ABSTRACT
OBJECTIVE: As hospital admission rate is high during the COVID-19 pandemic, hospital length of stay (LOS) is a key indicator of medical resource allocation. This study aimed to elucidate specific dynamic longitudinal computed tomography (CT) imaging changes for patients with COVID-19 over in-hospital and predict individual LOS of COVID-19 patients with Delta variant of SARS-CoV-2 using the machine learning method. MATERIALS AND METHODS: This retrospective study recruited 448 COVID-19 patients with a total of 1761 CT scans from July 14, 2021 to August 20, 2021 with an averaged hospital LOS of 22.5 ± 7.0 days. Imaging features were extracted from each CT scan, including CT morphological characteristics and artificial intelligence (AI) extracted features. Clinical features were obtained from each patient's initial admission. The infection distribution in lung fields and progression pattern tendency was analyzed. Then, to construct a model to predict patient LOS, each CT scan was considered as an independent sample to predict the LOS from the current CT scan time point to hospital discharge combining with the patients' corresponding clinical features. The 1761 follow-up CT data were randomly split into training set and testing set with a ratio of 7:3 at patient-level. A total of 85 most related clinical and imaging features selected by Least Absolute Shrinkage and Selection Operator were used to construct LOS prediction model. RESULTS: Infection-related features were obtained, such as the percentage of the infected region of lung, ground-glass opacity (GGO), consolidation and crazy-paving pattern, and air bronchograms. Their longitudinal changes show that the progression changes significantly in the earlier stages (0–3 days to 4–6 days), and then, changes tend to be statistically subtle, except for the intensity range between (−470 and −70) HU which exhibits a significant increase followed by a continuous significant decrease. Furthermore, the bilateral lower lobes, especially the right lower lobe, present more severe. Compared with other models, combining the clinical, imaging reading, and AI features to build the LOS prediction model achieved the highest R2 of 0.854 and 0.463, Pearson correlation coefficient of 0.939 and 0.696, and lowest mean absolute error of 2.405 and 4.426, and mean squared error of 9.176 and 34.728 on the training and testing set. CONCLUSION: The most obvious progression changes were significantly in the earlier stages (0–3 days to 4–6 days) and the bilateral lower lobes, especially the right lower lobe. GGO, consolidation, and crazy-paving pattern and air bronchograms are the most main CT findings according to the longitudinal changes of infection-related features with LOS (day). The LOS prediction model of combining clinical, imaging reading, and AI features achieved optimum performance.
ABSTRACT
BACKGROUND: This study aimed to build a radiomics model with deep learning (DL) and human auditing and examine its diagnostic value in differentiating between coronavirus disease 2019 (COVID-19) and community-acquired pneumonia (CAP). METHODS: Forty-three COVID-19 patients, whose diagnoses had been confirmed with reverse-transcriptase polymerase-chain-reaction (RT-PCR) tests, and 60 CAP patients, whose diagnoses had been confirmed with sputum cultures, were enrolled in this retrospective study. The candidate regions of interest (ROIs) on the computed tomography (CT) images of the 103 patients were determined using a DL-based segmentation model powered by transfer learning. These ROIs were manually audited and corrected by 3 radiologists (with an average of 12 years of experience; range 6-17 years) to check the segmentation acceptance for the radiomics analysis. ROI-derived radiomics features were subsequently extracted to build the classification model and processed using 4 different algorithms (L1 regularization, Lasso, Ridge, and Z test) and 4 classifiers, including the logistic regression (LR), multi-layer perceptron (MLP), support vector machine (SVM), and extreme Gradient Boosting (XGboost). A receiver operating characteristic curve (ROC) analysis was conducted to evaluate the performance of the model. RESULTS: Quantitative CT measurements derived from human-audited segmentation results showed that COVID-19 patients had significantly decreased numbers of infected lobes compared to patients in the CAP group {median [interquartile range (IQR)]: 4 [3, 4] and 4 [4, 5]; P=0.031}. The infected percentage (%) of the whole lung was significantly more elevated in the CAP group [6.40 (2.77, 11.11)] than the COVID-19 group [1.83 (0.65, 4.42); P<0.001], and the same trend applied to each lobe, except for the superior lobe of the right lung [1.81 (0.09, 5.28) for COVID-19 vs. 1.32 (0.14, 7.02) for CAP; P=0.649]. Additionally, the highest proportion of infected lesions were observed in the CT value range of (-470, -370) Hounsfield units (HU) in the COVID-19 group. Conversely, the CAP group had a value range of (30, 60) HU. Radiomic model using corrected ROIs exhibited the highest area under ROC (AUC) of 0.990 [95% confidence interval (CI): 0.962-1.000] using Lasso for feature selection and MLP for classification. CONCLUSIONS: The proposed radiomics model based on human-audited segmentation made accurate differential diagnoses of COVID-19 and CAP. The quantification of CT measurements derived from DL could potentially be used as effective biomarkers in current clinical practice.