Predicted success rate of fertility in patients treated with assisted reproductive techniques using neuro fuzzy

International Journal of Reproductive BioMedicine

Volume 11 - Number Supple.2

Article Type: Original Article

Introduction: Infertility is one of the major problems involving 15% of young couples. Some couples solve their problem with regular clinical treatments. However, it is untreatable in 48% of them and they need special laboratory techniques called ART. There have been interests in cognitive sciences, neuronal networks, fuzzy theory and statistical neuronal models as the most effective tools in prediction problems. The purpose of this study was to evaluate the efficacy of Neuro fuzzy method in predicting the success rate of ART methods in treatment of infertile patients. Materials and Methods: 300 infertile women aged between 20-38 years old, who underwent ICSI surgery and were in oocyte retrieval and embryo transfer stage entered in this retrospective study. After data collection and initial analysis using univariate statistical tests, logistic regression model wasperformed for synchronic modeling of obtained variants from previous steps and important clinical variants. Besides, to practice neural fuzzy model, data were divided into training and testing groups Then, appropriate model was resulted from training data and was tested against testing group to assess its accuracy. Finally, we used ROC curve to evaluate the prediction power of various models. Results: Logistic regression model showed that of entered variables into model, wife s age, kind of infertility (primary or secondary), and duration of infertility can predict success of infertility treatment. Prediction value was 52.4% for logistic regression model and 84% for fuzzy neuronal networks. Conclusion: Although both models had low reliability, fuzzy neuronal model was more reliable. Moreover, fuzzy neuronal network model presented more effective variables as effective factors.