Ensemble Deep Learning Derived from Transfer Learning for Classification of COVID-19 Patients on Hybrid Deep-Learning-Based Lung Segmentation: A Data Augmentation and Balancing Framework.

BACKGROUND AND MOTIVATION: Lung computed tomography (CT) techniques are high-resolution and are well adopted in the intensive care unit (ICU) for COVID-19 disease control classification. Most artificial intelligence (AI) systems do not undergo generalization and are typically overfitted. Such trained AI systems are not practical for clinical settings and therefore do not give accurate results when executed on unseen data sets. We hypothesize that ensemble deep learning (EDL) is superior to deep transfer learning (TL) in both non-augmented and augmented frameworks. METHODOLOGY: The system consists of a cascade of quality control, ResNet-UNet-based hybrid deep learning for lung segmentation, and seven models using TL-based classification followed by five types of EDL's. To prove our hypothesis, five different kinds of data combinations (DC) were designed using a combination of two multicenter cohorts-Croatia (80 COVID) and Italy (72 COVID and 30 controls)-leading to 12,000 CT slices. As part of generalization, the system was tested on unseen data and statistically tested for reliability/stability. RESULTS: Using the K5 (80:20) cross-validation protocol on the balanced and augmented dataset, the five DC datasets improved TL mean accuracy by 3.32%, 6.56%, 12.96%, 47.1%, and 2.78%, respectively. The five EDL systems showed improvements in accuracy of 2.12%, 5.78%, 6.72%, 32.05%, and 2.40%, thus validating our hypothesis. All statistical tests proved positive for reliability and stability. CONCLUSION: EDL showed superior performance to TL systems for both (a) unbalanced and unaugmented and (b) balanced and augmented datasets for both (i) seen and (ii) unseen paradigms, validating both our hypotheses.

May an incidental finding on chest CT be a predictor of access in intensive care unit? Role of hepatic steatosis in patients affected by SARS-CoV-2.

Background: Due to the high prevalence of hepatic steatosis (HS), the aim of the study is to verify the frequency of HS incidentally detected in chest computed tomography (CT) imaging in our population affected by SARS-CoV-2 and to investigate its association with the severity of the infection and outcome in terms of hospitalization. Design and methods: We retrospectively analyzed 500 patients with flu syndrome and clinically suspected of having Sars-CoV-2 infection who underwent unenhanced chest CT and have positive RT-PCR tests for Sars-CoV-2 RNA. Two radiologists both with >5 years of thoracic imaging experience, evaluated the images in consensus, without knowing the RT-PCR results. Liver density was measured by a region of interest (ROI), using a liver attenuation value ≤40 Hounsfield units (HU). Results: On 480 patients, 23.1% (111/480) had an incidental findings of HS on chest CT. The steatosis group, included 83 (74.7%) males and 28 (25.3%) females. Patients with HS were more likely to be hospitalized in the intensive care unit (ICU). On univariate analysis, there is a correlation between probability to be intubate (access in the ICU) and HS: patients with HS are twice as likely to be intubated (OR 2.04, CI 95% 1.11-3.73). Conclusion: Chest CT is an important diagnostic tool for COVID-19 and can provide information about the prognosis of the disease. HS can easily be detected on chest CT taken for the diagnosis of the COVID-19 disease, is an important sign for a poor prognosis and possible predictor of admission in ICU.

COVLIAS 2.0-cXAI: Cloud-Based Explainable Deep Learning System for COVID-19 Lesion Localization in Computed Tomography Scans.

Background: The previous COVID-19 lung diagnosis system lacks both scientific validation and the role of explainable artificial intelligence (AI) for understanding lesion localization. This study presents a cloud-based explainable AI, the "COVLIAS 2.0-cXAI" system using four kinds of class activation maps (CAM) models. Methodology: Our cohort consisted of ~6000 CT slices from two sources (Croatia, 80 COVID-19 patients and Italy, 15 control patients). COVLIAS 2.0-cXAI design consisted of three stages: (i) automated lung segmentation using hybrid deep learning ResNet-UNet model by automatic adjustment of Hounsfield units, hyperparameter optimization, and parallel and distributed training, (ii) classification using three kinds of DenseNet (DN) models (DN-121, DN-169, DN-201), and (iii) validation using four kinds of CAM visualization techniques: gradient-weighted class activation mapping (Grad-CAM), Grad-CAM++, score-weighted CAM (Score-CAM), and FasterScore-CAM. The COVLIAS 2.0-cXAI was validated by three trained senior radiologists for its stability and reliability. The Friedman test was also performed on the scores of the three radiologists. Results: The ResNet-UNet segmentation model resulted in dice similarity of 0.96, Jaccard index of 0.93, a correlation coefficient of 0.99, with a figure-of-merit of 95.99%, while the classifier accuracies for the three DN nets (DN-121, DN-169, and DN-201) were 98%, 98%, and 99% with a loss of ~0.003, ~0.0025, and ~0.002 using 50 epochs, respectively. The mean AUC for all three DN models was 0.99 (p < 0.0001). The COVLIAS 2.0-cXAI showed 80% scans for mean alignment index (MAI) between heatmaps and gold standard, a score of four out of five, establishing the system for clinical settings. Conclusions: The COVLIAS 2.0-cXAI successfully showed a cloud-based explainable AI system for lesion localization in lung CT scans.

COVLIAS 1.0Lesion vs. MedSeg: An Artificial Intelligence Framework for Automated Lesion Segmentation in COVID-19 Lung Computed Tomography Scans.

Background: COVID-19 is a disease with multiple variants, and is quickly spreading throughout the world. It is crucial to identify patients who are suspected of having COVID-19 early, because the vaccine is not readily available in certain parts of the world. Methodology: Lung computed tomography (CT) imaging can be used to diagnose COVID-19 as an alternative to the RT-PCR test in some cases. The occurrence of ground-glass opacities in the lung region is a characteristic of COVID-19 in chest CT scans, and these are daunting to locate and segment manually. The proposed study consists of a combination of solo deep learning (DL) and hybrid DL (HDL) models to tackle the lesion location and segmentation more quickly. One DL and four HDL models­namely, PSPNet, VGG-SegNet, ResNet-SegNet, VGG-UNet, and ResNet-UNet­were trained by an expert radiologist. The training scheme adopted a fivefold cross-validation strategy on a cohort of 3000 images selected from a set of 40 COVID-19-positive individuals. Results: The proposed variability study uses tracings from two trained radiologists as part of the validation. Five artificial intelligence (AI) models were benchmarked against MedSeg. The best AI model, ResNet-UNet, was superior to MedSeg by 9% and 15% for Dice and Jaccard, respectively, when compared against MD 1, and by 4% and 8%, respectively, when compared against MD 2. Statistical tests­namely, the Mann−Whitney test, paired t-test, and Wilcoxon test­demonstrated its stability and reliability, with p < 0.0001. The online system for each slice was <1 s. Conclusions: The AI models reliably located and segmented COVID-19 lesions in CT scans. The COVLIAS 1.0Lesion lesion locator passed the intervariability test.

Lung vessel volume evaluated with CALIPER software is an independent predictor of mortality in COVID-19 patients: a multicentric retrospective analysis.

INTRODUCTION: Computer-Aided Lung Informatics for Pathology Evaluation and Ratings (CALIPER) software has already been widely used in the evaluation of interstitial lung diseases (ILD) but has not yet been tested in patients affected by COVID-19. Our aim was to use it to describe the relationship between Coronavirus Disease 2019 (COVID-19) outcome and the CALIPER-detected pulmonary vascular-related structures (VRS). MATERIALS AND METHODS: We performed a multicentric retrospective study enrolling 570 COVID-19 patients who performed a chest CT in emergency settings in two different institutions. Fifty-three age- and sex-matched healthy controls were also identified. Chest CTs were analyzed with CALIPER identifying the percentage of VRS over the total lung parenchyma. Patients were followed for up to 72 days recording mortality and required intensity of care. RESULTS: There was a statistically significant difference in VRS between COVID-19-positive patients and controls (median (iqr) 4.05 (3.74) and 1.57 (0.40) respectively, p = 0.0001). VRS showed an increasing trend with the severity of care, p < 0.0001. The univariate Cox regression model showed that VRS increase is a risk factor for mortality (HR 1.17, p < 0.0001). The multivariate analysis demonstrated that VRS is an independent explanatory factor of mortality along with age (HR 1.13, p < 0.0001). CONCLUSION: Our study suggests that VRS increases with the required intensity of care, and it is an independent explanatory factor for mortality. KEY POINTS: • The percentage of vascular-related structure volume (VRS) in the lung is significatively increased in COVID-19 patients. • VRS showed an increasing trend with the required intensity of care, test for trend p< 0.0001. • Univariate and multivariate Cox models showed that VRS is a significant and independent explanatory factor of mortality.

COVLIAS 1.0 vs. MedSeg: Artificial Intelligence-Based Comparative Study for Automated COVID-19 Computed Tomography Lung Segmentation in Italian and Croatian Cohorts.

(1) Background: COVID-19 computed tomography (CT) lung segmentation is critical for COVID lung severity diagnosis. Earlier proposed approaches during 2020-2021 were semiautomated or automated but not accurate, user-friendly, and industry-standard benchmarked. The proposed study compared the COVID Lung Image Analysis System, COVLIAS 1.0 (GBTI, Inc., and AtheroPointTM, Roseville, CA, USA, referred to as COVLIAS), against MedSeg, a web-based Artificial Intelligence (AI) segmentation tool, where COVLIAS uses hybrid deep learning (HDL) models for CT lung segmentation. (2) Materials and Methods: The proposed study used 5000 ITALIAN COVID-19 positive CT lung images collected from 72 patients (experimental data) that confirmed the reverse transcription-polymerase chain reaction (RT-PCR) test. Two hybrid AI models from the COVLIAS system, namely, VGG-SegNet (HDL 1) and ResNet-SegNet (HDL 2), were used to segment the CT lungs. As part of the results, we compared both COVLIAS and MedSeg against two manual delineations (MD 1 and MD 2) using (i) Bland-Altman plots, (ii) Correlation coefficient (CC) plots, (iii) Receiver operating characteristic curve, and (iv) Figure of Merit and (v) visual overlays. A cohort of 500 CROATIA COVID-19 positive CT lung images (validation data) was used. A previously trained COVLIAS model was directly applied to the validation data (as part of Unseen-AI) to segment the CT lungs and compare them against MedSeg. (3) Result: For the experimental data, the four CCs between COVLIAS (HDL 1) vs. MD 1, COVLIAS (HDL 1) vs. MD 2, COVLIAS (HDL 2) vs. MD 1, and COVLIAS (HDL 2) vs. MD 2 were 0.96, 0.96, 0.96, and 0.96, respectively. The mean value of the COVLIAS system for the above four readings was 0.96. CC between MedSeg vs. MD 1 and MedSeg vs. MD 2 was 0.98 and 0.98, respectively. Both had a mean value of 0.98. On the validation data, the CC between COVLIAS (HDL 1) vs. MedSeg and COVLIAS (HDL 2) vs. MedSeg was 0.98 and 0.99, respectively. For the experimental data, the difference between the mean values for COVLIAS and MedSeg showed a difference of <2.5%, meeting the standard of equivalence. The average running times for COVLIAS and MedSeg on a single lung CT slice were ~4 s and ~10 s, respectively. (4) Conclusions: The performances of COVLIAS and MedSeg were similar. However, COVLIAS showed improved computing time over MedSeg.

Machine Learning to Predict In-Hospital Mortality in COVID-19 Patients Using Computed Tomography-Derived Pulmonary and Vascular Features.

Pulmonary parenchymal and vascular damage are frequently reported in COVID-19 patients and can be assessed with unenhanced chest computed tomography (CT), widely used as a triaging exam. Integrating clinical data, chest CT features, and CT-derived vascular metrics, we aimed to build a predictive model of in-hospital mortality using univariate analysis (Mann-Whitney U test) and machine learning models (support vectors machines (SVM) and multilayer perceptrons (MLP)). Patients with RT-PCR-confirmed SARS-CoV-2 infection and unenhanced chest CT performed on emergency department admission were included after retrieving their outcome (discharge or death), with an 85/15% training/test dataset split. Out of 897 patients, the 229 (26%) patients who died during hospitalization had higher median pulmonary artery diameter (29.0 mm) than patients who survived (27.0 mm, p < 0.001) and higher median ascending aortic diameter (36.6 mm versus 34.0 mm, p < 0.001). SVM and MLP best models considered the same ten input features, yielding a 0.747 (precision 0.522, recall 0.800) and 0.844 (precision 0.680, recall 0.567) area under the curve, respectively. In this model integrating clinical and radiological data, pulmonary artery diameter was the third most important predictor after age and parenchymal involvement extent, contributing to reliable in-hospital mortality prediction, highlighting the value of vascular metrics in improving patient stratification.

CT-derived Chest Muscle Metrics for Outcome Prediction in Patients with COVID-19.

Background Lower muscle mass is a known predictor of unfavorable outcomes, but its prognostic impact on patients with COVID-19 is unknown. Purpose To investigate the contribution of CT-derived muscle status in predicting clinical outcomes in patients with COVID-19. Materials and Methods Clinical or laboratory data and outcomes (intensive care unit [ICU] admission and death) were retrospectively retrieved for patients with reverse transcriptase polymerase chain reaction-confirmed SARS-CoV-2 infection, who underwent chest CT on admission in four hospitals in Northern Italy from February 21 to April 30, 2020. The extent and type of pulmonary involvement, mediastinal lymphadenopathy, and pleural effusion were assessed. Cross-sectional areas and attenuation by paravertebral muscles were measured on axial CT images at the T5 and T12 vertebral level. Multivariable linear and binary logistic regression, including calculation of odds ratios (ORs) with 95% CIs, were used to build four models to predict ICU admission and death, which were tested and compared by using receiver operating characteristic curve analysis. Results A total of 552 patients (364 men and 188 women; median age, 65 years [interquartile range, 54-75 years]) were included. In a CT-based model, lower-than-median T5 paravertebral muscle areas showed the highest ORs for ICU admission (OR, 4.8; 95% CI: 2.7, 8.5; P < .001) and death (OR, 2.3; 95% CI: 1.0, 2.9; P = .03). When clinical variables were included in the model, lower-than-median T5 paravertebral muscle areas still showed the highest ORs for both ICU admission (OR, 4.3; 95%: CI: 2.5, 7.7; P < .001) and death (OR, 2.3; 95% CI: 1.3, 3.7; P = .001). At receiver operating characteristic analysis, the CT-based model and the model including clinical variables showed the same area under the receiver operating characteristic curve (AUC) for ICU admission prediction (AUC, 0.83; P = .38) and were not different in terms of predicting death (AUC, 0.86 vs AUC, 0.87, respectively; P = .28). Conclusion In hospitalized patients with COVID-19, lower muscle mass on CT images was independently associated with intensive care unit admission and in-hospital mortality. © RSNA, 2021 Online supplemental material is available for this article.

Chest CT accuracy in diagnosing COVID-19 during the peak of the Italian epidemic: A retrospective correlation with RT-PCR testing and analysis of discordant cases.

OBJECTIVES: The goal of this study was to assess chest computed tomography (CT) diagnostic accuracy in clinical practice using RT-PCR as standard of reference. METHODS: From March 4th to April 9th 2020, during the peak of the Italian COVID-19 epidemic, we enrolled a series of 773 patients that performed both non-contrast chest CT and RT-PCR with a time interval no longer than a week due to suspected SARS-CoV-2 infection. The diagnostic performance of CT was evaluated according to sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic accuracy, considering RT-PCR as the reference standard. An analysis on the patients with discrepant CT scan and RT-PCR result and on the patient with both negative tests was performed. RESULTS: RT-PCR testing showed an overall positive rate of 59.8 %. CT sensitivity, specificity, PPV, NPV, and accuracy for SARS-CoV-2 infection were 90.7 % [95 % IC, 87.7%-93.2%], 78.8 % [95 % IC, 73.8-83.2%], 86.4 % [95 % IC, 76.1 %-88.9 %], 85.1 % [95 % IC, 81.0 %-88.4] and 85.9 % [95 % IC 83.2-88.3%], respectively. Twenty-five/66 (37.6 %) patients with positive CT and negative RT-PCR results and 12/245 (4.9 %) patients with both negative tests were nevertheless judged as positive cases by the clinicians based on clinical and epidemiological criteria and consequently treated. CONCLUSIONS: In our experience, in a context of high pre-test probability, CT scan shows good sensitivity and a consistently higher specificity for the diagnosis of COVID-19 pneumonia than what reported by previous studies, especially when clinical and epidemiological features are taken into account.