Figure 2
From: Sperm motility assessed by deep convolutional neural networks into WHO categories
Chart of workflow. (1) Semen samples from 65 individuals were placed on slides, and sperm motility was recorded by a microscope-mounted video camera. (2) Videos were distributed to reference laboratories for manual sperm motility assessment according to WHO recommendations. (3) Videos (n = 65) and mean values (n = 65) from the manual sperm motility assessment were used to train and test the three-category and four-category DCNN models. Optical flow generated from batches of 30 frames collected across the entire length of each video, was used to develop the models. (4) DCNN-predicted values were compared to manual results. The figure was created by the authors with Biorender.com.
