Figure 1 | Scientific Reports

Figure 1

From: Comparison between a deep-learning and a pixel-based approach for the automated quantification of HIV target cells in foreskin tissue

Figure 1

Training images rapidly improve the accuracy of a custom StarDist model used for the quantification of HIV target cells in foreskin tissue. Percent difference of automated counts generated by StarDist from manual counts for nuclei (all cells), CD3 + cells, CD4 + CCR5 + cells, and CD3 + CD4 + CCR5 + cells after training with 10 images, 20 images, 30 images, and 40 images. Automated counting was completed on 10 validation field-of-view images (600 × 600 µm) randomly generated from a set of 232 full foreskin tissue section scans stain for CD3, CD4, CCR5, and nuclei to identify HIV susceptible cells in the tissue.

Back to article page