Steroids Therapy in Patients With Severe COVID-19: Association With Decreasing of Pneumonia Fibrotic Tissue Volume

Abstract

Background: We use longitudinal chest CT images to explore the effect of steroids therapy in COVID-19 pneumonia which caused pulmonary lesion progression.

Materials and methods: We retrospectively enrolled 78 patients with severe to critical COVID-19 pneumonia, among which 25 patients (32.1%) who received steroid therapy. Patients were further divided into two groups with severe and significant-severe illness based on clinical symptoms. Serial longitudinal chest CT scans were performed for each patient. Lung tissue was segmented into the five lung lobes and mapped into the five pulmonary tissue type categories based on Hounsfield unit value. The volume changes of normal tissue and pneumonia fibrotic tissue in the entire lung and each five lung lobes were the primary outcomes. In addition, this study calculated the changing percentage of tissue volume relative to baseline value to directly demonstrate the disease progress.

Results: Steroid therapy was associated with the decrease of pneumonia fibrotic tissue (PFT) volume proportion. For example, after four CT cycles of treatment, the volume reduction percentage of PFT in the entire lung was -59.79[±12.4]% for the steroid-treated patients with severe illness, and its p-value was 0.000 compared to that (-27.54[±85.81]%) in non-steroid-treated ones. However, for the patient with a significant-severe illness, PFT reduction in steroid-treated patients was -41.92[±52.26]%, showing a 0.275 p-value compared to -37.18[±76.49]% in non-steroid-treated ones. The PFT evolution analysis in different lung lobes indicated consistent findings as well.

Conclusion: Steroid therapy showed a positive effect on the COVID-19 recovery, and its effect was related to the disease severity.

Keywords: COVID-19; chest-CT; pneumonia treatment; quantitative analysis; steroids.

FIGURE 1

Pipeline of chest CT image processing and quantification. (A) Collect the original CT images of patients; (D) Generate the full lung mask and lung lobe mask through a pre-trained neural network; (B) Extract the CT data of ROI from the original CT image based on the ROI masks; (C) Quantify the ROI CT data based on the HU values; (E) calculate the data outcomes. The primary outcomes are different tissue volume proportion in the entire lung region and each lung lobe along longitudinal CT time points, which are generated in (E) Feature Output step as: $\%Tissu{e}_{C{T}_i}={\displaystyle \frac{TissuevolumeatC{T}_{i}inROI}{TotalvolumeofROIatC{T}_i}}$, where CT_i = {CT₀,⋯,CT₄}, Tissue = {ET, NT, GGO, RLO, SO}, and ROI {fulllung, LUL, LLL, RUL, RML, RLL}. ET, emhpysema lung tissue; NT, normal lung tissue; PFT, pneumonia fibrotic tissue; GGO, ground-glass opacity; RLO, reticular and linear opacification; SO, consolidations.