Abstract
Introduction: Tweris Mini is a visual analysis Artificial Intelligence (AI) application designed to detect different types of fetal hypoxic stress, and common abnormal fetal heart rate patterns on Cardiotocograph (CTG) images. It is based on the classification system recommended by the International Expert Consensus Guidelines on Physiological Interpretation of Cardiotocograph published by over 50 CTG experts from more than 20 countries.
Objective: This study aims to evaluate the degree of agreement between the expert who pioneered the physiological interpretation of Cardiotocograph in 2006, and was on the editorial board of the international expert consensus guidelines, and the Tweris Mini App in classifying different types of fetal hypoxic stress and abnormal Cardiotocograph patterns, along with their proposed management.
Materials & Methods: A total of 100 anonymized CTG traces were randomly selected, representing no hypoxia (NH) and various types of fetal hypoxic stress: chronic (C), gradually evolving compensated (GC), gradually evolving decompensated (GD), subacute (S), and acute (A). Specific abnormal CTG patterns, including atypical sinusoidal (AS), typical sinusoidal (TS), and ZigZag (ZZ) patterns. The expert classified the traces and provided the following management recommendations: continue labour (CO), reduce stress (RS), or expedite birth (EB). Two independent obstetricians used the Tweris Mini App to classify the same CTG traces and the recommended management by the Tweris Mini AI App. The Cohen Kappa was used for statistical analysis to determine the level of agreement.
Results: The overall degree of agreement between the Tweris Mini App and the expert was 94%, with complete (100%) agreement relating to acute hypoxic stress, ZigZag and atypical sinusoidal (Poole Shark Teeth) patterns. The Cohen Kappa statistic for diagnostic agreement was 0.92 (95% CI: 0.83-0.99, p<.001). The agreement between the Tweris Mini App and the expert regarding the recommended management reached 98%. The Cohen Kappa statistic for management was 0.96 (95% CI: 0.94-0.98, p<.001.
Conclusion: There was excellent agreement (>90%) between the Tweris Mini App and the expert, who pioneered Physiological Interpretation of CTG in both diagnosing different types of fetal hypoxic stress and recommending optimum management. Our findings suggest that the Tweris AI Mini App has a high reliability to be used as diagnosis & decision support tool in clinical practice.