Time quantified detection of fetal movements using a new fetal movement algorithm

Primarily, the objective is to develop an automated ultrasound fetal movement detection system that will better characterize fetal movements. Secondarily, the objective is to develop an improved method of quantifying the performance of fetal movement detectors. We recorded 20-minute segments of fetal movement on 101 patients using a UAMS-developed fetal movement detection algorithm (Russell algorithm) and compared this to a Hewlett-Packard (HP) M-1350-A. Movements were recorded on a second-per-second basis by an expert examiner reviewing videotaped real-time ultrasound images. Videotape (86,592 seconds) was scored and compared with the electronic movement-detection systems. The Russell algorithm detected 95.53% of the discrete movements greater than 5 seconds, while the HP system (M-1350-A) detected only 86.08% of the discrete movements (p = 0.012). Both devices were less efficient at detecting the short discrete movements, obtaining sensitivities of 57.39 and 35.22, respectively. Neither system fully identifies fetal movement based on the second-per-second system. Improved methods of quantifying performance indicated that the Russell algorithm performed better than the HP on these patients.