Clinical usability of 3D gradient-echo-based ultrashort echo time imaging: Is it enough to facilitate diagnostic decision in real-world practice?

BACKGROUND: With recent advances in magnetic resonance imaging (MRI) technology, the practical role of lung MRI is expanding despite the inherent challenges of the thorax. The purpose of our study was to evaluate the current status of the concurrent dephasing and excitation (CODE) ultrashort echo-time sequence and the T1-weighted volumetric interpolated breath-hold examination (VIBE) sequence in the evaluation of thoracic disease by comparing it with the gold standard computed tomography (CT). METHODS: Twenty-four patients with lung cancer and mediastinal masses underwent both CT and MRI including T1-weighted VIBE and CODE. For CODE images, data were acquired in free breathing and end-expiratory images were reconstructed using retrospective respiratory gating. All images were evaluated through qualitative and quantitative approaches regarding various anatomical structures and lesions (nodule, mediastinal mass, emphysema, reticulation, honeycombing, bronchiectasis, pleural plaque and lymphadenopathy) inside the thorax in terms of diagnostic performance in making specific decisions. RESULTS: Depiction of the lung parenchyma, mediastinal and pleural lesion was not significant different among the three modalities (p > 0.05). Intra-tumoral and peritumoral features of lung nodules were not significant different in the CT, VIBE or CODE images (p > 0.05). However, VIBE and CODE had significantly lower image quality and poorer depiction of airway, great vessels, and emphysema compared to CT (p < 0.05). Image quality of central airways and depiction of bronchi were significantly better in CODE than in VIBE (p < 0.001 and p = 0.005). In contrast, the depiction of the vasculature was better for VIBE than CODE images (p = 0.003). The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were significant greater in VIBE than CODE except for SNRlung and SNRnodule (p < 0.05). CONCLUSIONS: Our study showed the potential of CODE and VIBE sequences in the evaluation of localized thoracic abnormalities including solid pulmonary nodules.

Predictions of PD-L1 Expression Based on CT Imaging Features in Lung Squamous Cell Carcinoma.

Purpose: To develop models to predict programmed death ligand 1 (PD-L1) expression in pulmonary squamous cell carcinoma (SCC) using CT. Materials and Methods: A total of 97 patients diagnosed with SCC who underwent PD-L1 expression assay were included in this study. We performed a CT analysis of the tumors using pretreatment CT images. Multiple logistic regression models were constructed to predict PD-L1 positivity in the total patient group and in the 40 advanced-stage (≥ stage IIIB) patients. The area under the receiver operating characteristic curve (AUC) was calculated for each model. Results: For the total patient group, the AUC of the 'total significant features model' (tumor stage, tumor size, pleural nodularity, and lung metastasis) was 0.652, and that of the 'selected feature model' (pleural nodularity) was 0.556. For advanced-stage patients, the AUC of the 'selected feature model' (tumor size, pleural nodularity, pulmonary oligometastases, and absence of interstitial lung disease) was 0.897. Among these factors, pleural nodularity and pulmonary oligometastases had the highest odds ratios (8.78 and 16.35, respectively). Conclusion: Our model could predict PD-L1 expression in patients with lung SCC, and pleural nodularity and pulmonary oligometastases were notable predictive CT features of PD-L1.

CT-Derived Deep Learning-Based Quantification of Body Composition Associated with Disease Severity in Chronic Obstructive Pulmonary Disease.

Purpose: Our study aimed to evaluate the association between automated quantified body composition on CT and pulmonary function or quantitative lung features in patients with chronic obstructive pulmonary disease (COPD). Materials and Methods: A total of 290 patients with COPD were enrolled in this study. The volume of muscle and subcutaneous fat, area of muscle and subcutaneous fat at T12, and bone attenuation at T12 were obtained from chest CT using a deep learning-based body segmentation algorithm. Parametric response mapping-derived emphysema (PRMemph), PRM-derived functional small airway disease (PRMfSAD), and airway wall thickness (AWT)-Pi10 were quantitatively assessed. The association between body composition and outcomes was evaluated using Pearson's correlation analysis. Results: The volume and area of muscle and subcutaneous fat were negatively associated with PRMemph and PRMfSAD (p < 0.05). Bone density at T12 was negatively associated with PRMemph (r = -0.1828, p = 0.002). The volume and area of subcutaneous fat and bone density at T12 were positively correlated with AWT-Pi10 (r = 0.1287, p = 0.030; r = 0.1668, p = 0.005; r = 0.1279, p = 0.031). However, muscle volume was negatively correlated with the AWT-Pi10 (r = -0.1966, p = 0.001). Muscle volume was significantly associated with pulmonary function (p < 0.001). Conclusion: Body composition, automatically assessed using chest CT, is associated with the phenotype and severity of COPD.

Correction: Deep Learning With Chest Radiographs for Making Prognoses in Patients With COVID-19: Retrospective Cohort Study.

[This corrects the article DOI: 10.2196/42717.].

Erratum: Correction of Affiliations in the Article "Establishment of a Nationwide Korean Imaging Cohort of Coronavirus Disease 2019".

This corrects the article on p. e413 in vol. 35, PMID: 33258333.

Deep Learning With Chest Radiographs for Making Prognoses in Patients With COVID-19: Retrospective Cohort Study.

BACKGROUND: An artificial intelligence (AI) model using chest radiography (CXR) may provide good performance in making prognoses for COVID-19. OBJECTIVE: We aimed to develop and validate a prediction model using CXR based on an AI model and clinical variables to predict clinical outcomes in patients with COVID-19. METHODS: This retrospective longitudinal study included patients hospitalized for COVID-19 at multiple COVID-19 medical centers between February 2020 and October 2020. Patients at Boramae Medical Center were randomly classified into training, validation, and internal testing sets (at a ratio of 8:1:1, respectively). An AI model using initial CXR images as input, a logistic regression model using clinical information, and a combined model using the output of the AI model (as CXR score) and clinical information were developed and trained to predict hospital length of stay (LOS) ≤2 weeks, need for oxygen supplementation, and acute respiratory distress syndrome (ARDS). The models were externally validated in the Korean Imaging Cohort of COVID-19 data set for discrimination and calibration. RESULTS: The AI model using CXR and the logistic regression model using clinical variables were suboptimal to predict hospital LOS ≤2 weeks or the need for oxygen supplementation but performed acceptably in the prediction of ARDS (AI model area under the curve [AUC] 0.782, 95% CI 0.720-0.845; logistic regression model AUC 0.878, 95% CI 0.838-0.919). The combined model performed better in predicting the need for oxygen supplementation (AUC 0.704, 95% CI 0.646-0.762) and ARDS (AUC 0.890, 95% CI 0.853-0.928) compared to the CXR score alone. Both the AI and combined models showed good calibration for predicting ARDS (P=.079 and P=.859). CONCLUSIONS: The combined prediction model, comprising the CXR score and clinical information, was externally validated as having acceptable performance in predicting severe illness and excellent performance in predicting ARDS in patients with COVID-19.

A Pictorial Review of Radiologic Findings of Foreign Bodies in the Thorax.

Thoracic foreign bodies (FBs) are serious and relatively frequent in emergency departments. Thoracic FBs may occur in association with aspiration, ingestion, trauma, or iatrogenic causes. Imaging plays an important role in the identification of FBs and their dimensions, structures, and locations, before the initiation of interventional treatment. To guide proper clinical management, radiologists should be aware of the radiologic presentations and the consequences of thoracic FBs. In this pictorial essay, we reviewed the optimal imaging settings to identify FBs in the thorax, classified thoracic FBs into four types according to their etiology, and reviewed the characteristic imaging features and the possible complications.

Genome-Wide Association Study of Airway Wall Thickening in a Korean Chronic Obstructive Pulmonary Disease Cohort.

Airway wall thickening (AWT) plays an important pathophysiological role in airway diseases such as chronic obstructive pulmonary disease (COPD). There are only a few studies on the genetic components contributing to AWT in the Korean population. This study aimed to identify AWT-related single-nucleotide polymorphisms (SNPs) using a genome-wide association study (GWAS). We performed GWAS for AWT using the CODA and KUCOPD cohorts. Thereafter, a meta-analysis was performed. Airway wall thickness was measured using automatic segmentation software. The AWT at an internal perimeter of 10 mm (AWT-Pi10) was calculated by the square root of the theoretical airway wall area using the full-width-half-maximum method. We identified a significant SNP (rs11648772, p = 1.41 × 10-8) located in LINC02127, near SALL1. This gene is involved in the inhibition of epithelial-mesenchymal transition in glial cells, and it affects bronchial wall depression in COPD patients. Additionally, we identified other SNPs (rs11970854, p = 1.92 × 10-6; rs16920168, p = 5.29 × 10-6) involved in airway inflammation and proliferation and found that AWT is influenced by these genetic variants. Our study helps identify the genetic cause of COPD in an Asian population and provides a potential basis for treatment.

How Many Private Data Are Needed for Deep Learning in Lung Nodule Detection on CT Scans? A Retrospective Multicenter Study.

Early detection of lung nodules is essential for preventing lung cancer. However, the number of radiologists who can diagnose lung nodules is limited, and considerable effort and time are required. To address this problem, researchers are investigating the automation of deep-learning-based lung nodule detection. However, deep learning requires large amounts of data, which can be difficult to collect. Therefore, data collection should be optimized to facilitate experiments at the beginning of lung nodule detection studies. We collected chest computed tomography scans from 515 patients with lung nodules from three hospitals and high-quality lung nodule annotations reviewed by radiologists. We conducted several experiments using the collected datasets and publicly available data from LUNA16. The object detection model, YOLOX was used in the lung nodule detection experiment. Similar or better performance was obtained when training the model with the collected data rather than LUNA16 with large amounts of data. We also show that weight transfer learning from pre-trained open data is very useful when it is difficult to collect large amounts of data. Good performance can otherwise be expected when reaching more than 100 patients. This study offers valuable insights for guiding data collection in lung nodules studies in the future.

Incidence and risk factors for sternal osteomyelitis after median sternotomy.

Background: Sternal osteomyelitis (OM) after median sternotomy is the rarest form of deep sternal wound infections (DSWIs). A retrospective study was implemented to evaluate the incidence and potential risk factors of sternal OM after median sternotomy. Methods: We analyzed 3,410 consecutive patients who underwent cardiothoracic surgery via median sternotomy from January 2005 to December 2019 at our institution. A sternal OM and control group without any sign of wound infections after median sternotomy were selected. Comparisons of the variables between the two groups were performed using the Student's t-test and Fisher's exact tests. The association of potential risk factors with sternal OM was tested by logistic regression analysis. Results: A total of 16 patients (0.47%) had sternal OM after median sternotomy. None of the variables were different between the sternal OM patients and the control group including body mass index (BMI), diabetes mellitus (DM), hypertension (HTN), left ventricle (LV) function, transfusion, operation time, cardiopulmonary bypass (CPB) time and intensive care unit and ventilator days. By univariate analysis, none of the variables were associated with an increased risk of sternal OM. Conclusions: The incidence of sternal OM after median sternotomy in our institution was 0.47% and there was no correlation between the known risk factors of DSWI and sternal OM in our study.

Structural and functional alterations of subjects with cement dust exposure: A longitudinal quantitative computed tomography-based study.

Cement dust exposure (CDE) can be a risk factor for pulmonary disease, causing changes in segmental airways and parenchymal lungs. This study investigates longitudinal alterations in quantitative computed tomography (CT)-based metrics due to CDE. We obtained CT-based airway structural and lung functional metrics from CDE subjects with baseline CT and follow-up CT scans performed three years later. From the CT, we extracted wall thickness (WT) and bifurcation angle (θ) at total lung capacity (TLC) and functional residual capacity (FRC), respectively. We also computed air volume (Vair), tissue volume (Vtissue), global lung shape, percentage of emphysema (Emph%), and more. Clinical measures were used to associate with CT-based metrics. Three years after their baseline, the pulmonary function tests of CDE subjects were similar or improved, but there were significant alterations in the CT-based structural and functional metrics. The follow-up CT scans showed changes in θ at most of the central airways; increased WT at the subgroup bronchi; smaller Vair at TLC at all except the right upper and lower lobes; smaller Vtissue at all lobes in TLC and FRC except for the upper lobes in FRC; smaller global lung shape; and greater Emph% at the right upper and lower lobes. CT-based structural and functional variables are more sensitive to the early identification of CDE subjects, while most clinical lung function changes were not noticeable. We speculate that the significant long-term changes in CT are uniquely observed in CDE subjects, different from smoking-induced structural changes.

Analysis of treatment pattern of anti-dementia medications in newly diagnosed Alzheimer's dementia using OMOP CDM.

Anti-dementia medications are widely prescribed to patients with Alzheimer's dementia (AD) in South Korea. This study investigated the pattern of medical management in newly diagnosed patients with AD using a standardized data format-the Observational Medical Outcome Partnership Common Data Model from five hospitals. We examined the anti-dementia treatment patterns from datasets that comprise > 5 million patients during 2009-2019. The medication utility information was analyzed with respect to treatment trends and persistence across 11 years. Among the 8653 patients with newly diagnosed AD, donepezil was the most commonly prescribed anti-dementia medication (4218; 48.75%), followed by memantine (1565; 18.09%), rivastigmine (1777; 8.98%), and galantamine (494; 5.71%). The rising prescription trend during observation period was found only with donepezil. The treatment pathways for the three cholinesterase inhibitors combined with N-methyl-D-aspartate receptor antagonist were different according to the drugs (19.6%; donepezil; 28.1%; rivastigmine, and 17.2%; galantamine). A 12-month persistence analysis showed values of approximately 50% for donepezil and memantine and approximately 40% for rivastigmine and galantamine. There were differences in the prescribing pattern and persistence among anti-dementia medications from database using the Observational Medical Outcome Partnership Common Data Model on the Federated E-health Big Data for Evidence Renovation Network platform in Korea.

Association between plasma sRAGE and emphysema according to the genotypes of AGER gene.

BACKGROUND: Higher soluble receptor for advanced glycation end product (sRAGE) levels are considered to be associated with severe emphysema. However, the relationship remains uncertain when the advanced glycation end-product specific receptor (AGER) gene is involved. We aimed to analyse the association between sRAGE levels and emphysema according to the genotypes of rs2070600 in the AGER gene. METHODS: We genotyped rs2070600 and measured the plasma concentration of sRAGE in each participant. Emphysema was quantified based on the chest computed tomography findings. We compared sRAGE levels based on the presence or absence and severity of emphysema in each genotype. Multiple logistic and linear regression models were used for the analyses. RESULTS: A total of 436 participants were included in the study. Among them, 64.2% had chronic obstructive pulmonary disease and 34.2% had emphysema. Among the CC-genotyped participants, the sRAGE level was significantly higher in participants without emphysema than in those with emphysema (P < 0.001). In addition, sRAGE levels were negatively correlated with emphysema severity in CC-genotyped patients (r = - 0.268 P < 0.001). Multiple regression analysis revealed that sRAGE was an independent protective factor for the presence of emphysema (adjusted odds ratio, 0.24; 95% confidence interval (CI) 0.11-0.51) and severity of emphysema (ß = - 3.28, 95% CI - 4.86 to - 1.70) in CC-genotyped participants. CONCLUSION: Plasma sRAGE might be a biomarker with a protective effect on emphysema among CC-genotyped patients of rs2070600 on the AGER gene. This is important in determining the target group for the future prediction and treatment of emphysema.

Quantitative assessment the longitudinal changes of pulmonary vascular counts in chronic obstructive pulmonary disease.

BACKGROUND: Chest computed tomography (CT) is a widely used method to assess morphological and dynamic abnormalities in chronic obstructive pulmonary disease (COPD). The small pulmonary vascular cross-section (CSA), quantitatively extracted from volumetric CT, is a reliable indicator for predicting pulmonary vascular changes. CSA is associated with the severity of symptoms, pulmonary function tests (PFT) and emphysema and in COPD patients the severity increases over time. We analyzed the correlation longitudinal changes in pulmonary vascular parameters with clinical parameters in COPD patients. MATERIALS AND METHODS: A total of 288 subjects with COPD were investigated during follow up period up to 6 years. CT images were classified into five subtypes from normal to severe emphysema according to percentage of low-attenuation areas less than -950 and -856 Hounsfield units (HU) on inspiratory and expiratory CT (LAA-950, LAA-856exp). Total number of vessels (Ntotal) and total number of vessels with area less than 5 mm2 (N<5 mm) per 1 cm2 of lung surface area (LSA) were measured at 6 mm from the pleural surface. RESULTS: Ntotal/LSA and N<5 mm/LSA changed from 1.16 ± 0.27 to 0.87 ± 0.2 and from 1.02 ± 0.22 to 0.78 ± 0.22, respectively, during Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage progression. Both parameters changed from normal to severe emphysema according to CT subtype from 1.39 ± 0.21 to 0.74 ± 0.17 and from 1.18 ± 0.19 to 0.67 ± 0.15, respectively. LAA-950 and LAA-856exp were negatively correlated with Ntotal/LSA (r = - 0.738, - 0.529) and N<5 mm /LSA (r = - 0.729, -- .497). On the other hand, pulmonary function test (PFT) results showed a weak correlation with Ntotal/LSA and N<5 mm/LSA (r = 0.205, 0.210). The depth in CT subtypes for longitudinal change both Ntotal/LSA and N<5 mm/LSA was (- 0.032, - 0.023) and (- 0.027) in normal and SAD, respectively. CONCLUSIONS: Quantitative computed tomography features faithfully reflected pulmonary vessel alterations, showing in particular that pulmonary vascular alteration started.

Quantitative computed tomography imaging-based classification of cement dust-exposed subjects with an artificial neural network technique.

BACKGROUND AND OBJECTIVE: Cement dust exposure is likely to affect the structural and functional alterations in segmental airways and parenchymal lungs. This study develops an artificial neural network (ANN) model for identifying cement dust-exposed (CDE) subjects using quantitative computed tomography-based airway structural and functional features. METHODS: We obtained the airway features in five central and five sub-grouped segmental regions and the lung features in five lobar regions and one total lung region from 311 CDE and 298 non-CDE (NCDE) subjects. The five-fold cross-validation method was adopted to train the following classification models:ANN, support vector machine (SVM), logistic regression (LR), and decision tree (DT). For all the classification models, linear discriminant analysis (LDA) and genetic algorithm (GA) were applied for dimensional reduction and hyperparameterization, respectively. The ANN model without LDA was also optimized by the GA method to observe the effect of the dimensional reduction. RESULTS: The genetically optimized ANN model without the LDA method was the best in terms of the classification accuracy. The accuracy, sensitivity, and specificity of the GA-ANN model with four layers were greater than those of the other classification models (i.e., ANN, SVM, LR, and DT using LDA and GA methods) in the five-fold cross-validation. The average values of accuracy, sensitivity, and specificity for the five-fold cross-validation were 97.0%, 98.7%, and 98.6%, respectively. CONCLUSIONS: We demonstrated herein that a quantitative computed tomography-based ANN model could more effectively detect CDE subjects when compared to their counterpart models. By employing the model, the CDE subjects may be identified early for therapeutic intervention.

Word Embedding Reveals Cyfra 21-1 as a Biomarker for Chronic Obstructive Pulmonary Disease.

BACKGROUND: Although patients with chronic obstructive pulmonary disease (COPD) experience high morbidity and mortality worldwide, few biomarkers are available for COPD. Here, we analyzed potential biomarkers for the diagnosis of COPD by using word embedding. METHODS: To determine which biomarkers are likely to be associated with COPD, we selected respiratory disease-related biomarkers. Degrees of similarity between the 26 selected biomarkers and COPD were measured by word embedding. And we infer the similarity with COPD through the word embedding model trained in the large-capacity medical corpus, and search for biomarkers with high similarity among them. We used Word2Vec, Canonical Correlation Analysis, and Global Vector for word embedding. We evaluated the associations of selected biomarkers with COPD parameters in a cohort of patients with COPD. RESULTS: Cytokeratin 19 fragment (Cyfra 21-1) was selected because of its high similarity and its significant correlation with the COPD phenotype. Serum Cyfra 21-1 levels were determined in patients with COPD and controls (4.3 ± 5.9 vs. 3.9 ± 3.6 ng/mL, P = 0.611). The emphysema index was significantly correlated with the serum Cyfra 21-1 level (correlation coefficient = 0.219, P = 0.015). CONCLUSION: Word embedding may be used for the discovery of biomarkers for COPD and Cyfra 21-1 may be used as a biomarker for emphysema. Additional studies are needed to validate Cyfra 21-1 as a biomarker for COPD.

A genome-wide association study of quantitative computed tomographic emphysema in Korean populations.

Emphysema is an important feature of chronic obstructive pulmonary disease (COPD). Genetic factors likely affect emphysema pathogenesis, but this question has predominantly been studied in those of European ancestry. In this study, we sought to determine genetic components of emphysema severity and characterize the potential function of the associated loci in Korean population. We performed a genome-wide association study (GWAS) on quantitative emphysema in subjects with or without COPD from two Korean COPD cohorts. We investigated the functional consequences of the loci using epigenetic annotation and gene expression data. We also compared our GWAS results with an epigenome-wide association study and previous differential gene expression analysis. In total, 548 subjects (476 [86.9%] male) including 514 COPD patients were evaluated. We identified one genome-wide significant SNP (P < 5.0 × 10-8), rs117084279, near PIBF1. We identified an additional 57 SNPs (P < 5.0 × 10-6) associated with emphysema in all subjects, and 106 SNPs (P < 5.0 × 10-6) in COPD patients. Of these candidate SNPs, 2 (rs12459249, rs11667314) near CYP2A6 were expression quantitative trait loci in lung tissue and a SNP (rs11214944) near NNMT was an expression quantitative trait locus in whole blood. Of note, rs11214944 was in linkage disequilibrium with variants in enhancer histone marks in lung tissue. Several genes near additional SNPs were identified in our previous EWAS study with nominal level of significance. We identified a novel SNP associated with quantitative emphysema on CT. Including the novel SNP, several candidate SNPs in our study may provide clues to the genetic etiology of emphysema in Asian populations. Further research and validation of the loci will help determine the genetic factors for the development of emphysema.

Computed tomography-based visual assessment of chronic obstructive pulmonary disease: comparison with pulmonary function test and quantitative computed tomography.

BACKGROUND: Chronic obstructive pulmonary disease (COPD) has variable subtypes involving mixture of large airway inflammation, small airway disease, and emphysema. This study evaluated the relationship between visually assessed computed tomography (CT) subtypes and clinical/imaging characteristics. METHODS: In total, 452 participants were enrolled in this study between 2012 and 2017. Seven subtypes were defined by visual evaluation of CT images using Fleischner Society classification: normal, paraseptal emphysema (PSE), bronchial disease, and centrilobular emphysema (trace, mild, moderate and confluent/advanced destructive). The differences in several variables, including clinical, laboratory, spirometric, and quantitative CT features among CT-based visual subtypes, were compared using the chi-square tests and one-way analysis of variance. RESULTS: Subjects who had PSE had better forced expiratory volume in 1 second (FEV1) (P=0.03) percentage and higher lung density (P<0.05) than those with moderate to confluent/advanced destructive centrilobular emphysema. As the visual grade of centrilobular emphysema worsened, pulmonary function declined and modified Medical Research Council, COPD assessment test (CAT) score, and quantitative assessment (emphysema index and air trapping) increased. The bronchial subtype was associated with higher body mass index (BMI), better lung function and higher lung density. Participants with trace emphysema showed a rapid increase in functional small airway disease. CONCLUSIONS: Classifying subtypes using visual CT imaging features can reflect heterogeneity and pathological processes of COPD.

A comparative study of chest CT findings regarding the effects of regional dust exposure on patients with COPD living in urban areas and rural areas near cement plants.

BACKGROUND: The clinical and radiological presentation of chronic obstructive pulmonary disease (COPD) is heterogenous depending on the characterized sources of inflammation. This study aimed to evaluate COPD phenotypes associated with specific dust exposure. METHODS: This study was designed to compare the characteristics, clinical outcomes and radiological findings between two prospective COPD cohorts representing two distinguishing regions in the Republic of Korea; COPD in Dusty Area (CODA) and the Korean Obstructive Lung Disease (KOLD) cohort. A total of 733 participants (n = 186 for CODA, and n = 547 for KOLD) were included finally. A multivariate analysis to compare lung function and computed tomography (CT) measurements of both cohort studies after adjusting for age, sex, education, body mass index, smoking status, and pack-year, Charlson comorbidity index, and frequency of exacerbation were performed by entering the level of FEV1(%), biomass exposure and COPD medication into the model in stepwise. RESULTS: The mean wall area (MWA, %) became significantly lower in COPD patients in KOLD from urban and metropolitan area than those in CODA cohort from cement dust area (mean ± standard deviation [SD]; 70.2 ± 1.21% in CODA vs. 66.8 ± 0.88% in KOLD, p = 0.028) after including FEV1 in the model. COPD subjects in KOLD cohort had higher CT-emphysema index (EI, 6.07 ± 3.06 in CODA vs. 20.0 ± 2.21 in KOLD, p < 0.001, respectively). The difference in the EI (%) was consistently significant even after further adjustment of FEV1 (6.12 ± 2.88% in CODA vs. 17.3 ± 2.10% in KOLD, p = 0.002, respectively). However, there was no difference in the ratio of mean lung density (MLD) between the two cohorts (p = 0.077). Additional adjustment for biomass parameters and medication for COPD did not alter the statistical significance after entering into the analysis with COPD medication. CONCLUSIONS: Higher MWA and lower EI were observed in COPD patients from the region with dust exposure. These results suggest that the imaging phenotype of COPD is influenced by specific environmental exposure.

A 3D-CNN model with CT-based parametric response mapping for classifying COPD subjects.

Chronic obstructive pulmonary disease (COPD) is a respiratory disorder involving abnormalities of lung parenchymal morphology with different severities. COPD is assessed by pulmonary-function tests and computed tomography-based approaches. We introduce a new classification method for COPD grouping based on deep learning and a parametric-response mapping (PRM) method. We extracted parenchymal functional variables of functional small airway disease percentage (fSAD%) and emphysema percentage (Emph%) with an image registration technique, being provided as input parameters of 3D convolutional neural network (CNN). The integrated 3D-CNN and PRM (3D-cPRM) achieved a classification accuracy of 89.3% and a sensitivity of 88.3% in five-fold cross-validation. The prediction accuracy of the proposed 3D-cPRM exceeded those of the 2D model and traditional 3D CNNs with the same neural network, and was comparable to that of 2D pretrained PRM models. We then applied a gradient-weighted class activation mapping (Grad-CAM) that highlights the key features in the CNN learning process. Most of the class-discriminative regions appeared in the upper and middle lobes of the lung, consistent with the regions of elevated fSAD% and Emph% in COPD subjects. The 3D-cPRM successfully represented the parenchymal abnormalities in COPD and matched the CT-based diagnosis of COPD.