Prediction of the basic gonadotrophic hormone levels in girls with precocious puberty using ultrasonic union artificial neural network

Objective To Predict the basic luteinizing hormone (LH) and follicle stimulating hormone (FSH) levels in girls with precocious puberty using ultrasonic union artificial neural network. Methods In 71 girls with precocious puberty, the uterine and ovarian were examined with ultrasound. The back-propagation (BP) neural network was established using uterine volume, ovarian volume and the largest diameter of bilateral ovarian follicles as the input variables and the basic LH and FSH levels as the output variable. The data of 61 cases were used to train the model, and the other 10 cases were used to test the model. Results The predicted values of basic LH levels were correlated with the actual values with the correlation coefficient of 0.997 and the regression slope of 1.0088. The predicted values of basic FSH levels were correlated with the actual values with the correlation coefficient of 0.63 and the regression slope of 1.0054. Conclusions The ultrasonic union artificial neural network can be used to predict the basic LH level quite well and effective.