Development and Validation of a Deep Learning-Based Model Using Computed Tomography Imaging for Predicting Disease Severity of Coronavirus Disease 2019

Xiao, Lu; Li, Pu; Sun, Fenglong; Zhang, Yan-Pei; Xu, Chunxiao; Zhu, Hongbo; Cai, Feng Qin; He, Yu; Zhang, Wen Feng; Ma, Si-Cong; Hu, Chenyi; Gong, Mengchun; Liu, Li; Shi, Wenzhao; Zhu, Hong

doi:10.3389/fbioe.2020.00898

Cited by 64 publications

(52 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The classification accuracy obtained from this model was 94%. Xiao et al [ 26 ] implemented a pre-trained ResNet34 to envisage the severity of COVID-19 using chest CT scans. 23,812 CT scans were used for experimentation.…”

Section: Related Workmentioning

confidence: 99%

Densely connected convolutional networks-based COVID-19 screening model

2021

View full text Add to dashboard Cite

The extensively utilized tool to detect novel coronavirus (COVID-19) is a real-time polymerase chain reaction (RT-PCR). However, RT-PCR kits are costly and consume critical time, around 6 to 9 hours to classify the subjects as COVID-19(+) or COVID-19(-). Due to the less sensitivity of RT-PCR, it suffers from high false-negative results. To overcome these issues, many deep learning models have been implemented in the literature for the early-stage classification of suspected subjects. To handle the sensitivity issue associated with RT-PCR, chest CT scans are utilized to classify the suspected subjects as COVID-19 (+), tuberculosis, pneumonia, or healthy subjects. The extensive study on chest CT scans of COVID-19 (+) subjects reveals that there are some bilateral changes and unique patterns. But the manual analysis from chest CT scans is a tedious task. Therefore, an automated COVID-19 screening model is implemented by ensembling the deep transfer learning models such as Densely connected convolutional networks (DCCNs), ResNet152V2, and VGG16. Experimental results reveal that the proposed ensemble model outperforms the competitive models in terms of accuracy, f-measure, area under curve, sensitivity, and specificity.

show abstract

Section: Related Workmentioning

confidence: 99%

Densely connected convolutional networks-based COVID-19 screening model

2021

View full text Add to dashboard Cite

show abstract

“…Chest computed tomography (CT) scans provide important diagnostic and prognostic information [ 7 , 8 ]; consequently, they have been the focus of numerous recent studies using machine learning techniques for COVID-19–related prediction tasks [ 9 - 21 ]. Previous studies have focused on mortality predictions [ 9 ], diagnosis (identifying COVID-19 cases and differentiating them from other pulmonary diseases or no disease) [ 10 - 15 , 19 , 22 - 25 ], and severity assessment and disease progression [ 16 - 18 , 23 ]. Most current approaches have used deep learning methods and imaging features from CT scans [ 10 - 15 , 19 , 22 - 24 ] and X-ray imaging [ 18 , 20 , 21 ] with popular architectures including ResNet [ 10 , 12 , 14 , 23 ], U-Net [ 11 , 17 ], Inception [ 15 , 22 ], Darknet [ 20 ], and other convolutional neural networks (NNs) [ 18 , 21 , 26 , 27 ].…”

Section: Introductionmentioning

confidence: 99%

“…Although automated assessment of chest CT scans to predict COVID-19 severity is of great clinical importance, few studies have focused on it [ 16 - 18 , 23 ]. Automated assessment of chest CT scans can substantially reduce the image reading time for radiologists, provide quantitative data that can be compared across patients and time points, and can be clinically applicable in disease detection and diagnosis, progression tracking, and prognosis [ 8 ].…”

Section: Introductionmentioning

confidence: 99%

Development and Validation of a Machine Learning Approach for Automated Severity Assessment of COVID-19 Based on Clinical and Imaging Data: Retrospective Study

Quiroz¹,

Feng²,

Cheng³

et al. 2021

JMIR Med Inform

View full text Add to dashboard Cite

Background COVID-19 has overwhelmed health systems worldwide. It is important to identify severe cases as early as possible, such that resources can be mobilized and treatment can be escalated. Objective This study aims to develop a machine learning approach for automated severity assessment of COVID-19 based on clinical and imaging data. Methods Clinical data—including demographics, signs, symptoms, comorbidities, and blood test results—and chest computed tomography scans of 346 patients from 2 hospitals in the Hubei Province, China, were used to develop machine learning models for automated severity assessment in diagnosed COVID-19 cases. We compared the predictive power of the clinical and imaging data from multiple machine learning models and further explored the use of four oversampling methods to address the imbalanced classification issue. Features with the highest predictive power were identified using the Shapley Additive Explanations framework. Results Imaging features had the strongest impact on the model output, while a combination of clinical and imaging features yielded the best performance overall. The identified predictive features were consistent with those reported previously. Although oversampling yielded mixed results, it achieved the best model performance in our study. Logistic regression models differentiating between mild and severe cases achieved the best performance for clinical features (area under the curve [AUC] 0.848; sensitivity 0.455; specificity 0.906), imaging features (AUC 0.926; sensitivity 0.818; specificity 0.901), and a combination of clinical and imaging features (AUC 0.950; sensitivity 0.764; specificity 0.919). The synthetic minority oversampling method further improved the performance of the model using combined features (AUC 0.960; sensitivity 0.845; specificity 0.929). Conclusions Clinical and imaging features can be used for automated severity assessment of COVID-19 and can potentially help triage patients with COVID-19 and prioritize care delivery to those at a higher risk of severe disease.

show abstract

“…Hence, it is extremely important to understand the critical factors associated with the severity of COVID-19 and provide convenient and efficient diagnostic methods. Xiao et al (12) developed an artificial intelligence-assisted tool using computed tomography (CT) imaging to predict disease severity and further estimate the risk of developing severe disease in patients suffering from COVID-19.…”

Section: Introductionmentioning

confidence: 99%

Prediction of COVID-19 Patients at High Risk of Progression to Severe Disease

Dai

Zeng

Cui

et al. 2020

Front. Public Health

View full text Add to dashboard Cite

In order to develop a novel scoring model for the prediction of coronavirus disease-19 (COVID-19) patients at high risk of severe disease, we retrospectively studied 419 patients from five hospitals in Shanghai, Hubei, and Jiangsu Provinces from January 22 to March 30, 2020. Multivariate Cox regression and orthogonal projections to latent structures discriminant analysis (OPLS-DA) were both used to identify high-risk factors for disease severity in COVID-19 patients. The prediction model was developed based on four high-risk factors. Multivariate analysis showed that comorbidity [hazard ratio (HR) 3.17, 95% confidence interval (CI) 1.96–5.11], albumin (ALB) level (HR 3.67, 95% CI 1.91–7.02), C-reactive protein (CRP) level (HR 3.16, 95% CI 1.68–5.96), and age ≥60 years (HR 2.31, 95% CI 1.43–3.73) were independent risk factors for disease severity in COVID-19 patients. OPLS-DA identified that the top five influencing parameters for COVID-19 severity were CRP, ALB, age ≥60 years, comorbidity, and lactate dehydrogenase (LDH) level. When incorporating the above four factors, the nomogram had a good concordance index of 0.86 (95% CI 0.83–0.89) and had an optimal agreement between the predictive nomogram and the actual observation with a slope of 0.95 (R2 = 0.89) in the 7-day prediction and 0.96 (R2 = 0.92) in the 14-day prediction after 1,000 bootstrap sampling. The area under the receiver operating characteristic curve of the COVID-19-American Association for Clinical Chemistry (AACC) model was 0.85 (95% CI 0.81–0.90). According to the probability of severity, the model divided the patients into three groups: low risk, intermediate risk, and high risk. The COVID-19-AACC model is an effective method for clinicians to screen patients at high risk of severe disease.

show abstract

Development and Validation of a Deep Learning-Based Model Using Computed Tomography Imaging for Predicting Disease Severity of Coronavirus Disease 2019

Cited by 64 publications

References 32 publications

Densely connected convolutional networks-based COVID-19 screening model

Densely connected convolutional networks-based COVID-19 screening model

Development and Validation of a Machine Learning Approach for Automated Severity Assessment of COVID-19 Based on Clinical and Imaging Data: Retrospective Study

Prediction of COVID-19 Patients at High Risk of Progression to Severe Disease

Contact Info

Product

Resources

About