Figure 3
From: Sperm motility assessed by deep convolutional neural networks into WHO categories
Comparison of manually and DCNN-predicted sperm motility in a three-category model. Correlation graphs (A–C) and difference plots (D,E) between sperm motility assessed manually by reference laboratories and motility values predicted for three categories by deep convolutional neural network (DCNN); (A) and (D) progressive spermatozoa, (B) and (E) non-progressive spermatozoa, (C) and (F) immotile spermatozoa. In the correlation graphs, solid lines represent the trendlines while dashed lines the 95% confidence interval. Pearson’s correlation coefficient (r) and p-value (p) between results are also depicted. In the difference plots, the bias estimated as the median of the differences between predicted and manually assessed motility is shown as a solid line, while the 95% limits of agreement (LoA) are represented in the figure and shown in dashed lines to display outliers.
