AI-CNN empowered micro-CT Imaging: deciphering phenotypic signatures in bleomycin-induced lung injury

Bleomycin (BLM)-induced lung injury in mice is a valuable model for investigating the molecular mechanisms that drive inflammation and fibrosis and for evaluating potential therapeutic approaches to treat the disease. Given high variability in the BLM model, it is critical to accurately phenotype the animals in the course of an experiment. In the current study, we aimed to demonstrate the utility of microscopic computed tomography (µCT) imaging combined with an artificial intelligence (AI) convolutional neural network (CNN)-powered lung segmentation for rapid phenotyping of BLM mice. µCT was performed in freely breathing C57Bl/6J mice under isoflurane anaesthesia on days 7, 14 and 21 post BLM administration, followed by terminal invasive lung function measurement. Histological assessment of lung collagen contents and CD45 staining were performed. BLM administration triggered formation of fibrotic lesions as evident by alterations in µCT-derived lung density, air to tissue volume ratio and lung aeration. In addition, the texture parameters, such as skewness and kurtosis were significantly altered in BLM mice in comparision to controls. AI CNN µCT-derived parameters strongly correlated with invasive lung function measurements and collagen contents. Our study demonstrates the utility of non-invasive µCT measurements and AI-CNN-powered image analysis for rapid, non-invasive, and accurate phenotyping of BLM mice in the course of the disease development and progression. Thus, the use of µCT imaging should be considered complementary to superior for ensuring proper randomisation of the affected animals prior to initiating any therapeutic application.