Automated fetal cardiac valve movement detection for modified myocardial performance index calculation

The Modified Myocardial Performance Index (Mod-MPI) is becoming an important index in fetal cardiac function evaluation. However, the current method for Mod-MPI calculation can be time-consuming and demonstrates poor inter-operator repeatability. This paper presents an automated method for detecting the opening and closing events of fetal cardiac valves with the aim of automating the Mod-MPI calculation. Fifty-four Doppler ultrasound images, showing blood inflow and outflow for the left ventricle, are analyzed to attempt to automatically detect the timings of a total of 905 opening and closing events for both aortic and mitral valves. Timings are found according to the morphological characteristics of waveforms as well as intensity information of images. The proposed method can detect the four valve movement events with high sensitivity (95.60-98.64%) and precision (96.85-100.00%). Results are verified by comparison with manual annotation of same images from an expert.