Murine studies have linked TGF-β signaling to emphysema, and human genome-wide association studies (GWAS) studies of lung function and COPD have identified associated regions near genes in the TGF-β superfamily. However, the functional regulatory mechanisms at these loci have not been identified. We performed the largest GWAS of emphysema patterns to date, identifying 10 GWAS loci including an association peak spanning a 200 kb region downstream from TGFB2. Integrative analysis of publicly available eQTL, DNaseI, and chromatin conformation data identified a putative functional variant, rs1690789, that may regulate TGFB2 expression in human fibroblasts. Using chromatin conformation capture, we confirmed that the region containing rs1690789 contacts the TGFB2 promoter in fibroblasts, and CRISPR/Cas-9 targeted deletion of a ~ 100 bp region containing rs1690789 resulted in decreased TGFB2 expression in primary human lung fibroblasts. These data provide novel mechanistic evidence linking genetic variation affecting the TGF-β pathway to emphysema in humans.

RATIONALE: Occupational exposures at the WTC site after September 11, 2001 have been associated with several presumably inflammatory lower airway diseases. Pulmonary arterial enlargement, as suggested by an increased ratio of the diameter of the pulmonary artery to the diameter of the aorta (PAAr) has been reported as a computed tomographic (CT) scan marker of adverse respiratory health outcomes, including WTC-related disease. In this study, we sought to utilize a novel quantitative CT (QCT) measurement of PAAr to test the hypothesis that an increased ratio is associated with FEV1 below each subject's statistically determined lower limit of normal (FEV1  < LLN). METHODS: In a group of 1,180 WTC workers and volunteers, we examined whether FEV1  < LLN was associated with an increased QCT-measured PAAr, adjusting for previously identified important covariates. RESULTS: Unadjusted analyses showed a statistically significant association of FEV1  < LLN with PAAr (35.3% vs 24.7%, P = 0.0001), as well as with height, body mass index, early arrival at the WTC disaster site, shorter WTC exposure duration, post-traumatic stress disorder checklist (PCL) score, wall area percent and evidence of bronchodilator response. The multivariate logistic regression model confirmed the association of FEV1  < LLN with PAAr (OR 1.63, 95% CI 1.21, 2.20, P = 0.0015) and all the unadjusted associations, except for PCL score. CONCLUSIONS: In WTC workers, FEV1  < LLN is associated with elevated PAAr which, although likely multifactorial, may be related to distal vasculopathy, as has been hypothesized for chronic obstructive pulmonary disease.

Biomarker inference from biomedical images is one of the main tasks of medical image analysis. Standard techniques follow a segmentation-and-measure strategy, where the structure is first segmented and then the measurement is performed. Recent work has shown that such strategy could be replaced by a direct regression of the biomarker value in using regression networks. While achieving high correlation coefficients, such techniques operate as a 'black-box', not offering quality-control images. We present a methodology to regress the biomarker from the image while simultaneously computing the quality control image. Our proposed methodology does not require segmentation masks for training, but infers the segmentations directly from the pixels that used to compute the biomarker value. The network proposed consists of two steps: a segmentation method to an unknown reference and a summation method for the biomarker estimation. The network is optimized using a dual loss function, L2 for the biomarkers and an L1 to enforce sparsity. We showcase our methodology in the problem of pectoralis muscle area (PMA) and subcutaneous fat area (SFA) inference in a single slice from chest-CT images. We use a database of 7000 cases to which only the value of the biomarker is known for training and a test set of 3000 cases with both, biomarkers and segmentations. We achieve a correlation coefficient of 0.97 for PMA and 0.98 for SFA with respect to the reference standard. The average DICE coefficient is of 0.88 (PMA) and 0.89 (SFA). Comparing with standard segment-and-measure techniques, we achieve the same correlation for the biomarkers but smaller DICE coefficients in segmentation. Such is of little surprise, since segmentation networks are the upper limit of performance achievable, and we are not using segmentation masks for training. We can conclude that it is possible to infer segmentation masks from biomarker regression networks.

BACKGROUND: Tobacco smoke exposure is associated with emphysema and pulmonary fibrosis, both of which are irreversible. We have developed a new objective CT analysis tool that combines densitometry with machine learning to detect high attenuation changes in visually normal appearing lung (NormHA) that may precede these diseases. METHODS: We trained the classification tool by placing 34,528 training points in chest CT scans from 297 COPDGene participants. The tool was then used to classify lung tissue in 9,038 participants as normal, emphysema, fibrotic/interstitial, or NormHA. Associations between the quartile of NormHA and plasma-based biomarkers, clinical severity, and mortality were evaluated using Jonckheere-Terpstra, pairwise Wilcoxon rank-sum tests, and multivariable linear and Cox regression. RESULTS: A higher percentage of lung occupied by NormHA was associated with higher C-reactive protein and intercellular adhesion molecule 1 (P for trend for both < .001). In analyses adjusted for multiple covariates, including high and low attenuation area, compared with those in the lowest quartile of NormHA, those in the highest quartile had a 6.50 absolute percent lower percent predicted lower FEV1 (P < .001), an 8.48 absolute percent lower percent predicted forced expiratory volume, a 10.78-meter shorter 6-min walk distance (P = .011), and a 56% higher risk of death (P = .003). These findings were present even in those individuals without visually defined interstitial lung abnormalities. CONCLUSIONS: A new class of NormHA on CT may represent a unique tissue class associated with adverse outcomes, independent of emphysema and fibrosis.

BACKGROUND: Exposure to cigarette smoke has been shown to lead to vascular remodelling. Computed tomography (CT) imaging measures of vascular pruning have been associated with pulmonary vascular disease, an important morbidity associated with smoking. In this study we compare CT-based measures of distal vessel loss to histological vascular and parenchymal changes. METHODS: A retrospective review of 80 patients who had undergone lung resection identified patients with imaging appropriate for three-dimensional (3D) vascular reconstruction (n=18) and a second group for two-dimensional (2D) analysis (n=19). Measurements of the volume of the small vessels (3D) and the cross-sectional area of the small vessels (<5 mm2 cross-section) were computed. Histological measures of cross-sectional area of the vasculature and loss of alveoli septa were obtained for all subjects. RESULTS: The 2D cross-sectional area of the vasculature on CT imaging was associated with the histological vascular cross-sectional area (r=0.69; p=0.001). The arterial small vessel volume assessed by CT correlated with the histological vascular cross-sectional area (r=0.50; p=0.04), a relationship that persisted even when adjusted for CT-derived measures of emphysema in a regression model. CONCLUSIONS: Loss of small vessel volume in CT imaging of smokers is associated with histological loss of vascular cross-sectional area. Imaging-based quantification of pulmonary vasculature provides a noninvasive method to study the multiscale effects of smoking on the pulmonary circulation.

Radiographic abnormalities of the pulmonary vessels, such as vascular pruning, are common in advanced airways disease, but it is unknown if pulmonary vascular volumes are related to measures of lung health and airways disease in healthier populations.In 2388 participants of the Framingham Heart Study computed tomography (CT) sub-study, we calculated total vessel volumes and the small vessel fraction using automated CT image analysis. We evaluated associations with measures of lung function, airflow obstruction on spirometry and emphysema on CT. We further tested if associations of vascular volumes with lung function were present among those with normal forced expiratory volume in 1 s and forced vital capacity.In fully adjusted linear and logistic models, we found that lower total and small vessel volumes were consistently associated with worse measures of lung health, including lower spirometric volumes, lower diffusing capacity and/or higher odds of airflow obstruction. For example, each standard deviation lower small vessel fraction (indicating more severe pruning) was associated with a 37% greater odds of obstruction (OR 1.37, 95% CI 1.11-1.71, p=0.004). A similar pattern was observed in the subset of participants with normal spirometry.Lower total and small vessel pulmonary vascular volumes were associated with poorer measures of lung health and/or greater odds of airflow obstruction in this cohort of generally healthy adults without high burdens of smoking or airways disease. Our findings suggest that quantitative CT assessment may detect subtle pulmonary vasculopathy that occurs in the setting of subclinical and early pulmonary and airways pathology.

BACKGROUND: Respiratory pathology is a major driver of mortality in the intensive care unit (ICU), even in the absence of a primary respiratory diagnosis. Prior work has demonstrated that a visual scoring system applied to chest radiographs (CXR) is associated with adverse outcomes in ICU patients with Acute Respiratory Distress Syndrome (ARDS). We hypothesized that a simple, semi-quantitative CXR score would be associated with clinical outcomes for the general ICU population, regardless of underlying diagnosis. METHODS: All individuals enrolled in the Registry of Critical Illness at Brigham and Women's Hospital between June 2008 and August 2018 who had a CXR within 24 h of admission were included. Each patient's CXR was assigned an opacification score of 0-4 in each of four quadrants with the total score being the sum of all four quadrants. Multivariable negative binomial, logistic, and Cox regression, adjusted for age, sex, race, immunosuppression, a history of chronic obstructive pulmonary disease, a history of congestive heart failure, and APACHE II scores, were used to assess the total score's association with ICU length of stay (LOS), duration of mechanical ventilation, in-hospital mortality, 60-day mortality, and overall mortality, respectively. RESULTS: A total of 560 patients were included. Higher CXR scores were associated with increased mortality; for every one-point increase in score, in-hospital mortality increased 10% (OR 1.10, CI 1.05-1.16, p < 0.001) and 60-day mortality increased by 12% (OR 1.12, CI 1.07-1.17, p < 0.001). CXR scores were also independently associated with both ICU length of stay (rate ratio 1.06, CI 1.04-1.07, p < 0.001) and duration of mechanical ventilation (rate ratio 1.05, CI 1.02-1.07, p < 0.001). CONCLUSIONS: Higher values on a simple visual score of a patient's CXR on admission to the medical ICU are associated with increased in-hospital mortality, 60-day mortality, overall mortality, length of ICU stay, and duration of mechanical ventilation.

In the last five years, deep learning (DL) has become the state-of-the-art tool for solving various tasks in medical image analysis. Among the different methods that have been proposed to improve the performance of Convolutional Neural Networks (CNNs), one typical approach is the augmentation of the training data set through various transformations of the input image. Data augmentation is typically used in cases where a small amount of data is available, such as the majority of medical imaging problems, to present a more substantial amount of data to the network and improve the overall accuracy. However, the ability of the network to improve the accuracy of the results when a slightly modified version of the same input is presented is often overestimated. This overestimation is the result of the strong correlation between data samples when they are considered independently in the training phase. In this paper, we emphasize the importance of optimizing for accuracy as well as precision among multiple replicates of the same training data in the context of data augmentation. To this end, we propose a new approach that leverages the augmented data to help the network focus on the precision through a specifically-designed loss function, with the ultimate goal to improve both the overall performance and the network's precision at the same time. We present two different applications of DL (regression and segmentation) to demonstrate the strength of the proposed strategy. We think that this work will pave the way to a explicit use of data augmentation within the loss function that helps the network to be invariant to small variations of the same input samples, a characteristic that is always required to every application in the medical imaging field.

Pulmonary emphysema is a destructive disease of the lungs that is currently diagnosed via visual assessment of lung Computed Tomography (CT) scans by a radiologist. Visual assessment can have poor inter-rater reliability, is time consuming, and requires access to trained assessors. Quantitative methods that reliably summarize the biologically relevant characteristics of an image are needed to improve the way lung diseases are characterized. The goal of this work was to show how spatial point process models can be used to create a set of radiologically derived quantitative lung biomarkers of emphysema. We formalized a general framework for applying spatial point processes to lung CT scans, and developed a Shot Noise Cox Process to quantify how radiologically based emphysematous tissue clusters into larger structures. Bayesian estimation of model parameters was done using spatial Birth-Death MCMC (BD-MCMC). In simulations, we showed the BD-MCMC estimation algorithm is able to accurately recover model parameters. In an application to real lung CT scans from the COPDGene cohort, we showed variability in the clustering characteristics of emphysematous tissue across disease subtypes that were based on visual assessments of the CT scans.

PURPOSE: To develop and validate a computed tomography (CT) harmonization technique by combining noise-stabilization and autocalibration methodologies to provide reliable densitometry measurements in heterogeneous acquisition protocols. METHODS: We propose to reduce the effects of spatially variant noise such as nonuniform patterns of noise and biases. The method combines the statistical characterization of the signal-to-noise relationship in the CT image intensities, which allows us to estimate both the signal and spatially variant variance of noise, with an autocalibration technique that reduces the nonuniform biases caused by noise and reconstruction techniques. The method is firstly validated with anthropomorphic synthetic images that simulate CT acquisitions with variable scanning parameters: different dosage, nonhomogeneous variance of noise, and various reconstruction methods. We finally evaluate these effects and the ability of our method to provide consistent densitometric measurements in a cohort of clinical chest CT scans from two vendors (Siemens, n = 54 subjects; and GE, n = 50 subjects) acquired with several reconstruction algorithms (filtered back-projection and iterative reconstructions) with high-dose and low-dose protocols. RESULTS: The harmonization reduces the effect of nonhomogeneous noise without compromising the resolution of the images (25% RMSE reduction in both clinical datasets). An analysis through hierarchical linear models showed that the average biases induced by differences in dosage and reconstruction methods are also reduced up to 74.20%, enabling comparable results between high-dose and low-dose reconstructions. We also assessed the statistical similarity between acquisitions obtaining increases of up to 30% points and showing that the low-dose vs high-dose comparisons of harmonized data obtain similar and even higher similarity than the observed for high-dose vs high-dose comparisons of nonharmonized data. CONCLUSION: The proposed harmonization technique allows to compare measures of low-dose with high-dose acquisitions without using a specific reconstruction as a reference. Since the harmonization does not require a precalibration with a phantom, it can be applied to retrospective studies. This approach might be suitable for multicenter trials for which a reference reconstruction is not feasible or hard to define due to differences in vendors, models, and reconstruction techniques.

Rationale: Cigarette smoke exposure is a risk factor for many lung diseases, and histologic studies suggest that tobacco-related vasoconstriction and vessel loss plays a role in the development of emphysema. However, it remains unclear how tobacco affects the pulmonary vasculature in general populations with a typical range of tobacco exposure, and whether these changes are detectable by radiographic methods. Objectives: To determine whether tobacco exposure in a generally healthy population manifests as lower pulmonary blood vessel volumes and vascular pruning on imaging. Methods: A total of 2,410 Framingham Heart Study participants with demographic data and smoking history underwent volumetric whole-lung computed tomography from 2008 to 2011. Automated algorithms calculated the total blood volume of all intrapulmonary vessels (TBV), smaller peripheral vessels (defined as cross-sectional area <5 mm2 [BV5]), and the relative fraction of small vessels (BV5/TBV). Tobacco exposure was assessed as smoking status, cumulative pack-years, and second-hand exposure. We constructed multivariable linear regression models to evaluate associations of cigarette exposure and pulmonary blood vessel volume measures, adjusting for demographic covariates, including age, sex, height, weight, education, occupation, and median neighborhood income. Results: All metrics of tobacco exposure (including smoking status, pack-years, and second-hand exposure) were consistently associated with higher absolute pulmonary blood vessel volume, higher small vessel volume, and/or higher small vessel fraction. For example, ever-smokers had a 4.6 ml higher TBV (95% confidence interval [CI] = 2.9-6.3, P < 0.001), 2.1 ml higher BV5 (95% CI = 1.3-2.9, P < 0.001), and 0.28 percentage-point-higher BV5/TBV (95% CI = 0.03-0.52, P = 0.03) compared with never-smokers. These associations remained significant after adjustment for percent predicted forced expiratory volume in 1 second, cardiovascular comorbidities, and did not differ based on presence or absence of airflow obstruction. Conclusions: Using computed tomographic imaging, we found that cigarette exposure was associated with higher pulmonary blood vessel volumes, especially in the smaller peripheral vessels. Although, histologically, tobacco-related vasculopathy is characterized by vessel narrowing and loss, our results suggest that radiographic vascular pruning may not be a surrogate of these pathologic changes.

Rationale: Interstitial lung abnormalities (ILA) are radiologic abnormalities on chest computed tomography scans that have been associated with an early or mild form of pulmonary fibrosis. Although ILA have been associated with radiologic progression, it is not known if specific imaging patterns are associated with progression or risk of mortality. Objectives: To determine the role of imaging patterns on the risk of death and ILA progression. Methods: ILA (and imaging pattern) were assessed in 5,320 participants from the AGES-Reykjavik Study, and ILA progression was assessed in 3,167 participants. Multivariable logistic regression was used to assess factors associated with ILA progression, and Cox proportional hazards models were used to assess time to mortality. Measurements and Main Results: Over 5 years, 327 (10%) had ILA on at least one computed tomography, and 1,435 (45%) did not have ILA on either computed tomography. Of those with ILA, 238 (73%) had imaging progression, whereas 89 (27%) had stable to improved imaging; increasing age and copies of MUC5B genotype were associated with imaging progression. The definite fibrosis pattern was associated with the highest risk of progression (odds ratio, 8.4; 95% confidence interval, 2.7-25; P = 0.0003). Specific imaging patterns were also associated with an increased risk of death. After adjustment, both a probable usual interstitial pneumonia and usual interstitial pneumonia pattern were associated with an increased risk of death when compared with those indeterminate for usual interstitial pneumonia (hazard ratio, 1.7; 95% confidence interval, 1.2-2.4; P = 0.001; hazard ratio, 3.9; 95% confidence interval, 2.3-6.8;P < 0.0001), respectively. Conclusions: In those with ILA, imaging patterns can be used to help predict who is at the greatest risk of progression and early death.

RATIONALE: There is increasing evidence that aberrant processes occurring in the airways may precede the development of idiopathic pulmonary fibrosis (IPF); however, there has been no prior confirmatory data derived from imaging studies. OBJECTIVES: To assess quantitative measures of airway wall thickness (AWT) in populations characterized for interstitial lung abnormalities (ILA) and for IPF. METHODS: Computed tomographic imaging of the chest and measures of AWT were available for 6,073, 615, 1,167, and 38 participants from COPDGene (Genetic Epidemiology of COPD study), ECLIPSE (Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints study), and the Framingham Heart Study (FHS) and in patients with IPF from the Brigham and Women's Hospital Herlihy Registry, respectively. To evaluate these associations, we used multivariable linear regression to compare a standardized measure of AWT (the square root of AWT for airways with an internal perimeter of 10 mm [Pi10]) and characterizations of ILA and IPF by computed tomographic imaging of the chest. RESULTS: In COPDGene, ECLIPSE, and FHS, research participants with ILA had increased measures of Pi10 compared with those without ILA. Patients with IPF had mean measures of Pi10 that were even greater than those noted in research participants with ILA. After adjustment for important covariates (e.g., age, sex, race, body mass index, smoking behavior, and chronic obstructive pulmonary disease severity when appropriate), research participants with ILA had increased measures of Pi10 compared with those without ILA (0.03 mm in COPDGene, 95% confidence interval [CI], 0.02-0.03; P < 0.001; 0.02 mm in ECLIPSE, 95% CI, 0.005-0.04; P = 0.01; 0.07 mm in FHS, 95% CI, 0.01-0.1; P = 0.01). Compared with COPDGene participants without ILA older than 60 years of age, patients with IPF were also noted to have increased measures of Pi10 (2.0 mm, 95% CI, 2.0-2.1; P < 0.001). Among research participants with ILA, increases in Pi10 were correlated with reductions in lung volumes in some but not all populations. CONCLUSIONS: These results demonstrate that measurable increases in AWT are consistently noted in research participants with ILA and in patients with IPF. These findings suggest that abnormalities of the airways may play a role in, or be correlated with, early pathogenesis of pulmonary fibrosis.

The characterization of the vasculature in the mediastinum, more specifically the pulmonary artery, is of vital importance for the evaluation of several pulmonary vascular diseases. Thus, the goal of this study is to automatically segment the pulmonary artery (PA) from computed tomography angiography images, which opens up the opportunity for more complex analysis of the evolution of the PA geometry in health and disease and can be used in complex fluid mechanics models or individualized medicine. For that purpose, a new 3D convolutional neural network architecture is proposed, which is trained on images coming from different patient cohorts. The network makes use a strong data augmentation paradigm based on realistic deformations generated by applying principal component analysis to the deformation fields obtained from the affine registration of several datasets. The network is validated on 91 datasets by comparing the automatic segmentations with semi-automatically delineated ground truths in terms of mean Dice and Jaccard coefficients and mean distance between surfaces, which yields values of 0.89, 0.80 and 1.25 mm, respectively. Finally, a comparison against a Unet architecture is also included.

Imaging continues to be the modality of choice for the diagnosis of venous thromboembolic disease, particularly when incorporated into diagnostic algorithms. Improvement in imaging techniques as well as new imaging modalities and processing methods have improved diagnostic accuracy and additionally are being leveraged in prognostication and decision making for choice of intervention. In this article, we review the role of imaging in diagnosis and prognostication of venous thromboembolism. We also discuss emerging imaging approaches that may in the near future find clinical usefulness in improving diagnosis and prognostication as well as differentiating disease phenotypes.