An Improved U-Net for Human Sperm Head Segmentation

Sperm morphology analysis is an important step in the clinical diagnosis of male infertility, which means that the shape of sperm head is an important indicator in sperm morphology analysis. Therefore the accurate and efficient segmentation of human sperm head is essential for accurate and objective analysis of sperm morphology. In this paper, we have proposed an efficient deep learning algorithm for fully automatic segmentation of human sperm head based on the U-Net network structure. First of all, we performed sperm cell image collection and built a new dataset that is suitable for segmentation of human sperm heads in deep learning algorithms. Our dataset consists of 1207 sperm cell images from more than 20 male infertility patients. Then we improved the U-Net architecture by integrating the dilated convolution into it and replaced the long skip layer in the original network with the block we designed, which finally formed our final deep convolutional neural network. We use our dataset to train our proposed network so that we can segment the sperm head. Our algorithm is one of the few methods for segmentation of sperm head using deep learning algorithms. And compared with previous methods, our model not only achieve good results in unstained and low-resolution images containing only individual sperm cells, but also shows excellent performance in complex images containing multiple sperm cells. Our experimental results have confirmed that the HDC (Hybrid Dilated Convolution) module and our designed Block have a noticeable improvement on segmentation results. Meanwhile, we have achieved a high Dice coefficient of 95.14\(\%\). The segmentation results we tested on the prostate dataset prove that our model has good generalization ability and robustness. It is worth noting that our algorithm processed the images showing the original true morphology of the sperm cells and achieved high segmentation accuracy. It’s is very important for doctors to diagnose whether the sperm morphology is abnormal in the clinic.