Skip to main content
Fig. 1 | Respiratory Research

Fig. 1

From: A fully automated micro‑CT deep learning approach for precision preclinical investigation of lung fibrosis progression and response to therapy

Fig. 1

Comparative imaging of a representative lung fibrosis mouse model at 21 days of BLM treatment. A A representative 2D coronal µCT slice of a fibrotic lung is shown on the left. The semi-automatic segmentation approach failed to visualize the entire lung (light-green line), while DL-based segmentation allowed the automated segmentation of the whole lung parenchyma (dark-green line). The 3D renderings shown on the right were generated by combining the images obtained with the semi-automatic (light green volume) and the DL-based segmentation approach (dark-green volume). The right and the left lung are indicated as RL and LL, respectively. Red pointed arrows indicate the more severe fibrotic lesions. B Histological slice of the lung with a dashed line separating the left and the right lungs (the bar corresponds to 1 mm). Regions of increasingly severe fibrosis are shown as magnified images (bars correspond to 25 µm), boxed by increasingly dark frames outside of the main picture. Fibrosis severity was assessed by Ashcroft score (AS) measurements: no/mild (light green), moderate (green), and severe (dark green). C Hounsfield Units (HU) frequency distributions for the left (LL, light blue line) and right (RL, orange line) lungs were determined with the DL-based model

Back to article page