A Pregnancy Prediction System based on Uterine Peristalsis from Ultrasonic Images

In infertility treatment, it is required to improve a success rate of the treatment. A purpose this study develop prediction system for pregnancy outcomes using ultrasonic images. typical evaluate endometrial shape by The convolutional neural network (CNN) developed in current predicted outcome velocity information. information has movement feature uterine. It known that uterine deep related infertility. Experiments compared velocity-based and shape-based systems. systems predict optimal features based on inputting original images CNN model. findings revealed provided similar accuracy However, output system, area under curve (AUC) receiver operating characteristic (ROC) curve, higher value than AUC values were 0.65, 0.72, respectively. These results showed analysis movements was effective prediction. Previous clinical evaluation did not target but only shape. Therefore, new treatment approach