Uterine Movement Analysis Based on Velocity Information by Convolutional Neural Networks

Female infertility patients have characteristic uterine movements. Since the movements are related to fertility, it is important to analyze the movements. We developed an automated analysis system for the uterine movements. The system was composed of convolutional neural networks. The system predicted the occurred movements from images with velocity information. The images were generated from Cine MRI images by calculating optical flow for each frame image. We confirmed that the accuracy of the model by the velocity images is higher than the model by the original images. We found that the uterine movement velocity is very important information to analyze the uterine movement patterns.