Initial CT features of COVID-19 predicting clinical category

Li Fan, Wenqing Le, Qin Zou, Xiuxiu Zhou, Yun Wang, Hao Tang, Jiafa Han, Shiyuan LiuChinese 2021 · DOI

Abstract

During the initial waves of the COVID-19 pandemic, rapid triage was essential for managing healthcare resources. While chest CT scans provided a quick method for identifying pneumonia, their ability to predict clinical progression remained a subject of investigation. This study evaluates the efficacy of initial chest CT features in predicting the clinical severity of COVID-19. Through a retrospective analysis of 86 patients, researchers identified key imaging markers, such as ground-glass opacities (GGO) and upper-lobe involvement, that correlate with severe disease. However, the study concludes that the most effective predictive model combines these imaging features with laboratory results, specifically a low lymphocyte count, which emerged as the strongest indicator of critical illness.

Introduction

The rapid spread of COVID-19 placed unprecedented strain on global healthcare systems, necessitating efficient methods for patient triaging. While Reverse Transcription Polymerase Chain Reaction (RT-PCR) testing remains the diagnostic gold standard, its turnaround time can be a bottleneck in emergency settings. Chest Computed Tomography (CT) emerged as a rapid alternative for detecting lung involvement. However, a critical question remained: could the initial CT scan accurately predict which patients would progress to severe or critical illness? Identifying high-risk individuals early is paramount for optimizing the allocation of intensive care unit (ICU) beds and mechanical ventilators. This research aimed to determine if specific "signatures" in initial CT scans could distinguish between mild/moderate cases and those destined for severe clinical outcomes.

Methodology

The researchers conducted a retrospective analysis of 86 RT-PCR-confirmed COVID-19 patients treated at Hankou Hospital in Wuhan between January and February 2020. The cohort consisted of 40 males and 46 females. Patients were categorized into two primary groups based on national treatment protocols:

  1. Mild/Moderate: Patients presenting with mild symptoms or non-severe pneumonia.
  2. Severe/Critical: Patients experiencing significant respiratory distress, low oxygen saturation (≤93%), or multi-organ failure.

High-resolution CT scans were performed using a Siemens Definition AS+ 128-slice scanner. Two experienced radiologists, blinded to the clinical status of the patients, analyzed the scans for various features, including lesion density (GGO vs. consolidation), distribution (focal vs. multifocal), and secondary signs like the "crazy-paving" pattern. Clinical and laboratory data, including white blood cell (WBC) and lymphocyte counts, were also collected. A binary logistic regression model was then developed to estimate the likelihood of severe/critical illness based on these combined variables.

Results and Discussion

The study revealed significant differences between the clinical groups. Patients in the severe/critical category were generally older (median age 63 vs. 49) and more likely to be male. Imaging analysis showed that ground-glass opacities (GGO) were significantly more prevalent in severe cases (57.8%) compared to mild/moderate cases (31.7%).

Initial CT scan showing focal ground-glass opacity, a common finding in early or less severe COVID-19 presentations.

A key finding was the significance of lung lobe involvement. While COVID-19 typically affects the lower lobes, involvement of the Right Upper Lobe (RUL) and Left Upper Lobe (LUL) was significantly higher in the severe/critical group. Extensive involvement across all lobes served as a major indicator of high risk. Despite these trends, the researchers noted a substantial overlap in CT features between categories, suggesting that a single scan in isolation may not be definitive for prognosis.

CT scans (a and b) demonstrating extensive, bilateral, and multifocal lung involvement, which is strongly associated with severe or critical clinical categories.

The most predictive factor identified was not an imaging feature alone, but a laboratory marker. The binary logistic regression model highlighted lymphocyte count as the most critical predictor. Patients with a decreased lymphocyte count were 7.58 times more likely to fall into the severe/critical category. In this cohort, 20% of the severe/critical group succumbed to the disease, while no deaths occurred in the mild/moderate group.

Conclusion

This research underscores that while initial chest CT scans provide valuable insights into the extent of lung involvement in COVID-19, they are most effective when integrated with clinical laboratory data. The combination of extensive lung involvement on CT and a low lymphocyte count provides a powerful tool for early triage. Future work should focus on larger, multi-center studies and the inclusion of temporal data from follow-up scans to further refine these predictive models and account for patient comorbidities.

Summary updated

BibTeX
@inproceedings{fan2021initial,
  author = {Li Fan and Wenqing Le and Qin Zou and Xiuxiu Zhou and Yun Wang and Hao Tang and Jiafa Han and Shiyuan Liu},
  title = {{Initial CT features of COVID-19 predicting clinical category}},
  booktitle = {Chinese},
  year = {2021},
  doi = {10.1007/s42058-021-00056-4},
}
Figure