Diagnosis of COVID-19 Using CT image Radiomics Features: A Comprehensive Machine Learning Study Involving 26,307 Patients (preprint)

PurposeTo derive and validate an effective radiomics-based model for differentiation of COVID-19 pneumonia from other lung diseases using a very large cohort of patients. MethodsWe collected 19 private and 5 public datasets, accumulating to 26,307 individual patient images (15,148 COVID-19; 9,657 with other lung diseases e.g. non-COVID-19 pneumonia, lung cancer, pulmonary embolism; 1502 normal cases). Images were automatically segmented using a validated deep learning (DL) model and the results carefully reviewed. Images were first cropped into lung-only region boxes, then resized to 296x216 voxels. Voxel dimensions was resized to 1x1x1mm3 followed by 64-bin discretization. The 108 extracted features included shape, first-order histogram and texture features. Univariate analysis was first performed using simple logistic regression. The thresholds were fixed in the training set and then evaluation performed on the test set. False discovery rate (FDR) correction was applied to the p-values. Z-Score normalization was applied to all features. For multivariate analysis, features with high correlation (R2>0.99) were eliminated first using Pearson correlation. We tested 96 different machine learning strategies through cross-combining 4 feature selectors or 8 dimensionality reduction techniques with 8 classifiers. We trained and evaluated our models using 3 different datasets: 1) the entire dataset (26,307 patients: 15,148 COVID-19; 11,159 non-COVID-19); 2) excluding normal patients in non-COVID-19, and including only RT-PCR positive COVID-19 cases in the COVID-19 class (20,697 patients including 12,419 COVID-19, and 8,278 non-COVID-19)); 3) including only non-COVID-19 pneumonia patients and a random sample of COVID-19 patients (5,582 patients: 3,000 COVID-19, and 2,582 non-COVID-19) to provide balanced classes. Subsequently, each of these 3 datasets were randomly split into 70% and 30% for training and testing, respectively. All various steps, including feature preprocessing, feature selection, and classification, were performed separately in each dataset. Classification algorithms were optimized during training using grid search algorithms. The best models were chosen by a one-standard-deviation rule in 10-fold cross-validation and then were evaluated on the test sets. ResultsIn dataset #1, Relief feature selection and RF classifier combination resulted in the highest performance (Area under the receiver operating characteristic curve (AUC) = 0.99, sensitivity = 0.98, specificity = 0.94, accuracy = 0.96, positive predictive value (PPV) = 0.96, and negative predicted value (NPV) = 0.96). In dataset #2, Recursive Feature Elimination (RFE) feature selection and Random Forest (RF) classifier combination resulted in the highest performance (AUC = 0.99, sensitivity = 0.98, specificity = 0.95, accuracy = 0.97, PPV = 0.96, and NPV = 0.98). In dataset #3, the ANOVA feature selection and RF classifier combination resulted in the highest performance (AUC = 0.98, sensitivity = 0.96, specificity = 0.93, accuracy = 0.94, PPV = 0.93, NPV = 0.96). ConclusionRadiomic features extracted from entire lung combined with machine learning algorithms can enable very effective, routine diagnosis of COVID-19 pneumonia from CT images without the use of any other diagnostic test.

COVID-19 Prognostic Modeling Using CT Radiomic Features and Machine Learning Algorithms: Analysis of a Multi-Institutional Dataset of 14,339 Patients (preprint)

Objective In this large multi-institutional study, we aimed to analyze the prognostic power of computed tomography (CT)-based radiomics models in COVID-19 patients. Methods CT images of 14,339 COVID-19 patients with overall survival outcome were collected from 19 medical centers. Whole lung segmentations were performed automatically using a previously validated deep learning-based model, and regions of interest were further evaluated and modified by a human observer. All images were resampled to an isotropic voxel size, intensities were discretized into 64-binning size, and 105 radiomics features, including shape, intensity, and texture features were extracted from the lung mask. Radiomics features were normalized using Z-score normalization. High-correlated features using Pearson (R 2 >0.99) were eliminated. We applied the Synthetic Minority Oversampling Technique (SMOT) algorithm in only the training set for different models to overcome unbalance classes. We used 4 feature selection algorithms, namely Analysis of Variance (ANOVA), Kruskal- Wallis (KW), Recursive Feature Elimination (RFE), and Relief. For the classification task, we used seven classifiers, including Logistic Regression (LR), Least Absolute Shrinkage and Selection Operator (LASSO), Linear Discriminant Analysis (LDA), Random Forest (RF), AdaBoost (AB), Naïve Bayes (NB), and Multilayer Perceptron (MLP). The models were built and evaluated using training and testing sets, respectively. Specifically, we evaluated the models using 10 different splitting and cross-validation strategies, including different types of test datasets (e.g. non-harmonized vs. ComBat-harmonized datasets). The sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC) were reported for models evaluation. Results In the test dataset (4301) consisting of CT and/or RT-PCR positive cases, AUC, sensitivity, and specificity of 0.83±0.01 (CI95%: 0.81-0.85), 0.81, and 0.72, respectively, were obtained by ANOVA feature selector + RF classifier. In RT-PCR-only positive test sets (3644), similar results were achieved, and there was no statistically significant difference. In ComBat harmonized dataset, Relief feature selector + RF classifier resulted in highest performance of AUC, reaching 0.83±0.01 (CI95%: 0.81-0.85), with sensitivity and specificity of 0.77 and 0.74, respectively. At the same time, ComBat harmonization did not depict statistically significant improvement relevant to non-harmonized dataset. In leave-one-center-out, the combination of ANOVA feature selector and LR classifier resulted in the highest performance of AUC (0.80±0.084) with sensitivity and specificity of 0.77 ± 0.11 and 0.76 ± 0.075, respectively. Conclusion Lung CT radiomics features can be used towards robust prognostic modeling of COVID-19 in large heterogeneous datasets gathered from multiple centers. As such, CT radiomics-based model has significant potential for use in prospective clinical settings towards improved management of COVID-19 patients.

COLI-NET: Fully Automated COVID-19 Lung and Infection Pneumonia Lesion Detection and Segmentation from Chest CT Images (preprint)

BackgroundWe present a deep learning (DL)-based automated whole lung and COVID-19 pneumonia infectious lesions (COLI-Net) detection and segmentation from chest CT images. MethodsWe prepared 2358 (347259, 2D slices) and 180 (17341, 2D slices) volumetric CT images along with their corresponding manual segmentation of lungs and lesions, respectively, in the framework of a multi-center/multi-scanner study. All images were cropped, resized and the intensity values clipped and normalized. A residual network (ResNet) with non-square Dice loss function built upon TensorFlow was employed. The accuracy of lung and COVID-19 lesions segmentation was evaluated on an external RT-PCR positive COVID-19 dataset (7333, 2D slices) collected at five different centers. To evaluate the segmentation performance, we calculated different quantitative metrics, including radiomic features. ResultsThe mean Dice coefficients were 0.98{+/-}0.011 (95% CI, 0.98-0.99) and 0.91{+/-}0.038 (95% CI, 0.90-0.91) for lung and lesions segmentation, respectively. The mean relative Hounsfield unit differences were 0.03{+/-}0.84% (95% CI, -0.12 - 0.18) and -0.18{+/-}3.4% (95% CI, -0.8 - 0.44) for the lung and lesions, respectively. The relative volume difference for lung and lesions were 0.38{+/-}1.2% (95% CI, 0.16-0.59) and 0.81{+/-}6.6% (95% CI, -0.39-2), respectively. Most radiomic features had a mean relative error less than 5% with the highest mean relative error achieved for the lung for the Range first-order feature (- 6.95%) and least axis length shape feature (8.68%) for lesions. ConclusionWe set out to develop an automated deep learning-guided three-dimensional whole lung and infected regions segmentation in COVID-19 patients in order to develop fast, consistent, robust and human error immune framework for lung and pneumonia lesion detection and quantification.

COVID-19 Antibody Seroprevalence Among General Population and High-Risk Occupational Groups in Eighteen Cities of Iran (preprint)

Background: As the COVID-19 pandemic unfolded, rapid case increase was observed in multiple cities in Iran. However, in the absence of seroprevalence surveys, the true infection rate remains unknown. In this population-based study we assessed the seroprevalence of SARS-CoV-2 antibodies in eighteen cities of Iran.Methods: We randomly selected and invited study participants from the general population (N = 3,547) and occupations with high risk of COVID-19 exposure, defined as high-risk population (e.g., supermarket employees) (N = 5,391), in eighteen cities of Iran. SARS-CoV-2 ELISA kits were used to detect antibody against COVID-19. Crude, population weight adjusted, and test performance adjusted seroprevalence rates were estimated.Findings: The population weight adjusted and test performance adjusted prevalence rates of antibody seropositivity in general population were 13·1% (95% CI 11·6-14·8%) and 18·5% (95% CI 16·1-21·3%), respectively. The population-weighted seroprevalence estimate implies that 3,290,633 (95% CI 2,907185-3,709,167) individuals, from the eighteen included cities in this study, were infected by end of April 2020.The overall prevalence rate was higher among individuals aged ≥ 60 years (32·0%, 95% CI 23·9-40·8%) and with comorbidity condition (23·7%, 95% CI 18·5-28·8%). The estimated seroprevalence of SARS-CoV-2 antibodies varied greatly by city and the highest population test-adjusted prevalence rates were in Rasht 78·1% (95% CI 58·3-98·3%) and Qom (66·5%, 95% CI 39·9-95·4%) cities. The test-adjusted prevalence did not differ between low and high-risk populations and was about 20.0%.Interpretations: The findings of this study imply that prevalence of seropositivity is likely much higher than the reported prevalence rates based on confirmed COVID-19 cases in Iran. Despite the high seroprevalence rates in a few cities, the low overall prevalence estimates indicate that a large proportion of population is still susceptible to the virus. The similar seroprevalence estimates between low and high-risk occupations might be an indicator of inadequate or low adherence to infection control measures among general population.Funding Statement: Iranian Ministry of Health and Medical Education COVID-19 Grant (number 99-1-97-47964).Declaration of Interests: None to disclose.Ethics Approval Statement: Ethics approval for this study was granted by Vice-Chancellor in Research Affairs-Tehran University of Medical Sciences (IR. TUMS.VCR.REC.1399.308)

Screening COVID-19 Based on CT/CXR Images & Building a Publicly Available CT-scan Dataset of COVID-19 (preprint)

The rapid outbreak of COVID-19 threatens humans life all around the world. Due to insufficient diagnostic infrastructures, developing an accurate, efficient, inexpensive, and quick diagnostic tool is of great importance. As chest radiography, such as chest X-ray (CXR) and CT computed tomography (CT), is a possible way for screening COVID-19, developing an automatic image classification tool is immensely helpful for detecting the patients with COVID-19. To date, researchers have proposed several different screening methods; however, none of them could achieve a reliable and highly sensitive performance yet. The main drawbacks of current methods are the lack of having enough training data, low generalization performance, and a high rate of false-positive detection. To tackle such limitations, this study firstly builds a large-size publicly available CT-scan dataset, consisting of more than 13k CT-images of more than 1000 individuals, in which 8k images are taken from 500 patients infected with COVID-19. Secondly, we propose a deep learning model for screening COVID-19 using our proposed CT dataset and report the baseline results. Finally, we extend the proposed CT model for screening COVID-19 from CXR images using a transfer learning approach. The experimental results show that the proposed CT and CXR methods achieve the AUC scores of 0.886 and 0.984 respectively.

Clinical characteristics, outcomes, and risk factors for mortality in hospitalized patients with COVID-19 and cancer history: a propensity score-matched study (preprint)

Background: COVID-19 has caused great concern for patients with underlying medical conditions. We aimed to determine the prognosis of patients with current or previous cancer with either a PCR-confirmed COVID-19 infection or a probable diagnosis according to chest CT scan. Methods: We conducted a case control study in a referral hospital on confirmed COVID-19 adult patients with and without a history of cancer from February 25th to April 21st , 2020. Patients were matched according to age, gender, and underlying diseases. Demographic features, clinical and Para clinical data have been extracted from medical records. Multivariable logistic regression was used to estimate odd ratios and 95% confidence intervals of each factor of interest with outcomes. Results: Fifty-three confirmed COVID-19 patients with history of cancer were recruited and compared with 106 non-cancerous COVID-19 patients. Male to female ratio was 1.33 and 45% were older than 65. Dyspnea was significantly associated with an increased rate of mortality in the cancer subgroup (p=0.013). Twenty-six patients (49%) survived among the cancer group while 89 patients (84%) survived in control (p=0.000). Patients with hematologic cancer had 63% mortality while those with solid tumors had 37%. Multivariate analysis showed that cancer, impaired consciousness, tachypnea, tachycardia, leukocytosis and thrombocytopenia were associated with an increased risk of death. Conclusion: Cancer increased mortality rate and hospital stay of COVID-19 patients and remained significant after adjustment of confounders. Compared to solid tumors, hematologic malignancies have been associated with worse consequences and higher mortality. Clinical and Para clinical indicators were not appropriate to predict death.

COVID-19 and heart medications: What's the connection? (preprint)

Aim: To determine association between clinical outcome of COVID-19 and prior usage of cardiovascular and metabolic drugs, including, Aspirin, ACEIs, ARBs, Clopidogrel, metformin, and Statins. Methods: Statistical examination of the demographic, clinical, laboratory and imaging features of 353 patients with SARS-CoV-2 disease admitted from February to April 2020. Result: Minor discrepancies were observed in the clinical presentations, radiologic involvement and laboratory results across groups of patients under treatment with specific drugs. Aspirin-users had better clinical outcome with lower need of ventilation support, whereas, metformin-users had increased chance of intubation and of mortality. Conclusion: Although not being conclusive, our findings suggest the possibility of the effect of previous drug usages on the various presentations and clinical course of COVID-19 infection.

CovidCTNet: An Open-Source Deep Learning Approach to Identify Covid-19 Using CT Image (preprint)

Coronavirus disease 2019 (Covid-19) is highly contagious with limited treatment options. Early and accurate diagnosis of Covid-19 is crucial in reducing the spread of the disease and its accompanied mortality. Currently, detection by reverse transcriptase polymerase chain reaction (RT-PCR) is the gold standard of outpatient and inpatient detection of Covid-19. RT-PCR is a rapid method, however, its accuracy in detection is only ~70-75%. Another approved strategy is computed tomography (CT) imaging. CT imaging has a much higher sensitivity of ~80-98%, but similar accuracy of 70%. To enhance the accuracy of CT imaging detection, we developed an open-source set of algorithms called CovidCTNet that successfully differentiates Covid-19 from community-acquired pneumonia (CAP) and other lung diseases. CovidCTNet increases the accuracy of CT imaging detection to 90% compared to radiologists (70%). The model is designed to work with heterogeneous and small sample sizes independent of the CT imaging hardware. In order to facilitate the detection of Covid-19 globally and assist radiologists and physicians in the screening process, we are releasing all algorithms and parametric details in an open-source format. Open-source sharing of our CovidCTNet enables developers to rapidly improve and optimize services, while preserving user privacy and data ownership.