Abstract
Background: New coronavirus disease (COVID-19) has caused a global pandemic that has raised worldwide concern. This study aims to investigate the correlation between the extent of lung infection and relevant clinical laboratory testing indicators in COVID-19 and to analyse its underlying mechanism. Methods: Chest high-resolution computer tomography (CT) images and laboratory examination data of 31 patients with COVID-19 were extracted, and the lesion areas in CT images were quantitatively segmented and calculated using the deep learning (DL) system. A cross-sectional study method was used to explore the differences among the proportions of lung lobe infection and to correlate the percentage of infection (POI) of the whole lung in all patients with clinical laboratory examination values. Results: No significant difference in the proportion of infection was noted among various lung lobes ($P > 0.05$). The POI of total lung infection was negatively correlated with the peripheral blood lymphocyte percentage (L%) ($r = -0.633, P < 0.0001$) and lymphocyte count (LY) ($r = -0.555, P = 0.001$) but positively correlated with the neutrophil percentage (G%) ($r = 0.565, P = 0.001$). Otherwise, the POI was not significantly correlated with the peripheral blood leukocyte count (WBC), monocyte percentage (M%) or haemoglobin content (HGB). In some patients, the G% increased continuously and was accompanied by a progressive decrease in the L% and LY. Conclusions: No significant difference in infection propensity was noted among the lung lobes in COVID-19 patients. The peripheral blood lymphocyte and neutrophil levels are significantly correlated with the extent of lung lesions, and related indicator abnormalities serve as a warning, thus guiding the implications of clinical interventions in patients.
Introduction
Early in the COVID-19 pandemic, clinicians relied heavily on RT-PCR nucleic acid tests and chest CT scans for diagnosis and monitoring. While CT scans are highly sensitive, traditional visual inspection by radiologists is qualitative and subjective. It is challenging for a human observer to precisely quantify the percentage of lung compromise, which complicates efforts to correlate physical pulmonary damage with clinical laboratory indicators.
This study addresses this gap by employing Artificial Intelligence (AI) to provide a quantitative "Percentage of Infection" (POI). By transforming qualitative imaging into quantitative data, the researchers aimed to identify which common blood markers serve as the most reliable proxies for the actual severity of lung infection, potentially providing a roadmap for clinical intervention.
Methods
The study utilized a cross-sectional design involving 31 laboratory-confirmed COVID-19 patients treated at Jingmen First People's Hospital in Hubei Province between January and February 2020. The cohort included 18 males and 13 females, with an average age of approximately 42 years.
The technical core of the research is the use of a Deep Learning (DL) system (Full-uAI-Discover-NCP) to perform automated segmentation of lung lesions from high-resolution CT (HRCT) scans. The AI divided each patient's lungs into 20 bronchopulmonary segments and calculated the Volume of Infection (VOI) and POI for the whole lung and individual lobes. These AI-generated metrics were then correlated with peripheral blood markers taken on the same day, including leukocyte count (WBC), neutrophil percentage (G%), lymphocyte percentage (L%), and lymphocyte count (LY), using Spearman correlation tests.
Results
Infection Uniformity
A key finding was the lack of significant difference in infection propensity among the different lung lobes ($P > 0.05$). COVID-19 appears to be an "equal opportunity" infector within the respiratory system, suggesting a systemic rather than localized vulnerability across the upper, middle, and lower lobes of both lungs.
Statistical Correlations
The study revealed a clear "seesaw" effect between lung damage and specific blood markers. The POI of the whole lung showed strong statistical correlations with lymphocyte and neutrophil levels, while other markers like total white blood cell count and hemoglobin showed no significant relationship.
As shown in the data, the POI was strongly negatively correlated with lymphocyte percentage ($r = -0.633$) and lymphocyte count ($r = -0.555$), while showing a positive correlation with neutrophil percentage ($r = 0.565$).
Dynamic Tracking
In longitudinal observations of patients whose conditions worsened, the AI recorded substantial increases in infected lung volume. These physical changes were precisely mirrored by a progressive decrease in lymphocytes and an increase in neutrophils over time.
For example, in Patient 18, as the infected lung volume jumped from 572.2 $cm^3$ to 1596.4 $cm^3$ across three time points, the lymphocyte percentage plummeted from 28.8% to 1.7%, while the neutrophil percentage rose from 52% to 95.9%.
Discussion
The strong correlation between POI and blood markers suggests that AI-driven quantification can bridge the gap between imaging and molecular biology. The authors hypothesize that the observed lymphopenia (drop in lymphocytes) might be due to an "inflammatory storm" (cytokine storm) causing direct cell destruction or the redistribution of these cells from the blood into the infected lung tissue.
Limitations
The study is limited by a small sample size (n=31) and potential treatment bias, as patients were at different stages of clinical care and may have received medications (like glucocorticoids) that influence blood marker levels independently of the viral infection.
Conclusion
This research demonstrates that peripheral blood lymphocyte and neutrophil levels are significant indicators of the extent of lung lesions in COVID-19. AI-based quantitative CT analysis provides a rigorous method for monitoring disease progression, where abnormalities in these blood markers can serve as early warning signs to guide clinical interventions.
Summary updated