Joyce D. Schroeder

BACKGROUND: Although chronic obstructive pulmonary disease (COPD) and heart failure with preserved ejection fraction often coexist with overlapping clinical features, they are usually studied separately. The SPIROMICS HF (Subpopulations and Intermediate Outcome Measures in COPD and Heart Failure Study) is testing hypotheses that new computed tomography emphysema subtypes are associated with specific cardiovascular phenotypes (eg, cor pulmonale , cor pulmonale parvus ), common airway branch variants are associated with right heart dysfunction, and symptomatic tobacco-exposed persons with preserved spirometry have signs of increased left ventricular afterload. METHODS: SPIROMICS is a multicenter observational study of COPD with extensive pulmonary phenotyping of participants with ≥20 pack-years smoking and nonsmoking controls. COPD and COPD severity were defined by standard spirometric criteria and symptomatic tobacco-exposed persons with preserved spirometry by ≥20 pack-years, normal spirometry, and COPD Assessment Test score >10. SPIROMICS HF selected all participants in SPIROMICS visit 5 at major sites. Its comprehensive speckle-tracking echocardiography, which included physiological perturbations of leg raise and low-intensity exercise, was harmonized prospectively with the Multi-Ethnic Study of Atherosclerosis Early Heart Failure and HeartSHARE (Combining Omics, Deep Phenotyping, and Electronic Health Records for Heart Failure Subtypes and Treatment Targets) studies. The cardiopulmonary magnetic resonance imaging protocol with gadolinium administration included myocardial fibrosis sequences, pulmonary angiography, time-resolved 3-dimensional cine magnetic resonance imaging (4-dimensional flow) of venous return, and metronome-paced tachypnea to induce dynamic hyperinflation. Coronary artery calcium was assessed on computed tomography scans. The Kansas City Cardiomyopathy Questionnaire was administered. RESULTS: Of the final sample of 753 participants, 57% had COPD (15% mild, 27% moderate, and 15% severe), 18% had symptomatic tobacco-exposed persons with preserved spirometry, 16% were smoking controls, and 8% were nonsmoking controls. Reproducibility of the main measures from speckle-tracking echocardiography (intraclass correlation coefficient, 0.83–0.99), exercise echocardiography (intraclass correlation coefficient, 0.71–0.99) and magnetic resonance imaging (intraclass correlation coefficient, 0.57–0.99) were good-to-excellent, including in severe COPD. CONCLUSIONS: SPIROMICS HF aims to characterize and understand cardiopulmonary interactions in COPD and COPD-related phenotypes to inform targeted treatments for combined cardiopulmonary failure.

Metronome-paced tachypnea during real-time cine pulmonary MRI is a newly proposed technique to measure dynamic hyperinflation in chronic obstructive pulmonary disease, which yielded reproducible results.

Adversarial training, especially projected gradient descent (PGD), has proven to be a successful approach for improving robustness against adversarial attacks. After adversarial training, gradients of models with respect to their inputs have a preferential direction. However, the direction of alignment is not mathematically well established, making it difficult to evaluate quantitatively. We propose a novel definition of this direction as the direction of the vector pointing toward the closest point of the support of the closest inaccurate class in decision space. To evaluate the alignment with this direction after adversarial training, we apply a metric that uses generative adversarial networks to produce the smallest residual needed to change the class present in the image. We show that PGD-trained models have a higher alignment than the baseline according to our definition, that our metric presents higher alignment values than a competing metric formulation, and that enforcing this alignment increases the robustness of models.

Deep learning has shown recent success in classifying anomalies in chest x-rays, but datasets are still small compared to natural image datasets. Supervision of abnormality localization has been shown to improve trained models, partially compensating for dataset sizes. However, explicitly labeling these anomalies requires an expert and is very time-consuming. We propose a potentially scalable method for collecting implicit localization data using an eye tracker to capture gaze locations and a microphone to capture a dictation of a report, imitating the setup of a reading room. The resulting REFLACX (Reports and Eye-Tracking Data for Localization of Abnormalities in Chest X-rays) dataset was labeled across five radiologists and contains 3,032 synchronized sets of eye-tracking data and timestamped report transcriptions for 2,616 chest x-rays from the MIMIC-CXR dataset. We also provide auxiliary annotations, including bounding boxes around lungs and heart and validation labels consisting of ellipses localizing abnormalities and image-level labels. Furthermore, a small subset of the data contains readings from all radiologists, allowing for the calculation of inter-rater scores.

Objectives: Magnetic resonance angiography (MRA) has been established as an important imaging method in cardiac ablation procedures. In pulmonary vein (PV) isolation procedures, MRA has the potential to minimize the risk of severe complications, such as atrio-esophageal fistula, by providing detailed information on esophageal position relatively to cardiac structures. However, traditional non-gated, first-pass (FP) MRA approaches have several limitations, such as long breath-holds, non-uniform signal intensity throughout the left atrium (LA), and poor esophageal visualization. The aim of this observational study was to validate a respiratory-navigated, ECG-gated (EC), saturation recovery-prepared MRA technique for simultaneous imaging of LA, LA appendage, PVs, esophagus, and adjacent anatomical structures. Methods: Before PVI, 106 consecutive patients with a history of AF underwent either conventional FP-MRA (n = 53 patients) or our new EC-MRA (n = 53 patients). Five quality scores (QS) of LA and esophagus visibility were assessed by two experienced readers. The non-parametric Mann–Whitney U-test was used to compare QS between FP-MRA and EC-MRA groups, and linear regression was applied to assess clinical contributors to image quality. Results: EC-MRA demonstrated significantly better image quality than FP-MRA in every quality category. Esophageal visibility using the new MRA technique was markedly better than with the conventional FP-MRA technique (median 3.5 [IQR 1] vs median 1.0, p < 0.001). In contrast to FP-MRA, overall image quality of EC-MRA was not influenced by heart rate. Conclusion: Our ECG-gated, respiratory-navigated, saturation recovery-prepared MRA technique provides significantly better image quality and esophageal visibility than the established non-gated, breath-holding FP-MRA. Image quality of EC-MRA technique has the additional advantage of being unaffected by heart rate. Advances in knowledge: Detailed information of cardiac anatomy has the potential to minimize the risk of severe complications and improve success rates in invasive electrophysiological studies. Our novel ECG-gated, respiratory-navigated, saturation recovery-prepared MRA technique provides significantly better image quality of LA and esophageal structures than the traditional first-pass algorithm. This new MRA technique is robust to arrhythmia (tachycardic, irregular heart rates) frequently observed in AF patients.

Background: Chronic obstructive pulmonary disease (COPD), the third leading cause of death worldwide, is often underdiagnosed. Purpose: To develop machine learning methods to predict COPD using chest radiographs and a convolutional neural network (CNN) trained with near-concurrent pulmonary function test (PFT) data. Comparison is made to natural language processing (NLP) of the associated radiologist text reports. Materials and Methods: This IRB-approved single-institution retrospective study uses 6749 two-view chest radiograph exams (2012– 2017, 4436 unique subjects, 54% female, 46% male), same-day associated radiologist text reports, and PFT exams acquired within 180 days. The Image Model (Resnet18 pre-trained with ImageNet CNN) is trained using frontal and lateral radiographs and PFTs with 10% of the subjects for validation and 19% for testing. The NLP Model is trained using radiologist text reports and PFTs. The primary metric of model comparison is the area under the receiver operating characteristic curve (AUC). Results: The Image Model achieves an AUC of 0.814 for prediction of obstructive lung disease (FEV1/FVC < 0.7) from chest radiographs and performs better than the NLP Model (AUC 0.704, p< 0.001) from radiologist text reports where FEV1 = forced expiratory volume in 1 second and FVC = forced vital capacity. The Image Model performs better for prediction of severe or very severe COPD (FEV1 < 0.5) with an AUC of 0.837 versus the NLP model AUC of 0.770 (p< 0.001). Conclusion: A CNN Image Model trained on physiologic lung function data (PFTs) can be applied to chest radiographs for quantitative prediction of obstructive lung disease with good accuracy.