Exploring New Characteristics: Using Deep Learning and 3D Reconstruction to Compare the Original COVID-19 and Its Delta Variant Based on Chest CT

Na Bai, Ruikai Lin, Zhiwei Wang, Shengyan Cai, Jianliang Huang, Zhongrui Su, Yuanzhen Yao, Fang Wen, Han Li, Yuxin Huang, Yi Zhao, Tao Xia, Mingsheng Lei, Weizhen Yang, Zhaowen QiuFrontiers in Molecular Biosciences 2022 · DOI

Introduction

While RT-PCR remains the gold standard for COVID-19 diagnosis, its binary results fail to capture the nuances of disease progression or the severity of lung involvement. Radiologists frequently turn to chest CT scans to fill this gap; however, traditional visual scoring is subjective, time-consuming, and prone to inter-observer variability. As the SARS-CoV-2 virus evolved from the original strain to the Delta variant, understanding how these physical manifestations changed became critical for clinical staging and recovery prediction. This study introduces an automated, AI-driven pipeline to objectively quantify and compare the lung damage caused by these two variants.

Automated AI Pipeline

To move beyond subjective visual assessment, researchers developed a multi-stage computational framework. The core of this system is a 2.5D Segmentation Model. By feeding several adjacent CT slices into a U-Net-based architecture, the model utilizes spatial context to accurately segment lungs and infections without the prohibitive computational costs of full 3D convolutional neural networks (CNNs).

The online COVID-19 modeling platform showing (A) 2D CT reading and (B) 3D model visualization.

Once segmented, the regions are converted into 3D models (STL format). A novel stereoscopic segmentation algorithm then automatically subdivides these models into five lobes and 18 segments based on the branching patterns of pulmonary arteries and bronchi. This allows for the calculation of the Percentage of Infection (POI) for each specific lobe, which is then mapped to a digitized version of the Kazerooni semi-quantitative scoring system (0–5).

The principle flowchart illustrating the automated lung lobe segmentation process.

Comparative Analysis: Original vs. Delta

The study analyzed retrospective data from two cohorts: 35 patients with the original strain (91 scans) and 72 patients with the Delta variant (358 scans). Using non-linear regression to fit CT scores against time since symptom onset, several key differences emerged:

  • Delayed Peak: The Delta variant exhibited a "slow burn" progression. Lung involvement peaked on day 15 for Delta patients, compared to day 11 for those with the original strain.
  • Prolonged Clearance: The median time to reach a stable negative RT-PCR result was significantly longer for Delta (39 days) than for the original strain (22 days, $p < 0.001$). Consequently, Delta patients required longer hospital stays (median 42 vs. 28 days).
  • Viral Persistence: Delta patients were nearly twice as likely to experience "re-detectable" positive RT-PCR results after an initial negative test (30.5% vs. 17.1%), suggesting different viral shedding dynamics.
  • Spatial Distribution: While both strains primarily affect the lower lobes, the original strain showed higher initial involvement in the right lower lobe. In contrast, the Delta variant showed more persistent lesions in the right middle lobe during the absorption (recovery) phase.

Clinical Implications and Limitations

Based on these findings, the authors propose a revised clinical timeline for the Delta variant: an Early Stage (0–3 days), a Progressive and Peak Stage (4–16 days), and an extended Absorption Stage (17–42 days). This longer duration of lung change and the higher rate of re-detectable viral loads suggest that Delta requires more cautious discharge criteria and longer follow-up periods.

However, the study notes limitations, including the mixed vaccination status of the Delta cohort (41.7% vaccinated vs. 0% in the original group) and data heterogeneity across different hospital CT equipment. Future research will focus on applying this 3D quantification pipeline to newer variants, such as Omicron, to further refine our understanding of COVID-19's evolving clinical signature.

Summary updated

Abstract

<b>Purpose:</b> Computer-aided diagnostic methods were used to compare the characteristics of the Original COVID-19 and its Delta Variant. <b>Methods:</b> This was a retrospective study. A deep learning segmentation model was applied to segment lungs and infections in CT. Three-dimensional (3D) reconstruction was used to create 3D models of the patient's lungs and infections. A stereoscopic segmentation method was proposed, which can subdivide the 3D lung into five lobes and 18 segments. An expert-based CT scoring system was improved and artificial intelligence was used to automatically score instead of visual score. Non-linear regression and quantitative analysis were used to analyze the dynamic changes in the percentages of infection (POI). <b>Results:</b> The POI in the five lung lobes of all patients were calculated and converted into CT scores. The CT scores of Original COVID-19 patients and Delta Variant patients since the onset of initial symptoms were fitted over time, respectively. The peak was found to occur on day 11 in Original COVID-19 patients and on day 15 in Delta Variant patients. The time course of lung changes in CT of Delta Variant patients was redetermined as early stage (0-3 days), progressive and peak stage (4-16 days), and absorption stage (17-42 days). The first RT-PCR negative time in Original COVID-19 patients appeared earlier than in Delta Variant patients (22 [17-30] vs. 39 [31-44], <i>p</i> < 0.001). Delta Variant patients had more re-detectable positive RT-PCR test results than Original COVID-19 patients after the first negative RT-PCR time (30.5% vs. 17.1%). In the early stage, CT scores in the right lower lobe were significantly different (Delta Variant vs. Original COVID-19, 0.8 ± 0.6 vs. 1.3 ± 0.6, <i>p</i> = 0.039). In the absorption stage, CT scores of the right middle lobes were significantly different (Delta Variant vs. Original COVID-19, 0.6 ± 0.7 vs. 0.3 ± 0.4, <i>p</i> = 0.012). The left and the right lower lobes contributed most to lung involvement at any given time. <b>Conclusion:</b> Compared with the Original COVID-19, the Delta Variant has a longer lung change duration, more re-detectable positive RT-PCR test results, different locations of pneumonia, and more lesions in the early stage, and the peak of infection occurred later.
BibTeX
@inproceedings{bai2022exploring,
  author = {Na Bai and Ruikai Lin and Zhiwei Wang and Shengyan Cai and Jianliang Huang and Zhongrui Su and Yuanzhen Yao and Fang Wen and Han Li and Yuxin Huang and Yi Zhao and Tao Xia and Mingsheng Lei and Weizhen Yang and Zhaowen Qiu},
  title = {{Exploring New Characteristics: Using Deep Learning and 3D Reconstruction to Compare the Original COVID-19 and Its Delta Variant Based on Chest CT}},
  booktitle = {Frontiers in Molecular Biosciences},
  year = {2022},
  doi = {10.3389/fmolb.2022.836862},
}
Figure