Severity Assessment of Coronavirus Disease 2019 (COVID-19) Using Quantitative Features from Chest CT Images (2003.11988v1)

Abstract: Background: Chest computed tomography (CT) is recognized as an important tool for COVID-19 severity assessment. As the number of affected patients increase rapidly, manual severity assessment becomes a labor-intensive task, and may lead to delayed treatment. Purpose: Using machine learning method to realize automatic severity assessment (non-severe or severe) of COVID-19 based on chest CT images, and to explore the severity-related features from the resulting assessment model. Materials and Method: Chest CT images of 176 patients (age 45.3$\pm$16.5 years, 96 male and 80 female) with confirmed COVID-19 are used, from which 63 quantitative features, e.g., the infection volume/ratio of the whole lung and the volume of ground-glass opacity (GGO) regions, are calculated. A random forest (RF) model is trained to assess the severity (non-severe or severe) based on quantitative features. Importance of each quantitative feature, which reflects the correlation to the severity of COVID-19, is calculated from the RF model. Results: Using three-fold cross validation, the RF model shows promising results, i.e., 0.933 of true positive rate, 0.745 of true negative rate, 0.875 of accuracy, and 0.91 of area under receiver operating characteristic curve (AUC). The resulting importance of quantitative features shows that the volume and its ratio (with respect to the whole lung volume) of ground glass opacity (GGO) regions are highly related to the severity of COVID-19, and the quantitative features calculated from the right lung are more related to the severity assessment than those of the left lung. Conclusion: The RF based model can achieve automatic severity assessment (non-severe or severe) of COVID-19 infection, and the performance is promising. Several quantitative features, which have the potential to reflect the severity of COVID-19, were revealed.

Severity Assessment of COVID-19 Using Quantitative Features from Chest CT Imaging

The paper "Severity Assessment of Coronavirus Disease 2019 (COVID-19) Using Quantitative Features from Chest CT Images" explores the development and application of a Random Forest (RF) model designed to facilitate automatic severity assessment of COVID-19 from chest computed tomography (CT) images. The authors leverage machine learning techniques to classify the severity of infection into non-severe and severe categories, based on quantitative features extracted from CT scans. This approach addresses the bottleneck caused by manual assessments, which are labor-intensive and may be subject to delays.

In this paper, CT images from 176 patients diagnosed with COVID-19 were assessed, from which 63 quantitative features were extracted. These features include metrics such as infection volume and its ratio relative to the whole lung, with a specific focus on ground glass opacity (GGO) regions, which have shown significant correlation with disease severity. The authors trained the RF model using these quantitative features to distinguish between severe and non-severe cases, obtaining promising results through a three-fold cross-validation process.

The RF model achieved a true positive rate (TPR) of 0.933 and a true negative rate (TNR) of 0.745, yielding an overall accuracy of 0.875 and an area under the receiver operating characteristic (ROC) curve (AUC) of 0.91. These results underscore the model's potential utility in clinical settings, where rapid and accurate severity assessments are critical. Notably, the analysis revealed that features derived from the right lung were more indicative of severity than those from the left lung.

The authors highlight several key findings from their paper. The volume and ratio of GGO regions emerged as the most significant indicators of COVID-19 severity, contrasting with previous studies that prioritized consolidation regions typically seen in ICU patients or in the later disease stages. This discrepancy may arise from the paper's focus on initial CT images, suggesting GGO predominance at baseline. Furthermore, the authors reported that the top 30 quantitative features, rather than the full set of 63, provided optimal model performance.

This research holds implications for both practical and theoretical domains. Practically, it provides a framework for the automated and timely assessment of COVID-19 severity, potentially leading to faster clinical decision-making and resource allocation. Theoretically, it contributes to the body of knowledge surrounding the interpretation and significance of GGO in COVID-19 pathology.

In terms of future directions, the paper's limitations—specifically the binary classification of severity—offer opportunities for refinement. Expanding the classification to include the full spectrum of clinical presentations (mild, common, severe, and critical) could enhance the model's applicability. Additionally, further collaborations across multiple centers could enrich the dataset, addressing class imbalances and enhancing the model's robustness.

Overall, this paper provides valuable insights into the application of machine learning in medical imaging and severity assessment, reinforcing the role of quantitative imaging features in enhancing the precision of automated diagnostic tools.