Artificial intelligence-based vessel suppression for detection of sub-solid nodules in lung cancer screening computed tomography

Background

Lung cancer screening (LCS) with low-dose computed tomography (LDCT) helps early lung cancer detection, commonly presenting as small pulmonary nodules. Artificial intelligence (AI)-based vessel suppression (AI-VS) and automatic detection (AI-AD) algorithm can improve detection of subsolid nodules (SSNs) on LDCT. We assessed the impact of AI-VS and AI-AD in detection and classification of SSNs [ground-glass nodules (GGNs) and part-solid nodules (PSNs)], on LDCT performed for LCS.

Methods

Following regulatory approval, 123 LDCT examinations with sub-solid pulmonary nodules (average diameter ≥6 mm) were processed to generate three image series for each examination-unprocessed, AI-VS, and AI-AD series with annotated lung nodules. Two thoracic radiologists in consensus formed the standard of reference (SOR) for this study. Two other thoracic radiologists (R1 and R2; 5 and 10 years of experience in thoracic CT image interpretation) independently assessed the unprocessed images alone, then together with AI-VS series, and finally with AI-AD for detecting all ≥6 mm GGN and PSN. We performed receiver operator characteristics (ROC) and Cohen's Kappa analyses for statistical analyses.

Results

On unprocessed images, R1 and R2 detected 232/310 nodules (R1: 114 GGN, 118 PSN) and 255/310 nodules (R2: 122 GGN, 133 PSN), respectively (P>0.05). On AI-VS images, they detected 249/310 nodules (119 GGN, 130 PSN) and 277/310 nodules (128 GGN, 149 PSN), respectively (P≥0.12). When compared to the SOR, accuracy (AUC) for detection of PSN on the AI-VS images (AUC 0.80-0.81) was greater than on the unprocessed images (AUC 0.70-0.76). AI-VS images enabled detection of solid components in five nodules deemed as GGN on the unprocessed images. Accuracy of AI-AD was lower than both the radiologists (AUC 0.60-0.72).

Conclusions

AI-VS improved the detection and classification of SSN into GGN and PSN on LDCT of the chest for the two radiologist (R1 and R2) readers.

Multicenter Assessment of CT Pneumonia Analysis Prototype for Predicting Disease Severity and Patient Outcome

To perform a multicenter assessment of the CT Pneumonia Analysis prototype for predicting disease severity and patient outcome in COVID-19 pneumonia both without and with integration of clinical information. Our IRB-approved observational study included consecutive 241 adult patients (> 18 years; 105 females; 136 males) with RT-PCR-positive COVID-19 pneumonia who underwent non-contrast chest CT at one of the two tertiary care hospitals (site A: Massachusetts General Hospital, USA; site B: Firoozgar Hospital Iran). We recorded patient age, gender, comorbid conditions, laboratory values, intensive care unit (ICU) admission, mechanical ventilation, and final outcome (recovery or death). Two thoracic radiologists reviewed all chest CTs to record type, extent of pulmonary opacities based on the percentage of lobe involved, and severity of respiratory motion artifacts. Thin-section CT images were processed with the prototype (Siemens Healthineers) to obtain quantitative features including lung volumes, volume and percentage of all-type and high-attenuation opacities (≥ -200 HU), and mean HU and standard deviation of opacities within a given lung region. These values are estimated for the total combined lung volume, and separately for each lung and each lung lobe. Multivariable analyses of variance (MANOVA) and multiple logistic regression were performed for data analyses. About 26% of chest CTs (62/241) had moderate to severe motion artifacts. There were no significant differences in the AUCs of quantitative features for predicting disease severity with and without motion artifacts (AUC 0.94-0.97) as well as for predicting patient outcome (AUC 0.7-0.77) (p > 0.5). Combination of the volume of all-attenuation opacities and the percentage of high-attenuation opacities (AUC 0.76-0.82, 95% confidence interval (CI) 0.73-0.82) had higher AUC for predicting ICU admission than the subjective severity scores (AUC 0.69-0.77, 95% CI 0.69-0.81). Despite a high frequency of motion artifacts, quantitative features of pulmonary opacities from chest CT can help differentiate patients with favorable and adverse outcomes.

Qualitative and quantitative DECT pulmonary angiography in COVID-19 pneumonia and pulmonary embolism

AIM

To assess differences in qualitative and quantitative parameters of pulmonary perfusion from dual-energy computed tomography (CT) pulmonary angiography (DECT-PA) in patients with COVID-19 pneumonia with and without pulmonary embolism (PE).

MATERIALS AND METHODS

This retrospective institutional review board-approved study included 74 patients (mean age 61±18 years, male:female 34:40) with COVID-19 pneumonia in two countries (one with 68 patients, and the other with six patients) who underwent DECT-PA on either dual-source (DS) or single-source (SS) multidetector CT machines. Images from DS-DECT-PA were processed to obtain virtual mono-energetic 40 keV (Mono40), material decomposition iodine (MDI) images and quantitative perfusion statistics (QPS). Two thoracic radiologists determined CT severity scores based on type and extent of pulmonary opacities, assessed presence of PE, and pulmonary parenchymal perfusion on MDI images. The QPS were calculated from the CT Lung Isolation prototype (Siemens). The correlated clinical outcomes included duration of hospital stay, intubation, SpO₂ and death. The significance of association was determined by receiver operating characteristics and analysis of variance.

RESULTS

One-fifth (20.2%, 15/74 patients) had pulmonary arterial filling defects; most filling defects were occlusive (28/44) located in the segmental and sub-segmental arteries. The parenchymal opacities were more extensive and denser (CT severity score 24±4) in patients with arterial filling defects than without filling defects (20±8; p=0.028). Ground-glass opacities demonstrated increased iodine distribution; mixed and consolidative opacities had reduced iodine on DS-DECT-PA but increased or heterogeneous iodine content on SS-DECT-PA. QPS were significantly lower in patients with low SpO₂ (p=0.003), intubation (p=0.006), and pulmonary arterial filling defects (p=0.007).

CONCLUSION

DECT-PA QPS correlated with clinical outcomes in COVID-19 patients.

CovidCTNet: an open-source deep learning approach to diagnose covid-19 using small cohort of CT images

Coronavirus disease 2019 (Covid-19) is highly contagious with limited treatment options. Early and accurate diagnosis of Covid-19 is crucial in reducing the spread of the disease and its accompanied mortality. Currently, detection by reverse transcriptase-polymerase chain reaction (RT-PCR) is the gold standard of outpatient and inpatient detection of Covid-19. RT-PCR is a rapid method; however, its accuracy in detection is only ~70-75%. Another approved strategy is computed tomography (CT) imaging. CT imaging has a much higher sensitivity of ~80-98%, but similar accuracy of 70%. To enhance the accuracy of CT imaging detection, we developed an open-source framework, CovidCTNet, composed of a set of deep learning algorithms that accurately differentiates Covid-19 from community-acquired pneumonia (CAP) and other lung diseases. CovidCTNet increases the accuracy of CT imaging detection to 95% compared to radiologists (70%). CovidCTNet is designed to work with heterogeneous and small sample sizes independent of the CT imaging hardware. To facilitate the detection of Covid-19 globally and assist radiologists and physicians in the screening process, we are releasing all algorithms and model parameter details as open-source. Open-source sharing of CovidCTNet enables developers to rapidly improve and optimize services while preserving user privacy and data ownership.

Deep metric learning-based image retrieval system for chest radiograph and its clinical applications in COVID-19

In recent years, deep learning-based image analysis methods have been widely applied in computer-aided detection, diagnosis and prognosis, and has shown its value during the public health crisis of the novel coronavirus disease 2019 (COVID-19) pandemic. Chest radiograph (CXR) has been playing a crucial role in COVID-19 patient triaging, diagnosing and monitoring, particularly in the United States. Considering the mixed and unspecific signals in CXR, an image retrieval model of CXR that provides both similar images and associated clinical information can be more clinically meaningful than a direct image diagnostic model. In this work we develop a novel CXR image retrieval model based on deep metric learning. Unlike traditional diagnostic models which aim at learning the direct mapping from images to labels, the proposed model aims at learning the optimized embedding space of images, where images with the same labels and similar contents are pulled together. The proposed model utilizes multi-similarity loss with hard-mining sampling strategy and attention mechanism to learn the optimized embedding space, and provides similar images, the visualizations of disease-related attention maps and useful clinical information to assist clinical decisions. The model is trained and validated on an international multi-site COVID-19 dataset collected from 3 different sources. Experimental results of COVID-19 image retrieval and diagnosis tasks show that the proposed model can serve as a robust solution for CXR analysis and patient management for COVID-19. The model is also tested on its transferability on a different clinical decision support task for COVID-19, where the pre-trained model is applied to extract image features from a new dataset without any further training. The extracted features are then combined with COVID-19 patient's vitals, lab tests and medical histories to predict the possibility of airway intubation in 72 hours, which is strongly associated with patient prognosis, and is crucial for patient care and hospital resource planning. These results demonstrate our deep metric learning based image retrieval model is highly efficient in the CXR retrieval, diagnosis and prognosis, and thus has great clinical value for the treatment and management of COVID-19 patients.

Artificial intelligence matches subjective severity assessment of pneumonia for prediction of patient outcome and need for mechanical ventilation: a cohort study

To compare the performance of artificial intelligence (AI) and Radiographic Assessment of Lung Edema (RALE) scores from frontal chest radiographs (CXRs) for predicting patient outcomes and the need for mechanical ventilation in COVID-19 pneumonia. Our IRB-approved study included 1367 serial CXRs from 405 adult patients (mean age 65 ± 16 years) from two sites in the US (Site A) and South Korea (Site B). We recorded information pertaining to patient demographics (age, gender), smoking history, comorbid conditions (such as cancer, cardiovascular and other diseases), vital signs (temperature, oxygen saturation), and available laboratory data (such as WBC count and CRP). Two thoracic radiologists performed the qualitative assessment of all CXRs based on the RALE score for assessing the severity of lung involvement. All CXRs were processed with a commercial AI algorithm to obtain the percentage of the lung affected with findings related to COVID-19 (AI score). Independent t- and chi-square tests were used in addition to multiple logistic regression with Area Under the Curve (AUC) as output for predicting disease outcome and the need for mechanical ventilation. The RALE and AI scores had a strong positive correlation in CXRs from each site (r2 = 0.79-0.86; p < 0.0001). Patients who died or received mechanical ventilation had significantly higher RALE and AI scores than those with recovery or without the need for mechanical ventilation (p < 0.001). Patients with a more substantial difference in baseline and maximum RALE scores and AI scores had a higher prevalence of death and mechanical ventilation (p < 0.001). The addition of patients' age, gender, WBC count, and peripheral oxygen saturation increased the outcome prediction from 0.87 to 0.94 (95% CI 0.90-0.97) for RALE scores and from 0.82 to 0.91 (95% CI 0.87-0.95) for the AI scores. AI algorithm is as robust a predictor of adverse patient outcome (death or need for mechanical ventilation) as subjective RALE scores in patients with COVID-19 pneumonia.

Severity and Consolidation Quantification of COVID-19 From CT Images Using Deep Learning Based on Hybrid Weak Labels

Early and accurate diagnosis of Coronavirus disease (COVID-19) is essential for patient isolation and contact tracing so that the spread of infection can be limited. Computed tomography (CT) can provide important information in COVID-19, especially for patients with moderate to severe disease as well as those with worsening cardiopulmonary status. As an automatic tool, deep learning methods can be utilized to perform semantic segmentation of affected lung regions, which is important to establish disease severity and prognosis prediction. Both the extent and type of pulmonary opacities help assess disease severity. However, manually pixel-level multi-class labelling is time-consuming, subjective, and non-quantitative. In this article, we proposed a hybrid weak label-based deep learning method that utilize both the manually annotated pulmonary opacities from COVID-19 pneumonia and the patient-level disease-type information available from the clinical report. A UNet was firstly trained with semantic labels to segment the total infected region. It was used to initialize another UNet, which was trained to segment the consolidations with patient-level information using the Expectation-Maximization (EM) algorithm. To demonstrate the performance of the proposed method, multi-institutional CT datasets from Iran, Italy, South Korea, and the United States were utilized. Results show that our proposed method can predict the infected regions as well as the consolidation regions with good correlation to human annotation.

Cause determination of missed lung nodules and impact of reader training and education: Simulation study with nodule insertion software

Background

There are "blind spots" on chest computed tomography (CT) where pulmonary nodules can easily be overlooked. The number of missed pulmonary nodules can be minimized by instituting a training program with particular focus on the depiction of nodules at blind spots.

Purpose

The purpose of this study was to assess the variation in lung nodule detection in chest CT based on location, attenuation characteristics, and reader experience.

Materials and Methods

We selected 18 noncalcified lung nodules (6-8 mm) suspicious of primary and metastatic lung cancer with solid (n = 7), pure ground-glass (6), and part-solid ground-glass (5) attenuation from 12 chest CT scans. These nodules were randomly inserted in chest CT of 34 patients in lung hila, 1^st costochondral junction, branching vessels, paramediastinal lungs, lung apices, juxta-diaphragm, and middle and outer thirds of the lungs. Two residents and two chest imaging clinical fellows evaluated the CT images twice, over a 4-month interval. Before the second reading session, the readers were trained and made aware of the potential blind spots. Chi-square test was used to assess statistical significance.

Results

Pretraining session: Fellows detected significantly more part-solid ground-glass nodules compared to residents (P = 0.008). A substantial number of nodules adjacent to branching vessels and posterior mediastinum were missed. Posttraining session: There was a significant increase in detectability independent of attenuation and location of nodules for all readers (P < 0.0008).

Conclusion

Dedicated chest CT training improves detection of lung nodules, especially the part-solid ground-glass nodules. Detection of nodules adjacent to branching vessels and the posterior mediastinal lungs is difficult even for fellowship-trained radiologists.

Variations in CT Utilization, Protocols, and Radiation Doses in COVID-19 Pneumonia: Results from 28 Countries in the IAEA Study

Background

There is lack of guidance on specific CT protocols for imaging patients with coronavirus disease 2019 (COVID-19) pneumonia.

Purpose

To assess international variations in CT utilization, protocols, and radiation doses in patients with COVID-19 pneumonia.

Materials and Methods

In this retrospective data collection study, the International Atomic Energy Agency (IAEA) coordinated a survey between May and July 2020 regarding CT utilization, protocols, and radiation doses from 62 healthcare sites in 34 countries across five continents for CT exams performed in COVID-19 pneumonia. The questionnaire obtained information on local prevalence, method of diagnosis, most frequent imaging, indications for CT, and specific policies on use of CT in COVID-19 pneumonia. Collected data included general information (patient age, weight, clinical indication), CT equipment (CT make and model, year of installation, number of detector rows), scan protocols (body region, scan phases, tube current and potential), and radiation dose descriptors (CT dose index (CTDI_vol) and dose length product (DLP)). Descriptive statistics and generalized estimating equations were performed.

Results

Data from 782 patients (median age (interquartile range) of 59(15) years) from 54 healthcare sites in 28 countries were evaluated. Less than one-half of the healthcare sites used CT for initial diagnosis of COVID-19 pneumonia and three-fourth used CT for assessing disease severity. CTDI_vol varied based on CT vendors (7-11mGy, p<0.001), number of detector-rows (8-9mGy, p<0.001), year of CT installation (7-10mGy, p=0.006), and reconstruction techniques (7-10mGy, p=0.03). Multiphase chest CT exams performed in 20% of sites (11 of 54) were associated with higher DLP compared with single-phase chest CT exams performed in 80% (43 of 54 sites) (p=0.008).

Conclusion

CT use, scan protocols, and radiation doses in patients with COVID-19 pneumonia showed wide variation across healthcare sites within the same and different countries. Many patients were scanned multiple times and/or with multiphase CT scan protocols.

See also the editorial by Lee.

Quantifying and leveraging predictive uncertainty for medical image assessment

The interpretation of medical images is a challenging task, often complicated by the presence of artifacts, occlusions, limited contrast and more. Most notable is the case of chest radiography, where there is a high inter-rater variability in the detection and classification of abnormalities. This is largely due to inconclusive evidence in the data or subjective definitions of disease appearance. An additional example is the classification of anatomical views based on 2D Ultrasound images. Often, the anatomical context captured in a frame is not sufficient to recognize the underlying anatomy. Current machine learning solutions for these problems are typically limited to providing probabilistic predictions, relying on the capacity of underlying models to adapt to limited information and the high degree of label noise. In practice, however, this leads to overconfident systems with poor generalization on unseen data. To account for this, we propose a system that learns not only the probabilistic estimate for classification, but also an explicit uncertainty measure which captures the confidence of the system in the predicted output. We argue that this approach is essential to account for the inherent ambiguity characteristic of medical images from different radiologic exams including computed radiography, ultrasonography and magnetic resonance imaging. In our experiments we demonstrate that sample rejection based on the predicted uncertainty can significantly improve the ROC-AUC for various tasks, e.g., by 8% to 0.91 with an expected rejection rate of under 25% for the classification of different abnormalities in chest radiographs. In addition, we show that using uncertainty-driven bootstrapping to filter the training data, one can achieve a significant increase in robustness and accuracy. Finally, we present a multi-reader study showing that the predictive uncertainty is indicative of reader errors.

Integrative analysis for COVID-19 patient outcome prediction

While image analysis of chest computed tomography (CT) for COVID-19 diagnosis has been intensively studied, little work has been performed for image-based patient outcome prediction. Management of high-risk patients with early intervention is a key to lower the fatality rate of COVID-19 pneumonia, as a majority of patients recover naturally. Therefore, an accurate prediction of disease progression with baseline imaging at the time of the initial presentation can help in patient management. In lieu of only size and volume information of pulmonary abnormalities and features through deep learning based image segmentation, here we combine radiomics of lung opacities and non-imaging features from demographic data, vital signs, and laboratory findings to predict need for intensive care unit (ICU) admission. To our knowledge, this is the first study that uses holistic information of a patient including both imaging and non-imaging data for outcome prediction. The proposed methods were thoroughly evaluated on datasets separately collected from three hospitals, one in the United States, one in Iran, and another in Italy, with a total 295 patients with reverse transcription polymerase chain reaction (RT-PCR) assay positive COVID-19 pneumonia. Our experimental results demonstrate that adding non-imaging features can significantly improve the performance of prediction to achieve AUC up to 0.884 and sensitivity as high as 96.1%, which can be valuable to provide clinical decision support in managing COVID-19 patients. Our methods may also be applied to other lung diseases including but not limited to community acquired pneumonia. The source code of our work is available at https://github.com/DIAL-RPI/COVID19-ICUPrediction.

Optimization of paranasal sinus CT procedure: Ultra-low dose CT as a roadmap for pre-functional endoscopic sinus surgery

OBJECTIVE

To assess image quality and radiation dose associated with ultra-low dose CT protocol for patients with benign paranasal sinus diseases undergoing functional endoscopic surgery (FESS).

METHODS

We scanned the head portion of Alderson RANDO phantom on a second generation, dual-source, multidetector-row CT scanner (Siemens Definition Flash) using standard-dose and five low-dose protocols. Two radiologists assessed the image quality for each protocol to determine best ultra-low-dose protocols for imaging patients with benign paranasal sinus diseases undergoing FESS. The ultra-low-dose CT protocols were then used for scanning. Thereafter, 40 adult patients (age range 18-54 years, M:F 23:17) were scanned with the four low dose scanning protocols (10 patients per protocol). On both transverse and coronal reformatted CT images, two radiologists assessed visibility of key anatomic landmarks for FESS on a 2-point scale (1 = clear and complete visualization; 2 = suboptimal visualization). Data were analyzed with descriptive statistics and Cohen's kappa coefficient for interobserver agreement.

RESULTS

In phantom study, the lowest dose scan protocol (CTDI_vol 2.1 mGy, 70 kV, 75 mAs) was unacceptable due to poor image quality. For patient studies, both radiologists gave acceptable image quality scores for ultra-low-dose scan protocol with axial scan mode, automatic tube potential selection and tube current modulation (CTDI_vol 2.2 mGy; DLP 22.9 mGy.cm) with up to 60% lower dose compared to prior standard-dose CT (CTDI_vol 5.3 mGy; DLP 73.5 mGy.cm).

CONCLUSIONS

Ultra-low-dose CT protocol provides sufficient image quality for scanning patients undergoing functional endoscopic surgery for benign paranasal sinus diseases.

Low contrast volume dual-energy CT of the chest: Quantitative and qualitative assessment

PURPOSE

To evaluate the image quality of chest CT performed on dual-energy scanners using low contrast volume for routine chest (DECT-R) and pulmonary angiography (DECTPA) protocols.

MATERIALS AND METHODS

This retrospective study included dual-energy CT scans of chest performed with low contrast volume in 84 adults (34M:50F; Age 69 ± 16 years: Weight 71 ± 16kg). There were 42 patients with DECT-R and 42 patients with DECT-PA protocols. Images were reviewed by two thoracic radiologists. Qualitative assessment was done on a four-point scale, for subjective assessment of contrast enhancement and artifacts (1 = Excellent, 2 = optimal, 3 = suboptimal, and 4 = Limited) in the pulmonary arteries and thoracic aorta, on virtual monoenergetic and material decomposition iodine (MDI) images. Quantitative assessment was performed by measuring the CT (Hounsfield) units in aorta and pulmonary arteries. The estimated glomerular filtration rate (eGFR) was calculated before and after CT scans. Two tailed student's t-test was performed to assess the significance of findings, and strength of correlation between readers was determined by Cohen's kappa test.

RESULTS

DECT-PA and DECT-R demonstrated excellent/adequate contrast density within the pulmonary arteries (up to segmental branch), and aorta. There was no suboptimal or limited examination. There was strong interobserver agreement for arterial enhancement in pulmonary arteries (kappa = 0.62-0.89) and for thoracic aorta (kappa = 0.62-0.94). Pulmonary emboli were seen in 3/42(7%) in DECT-R and in 5/42(12%) in DECT-PA. There was no significant change in eGFR before and after IV contrast injection (p = 0.46-0.52).

CONCLUSION

DECT-R and DECT-PA performed with low contrast volume provide diagnostic quality opacification of the pulmonary vessels and aorta vessels.

Clinical and imaging features predict mortality in COVID-19 infection in Iran

The new coronavirus disease 2019 (COVID-19) pandemic has challenged many healthcare systems around the world. While most of the current understanding of the clinical features of COVID-19 is derived from Chinese studies, there is a relative paucity of reports from the remaining global health community. In this study, we analyze the clinical and radiologic factors that correlate with mortality odds in COVID-19 positive patients from a tertiary care center in Tehran, Iran. A retrospective cohort study of 90 patients with reverse transcriptase-polymerase chain reaction (RT-PCR) positive COVID-19 infection was conducted, analyzing demographics, co-morbidities, presenting symptoms, vital signs, laboratory values, chest radiograph findings, and chest CT features based on mortality. Chest radiograph was assessed using the Radiographic Assessment of Lung Edema (RALE) scoring system. Chest CTs were assessed according to the opacification pattern, distribution, and standardized severity score. Initial and follow-up Chest CTs were compared if available. Multiple logistic regression was used to generate a prediction model for mortality. The 90 patients included 59 men and 31 women (59.4 ± 16.6 years), including 21 deceased and 69 surviving patients. Among clinical features, advanced age (p = 0.02), low oxygenation saturation (p<0.001), leukocytosis (p = 0.02), low lymphocyte fraction (p = 0.03), and low platelet count (p = 0.048) were associated with increased mortality. High RALE score on initial chest radiograph (p = 0.002), presence of pleural effusions on initial CT chest (p = 0.005), development of pleural effusions on follow-up CT chest (p = 0.04), and worsening lung severity score on follow-up CT Chest (p = 0.03) were associated with mortality. A two-factor logistic model using patient age and oxygen saturation was created, which demonstrates 89% accuracy and area under the ROC curve of 0.86 (p<0.0001). Specific demographic, clinical, and imaging features are associated with increased mortality in COVID-19 infections. Attention to these features can help optimize patient management.

Integrative Analysis for COVID-19 Patient Outcome Prediction

While image analysis of chest computed tomography (CT) for COVID-19 diagnosis has been intensively studied, little work has been performed for image-based patient outcome prediction. Management of high-risk patients with early intervention is a key to lower the fatality rate of COVID-19 pneumonia, as a majority of patients recover naturally. Therefore, an accurate prediction of disease progression with baseline imaging at the time of the initial presentation can help in patient management. In lieu of only size and volume information of pulmonary abnormalities and features through deep learning based image segmentation, here we combine radiomics of lung opacities and non-imaging features from demographic data, vital signs, and laboratory findings to predict need for intensive care unit (ICU) admission. To our knowledge, this is the first study that uses holistic information of a patient including both imaging and non-imaging data for outcome prediction. The proposed methods were thoroughly evaluated on datasets separately collected from three hospitals, one in the United States, one in Iran, and another in Italy, with a total 295 patients with reverse transcription polymerase chain reaction (RT-PCR) assay positive COVID-19 pneumonia. Our experimental results demonstrate that adding non-imaging features can significantly improve the performance of prediction to achieve AUC up to 0.884 and sensitivity as high as 96.1%, which can be valuable to provide clinical decision support in managing COVID-19 patients. Our methods may also be applied to other lung diseases including but not limited to community acquired pneumonia. The source code of our work is available at https://github.com/DIAL-RPI/COVID19-ICUPrediction.

Validation of a Deep Learning Algorithm for the Detection of Malignant Pulmonary Nodules in Chest Radiographs

IMPORTANCE

The improvement of pulmonary nodule detection, which is a challenging task when using chest radiographs, may help to elevate the role of chest radiographs for the diagnosis of lung cancer.

OBJECTIVE

To assess the performance of a deep learning-based nodule detection algorithm for the detection of lung cancer on chest radiographs from participants in the National Lung Screening Trial (NLST).

DESIGN, SETTING, AND PARTICIPANTS

This diagnostic study used data from participants in the NLST ro assess the performance of a deep learning-based artificial intelligence (AI) algorithm for the detection of pulmonary nodules and lung cancer on chest radiographs using separate training (in-house) and validation (NLST) data sets. Baseline (T0) posteroanterior chest radiographs from 5485 participants (full T0 data set) were used to assess lung cancer detection performance, and a subset of 577 of these images (nodule data set) were used to assess nodule detection performance. Participants aged 55 to 74 years who currently or formerly (ie, quit within the past 15 years) smoked cigarettes for 30 pack-years or more were enrolled in the NLST at 23 US centers between August 2002 and April 2004. Information on lung cancer diagnoses was collected through December 31, 2009. Analyses were performed between August 20, 2019, and February 14, 2020.

EXPOSURES

Abnormality scores produced by the AI algorithm.

MAIN OUTCOMES AND MEASURES

The performance of an AI algorithm for the detection of lung nodules and lung cancer on radiographs, with lung cancer incidence and mortality as primary end points.

RESULTS

A total of 5485 participants (mean [SD] age, 61.7 [5.0] years; 3030 men [55.2%]) were included, with a median follow-up duration of 6.5 years (interquartile range, 6.1-6.9 years). For the nodule data set, the sensitivity and specificity of the AI algorithm for the detection of pulmonary nodules were 86.2% (95% CI, 77.8%-94.6%) and 85.0% (95% CI, 81.9%-88.1%), respectively. For the detection of all cancers, the sensitivity was 75.0% (95% CI, 62.8%-87.2%), the specificity was 83.3% (95% CI, 82.3%-84.3%), the positive predictive value was 3.8% (95% CI, 2.6%-5.0%), and the negative predictive value was 99.8% (95% CI, 99.6%-99.9%). For the detection of malignant pulmonary nodules in all images of the full T0 data set, the sensitivity was 94.1% (95% CI, 86.2%-100.0%), the specificity was 83.3% (95% CI, 82.3%-84.3%), the positive predictive value was 3.4% (95% CI, 2.2%-4.5%), and the negative predictive value was 100.0% (95% CI, 99.9%-100.0%). In digital radiographs of the nodule data set, the AI algorithm had higher sensitivity (96.0% [95% CI, 88.3%-100.0%] vs 88.0% [95% CI, 75.3%-100.0%]; P = .32) and higher specificity (93.2% [95% CI, 89.9%-96.5%] vs 82.8% [95% CI, 77.8%-87.8%]; P = .001) for nodule detection compared with the NLST radiologists. For malignant pulmonary nodule detection on digital radiographs of the full T0 data set, the sensitivity of the AI algorithm was higher (100.0% [95% CI, 100.0%-100.0%] vs 94.1% [95% CI, 82.9%-100.0%]; P = .32) compared with the NLST radiologists, and the specificity (90.9% [95% CI, 89.6%-92.1%] vs 91.0% [95% CI, 89.7%-92.2%]; P = .91), positive predictive value (8.2% [95% CI, 4.4%-11.9%] vs 7.8% [95% CI, 4.1%-11.5%]; P = .65), and negative predictive value (100.0% [95% CI, 100.0%-100.0%] vs 99.9% [95% CI, 99.8%-100.0%]; P = .32) were similar to those of NLST radiologists.

CONCLUSIONS AND RELEVANCE

In this study, the AI algorithm performed better than NLST radiologists for the detection of pulmonary nodules on digital radiographs. When used as a second reader, the AI algorithm may help to detect lung cancer.

Radiomic features of primary tumor by lung cancer stage: analysis in BRAF mutated non-small cell lung cancer

Background

The clinical features and traditional semantic imaging characteristics of BRAF-mutated non-small cell lung cancer (NSCLC) have been previously reported. The radiomic features of BRAF-mutated NSCLC and their role in predicting cancer stage, however, have yet to be investigated. This study’s goal is to assess the differences in CT radiomic features of primary NSCLC driven by BRAF mutation and stratified by tumor-node-metastasis (TNM) staging.

Methods

Our IRB approved study included 62 patients with BRAF mutations (V600 in 27 and non-V600 in 35 patients), who underwent contrast-enhanced chest CT. Tumor stage was determined based on the 8^th edition of TNM staging. Two thoracic radiologists assessed the primary tumor imaging features such, including tumor size (maximum and minimum dimensions) and density (Hounsfield units, HU). De-identified transverse CT images (DICOM) were processed with 3D slicer (Version 4.7) for manual lesion segmentation and estimation of radiomic features. Descriptive statistics, multivariate logistic regression, and receiver operating characteristics (ROC) were performed.

Results

There were significant differences in the radiomic features based on cancer stages I-IV with the most significant differences between stage IV and stage I lesions [AUC 0.94 (95% CI: 0.86–0.99), P<0.04]. There were also significant differences in radiomic features between stage IV and combined stages I-III [40/113 radiomic features; AUC 0.71 (95% CI: 0.59–0.85); P<0.04–0.0001]. None of the clinical (0/6) or imaging (0/3) features were significantly different between stage IV and combined stages I–III.

Conclusions

The radiomic features of primary tumor in BRAF driven NSCLC significantly vary with cancer stage, independent of standard imaging and clinical features.

CT Radiomics, Radiologists, and Clinical Information in Predicting Outcome of Patients with COVID-19 Pneumonia

Purpose

To compare prediction of disease outcome, severity, and patient triage in coronavirus disease 2019 (COVID-19) pneumonia with whole lung radiomics, radiologists' interpretation, and clinical variables.

Materials and Methods

This institutional review board-approved retrospective study included 315 adult patients (mean age, 56 years [range, 21-100 years], 190 men, 125 women) with COVID-19 pneumonia who underwent noncontrast chest CT. All patients (inpatients, n = 210; outpatients, n = 105) were followed-up for at least 2 weeks to record disease outcome. Clinical variables, such as presenting symptoms, laboratory data, peripheral oxygen saturation, and comorbid diseases, were recorded. Two radiologists assessed each CT in consensus and graded the extent of pulmonary involvement (by percentage of involved lobe) and type of opacities within each lobe. Radiomics were obtained for the entire lung, and multiple logistic regression analyses with areas under the curve (AUCs) as outputs were performed.

Results

Most patients (276/315, 88%) recovered from COVID-19 pneumonia; 36/315 patients (11%) died, and 3/315 patients (1%) remained admitted in the hospital. Radiomics differentiated chest CT in outpatient versus inpatient with an AUC of 0.84 (P < .005), while radiologists' interpretations of disease extent and opacity type had an AUC of 0.69 (P < .0001). Whole lung radiomics were superior to the radiologists' interpretation for predicting patient outcome in terms of intensive care unit (ICU) admission (AUC: 0.75 vs 0.68) and death (AUC: 0.81 vs 0.68) (P < .002). The addition of clinical variables to radiomics improved the AUC to 0.84 for predicting ICU admission.

Conclusion

Radiomics from noncontrast chest CT were superior to radiologists' assessment of extent and type of pulmonary opacities in predicting COVID-19 pneumonia outcome, disease severity, and patient triage.© RSNA, 2020.

Relation between age and CT radiation doses: Dose trends in 705 pediatric head CT

PURPOSE

To evaluate the relationship between patient age and radiation doses associated with routine pediatric head CT performed with automatic tube potential selection and tube current modulation techniques.

METHODS

We obtained patient demographics, scan parameters, and radiation dose descriptors (CT dose index volume -CTDIvol and dose length product -DLP) associated with consecutive routine head CT in 705 children (mean age 6.9 ± 5 years). Children were scanned on one of the three multidetector-row CTs (64-128 slices, Siemens) over 6 months period in a tertiary hospital. All head CT exams were performed in helical scan mode using automatic tube potential selection (Care kV) and automatic tube current modulation (Care Dose 4D) techniques. The information was obtained from a radiation dose monitoring software. Data were analyzed using linear correlation and analysis of variance.

RESULTS

Most age-wise median CTDIvol (9-27 mGy; 703/705 pediatric head CT, >99 %) from our institution were lower than the European Diagnostic Reference Levels (EDRL, CTDIvol 24-50 mGy) but median DLP (151-586 mGy cm) from 201/705 children (28 %) was higher than the EDRL (DLP 300-650 mGy cm). Unlike the age-stratified EDRL, a combination of automatic tube potential selection and tube current modulation for pediatric head results in a significant linear correlation between radiation doses and patient age (r2 = 0.66, p < 0.001).

CONCLUSIONS

Radiation doses for head CT change linearly with children's age. Despite lower CTDIvol and DLP for most children, longer scan length resulted in higher DLP for some pediatric head CT compared to the corresponding EDRL; this result underscores the need to promote clear guidelines for technologists operating CT.

Chest CT practice and protocols for COVID-19 from radiation dose management perspective

The global pandemic of coronavirus disease 2019 (COVID-19) has upended the world with over 6.6 million infections and over 391,000 deaths worldwide. Reverse-transcription polymerase chain reaction (RT-PCR) assay is the preferred method of diagnosis of COVID-19 infection. Yet, chest CT is often used in patients with known or suspected COVID-19 due to regional preferences, lack of availability of PCR assays, and false-negative PCR assays, as well as for monitoring of disease progression, complications, and treatment response. The International Atomic Energy Agency (IAEA) organized a webinar to discuss CT practice and protocol optimization from a radiation protection perspective on April 9, 2020, and surveyed participants from five continents. We review important aspects of CT in COVID-19 infection from the justification of its use to specific scan protocols for optimizing radiation dose and diagnostic information.

Key Points

• Chest CT provides useful information in patients with moderate to severe COVID-19 pneumonia.

• When indicated, chest CT in most patients with COVID-19 pneumonia must be performed with non-contrast, low-dose protocol.

• Although chest CT has high sensitivity for diagnosis of COVID-19 pneumonia, CT findings are non-specific and overlap with other viral infections including influenza and H1N1.

Quadratic Autoencoder (Q-AE) for Low-Dose CT Denoising

Inspired by complexity and diversity of biological neurons, our group proposed quadratic neurons by replacing the inner product in current artificial neurons with a quadratic operation on input data, thereby enhancing the capability of an individual neuron. Along this direction, we are motivated to evaluate the power of quadratic neurons in popular network architectures, simulating human-like learning in the form of "quadratic-neuron-based deep learning". Our prior theoretical studies have shown important merits of quadratic neurons and networks in representation, efficiency, and interpretability. In this paper, we use quadratic neurons to construct an encoder-decoder structure, referred as the quadratic autoencoder, and apply it to low-dose CT denoising. The experimental results on the Mayo low-dose CT dataset demonstrate the utility and robustness of quadratic autoencoder in terms of image denoising and model efficiency. To our best knowledge, this is the first time that the deep learning approach is implemented with a new type of neurons and demonstrates a significant potential in the medical imaging field.

Multiplatform, Non-Breath-Hold Fast Scanning Protocols: Should We Stop Giving Breath-Hold Instructions for Routine Chest CT? [Formula: see text]

OBJECTIVE

We assessed if non-breath-hold (NBH) fast scanning protocol can provide respiratory motion-free images for interpretation of chest computed tomography (CT).

MATERIALS AND METHODS

In our 2-phase project, we first collected baseline data on frequency of respiratory motion artifacts on breath-hold chest CT in 826 adult patients. The second phase included 62 patients (mean age 66 ± 15 years; 21 females, 41 males) who underwent an NBH chest CT on either single-source (n = 32) or dual-source (n = 30) multidetector-row CT scanners. Clinical indications for chest CT, reason for using NBH CT, scanner type, scan duration, and radiation dose (CT dose index volume, dose length product) were recorded. Two thoracic radiologists (R1 and R2) independently graded respiratory motion artifacts (1 = no respiratory motion artifacts with unrestricted evaluation; 2 = minor motion artifacts limited to one lung lobe or less with good diagnostic quality; 3 = moderate motion artifacts limited to 2 to 3 lung lobes but adequate for clinical diagnosis; 4 = poor evaluability or unevaluable from severe motion artifacts; and 5 = limited quality due to other causes like high noise, beam hardening, or metallic artifacts), and recorded pulmonary and mediastinal findings. Descriptive analyses, Cohen κ test for interobserver agreement, and Student t test were performed for statistical analysis.

RESULTS

No NBH chest CT were deemed uninterpretable by either radiologist; most NBH CT (R1-59 of 62, 95%; R2-62 of 62, 100%) had no or minimal motion artifacts. Only 3 of 62 (R1) NBH chest CT had motion artifacts limiting diagnostic evaluation for lungs but not in the mediastinum.

CONCLUSION

Non-breath-hold fast protocol enables acquisition of diagnostic quality chest CT free of respiratory motion artifacts in patients who cannot hold their breath.