Convolutional neural networks for segmentation and object detection of\n human semen

2017/04/03 by Malte Stær Nissen, Oswin Krause, Nissen, Malte Stær +9
Computer Science · Medicine · #Advanced Neural Network Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Reproductive Biology and Fertility #Sperm and Testicular Function

paper · pdf · doi:10.48550/arxiv.1704.00498

openalex publication_date 2017/04/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We compare a set of convolutional neural network (CNN) architectures for the\ntask of segmenting and detecting human sperm cells in an image taken from a\nsemen sample. In contrast to previous work, samples are not stained or washed\nto allow for full sperm quality analysis, making analysis harder due to\nclutter. Our results indicate that training on full images is superior to\ntraining on patches when class-skew is properly handled. Full image training\nincluding up-sampling during training proves to be beneficial in deep CNNs for\npixel wise accuracy and detection performance. Predicted sperm cells are found\nby using connected components on the CNN predictions. We investigate\noptimization of a threshold parameter on the size of detected components. Our\nbest network achieves 93.87% precision and 91.89% recall on our test dataset\nafter thresholding outperforming a classical mage analysis approach.\n

Related

• Sperm Detection and Tracking in Phase-Contrast Microscopy Image Sequences using Deep Learning and Modified CSR-DCF
• TOD-CNN: An Effective Convolutional Neural Network for Tiny Object Detection in Sperm Videos
• DeepSperm: A robust and real-time bull sperm-cell detection in densely populated semen videos
• Predicting Semen Motility using three-dimensional Convolutional Neural Networks
• Machine Learning-Based Analysis of Sperm Videos and Participant Data for\n Male Fertility Prediction