Deep learning parametric response mapping from inspiratory chest CT scans: a new approach for small airway disease screening.

OBJECTIVES: Parametric response mapping (PRM) enables the evaluation of small airway disease (SAD) at the voxel level, but requires both inspiratory and expiratory chest CT scans. We hypothesize that deep learning PRM from inspiratory chest CT scans can effectively evaluate SAD in individuals with normal spirometry. METHODS: We included 537 participants with normal spirometry, a history of smoking or secondhand smoke exposure, and divided them into training, tuning, and test sets. A cascaded generative adversarial network generated expiratory CT from inspiratory CT, followed by a UNet-like network predicting PRM using real inspiratory CT and generated expiratory CT. The performance of the prediction is evaluated using SSIM, RMSE and dice coefficients. Pearson correlation evaluated the correlation between predicted and ground truth PRM. ROC curves evaluated predicted PRMfSAD (the volume percentage of functional small airway disease, fSAD) performance in stratifying SAD. RESULTS: Our method can generate expiratory CT of good quality (SSIM 0.86, RMSE 80.13 HU). The predicted PRM dice coefficients for normal lung, emphysema, and fSAD regions are 0.85, 0.63, and 0.51, respectively. The volume percentages of emphysema and fSAD showed good correlation between predicted and ground truth PRM (|r| were 0.97 and 0.64, respectively, p < 0.05). Predicted PRMfSAD showed good SAD stratification performance with ground truth PRMfSAD at thresholds of 15%, 20% and 25% (AUCs were 0.84, 0.78, and 0.84, respectively, p < 0.001). CONCLUSION: Our deep learning method generates high-quality PRM using inspiratory chest CT and effectively stratifies SAD in individuals with normal spirometry.

Structural features on quantitative chest computed tomography of patients with maximal mid-expiratory flow impairment in a normal lung function population.

BACKGROUND: Maximal mid-expiratory flow (MMEF) is an earlier predictor of chronic obstructive pulmonary disease (COPD) development than forced expiratory volume in 1 s (FEV1). Changes of lung structure in patients with MMEF impairment only is still not clear. Therefore, this study aimed to investigate the structural features of patients with decreased MMEF by quantitative computed tomography (QCT) and develop a predictive model for predicting patients with reduced MMEF in normal lung function population. METHODS: In this study, 131 patients with normal spirometry results and available volumetric chest CT images were enrolled and divided into the reduced MMEF group (FEV1/forced expiratory vital capacity (FEV1/FVC) > 0.7, FEV1% predictive values (FEV1%pred) > 80%, MMEF%pred < 80%, n = 52) and the normal MMEF group (FEV1/FVC > 0.7, FEV1%pred > 80%, MMEF%pred ≥ 80%, n = 79). The emphysema, small airway disease and medium-size airway parameters were measured by a commercial software. The differences were investigated in clinical features, spirometrical parameters and QCT parameters between the two groups. A nomogram model was constructed based on the results of the multivariable logistic regression model. Spearman's correlation coefficients were calculated between QCT measurements and spirometrical parameters. RESULTS: There were more males in reduced MMEF group than normal group (P < 0.05). Lung parenchyma parameter (PRMEmph) and airway-related parameters (functional small airway disease (PRMfSAD), luminal area of fifth- and sixth- generation airway (LA5, LA6) were significantly different between the reduced MMEF group and the normal group (20.2 ± 17.4 vs 9.4 ± 6.7, 3.4 ± 3.5 vs 1.9 ± 2.0, 12.2 ± 2.5 vs 13.7 ± 3.4, 7.7 ± 2.4 vs 8.9 ± 2.8, respectively, all P < 0.01). After multivariable logistical regression, only sex (odds ratio [OR]: 2.777; 95% confidence interval [CI]:1.123-3.867), PRMfSAD (OR:1.102, 95%CI:1.045-1.162) and LA6 (OR:0.650, 95%CI:0.528-0.799) had significant differences between the two groups (P < 0.05) and a model incorporating with the three indicators was constructed (area under curve, 0.836). Correlation analysis showed MMEF%pred had mild to moderate correlation with airway-related measurements. CONCLUSION: In normal lung function population, patients with reduced MMEF have potential medium-size and small airway changes, and MMEF%pred is significantly associated with airway-related CT parameters. The nomogram incorporating with sex, PRMfSAD and LA6 has good predictive value and offers more objective evidences in a group with reduced MMEF.

Subtyping preserved ratio impaired spirometry (PRISm) by using quantitative HRCT imaging characteristics.

BACKGROUND: Preserved Ratio Impaired Spirometry (PRISm) is defined as FEV1/FVC ≥ 70% and FEV1 < 80%pred by pulmonary function test (PFT). It has highly prevalence and is associated with increased respiratory symptoms, systemic inflammation, and mortality. However, there are few radiological studies related to PRISm. The purpose of this study was to investigate the quantitative high-resolution computed tomography (HRCT) characteristics of PRISm and to evaluate the correlation between quantitative HRCT parameters and pulmonary function parameters, with the goal of establishing a nomogram model for predicting PRISm based on quantitative HRCT. METHODS: A prospective and continuous study was performed in 488 respiratory outpatients from February 2020 to February 2021. All patients underwent both deep inspiratory and expiratory CT examinations, and received pulmonary function test (PFT) within 1 month. According to the exclusion criteria and Global Initiative for Chronic Obstructive Lung Disease (GOLD) classification standard, 94 cases of normal pulmonary function, 51 cases of PRISm and 48 cases of mild to moderate chronic obstructive lung disease (COPD) were included in the study. The lung parenchyma, parametric response mapping (PRM), airway and vessel parameters were measured by automatic segmentation software (Aview). One-way analysis of variance (ANOVA) was used to compare the differences in clinical features, pulmonary function parameters and quantitative CT parameters. Spearman rank correlation analysis was used to evaluate the correlation between CT quantitative index and pulmonary function parameters. The predictors were obtained by binary logistics regression analysis respectively in normal and PRISm as well as PRISm and mild to moderate COPD, and the nomogram model was established. RESULTS: There were significant differences in pulmonary function parameters among the three groups (P < 0.001). The differences in pulmonary parenchyma parameters such as emphysema index (EI), pixel indices-1 (PI-1) and PI-15 were mainly between mild to moderate COPD and the other two groups. The differences of airway parameters and pulmonary vascular parameters were mainly between normal and the other two groups, but were not found between PRISm and mild to moderate COPD. Especially there were significant differences in mean lung density (MLD) and the percent of normal in PRM (PRMNormal) among the three groups. Most of the pulmonary quantitative CT parameters had mild to moderate correlation with pulmonary function parameters. The predictors of the nomogram model using binary logistics regression analysis to distinguish normal from PRISm were smoking, MLD, the percent of functional small airways disease (fSAD) in PRM (PRMfSAD) and Lumen area. It had a good goodness of fit (χ2 = 0.31, P < 0.001) with the area under curve (AUC) value of 0.786. The predictor of distinguishing PRISm from mild to moderate COPD were PRMEmph (P < 0.001, AUC = 0.852). CONCLUSIONS: PRISm was significantly different from subjects with normal pulmonary function in small airway and vessel lesions, which was more inclined to mild to moderate COPD, but there was no increase in pulmonary parenchymal attenuation. The nomogram based on quantitative HRCT parameters has good predictive value and provide more objective evidence for the early screening of PRISm.

Three-Dimensional Convolutional Neural Network-Based Prediction of Epidermal Growth Factor Receptor Expression Status in Patients With Non-Small Cell Lung Cancer.

OBJECTIVES: EGFR testing is a mandatory step before targeted therapy for non-small cell lung cancer patients. Combining some quantifiable features to establish a predictive model of EGFR expression status, break the limitations of tissue biopsy. MATERIALS AND METHODS: We retrospectively analyzed 1074 patients of non-small cell lung cancer with complete reports of EGFR gene testing. Then manually segmented VOI, captured the clinicopathological features, analyzed traditional radiology features, and extracted radiomic, and deep learning features. The cases were randomly divided into training and test set. We carried out feature screening; then applied the light GBM algorithm, Resnet-101 algorithm, logistic regression to develop sole models, and fused models to predict EGFR mutation conditions. The efficiency of models was evaluated by ROC and PRC curves. RESULTS: We successfully established Modelclinical, Modelradiomic, ModelCNN (based on clinical-radiology, radiomic and deep learning features respectively), Modelradiomic+clinical (combining clinical-radiology and radiomic features), and ModelCNN+radiomic+clinical (combining clinical-radiology, radiomic, and deep learning features). Among the prediction models, ModelCNN+radiomic+clinical showed the highest performance, followed by ModelCNN, and then Modelradiomic+clinical. All three models were able to accurately predict EGFR mutation with AUC values of 0.751, 0.738, and 0.684, respectively. There was no significant difference in the AUC values between ModelCNN+radiomic+clinical and ModelCNN. Further analysis showed that ModelCNN+radiomic+clinical effectively improved the efficacy of Modelradiomic+clinical and showed better efficacy than ModelCNN. The inclusion of clinical-radiology features did not effectively improve the efficacy of Modelradiomic. CONCLUSIONS: Either deep learning or radiomic signature-based models can provide a fairly accurate non-invasive prediction of EGFR expression status. The model combined both features effectively enhanced the performance of radiomic models and provided marginal enhancement to deep learning models. Collectively, fusion models offer a novel and more reliable way of providing the efficacy of currently developed prediction models, and have far-reaching potential for the optimization of noninvasive EGFR mutation status prediction methods.

A Computed Tomography-Derived Radiomics Approach for Predicting Uncommon EGFR Mutation in Patients With NSCLC.

PURPOSE: This study aims to develop a CT-based radiomics approach for identifying the uncommon epidermal growth factor receptor (EGFR) mutation in patients with non-small cell lung cancer (NSCLC). METHODS: This study involved 223 NSCLC patients (107 with uncommon EGFR mutation-positive and 116 with uncommon EGFR mutation-negative). A total of 1,269 radiomics features were extracted from the non-contrast-enhanced CT images after image segmentation and preprocessing. Support vector machine algorithm was used for feature selection and model construction. Receiver operating characteristic curve analysis was applied to evaluate the performance of the radiomics signature, the clinicopathological model, and the integrated model. A nomogram was developed and evaluated by using the calibration curve and decision curve analysis. RESULTS: The radiomics signature demonstrated a good performance for predicting the uncommon EGFR mutation in the training cohort (area under the curve, AUC = 0.802; 95% confidence interval, CI: 0.736-0.858) and was verified in the validation cohort (AUC = 0.791, 95% CI: 0.642-0.899). The integrated model combined radiomics signature with clinicopathological independent predictors exhibited an incremental performance compared with the radiomics signature or the clinicopathological model. A nomogram based on the integrated model was developed and showed good calibration (Hosmer-Lemeshow test, P = 0.92 in the training cohort and 0.608 in the validation cohort) and discrimination capacity (AUC of 0.816 in the training cohort and 0.795 in the validation cohort). CONCLUSION: Radiomics signature combined with the clinicopathological features can predict uncommon EGFR mutation in NSCLC patients.

Prediction of the Growth Rate of Early-Stage Lung Adenocarcinoma by Radiomics.

OBJECTIVES: To investigate the value of imaging in predicting the growth rate of early lung adenocarcinoma. METHODS: From January 2012 to June 2018, 402 patients with pathology-confirmed lung adenocarcinoma who had two or more thin-layer CT follow-up images were retrospectively analyzed, involving 407 nodules. Two complete preoperative CT images and complete clinical data were evaluated. Training and validation sets were randomly assigned according to an 8:2 ratio. All cases were divided into fast-growing and slow-growing groups. Researchers extracted 1218 radiomics features from each volumetric region of interest (VOI). Then, radiomics features were selected by repeatability analysis and Analysis of Variance (ANOVA); Based on the Univariate and multivariate analyses, the significant radiographic features is selected in training set. A decision tree algorithm was conducted to establish the radiographic model, radiomics model and the combined radiographic-radiomics model. Model performance was assessed by the area under the curve (AUC) obtained by receiver operating characteristic (ROC) analysis. RESULTS: Sixty-two radiomics features and one radiographic features were selected for predicting the growth rate of pulmonary nodules. The combined radiographic-radiomics model (AUC 0.78) performed better than the radiographic model (0.727) and the radiomics model (0.710) in the validation set. CONCLUSIONS: The model has good clinical application value and development prospects to predict the growth rate of early lung adenocarcinoma through the combined radiographic-radiomics model.

Radiomics signature on CECT as a predictive factor for invasiveness of lung adenocarcinoma manifesting as subcentimeter ground glass nodules.

Controversy and challenges remain regarding the cognition of lung adenocarcinomas presented as subcentimeter ground glass nodules (GGNs). Postoperative lymphatic involvement or intrapulmonary metastasis is found in approximately 15% to 20% of these cases. This study aimed to develop and validate a radiomics signature to identify the invasiveness of lung adenocarcinoma appearing as subcentimeter ground glass nodules. We retrospectively enrolled 318 subcentimeter GGNs with histopathology-confirmed adenocarcinomas in situ (AIS), minimally invasive adenocarcinomas (MIA) and invasive adenocarcinomas (IAC). The radiomics features were extracted from manual segmentation based on contrast-enhanced CT (CECT) and non-contrast enhanced CT (NCECT) images after imaging preprocessing. The Lasso algorithm was applied to construct radiomics signatures. The predictive performance of radiomics models was evaluated by receiver operating characteristic (ROC) analysis. A radiographic-radiomics combined nomogram was developed to evaluate its clinical utility. The radiomics signature on CECT (AUC: 0.896 [95% CI 0.815-0.977]) performed better than the radiomics signature on NCECT data (AUC: 0.851[95% CI 0.712-0.989]) in the validation set. An individualized prediction nomogram was developed using radiomics model on CECT and radiographic model including type, shape and vascular change. The C index of the nomogram was 0.915 in the training set and 0.881 in the validation set, demonstrating good discrimination. Decision curve analysis (DCA) revealed that the proposed model was clinically useful. The radiomics signature built on CECT could provide additional benefit to promote the preoperative prediction of invasiveness in patients with subcentimeter lung adenocarcinomas.

Predictors of CT Morphologic Features to Identify Spread Through Air Spaces Preoperatively in Small-Sized Lung Adenocarcinoma.

OBJECTIVES: This study aimed to explore the predictive CT features of spread through air spaces (STAS) in patients with small-sized lung adenocarcinoma. METHODS: From January 2017 to May 2019, patients with confirmed pathology of small-sized lung adenocarcinoma (less than or equal to 2 cm) and who underwent surgery were retrospectively analyzed. The clinical, pathological, and surgical information and CT features were analyzed. RESULTS: A total of 47 patients with STAS (males, 61.7%; mean age, 56 ± 8years) and 143 patients without STAS (males, 58%; mean age, 53 ± 11 years) were included. Pathologically, papillary, micropapillary, solid predominant subtypes, and vascular and pleural invasion were most commonly observed features in the STAS group. Radiologically, higher consolidation tumor ratio (CTR), presence of spiculation, satellites, ground glass ribbon sign, pleural attachment, and unclear tumor-lung interface were more commonly observed features in the STAS group. CTR, presence of ground glass ribbons and pleural connection, and absence of cystic airspaces were considered as stable predictors of STAS in multivariate logistic models. The receiver operating characteristic curve (ROC) analysis for predicting STAS demonstrated higher area under the curve (AUC) in the model that used CTR (0.760, 95% confidence interval, 0.69-0.83) for predicting STAS than in the model that used long diameter of entire lesion (0.640). CONCLUSIONS: CTR is the best CT sign for predicting STAS in small-sized lung adenocarcinoma. The ground glass ribbon is a newly found indicator and has the potential for predicting STAS.