Search

Background Aortic diameter measurements in patients with a thoracic aortic aneurysm (TAA) show wide variation. There is no technique to quantify aortic growth in a three-dimensional (3D) manner. Purpose To validate a CT-based technique for quantification of 3D growth based on deformable registration in patients with TAA. Materials and Methods Patients with ascending and descending TAA with two or more CT angiography studies between 2006 and 2020 were retrospectively identified. The 3D aortic growth was quantified using vascular deformation mapping (VDM), a technique that uses deformable registration to warp a mesh constructed from baseline aortic anatomy. Growth assessments between VDM and clinical CT diameter measurements were compared. Aortic growth was quantified as the ratio of change in surface area at each mesh element (area ratio). Manual segmentations were performed by independent raters to assess interrater reproducibility. Registration error was assessed using manually placed landmarks. Agreement between VDM and clinical diameter measurements was assessed using Pearson correlation and Cohen κ coefficients. Results A total of 38 patients (68 surveillance intervals) were evaluated (mean age, 69 years ± 9 [standard deviation]; 21 women), with TAA involving the ascending aorta (n = 26), descending aorta (n = 10), or both (n = 2). VDM was technically successful in 35 of 38 (92%) patients and 58 of 68 intervals (85%). Median registration error was 0.77 mm (interquartile range, 0.54-1.10 mm). Interrater agreement was high for aortic segmentation (Dice similarity coefficient = 0.97 ± 0.02) and VDM-derived area ratio (bias = 0.0, limits of agreement: -0.03 to 0.03). There was strong agreement (r = 0.85, P < .001) between peak area ratio values and diameter change. VDM detected growth in 14 of 58 (24%) intervals. VDM revealed growth outside the maximally dilated segment in six of 14 (36%) growth intervals, none of which were detected with diameter measurements. Conclusion Vascular deformation mapping provided reliable and comprehensive quantitative assessment of three-dimensional aortic growth and growth patterns in patients with thoracic aortic aneurysms undergoing CT surveillance. Published under a CC BY 4.0 license Online supplemental material is available for this article. See also the editorial by Wieben in this issue.

Background Lung cancer risk prediction models do not routinely incorporate imaging metrics available on low-dose CT (LDCT) imaging of the chest ordered for lung cancer screening. Research Question What is the association between quantitative emphysema measured on LDCT imaging and lung cancer incidence and mortality, all-cause mortality, and airflow obstruction in individuals who currently or formerly smoked and are undergoing lung cancer screening? Study Design and Methods In 7,262 participants in the CT arm of the National Lung Screening Trial, percent low attenuation area (%LAA) was defined as the percentage of lung volume with voxels less than –950 Hounsfield units on the baseline examination. Multivariable Cox proportional hazards models, adjusting for competing risks where appropriate, were built to test for association between %LAA and lung cancer incidence, lung cancer mortality, and all-cause mortality with censoring at 6 years. In addition, multivariable logistic regression models were built to test the cross-sectional association between %LAA and airflow obstruction on spirometry, which was available in 2,700 participants. Results The median %LAA was 0.8% (interquartile range, 0.2%-2.7%). Every 1% increase in %LAA was independently associated with higher hazards of lung cancer incidence (hazard ratio [HR], 1.02; 95% CI, 1.01-1.03; P = .004), lung cancer mortality (HR, 1.02; 95% CI, 1.00-1.05; P = .045), and all-cause mortality (HR, 1.01; 95% CI, 1.00-1.03; P = .042). Among participants with spirometry, 892 had airflow obstruction. The likelihood of airflow obstruction increased with every 1% increase in %LAA (odds ratio, 1.07; 95% CI, 1.06-1.09; P 65 years. Interpretation Quantitative emphysema measured on LDCT imaging of the chest can be leveraged to improve lung cancer risk prediction and help diagnose COPD in individuals who currently or formerly smoked and are undergoing lung cancer screening.

Background: Little is known about factors associated with emphysema progression in cigarette smokers. We evaluated factors associated with change in emphysema and forced expiratory volume in 1 second (FEV(1)) in participants with and without chronic obstructive pulmonary disease (COPD). Methods: This retrospective study included individuals participating in the COPD Genetic Epidemiology study who completed the 5-year follow-up, including inspiratory and expiratory computed tomography (CT) and spirometry. All paired CT scans were analyzed using micro-mapping, which classifies individual voxels as emphysema or functional small airway disease (fSAD). Presence and progression of emphysema and FEV(1) were determined based on comparison to nonsmoker values. Logistic regression analyses were used to identify clinical parameters associated with disease progression. Results: A total of 3088 participants were included with a mean ± SD age of 60.7±8.9 years, including 72 nonsmokers. In all Global initiative for chronic Obstructive Lung Disease (GOLD) stages, the presence of emphysema at baseline was associated with emphysema progression (odds ratio [OR]: GOLD 0: 4.32; preserved ratio-impaired spirometry [PRISm]; 5.73; GOLD 1: 5.16; GOLD 2: 5.69; GOLD 3/4: 5.55; all p ≤0.01). If there was no emphysema at baseline, the amount of fSAD at baseline was associated with emphysema progression (OR for 1% increase: GOLD 0: 1.06; PRISm: 1.20; GOLD 1: 1.7; GOLD 3/4: 1.08; all p ≤ 0.03).In 1735 participants without spirometric COPD, progression in emphysema occurred in 105 (6.1%) participants and only 21 (1.2%) had progression in both emphysema and FEV(1). Conclusions: The presence of emphysema is an important predictor of emphysema progression. In patients without emphysema, fSAD is associated with the development of emphysema. In participants without spirometric COPD, emphysema progression occurred independently of FEV(1) decline.

Parametric response mapping (PRM) is a novel computed tomography (CT) technology that has shown potential for assessment of bronchiolitis obliterans syndrome (BOS) after hematopoietic stem cell transplantation (HCT). The primary aim of this study was to evaluate whether variations in image acquisition under real-world conditions affect the PRM measurements of clinically diagnosed BOS. CT scans were obtained retrospectively from 72 HCT recipients with BOS and graft-versus-host disease from Fred Hutchinson Cancer Research Center, Karolinska Institute, and the University of Michigan. Whole lung volumetric scans were performed at inspiration and expiration using site-specific acquisition and reconstruction protocols. PRM and pulmonary function measurements were assessed. Patients with moderately severe BOS at diagnosis (median forced expiratory volume at 1 second [FEV1] 53.5% predicted) had similar characteristics between sites. Variations in site-specific CT acquisition protocols had a negligible effect on the PRM-derived small airways disease (SAD), that is, BOS measurements. PRM-derived SAD was found to correlate with FEV1% predicted and FEV1/ forced vital capacity (R = -0.236, P = .046; and R = -0.689, P < .0001, respectively), which suggests that elevated levels in the PRM measurements are primarily affected by BOS airflow obstruction and not CT scan acquisition parameters. Based on these results, PRM may be applied broadly for post-HCT diagnosis and monitoring of BOS.

Rationale: Evidence suggests damage to small airways is a key pathologic lesion in chronic obstructive pulmonary disease (COPD). Computed tomography densitometry has been demonstrated to identify e...

Thoracic aortic aneurysm (TAA) growth is currently assessed by changes in maximal aortic diameter (i.e., radial growth of the vessel). However, there is growing awareness that longitudinal aortic growth (i.e., elongation) is an important metric of disease status, albeit one that is difficult to measure using current clinical techniques. Previously we have proposed a method to assess 3D changes in aortic wall growth/deformation using deformable image registration with interpolation of the spatial Jacobian determinant to the aortic surface. Here we propose a method to re-orient the Jacobian into directional components relative to the aortic surface rather than the image space, allowing clinicians and researchers to isolate and study the pathologic effects of each directional component of growth separately. To this end, we first perform a deformable image registration between two aortic geometries. Second, we segment the aortic surface and centerline in the fixed image, and use the resulting geometry to construct anatomically-based local coordinate systems at each voxel of the aortic surface. Using the Jacobian matrix field resulting from the deformable registration, we can obtain the anatomically oriented Jacobian components by rotating the Jacobian matrix at each voxel so that it is aligned with the anatomically based local coordinate system. Through experiments on toy cylinders and real clinical cases, we show clear differences between the Jacobian determinant and its directional components, with the directional Jacobian component being able to remove one directional change (e.g., longitudinal) while maintaining the other (e.g., cross-sectional).

Accurate segmentation of the aorta in computed tomography angiography (CTA) images is the first step for analysis of diseases such as aortic aneurysm, but manual segmentation can be prohibitively time-consuming and error prone. Convolutional neural network (CNN) based models have been utilized for automated segmentation of anatomy in CTA scans, with the ubiquitous U-Net being one of the most popular architectures. For many downstream image analysis tasks (e.g., registration, diameter measurement) very accurate segmentation accuracy may be required. In this work, we developed and tested a U-Net model with attention gating for segmentation of the thoracic aorta in clinical CTA data of patients with thoracic aortic aneurysm. Attention gating helps the model focus on difficult to segment target structures automatically and has been previously shown to increase segmentation accuracy in other applications. We trained U-Nets both with and without attention gating on 145 CTAs. Performance of the models were evaluated by calculating the DCS and Average Hausdorff Distance (AHD) on a test set of 20 CTAs. We found that the U-Net with attention gating yields more accurate segmentation than the U-Net without attention gating (DCS 0.966±0.028 vs. 0.944±0.022, AHD 0.189±0.134mm vs. 0.247±0.155mm). Furthermore, we explored the segmentation accuracy of this U-Net for multi-class labeling of various anatomic segments of the thoracic aorta, and found an average DCS of 0.86 for across 7 different labels. We conclude that the U-Net with attention gating improves segmentation performance and may aid segmentation tasks that require high levels of accuracy.

Despite good performance for medical image segmentation, deep convolutional neural networks (CNNs) have not been widely accepted in clinical practice as they are complex and tend to fail silently. Additionally, uncertainty in their predictions are not well understood, making them obscure and challenging to interpret. Automatically detecting possible failures in network predictions is important, as we can refer such cases for manual inspection or correction by human observers. In this paper, we analyse the uncertainty for deep CNN-based lung lobe segmentation in computed tomography (CT) scans by proposing a test-time augmentation-based aleatoric uncertainty measure. Through this analysis, we produce spatial uncertainty maps, from which a clinician can observe where and why a system thinks it is failing, and quantify the image-level prediction of failure. Our results show that such an uncertainty measure is highly correlated to segmentation accuracy and therefore presents an inherent measure of segmentation quality.

Abstract Impaired single breath carbon monoxide diffusing capacity (DLCO) is associated with emphysema. Small airways disease (SAD) may be a precursor lesion to emphysema, but the relationship between SAD and DLCO is undescribed. We hypothesized that in mild COPD, functional SAD (fSAD) defined by computed tomography (CT) and Parametric Response Mapping methodology would correlate with impaired DLCO. Using data from ever-smokers in the COPDGene cohort, we established that fSAD correlated significantly with lower DLCO among both non-obstructed and GOLD 1–2 subjects. The relationship between DLCO with CT-defined emphysema was present in all GOLD stages, but most prominent in severe disease. Trial registration NCT00608764. Registry: COPDGene. Registered 06 February 2008, retrospectively registered.

PURPOSE: Multiple commercial, open-source, and academic software tools exist for objective quantification of lung density in computed tomography (CT) images. The purpose of this study was to evaluate the intersoftware reproducibility of CT lung density measurements. METHODS: Computed tomography images from 50 participants from the COPDGene™ cohort study were randomly selected for analysis; n = 10 participants across each global initiative for chronic obstructive lung disease (GOLD) grade (GOLD 0–IV). Academic-based groups (n = 4) and commercial vendors (n = 4) participated anonymously to generate CT lung density measurements using their software tools. Computed tomography total lung volume (TLV), percentage of the low attenuation areas in the lung with Hounsfield unit (HU) values below −950HU (LAA(950)), and the HU value corresponding to the 15th percentile on the parenchymal density histogram (Perc15) were included in the analysis. The intersoftware bias and reproducibility coefficient (RDC) was generated with and without quality assurance (QA) for manual correction of the lung segmentation; intrasoftware bias and RDC was also generated by repeated measurements on the same images. RESULTS: Intersoftware mean bias was within ±0.22 mL, ±0.46%, and ±0.97 HU for TLV, LAA(950) and Perc15, respectively. The RDC was 0.35 L, 1.2% and 1.8 HU for TLV, LAA(950) and Perc15, respectively. Intersoftware RDC remained unchanged following QA: 0.35 L, 1.2% and 1.8 HU for TLV, LAA(950) and Perc15, respectively. All software investigated had an intrasoftware RDC of 0. The RDC was comparable for TLV, LAA(950) and Perc15 measurements, respectively, for academic-based groups/commercial vendor-based software tools: 0.39 L/0.32 L, 1.2%/1.2%, and 1.7 HU/1.6 HU. Multivariable regression analysis showed that academic-based software tools had greater within-subject standard deviation of TLV than commercial vendors, but no significant differences between academic and commercial groups were found for LAA(950) or Perc15 measurements. CONCLUSIONS: Computed tomography total lung volume and lung density measurement bias and reproducibility was reported across eight different software tools. Bias was negligible across vendors, reproducibility was comparable for software tools generated by academic-based groups and commercial vendors, and segmentation QA had negligible impact on measurement variability between software tools. In summary, results from this study report the amount of additional measurement variability that should be accounted for when using different software tools to measure lung density longitudinally with well-standardized image acquisition protocols. However, intrasoftware reproducibility was deterministic for all cases so use of the same software tool to reduce variability for serial studies is highly recommended.

In this work, a convolutional neural network (CNN) based method for automated lung boundary estimation from computed tomography (CT) scans is presented and validated. The CNN model was trained to regress the locations of the superior and inferior borders of the lungs from multiple tissue-specific 2D projections of thoracic CT images. The model utilized a DenseNet architecture and was trained and evaluated on CT images from the COPDGene study. The median (95th percentile) localization error was 2.51 (11.18) for the inferior border and 1.52 (7.21) for the superior border of the lungs.

Rationale and Objectives Chronic obstructive pulmonary disease is a heterogeneous disease characterized by small airway abnormality and emphysema. We hypothesized that a voxel-wise computed tomography analytic approach would identify patterns of disease progression in smokers. Materials and Methods We analyzed 725 smokers in spirometric GOLD stages 0-4 with two chest CTs 5 years apart. Baseline inspiration, follow-up inspiration and follow-up expiration images were spatially registered to baseline expiration so that each voxel had correspondences across all time points and respiratory phases. Voxel-wise Parametric Response Mapping (PRM) was then generated for the baseline and follow-up scans. PRM classifies lung as normal, functional small airway disease (PRMfSAD), and emphysema (PRMEMPH). Results Subjects with low baseline PRMfSAD and PRMEMPH predominantly had an increase in PRMfSAD on follow-up; those with higher baseline PRMfSAD and PRMEMPH mostly had increases in PRMEMPH. For GOLD 0 participants (n = 419), mean 5-year increases in PRMfSAD and PRMEMPH were 0.3% for both; for GOLD 1–4 participants (n = 306), they were 0.6% and 1.6%, respectively. Eighty GOLD 0 subjects (19.1%) had overall radiologic progression (30.0% to PRMfSAD, 52.5% to PRMEMPH, and 17.5% to both); 153 GOLD 1–4 subjects (50.0%) experienced progression (17.6% to PRMfSAD, 48.4% to PRMEMPH, and 34.0% to both). In a multivariable model, both baseline PRMfSAD and PRMEMPH were associated with development of PRMEMPH on follow-up, although this relationship was diminished at higher levels of baseline PRMEMPH. Conclusion A voxel-wise longitudinal PRM analytic approach can identify patterns of disease progression in smokers with and without chronic obstructive pulmonary disease.

PURPOSE: To evaluate the reproducibility and predicted clinical outcomes of CT-based quantitative lung density measurements using standard fixed-dose (FD) and reduced-dose (RD) scans. MATERIALS AND METHODS: In this retrospective analysis of prospectively acquired data, 1205 participants (mean age, 65 years ± 9 [standard deviation]; 618 men) enrolled in the COPDGene study who underwent FD and RD CT image acquisition protocols between November 2014 and July 2017 were included. Of these, the RD scans of 640 participants were also reconstructed using iterative reconstruction (IR). Median filtering was applied to the RD scans (RD-MF) to investigate an alternative noise reduction strategy. CT attenuation at the 15th percentile of the lung CT histogram (Perc15) was computed for all image types (FD, RD, RD-MF, and RD-IR). Reproducibility coefficients were calculated to quantify the measurement differences between FD and RD scans. The ability of Perc15 to predict chronic obstructive pulmonary disease (COPD) diagnosis and exacerbation frequency was investigated using receiver operating characteristic analysis. RESULTS: The Perc15 reproducibility coefficients with and without volume adjustment were as follows: RD, 29.43 HU ± 0.62 versus 32.81 HU ± 1.70; RD-MF, 7.42 HU ± 0.42 versus 19.40 HU ± 2.65; and RD-IR, 7.10 HU ± 0.52 versus 22.46 HU ± 3.91. Receiver operating characteristic curve analysis indicated that Perc15 on volume-adjusted FD and RD scans were both predictive for COPD diagnosis (area under the receiver operating characteristic curve [AUC]: FD, 0.724 ± 0.045; RD, 0.739 ± 0.045) and for having one or more exacerbation per year (AUCs: FD, 0.593 ± 0.068; RD, 0.589 ± 0.066). Similar trends were observed when volume adjustment was not applied. CONCLUSION: A combination of volume adjustment and noise reduction filtering improved the reproducibility of lung density measurements computed using serial FD and RD CT scans. Supplemental material is available for this article. © RSNA, 2021

Background Radiologic evidence of air trapping (AT) on expiratory computed tomography (CT) scans is associated with early pulmonary dysfunction in patients with cystic fibrosis (CF). However, standard techniques for quantitative assessment of AT are highly variable, resulting in limited efficacy for monitoring disease progression. Objective To investigate the effectiveness of a convolutional neural network (CNN) model for quantifying and monitoring AT, and to compare it with other quantitative AT measures obtained from threshold-based techniques. Materials and methods Paired volumetric whole lung inspiratory and expiratory CT scans were obtained at four time points (0, 3, 12 and 24 months) on 36 subjects with mild CF lung disease. A densely connected CNN (DN) was trained using AT segmentation maps generated from a personalized threshold-based method (PTM). Quantitative AT (QAT) values, presented as the relative volume of AT over the lungs, from the DN approach were compared to QAT values from the PTM method. Radiographic assessment, spirometric measures, and clinical scores were correlated to the DN QAT values using a linear mixed effects model. Results QAT values from the DN were found to increase from 8.65% ± 1.38% to 21.38% ± 1.82%, respectively, over a two-year period. Comparison of CNN model results to intensity-based measures demonstrated a systematic drop in the Dice coefficient over time (decreased from 0.86 ± 0.03 to 0.45 ± 0.04). The trends observed in DN QAT values were consistent with clinical scores for AT, bronchiectasis, and mucus plugging. In addition, the DN approach was found to be less susceptible to variations in expiratory deflation levels than the threshold-based approach. Conclusion The CNN model effectively delineated AT on expiratory CT scans, which provides an automated and objective approach for assessing and monitoring AT in CF patients.

Purpose: Lung cancer has been the leading cause of cancer-related deaths worldwide. To address the clinical need for efficacious treatments, genetically engineered mouse models (GEMMs) have become integral in identifying and evaluating unique pathways that may be exploited as therapeutic targets. Assessment of GEMM tumor burden on histopathological sections performed by manual inspection is both time consuming and prone to subjective bias. Therefore, an interplay of needs and challenges exists for computer aided detection tools, for the accurate and efficient analysis of these histopathology images. Our work demonstrates a simple machine learning approach called sparse principal component analysis (PCA) network, for automated detection of cancerous lesions on histological lung slides stained by hematoxylin and eosin (H&E). Methods: Our method comprises four steps: 1) cascaded sparse PCA; 2) graph-based PCA hashing; 3) block-wise histograms; and 4) support vector machine (SVM) classification. In our proposed architecture, sparse PCA is employed to learn the filter banks of the multiple stages. This is followed by a graph-based PCA hashing and block histograms for indexing and pooling. The meaningful features extracted from this sparse PCA are then fed to an SVM classifier. We tested the proposed sparse PCA network on H&E slides obtained from an inducible KrasG12D lung cancer mouse model. Our dataset consists of N = 21 whole slide histopathology lung images with 9 non-tumor bearing control mice and 12 mice with visible lung tumors. Tumor lesions from 12 lung images with visible tumors were visually identified by three trained individuals, which served as ground truth. The size of each image in our dataset is 2048 × 2048 pixels. Each image was divided into non-overlapping image patches of size 20 × 20 pixels consisting of a total of 12,361 cancer lesion patches and 207,839 non-cancer patches. We used 50% of the data for training and 50% of the data for testing our proposed sparse PCA network. We evaluated our algorithm using conventional metrics that have been used for evaluation of classification algorithms, namely precision (P), recall (R), and coverage measure (F-score). Results: The automatic cancer lesion detection results were compared with manually annotated ground truth. The proposed method achieves a cancer lesion detection accuracy of 97.98% with P = 0.8624, R = 0.9062 and F-score = 0.8790. The proposed method was found to take on average 17 minutes to train and learn a good representation for accurate and efficient classification of cancerous lesions within the images. Conclusion: We demonstrated a simple machine learning methodology for detection of cancerous lesions within histopathological lung images. Experimental results show that the proposed method is able to classify the regions of interest both efficiently and accurately. Future work will focus on feature extraction of individual tumors and tumor location within lungs. Citation Format: Sundaresh Ram, Wenfei Tang, Alexander J. Bell, Cara Spencer, Alexander Buschhuas, Charles R. Hatt, Marina P. di Magliano, Stefanie Galban, Craig J. Galban. Detection of cancer lesions in histopathological lung images using a sparse PCA network [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-086.

The aim of this study is to determine the effects of early intravenous (IV) infusion later followed by transendocardial (TE) injection of allogeneic mesenchymal stem cells (MSCs) following myocardial infarction (MI). Twenty-four swine underwent balloon occlusion reperfusion MI and were randomized into 4 groups: IV MSC (or placebo) infusion (post-MI day 2) and TE MSC (or placebo) injection targeting the infarct border with 2D X-ray fluoroscopy fused to 3D magnetic resonance (XFM) co-registration (post-MI day 14). Continuous ECG recording, MRI, and invasive pressure-volume analyses were performed. IV MSC plus TE MSC treated group was superior to other groups for contractility reserve (p = 0.02) and freedom from VT (p = 0.03) but had more lymphocytic foci localized to the peri-infarct region (p = 0.002). No differences were observed in post-MI remodeling parameters. IV followed by XFM targeted TE MSC therapy improves contractility reserve and suppresses VT but does not affect post-MI remodeling and may cause an immune response.

Lung cancer is a leading cause of mortality and morbidity for patients suffering from Chronic Obstructive Pulmonary Disease (COPD). Both the presence of visually assessed emphysema on CT scans and abnormal pulmonary function tests are associated with the development of lung cancer. Based on recent results showing that convolutional neural networks (CNNs) applied to CT scans can predict spirometrically-defined COPD (\(\frac{FEV_{1}}{FVC}

Small airways disease (SAD) is one of the leading causes of airflow limitations in patients diagnosed with chronic obstructive pulmonary disease (COPD). Parametric response mapping (PRM) of computed tomography (CT) scans allows for the quantification of this previously invisible COPD component. Although PRM is being investigated as a diagnostic tool for COPD, variability in the longitudinal measurements of SAD by PRM has been reported. Here, we show a method for correcting longitudinal PRM data because of nonpathological variations in serial CT scans. In this study, serial whole-lung high-resolution CT scans over a 30-day interval were obtained from 90 subjects with and without COPD accrued as part of SPIROMICS. It was assumed in all subjects that the COPD did not progress between examinations. CT scans were acquired at inspiration and expiration, spatially aligned to a single geometric frame, and analyzed using PRM. By modeling variability in longitudinal CT scans, our method could identify, at the voxel-level, shifts in PRM classification over the 30-day interval. In the absence of any correction, PRM generated serial percent volumes of functional SAD with differences as high as 15%. Applying the correction strategy significantly mitigated this effect with differences ∼1%. At the voxel-level, significant differences were found between baseline PRM classifications and the follow-up map computed with and without correction (P &lt, 01 over GOLD). This strategy of accounting for nonpathological sources of variability in longitudinal PRM may improve the quantification of COPD phenotypes transitioning with disease progression.

Background: Upper lobe-predominant is associated with poor outcomes. It is not clear if greater emphysema in central (core) versus outer portions of the lung (rind) is associated with differences in symptoms. Methods: We analyzed current/former smokers with airflow obstruction (FEV1/FVC 1 identifying core-predominance. We tested associations of core-predominant emphysema (higher quartile of the core:rind distribution) with respiratory symptoms (dyspnea score ≥2, 6-minute walk distance and Saint George’s Respiratory Questionnaire [SGRQ]), adjusting for age, sex, FEV1% predicted, smoking history, smoking status and exacerbation history. Results: We analyzed 758 COPD subjects (mean age 63.4 years, 44% females, 37% current smokers, mean FEV1% predicted 60.8%, spirometry stage GOLD 3/4 33.7%), who had average 15.5% emphysema in the core and 8.7% in the rind. In adjusted models, having core-predominant emphysema (higher quartile) was associated with 1.7 higher odds of mMRC score ≥2 (95% CI 1.2, 2.4), higher SGRQ total (Beta 3.8; 95% CI 1.1, 6.5), higher SGRQ activity (Beta 7.1; 95% CI3.5, 10.6) and shorter walk distance (Beta -52; 95% CI -100, -1.0). Conclusion: PRM-defined core-predominant emphysema is associated with more respiratory-related limitations.

Purpose Transcatheter aortic valve replacement (TAVR) is a minimally invasive procedure in which a prosthetic heart valve is placed and expanded within a defective aortic valve. The device placement is commonly performed using two-dimensional (2D) fluoroscopic imaging. Within this work, we propose a novel technique to track the motion and deformation of the prosthetic valve in three dimensions based on biplane fluoroscopic image sequences. Methods The tracking approach uses a parameterized point cloud model of the valve stent which can undergo rigid three-dimensional (3D) transformation and different modes of expansion. Rigid elements of the model are individually rotated and translated in three dimensions to approximate the motions of the stent. Tracking is performed using an iterative 2D-3D registration procedure which estimates the model parameters by minimizing the mean-squared image values at the positions of the forward-projected model points. Additionally, an initialization technique is proposed, which locates clusters of salient features to determine the initial position and orientation of the model. Results The proposed algorithms were evaluated based on simulations using a digital 4D CT phantom as well as experimentally acquired images of a prosthetic valve inside a chest phantom with anatomical background features. The target registration error was 0.12 ± 0.04 mm in the simulations and 0.64 ± 0.09 mm in the experimental data. Conclusions The proposed algorithm could be used to generate 3D visualization of the prosthetic valve from two projections. In combination with soft-tissue sensitive-imaging techniques like transesophageal echocardiography, this technique could enable 3D image guidance during TAVR procedures.

Myocardial infarction (MI) is one of the leading causes of death in the world. Small animal studies have shown that stem-cell therapy offers dramatic functional improvement post-MI. An endomyocardial catheter injection approach to therapeutic agent delivery has been proposed to improve efficacy through increased cell retention. Accurate targeting is critical for reaching areas of greatest therapeutic potential while avoiding a life-threatening myocardial perforation. Multimodal image fusion has been proposed as a way to improve these procedures by augmenting traditional intra-operative imaging modalities with high resolution pre-procedural images. Previous approaches have suffered from a lack of real-time tissue imaging and dependence on X-ray imaging to track devices, leading to increased ionizing radiation dose. In this paper, we present a new image fusion system for catheter-based targeted delivery of therapeutic agents. The system registers real-time 3D echocardiography, magnetic resonance, X-ray, and electromagnetic sensor tracking within a single flexible framework. All system calibrations and registrations were validated and found to have target registration errors less than 5 mm in the worst case. Injection accuracy was validated in a motion enabled cardiac injection phantom, where targeting accuracy ranged from 0.57 to 3.81 mm. Clinical feasibility was demonstrated with in-vivo swine experiments, where injections were successfully made into targeted regions of the heart.

Accurate and artifact-free reconstruction of tomographic images requires precise knowledge of the imaging system geometry. A projection matrix-based calibration method to enable C-arm inverse geometry CT (IGCT) is proposed. The method is evaluated for scanning-beam digital x-ray (SBDX), a C-arm mounted inverse geometry fluoroscopic technology. A helical configuration of fiducials is imaged at each gantry angle in a rotational acquisition. For each gantry angle, digital tomosynthesis is performed at multiple planes and a composite image analogous to a cone-beam projection is generated from the plane stack. The geometry of the C-arm, source array, and detector array is determined at each angle by constructing a parameterized three-dimensional-to-two-dimensional projection matrix that minimizes the sum-of-squared deviations between measured and projected fiducial coordinates. Simulations were used to evaluate calibration performance with translations and rotations of the source and detector. The relative root-mean-square error in a reconstruction of a numerical thorax phantom was 0.4% using the calibration method versus 7.7% without calibration. In phantom studies, reconstruction of SBDX projections using the proposed method eliminated artifacts present in noncalibrated reconstructions. The proposed IGCT calibration method reduces image artifacts when uncertainties exist in system geometry.

In recent years, registration between x-ray fluoroscopy (XRF) and transesophageal echocardiography (TEE) has been rapidly developed, validated, and translated to the clinic as a tool for advanced image guidance of structural heart interventions. This technology relies on accurate pose-estimation of the TEE probe via standard 2D/3D registration methods. It has been shown that latencies caused by slow registrations can result in errors during untracked frames, and a real-time ( > 15 hz) tracking algorithm is needed to minimize these errors. This paper presents two novel similarity metrics designed for accurate, robust, and extremely fast pose-estimation of devices from XRF images: Direct Splat Correlation (DSC) and Patch Gradient Correlation (PGC). Both metrics were implemented in CUDA C, and validated on simulated and clinical datasets against prior methods presented in the literature. It was shown that by combining DSC and PGC in a hybrid method (HYB), target registration errors comparable to previously reported methods were achieved, but at much higher speeds and lower failure rates. In simulated datasets, the proposed HYB method achieved a median projected target registration error (pTRE) of 0.33 mm and a mean registration frame-rate of 12.1 hz, while previously published methods produced median pTREs greater than 1.5 mm and mean registration frame-rates less than 4 hz. In clinical datasets, the HYB method achieved a median pTRE of 1.1 mm and a mean registration frame-rate of 20.5 hz, while previously published methods produced median pTREs greater than 1.3 mm and mean registration frame-rates less than 12 hz. The proposed hybrid method also had much lower failure rates than previously published methods.

Accurate and artifact free reconstruction of tomographic images requires precise knowledge of the imaging system geometry. This work proposes a novel projection matrix (P-matrix) based calibration method to enable C-arm inverse geometry CT (IGCT). The method is evaluated for scanning-beam digital x-ray (SBDX), a C-arm mounted inverse geometry fluoroscopic technology. A helical configuration of fiducials is imaged at each gantry angle in a rotational acquisition. For each gantry angle, digital tomosynthesis is performed at multiple planes and a composite image analogous to a cone-beam projection is generated from the plane stack. The geometry of the C-arm, source array, and detector array is determined at each angle by constructing a parameterized 3D-to-2D projection matrix that minimizes the sum-of-squared deviations between measured and projected fiducial coordinates. Simulations were used to evaluate calibration performance with translations and rotations of the source and detector. In a geometry with 1 mm translation of the central ray relative to the axis-of-rotation and 1 degree yaw of the detector and source arrays, the maximum error in the recovered translational parameters was 0.4 mm and maximum error in the rotation parameter was 0.02 degrees. The relative rootmean- square error in a reconstruction of a numerical thorax phantom was 0.4% using the calibration method, versus 7.7% without calibration. Changes in source-detector-distance were the most challenging to estimate. Reconstruction of experimental SBDX data using the proposed method eliminated double contour artifacts present in a non-calibrated reconstruction. The proposed IGCT geometric calibration method reduces image artifacts when uncertainties exist in system geometry.

A method for real-time localization of devices in fluoroscopic images is presented. Device pose is estimated using a Hough forest based detection framework. The method was applied to two types of devices used for transcatheter aortic valve replacement: a transesophageal echo TEE probe and prosthetic valve PV. Validation was performed on clinical datasets, where both the TEE probe and PV were successfully detected in 95.8% and 90.1% of images, respectively. TEE probe position and orientation errors were 1.42 ± 0.79 mm and 2.59i¾? ± 1.87i¾?, while PV position and orientation errors were 1.04 ± 0.77 mm and 2.90i¾? ± 2.37i¾?. The Hough forest was implemented in CUDA C, and was able to generate device location hypotheses in less than 50 ms for all experiments.

X-ray fluoroscopy (XRF) is considered the primary imaging modality for guidance during cardiac interventions. However, visualization of soft tissue is limited. Image registration between transthoracic echocardiography (TTE) and XRF could potentially be used to fuse soft-tissue imaging with device imaging in minimally sedated patients. We propose a method of XRF/TTE image registration in which a fiducial apparatus is rigidly attached to the TTE probe, and the 3D pose of the fiducial is determined based on its appearance in the XRF image. The fiducial consists of a constellation of radio-opaque markers with known geometry. 3D pose estimation accuracy depends on 1) accurate marker localization in the XRF image, 2) geometric properties of the apparatus such as marker diameter, overall apparatus dimensions, marker count, manufacturing tolerances, and 3) pose estimation method. In this work, we performed simulation studies to examine the effect that each variable had on the 3D target registration error (TRE). In the first experiment, we generated simulated XRF images of a spherical marker with varying levels of peak signal-difference-to-noise ratio (pSDNR) and varying marker diameters, and measured the effect on marker localization error (MLE). In the second experiment, using a circular con- stellation of markers, we tested the effect that MLE, manufacturing tolerance, apparatus dimensions, marker count, and shape constraints during optimization had on TRE. Our results showed that MLE was roughly 0.14, 0.05, 0.02 mm for pSDNRs of 10, 20 and 30 using a 2 mm diameter spherical marker. TRE decreased as the marker count and apparatus diameter increased. Manufacturing error increased TRE, but had minor effects for clinically relevant levels of MLE. Results indicate that, for a “medium” level of pSDNR ( 15), the optimal fiducial configuration can achieve a 3DTRE of 0.32 ± 0.22 mm, and can achieve a 3DTRE of less than 1.5 mm for noisier images.

We present a novel 2D/ 3D registration algorithm for fusion between transesophageal echocardiography (TEE) and X-ray fluoroscopy (XRF). The TEE probe is modeled as a subset of 3D gradient and intensity point features, which facilitates efficient 3D-to-2D perspective projection. A novel cost-function, based on a combination of intensity and edge features, evaluates the registration cost value without the need for time-consuming generation of digitally reconstructed radiographs (DRRs). Validation experiments were performed with simulations and phantom data. For simulations, in silica XRF images of a TEE probe were generated in a number of different pose configurations using a previously acquired CT image. Random misregistrations were applied and our method was used to recover the TEE probe pose and compare the result to the ground truth. Phantom experiments were performed by attaching fiducial markers externally to a TEE probe, imaging the probe with an interventional cardiac angiographic x-ray system, and comparing the pose estimated from the external markers to that estimated from the TEE probe using our algorithm. Simulations found a 3D target registration error of 1.08(1.92) mm for biplane (monoplane) geometries, while the phantom experiment found a 2D target registration error of 0.69mm. For phantom experiments, we demonstrated a monoplane tracking frame-rate of 1.38 fps. The proposed feature-based registration method is computationally efficient, resulting in near real-time, accurate image based registration between TEE and XRF.

Segmentation of needles in ultrasound images remains a challenging problem. In this paper, we introduce a machine learning-based method for needle segmentation in 2D beam-steered ultrasound images. We used a statistical boosting approach to train a pixel-wise classifier for needle segmentation. The Radon transform was then used to find the needle position and orientation from the segmented image. We validated our method with data from ex vivo specimens and clinical nerve block procedures, and compared the results to those obtained using previously reported needle segmentation methods. Results show improved localization success and accuracy using the proposed method. For the ex vivo datasets, assuming that the needle orientation was known a priori, the needle was successfully localized in 86.2% of the images, with a mean targeting error of 0.48mm. The robustness of the proposed method to a lack of a priori knowledge of needle orientation was also demonstrated. For the clinical datasets, assuming that the needle orientation was closely aligned with the beam steering angle selected by the physician, the needle was successfully localized in 99.8% of the images, with a mean targeting error 0.19mm. These results indicate that the learning-based segmentation method may allow for increased targeting accuracy and enhanced visualization during ultrasound-guided needle procedures.

This paper describes a clinically translatable interventional guidance platform to improve the accuracy and precision of stem cell injections into a beating heart. The proposed platform overlays live position of an injection catheter onto a fusion of a pre-procedural MR roadmap with real-time 3D transesophageal echocardiography (TEE). Electromagnetic (EM) tracking is used to initialize the fusion. The fusion is intra-operatively compensated for respiratory motion using a novel algorithm that uses peri-operative full volume ultrasound images. Validation of the system on a moving heart phantom produced a landmark registration accuracy of 2.8 ± 1.45mm. Validation on animal in vivo data produced an average registration accuracy of 2.2 ± 1.8 mm; indicating that it is feasible to reliably and robustly fuse the MR road-map with catheter position using 3D ultrasound in a clinical setting.