Integrating genetic algorithm and decision tree learning for assistance in predicting in vitro fertilization outcomes

Accurate and early prediction of the outcome of an in vitro fertilization (IVF) treatment is important for both patients and physicians. The most common question asked by IVF patients is What are my chances of conceiving? The answer to this difficult question typically considers patient age, day 3 serum follicle stimulating hormone (FSH) levels, and infertility diagnosis. However, many more parameters are known to affect IVF outcome. It is difficult for the clinician to recognize trends and intuitively decide how to optimize success rates for each infertile couple. This paper presents a hybrid intelligence method which integrating genetic algorithm and decision learning techniques for knowledge mining of an IVF medical database. The proposed method can not only assist the IVF physician in predicting the IVF outcome, but also find useful knowledge that can help the IVF physician tailor the IVF treatment to the individual patient with I he aim of improving the pregnancy success rate. The twenty-eight most significant attributes for determining the pregnancy rate (e.g., patient's age, number of embryo transferred, number of frozen embryos, and culture days of embryo) and their combinative relationships (represented by if-then rules) were identified through the proposed method. The knowledge discovered in this study is currently accepted as an interesting discovery from the viewpoint of domain experts. For the results from this study to be conveniently accessed by IVF physicians and patients, an expert system tool equipped with the proposed IVF outcome prediction model was built. (C) 2010 Elsevier Ltd. All rights reserved.