In male reproducible health and fertility and IVF (in-vitro fertilization), morphological analysis of sperm has been most important. But the traditional tools for semen analysis are subjective, imprecise, inaccurate, difficult to standardize, and reproduce mainly due to their manually oriented operations. The purpose of morphological analysis of sperm is to microscopically type-classify sperm according to their morphological characteristics of heads. Until now, the strict criteria method has long been used in clinic to discriminate normal sperm from abnormal. This method cannot classify the diverse groups of abnormal sperm in detail and shows large variations in inter-operators and intra-operator. In this paper, we have studied a new method of sperm morphological classification using artificial neural networks that are widely used in pattern recognition and image processing. With a multi-layer perceptron trained by the error back-propagation algorithm, profile features from digitized sperm images were classified into four classes that consisted of one normal group and three abnormal groups according to their morphological characteristics.