Shortcoming of Visual Interpretation of Cardiotocography: A Comparative Study with Automated Method and Established Guideline Using Statistical Analysis

• Published in: SN computer science 2020-05, Vol.1 (3), p.179, Article 179
• Main Authors: Das, Sahana, Mukherjee, Himadri, Roy, Kaushik, Saha, Chanchal K.
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Clinical decision making; Comparative studies; Computer Imaging; Computer Science; Computer Systems Organization and Communication Networks; Data Structures and Information Theory; Decision analysis; Decision support systems; Decision theory; Decision trees; Fuzzy logic; Guidelines; Heart rate; Hypoxia; Information Systems and Communication Service; Parameters; Pattern Recognition and Graphics; Software Engineering/Programming and Operating Systems; Statistical analysis; Survey Article; Vision
• ISSN: 2662-995X; 2661-8907
• DOI: 10.1007/s42979-020-00188-x

Cardiotocography (CTG) has been the primary tool for monitoring the fetal health during antepartum and intra-partum period since the 1960s. It works by recording both fetal heart rate and mother’s uterine contraction pressure simultaneously. However, due to lack of consensus in the interpretation of features and due to inter- and intra-clinician variation the introduction of CTG in fetal care did little to reduce the fetal mortality and morbidity. In order to ensure that the signs of hypoxia are recognised at the onset it is necessary to have a robust clinical decision support system. Visual analysis by clinicians has been known to produce erroneous result as it is not possible to detect subtle changes with naked eye. Though many guidelines have been proposed, NICHD guidelines for the analysis of CTG are the widely acceptable one. In this work, authors compared inter- and intra-clinician variation using statistical measures and found poor agreement between them. Parameters of CTG were compared using fuzzy-logic-based algorithms developed by authors and the algorithm based on NICHD guidelines. Kappa coefficient, Bland–Altman and Deming regression showed that there was a good agreement between the two methods in measuring both quantitative parameters and qualitative parameters. For the overall classification of CTG, majority voting showed the disagreement between the clinicians. There was a good agreement between the three methods in identifying Normal CTG, but the agreement was poor for Pathological CTG when visual interpretation was compared with the other two methods using Deming regression, sensitivity and specificity analysis and Bland–Altman plot. The analyses established that visual interpretation fails to perform well when compared to the automated procedure and the guidelines of NICHD. Among the three methods, the automated system performed better as it had high true positive (TP) and true negative (TN) values while identifying Pathological CTGs.