A Distance Transformation Deep Forest Framework With Hybrid-Feature Fusion for CXR Image Classification.

Detecting pneumonia, especially coronavirus disease 2019 (COVID-19), from chest X-ray (CXR) images is one of the most effective ways for disease diagnosis and patient triage. The application of deep neural networks (DNNs) for CXR image classification is limited due to the small sample size of the well-curated data. To tackle this problem, this article proposes a distance transformation-based deep forest framework with hybrid-feature fusion (DTDF-HFF) for accurate CXR image classification. In our proposed method, hybrid features of CXR images are extracted in two ways: hand-crafted feature extraction and multigrained scanning. Different types of features are fed into different classifiers in the same layer of the deep forest (DF), and the prediction vector obtained at each layer is transformed to form distance vector based on a self-adaptive scheme. The distance vectors obtained by different classifiers are fused and concatenated with the original features, then input into the corresponding classifier at the next layer. The cascade grows until DTDF-HFF can no longer gain benefits from the new layer. We compare the proposed method with other methods on the public CXR datasets, and the experimental results show that the proposed method can achieve state-of-the art (SOTA) performance. The code will be made publicly available at https://github.com/hongqq/DTDF-HFF.

Two-stage hybrid network for segmentation of COVID-19 pneumonia lesions in CT images: a multicenter study.

COVID-19 has been spreading continuously since its outbreak, and the detection of its manifestations in the lung via chest computed tomography (CT) imaging is essential to investigate the diagnosis and prognosis of COVID-19 as an indispensable step. Automatic and accurate segmentation of infected lesions is highly required for fast and accurate diagnosis and further assessment of COVID-19 pneumonia. However, the two-dimensional methods generally neglect the intraslice context, while the three-dimensional methods usually have high GPU memory consumption and calculation cost. To address these limitations, we propose a two-stage hybrid UNet to automatically segment infected regions, which is evaluated on the multicenter data obtained from seven hospitals. Moreover, we train a 3D-ResNet for COVID-19 pneumonia screening. In segmentation tasks, the Dice coefficient reaches 97.23% for lung segmentation and 84.58% for lesion segmentation. In classification tasks, our model can identify COVID-19 pneumonia with an area under the receiver-operating characteristic curve value of 0.92, an accuracy of 92.44%, a sensitivity of 93.94%, and a specificity of 92.45%. In comparison with other state-of-the-art methods, the proposed approach could be implemented as an efficient assisting tool for radiologists in COVID-19 diagnosis from CT images.

Measurement report: The importance of biomass burning in light extinction and direct radiative effect of urban aerosol during the COVID-19 lockdown in Xi'an, China

Due to the complexity of emission sources, a better understanding of aerosol optical properties is required to mitigate climate change in China. Here, an intensive real-time measurement campaign was conducted in an urban area of China before and during the COVID-19 lockdown in order to explore the impacts of anthropogenic activities on aerosol light extinction and the direct radiative effect (DRE). The mean light extinction coefficient (bext) decreased from 774.7 ± 298.1 Mm-1 during the normal period to 544.3 ± 179.4 Mm-1 during the lockdown period. A generalised additive model analysis indicated that the large decline in bext (29.7 %) was due to sharp reductions in anthropogenic emissions. Chemical calculation of bext based on a ridge regression analysis showed that organic aerosol (OA) was the largest contributor to bext in both periods (45.1 %–61.4 %), and the contributions of two oxygenated OAs to bext increased by 3.0 %–14.6 % during the lockdown. A hybrid environmental receptor model combined with chemical and optical variables identified six sources of bext. It was found thatbext from traffic-related emissions, coal combustion, fugitive dust, the nitrate and secondary OA (SOA) source, and the sulfate and SOA source decreased by 21.4 %–97.9 % in the lockdown, whereas bext from biomass burning increased by 27.1 %, mainly driven by the undiminished need for residential cooking and heating. An atmospheric radiative transfer model was further used to illustrate that biomass burning, rather than traffic-related emissions, became the largest positive effect (10.0 ± 10.9 W m-2) on aerosol DRE in the atmosphere during the lockdown. Our study provides insights into aerosol bext and DRE from anthropogenic sources, and the results imply the importance of controlling biomass burning for tackling climate change in China in the future.

Impact of reduced anthropogenic emissions on chemical characteristics of urban aerosol by individual particle analysis.

A single particle aerosol mass spectrometer was deployed in a heavily polluted area of China during a coronavirus lockdown to explore the impact of reduced anthropogenic emissions on the chemical composition, size distributions, mixing state, and secondary formation of urban aerosols. Ten particle groups were identified using an adaptive resonance network algorithm. Increased atmospheric oxidation during the lockdown period (LP) resulted in a 42.2%-54% increase in the major NaK-SN particle fraction relative to the normal period (NP). In contrast, EC-aged particles decreased from 31.5% (NP) to 23.7% (LP), possibly due to lower emissions from motor vehicles and coal combustion. The peak particle size diameter increased from 440 nm during the NP to 500 nm during LP due to secondary particle formation. High proportions of mixed 62NO3- indicate extensive particle aging. Correlations between secondary organic (43C2H3O+, oxalate) and secondary inorganic species (62NO3-, 97HSO4- and 18NH4+) versus oxidants (Ox = O3 + NO2) and relative humidity (RH) indicate that increased atmospheric oxidation promoted the generation of secondary species, while the effects of RH were more complex. Differences between the NP and LP show that reductions in primary emissions had a remarkable impact on the aerosol particles. This study provides new insights into the effects of pollution emissions on atmospheric reactions and the specific aerosol types in urban regions.

Progression of severity in coronavirus disease 2019 patients before treatment and a self-assessment scale to predict disease severity.

OBJECTIVES: This study aims to further investigate the association of COVID-19 disease severity with numerous patient characteristics, and to develop a convenient severity prediction scale for use in self-assessment at home or in preliminary screening in community healthcare settings. SETTING AND PARTICIPANTS: Data from 45,450 patients infected with COVID-19 from January 1 to February 27, 2020 were extracted from the municipal Notifiable Disease Report System in Wuhan, China. PRIMARY AND SECONDARY OUTCOME MEASURES: We categorized COVID-19 disease severity, based on The Chinese Diagnosis and Treatment Protocol for COVID-19, as "nonsevere" (which grouped asymptomatic, mild, and ordinary disease) versus "severe" (grouping severe and critical illness). RESULTS: Twelve scale items-age, gender, illness duration, dyspnea, shortness of breath (clinical evidence of altered breathing), hypertension, pulmonary disease, diabetes, cardio/cerebrovascular disease, number of comorbidities, neutrophil percentage, and lymphocyte percentage-were identified and showed good predictive ability (area under the curve = 0·72). After excluding the community healthcare laboratory parameters, the remaining model (the final self-assessment scale) showed similar area under the curve (= 0·71). CONCLUSIONS: Our COVID-19 severity self-assessment scale can be used by patients in the community to predict their risk of developing severe illness and the need for further medical assistance. The tool is also practical for use in preliminary screening in community healthcare settings. Our study constructed a COVID-19 severity self-assessment scale that can be used by patients in the community to predict their risk of developing severe illness and the need for further medical assistance.

Extrapolation of anthropogenic disturbances on hazard elements in PM2.5 in a typical heavy industrial city in northwest China.

Baoji is a typical heavy industrial city in northwest China. Its air quality is greatly impacted by the emission from the factories. Elements in fine particulate matter (PM2.5) that are greatly emitted from anthropogenic sources could pose diverse health impacts on humans. In this study, an online AMMS-100 atmospheric heavy metal analyzer was used to quantify 30 elements in PM2.5 under the weak and strong anthropogenic disturbance scenarios before the city lockdown period (from January 9th to 23rd) and the lockdown period (from January 26th to February 9th) due to the outbreak of COVID-19 in 2020. During the lockdown period, the average total concentration of total quantified elements was 3475.0 ng/m3, which was 28% and 33% lower than that of the week and strong anthropogenic disturbance scenarios during the pre-lockdown period. The greatest reductions were found for the elements of chromium (Cr), titanium (Ti), manganese (Mn), and Zinc (Zn), consistent with the industrial structure of Baoji. The mass concentrations of most elements showed obvious reductions when the government post-alerted the industries to reduce the operations and production. Dust, traffic sources, combustion, non-ferrous metal processing, and Ti-related industrial processing that are the contributors of the elements in the pre-lockdown period were apportioned by the positive matrix factorization (PMF) model. Substantial changes in the quantified elements' compositions and sources were found in the lockdown period. Health assessment was conducted and characterized by apportioned sources. The highest non-carcinogenic risk (HQ) was seen for Zn, demonstrating the high emissions from the related industrial activities. The concentration level of arsenic (As) exceeded the incremental lifetime carcinogenic risk (ILCR) in the lockdown period. This could be attributed to the traditional firework activities for the celebration of the Chinese New Year within the lockdown period.

The Heavy Particulate Matter Pollution During the COVID-19 Lockdown Period in the Guanzhong Basin, China.

Nationwide restrictions on human activities (lockdown) in China since 23 January 2020, to control the 2019 novel coronavirus disease pandemic (COVID-19), has provided an opportunity to evaluate the effect of emission mitigation on particulate matter (PM) pollution. The WRF-Chem simulations of persistent heavy PM pollution episodes from 20 January to 14 February 2020, in the Guanzhong Basin (GZB), northwest China, reveal that large-scale emission reduction of primary pollutants has not substantially improved the air quality during the COVID-19 lockdown period. Simultaneous reduction of primary precursors during the lockdown period only decreases the near-surface PM2.5 mass concentration by 11.6% (12.6 µg m-3), but increases ozone (O3) concentration by 9.2% (5.5 µg m-3) in the GZB. The primary organic aerosol and nitrate are the major contributor to the decreased PM2.5 in the GZB, with the reduction of 28.0% and 21.8%, respectively, followed by EC (10.1%) and ammonium (7.2%). The increased atmospheric oxidizing capacity by the O3 enhancement facilitates the secondary aerosol (SA) formation in the GZB, increasing secondary organic aerosol and sulphate by 6.5% and 3.3%, respectively. Furthermore, sensitivity experiments suggest that combined emission reduction of NOX and VOCs following the ratio of 1:1 is conducive to lowering the wintertime SA and O3 concentration and further alleviating the PM pollution in the GZB.

Vaccination as an alternative to non-drug interventions to prevent local resurgence of COVID-19.

BACKGROUND: While a COVID-19 vaccine protects people from serious illness and death, it remains a concern when and how to lift the high-cost and strict non-pharmaceutical interventions (NPIs). This study examined the joint effect of vaccine coverage and NPIs on the control of local and sporadic resurgence of COVID-19 cases. METHODS: Between July 2021 and January 2022, we collected the large-scale testing information and case number of imported COVID-19 patients from the website of the National Health Commission of China. A compartment model was developed to identify the level of vaccine coverage that would allow safe relaxation of NPIs, and vaccination strategies that can best achieve this level of coverage. We applied Monte Carlo simulation 50 000 times to remove random fluctuation effects and obtain fitted/predicted epidemic curve based on various parameters with 95% confidence interval at each time point. RESULTS: We found that a vaccination coverage of 50.4% was needed for the safe relaxation of NPIs, if the vaccine effectiveness was 79.3%. The total number of incidence cases under the key groups firstly strategy was 103 times higher than that of accelerated vaccination strategy. It needed 35 months to fully relax NPIs if the key groups firstly strategy was implemented, and 27 months were needed with the accelerated vaccination strategy. If combined the two strategies, only 8 months are needed to achieve the vaccine coverage threshold for the fully relaxation of NPIs. Sensitivity analyses results shown that the higher the transmission rate of the virus and the lower annual vaccine supply, the more difficult the epidemic could be under control. When the transmission rate increased 25% or the vaccination effectiveness rate decreased 20%, 33 months were needed to reduce the number of total incidence cases below 1000. CONCLUSIONS: As vaccine coverage improves, the NPIs can be gradually relaxed. Until that threshold is reached, however, strict NPIs are still needed to control the epidemic. The more transmissible SARS-CoV-2 variant led to higher resurgence probability, which indicates the importance of accelerated vaccination and achieving the vaccine coverage earlier.

The Role of Imaging in the Detection and Management of COVID-19: A Review.

Coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is spreading rapidly around the world, resulting in a massive death toll. Lung infection or pneumonia is the common complication of COVID-19, and imaging techniques, especially computed tomography (CT), have played an important role in diagnosis and treatment assessment of the disease. Herein, we review the imaging characteristics and computing models that have been applied for the management of COVID-19. CT, positron emission tomography - CT (PET/CT), lung ultrasound, and magnetic resonance imaging (MRI) have been used for detection, treatment, and follow-up. The quantitative analysis of imaging data using artificial intelligence (AI) is also explored. Our findings indicate that typical imaging characteristics and their changes can play crucial roles in the detection and management of COVID-19. In addition, AI or other quantitative image analysis methods are urgently needed to maximize the value of imaging in the management of COVID-19.

Impact of COVID-19 lockdown on the optical properties and radiative effects of urban brown carbon aerosol

Intensive measurements were conducted in Xi’an, China before and during a COVID-19 lockdown period to investigate how changes in anthropogenic emissions affected the optical properties and radiative effects of brown carbon (BrC) aerosol. The contribution of BrC to total aerosol light absorption during the lockdown (13%–49%) was higher compared with the normal period (4%–29%). Mass absorption cross-sections (MAC) of specific organic aerosol (OA) factors were calculated from a ridge regression model. Of the primary OA (POA), coal combustion OA (CCOA) had the largest MACs at all tested wavelengths during both periods due to high molecular-weight BrC chromophores;that was followed by biomass burning OA (BBOA) and hydrocarbon-like OA (HOA). For secondary OA (SOA), the MACs of the less-oxidized oxygenated OA (OOA) species (LO-OOA) at λ = 370–590 nm were higher than those of more-oxidized OOA (MO-OOA) during both periods, presumably due to chromophore bleaching. The largest contributor to BrC absorption at the short wavelengths was CCOA during both periods, but BrC absorption by LO-OOA and MO-OOA became dominant at longer wavelengths during the lockdown. The estimated radiation forcing efficiency of BrC over 370–600 nm increased from 37.5 W· g-1 during the normal period to 50.2 W·g-1 during the lockdown, and that enhancement was mainly caused by higher MACs for both LO-OOA and MO-OOA. This study provides insights into the optical properties and radiative effects of source-specific BrC aerosol when pollution emissions are reduced.

Mix Contrast for COVID-19 Mild-to-Critical Prediction.

OBJECTIVE: In a few patients with mild COVID-19, there is a possibility of the infection becoming severe or critical in the future. This work aims to identify high-risk patients who have a high probability of changing from mild to critical COVID-19 (only account for 5% of cases). METHODS: Using traditional convolutional neural networks for classification may not be suitable to identify this 5% of high risk patients from an entire dataset due to the highly imbalanced label distribution. To address this problem, we propose a Mix Contrast model, which matches original features with mixed features for contrastive learning. Three modules are proposed for training the model: 1) a cumulative learning strategy for synthesizing the mixed feature; 2) a commutative feature combination module for learning the commutative law of feature concatenation; 3) a united pairwise loss assigning adaptive weights for sample pairs with different class anchors based on their current optimization status. RESULTS: We collect a multi-center computed tomography dataset including 918 confirmed COVID-19 patients from four hospitals and evaluate the proposed method on both the COVID-19 mild-to-critical prediction and COVID-19 diagnosis tasks. For mild-to-critical prediction, the experimental results show a recall of 0.80 and a specificity of 0.815. For diagnosis, the model shows comparable results with deep neural networks using a large dataset. Our method demonstrates improvements when the amount of training data is small or imbalanced. SIGNIFICANCE: Identifying mild-to-critical COVID-19 patients is important for early prevention and personalized treatment planning.

An enhanced method for nucleic acid detection with CRISPR-Cas12a using phosphorothioate modified primers and optimized gold-nanopaticle strip.

CRISPR-Cas12a system has been shown promising for nucleic acid diagnostics due to its rapid, portable and accurate features. However, cleavage of the amplicons and primers by the cis- and trans-activity of Cas12a hinders the attempts to integrate the amplification and detection into a single reaction. Through phosphorothioate modification of primers, we realized onepot detection with high sensitivity using plasmids of SARS-CoV-2, HPV16 and HPV18. We also identified the activated Cas12a has a much higher affinity to C nucleotide-rich reporter than others. By applying such reporters, the reaction time required for a lateral-flow readout was significantly reduced. Furthermore, to improve the specificity of the strip-based assay, we created a novel reporter and, when combined with a customized gold-nanopaticle strip, the readout was greatly enhanced owing to the elimination of the nonspecific signal. This established system, termed Targeting DNA by Cas12a-based Eye Sight Testing in an Onepot Reaction (TESTOR), was validated using clinical cervical scrape samples for human papillomaviruses (HPVs) detection. Our system represents a general approach to integrating the nucleic acid amplification and detection into a single reaction in CRISPR-Cas systems, highlighting its potential as a rapid, portable and accurate detection platform of nucleic acids.

Efficacy and safety of Xiyanping injection in the treatment of COVID-19: A multicenter, prospective, open-label and randomized controlled trial.

Xiyanping (XYP) is a Chinese herbal medicine used in the clinic to treat respiratory infection and pneumonia. Recent evidence identified XYP as a potential inhibitor of severe acute respiratory syndrome coronavirus 2, implying XYP as a possible treatment for the coronavirus disease 2019 (COVID-19). Here, we conducted a prospective, multicenter, open-label and randomized controlled trial to evaluate the safety and effectiveness of XYP injection in patients with mild to moderate COVID-19. We consecutively recruited 130 COVID-19 patients with mild to moderate symptoms from five study sites, and randomized them in 1:1 ratio to receive XYP injection in combination with standard therapy or receive standard supportive therapy alone. We found that XYP injection significantly reduced the time to cough relief, fever resolution and virus clearance. Less patients receiving XYP injection experienced disease progression to the severe stage during the treatment process. No severe adverse events were reported during the study. Taken together, XYP injection is safe and effective in improving the recovery of patients with mild to moderate COVID-19. However, further studies are warranted to evaluate the efficacy of XYP in an expanded cohort comprising COVID-19 patients at different disease stages.

A Deep Learning Radiomics Model to Identify Poor Outcome in COVID-19 Patients With Underlying Health Conditions: A Multicenter Study.

OBJECTIVE: Coronavirus disease 2019 (COVID-19) has caused considerable morbidity and mortality, especially in patients with underlying health conditions. A precise prognostic tool to identify poor outcomes among such cases is desperately needed. METHODS: Total 400 COVID-19 patients with underlying health conditions were retrospectively recruited from 4 centers, including 54 dead cases (labeled as poor outcomes) and 346 patients discharged or hospitalized for at least 7 days since initial CT scan. Patients were allocated to a training set (n = 271), a test set (n = 68), and an external test set (n = 61). We proposed an initial CT-derived hybrid model by combining a 3D-ResNet10 based deep learning model and a quantitative 3D radiomics model to predict the probability of COVID-19 patients reaching poor outcome. The model performance was assessed by area under the receiver operating characteristic curve (AUC), survival analysis, and subgroup analysis. RESULTS: The hybrid model achieved AUCs of 0.876 (95% confidence interval: 0.752-0.999) and 0.864 (0.766-0.962) in test and external test sets, outperforming other models. The survival analysis verified the hybrid model as a significant risk factor for mortality (hazard ratio, 2.049 [1.462-2.871], P < 0.001) that could well stratify patients into high-risk and low-risk of reaching poor outcomes (P < 0.001). CONCLUSION: The hybrid model that combined deep learning and radiomics could accurately identify poor outcomes in COVID-19 patients with underlying health conditions from initial CT scans. The great risk stratification ability could help alert risk of death and allow for timely surveillance plans.

Impacts of primary emissions and secondary aerosol formation on air pollution in an urban area of China during the COVID-19 lockdown.

Restrictions on human activities were implemented in China to cope with the outbreak of the Coronavirus Disease 2019 (COVID-19), providing an opportunity to investigate the impacts of anthropogenic emissions on air quality. Intensive real-time measurements were made to compare primary emissions and secondary aerosol formation in Xi'an, China before and during the COVID-19 lockdown. Decreases in mass concentrations of particulate matter (PM) and its components were observed during the lockdown with reductions of 32-51%. The dominant contributor of PM was organic aerosol (OA), and results of a hybrid environmental receptor model indicated OA was composed of four primary OA (POA) factors (hydrocarbon-like OA (HOA), cooking OA (COA), biomass burning OA (BBOA), and coal combustion OA (CCOA)) and two oxygenated OA (OOA) factors (less-oxidized OOA (LO-OOA) and more-oxidized OOA (MO-OOA)). The mass concentrations of OA factors decreased from before to during the lockdown over a range of 17% to 58%, and they were affected by control measures and secondary processes. Correlations of secondary aerosols/ΔCO with Ox (NO2 + O3) and aerosol liquid water content indicated that photochemical oxidation had a greater effect on the formation of nitrate and two OOAs than sulfate; however, aqueous-phase reaction presented a more complex effect on secondary aerosols formation at different relative humidity condition. The formation efficiencies of secondary aerosols were enhanced during the lockdown as the increase of atmospheric oxidation capacity. Analyses of pollution episodes highlighted the importance of OA, especially the LO-OOA, for air pollution during the lockdown.

Risk assessment and evaluation of China's policy to prevent COVID-19 cases imported by plane.

As of October 5, 2020, China has reported 2,921 cases imported from overseas. Assessing the effectiveness of China's current policies on imported cases abroad is very important for China and other countries that are facing or will face overseas imported cases. In April, we used a susceptible-exposed-infectious-recovered metapopulation model to simulate the epidemic in seven foreign countries, China and the three Chinese key cities. Based on the model outside China, we estimated the proportion of people in incubation period and calculated the risk indexes for Chinese cities through analyzing aviation traffic data from these countries. Based on the model in China and the three key cities, we collected information on control measures and quantified the effectiveness of implementing the current policies at different times and intensities. Our model results showed that Shanghai, Beijing, Qingdao, Guangzhou, and Tianjin have the top five risk indexes. As of April 20, 2020, under current measures, the number of confirmed cases could be reduced by 99% compared with no air traffic restrictions and isolation measures; the reduction could be 93% with isolation of passengers only from key countries. If the current policy were postponed for 7, 10, or 20 days, the increase in the number of confirmed cases would be 1,329, 5,524, and 779,245 respectively, which is 2.1, 5.7, and 662.9 times the number of confirmed cases under current measures. Our research indicates that the importation control measures currently taken by China were implemented at an appropriate time to prevent the epidemic spreading and have achieved relatively good control results. However, it is necessary to remain vigilant; otherwise, another outbreak peak could occur.

Classification of Severe and Critical Covid-19 Using Deep Learning and Radiomics.

OBJECTIVE: The coronavirus disease 2019 (COVID-19) is rapidly spreading inside China and internationally. We aimed to construct a model integrating information from radiomics and deep learning (DL) features to discriminate critical cases from severe cases of COVID-19 using computed tomography (CT) images. METHODS: We retrospectively enrolled 217 patients from three centers in China, including 82 patients with severe disease and 135 with critical disease. Patients were randomly divided into a training cohort (n = 174) and a test cohort (n = 43). We extracted 102 3-dimensional radiomic features from automatically segmented lung volume and selected the significant features. We also developed a 3-dimensional DL network based on center-cropped slices. Using multivariable logistic regression, we then created a merged model based on significant radiomic features and DL scores. We employed the area under the receiver operating characteristic curve (AUC) to evaluate the model's performance. We then conducted cross validation, stratified analysis, survival analysis, and decision curve analysis to evaluate the robustness of our method. RESULTS: The merged model can distinguish critical patients with AUCs of 0.909 (95% confidence interval [CI]: 0.859-0.952) and 0.861 (95% CI: 0.753-0.968) in the training and test cohorts, respectively. Stratified analysis indicated that our model was not affected by sex, age, or chronic disease. Moreover, the results of the merged model showed a strong correlation with patient outcomes. SIGNIFICANCE: A model combining radiomic and DL features of the lung could help distinguish critical cases from severe cases of COVID-19.

A Deep Learning Prognosis Model Help Alert for COVID-19 Patients at High-Risk of Death: A Multi-Center Study.

Since its outbreak in December 2019, the persistent coronavirus disease (COVID-19) became a global health emergency. It is imperative to develop a prognostic tool to identify high-risk patients and assist in the formulation of treatment plans. We retrospectively collected 366 severe or critical COVID-19 patients from four centers, including 70 patients who died within 14 days (labeled as high-risk patients) since their initial CT scan and 296 who survived more than 14 days or were cured (labeled as low-risk patients). We developed a 3D densely connected convolutional neural network (termed De-COVID19-Net) to predict the probability of COVID-19 patients belonging to the high-risk or low-risk group, combining CT and clinical information. The area under the curve (AUC) and other evaluation techniques were used to assess our model. The De-COVID19-Net yielded an AUC of 0.952 (95% confidence interval, 0.928-0.977) on the training set and 0.943 (0.904-0.981) on the test set. The stratified analyses indicated that our model's performance is independent of age, sex, and with/without chronic diseases. The Kaplan-Meier analysis revealed that our model could significantly categorize patients into high-risk and low-risk groups (p < 0.001). In conclusion, De-COVID19-Net can non-invasively predict whether a patient will die shortly based on the patient's initial CT scan with an impressive performance, which indicated that it could be used as a potential prognosis tool to alert high-risk patients and intervene in advance.

Radiomics Analysis of Computed Tomography helps predict poor prognostic outcome in COVID-19.

Rationale: Given the rapid spread of COVID-19, an updated risk-stratify prognostic tool could help clinicians identify the high-risk patients with worse prognoses. We aimed to develop a non-invasive and easy-to-use prognostic signature by chest CT to individually predict poor outcome (death, need for mechanical ventilation, or intensive care unit admission) in patients with COVID-19. Methods: From November 29, 2019 to February 19, 2020, a total of 492 patients with COVID-19 from four centers were retrospectively collected. Since different durations from symptom onsets to the first CT scanning might affect the prognostic model, we designated the 492 patients into two groups: 1) the early-phase group: CT scans were performed within one week after symptom onset (0-6 days, n = 317); and 2) the late-phase group: CT scans were performed one week later after symptom onset (≥7 days, n = 175). In each group, we divided patients into the primary cohort (n = 212 in the early-phase group, n = 139 in the late-phase group) and the external independent validation cohort (n = 105 in the early-phase group, n = 36 in the late-phase group) according to the centers. We built two separate radiomics models in the two patient groups. Firstly, we proposed an automatic segmentation method to extract lung volume for radiomics feature extraction. Secondly, we applied several image preprocessing procedures to increase the reproducibility of the radiomics features: 1) applied a low-pass Gaussian filter before voxel resampling to prevent aliasing; 2) conducted ComBat to harmonize radiomics features per scanner; 3) tested the stability of the features in the radiomics signature by several image transformations, such as rotating, translating, and growing/shrinking. Thirdly, we used least absolute shrinkage and selection operator (LASSO) to build the radiomics signature (RadScore). Afterward, we conducted a Fine-Gray competing risk regression to build the clinical model and the clinic-radiomics signature (CrrScore). Finally, performances of the three prognostic signatures (clinical model, RadScore, and CrrScore) were estimated from the two aspects: 1) cumulative poor outcome probability prediction; 2) 28-day poor outcome prediction. We also did stratified analyses to explore the potential association between the CrrScore and the poor outcomes regarding different age, type, and comorbidity subgroups. Results: In the early-phase group, the CrrScore showed the best performance in estimating poor outcome (C-index = 0.850), and predicting the probability of 28-day poor outcome (AUC = 0.862). In the late-phase group, the RadScore alone achieved similar performance to the CrrScore in predicting poor outcome (C-index = 0.885), and 28-day poor outcome probability (AUC = 0.976). Moreover, the RadScore in both groups successfully stratified patients with COVID-19 into low- or high-RadScore groups with significantly different survival time in the training and validation cohorts (all P < 0.05). The CrrScore in both groups can also significantly stratify patients with different prognoses regarding different age, type, and comorbidities subgroups in the combined cohorts (all P < 0.05). Conclusions: This research proposed a non-invasive and quantitative prognostic tool for predicting poor outcome in patients with COVID-19 based on CT imaging. Taking the insufficient medical recourse into account, our study might suggest that the chest CT radiomics signature of COVID-19 is more effective and ideal to predict poor outcome in the late-phase COVID-19 patients. For the early-phase patients, integrating radiomics signature with clinical risk factors can achieve a more accurate prediction of individual poor prognostic outcome, which enables appropriate management and surveillance of COVID-19.