The inflammatory response of the lung in COVID-19 : correlations between imaging-based methods and clinical outcomes

<p dir="ltr"><b>Background</b> </p><p dir="ltr">Coronavirus disease 2019 (COVID-19), caused by the novel coronavirus SARS- CoV-2, was first identified in China in late 2019. Its rapid global spread led to a pandemic. Clinical presentation varies from asymptomatic to critical illness. Although multiple organs may be affected, severe cases are marked by a dysregulated immune response that primarily targets the lungs and can progress to acute respiratory failure and death.</p><p dir="ltr">As pulmonary involvement is central to severe disease, chest computed tomography (CT) rapidly emerged as an essential tool for assessing lung injury in hospitalised patients. CT provided higher sensitivity than chest radiography and supported clinical decision-making when molecular testing was limited. Guidance from the European Society of Thoracic Imaging (ESTI) and the European Society of Radiology (ESR) emphasised that CT should be reserved for hospitalised patients with moderate to severe disease or when differential diagnoses needed to be excluded.</p><p dir="ltr">However, the prognostic value of specific radiological patterns on CT and their relationships to pathology, clinical outcomes, and long-term sequelae remained unclear. Studies in this thesis included hospitalised COVID-19 cohorts, spanning the acute phase, intensive care unit (ICU) survivors, and post-discharge follow- up. The work integrates radiological, pathological, functional, and computational imaging perspectives to clarify how pulmonary inflammation presents on imaging and how image derived measures relate to clinical trajectories from acute illness through to convalescence.</p><p dir="ltr">The overall aim was to characterise the extent of pulmonary inflammation as visualised by CT in hospitalised COVID-19 patients, and to investigate the underlying structural and physiological mechanisms reflected in imaging findings, in relation to pathology, lung function, and clinical outcomes across different disease phases. Specifically, the thesis sought to: (i) determine the prognostic significance of widespread parenchymal abnormalities (WPA) with or without pulmonary embolism (PE) on contrast-enhanced CT pulmonary angiography (CTPA) in the acute setting; (ii) evaluate whether artificial intelligence (AI) can reproducibly quantify opacity burden on contrast-enhanced CT compared with radiologists; (iii) assess the prevalence and functional impact of residual radiological abnormalities in ICU survivors during follow-up; (iv) identify structural vascular and airway correlates of reduced diffusion capacity using functional respiratory imaging (FRI) on non-contrast high-resolution CT (HRCT); and (v) validate in-vivo CT patterns against post-mortem histopathology in fatal cases.</p><p dir="ltr"><b>Material and Methods</b></p><p dir="ltr">The thesis comprises five studies conducted primarily in hospitalised COVID-19 cohorts, ranging from acutely admitted patients to ICU survivors and selected follow-up groups. In paper 1, a retrospective analysis of contrast-enhanced CTPA for suspected PE during acute COVID-19 was performed. Extent of WPA, presence and distribution of PE, and associations with markers and 60-day mortality were assessed. In paper 2, AI-based opacity quantification was tested on contrast-enhanced CTPA in prospectively included acute-phase patients, by comparison with two radiologists. In paper 3, a cohort of ICU-treated survivors underwent non-contrast CT, lung function testing, and exercise assessment at ~10 months, with a subset re-imaged at two years to evaluate fibrotic-like changes and their correlates. In paper 4, FRI was applied to non-contrast inspiratory and expiratory HRCT 4-11 months post-discharge in a homogeneous male cohort, deriving vascular and airway metrics and related them to diffusion capacity (DLCO) and walking performance. In paper 5, systematic radiology- pathology correlation of in-vivo chest CT shortly before death and autopsy was undertaken, examining how CT patterns corresponded to histopathological features in fatal COVID-19.</p><p dir="ltr"><b>Main Results</b></p><p dir="ltr">Paper 1: In the acute phase, WPA on CT, particularly when combined with PE, were strongly associated with severity and 60-day mortality. Emboli were frequently small and peripheral, consistent with recognition of pulmonary vascular involvement in COVID-19.</p><p dir="ltr">Paper 2: On contrast-enhanced CTPA, an AI tool trained on non-contrast data showed strong agreement with radiologists in estimating opacity burden. Performance held despite intravenous contrast, supporting its applicability in real-world CTPA workflows where contrast is required to exclude PE.</p><p dir="ltr">Paper 3: Beyond the acute phase, residual widespread opacities on follow-up CT were common. At two years, "fibrotic-like" changes were frequent (92%). These abnormalities correlated with reduced lung volumes, lower DLCO, and exercise-induced desaturation, though the relation between radiology and function was heterogeneous.</p><p dir="ltr">Paper 4: FRI revealed vascular and airway remodelling 4-11 months post- discharge. Patients with reduced DLCO had lower BV5% and higher siVaw% aligned with reduced lobar volumes, supporting small-vessel loss and airway calibre changes as contributors to impaired gas transfer.</p><p dir="ltr">Paper 5: Correlation of in-vivo CT with post-mortem histopathology showed that typical CT patterns (ground-glass opacities, consolidations, organising pneumonia, and crazy paving), corresponded to diffuse alveolar damage with oedema, hyaline membranes, and organising pneumonia. Thromboembolic events were identified in larger pulmonary arteries at autopsy, while microvascular thrombi were not consistently visible on CT, reflecting resolution limits and timing relative to autopsy.</p><p dir="ltr"><b>Conclusions</b></p><p dir="ltr">Across hospitalised cohorts and disease stages, CT and advanced imaging capture COVID-19 lung injury from acute inflammation to chronic sequelae. In the acute setting, WPA, especially with PE on CTPA, are strongly associated with mortality, supporting CT for risk stratification in addition to diagnosis. AI offers reliable quantification of parenchymal involvement on contrast-enhanced scans, integrating with clinical workflows where contrast is required. During long-term follow-up of ICU survivors, persistent abnormalities align with physiological impairment, underlining the value of structured surveillance and the need to interpret imaging alongside function. FRI revealed structural changes, with loss of pulmonary vessels (reduced BV5%) and airway calibre changes (higher siVaw with reduced lobar volumes) as correlates of reduced diffusion capacity. Finally, radiology-pathology correlation validates CT features of diffuse alveolar damage and organising pneumonia, while clarifying CT's limits for microvascular thrombosis.</p><p dir="ltr">Taken together, these complementary approaches delineate how acute inflammation can evolve into chronic structural and functional sequelae. Overall, this thesis advances our understanding of COVID-19-related pulmonary inflammation by linking imaging findings to clinical, functional, and pathological outcomes.</p><h3>List of scientific papers</h3><p dir="ltr">I. Jalde FC, Beckman MO, <b>Svensson AM</b>, Bell M, Sköld M, Strand F, Nyren S, Kistner A. Widespread Parenchymal Abnormalities and Pulmonary Embolism on Contrast-Enhanced CT Predict Disease Severity and Mortality in Hospitalized COVID-19 Patients. Front Med (Lausanne). 2021 Jun 29;8:666723. PMID: 34268322 <a href="https://doi.org/10.3389/fmed.2021.666723">https://doi.org/10.3389/fmed.2021.666723</a></p><p dir="ltr">II. <b>Svensson AM</b>, Kistner A, Kairaitis K, Prisk GK, Farrow C, Amis T, Wagner PD, Malhotra A, Harbut P. Quantitative assessment of lung opacities from CT of pulmonary artery imaging data in COVID-19 patients: artificial intelligence versus radiologist. BJR Open. 2025 Apr 29;7(1):tzaf008. PMID: 40370862 <a href="https://doi.org/10.1093/bjro/tzaf008">https://doi.org/10.1093/bjro/tzaf008</a></p><p dir="ltr">III. Björnson M, <b>Svensson AM</b>, He C, Sköld M, Nyrén S, Nygren- Bonnier M, Bruchfeld J, Runold M, Jalde FC, Kistner A. Residual radiological opacities correlate with disease outcomes in ICU- treated COVID-19. Front Med (Lausanne). 2024 Apr 3;11:1263511. PMID: 38633311 <a href="https://doi.org/10.3389/fmed.2024.1263511">https://doi.org/10.3389/fmed.2024.1263511</a></p><p dir="ltr">IV. <b>Svensson AM</b>, Björnson M, Sköld M, Kauczor HU, Nygren- Bonnier M, Bruchfeld J, Runold M, Kistner A, Nyrén S. Pulmonary vascular and airway changes in previously hospitalised COVID- 19 patients: long-term functional respiratory imaging findings correlate with reduced DLCO. [Accepted]</p><p dir="ltr">V. <b>Svensson AM</b>, Sakely L, Brunnström H, Nyrén S, Kistner A, Ortiz- Villalón C. Pulmonary findings in COVID-19: radiology correlates with the histopathological post-mortem findings in patients with fatal acute disease. [Submitted]</p>