Estimation of sperm morphology using a new CASA system

The rate of morphologically normal and abnormal spermatozoa plays a role in the prediction of male fertility and in the description of the integrity of spermatogenesis. It is thus important to estimate sperm morphology with reproducible results. For this purpose, several systems using computer-assisted image analysis were developed, similar to those used for the determination of sperm motility, which are known as CASA-systems. We tested an advanced CASA-system using a neuronal network, which could be trained for the determination of classes of morphology defined by the WHO group. For the first time, this system also provided the automatic determination of midpieces and sperm tails. The system revealed no significant differences of the percentage of normal forms when compared to the direct microscopical estimation. However, nearly all the classes of pathological sperm heads were estimated by the two methods as being significantly different. Concerning midpieces, the neuronal net was unable to classify the normal forms correctly. The estimation of normal sperm tails by the two methods led to a sufficient correlation of values. Other classes of sperm tails, like curled or kinked tails, were not recognized by the software. The investigation time with the CASA system was twice as long as with the direct microscopical evaluation. Since functional characteristics of spermatozoa cannot be read from morphological parameters, it would be of use to measure the functional parameters directly. The main advantages of a CASA-system for sperm morphometry concerns laboratory quality controls.