Artificial intelligence-based lung nodule detection for pulmonary arteriovenous fistulas on chest computed tomography

Clinical application of the COVID-19 Reporting and Data System (CO-RADS) in patients with suspected SARS-CoV-2 infection: observational study in an emergency department

Although traditional teachings in regard to pneumothorax and hemothorax generally recommend chest tube placement and hospital admission, the increasing use of chest computed tomography (CT) in blunt trauma evaluation may detect more minor pneumothorax and hemothorax that might indicate a need to modify these traditional practices. We determine the incidence of pneumothorax and hemothorax observed on CT only and the incidence of isolated pneumothorax and hemothorax (pneumothorax and hemothorax occurring without other thoracic injuries), and describe the clinical implications of these injuries. This was a planned secondary analysis of 2 prospective, observational studies of adult patients with blunt trauma, NEXUS Chest (January 2009 to December 2012) and NEXUS Chest CT (August 2011 to May 2014), set in 10 Level I US trauma centers. Participants' inclusion criteria were older than 14 years, presentation to the emergency department (ED) within 6 hours of blunt trauma, and receipt of chest imaging (chest radiograph, chest CT, or both) during their ED evaluation. Exposure(s) (for observational studies) were that patients had trauma and chest imaging. Primary measures and outcomes included the incidence of pneumothorax and hemothorax observed on CT only versus on both chest radiograph and chest CT, the incidence of isolated pneumothorax and hemothorax (pneumothorax and hemothorax occurring without other thoracic injuries), and admission rates, hospital length of stay, mortality, and frequency of chest tube placement for these injuries. Of 21,382 enrolled subjects, 1,064 (5%) had a pneumothorax and 384 (1.8%) had a hemothorax. Of the 8,661 patients who received both a chest radiograph and a chest CT, 910 (10.5%) had a pneumothorax, with 609 (67%) observed on CT only; 319 (3.7%) had a hemothorax, with 254 (80%) observed on CT only. Of 1,117 patients with pneumothorax, hemothorax, or both, 108 (10%) had isolated pneumothorax or hemothorax. Patients with pneumothorax observed on CT only had a lower chest tube placement rate (30% versus 65%; difference in proportions [Δ] -35%; 95% confidence interval [CI] -28% to 42%), admission rate (94% versus 99%; Δ 5%; 95% CI 3% to 8%), and median length of stay (5 versus 6 days; difference 1 day; 95% CI 0 to 2 days) but similar mortality compared with patients with pneumothorax observed on chest radiograph and CT. Patients with hemothorax observed on CT had only a lower chest tube placement rate (49% versus 68%; Δ -19%; 95% CI -31% to -5%) but similar admission rate, mortality, and median length of stay compared with patients with hemothorax observed on chest radiograph and CT. Compared with patients with other thoracic injury, those with isolated pneumothorax or hemothorax had a lower chest tube placement rate (20% versus 43%; Δ -22%; 95% CI -30% to -13%), median length of stay (4 versus 5 days; difference -1 day; 95% CI -3 to 1 days), and admission rate (44% versus 97%; Δ -53%; 95% CI -62% to -43%), with an admission rate comparable to that of patients without pneumothorax or hemothorax (49%). Under current imaging protocols for adult blunt trauma evaluation, most pneumothoraces and hemothoraces are observed on CT only and few occur as isolated thoracic injury. The clinical implications (admission rates and frequency of chest tube placement) of pneumothorax and hemothorax observed on CT only and isolated pneumothorax or hemothorax are lower than those of patients with pneumothorax and hemothorax observed on chest radiograph and CT and of those who have other thoracic injury, respectively.

Computer-Aided Nodule Detection System: Results in an Unselected Series of Consecutive Chest Radiographs

BackgroundLung cancer is the leading cause of cancer-related deaths throughout the world. Chest computed tomography (CT) is now widely used in the screening and diagnosis of lung cancer due to its effectiveness. Radiologists must identify each small nodule shadow from 3D volume images, which is very burdensome and often results in missed nodules. To address these challenges, we developed a computer-aided detection (CAD) system that automatically detects lung nodules in CT images.MethodsA total of 1997 chest CT scans were collected for algorithm development. The algorithm was designed using deep learning technology. In addition to evaluating detection performance on various public datasets, its robustness to changes in radiation dose was assessed by a phantom study. To investigate the clinical usefulness of the CAD system, a reader study was conducted with 10 doctors, including inexperienced and expert readers. This study investigated whether the use of the CAD as a second reader could prevent nodular lesions in lungs that require follow-up examinations from being overlooked. Analysis was performed using the Jackknife Free-Response Receiver-Operating Characteristic (JAFROC).ResultsThe CAD system achieved sensitivity of 0.98/0.96 at 3.1/7.25 false positives per case on two public datasets. Sensitivity did not change within the range of practical doses for a study using a phantom. A second reader study showed that the use of this system significantly improved the detection ability of nodules that could be picked up clinically (p = 0.026).ConclusionsWe developed a deep learning-based CAD system that is robust to imaging conditions. Using this system as a second reader increased detection performance.

The aim of this study was to investigate the reliability and accuracy of automatic coronary artery calcium (CAC) scoring and risk classification in non-gated, non-contrast chest computed tomography (CT) of different slice thicknesses using a deep learning algorithm. This retrospective study was performed at 2 tertiary hospitals. Paired, dedicated calcium-scoring CT scans and non-gated, non-contrast chest CT scans taken within a month from the same patients were included. Chest CT images were grouped according to the slice thickness (group A: 1 mm; group B: 3 mm). For internal scans, the CAC score manually measured on dedicated calcium scoring CT images was used as the gold standard. The deep learning algorithm for group A was trained using 150 chest CT scans and tested using 144 scans, and that for group B was trained using 170 chest CT scans and tested using 144 scans. The intraclass correlation coefficient (ICC) was used to evaluate the correlation between the algorithm and the gold standard. Agreement between the deep learning algorithm, the manual results on chest CT, and the gold standard was determined by Bland-Altman analysis. Cardiac risk categories were compared. External validation was performed on 334 paired scans from a different organization. A total of 608 internal paired scans (1 mm: 294; 3 mm: 314) of 406 individuals and 334 external paired scans (1 mm: 117; 3 mm: 117) of 117 individuals were included in the analysis. The ICCs between the deep learning algorithm and the gold standard were excellent in both group A (0.90; 95% CI: 0.85-0.93) and group B (0.94; 95% CI: 0.92-0.96). The Bland-Altman plots showed good agreement in both groups. For the cardiovascular risk category, the deep learning algorithm accurately classified 71% of cases in group A and 81% of cases in group B. The Kappa values for risk classification were 0.72 in group A and 0.82 in group B. External validation yielded equally good results. The automatic calculation of CAC score and cardiovascular risk stratification on non-gated chest CT using a deep learning algorithm was reliable and accurate on both 1 and 3 mm scans. Chest CT with a slice thickness of 3 mm was slightly more accurate in CAC detection and risk classification.

Screening Chest CT Prior to Allogenic Transplantation - High Rates of Occult Abnormalities

BackgroundCharacteristic chest computed tomography (CT) manifestation of 2019 novel coronavirus (COVID-19) was added as a diagnostic criterion in the Chinese National COVID-19 management guideline. Whether the characteristic findings of Chest CT could differentiate confirmed COVID-19 cases from other positive nucleic acid test (NAT)-negative patients has not been rigorously evaluated.PurposeWe aim to test whether chest CT manifestation of 2019 novel coronavirus (COVID-19) can be differentiated by a radiologist or a computer-based CT image analysis system.MethodsWe conducted a retrospective case-control study that included 52 laboratory-confirmed COVID-19 patients and 80 non-COVID-19 viral pneumonia patients between 20 December, 2019 and 10 February, 2020. The chest CT images were evaluated by radiologists in a double blind fashion. A computer-based image analysis system (uAI System, Lianying Inc., Shanghai, China) detected the lesions in 18 lung segments defined by Boyden classification system and calculated the infected volume in each segment. The number and volume of lesions detected by radiologist and computer system was compared with Chi-square test or Mann-Whitney U test as appropriate.ResultsThe main CT manifestations of COVID-19 were multi-lobar/segmental peripheral ground-glass opacities and patchy air space infiltrates. The case and control groups were similar in demographics, comorbidity, and clinical manifestations. There was no significant difference in eight radiologist identified CT image features between the two groups of patients. There was also no difference in the absolute and relative volume of infected regions in each lung segment.ConclusionWe documented the non-differentiating nature of initial chest CT image between COVID-19 and other viral pneumonia with suspected symptoms. Our results do not support CT findings replacing microbiological diagnosis as a critical criterion for COVID-19 diagnosis. Our findings may prompt re-evaluation of isolated patients without laboratory confirmation.

The suppression of lung vessels in chest computed tomography (CT) scans can enhance the conspicuity of lung nodules, thereby may improve the detection rate of early lung cancer. This study aimed to verify the effect of lung vessel suppression on the performance of the lung nodule detector. Firstly, a lung vessel suppression technique was developed to remove the vessels while preserving the nodules. Then, a lung nodule detector was developed with two stages: nodule candidate generation and false positive reduction. The vessel suppression and nodule detection methods were validated respectively in 50 three-dimensional (3D) chest CT images with manually-labeled vessel trees and 888 3D chest CT images with manually-located nodules (LUNA16). The lung vessel suppression results were quantitatively evaluated by using the Dice coefficient (DICE) and the contrast-to-noise ratio (CNR), and the lung nodule detection results were quantitatively evaluated by using the sensitivity under two conditions: “without” and “with vessel suppression”. The lung vessel suppression accurately removed vessels with a DICE of 0.943 and improved the CNR for nodules from 4.24 (6.27 dB) to 7.02 (8.46 dB), which subsequently improved the average sensitivity from 0.948 to 0.969 under 7 specified false positives for lung nodule detection.

Chronic airway diseases, including chronic obstructive pulmonary disease (COPD), asthma, and bronchiectasis, impose substantial morbidity and mortality worldwide. Precise phenotyping of their complex pathophysiological manifestations is essential for effective management. Chest computed tomography (CT) allows qualitative and quantitative assessments of emphysema, airway structure, mucus plugs, vascular abnormalities, bronchiectasis, and comorbid interstitial lung abnormality, making it central to charactering these conditions. From perspectives of pulmonologists who use chest CT with guidance from radiologists, this review describes the advances in CT image analysis and their implications for patients with chronic airway disease. On inspiratory CT, low-attenuation regions reflect emphysema and are associated with clinical outcomes in smokers. Airway lumen, wall size, branch count, and fractal dimension correlate with disease severity and lung function impairment in COPD and asthma. Airway mucus plugs reflect inflammatory patterns and are associated with reduced lung function and exacerbations; however, mucus plugs are increasingly recognized as a treatable trait in the era of biologics treatment. Pulmonary vascular abnormalities are quantified using pulmonary artery-to-aorta diameter ratio and small vessel volume proportions. Along with chronic symptoms, bronchiectasis is diagnosed radiologically by an increased broncho-arterial ratio, absent bronchial tapering, and peripheral airway visibility. Although limited spatial resolution precludes direct evaluation of small airway disease, air-trapping on expiratory CT or registered Inspiratory-expiratory CT allows indirect estimation of small airway disease. Despite these advances, many research findings remain unapplied to routine clinical image analysis. Radiation exposure is an inherent limitation. Nonetheless, chest CT provides greater diagnostic information than chest radiography and is more accessible than magnetic resonance imaging and nuclear imaging. Further studies are needed to maximize the potential of chest CT for early detection, risk stratification, and treatment monitoring in airway disease management.

Coronavirus disease (COVID-19) is an infectious disease that can lead to pneumonia, pulmonary oedema, acute respiratory distress syndrome, multiple organ and system dysfunction, and death. This study aimed to verify the efficacy of chest computed tomography (CT) for the initial diagnosis of COVID-19. This observational, retrospective, cross-sectional study included 259 individuals who underwent clinical evaluation, blood collection, chest CT, and a reverse transcription polymerase chain reaction (RT-PCR) diagnostic test for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) during their course of treatment at a reference hospital in Belém, Pará, Brazil between April and June 2020. Inclusion criteria were flu-like symptoms in adults of both sexes. Individuals with an inconclusive COVID-19 molecular test or who had artifacts in the chest CT images were excluded. Parametric data were analyzed using Student-t-test and non-parametric data were analyzed using average test and Fisher exact test. Participants were divided into two groups: Group 1 (COVID-19 positive), n = 211 (124 males, 87 females), 51.8 ± 17.9 years old and Group 2 (COVID-19 negative), n = 48 (22 males, 26 females), 47.6 ± 18.6 years old. Most frequent symptoms were cough [Group 1 n = 199 (94%)/Group 2 n = 46 (95%)], fever [Group 1 n = 154 (72%)/Group 2 n = 28 (58%)], myalgia [Group 1 n = 172 (81%)/Group 2 n = 38 (79%)], dyspnoea [Group 1 n = 169 (80%) / Group 2 n = 37 (77%)], headache [Group 1 n = 163 (77%)/Group 2 n = 32 (66%)], and anosmia [Group 1 n = 154 (73%)/Group 2 n = 29 (60%)]. Group 1 had a higher proportion of ground-glass opacity [Group 1 n = 175 (83%)/Group 2 n = 24 (50%), 0.00], vascular enhancement sign [Group 1 n = 128 (60%)/Group 2 n = 15 (31%), 0.00], septal thickening [Group 1 n = 99 (47%)/Group 2 n = 13 (27%), 0.01], crazy-paving pattern [Group 1 n = 98 (46%) / Group 2 n = 13 (27%), 0.01], consolidations [Group 1 n = 92 (43%)/Group 2 n = 8 (16%), 0.00], and CO-RADS 4 and 5 [Group 1 n = 163 (77.25%)/Group 2 n = 24 (50%), 0.00] categories in chest CT. Chest CT, when available, was found to be an efficient method for the initial diagnosis and better management of individuals with COVID-19.

Impact of Preexisting Pulmonary Fibrosis Detected on Chest Radiograph and CT on the Development of Gefitinib-Related Interstitial Lung Disease

PurposeThe novel coronavirus COVID-19, which spread globally in late December 2019, is a global health crisis. Chest computed tomography (CT) has played a pivotal role in providing useful information for clinicians to detect COVID-19. However, segmenting COVID-19-infected regions from chest CT results is challenging. Therefore, it is desirable to develop an efficient tool for automated segmentation of COVID-19 lesions using chest CT. Hence, we aimed to propose 2D deep-learning algorithms to automatically segment COVID-19-infected regions from chest CT slices and evaluate their performance.Material and methodsHerein, 3 known deep learning networks: U-Net, U-Net++, and Res-Unet, were trained from scratch for automated segmenting of COVID-19 lesions using chest CT images. The dataset consists of 20 labelled COVID-19 chest CT volumes. A total of 2112 images were used. The dataset was split into 80% for training and validation and 20% for testing the proposed models. Segmentation performance was assessed using Dice similarity coefficient, average symmetric surface distance (ASSD), mean absolute error (MAE), sensitivity, specificity, and precision.ResultsAll proposed models achieved good performance for COVID-19 lesion segmentation. Compared with Res-Unet, the U-Net and U-Net++ models provided better results, with a mean Dice value of 85.0%. Compared with all models, U-Net gained the highest segmentation performance, with 86.0% sensitivity and 2.22 mm ASSD. The U-Net model obtained 1%, 2%, and 0.66 mm improvement over the Res-Unet model in the Dice, sensitivity, and ASSD, respectively. Compared with Res-Unet, U-Net++ achieved 1%, 2%, 0.1 mm, and 0.23 mm improvement in the Dice, sensitivity, ASSD, and MAE, respectively.ConclusionsOur data indicated that the proposed models achieve an average Dice value greater than 84.0%. Two-dimensional deep learning models were able to accurately segment COVID-19 lesions from chest CT images, assisting the radiologists in faster screening and quantification of the lesion regions for further treatment. Nevertheless, further studies will be required to evaluate the clinical performance and robustness of the proposed models for COVID-19 semantic segmentation.

We propose a generalized framework for developing computer-aided detection (CADe) systems whose characteristics depend only on those of the training dataset. The purpose of this study is to show the feasibility of the framework. Two different CADe systems were experimentally developed by a prototype of the framework, but with different training datasets. The CADe systems include four components; preprocessing, candidate area extraction, candidate detection, and candidate classification. Four pretrained algorithms with dedicated optimization/setting methods corresponding to the respective components were prepared in advance. The pretrained algorithms were sequentially trained in the order of processing of the components. In this study, two different datasets, brain MRA with cerebral aneurysms and chest CT with lung nodules, were collected to develop two different types of CADe systems in the framework. The performances of the developed CADe systems were evaluated by threefold cross-validation. The CADe systems for detecting cerebral aneurysms in brain MRAs and for detecting lung nodules in chest CTs were successfully developed using the respective datasets. The framework was shown to be feasible by the successful development of the two different types of CADe systems. The feasibility of this framework shows promise for a new paradigm in the development of CADe systems: development of CADe systems without any lesion specific algorithm designing.

AimsIncreasing evidence suggests that poor glycemic control in diabetic individuals is associated with poor coronavirus disease 2019 (COVID-19) pneumonia outcomes and influences chest computed tomography (CT) manifestations. This study aimed to explore the impact of diabetes mellitus (DM) and glycemic control on chest CT manifestations, acquired using an artificial intelligence (AI)-based quantitative evaluation system, and COVID-19 disease severity and to investigate the association between CT lesions and clinical outcome.MethodsA total of 126 patients with COVID-19 were enrolled in this retrospective study. According to their clinical history of DM and glycosylated hemoglobin (HbA1c) level, the patients were divided into 3 groups: the non-DM group (Group 1); the well-controlled blood glucose (BG) group, with HbA1c < 7% (Group 2); and the poorly controlled BG group, with HbA1c ≥ 7% (Group 3). The chest CT images were analyzed with an AI-based quantitative evaluation system. Three main quantitative CT features representing the percentage of total lung lesion volume (PLV), percentage of ground-glass opacity volume (PGV) and percentage of consolidation volume (PCV) in bilateral lung fields were used to evaluate the severity of pneumonia lesions.ResultsPatients in Group 3 had the highest percentage of severe or critical illness, with 12 (32%) cases, followed by 6 (11%) and 7 (23%) cases in Groups 1 and 2, respectively (p = 0.042). The composite endpoints, including death or using mechanical ventilation or admission to the intensive care unit (ICU), were 3 (5%), 5 (16%) and 10 (26%) in Groups 1, 2 and 3, respectively (p = 0.013). The PLV, PGV and PCV in bilateral lung fields were significantly different among the three groups (all p < 0.001): the median PLVs were 12.5% (Group 3), 3.8% (Group 2) and 2.4% (Group 1); the median PGVs were 10.2% (Group 3), 3.6% (Group 2) and 1.9% (Group 1); and the median PCVs were 1.8% (Group 3), 0.3% (Group 2) and 0.1% (Group 1). In the linear regression analyses, which were adjusted for age, sex, BMI, and comorbidities, HbA1c remained positively associated with PLV (β = 0.401, p < 0.001), PGV (β = 0.364, p = 0.001) and PCV (β = 0.472, p < 0.001); this relationship was also observed between fasting blood glucose (FBG) and the three CT quantitative parameters. In the logistic regression analyses, PLV [OR 1.067 (1.032, 1.103)], PGV [OR 1.076 (1.034, 1.120)] and PCV [OR 1.280 (1.110, 1.476)] levels were independent predictors of the composite endpoints, as well as the areas under the ROC (AUCs) for PLV [AUC 0.796 (0.691, 0.900)], PGV [AUC 0.783 (0.678, 0.889)] and PCV [AUC 0.816 (0.722, 0.911)]; the ORs were still significant for CT lesions after adjusting for age, sex and poorly controlled diabetes.ConclusionsIncreased blood glucose level was correlated with the severity of lung involvement, as evidenced by certain chest CT parameters, and clinical prognosis in diabetic COVID-19 patients. There was a positive correlation between blood glucose level (both HbA1c and FBG) on admission and lung lesions. Moreover, the CT lesion severity by AI quantitative analysis was correlated with clinical outcomes.

To evaluate the diagnostic yield of recommended chest computed tomography (CT) prompted by abnormalities detected on outpatient chest radiographic images. This HIPAA-compliant study had institutional review board approval; informed consent was waived. Reports of all outpatient chest radiographic examinations performed at a large academic center during 2008 (n = 29 138) were queried to identify studies that included a recommendation for a chest CT imaging. The radiology information system was queried for these patients to determine if a chest CT examination was obtained within 1 year of the index radiographic examination that contained the recommendation. For chest CT examinations obtained within 1 year of the index chest radiographic examination and that met inclusion criteria, chest CT images were reviewed to determine if there was an abnormality that corresponded to the chest radiographic finding that prompted the recommendation. All corresponding abnormalities were categorized as clinically relevant or not clinically relevant, based on whether further work-up or treatment was warranted. Groups were compared by using t test and Fisher exact test with a Bonferroni correction applied for multiple comparisons. There were 4.5% (1316 of 29138 [95% confidence interval {CI}: 4.3%, 4.8%]) of outpatient chest radiographic examinations that contained a recommendation for chest CT examination, and increasing patient age (P < .001) and positive smoking history (P = .001) were associated with increased likelihood of a recommendation for chest CT examination. Of patients within this subset who met inclusion criteria, 65.4% (691 of 1057 [95% CI: 62.4%, 68.2%) underwent a chest CT examination within the year after the index chest radiographic examination. Clinically relevant corresponding abnormalities were present on chest CT images in 41.4% (286 of 691 [95% CI: 37.7%, 45.2%]) of cases, nonclinically relevant corresponding abnormalities in 20.6% (142 of 691 [95% CI: 17.6%, 23.8%]) of cases, and no corresponding abnormalities in 38.1% (263 of 691 [95% CI: 34.4%, 41.8%]) of cases. Newly diagnosed, biopsy-proven malignancies were detected in 8.1% (56 of 691 [95% CI: 6.2%, 10.4%]) of cases. A radiologist recommendation for chest CT to evaluate an abnormal finding on an outpatient chest radiographic examination has a high yield of clinically relevant findings.