Development of AI based imaging tools for the analysis of lung disease in preclinical rodent models

In vivo lung micro-computed tomography (micro-CT) is gaining interest for the longitudinal assessment of experimental disease models in preclinical pulmonary research. Currently, there is a shift towards automated quantitative measuring in clinical imaging as it eliminates variability in qualitative readings and can accurately measure disease progress or treatment response when performing follow-up studies. In this context, artificial intelligence (AI) and more specifically deep learning (DL) has proven to be extremely useful. In the first stage of the PhD, state-of-the-art deep learning techniques will be deployed to automate the segmentation of lung structures in micro-CT images of rodent animals, currently requiring manual labour. Structures of interest will include the lungs as a whole and in their lobes, the airway tree, the lung vessels and parenchymal diseases. In the second stage, the changes of the above-mentioned lung structures will be analysed during disease initiation and evolvement, while comparing them with healthy controls. This approach will be evaluated in a variety of available mouse lung disease models, including interstitial lung disease (ILD), chronic obstructive pulmonary disease (COPD), lung infections, lung metastasis, asthma and pulmonary hypertension. For each of them, radiological biomarkers constituting measurable indicators of disease state and efficacy of treatment will be assessed.