Deep learning attention-guided radiomics for COVID-19 chest radiograph classification.

Background: Accurate assessment of coronavirus disease 2019 (COVID-19) lung involvement through chest radiograph plays an important role in effective management of the infection. This study aims to develop a two-step feature merging method to integrate image features from deep learning and radiomics to differentiate COVID-19, non-COVID-19 pneumonia and normal chest radiographs (CXR). Methods: In this study, a deformable convolutional neural network (deformable CNN) was developed and used as a feature extractor to obtain 1,024-dimensional deep learning latent representation (DLR) features. Then 1,069-dimensional radiomics features were extracted from the region of interest (ROI) guided by deformable CNN's attention. The two feature sets were concatenated to generate a merged feature set for classification. For comparative experiments, the same process has been applied to the DLR-only feature set for verifying the effectiveness of feature concatenation. Results: Using the merged feature set resulted in an overall average accuracy of 91.0% for three-class classification, representing a statistically significant improvement of 0.6% compared to the DLR-only classification. The recall and precision of classification into the COVID-19 class were 0.926 and 0.976, respectively. The feature merging method was shown to significantly improve the classification performance as compared to using only deep learning features, regardless of choice of classifier (P value <0.0001). Three classes' F1-score were 0.892, 0.890, and 0.950 correspondingly (i.e., normal, non-COVID-19 pneumonia, COVID-19). Conclusions: A two-step COVID-19 classification framework integrating information from both DLR and radiomics features (guided by deep learning attention mechanism) has been developed. The proposed feature merging method has been shown to improve the performance of chest radiograph classification as compared to the case of using only deep learning features.

Artificial intelligence-assisted multistrategy image enhancement of chest X-rays for COVID-19 classification.

Background: The coronavirus disease 2019 (COVID-19) led to a dramatic increase in the number of cases of patients with pneumonia worldwide. In this study, we aimed to develop an AI-assisted multistrategy image enhancement technique for chest X-ray (CXR) images to improve the accuracy of COVID-19 classification. Methods: Our new classification strategy consisted of 3 parts. First, the improved U-Net model with a variational encoder segmented the lung region in the CXR images processed by histogram equalization. Second, the residual net (ResNet) model with multidilated-rate convolution layers was used to suppress the bone signals in the 217 lung-only CXR images. A total of 80% of the available data were allocated for training and validation. The other 20% of the remaining data were used for testing. The enhanced CXR images containing only soft tissue information were obtained. Third, the neural network model with a residual cascade was used for the super-resolution reconstruction of low-resolution bone-suppressed CXR images. The training and testing data consisted of 1,200 and 100 CXR images, respectively. To evaluate the new strategy, improved visual geometry group (VGG)-16 and ResNet-18 models were used for the COVID-19 classification task of 2,767 CXR images. The accuracy of the multistrategy enhanced CXR images was verified through comparative experiments with various enhancement images. In terms of quantitative verification, 8-fold cross-validation was performed on the bone suppression model. In terms of evaluating the COVID-19 classification, the CXR images obtained by the improved method were used to train 2 classification models. Results: Compared with other methods, the CXR images obtained based on the proposed model had better performance in the metrics of peak signal-to-noise ratio and root mean square error. The super-resolution CXR images of bone suppression obtained based on the neural network model were also anatomically close to the real CXR images. Compared with the initial CXR images, the classification accuracy rates of the internal and external testing data on the VGG-16 model increased by 5.09% and 12.81%, respectively, while the values increased by 3.51% and 18.20%, respectively, for the ResNet-18 model. The numerical results were better than those of the single-enhancement, double-enhancement, and no-enhancement CXR images. Conclusions: The multistrategy enhanced CXR images can help to classify COVID-19 more accurately than the other existing methods.

Prognostic Model of COVID-19 Severity and Survival among Hospitalized Patients Using Machine Learning Techniques.

We conducted a statistical study and developed a machine learning model to triage COVID-19 patients affected during the height of the COVID-19 pandemic in Hong Kong based on their medical records and test results (features) collected during their hospitalization. The correlation between the values of these features is studied against discharge status and disease severity as a preliminary step to identify those features with a more pronounced effect on the patient outcome. Once identified, they constitute the inputs of four machine learning models, Decision Tree, Random Forest, Gradient and RUSBoosting, which predict both the Mortality and Severity associated with the disease. We test the accuracy of the models when the number of input features is varied, demonstrating their stability; i.e., the models are already highly predictive when run over a core set of (6) features. We show that Random Forest and Gradient Boosting classifiers are highly accurate in predicting patients' Mortality (average accuracy ∼99%) as well as categorize patients (average accuracy ∼91%) into four distinct risk classes (Severity of COVID-19 infection). Our methodical and broad approach combines statistical insights with various machine learning models, which paves the way forward in the AI-assisted triage and prognosis of COVID-19 cases, which is potentially generalizable to other seasonal flus.

Mass Screening Is Associated with Low Rates of Acute Kidney Injury among COVID-19 Patients in Hong Kong.

INTRODUCTION: Renal involvement in COVID-19 is less well characterized in settings with vigilant public health surveillance, including mass screening and early hospitalization. We assessed kidney complications among COVID-19 patients in Hong Kong, including the association with risk factors, length of hospitalization, critical presentation, and mortality. METHODS: Linked electronic records of all patients with confirmed COVID-19 from 5 major designated hospitals were extracted. Duplicated records due to interhospital transferal were removed. Primary outcome was the incidence of in-hospital acute kidney injury (AKI). Secondary outcomes were AKI-associated mortality, incident renal replacement therapy (RRT), intensive care admission, prolonged hospitalization and disease course (defined as >90th percentile of hospitalization duration [35 days] and duration from symptom onset to discharge [43 days], respectively), and change of estimated glomerular filtration rate (GFR). Patients were further stratified into being symptomatic or asymptomatic. RESULTS: Patients were characterized by young age (median: 38.4, IQR: 28.4-55.8 years) and short time (median: 5, IQR: 2-9 days) from symptom onset to admission. Among the 591 patients, 22 (3.72%) developed AKI and 4 (0.68%) required RRT. The median time from symptom onset to in-hospital AKI was 15 days. AKI increased the odds of prolonged hospitalization and disease course by 2.0- and 3.5-folds, respectively. Estimated GFR 24 weeks post-discharge reduced by 7.51 and 1.06 mL/min/1.73 m2 versus baseline (upon admission) in the AKI and non-AKI groups, respectively. The incidence of AKI was comparable between asymptomatic (4.8%, n = 3/62) and symptomatic (3.7%, n = 19/519) patients. CONCLUSION: The overall rate of AKI among COVID-19 patients in Hong Kong is low, which could be attributable to a vigilant screening program and early hospitalization. Among patients who developed in-hospital AKI, the duration of hospitalization is prolonged and kidney function impairment can persist for up to 6 months post-discharge. Mass surveillance for COVID-19 is warranted in identifying asymptomatic subjects for earlier AKI management.

Emergence of a Severe Acute Respiratory Syndrome Coronavirus 2 Virus Variant With Novel Genomic Architecture in Hong Kong.

Throughout the coronavirus disease 2019 (COVID-19) pandemic, divergent severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) lineages have emerged continuously, mostly through the genomic accumulation of substitutions. We report the discovery of a SARS-CoV-2 variant with a novel genomic architecture characterized by absent ORF7a, ORF7b, and ORF8, and a C-terminally modified ORF6 product resulting from partial 5'-untranslated region (UTR) duplication and transposition.

Posterior Oropharyngeal Saliva for the Detection of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2).

BACKGROUND: The coronavirus disease 2019 (COVID-19) pandemic has put tremendous pressure on the healthcare system worldwide. Diagnostic testing remained one of the limiting factors for early identification and isolation of infected patients. This study aimed to evaluate posterior oropharyngeal saliva (POPS) for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) detection among patients with confirmed or suspected COVID-19. METHODS: The laboratory information system was searched retrospectively for all respiratory specimens and POPS requested for SARS-CoV-2 RNA detection between 1 February 2020 and 15 April 2020. The agreement and diagnostic performance of POPS against NPsp were evaluated. RESULTS: A total of 13772 specimens were identified during the study period, including 2130 POPS and 8438 nasopharyngeal specimens (NPsp). Two hundred and twenty-nine same-day POPS-NPsp paired were identified with POPS and NPsp positivity of 61.5% (95% confidence interval [CI] 55.1-67.6%) and 53.3% (95% CI 46.8-59.6%). The overall, negative and positive percent agreement were 76.0% (95% CI 70.2-80.9%), 65.4% (95% CI 55.5-74.2%), 85.2% (95% CI 77.4-90.8%). Better positive percent agreement was observed in POPS-NPsp obtained within 7 days (96.6%, 95% CI 87.3-99.4%) compared with after 7 days of symptom onset (75.0%, 95% CI 61.4-85.2%). Among the 104 positive pairs, the mean difference in Cp value was 0.26 (range: 12.63 to -14.74), with an overall higher Cp value in NPsp (Pearson coefficient 0.579). No significant temporal variation was noted between the 2 specimen types. CONCLUSIONS: POPS is an acceptable alternative specimen to nasopharyngeal specimen for the detection of SARS-CoV-2.

SARS-CoV-2 shedding and seroconversion among passengers quarantined after disembarking a cruise ship: a case series.

BACKGROUND: A cruise ship is a closed-off environment that simulates the basic functioning of a city in terms of living conditions and interpersonal interactions. Thus, the Diamond Princess cruise ship, which was quarantined because of an onboard outbreak of COVID-19 in February, 2020, provides an opportunity to define the shedding pattern of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and patient antibody responses before and after the onset of symptoms. METHODS: We recruited adult (≥18 years) passengers from Hong Kong who had been on board the Diamond Princess cruise ship docked in Yokohama, Japan in February, 2020. All participants had been found to be negative for SARS-CoV-2 by RT-PCR 4 days before disembarking and were transferred to further quarantine in a public estate in Hong Kong, where they were recruited. Participants were prospectively screened by quantitative RT-PCR (RT-qPCR) of nasopharyngeal and throat swabs, and serum IgG and IgM against internal nucleoprotein and the surface spike receptor-binding protein (RBD) of SARS-CoV-2 at baseline (upon entering quarantine) and on days 4, 8, and 12 of quarantine. FINDINGS: On Feb 22, 2020, 215 adults were recruited, of whom nine (4%; 95% CI 2-8) were positive for SARS-CoV-2 by RT-qPCR or serology and were hospitalised. Of these nine patients, nasopharyngeal swab RT-qPCR was positive in eight patients (89%; 57-99) at baseline. All nine patients were positive for anti-RBD IgG by day 8. Eight (89%; 57-99) were simultaneously positive for nasopharyngeal swab RT-PCR and anti-RBD IgG. One patient who was positive for anti-RBD IgG and had a negative viral load had multifocal peripheral ground-glass changes on high-resolution CT that were typical of COVID-19. Five patients (56%; 27-81) with ground-glass changes on high-resolution CT were found to have higher anti-nucleoprotein-IgG OD values on day 8 and 12 and anti-RBD IgG OD value on day 12 than patients without ground-glass changes. Six (67%; 35-88) patients remained asymptomatic throughout the 14-day quarantine period. INTERPRETATION: Patients with COVID-19 can develop asymptomatic lung infection with viral shedding and those with evidence of pneumonia on imaging tend to have an increased antibody response. Positive IgG or IgM confirmed infection of COVID-19 in both symptomatic and asymptomatic patients. A combination of RT-PCR and serology should be implemented for case finding and contact tracing to facilitate early diagnosis, prompt isolation, and treatment. FUNDING: Shaw Foundation Hong Kong; Sanming-Project of Medicine (Shenzhen); High Level-Hospital Program (Guangdong Health Commission).

Triple combination of interferon beta-1b, lopinavir-ritonavir, and ribavirin in the treatment of patients admitted to hospital with COVID-19: an open-label, randomised, phase 2 trial.

BACKGROUND: Effective antiviral therapy is important for tackling the coronavirus disease 2019 (COVID-19) pandemic. We assessed the efficacy and safety of combined interferon beta-1b, lopinavir-ritonavir, and ribavirin for treating patients with COVID-19. METHODS: This was a multicentre, prospective, open-label, randomised, phase 2 trial in adults with COVID-19 who were admitted to six hospitals in Hong Kong. Patients were randomly assigned (2:1) to a 14-day combination of lopinavir 400 mg and ritonavir 100 mg every 12 h, ribavirin 400 mg every 12 h, and three doses of 8 million international units of interferon beta-1b on alternate days (combination group) or to 14 days of lopinavir 400 mg and ritonavir 100 mg every 12 h (control group). The primary endpoint was the time to providing a nasopharyngeal swab negative for severe acute respiratory syndrome coronavirus 2 RT-PCR, and was done in the intention-to-treat population. The study is registered with ClinicalTrials.gov, NCT04276688. FINDINGS: Between Feb 10 and March 20, 2020, 127 patients were recruited; 86 were randomly assigned to the combination group and 41 were assigned to the control group. The median number of days from symptom onset to start of study treatment was 5 days (IQR 3-7). The combination group had a significantly shorter median time from start of study treatment to negative nasopharyngeal swab (7 days [IQR 5-11]) than the control group (12 days [8-15]; hazard ratio 4·37 [95% CI 1·86-10·24], p=0·0010). Adverse events included self-limited nausea and diarrhoea with no difference between the two groups. One patient in the control group discontinued lopinavir-ritonavir because of biochemical hepatitis. No patients died during the study. INTERPRETATION: Early triple antiviral therapy was safe and superior to lopinavir-ritonavir alone in alleviating symptoms and shortening the duration of viral shedding and hospital stay in patients with mild to moderate COVID-19. Future clinical study of a double antiviral therapy with interferon beta-1b as a backbone is warranted. FUNDING: The Shaw-Foundation, Richard and Carol Yu, May Tam Mak Mei Yin, and Sanming Project of Medicine.

Temporal profiles of viral load in posterior oropharyngeal saliva samples and serum antibody responses during infection by SARS-CoV-2: an observational cohort study.

BACKGROUND: Coronavirus disease 2019 (COVID-19) causes severe community and nosocomial outbreaks. Comprehensive data for serial respiratory viral load and serum antibody responses from patients infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) are not yet available. Nasopharyngeal and throat swabs are usually obtained for serial viral load monitoring of respiratory infections but gathering these specimens can cause discomfort for patients and put health-care workers at risk. We aimed to ascertain the serial respiratory viral load of SARS-CoV-2 in posterior oropharyngeal (deep throat) saliva samples from patients with COVID-19, and serum antibody responses. METHODS: We did a cohort study at two hospitals in Hong Kong. We included patients with laboratory-confirmed COVID-19. We obtained samples of blood, urine, posterior oropharyngeal saliva, and rectal swabs. Serial viral load was ascertained by reverse transcriptase quantitative PCR (RT-qPCR). Antibody levels against the SARS-CoV-2 internal nucleoprotein (NP) and surface spike protein receptor binding domain (RBD) were measured using EIA. Whole-genome sequencing was done to identify possible mutations arising during infection. FINDINGS: Between Jan 22, 2020, and Feb 12, 2020, 30 patients were screened for inclusion, of whom 23 were included (median age 62 years [range 37-75]). The median viral load in posterior oropharyngeal saliva or other respiratory specimens at presentation was 5·2 log10 copies per mL (IQR 4·1-7·0). Salivary viral load was highest during the first week after symptom onset and subsequently declined with time (slope -0·15, 95% CI -0·19 to -0·11; R2=0·71). In one patient, viral RNA was detected 25 days after symptom onset. Older age was correlated with higher viral load (Spearman's ρ=0·48, 95% CI 0·074-0·75; p=0·020). For 16 patients with serum samples available 14 days or longer after symptom onset, rates of seropositivity were 94% for anti-NP IgG (n=15), 88% for anti-NP IgM (n=14), 100% for anti-RBD IgG (n=16), and 94% for anti-RBD IgM (n=15). Anti-SARS-CoV-2-NP or anti-SARS-CoV-2-RBD IgG levels correlated with virus neutralisation titre (R2>0·9). No genome mutations were detected on serial samples. INTERPRETATION: Posterior oropharyngeal saliva samples are a non-invasive specimen more acceptable to patients and health-care workers. Unlike severe acute respiratory syndrome, patients with COVID-19 had the highest viral load near presentation, which could account for the fast-spreading nature of this epidemic. This finding emphasises the importance of stringent infection control and early use of potent antiviral agents, alone or in combination, for high-risk individuals. Serological assay can complement RT-qPCR for diagnosis. FUNDING: Richard and Carol Yu, May Tam Mak Mei Yin, The Shaw Foundation Hong Kong, Michael Tong, Marina Lee, Government Consultancy Service, and Sanming Project of Medicine.

Consistent Detection of 2019 Novel Coronavirus in Saliva.

The 2019 novel coronavirus (2019-nCoV) was detected in the self-collected saliva of 91.7% (11/12) of patients. Serial saliva viral load monitoring generally showed a declining trend. Live virus was detected in saliva by viral culture. Saliva is a promising noninvasive specimen for diagnosis, monitoring, and infection control in patients with 2019-nCoV infection.