Figure 5 | Scientific Reports

Figure 5

From: Application of multiple-finding segmentation utilizing Mask R-CNN-based deep learning in a rat model of drug-induced liver injury

Figure 5

Comparison of original, annotated, and algorithm-predicted images in the multiple-finding model. (A) Portal area, connective tissue, infiltration, necrosis, fibrosis, and vacuolation in the algorithm prediction and ground-truth annotation. Representative slide images include portal area, connective tissue (including fibrosis), infiltration, necrosis, and vacuolation. Colors are as follows: portal area (yellow), connective tissue (blue), infiltration (green), and vacuolation (red). (B) Linear regression was used to compare the number of pixels annotated by an accredited toxicologic pathologist with those predicted by the established multiple-finding model for portal area, infiltration, necrosis, vacuolation, and connective tissue (including fibrosis).

Back to article page