Deep learning reveals lung shape differences on baseline chest CT between mild and severe COVID-19: A multi-site retrospective study.

Severe COVID-19 can lead to extensive lung disease causing lung architectural distortion. In this study we employed machine learning and statistical atlas-based approaches to explore possible changes in lung shape among COVID-19 patients and evaluated whether the extent of these changes was associated with COVID-19 severity. On a large multi-institutional dataset (N = 3443), three different populations were defined; a) healthy (no COVID-19), b) mild COVID-19 (no ventilator required), c) severe COVID-19 (ventilator required), and the presence of lung shape differences between them were explored using baseline chest CT. Significant lung shape differences were observed along mediastinal surfaces of the lungs across all severity of COVID-19 disease. Additionally, differences were seen on basal surfaces of the lung when compared between healthy and severe COVID-19 patients. Finally, an AI model (a 3D residual convolutional network) characterizing these shape differences coupled with lung infiltrates (ground-glass opacities and consolidation regions) was found to be associated with COVID-19 severity.

Enhancing cardiovascular risk prediction through AI-enabled calcium-omics.

Whole-heart coronary calcium Agatston score is a well-established predictor of major adverse cardiovascular events (MACE), but it does not account for individual calcification features related to the pathophysiology of the disease (e.g., multiple-vessel disease, spread of the disease along the vessel, stable calcifications, numbers of lesions, and density). We used novel, hand-crafted calcification features (calcium-omics); Cox time-to-event modeling; elastic net; and up and down synthetic sampling methods for imbalanced data, to assess MACE risk. We used 2457 CT calcium score (CTCS) images enriched for MACE events from our large no-cost CLARIFY program (ClinicalTrials.gov Identifier: NCT04075162). Among calcium-omics features, numbers of calcifications, LAD mass, and diffusivity (a measure of spatial distribution) were especially important determinants of increased risk, with dense calcification (> 1000HU, stable calcifications) associated with reduced risk Our calcium-omics model with (training/testing, 80/20) gave C-index (80.5%/71.6%) and 2-year AUC (82.4%/74.8%). Although the C-index is notoriously impervious to model improvements, calcium-omics compared favorably to Agatston and gave a significant difference (P < 0.001). The calcium-omics model identified 73.5% of MACE cases in the high-risk group, a 13.2% improvement as compared to Agatston, suggesting that calcium-omics could be used to better identity candidates for intensive follow-up and therapies. The categorical net-reclassification index was NRI = 0.153. Our findings from this exploratory study suggest the utility of calcium-omics in improved risk prediction. These promising results will pave the way for more extensive, multi-institutional studies of calcium-omics.

Emergency department imaging utilization post-transcatheter aortic valve replacement: single institution 7-year experience.

PURPOSE: This study aims to highlight presentations, acute findings and imaging phenotypes of patients presenting to the emergency department (ED) within 30 days of a transcatheter aortic valve replacement (TAVR). METHODS: A retrospective review of patients diagnosed with aortic valve disease who underwent a TAVR between Jan 2015 and Nov 2021 at a large academic medical center was completed. From an initial 1271 patients, 146 were included based on their presentation to the ED within 30 days post-TAVR procedure. Patient data, including ED presentation details and imaging results, were recorded and de-identified. RESULTS: Of the 146 post-TAVR patients, there were 168 ED visits within 30 days. The median time to ED after TAVR was 12 days. Respiratory symptoms were the most common complaint (27%). Neurological (23%) and cardiovascular symptoms (18%) followed. Cross-sectional imaging was conducted 250 times across visits, with an average of 1.7 scans per patient. CTs were most frequently used, followed by ultrasounds, especially echocardiograms and duplex extremity vasculature ultrasounds. 30.1% of patients had acute findings from imaging. Specific findings included heart failure (5.5%), access site complications (5.5%), pneumonia (5.5%), intracranial pathologies (3.4% for strokes and 0.7% for hematoma), and pleural effusion (3.4%). Echocardiograms and CTA chest were most associated with significant acute findings. CONCLUSION: Our study highlights the vital role of early and accurate imaging in post-TAVR patients within 30 days post-procedure. As transcatheter approaches rise in popularity, emergency radiologists become instrumental in diagnosing common post-procedural presentations. Continued research is essential to devise post-discharge strategies to curtail readmissions and related costs. Proper imaging ensures prompt, effective care, enhancing overall patient outcomes.

Enhancing cardiovascular risk prediction through AI-enabled calcium-omics.

Background: Coronary artery calcium (CAC) is a powerful predictor of major adverse cardiovascular events (MACE). Traditional Agatston score simply sums the calcium, albeit in a non-linear way, leaving room for improved calcification assessments that will more fully capture the extent of disease. Objective: To determine if AI methods using detailed calcification features (i.e., calcium-omics) can improve MACE prediction. Methods: We investigated additional features of calcification including assessment of mass, volume, density, spatial distribution, territory, etc. We used a Cox model with elastic-net regularization on 2457 CT calcium score (CTCS) enriched for MACE events obtained from a large no-cost CLARIFY program (ClinicalTrials.gov Identifier: NCT04075162). We employed sampling techniques to enhance model training. We also investigated Cox models with selected features to identify explainable high-risk characteristics. Results: Our proposed calcium-omics model with modified synthetic down sampling and up sampling gave C-index (80.5%/71.6%) and two-year AUC (82.4%/74.8%) for (80:20, training/testing), respectively (sampling was applied to the training set only). Results compared favorably to Agatston which gave C-index (71.3%/70.3%) and AUC (71.8%/68.8%), respectively. Among calcium-omics features, numbers of calcifications, LAD mass, and diffusivity (a measure of spatial distribution) were important determinants of increased risk, with dense calcification (>1000HU) associated with lower risk. The calcium-omics model reclassified 63% of MACE patients to the high risk group in a held-out test. The categorical net-reclassification index was NRI=0.153. Conclusions: AI analysis of coronary calcification can lead to improved results as compared to Agatston scoring. Our findings suggest the utility of calcium-omics in improved prediction of risk.

Clinical Implementation of an Artificial Intelligence Tool in the Detection and Management of Pneumothoraces in Patients With COVID-19.

In this report, we present a series involving critically ill patients with known coronavirus disease (COVID-19) infection where a portable X-ray machine equipped with artificial intelligence (AI) software aided in the urgent radiographic diagnosis of pneumothorax. These cases demonstrate how real-world clinical employment of AI tools capable of analyzing and prioritizing studies in the radiologist's worklist can potentially lead to earlier detection of emergent findings like pneumothorax. The use of AI tools in this manner has the potential to both improve radiology workflow and add significant clinical value in managing critically ill patient populations, such as those with severe COVID-19 infection.

An End-to-End Integrated Clinical and CT-Based Radiomics Nomogram for Predicting Disease Severity and Need for Ventilator Support in COVID-19 Patients: A Large Multisite Retrospective Study.

Objective: The disease COVID-19 has caused a widespread global pandemic with ~3. 93 million deaths worldwide. In this work, we present three models-radiomics (MRM), clinical (MCM), and combined clinical-radiomics (MRCM) nomogram to predict COVID-19-positive patients who will end up needing invasive mechanical ventilation from the baseline CT scans. Methods: We performed a retrospective multicohort study of individuals with COVID-19-positive findings for a total of 897 patients from two different institutions (Renmin Hospital of Wuhan University, D1 = 787, and University Hospitals, US D2 = 110). The patients from institution-1 were divided into 60% training, D 1 T ( N = 473 ) , and 40% test set D 1 V ( N = 314 ) . The patients from institution-2 were used for an independent validation test set D 2 V ( N = 110 ) . A U-Net-based neural network (CNN) was trained to automatically segment out the COVID consolidation regions on the CT scans. The segmented regions from the CT scans were used for extracting first- and higher-order radiomic textural features. The top radiomic and clinical features were selected using the least absolute shrinkage and selection operator (LASSO) with an optimal binomial regression model within D 1 T . Results: The three out of the top five features identified using D 1 T were higher-order textural features (GLCM, GLRLM, GLSZM), whereas the last two features included the total absolute infection size on the CT scan and the total intensity of the COVID consolidations. The radiomics model (MRM) was constructed using the radiomic score built using the coefficients obtained from the LASSO logistic model used within the linear regression (LR) classifier. The MRM yielded an area under the receiver operating characteristic curve (AUC) of 0.754 (0.709-0.799) on D 1 T , 0.836 on D 1 V , and 0.748 D 2 V . The top prognostic clinical factors identified in the analysis were dehydrogenase (LDH), age, and albumin (ALB). The clinical model had an AUC of 0.784 (0.743-0.825) on D 1 T , 0.813 on D 1 V , and 0.688 on D 2 V . Finally, the combined model, MRCM integrating radiomic score, age, LDH and ALB, yielded an AUC of 0.814 (0.774-0.853) on D 1 T , 0.847 on D 1 V , and 0.771 on D 2 V . The MRCM had an overall improvement in the performance of ~5.85% ( D 1 T : p = 0.0031; D 1 V p = 0.0165; D 2 V : p = 0.0369) over MCM. Conclusion: The novel integrated imaging and clinical model (MRCM) outperformed both models (MRM) and (MCM). Our results across multiple sites suggest that the integrated nomogram could help identify COVID-19 patients with more severe disease phenotype and potentially require mechanical ventilation.

Prevalence of thoracic aortic aneurysm in patients referred for no/low-charge coronary artery calcium scoring: Insights from the CLARIFY registry.

Objective: Low-dose cardiac-gated chest CTs allow for simultaneous evaluation of coronary artery calcification and aortic size. We sought to evaluate the prevalence of thoracic aortic dilation (TAD) and thoracic aortic aneurysm (TAA) in a large cohort of patients undergoing coronary artery calcium (CAC) screening. Methods: We reviewed all patients from a large, prospective no-charge CAC screening program (CLARIFY, Clinicaltrials.gov NCT04075162) for whom measurements of the ascending aorta were available. TAD was defined as an ascending aortic diameter ≥4.0cm, while TAA was defined as ascending aortic diameter ≥ 4.5cm. We explored associations between patient characteristics, CAC, and the prevalence of TAD/TAA. Results: A total of 36,356 patients enrolled in the CLARIFY program underwent analysis for TAD/TAA. 3,130 patients (8.6%) had TAD and 237 (0.7%) had TAA. Patients with TAA were older (63±8 vs 59±10 years, p < 0.001), more likely to be male (87% vs 49%, p < 0.001), have higher BMI (32 vs 30 kg/m2, p < 0.001), and 10-year atherosclerotic cardiovascular disease estimated risk (18% vs 12%, p < 0.001). Similar differences were observed for individuals with TAD compared to individuals without TAD with respect to age (63 vs 59 years, p < 0.001), percent male (76% vs 46%, p < 0.001), BMI (32 vs 30 kg/m2, p < 0.001), and 10-year predicted risk (17% vs 11%, p < 0.001). CAC score was associated with prevalence of TAD (4.9% in those with CAC 0 to 16.5% in those with CAC≥400) and TAA (0.3% in those with CAC of 0 to 1.5% in those with CAC ≥400). Conclusion: In this large, prospective study of patients undergoing no-charge CAC screening, 8.6% had TAD (≥4.0cm) and 0.7% had TAA (≥4.5cm). Our results highlight a high yield of TAD/TAA diagnosis in this targeted cohort with cardiovascular risk factors and supports the role of no-charge CAC as a population-level strategy.

Impact of low/no-charge coronary artery calcium scoring on statin eligibility and outcomes in women: The CLARIFY study.

Background: Prior studies have suggested significant underutilization of statins in women and worse cardiovascular outcomes. Data examining the impact of real-world coronary artery calcium (CAC) scoring to improve utilization of preventive therapies and outcomes is limited. Methods: In a prospective registry study of low cost or no-cost CAC scoring between 2014 and 19 (CLARIFY Study, Clinicaltrials.gov NCT04075162), we sought to study the association of CAC scoring on statin utilization, blood lipids (LDL, total cholesterol, triglycerides), downstream ischemic testing (coronary angiography and stress testing), coronary revascularization and outcomes (MI, stroke, death) in women compared with men. Eligibility for statin initiation was defined as atherosclerotic cardiovascular disease pooled cohort equation (ASCVD-PCE) ≥ 7.5% and CAC≥100/≥75th percentile. Results: A total of 52,151 patients (26,336 women and 25,815 men) were enrolled. Women were more likely to have CAC 0 (51% vs 30%, P<0.001). Among patients not eligible for statin by PCE, CAC reclassified statin eligibility in a smaller proportion of women than men (25.4% vs 30%, P<0.001), while among patients eligible for statin by PCE, CAC was more likely to downgrade risk/statin eligibility in women than men (30.1% vs 48.4%, P<0.001). After CAC scoring, statin initiation was similar in women and men, but high-intensity statin use was lower in women (CAC-adjusted HR 0.76 [0.70-0.83], P<0.001). Women had similar reduction in LDL cholesterol levels compared with men. There was no difference between men and women with respect to CAC-stratified major adverse cardiovascular events. Conclusion: CAC scoring primarily served to downgrade statin eligibility in women compared with men. Women had similar CAC risk-guided reductions in LDL cholesterol compared with men.

Incidental Finding of Secondary Tumoral Calcinosis Following Cardiothoracic Surgery: The Role of Multimodality Imaging Including Spectral Detector Computed Tomography.

Tumoral calcinosis is a rare syndrome that affects mostly soft tissues. It is characterized by calcium salt deposition in the periarticular soft tissue surrounding bony structures forming slow-growing, seldom asymptomatic masses. This case report describes a 41-year-old male with end-stage renal disease on home hemodialysis, who presented with an unusual rapidly progressive mass overlying the manubrium and suprasternal notch, following recent cardiothoracic surgery, which was initially felt to be a hematoma. The case highlights the role of spectral detector computed tomography (SDCT) in reaching the correct diagnosis of tumoral calcinosis as well as demonstrating additional changes of ectopic parathyroid hyperplasia in the anterior mediastinum.

"Digital Tomosynthesis" As a Technique for the Evaluation of Endobronchial Stents in Lung Transplant Recipients.

Lung transplant patients often suffer from posttransplant airway pathologies that require placement of endobronchial stents. In addition to surveillance bronchoscopy, patients often undergo radiographic stent evaluations. Chest x-rays are extremely limited in their ability to diagnose stent complications, so many patients require chest computed tomography (CT) scans for stent evaluation. Chest CT scans are costly and expose patients to higher cumulative radiation doses. Digital tomosynthesis (DTS) is an imaging modality that provides high-resolution images using limited angle tomography. The costs and radiation doses are comparable to conventional x-ray. We present a series of 4 postlung transplant patients with bronchial stents in whom we performed DTS and chest x-ray simultaneously. The DTS images were far superior to chest x-ray and comparable with CT in evaluating the placement and patency of the stents, especially in the case of silicone stents. Furthermore, the improved resolution provided clinically relevant diagnostic information that resulted in therapeutic bronchoscopy for suctioning of mucus impaction in one of the patients.

Coronary Artery Calcium Scoring: Current Status and Future Directions.

Coronary artery calcium (CAC) scores obtained from CT scans have been shown to be prognostic in assessment of the risk for development of cardiovascular diseases, facilitating the prediction of outcome in asymptomatic individuals. Currently, several methods to calculate the CAC score exist, and each has its own set of advantages and disadvantages. Agatston CAC scoring is the most extensively used method. CAC scoring is currently recommended for use in asymptomatic individuals to predict the risk of developing cardiovascular diseases and the disease-specific mortality. In specific subsets of patients, the CAC score has also been recommended for reclassifying cardiovascular risk and aiding in decision making when planning primary prevention interventions such as statin therapy. The progression of CAC scores on follow-up images has been shown to be linked to risk of myocardial infarction and cardiovascular mortality. While the CAC score is a validated tool used clinically, several challenges, including various pitfalls associated with the acquisition, calculation, and interpretation of the score, prevent more widespread adoption of this metric. Recent research has been focused extensively on strategies to improve existing scoring methods, including measuring calcium attenuation, detecting microcalcifications, and focusing on extracoronary calcifications, and on strategies to improve image acquisition. A better understanding of CAC scoring approaches will help radiologists and other physicians better use and interpret these scores in their workflows. An invited commentary by S. Gupta is available online. Online supplemental material is available for this article. ©RSNA, 2022.

Combined Radiomic and Visual Assessment for Improved Detection of Lung Adenocarcinoma Invasiveness on Computed Tomography Scans: A Multi-Institutional Study.

Objective: The timing and nature of surgical intervention for semisolid abnormalities are dependent upon distinguishing between adenocarcinoma-in-situ (AIS), minimally invasive adenocarcinoma (MIA), and invasive adenocarcinoma (INV). We sought to develop and evaluate a quantitative imaging method to determine invasiveness of small, ground-glass lesions on computed tomography (CT) chest scans. Methods: The study comprised 268 patients from 4 institutions with resected (<=3 cm) semisolid lesions with confirmed histopathological diagnosis of MIA/AIS or INV. A total of 248 radiomic texture features from within the tumor nodule (intratumoral) and adjacent to the nodule (peritumoral) were extracted from manually annotated lung nodules of chest CT scans. The datasets were randomly divided, with 40% of patients used for training and 60% used for testing the machine classifier (Training DTrain, N=106; Testing, DTest, N=162). Results: The top five radiomic stable features included four intratumoral (Laws and Haralick feature families) and one peritumoral feature within 3 to 6 mm of the nodule (CoLlAGe feature family), which successfully differentiated INV from MIA/AIS nodules with an AUC of 0.917 [0.867-0.967] on DTrain and 0.863 [0.79-0.931] on DTest. The radiomics model successfully differentiated INV from MIA cases (<1 cm AUC: 0.76 [0.53-0.98], 1-2 cm AUC: 0.92 [0.85-0.98], 2-3 cm AUC: 0.95 [0.88-1]). The final integrated model combining the classifier with the radiologists' score gave the best AUC on DTest (AUC=0.909, p<0.001). Conclusions: Addition of advanced image analysis via radiomics to the routine visual assessment of CT scans help better differentiate adenocarcinoma subtypes and can aid in clinical decision making. Further prospective validation in this direction is warranted.

Imaging of the Left Atrial Appendage Before Occluder Device Placement: Evaluation of Virtual Monoenergetic Images in a Single-Bolus Dual-Phase Protocol.

PURPOSE: Preimplantation cardiac computed tomography (CT) for assessment of the left atrial appendage (LAA) enables correct sizing of the device and the detection of contraindications, such as thrombi. In the arterial phase, distinction between false filling defects and true thrombi can be hampered by insufficient contrast medium distribution. A delayed scan can be used to further differentiate both conditions, but contrast in these acquisitions is relatively lower. In this study, we investigated whether virtual monoenergetic images (VMI) from dual-energy spectral detector CT (SDCT) can be used to enhance contrast and visualization in the delayed phase. MATERIALS AND METHODS: Forty-nine patients receiving SDCT imaging of the LAA were retrospectively enrolled. The imaging protocol comprised dual-phase acquisitions with single-bolus contrast injection. Conventional images (CI) from both phases and 40-keV VMI from the delayed phase were reconstructed. Attenuation, signal-, and contrast-to-noise ratios (SNR/CNR) were calculated by placing regions-of-interest in the LAA, left atrium, and muscular portion of interventricular septum. Two radiologists subjectively evaluated conspicuity and homogeneity of contrast distribution within the LAA. RESULTS: Contrast of the LAA decreased significantly in the delayed phase but was significantly improved by VMI, showing comparable attenuation, SNR, and CNR to CI from the arterial phase (attenuation/SNR/CNR, CI arterial phase: 266.0 ± 117.0 HU/14.2 ± 7.2/6.6 ± 3.9; CI-delayed phase: 107.6 ± 35.0 HU/5.9 ± 3.0/1.0 ± 1.0; VMI delayed phase: 260.3 ± 108.6 HU/18.2 ± 10.6/4.8 ± 3.4). The subjective reading confirmed the objective findings showing improved conspicuity and homogeneity in the delayed phase. CONCLUSIONS: The investigated single-bolus dual-phase acquisition protocol provided improved visualization of the LAA. Homogeneity of contrast media was higher in the delayed phase, while VMI maintained high contrast.

Large pulmonary artery pseudoaneurysm after CardioMEMS implantation: a case report.

Background: CardioMEMS heart failure (HF) system is an implantable wireless pressure sensor that is placed in a branch of the pulmonary artery (PA) for remote monitoring of PA pressures in patients with HF. Pulmonary artery injury/haemoptysis can occur during the sensor placement. Case summary: An 80-year-old male patient with HF with reduced ejection fraction (20%) underwent CardioMEMS HF system implantation for recurrent shortness of breath. He developed haemoptysis and dyspnoea during the procedure, which was managed with furosemide. The patient's computerized tomographic angiography showed a 3.4 cm pseudoaneurysm with active extravasation from the superior segmental branch of the left PA due to injury during device placement. The decision to embolize the pseudoaneurysm was made after a multi-disciplinary team meeting and discussion with the patient. The embolization procedure was carried out successfully with the final left pulmonary angiogram showed complete stasis and no further filling of the pseudoaneurysm sac. Discussion: The incidence of mortality in patients with PA injury from CardioMEMS devices is high, and therefore prompt diagnosis and management are critical. Pulmonary artery pseudoaneurysms are uncommon and present with haemoptysis. Transcatheter embolization has been shown to be a practical, effective, and safe therapeutic option in stable patients.

Effect of No-Charge Coronary Artery Calcium Scoring on Cardiovascular Prevention.

Prevention of cardiovascular disease is currently guided by probabilistic risk scores that may misclassify individual risk and commit many middle-aged patients to prolonged pharmacotherapy. The coronary artery calcium (CAC) score, although endorsed for intermediate-risk patients, is not widely adopted because of barriers in reimbursement. The impact of removing cost barrier on cardiovascular outcomes in real-world settings is not known. Within the University Hospitals Health System (Cleveland, Ohio), CAC was offered to patients with at least 1 cardiovascular risk factor at low charge between 2014 and 2017 ($99) and no charge from January 1, 2018 onward. CAC use and access, patient characteristics, reclassification of risk compared with the pooled cohort equations (PCEs) for atherosclerotic vascular disease, statin use, changes in parameters of cardiometabolic health, downstream cardiovascular testing, downstream coronary revascularization, and cardiovascular outcomes were evaluated. A total of 52,151 patients underwent CAC testing over the study period. Median 10-year PCE for atherosclerotic vascular disease, in the entire cohort was 8.3% (4.0% to 15.9%). Among patients with PCE >20%, 21% had CAC <100, whereas 37% of those with PCE <7.5% had CAC ≥100. Among patients who were not on statin before CAC testing, 1-year statin prescription was 24% and was significantly associated with higher CAC scores. Total cholesterol, low-density lipoprotein cholesterol, and triglycerides all decreased significantly 1-year after CAC, and the degree of decrease was strongly linked with CAC scores. One year after CAC, 14% underwent noninvasive ischemic evaluation, 1.4% underwent invasive coronary angiography, and 0.9% underwent revascularization. The majority (74%) of revascularization procedures occurred in patients with CAC >400. In conclusion, reducing or removing the cost burden of CAC leads to significant test uptake by patients, which is followed by reclassification of statin eligibility, increases in the use of preventive medications, and improvement in risk factors, with very low rates of invasive downstream testing.

Seamless Integration of Artificial Intelligence Into the Clinical Environment: Our Experience With a Novel Pneumothorax Detection Artificial Intelligence Algorithm.

Although interest in artificial intelligence (AI) has exploded in recent years and led to the development of numerous commercial and noncommercial algorithms, the process of implementing such tools into day-to-day clinical practice is rarely described in the burgeoning AI literature. In this report, we describe our experience with the successful integration of an AI-enabled mobile x-ray scanner with an FDA-approved algorithm for detecting pneumothoraces into an end-to-end solution capable of extracting, delivering, and prioritizing positive studies within our thoracic radiology clinical workflow. We also detail several sample cases from our AI algorithm and associated PACS workflow in action to highlight key insights from our experience. We hope this report can help inform other radiology enterprises seeking to evaluate and implement AI-related workflow solutions into daily clinical practice.

COVID-19 pandemic revisited: lessons the radiology community has learned a year later.

For more than 1 year, COVID-19 pandemic has impacted every aspect of our lives. This paper reviews the major challenges that the radiology community faced over the past year and the impact the pandemic had on the radiology practice, radiologist-in-training education, and radiology research. The lessons learned from COVID-19 pandemic can help the radiology community to be prepared for future outbreaks and new pandemics, preserve good habits, enhance cancer screening programs, and adapt to the changes in radiology education and scientific meetings.

Integrated Clinical and CT Based Artificial Intelligence Nomogram for Predicting Severity and Need for Ventilator Support in COVID-19 Patients: A Multi-Site Study.

Almost 25% of COVID-19 patients end up in ICU needing critical mechanical ventilation support. There is currently no validated objective way to predict which patients will end up needing ventilator support, when the disease is mild and not progressed. N = 869 patients from two sites (D1: N = 822, D2: N = 47) with baseline clinical characteristics and chest CT scans were considered for this study. The entire dataset was randomly divided into 70% training, D1train (N = 606) and 30% test-set (Dtest: D1test (N = 216) + D2 (N = 47)). An expert radiologist delineated ground-glass-opacities (GGOs) and consolidation regions on a subset of D1train, (D1train_sub, N = 88). These regions were automatically segmented and used along with their corresponding CT volumes to train an imaging AI predictor (AIP) on D1train to predict the need of mechanical ventilators for COVID-19 patients. Finally, top five prognostic clinical factors selected using univariate analysis were integrated with AIP to construct an integrated clinical and AI imaging nomogram (ClAIN). Univariate analysis identified lactate dehydrogenase, prothrombin time, aspartate aminotransferase, %lymphocytes, albumin as top five prognostic clinical features. AIP yielded an AUC of 0.81 on Dtest and was independently prognostic irrespective of other clinical parameters on multivariable analysis (p<0.001). ClAIN improved the performance over AIP yielding an AUC of 0.84 (p = 0.04) on Dtest. ClAIN outperformed AIP in predicting which COVID-19 patients ended up needing a ventilator. Our results across multiple sites suggest that ClAIN could help identify COVID-19 with severe disease more precisely and likely to end up on a life-saving mechanical ventilation.

A Novel Nodule Edge Sharpness Radiomic Biomarker Improves Performance of Lung-RADS for Distinguishing Adenocarcinomas from Granulomas on Non-Contrast CT Scans.

The aim of this study is to evaluate whether NIS radiomics can distinguish lung adenocarcinomas from granulomas on non-contrast CT scans, and also to improve the performance of Lung-RADS by reclassifying benign nodules that were initially assessed as suspicious. The screening or standard diagnostic non-contrast CT scans of 362 patients was divided into training (St, N = 145), validation (Sv, N = 145), and independent validation (Siv, N = 62) sets from different institutions. Nodules were identified and manually segmented on CT images by a radiologist. A series of 264 features relating to the edge sharpness transition from the inside to the outside of the nodule were extracted. The top 10 features were used to train a linear discriminant analysis (LDA) machine learning classifier on St. In conjunction with the LDA classifier, NIS radiomics classified nodules with an AUC of 0.82 ± 0.04, 0.77, and 0.71 respectively on St, Sv, and Siv. We evaluated the ability of the NIS classifier to determine the proportion of the patients in Sv that were identified initially as suspicious by Lung-RADS but were reclassified as benign by applying the NIS scores. The NIS classifier was able to correctly reclassify 46% of those lesions that were actually benign but deemed suspicious by Lung-RADS alone on Sv.