ABSTRACT

CONCLUSIONS:

A logistic regression model of clinical relevance was developed, and is deployed on the World Wide Web for clinical use.

RESULTS:

Forward regression of this model showed sperm morphology to be the most significant feature in predicting pregnancy (p=0.39), followed by maternal age (p=0.42), type of medication used for ovulation induction (p=0.6), sperm motility (p=0.61), semen volume (p=0.71) and sperm concentration (p=0.9). Reverse regression of the model revealed sperm motility to be the most significant feature in predicting pregnancy (p=0.37), followed by sperm morphology (p=0.39), maternal age (p=0.49), type of medication used for ovulation induction (0.61), sperm concentration (p=0.72) and semen volume (p=0.74).

MATERIALS-METHODS:

A dataset of 212 exemplars, derived from patients who underwent a first IUI cycle with ejaculated sperm, was divided into separate modeling and cross-validation sets, and retrospectively analyzed. The dataset contained input features of maternal age, type of medication used for ovulation induction, semen volume, sperm concentration, motility and morphology and output intra-uterine pregnancy, and was modeled using various mathematical methods, including linear and radial support vector machines, linear and quadratic discriminant function analysis, logistic regression, and neural computation. Various model architectures were used, in an attempt to achieve the highest model accuracy. A logistic regression model was found to have the highest accuracy, with a test set ROC area of 0.717.

OBJECTIVE:

Intrauterine insemination (IUI) using ejaculated sperm is a common option in the treatment of infertility of various etiologies. We sought to develop a computational model for the prediction of pregnancy following IUI.