Fig. 4

Module 2: cell detection and training augmentation. a Sample images from training augmentation. The original image (top left) underwent 14 different transformations including brightness and contrast (bc) adjustments, color inversion (flip color), noise addition and image rotation. b The precision, recall and average precision (A.P.) of our cell detection model after and before training augmentation. c The precision and recall metrics of our model at detecting 100 test images at increasing noise levels. d The frequency distributions of the SNR of the 100 test images at increasing noise levels. e A test sample illustrating the model’s ability to detect cells at increasing noise levels. White arrows, relevant ROIs that were not detected at higher noise. Red arrows, irrelevant ROIs detected