Fig. 3: Detection performance. | Communications Biology

Fig. 3: Detection performance.

From: A deep learning framework deploying segment anything to detect pan-cancer mitotic figures from haematoxylin and eosin-stained slides

Fig. 3

a The testing F1 scores in the human subsets of the proposed framework. The scores are presented as individual data points (n = 5 independent experiments), with yellow dashed lines marking the ensemble F1 scores and red dashed lines marking the mean F1 scores reported by MIDOG++. b The changes in the average F1 score as more mitotic figures (MFs) are included in training. The scores are presented as individual data points (n = 5 independent experiments), with error bars presenting the 95% confidence intervals. c The detection results of the OMG-Net and the Retina-Net used in the MIDOG++ in example regions. The green, yellow and red bounding boxes represent the true positives, false positives and false negatives.

Back to article page