Computed tomography-based COVID-19 triage through a deep neural network using mask-weighted global average pooling.

.Purpose: Coronavirus disease 2019 (COVID-19) is a new infection that has spread worldwide and with no automatic model to reliably detect its presence from images. We aim to investigate the potential of deep transfer learning to predict COVID-19 infection using chest computed tomography (CT) and x-ray images.Approach: Regions of interest (ROI) corresponding to ground-glass opacities (GGO), consolidations, and pleural effusions were labeled in 100 axial lung CT images from 60 COVID-19-infected subjects. These segmented regions were then employed as an additional input to six deep convolutional neural network (CNN) architectures (AlexNet, DenseNet, GoogleNet, NASNet-Mobile, ResNet18, and DarkNet), pretrained on natural images, to differentiate between COVID-19 and normal CT images. We also explored the model’s ability to classify x-ray images as COVID-19, non-COVID-19 pneumonia, or normal. Performance on test images was measured with global accuracy and area under the receiver operating characteristic curve (AUC).Results: When using raw CT images as input to the tested models, the highest accuracy of 82% and AUC of 88.16% is achieved. Incorporating the three ROIs as an additional model inputs further boosts performance to an accuracy of 82.30% and an AUC of 90.10% (DarkNet). For x-ray images, we obtained an outstanding AUC of 97% for classifying COVID-19 versus normal versus other. Combing chest CT and x-ray images, DarkNet architecture achieves the highest accuracy of 99.09% and AUC of 99.89% in classifying COVID-19 from non-COVID-19. Our results confirm the ability of deep CNNs with transfer learning to predict COVID-19 in both chest CT and x-ray images.Conclusions: The proposed method could help radiologists increase the accuracy of their diagnosis and increase efficiency in COVID-19 management.

Background: Coronavirus Disease 2019 (COVID-19) is associated with an increased risk for acute ischemic stroke but screening for COVID-19 based on clinical criteria or laboratory testing may be difficult in acute stroke evaluation. Chest computed tomographic (CT) scan may be another time-sensitive option for identification. of COVID-19 in such patients. Objective: We report our experience of incorporating chest CT scan in the initial neuroimaging protocol for evaluation of acute stroke patients. Methods: All acute stroke patients underwent chest CT scan concurrent to CT head, CT angiogram of head and neck and CT perfusion for 4 months. We identified patients who had chest CT scan findings that were suggestive of COVID-19 including bilateral, multilobar ground glass opacification with a peripheral or posterior distribution, and/or consolidation (mainly in the lower lobes). All patients subsequently underwent polymerase chain reaction (PCR) testing of nasopharyngeal swab with contact isolation until COVID-19 could be excluded. Results: A total of 224 consecutive patients (mean age 62.12 years±SD; 15.3) underwent acute stroke evaluation with a concurrent chest CT scan. The chest CT identified findings suggestive of COVID-19 in 11 (4.9%) patients. Subsequent PCR testing did not confirm the diagnosis of COVID-19 in any of the patient. Another 99 patients (44%) without any findings suggestive of COVID-19 on chest CT scan underwent PCR testing. PCR testing did not confirm the diagnosis of COVID-19 in any of the patients. Four patients (4.3%) with chest CT scan findings suggestive of COVID-19 were found to have an ischemic stroke while 7 patients (5.9%) with chest CT scan findings suggestive of COVID-19 did not have any ischemic stroke (stroke mimic). Conclusions: We found a very low yield for identifying COVID-19 in acute stroke patients by performing chest CT scan concurrent to standard acute stroke neuroimaging protocol.

status: Published

Background and Aims Non-contrast chest computed tomography (CT) scans can accurately evaluate the type and extent of lung lesions. The aim of this study was to investigate the chest CT features associated with critical and non-critical patients with coronavirus disease 2019 (COVID-19). Methods A total of 1078 patients with COVID-19 pneumonia who underwent chest CT scans, including 169 critical cases and 909 non-critical cases, were enrolled in this retrospective study. The scans of all participants were reviewed and compared in two groups of study. In addition, the risk factors associated with disease in critical and non-critical patients were analyzed. Results Chest CT scans showed bilateral and multifocal involvement in most (86.4%) of the participants, with 97.6 and 84.3% reported in critical and non-critical patients, respectively. The incidences of pure consolidation (p=0.019), mixed ground-glass opacities (GGOs) and consolidation (p<0.001), pleural effusion (p<0.001), and intralesional traction bronchiectasis (p=0.007) were significantly higher in critical compared to non-critical patients. However, non-critical patients showed higher incidence of pure GGOs than the critical patients (p<0.001). Finally, the total opacity scores of the critical patients were significantly higher than those of non-critical patients (13.71±6.26 versus 4.86±3.52, p<0.001), with an area under the curve of 0.91 (0.88-0.94) for COVID-19 detection. Conclusions Our results revealed that the chest CT examination was an effective means of detecting pulmonary parenchymal abnormalities in the natural course of COVID-19. It can distinguish the critical patients from the non-critical patients (AUC=0.91), which is helpful for the judgment of clinical condition and has important clinical value for the diagnosis and follow-up of COVID-19 pneumonia.

Notwithstanding the extensive research efforts directed towards devising a dependable approach for the diagnosis of coronavirus disease 2019 (COVID-19), the inherent complexity and capriciousness of the virus continue to pose a formidable challenge to the precise identification of affected individuals. In light of this predicament, it is essential to devise a model for COVID-19 prediction utilizing chest computed tomography (CT) scans. To this end, we present a hybrid quantum-classical convolutional neural network (HQCNN) model, which is founded on stochastic quantum circuits that can discern COVID-19 patients from chest CT images. Two publicly available chest CT image datasets were employed to evaluate the performance of our model. The experimental outcomes evinced diagnostic accuracies of 99.39% and 97.91%, along with precisions of 99.19% and 98.52%, respectively. These findings are indicative of the fact that the proposed model surpasses recently published works in terms of performance, thus providing a superior ability to precisely predict COVID-19 positive instances. Communicated by Ramaswamy H. Sarma

Background: Coronavirus disease 2019 (COVID-19) has since spread globally causing the ongoing pandemic. However, there are few studies on the symptoms and outcomes of pregnant women with COVID-19 and their newborns. Limited data are available comparing pregnant women with and without COVID-19 pneumonia. Our study aimed to find symptoms unique in pregnancy to help the early detection of COVID-19 in pregnant women and to evaluate its impact on perinatal women and their newborns. Methods: Clinical data were reviewed and collected for 11 pregnant women with laboratory-confirmed COVID-19 (pregnant confirmed cohort) who were consecutively admitted to Zhongnan Hospital of Wuhan University and Maternal and Child Health Hospital of Hubei Province, from January 26 to February 26, 2020. They were subsequently matched by age with 11 pregnant women suspected of COVID-19 (pregnant suspected group) who had typical changes in chest computed tomography (CT) scans but were repeatedly negative for reverse-transcriptase-polymerase-chain-reaction (RT-PCR), with 11 pregnant women who were not infected (pregnant non-infected group), and with 44 non-pregnant women with confirmed COVID-19 (Non-pregnant confirmed group). Results: The significant symptoms in pregnant confirmed women were mild fever and fatigue. Both the confirmed and suspected women had a fever (45.5% vs. 72.7%, P &gt; 0.05) which were barely distinguishable. The chest CT scans of confirmed women showed unspecific manifestations, such as bilateral pleural effusions and slightly increased densities. Eight of eleven (72.7%) confirmed women did not feel anything unusual until abnormalities were found on chest CT scans on the admission screening test. All three groups had elevated white blood cell count, neutrophil count, lactate dehydrogenase, and reduced total protein. Compared to the normal group, women in the confirmed group had significantly reduced hemoglobin, lymphocyte, albumin, and increased lactate dehydrogenase (P &lt; 0.05). The infection did not increase the risk of premature delivery, premature rupture of the membrane, or comorbidities in pregnancy. The neonates of the three groups of mothers had similar symptoms, except for those of the confirmed group had increased heart rate (130–143 bpm) and mean arterial pressure (93–112 mmHg) than those born to the normal group. The infection of the mother did lead to more complications in their neonates. And there was no evidence of vertical transmission during late pregnancy. Conclusions: Pregnant women were often asymptomatic and had incidental abnormal findings on a chest CT scan on admission. Screening of CT scans in disease prevalent areas is thus of crucial importance. Even after the laboratory confirmation, the manifestation of the CT scan could be atypical, which alerted the necessity of protection for healthcare workers. COVID-19 did not increase the risk of complications in pregnant women and their neonates.

This study aimed to explore the diagnostic value of chest computed tomography (CT) based on adaptive statistical iterative reconstruction (ASiR) in the optimal weight range for children with community-acquired pneumonia (CAP) and find the differences between chest CT and ultrasound in the diagnosis of CAP. 106 children who were diagnosed as CAP were selected as the research objects. All objects underwent CT and ultrasound scans to measure the noise of CT images after iterative reconstruction and the contrast-to-noise ratio (CNR) of the thoracic aorta and the back muscles. The correlation between ASiR weight, noise, and CNR was analyzed and compared further. In addition, the differences in imaging manifestations of lung consolidation, abnormal pleural line, alveolar interstitial syndrome, and pleural effusion were observed under the ultrasound and CT scans of the patients. The sensitivity, specificity, positive predicted value (PV), and negative PV of the CT scan and ultrasound were analyzed further for the diagnosis of CAP in children. The results showed that ASiR weight and image noise were extremely and negatively correlated ( P &lt; 0.001 ) and that CNR and ASiR weight were extremely and positively correlated ( P &lt; 0.001 ). 40–60% of ASiR image noise was in the middle position, and the muscle space and subcutaneous fat in CT images were clearer when ASiR was at 40% and 60%. 60.38% of children were diagnosed as CAP. The proportions of children with lung consolidation, interstitial changes, pleural effusion, and atelectasis after ultrasound and CT examination were not greatly different ( P &gt; 0.05 ). After ASiR, the sensitivity of CT in detecting the abnormal pleural line + B-line/consolidation feature was the highest, 93.75%. The results of ultrasound diagnosis were consistent with those of chest CT diagnosis ( P &lt; 0.01 ). This indicated that the sensitivity, specificity, positive PV, and negative PV of ASiR-based chest CT in the diagnosis of CAP were close to those of ultrasound.

Background: The aim of this study was to evaluate the performance of an automated COVID-19 detection method based on a transfer learning technique that makes use of chest computed tomography (CT) images. Method: In this study, we used a publicly available multiclass CT scan dataset containing 4171 CT scans of 210 different patients. In particular, we extracted features from the CT images using a set of convolutional neural networks (CNNs) that had been pretrained on the ImageNet dataset as feature extractors, and we then selected a subset of these features using the Information Gain filter. The resulting feature vectors were then used to train a set of k Nearest Neighbors classifiers with 10-fold cross validation to assess the classification performance of the features that had been extracted by each CNN. Finally, a majority voting approach was used to classify each image into two different classes: COVID-19 and NO COVID-19. Results: A total of 414 images of the test set (10% of the complete dataset) were correctly classified, and only 4 were misclassified, yielding a final classification accuracy of 99.04%. Conclusions: The high performance that was achieved by the method could make it feasible option that could be used to assist radiologists in COVID-19 diagnosis through the use of CT images.

The spread of abnormal opacity on chest computed tomography (CT) has been reported as a predictor of coronavirus disease 2019 (COVID-19) severity; however, the relationship between CT findings and prognosis in patients with severe COVID-19 remains unclear. The objective of this study was to evaluate the extent of abnormal opacity on chest CT and its association with prognosis in patients with COVID-19 in a critical care medical center, using a simple semi-quantitative method. This single-center case-control study included patients diagnosed with severe COVID-19 pneumonia who were admitted to a critical care center. The diagnosis of COVID-19 was based on positive results of a reverse transcription polymerase chain reaction test. All patients underwent non-contrast whole-body CT upon admission. Six representative axial chest CT images were selected for each patient to evaluate the extent of lung lesions. The percentage of the area involved in the representative CT images was visually assessed by 2 radiologists and scored on 4-point scale to obtain the bedside CT score, which was compared between patients who survived and those who died using the Mann–Whitney U test. A total of 63 patients were included in this study: 51 survived and 12 died after intensive treatment. The inter-rater reliability of bedside scores between the 2 radiologists was acceptable. The median bedside CT score of the survival group was 12.5 and that of the mortality group was 16.5; the difference between the 2 groups was statistically significant. The degree of opacity can be easily scored using representative CT images in patients with severe COVID-19 pneumonia, without sophisticated software. A greater extent of abnormal opacity is associated with poorer prognosis. Predicting the prognosis of patients with severe COVID-19 could facilitate prompt and appropriate treatment.

Pregnant women and children sometimes had to undergo chest computed tomography (CT) scans during the Corona Virus Disease 2019 (COVID-19) pandemic. This study estimated the fetal and pediatric doses from chest CT scans. Organ doses and effective doses were calculated using the VirtualDose-CT software. Two groups of computational human phantoms, pregnant females and pediatric patients were used in this study. The results of doses normalized to volumetric CT Dose Index (CTDIvol) can be used universally for other dosimetry studies. Based on our calculations and international survey data of CTDIvol, fetal absorbed doses from COVID-19-related chest CT were found to be 0.04-0.36, 0.05-0.44 and 0.07-0.61mGy for 3, 6 and 9months of pregnancy, respectively. When the scan range is extended to the abdominal region, fetal doses increase by almost 4-fold. Effective doses for COVID-19-related chest CT were 1.62-13.77, 1.58-13.46, 1.57-13.33 and 1.29-10.98mSv for the newborn, 1-, 5- and 10-y-old children, respectively. In addition, the effects of specific axial scan ranges exceeding the thorax region were evaluated. Although doses from chest CT scans are small, such data allow radiologists and patients to be informed of the dose levels and ways to avoid unnecessary radiation.

The coronavirus disease 2019 (COVID-19) pandemic had a major impact on global health and was associated with millions of deaths worldwide. During the pandemic, imaging characteristics of chest X-ray (CXR) and chest computed tomography (CT) played an important role in the screening, diagnosis and monitoring the disease progression. Various studies suggested that quantitative image analysis methods including artificial intelligence and radiomics can greatly boost the value of imaging in the management of COVID-19. However, few studies have explored the use of longitudinal multi-modal medical images with varying visit intervals for outcome prediction in COVID-19 patients. This study aims to explore the potential of longitudinal multimodal radiomics in predicting the outcome of COVID-19 patients by integrating both CXR and CT images with variable visit intervals through deep learning. 2274 patients who underwent CXR and/or CT scans during disease progression were selected for this study. Of these, 946 patients were treated at the University of Pennsylvania Health System (UPHS) and the remaining 1328 patients were acquired at Stony Brook University (SBU) and curated by the Medical Imaging and Data Resource Center (MIDRC). 532 radiomic features were extracted with the Cancer Imaging Phenomics Toolkit (CaPTk) from the lung regions in CXR and CT images at all visits. We employed two commonly used deep learning algorithms to analyze the longitudinal multimodal features, and evaluated the prediction results based on the area under the receiver operating characteristic curve (AUC). Our models achieved testing AUC scores of 0.816 and 0.836, respectively, for the prediction of mortality.

The ongoing global pandemic of Coronavirus Disease 2019 (COVID-19) poses a serious threat to public health and the economy. Rapid and accurate diagnosis of COVID-19 is essential to prevent the further spread of the disease and reduce its mortality. Chest Computed tomography (CT) is an effective tool for the early diagnosis of lung diseases including pneumonia. However, detecting COVID-19 from CT is demanding and prone to human errors as some early-stage patients may have negative findings on images. Recently, many deep learning methods have achieved impressive performance in this regard. Despite their effectiveness, most of these methods underestimate the rich spatial information preserved in the 3D structure or suffer from the propagation of errors. To address this problem, we propose a Dual-Attention Residual Network (DARNet) to automatically identify COVID-19 from other common pneumonia (CP) and healthy people using 3D chest CT images. Specifically, we design a dual-attention module consisting of channel-wise attention and depth-wise attention mechanisms. The former is utilized to enhance channel independence, while the latter is developed to recalibrate the depth-level features. Then, we integrate them in a unified manner to extract and refine the features at different levels to further improve the diagnostic performance. We evaluate DARNet on a large public CT dataset and obtain superior performance. Besides, the ablation study and visualization analysis prove the effectiveness and interpretability of the proposed method.

On morning rounds, a medical resident, 6 months into her training, proudly admitted, ‘‘I have never seen a real chest x-ray (CXR) before.’’ Although she seemed familiar and competent with the interpretation of the chest computed tomographic (CT) scan, her knowledge in CXR readings paled dismally in comparison. The era of plain CXR seems to be rapidly transforming into the sophisticated era of chest CT. Too many chest CT scans are ordered and taken above and beyond CXRs. Daily morning rounds now may not begin with CXRs. Radiology conferences rarely refer to CXRs. We wonder, ‘‘Why even order the lateral view?’’ In the last few years we have not encountered a clinician voluntarily pulling up a lateral view for inspection. On 1 occasion, my curiosity received a comment from a radiologist, ‘‘CXR is a test of low sensitivity.’’ I have always wondered whose sensitivity we were discussing. Is it the reimbursement or defensive medical practice? Whatever the cause, in most hospitals, a chest CT is quicker to obtain than a CXR. A patient entering the emergency room with shortness of breath invariably gets a CT angiogram almost as a screening test. A CT scan is commonly ordered for a patient with simple community-acquired pneumonia, than a superfluous CXR. More and more we see a CT scan ordered to confirm the position of chest tubes and on a weekly basis to follow the course of pulmonary illness. Have we collectively and deliberately killed the time-tested CXR? With the advent of computer technology, with its ever-expanding capabilities, we have developed the ‘‘digital watch syndrome’’: the inability to recognize the significance of the short and long arms on the clock face. The old-fashioned analytical approach, proper interpretation of a low-cost test before ordering an expensive one, seems to have fallen out of favor. I am certain that readers with this perspective will feel that our views are too old fashioned and that we are fighting a losing battle.

The recent medical applications of deep-learning (DL) algorithms have demonstrated their clinical efficacy in improving speed and accuracy of image interpretation. If the DL algorithm achieves a performance equivalent to that achieved by physicians in chest radiography (CR) diagnoses with Coronavirus disease 2019 (COVID-19) pneumonia, the automatic interpretation of the CR with DL algorithms can significantly reduce the burden on clinicians and radiologists in sudden surges of suspected COVID-19 patients. The aim of this study was to evaluate the efficacy of the DL algorithm for detecting COVID-19 pneumonia on CR compared with formal radiology reports. This is a retrospective study of adult patients that were diagnosed as positive COVID-19 cases based on the reverse transcription polymerase chain reaction among all the patients who were admitted to five emergency departments and one community treatment center in Korea from February 18, 2020 to May 1, 2020. The CR images were evaluated with a publicly available DL algorithm. For reference, CR images without chest computed tomography (CT) scans classified as positive for COVID-19 pneumonia were used given that the radiologist identified ground-glass opacity, consolidation, or other infiltration in retrospectively reviewed CR images. Patients with evidence of pneumonia on chest CT scans were also classified as COVID-19 pneumonia positive outcomes. The overall sensitivity and specificity of the DL algorithm for detecting COVID-19 pneumonia on CR were 95.6%, and 88.7%, respectively. The area under the curve value of the DL algorithm for the detection of COVID-19 with pneumonia was 0.921. The DL algorithm demonstrated a satisfactory diagnostic performance comparable with that of formal radiology reports in the CR-based diagnosis of pneumonia in COVID-19 patients. The DL algorithm may offer fast and reliable examinations that can facilitate patient screening and isolation decisions, which can reduce the medical staff workload during COVID-19 pandemic situations.

ObjectivesTo compare the performance of chest computed tomography (CT) scan versus reverse transcription polymerase chain reaction (RT-PCR) as the reference standard in the initial diagnostic assessment of coronavirus disease 2019 (COVID-19) patients.DesignA systematic review and meta-analysis were performed as per the Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines. A search of electronic information was conducted using the following databases: MEDLINE, EMBASE, EMCARE, CINAHL and the Cochrane Central Register of Controlled Trials.SettingStudies that compared the diagnostic performance within the same patient cohort of chest CT scan versus RT-PCR in COVID-19 suspected patients.ParticipantsThirteen non-randomised studies enrolling 4092 patients were identified.Main outcome measuresSensitivity, specificity and accuracy were primary outcome measures. Secondary outcomes included other test performance characteristics and discrepant findings between both investigations.ResultsChest CT had a median sensitivity, specificity and accuracy of 0.91 (range 0.82–0.98), 0.775 (0.25–1.00) and 0.87 (0.68–0.99), respectively, with RT-PCR as the reference. Importantly, early small, China-based studies tended to favour chest CT versus later larger, non-China studies.ConclusionsA relatively high false positive rate can be expected with chest CT. It is possible it may still be useful to provide circumstantial evidence, however, in some patients with a suspicious clinical presentation of COVID-19 and negative initial Severe Acute Respiratory Syndrome Coronavirus 2 RT-PCR tests, but more evidence is required in this context. In acute cardiorespiratory presentations, negative CT scan and RT-PCR tests is likely to be reassuring.