CT imaging determinants of persistent hypoxemia in acute intermediate-risk pulmonary embolism.

The factors associated with persistent hypoxemia after pulmonary embolus (PE) are not well understood. Predicting the need for oxygen post discharge at the time of diagnosis using available CT imaging will enable better discharge planning. To examine the relationship between CT derived imaging markers (automated computation of arterial small vessel fraction, pulmonary artery diameter to aortic diameter ratio (PA:A), right to left ventricular diameter ratio (RV:LV) and new oxygen requirement at the time of discharge in patients diagnosed with acute intermediate-risk PE. CT measurements were obtained in a retrospective cohort of patients with acute-intermediate risk PE admitted to Brigham and Women's Hospital between 2009 and 2017. Twenty one patients without a history of lung disease requiring home oxygen and 682 patients without discharge oxygen requirements were identified. There was an increased median PA:A ratio (0.98 vs. 0.92, p = 0.02) and arterial small vessel fraction (0.32 vs. 0.39, p = 0.001) in the oxygen-requiring group], but no difference in the median RV:LV ratio (1.20 vs. 1.20, p = 0.74). Being in the upper quantile for the arterial small vessel fraction was associated with decreased odds of oxygen requirement (OR 0.30 [0.10-0.78], p = 0.02). Loss of arterial small vessel volume as measured by arterial small vessel fraction and an increase in the PA:A ratio at the time of diagnosis were associated with the presence of persistent hypoxemia on discharge in acute intermediate-risk PE.

CNNs trained with adult data are useful in pediatrics. A pneumonia classification example.

BACKGROUND AND OBJECTIVES: The scarcity of data for training deep learning models in pediatrics has prompted questions about the feasibility of employing CNNs trained with adult images for pediatric populations. In this work, a pneumonia classification CNN was used as an exploratory example to showcase the adaptability and efficacy of such models in pediatric healthcare settings despite the inherent data constraints. METHODS: To develop a curated training dataset with reduced biases, 46,947 chest X-ray images from various adult datasets were meticulously selected. Two preprocessing approaches were tried to assess the impact of thoracic segmentation on model attention outside the thoracic area. Evaluation of our approach was carried out on a dataset containing 5,856 chest X-rays of children from 1 to 5 years old. RESULTS: An analysis of attention maps indicated that networks trained with thorax segmentation placed less attention on regions outside the thorax, thus eliminating potential bias. The ensuing network exhibited impressive performance when evaluated on an adult dataset, achieving a pneumonia discrimination AUC of 0.95. When tested on a pediatric dataset, the pneumonia discrimination AUC reached 0.82. CONCLUSIONS: The results of this study show that adult-trained CNNs can be effectively applied to pediatric populations. This could potentially shift focus towards validating adult models over pediatric population instead of training new CNNs with limited pediatric data. To ensure the generalizability of deep learning models, it is important to implement techniques aimed at minimizing biases, such as image segmentation or low-quality image exclusion.

COPD: pulmonary vascular volume associated with cardiac structure and function.

BACKGROUND: Early recognition of cardiac dysfunction in patients with chronic obstructive pulmonary disease (COPD) may prevent future cardiac impairment and improve prognosis. Quantitative assessment of subsegmental and segmental vessel volume by Computed Tomographic (CT) imaging can provide a surrogate of pulmonary vascular remodeling. We aimed to examine the relationship between lung segmental- and subsegmental vessel volume, and echocardiographic measures of cardiac structure and function in patients with COPD. METHODS: We studied 205 participants with COPD, included in a large cohort study of cardiovascular disease in COPD patients. Participants had an available CT scan and echocardiogram. Artificial intelligence (AI) algorithms calculated the subsegmental vessel fraction as the vascular volume in vessels below 10 mm2 in cross-sectional area, indexed to total intrapulmonary vessel volume. Linear regressions were conducted, and standardized ß-coefficients were calculated. Scatterplots were created to visualize the continuous correlations between the vessel fractions and echocardiographic parameters. RESULTS: We found that lower subsegmental vessel fraction and higher segmental vessel volume were correlated with higher left ventricular (LV) mass, LV diastolic dysfunction, and inferior vena cava (IVC) dilatation. Subsegmental vessel fraction was correlated with right ventricular (RV) remodeling, while segmental vessel fraction was correlated with higher pulmonary pressure. Measures of LV mass and right atrial pressure displayed the strongest correlations with pulmonary vasculature measures. CONCLUSION: Pulmonary vascular remodeling in patients with COPD, may negatively affect cardiac structure and function. AI-identified remodeling in pulmonary vasculature may provide a tool for early identification of COPD patients at higher risk for cardiac impairment.

Silent Airway Mucus Plugs in COPD and Clinical Implications.

BACKGROUND: Airway mucus plugs are frequently identified on CT scans of patients with COPD with a smoking history without mucus-related symptoms (ie, cough, phlegm [silent mucus plugs]). RESEARCH QUESTION: In patients with COPD, what are the risk and protective factors associated with silent airway mucus plugs? Are silent mucus plugs associated with functional, structural, and clinical measures of disease? STUDY DESIGN AND METHODS: We identified mucus plugs on chest CT scans of participants with COPD from the COPDGene study. The mucus plug score was defined as the number of pulmonary segments with mucus plugs, ranging from 0 to 18, and categorized into three groups (0, 1-2, and ≥ 3). We determined risk and protective factors for silent mucus plugs and the associations of silent mucus plugs with measures of disease severity using multivariable linear and logistic regression models. RESULTS: Of 4,363 participants with COPD, 1,739 had no cough or phlegm. Among the 1,739 participants, 627 (36%) had airway mucus plugs identified on CT scan. Risk factors of silent mucus plugs (compared with symptomatic mucus plugs) were older age (OR, 1.02), female sex (OR, 1.40), and Black race (OR, 1.93) (all P values < .01). Among those without cough or phlegm, silent mucus plugs (vs absence of mucus plugs) were associated with worse 6-min walk distance, worse resting arterial oxygen saturation, worse FEV1 % predicted, greater emphysema, thicker airway walls, and higher odds of severe exacerbation in the past year in adjusted models. INTERPRETATION: Mucus plugs are common in patients with COPD without mucus-related symptoms. Silent mucus plugs are associated with worse functional, structural, and clinical measures of disease. CT scan-identified mucus plugs can complement the evaluation of patients with COPD.

Functional-Consistent CycleGAN for CT to Iodine Perfusion Map Translation.

Image-to-image translation from a source to a target domain by means of generative adversarial neural network (GAN) has gained a lot of attention in the medical imaging field due to their capability to learn the mapping characteristics between different modalities. CycleGAN has been proposed for image-to-image translation with unpaired images by means of a cycle-consistency loss function, which is optimized to reduce the difference between the image reconstructed from the synthetically-generated domain and the original input. However, CycleGAN inherently implies that the mapping between both domains is invertible, i.e., given a mapping G (forward cycle) from domain A to B, there is a mapping F (backward cycle) that is the inverse of G. This is assumption is not always true. For example, when we want to learn functional activity from structural modalities. Although it is well-recognized the relation between structure and function in different physiological processes, the problem is not invertible as the original modality cannot be recovered from a given functional response. In this paper, we propose a functional-consistent CycleGAN that leverages the usage of a proxy structural image in a third domain, shared between source and target, to help the network learn fundamental characteristics while being cycle consistent. To demonstrate the strength of the proposed strategy, we present the application of our method to estimate iodine perfusion maps from contrast CT scans, and we compare the performance of this technique to a traditional CycleGAN framework.