are treated as rule miners from CTG data without any discretization involving expert knowledge. In this analysis, the evolutionary methods designed for a clinical recommendation system adjust and adapt to automatically discover numerical ARs without any data adjustment. Experimental results reveal that the MODENAR was superior in terms of key performance indicators according to the multi-target rule sets received.