Predicting pulmonary fibrosis and ventilation-perfusion mismatch in Covid-19 patients by CT and MR image analysis

Background: Two prominent lung-related findings have emerged during follow-up of post-Covid patients who required hospital stay: 1) residual fibrotic changes on chest computer tomography (CT) together with impaired lung function and 2) exertional dyspnea and unexplainable desaturation during exercise. Also, patients with an initial mild disease may progress into long term symptoms.

Aims: We believe that imaging techniques, mainly CT and magnetic resonance imaging (MRI) can predict outcome and development of chronic lung disease such as pulmonary fibrosis, in Covid-19, and elucidate mechanisms of exertional dyspnea. We propose three major aims:

1. use and systematically interpret chest CT for the prediction of long-term lung damage in hospitalized and non-hospitalized Covid-19 patients

2. develop machine learning methods, including Artificial Intelligence (AI), for automatic analysis of pulmonary CT images resulting in rapid and reliable prediction of chronic lung disease.

3. use MRI techniques for detection of regional impairments in ventilation, perfusion and assessment of lung-tissue compliance

Project plan: Hospitalized Covid-19 patients from the out-patients section at Karolinska University Hospital will undergo CT images at the time of acute disease and at 2, 6 and 12 months of follow-up. Images will be analyzed both by manual assessment and by using artificial intelligence (AI) algorithms in order to develop models for predicting long-term fibrosis.

Lung-MRI will be performed on those with marked exertional dyspnea and desaturation to visualize functional lung changes not detectable using conventional radiologic and physiologic exams.

Significance: Our data will add towards better understanding of long-term impact of Covid-19 on the lung including identification of patients who are prone to develop pulmonary fibrosis. We may provide mechanistic explanation of what causes the exertional dyspnea and desaturation by investigation ventilation-perfusion imbalance. Secondly, during multidisciplinary conferences, this information will contribute to improved patient prognostication and rehabilitation stratification.

Also, the quantitative information of structural and functional changes by AI enables us to follow the patients with high precision, which, in a near future can be exploited for therapeutic trials.