Abstract
Background
Chronic obstructive pulmonary disease (COPD) and asthma are two common chronic inflammatory diseases that are difficult to distinguish clinically. While pulmonary function tests (PFTs) are considered the Global Initiative for Chronic Obstructive Lung Disease (GOLD) standard for diagnosing these two diseases, they have inherent drawbacks, such as heavily relying on patient cooperation. This study aimed to explore the ability of impulse oscillometry system (IOS), quantitative computed tomography (QCT) imaging and radiomics features for differentiating between COPD and asthma.
Methods
This retrospective study enrolled 151 patients, including 67 COPD patients and 84 asthma patients who completed inspiratory high-resolution computed tomography (CT) scans with recorded demographics, PFTs and IOS data from January 2021 to September 2023 from The First Affiliated Hospital of Zhengzhou University. The NeuLungCARE-QA software (Neusoft Medical Systems Co., Ltd., Shenyang, China) was used to extract QCT parameters. The radiomics features were calculated using the Pyradiomics package. After features selection, the logistic regression method was used for model construction. An additional 37 patients were recruited as the external validation set to validate the model’s generalization.
Results
There were significant differences in PFT parameters (P<0.001), IOS indices (P<0.05), and emphysema parameters (P<0.001) between COPD and asthma groups. Model 6, which combined IOS indices, QCT parameters, and radiomics features, achieved the best performance with area under the receiver operating characteristic curve (AUC) values of 0.959 [95% confidence interval (CI): 0.918–0.989], 0.960 (95% CI: 0.909–0.999) and 0.949 (95% CI: 0.871–0.999) in the training set, testing set, external validation set, respectively. The DeLong test comparing the fusion features-based Model 6 with the PFTs-based Model 1 showed no significant differences in training set, testing set and external validation set, respectively.
Conclusions
The IOS indices, along with quantitative CT imaging and radiomics features, could provide a reliable tool for differentiating asthma from COPD patients. This approach could serve as an alternative and/or a complement method for patients who find it difficult to perform PFTs.
Keywords: Impulse oscillometry system (IOS), quantitative computed tomography (QCT), radiomics, chronic obstructive pulmonary disease (COPD), asthma
Highlight box.
Key findings
• Our study has demonstrated that significant differences in pulmonary function tests (PFTs), impulse oscillometry system (IOS), and quantitative computed tomography (QCT) parameters exist between the chronic obstructive pulmonary disease (COPD) group and the asthma group. Furthermore, the model integrating IOS, QCT, and radiomics parameters can distinguish COPD from asthma, achieving the best performance with area under the receiver operating characteristic curve values of 0.959 [95% confidence interval (CI): 0.918–0.989], 0.960 (95% CI: 0.909–0.999) and 0.949 (95% CI: 0.871–0.999) in the training set, testing set, and external validation set, respectively.
What is known and what is new?
• Compared with PFTs, IOS requires minimal effort and cooperation, making it suitable for a broader range of patients. However, IOS is less commonly used to differentiate COPD from asthma. Chest computed tomography (CT), by contrast, is widely used for lung cancer screening and diagnosis, and its role in differentiating between COPD and asthma is underestimated.
• This study innovatively constructs a model to differentiate COPD from asthma by integrating IOS, QCT, and radiomics parameters.
What is the implication, and what should change now?
• This model has the potential to act as an alternative or supplementary tool for patients with COPD and asthma who are unable to undergo conventional PFTs.
• For patients who experience difficulty in undergoing PFTs, clinicians may consider utilizing IOS and CT for the differential diagnosis of COPD and asthma.
Introduction
Chronic inflammatory diseases of the airways, represented by chronic obstructive pulmonary disease (COPD) and asthma, are common diseases of the respiratory system (1), affecting over 500 million people worldwide (2). They consume a large amount of healthcare resources every year and bring a heavy burden to patients and society (3,4). Asthma and COPD exhibit several similar clinical features including symptoms and physiologic impairments (5). In addition, there is also partial overlap in the pathological characteristics of the two diseases, such as promotion of airway hyper-reactivity, common susceptibility genes, lung growth during the fetal stage/early childhood and respiratory dysfunction (6), further increasing the difficulty of distinguishing between the two diseases clinically. A previous study revealed that around 15–45% of patients with COPD may have co-existed asthma-COPD (7). The 2024 Global Initiative for Chronic Obstructive Lung Disease (GOLD) reported that asthma may be a risk factor for the development of chronic airflow obstruction and COPD (8). Therefore, accurate differential diagnosis of the two diseases and the development of reasonable treatment measures are crucial for improving patient symptoms and reducing the burden of the disease.
Pulmonary function tests (PFTs) are considered the gold standard for the clinical diagnosis of COPD and asthma (9). However, PFTs rely on patient cooperation, have limited application scope, and exhibit poor reproducibility (10). The impulse oscillometry system (IOS) is an independent technique that can be used during spontaneous tidal breathing or respiratory support ventilation, recording gas flow velocity and airway resistance at different frequencies (11). Compared with PFTs, IOS requires minimal effort and cooperation, making it suitable for a wide range of patients including children, elders, and people with disabilities (12). Many studies have explored the clinical value of IOS technique. For example, Li et al. found that IOS indices were more sensitive in detecting small airway dysfunction than spirometry in subjects with chronic respiratory symptoms and preserved pulmonary function, and they correlated better with symptoms (13). Chen et al. identified the potential roles of IOS indices, such as airway resistance and reactance, in predicting COPD for chest physicians and clinicians (14). Additionally, Sugawara et al. reported that IOS indices were significantly correlated with the severity of asthma, suggesting that could be an effective tool for selecting inhaled corticosteroids and evaluating asthma phenotype (15). Xu et al. demonstrated that IOS indices had good long-term repeatability in both asthma and COPD (16). Therefore, IOS, as a non-invasive test, provides rich information that can sensitively reflect dynamic changes in respiratory status and mechanics. While these studies have primarily focused on a sole disease, the ability of IOS indicators to differentiate COPD from asthma warrants further exploration.
Increased airway resistance, narrowing, and remodeling are important manifestations of the progression of chronic respiratory diseases (17). Computed tomography (CT) scanning is a non-invasive and highly reproducible tool for the objective assessment of changes in lung structure. Recently, the quantitative computed tomography (QCT) imaging technique has been proposed to explore the pathophysiology, diagnosis and outcomes in chronic respiratory diseases, which could provide the quantitative analysis including the degree of emphysema, airway inflammation and remodeling (18,19). QCT measurements have been shown to correlate strongly with pulmonary function tests and predict exacerbation risk and mortality in patients with chronic respiratory diseases (20). In both COPD and asthma research, many studies have demonstrated the clinical significance of QCT technique. QCT technique plays a critical role in COPD management by improving early detection, addressing disease heterogeneity, and monitoring progression, as noted by the Asian Society of Thoracic Radiology (21). For asthma, particularly severe asthma, QCT technique helps identify imaging-based phenotypes and shows a good correlation with lung functional decline, thereby guiding clinicians in understanding the pathophysiological mechanisms underlying these phenotypes (22). Additionally, QCT parameters can reflect the similarity and disparity in lesions, such as airway remodeling, air trapping, and emphysema, between COPD patients and asthma patients (23,24). While QCT method provides valuable insights, radiomics has emerged as a complementary method that can extract rich information from medical images, bridging medical imaging and precision medicine (25). Radiomics has been applied in COPD research for early prediction (26,27), identification and severity staging (28,29), and prognostic analysis in recent years (30). However, we found that only one study explored the value of radiomics for staging patients with asthma, demonstrating that the radiomics model could offer promising method for the clinical diagnosis of asthma patients (31). Thus, the effectiveness of radiomics features in differentiating asthma from COPD remains to be determined. This underscores the critical need for further research in this area to improve clinical decision-making and enhance patient outcomes.
Therefore, this study aimed to explore the diagnostic value of IOS indices, QCT parameters, and radiomics features in distinguishing COPD from asthma, and seek an optimal method for the patients who find it difficult to perform PFTs. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1-2558/rc).
Methods
Participants
This study involves a retrospective analysis of patients who presented at The First Affiliated Hospital of Zhengzhou University between January 2021 and September 2023, exhibiting chronic respiratory symptoms primarily characterized by chest tightness, cough, and sputum production. The inclusion criteria and exclusion criteria are as follows. Inclusion criteria: (I) the presence of non-fully reversible airflow obstruction [i.e., forced expiratory volume in 1 second (FEV1)/forced vital capacity (FVC) <0.7 post-bronchodilation] measured by spirometry, confirming the diagnosis of COPD according to Global Strategy for The Diagnosis, Management, and Prevention of COPD (2024 Report), which can be found on the GOLD website (www.goldcopd.org); (II) the diagnosis of asthma is mainly based on typical clinical symptoms and reversible airflow restriction according to Global Strategy for Asthma Management and Prevention (2024 update), available on the GINA website (www.ginasthma.org); (III) all patients were diagnosed in our hospital and have complete PFTs data and inspiratory CT scans. Exclusion criteria: (I) patients with insufficient clinical data; (II) patients who have not undergone IOS or CT examinations; (III) patients with asthma-COPD overlap.
Finally, a total of 151 patients were included as the discovery set in this study, which were divided into COPD groups with 67 patients and asthma groups with 84 patients. Additionally, we enrolled an external validation set of 37 patients which included 21 patients with asthma and 16 patients with COPD. The framework for patient enrollment is shown in Figure 1.
Figure 1.
The framework of participant selection. COPD, chronic obstructive pulmonary disease; CT, computed tomography.
This study conformed with the principles outlined in the Declaration of Helsinki and its subsequent amendments. This study was approved by the ethics committee of The First Affiliated Hospital of Zhengzhou University (No. 2024-KY-1413-001) and individual consent for this retrospective analysis was waived.
CT imaging protocol
All participants underwent CT scans while placed in the supine position during end-inspiration. Patients with breathing difficulties received guidance and consultation before the scan, and all scans were performed after breath-hold practice. The discovery set comprised CT scans with a thickness of 1.00 mm, obtained using four CT systems: SOMATOM X. cite and SOMATOM go. Top (Siemens Healthineers AG, Forchheim, Germany); as well as CT 6000 and IQon-Spectral CT (Philips Medical Systems, Best, Netherlands). The external validation set included CT scans with a thickness of 1.25 mm, obtained using Revolution CT (GE Healthcare; Somatom Force, Siemens). The CT imaging protocol involved a voltage of 120 kV, a tube current ranging from 30 to 100 mA, and a field of view of 512 mm × 512 mm.
PFT and IOS parameters
The Jaeger MasterScreen PFTs system (Jaeger Co., Wurzburg, Germany) was used to quantify PFTs and IOS parameters, with the operating procedure following the American Thoracic Society/European Respiratory Society guidelines (32). The following PFTs parameters were evaluated: FEV1, predicted FEV1 (FEV1%_pred), FEV1/FVC, maximal expiratory flow at 50% (MEF50%), maximal expiratory flow at 75% (MEF75%), predicted MEF50% (MEF50%_pred) and predicted MEF75% (MEF75%_pred). For IOS indices, the evaluated parameters included resonant frequency (Fres), respiratory resistance at 5 Hz (R5), predicted R5 (R5_pred), respiratory resistance at 20 Hz (R20), predicted R20 (R20_pred), the difference between R5 and R20 (R5–R20), and respiratory reactance at 5 Hz (X5). Each measurement was repeated three times, and the average values were used for subsequent analyses.
Assessment of emphysema and airway change
The Quantitative CT metric for airways and emphysema were automatically computed using the NeuLungCARE-QA software (Neusoft Medical Systems Co., Ltd., Shenyang, China) (33). The software analyzes the degree of emphysema by assessing lung parenchymal density, including percentage of low attenuation areas of the whole lung and each lobe including right upper lobe, right middle lobe, right lower lobe, left lower lobe, and left upper lobe, defined as lung tissue with a density ≤−950 or −910 Hounsfield unit (HU) on inspiratory CT (LAA_950 or LAA_910), and the CT attenuation at the 15th percentile of the CT histogram (perc15). In addition, this software provides the cross-sectional measurement of 0–4 generation airways characteristics, including the maximum, mean, and minimum wall thickness, as well as the maximum, mean, and minimum wall diameters. It also measures square root of wall area of a hypothetical airway with an internal perimeter of 10 mm (Pi10), and normalized airway metrics such as the percentage of airway wall area (WA).
Radiomics feature extraction
The U-net (R231) deep learning model (https://github.com/JoHof/lungmask) was used for automatic lung segmentation (34). The PyRadiomics (version 3.0.1) package for Python (version 3.10.0) was used for the extraction of radiomics features from CT imaging. First, we converted the image format from Dicom to Nii.gz. Second, the images were normalized to a scale of 100 and resampled to 1 mm × 1 mm × 1 mm, with a bin-width set to 25. The original image features included 14 shape features, 18 first-order histogram-based features, and 75 texture features. Additional features derived from Laplacian of Gaussian images with sigma values of 3.0 and 5.0, and wavelet-transformed images filtered into eight subbands using a 3D wavelet filter, were computed, resulting in a total of 1,037 radiomics features extracted from each patient.
Feature analysis
A two-stage feature selection strategy, including F-test and LASSO regression, was employed to reduce the risk of model overfitting. The F-test was used to determine whether the variability between groups was greater than the variability within group observations. We combined the 50 clusters bootstrapping with F-test to selected features with P<0.05 for next step feature selection, ensuring reliable outcomes. To avoid selecting 0 features in the first step method, we will appropriately reduce the number of bootstrapping clusters. Then, L1 regularization of LASSO method combined with five-fold cross validations was used to select the final features.
Model development and validation
The discovery set was stratified into a training set and a testing set at a ratio of 7:3. First, the training set data were standardized using Z-score normalization to eliminate the dimensional influence between indices. After features selection, and 6 models were constructed to compare the classification ability of different feature sets. The validity of the model built by the selected features is tested in 30% of the testing set, and the results of the external validation set to verify the generalization of the model. The six models are defined as follows: Model 1 based on PFTs parameters; Model 2 based on IOS indices; Model 3 based on QCT parameters; Model 4 based on radiomics features; Model 5 based on both QCT parameters and radiomics features; Model 6 based on IOS indices, QCT parameters and radiomics features. The external validation set of 37 patients, which included 21 patients with asthma and 16 patients with COPD, was used to assess the robustness of the six models. All the analyses were performed using the “Scikit-learn” package in Python (https://scikit-learn.org/stable/index.html).
Statistical analysis
The normality of the data was assessed using the Shapiro-Wilk method. Continuous variables that followed a normal distribution were presented as mean ± standard deviation (SD), while those that did not follow a normal distribution were presented as median (interquartile range, IQR). Categorical variables were expressed as numbers and percentages (N, %). Continuous and categorical variables were compared using two samples t-tests, Mann-Whitney U tests, or Chi-squared tests, as appropriate. Five evaluation metrics including the area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, specificity, accuracy, and F1-score, were used to evaluate the reliability of models. The DeLong test was used to analyze the significant differences between the AUC values of different models. A nomogram and decision curve analysis (DCA) were drawn to illustrate the effectiveness of model. All the statistical analyses were two-tailed, and a P<0.05 was considered to be significant.
Results
Participants’ characteristics
As shown in Table 1, compared to the asthma group, patients in the COPD group were relatively older, with a significant difference between the two groups (P<0.001). In the COPD group, 76.12% of the patients were male, while in the asthma group, 48.80% of the patients were male, with a significant difference between the two groups (P<0.001). Regarding PFTs data, the values of FEV1, FEV1%_pred, FEV1/FVC, MEF75, MEF75%_pred, MEF50, and MEF50%_pred in the COPD group were significantly lower than those observed in the asthma group (P<0.001). Similarly, the IOS indictor s in the COPD group, including Fres, R5, R5_pred, R20, R20_pred, and R5-R20, were significantly higher compared to asthma group (P<0.05). However, the X5 value was significantly lower in the COPD group (P<0.001).
Table 1. Baseline characteristics analysis of the asthma patients and COPD patients.
| Clinical data | Asthma (n=84) | COPD (n=67) | P value |
|---|---|---|---|
| Sex, male, n (%) | 41 (48.80) | 51 (76.12) | <0.001 |
| Age (years) | 50.00 (22.00) | 60.85±8.93 | <0.001 |
| PFTs | |||
| FEV1 (mL) | 2.74 (0.92) | 1.36±0.58 | <0.001 |
| FEV1%_pred | 96.16±12.74 | 51.80±19.16 | <0.001 |
| FEV1/FVC | 74.55±7.79 | 51.42 (13.57) | <0.001 |
| MEF75% | 5.23±1.58 | 1.49 (1.50) | <0.001 |
| MEF75%_pred | 83.83±21.34 | 24.60 (18.80) | <0.001 |
| MEF50% | 2.55 (1.06) | 0.61 (0.56) | <0.001 |
| MEF50%_pred | 63.18±17.79 | 17.30 (14.3) | <0.001 |
| IOS | |||
| Fres | 13.43 (6.81) | 30.81±9.67 | <0.001 |
| R5 | 0.40 (0.18) | 0.73±0.26 | <0.001 |
| R5_pred | 126.10 (43.40) | 222.04±72.47 | <0.001 |
| R20 | 0.37±0.09 | 0.40 (0.12) | 0.01 |
| R20_pred | 128.34±27.61 | 146.08±33.27 | <0.001 |
| R5–R20 | 0.05 (0.06) | 0.30 (0.27) | <0.001 |
| X5 | −0.10 (0.04) | −0.34 (0.27) | <0.001 |
Data are presented as mean ± standard deviation or median (interquartile range) unless otherwise specified. COPD, chronic obstructive pulmonary disease; FEV1, forced expiratory volume in 1 second; FEV1%_pred, forced expiratory volume in 1 second percentage predicted; Fres, resonant frequency; FVC, forced vital capacity; IOS, impulse oscillometry system; MEF50%, maximal expiratory flow at 50%; MEF50%_pred, predicted MEF50%; MEF75%, maximal expiratory flow at 75%; MEF75%_pred, predicted MEF75%; PFTs, pulmonary function tests; R20, respiratory resistance at 20 Hz; R20_pred, predicted R20; R5, respiratory resistance at 5 Hz; R5_pred, predicted R5; R5–R20, the difference between R5 and R20; X5, respiratory reactance at 5 Hz.
The emphysema parameters in asthma groups and COPD groups
As shown in Table 2, compared with the lung attenuation parameters of the asthma group, the COPD group exhibited lower perc15 values (−931.8 vs. −908.31), and significant increased LAA_910 and LAA_950 parameters in both the total lungs and lobes (P<0.001), indicating that the COPD group has more severe emphysema. In terms of different lobes, emphysema in the right middle lobe was observed to be more severe than in other lobes for both asthma and COPD groups (LAA_910_right_middle: mean value of 0.3 in the asthma group and 0.42 in the COPD group; LAA_950_right_middle: median value of 0.02 in the asthma groups and 0.06 in the COPD group). In addition, the LAA_950 of the whole lung and each lung lobe was lower than LAA_910, which illustrated the CT threshold would affect the quantification of the emphysema area.
Table 2. The emphysema parameters analysis of the asthma patients and COPD patients.
| Emphysema parameters | Asthma (n=84) | COPD (n=67) | P value |
|---|---|---|---|
| perc15 | −908.31±20.09 | −931.8 (26.38) | <0.001 |
| LAA_910_left_upper | 0.17 (0.22) | 0.38±0.20 | <0.001 |
| LAA_910_left_lower | 0.10 (0.13) | 0.26 (0.36) | <0.001 |
| LAA_910_right_upper | 0.12 (0.19) | 0.36±0.21 | <0.001 |
| LAA_910_right_middle | 0.30±0.18 | 0.42±0.19 | <0.001 |
| LAA_910_right_lower | 0.08 (0.12) | 0.25 (0.32) | <0.001 |
| LAA_910_lung | 0.14 (0.15) | 0.34±0.19 | <0.001 |
| LAA_950_left_upper | 0.01 (0.03) | 0.05 (0.10) | <0.001 |
| LAA_950_left_lower | 0.01 (0.01) | 0.04 (0.10) | <0.001 |
| LAA_950_right_upper | 0.01 (0.01) | 0.04 (0.08) | <0.001 |
| LAA_950_right_middle | 0.02 (0.03) | 0.06 (0.09) | <0.001 |
| LAA_950_right_lower | 0.01 (0.01) | 0.02 (0.07) | <0.001 |
| LAA_950_lung | 0.01 (0.02) | 0.05 (0.07) | <0.001 |
Data are presented as mean ± standard deviation or median (interquartile range). COPD, chronic obstructive pulmonary disease; LAA_910, low attenuation area below-910 Hounsfield units; LAA_950, low attenuation area below-950 Hounsfield units; perc15, the CT attenuation at the 15th percentile of the CT histogram.
The airway parameters in asthma groups and COPD groups
As shown in Table 3, the thickness values showed no significant difference between COPD and asthma groups. However, the Pi10 value in the COPD group was higher than that in the asthma group (mean value of 6.1 in the asthma group and 6.17 in the COPD group), which indicated a thickening of the airway wall in the COPD group. The WA parameters for the 0th to 2nd generations were decreased in COPD groups while the WA parameters for the 3rd to 4th generations were increased. Additionally, four bronchial diameter measurement parameters including Mean_diameter_0, Max_diameter_0, Mean_diameter_4, Min_diameter_4 showed significant differences between two groups (P<0.05).
Table 3. The airway parameters analysis of asthma patients and COPD patients.
| Airway parameters | Asthma (n=84) | COPD (n=67) | P value |
|---|---|---|---|
| Pi10 | 6.10±0.78 | 6.17±0.76 | 0.63 |
| WA_0 | 0.43±0.05 | 0.42±0.06 | 0.08 |
| Mean_thickness_0 | 2.36 (0.43) | 2.35 (0.37) | 0.74 |
| Max_thickness_0 | 2.98±0.39 | 2.98±0.40 | 0.99 |
| Min_thickness_0 | 1.60 (0.29) | 1.60 (0.28) | 0.46 |
| Mean_diameter_0 | 15.31±2.08 | 16.47±2.02 | 0.001 |
| Max_diameter_0 | 16.63 (3.71) | 19.07±2.68 | <0.001 |
| Min_diameter_0 | 13.48±1.83 | 13.97 (2.57) | 0.19 |
| WA_1 | 0.49±0.05 | 0.48±0.06 | 0.28 |
| Mean_thickness_1 | 2.30 (0.44) | 2.32 (0.42) | 0.97 |
| Max_thickness_1 | 3.04±0.39 | 3.04±0.41 | 0.96 |
| Min_thickness_1 | 1.43 (0.28) | 1.43 (0.26) | 0.65 |
| Mean_diameter_1 | 14.78 (2.73) | 15.55±2.50 | 0.24 |
| Max_diameter_1 | 19.55 (4.31) | 20.47 (4.30) | 0.62 |
| Min_diameter_1 | 10.22±1.57 | 10.74±1.67 | 0.050 |
| WA_2 | 0.58±0.06 | 0.57±0.07 | 0.72 |
| Mean_thickness_2 | 2.22±0.31 | 2.20±0.32 | 0.66 |
| Max_thickness_2 | 2.98±0.37 | 2.97±0.39 | 0.80 |
| Min_thickness_2 | 1.38±0.25 | 1.37±0.26 | 0.65 |
| Mean_diameter_2 | 11.40±2.06 | 11.57±2.89 | 0.94 |
| Max_diameter_2 | 15.39±3.17 | 15.55±4.61 | 0.83 |
| Min_diameter_2 | 7.41±1.31 | 7.59±1.43 | 0.44 |
| WA_3 | 0.69±0.07 | 0.72 (0.10) | 0.17 |
| Mean_thickness_3 | 2.17 (0.37) | 2.15±0.31 | 0.93 |
| Max_thickness_3 | 2.84±0.34 | 2.80±0.35 | 0.45 |
| Min_thickness_3 | 1.41±0.25 | 1.40±0.27 | 0.81 |
| Mean_diameter_3 | 7.10±1.85 | 6.28 (2.48) | 0.09 |
| Max_diameter_3 | 9.25±2.75 | 7.59 (3.42) | 0.11 |
| Min_diameter_3 | 4.96±1.09 | 4.50 (1.65) | 0.17 |
| WA_4 | 0.75±0.06 | 0.78 (0.08) | 0.14 |
| Mean_thickness_4 | 1.72±0.22 | 1.72 (0.29) | 0.76 |
| Max_thickness_4 | 2.25±0.28 | 2.22±0.27 | 0.47 |
| Min_thickness_4 | 1.16±0.18 | 1.16±0.20 | 0.93 |
| Mean_diameter_4 | 4.73 (1.77) | 4.16 (1.37) | 0.04 |
| Max_diameter_4 | 6.12 (2.56) | 5.47 (1.89) | 0.047 |
| Min_diameter_4 | 3.25±0.68 | 3.05±0.72 | 0.08 |
Data are presented as mean ± standard deviation or median (interquartile range). COPD, chronic obstructive pulmonary disease; Pi10, square root of wall area of a hypothetical airway with an internal perimeter of 10 mm; WA, the percentage of airway wall area.
The process of feature selection
After two-step feature selection, the optimal features were determined for model construction as shown in Figures S1-S5. For PFT parameters, seven features were selected using the ANOVA F-test with 50 bootstrap iterations (Figure S1), and three of them were subsequently selected using LASSO with five-fold cross-validation (Figure S2). Using the same feature selection approach, three IOS indices (Figure S1 and Figure S3) and two QCT parameters (Figure S1 and Figure S4) were independently selected. For radiomics feature selection, ANOVA F-test with 10 bootstrap iterations followed by LASSO were employed to select two features (Figure S1 and Figure S5).
Comparison of performance in distinguishing COPD from asthma in different models
As shown in Table 4 and Figure 2, Model 1 based on PFTs parameters achieved the highest AUC in training set, testing set, and external validation set, indicating that PFTs parameters had the excellent ability for distinguishing COPD from asthma. Model 2 based on IOS indices obtained AUCs of 0.925 [95% confidence interval (CI): 0.855–0.967], 0.958 (95% CI: 0.915–0.994), and 0.932 (95% CI: 0.864–0.987) in the training set, testing set, and external validation set, respectively; the DeLong test results in the testing set and external validation set showed no significant difference compared to Model 1 (P=0.24 and 0.27). This means that IOS indices also have good ability for distinguishing COPD from asthma.
Table 4. The performance of distinguishing COPD from asthma in different models.
| Models | Accuracy | Recall | Precision | F1-score | AUC | 95% CI lower | 95% CI upper | Delong test (P) |
|---|---|---|---|---|---|---|---|---|
| Training set | ||||||||
| Model 1 | 0.943 | 0.936 | 0.936 | 0.936 | 0.989 | 0.973 | 0.997 | Reference |
| Model 2 | 0.876 | 0.809 | 0.905 | 0.854 | 0.925 | 0.855 | 0.967 | 0.03 |
| Model 3 | 0.762 | 0.681 | 0.762 | 0.719 | 0.815 | 0.721 | 0.872 | <0.001 |
| Model 4 | 0.724 | 0.723 | 0.680 | 0.701 | 0.762 | 0.695 | 0.829 | <0.001 |
| Model 5 | 0.762 | 0.681 | 0.762 | 0.719 | 0.849 | 0.789 | 0.912 | <0.001 |
| Model 6 | 0.886 | 0.809 | 0.927 | 0.864 | 0.959 | 0.918 | 0.989 | 0.11 |
| Testing set | ||||||||
| Model 1 | 0.957 | 0.950 | 0.950 | 0.950 | 0.985 | 0.949 | 1.000 | Reference |
| Model 2 | 0.826 | 0.650 | 0.929 | 0.765 | 0.958 | 0.915 | 0.994 | 0.24 |
| Model 3 | 0.717 | 0.550 | 0.733 | 0.629 | 0.746 | 0.593 | 0.893 | 0.002 |
| Model 4 | 0.609 | 0.400 | 0.571 | 0.471 | 0.727 | 0.559 | 0.878 | <0.001 |
| Model 5 | 0.739 | 0.550 | 0.786 | 0.647 | 0.785 | 0.710 | 0.876 | 0.004 |
| Model 6 | 0.891 | 0.800 | 0.941 | 0.865 | 0.960 | 0.909 | 0.999 | 0.31 |
| External validation set | ||||||||
| Model 1 | 0.865 | 0.813 | 0.867 | 0.839 | 0.979 | 0.935 | 1.000 | Reference |
| Model 2 | 0.838 | 0.750 | 0.857 | 0.800 | 0.932 | 0.864 | 0.987 | 0.27 |
| Model 3 | 0.703 | 0.500 | 0.727 | 0.593 | 0.836 | 0.676 | 0.933 | 0.04 |
| Model 4 | 0.676 | 0.875 | 0.583 | 0.700 | 0.753 | 0.612 | 0.912 | 0.005 |
| Model 5 | 0.757 | 0.688 | 0.733 | 0.710 | 0.836 | 0.675 | 0.941 | 0.03 |
| Model 6 | 0.838 | 0.813 | 0.813 | 0.813 | 0.949 | 0.871 | 0.999 | 0.39 |
The six models are defined as follows: Model 1 based on PFTs parameters; Model 2 based on IOS indices; Model 3 based on QCT parameters; Model 4 based on radiomics features; Model 5 based on both QCT parameters and radiomics features; Model 6 based on IOS indices, QCT parameters and radiomics features. AUC, area under the receiver operating characteristic curve; CI, confidence interval; COPD, chronic obstructive pulmonary disease; IOS, impulse oscillometry system; PFTs, pulmonary function tests; QCT, quantitative computed tomography.
Figure 2.
ROC curves of the six models in training set (A), testing set (B), and external validation set (C). AUC, area under the receiver operating characteristic curve; CI, confidence interval; IOS, impulse oscillometry system; PFT, pulmonary function test; QCT, quantitative computed tomography; ROC, receiver operating characteristic curve.
Model 3 based on QCT parameters yielded AUCs of 0.815 (95% CI: 0.721–0.872), 0.746 (95% CI: 0.593–0.893) and 0.836 (95% CI: 0.676–0.933) in the training, testing and external validation sets. However, the performance of Model 4 based on radiomics features was relatively poor with AUCs of 0.762 (95% CI: 0.695–0.829), 0.727(95% CI: 0.559–0.878), 0.753 (95% CI: 0.612–0.912) in the training, testing, and external validation sets.
Model 5 combined QCT parameters and radiomics features improved the prediction results in the testing set compared to Model 3 based only on QCT parameters and Model 4 based only on radiomics features (Model 5: 0.785 vs. Model 3: 0.746 vs. Model 4: 0.727). When not considering PFTs parameters, Model 6 combining IOS indices, QCT parameters, and radiomics features achieved the best results. In the training set, the accuracy, recall, precision, F1-score, AUC of Model 6 were 0.886, 0.809, 0.927, 0.864, and 0.959 (95% CI: 0.918–0.989) respectively. In the testing set, the accuracy, recall, precision, F1-score, AUC of Model 6 were 0.891, 0.800, 0.941, 0.865, and 0.960 (95% CI: 0.909–0.999), respectively. In the external validation set, the accuracy, recall, precision, F1-score, AUC of Model 6 were 0.838, 0.813, 0.813, 0.813, and 0.949 (95% CI: 0.871–0.999), respectively. The DeLong tests comparing Model 6 with Model 1 showed no significant difference in all sets (P=0.11, 0.31, and 0.39 for training, testing, and external validation sets).
The performance in distinguishing COPD from asthma based on multi-features fusion Model 6
A nomogram was constructed based on a multi-features fusion Model 6, which included IOS indices, QCT parameters and radiomics features, as shown in Figure 3. The decision curve of the Model 6 using multi-features is shown in Figure 4. The DCA results showed that the Model 6 achieved good net benefits in distinguishing COPD from asthma in the training, testing, and external validation sets.
Figure 3.
A nomogram was compiled based on selected features of Model 6. Model 6 based on IOS indices, QCT parameters and radiomics features. COPD, chronic obstructive pulmonary disease; Fres, resonant frequency; LAA_910, low attenuation area below-910 Hounsfield units; LAA_950, low attenuation area below-950 Hounsfield units; R5, respiratory resistance at 5 Hz; X5, respiratory reactance at 5 Hz.
Figure 4.
DCA of the Model 6 in training set (A), testing set (B), and external validation set (C). Model 6 based on IOS indices, QCT parameters and radiomics features. DCA, decision curve analysis.
Discussion
The current study assessed the differences in PFTs parameters, respiratory mechanics, emphysema parameters, and airway parameters between COPD patients and asthma patients. It was found that patients with COPD had worse airflow limitation, higher resistance values, more severe emphysema and airway lesions compared to asthma patients. In addition, we constructed the six models to automatically identify the COPD and asthma, among which the Model 2 based on IOS indices obtained AUCs of 0.925, 0.958, 0.932 in the training, testing, and external validation sets, demonstrating that IOS indices had the good ability for the differential diagnosis between asthma and COPD. In particular, the multi-feature fusion Model 6 including IOS indices, QCT parameter and radiomics features achieved the best AUC (training set: 0.959; testing set: 0.960; external validation set: 0.949) and the DeLong test indicated no significant differences between this model and Model 1 based on PFTs in the training, testing, and external validation sets; this demonstrated that the multi-feature fusion model may be a clinically feasible tool for COPD and asthma classification.
Previous studies showed that the predominant airway disorders in asthma and COPD were central and peripheral, respectively, and IOS can reflect the difference in airway disorders by R5 (central component) and R5–R20 (peripheral component) between asthma and COPD (35). Kanda et al. explored the differences in physiological properties of the airways between asthma and COPD using IOS. In that study, it was shown that the IOS indices R5–R20 increased in the COPD group compared with those in the asthma group; R5 increased in the COPD and asthma groups compared to the control group (36). Kim et al. found that IOS indices of Fres, R5, R20, and R5–R20 were higher, whereas X5 values were lower in asthmatic patients than in non-asthmatic patients (37). These remarkable pathological and physiological changes correlated with the severity of airways inflammation, which were consistent with our finding that higher Fres, R5, R20, R5–R20 and lower X5 were identified in COPD patients compared to asthma (Table 1) due to the more severe airways inflammation in COPD patients in our study. From a physiological perspective, the lesion of COPD always originates from the small airways, and a significant increase in small airway resistance will be observed when extensive lesions occur in the small airways, likely leading to the increase in total airway resistance in COPD patients (38). In COPD patients, as alveoli continue to expand, lung tissue overexpansion leads to decreased lung compliance. Therefore, the sum of elastic resistance and inertial resistance of the lung and thorax, i.e., X5 value, was significantly reduced. Asthma is generally considered to be a disease mainly involving the atmospheric tract, so during the acute episode of asthma, IOS parameters, such as R20 that reflect the dysfunction of the atmospheric tract, may be abnormal. Specifically, Kitaguchi et al. found that compared with COPD patients, R20 significantly increased in asthma patients with persistent airflow limitation (39). However, in the present study, the R20 levels in the COPD group were significantly higher than those in the asthma group, which may be related to the fact that most subjects in the asthma group were mild patients seeking outpatient treatment. These findings indicate that IOS can reflect the pulmonary characteristics of respiratory physiology and mechanics in patients with COPD, which plays an important role in the evaluation of COPD and asthma. In addition, a previous study has also demonstrated the clinical value of the IOS model in distinguishing COPD from asthma (40).
Furthermore, the evaluation of asthma and COPD could be performed using QCT parameters. Specifically, Hartley compared airway remodeling, air trapping, and emphysema between asthmatic patients and those with COPD; the relationships with airflow limitation were investigated using QCT parameters and it was shown that emphysema was largely restricted to patients with COPD, while asthmatic patients exhibited greater proximal airway remodeling compared to the COPD patients (23). In our study, the COPD group showed greater WA, smaller thickness values and diameters values in the 3rd to 4th generations airways compared to asthma patients. The emphysema parameters including perc 15, LAA_950, and LAA_910 in total lung and each lung lobe increased in COPD patients compare with those in asthma patients. These findings were consistent with previous studies, supporting the view that airflow limitation and obstruction in COPD may be attributed to a combination of small-airway obliteration and emphysema (41). To further explore the application of QCT parameters in distinguishing COPD from asthma, Moslemi et al. identified optimal QCT parameters to distinguish COPD from asthma, including the emphysema parameter of LAA_950, as well as some airway parameters such as outer airway perimeter, inner airway perimeter, total airway count, and low-attenuation cluster total hole count (42). Their model achieved an accuracy of 80%, but lacked external validation of its generalization. In contrast, our QCT model was not only validated on an external validation set, but also calculates AUC to comprehensively evaluate model performance, obtaining AUCs of 0.815, 0.746, and 0.836 in the training, testing, and external validation sets, further demonstrating that the QCT model can distinguish COPD from asthma.
Notably, radiomics have been demonstrated to be capable of conducting early prediction, identification and severity staging, prognostic analysis of COPD (26-30). Chen et al. achieved good results in predicting clinical staging of asthma patients based on radiomics features (31). Small airway dysfunction (fSAD), a precursor of COPD and asthma, was detected using radiomics feature, and the model showed satisfactory predictive performance (43). However, there is still a lack of research on the differentiation between COPD and asthma using radiomics features. Our study aimed to this research gap and the constructed radiomics model achieved good predictive performance in identification of COPD and asthma (AUCs of 0.762, 0.727, 0.753 for the training, testing, and external validation sets, respectively). Radiomics features can be categorized into three distinct types including first order features, shape features and texture features, which provide information about local and global heterogeneity of the region of interest (44). Specifically, as shown in Figure S5, the radiomics model, log-sigma-3-mm-3D_glrlm_GrayLevelNonUniformity and wavelet-LLL_glcm_Imc1 were the most discriminating signature in our study. GrayLevelNonUniformity is a second-order feature that reveals the distribution uniformity of different gray levels in the region of interest of an image and Imc represents an Information Measure of Correlation, which reflects the correlation between gray levels in the region of interest of the image. Wavelet transformation and the Laplacian of Gaussian were two spatial image filter that provide higher-order features that capture more complex texture patterns (45). Different features display the heterogeneity and complexity of lung disease in different ways, thus providing complementary value in the model (46).
Importantly, the present study considered the comprehensive information extracted from the IOS indices, QCT parameters, and radiomics features. Specifically, two selected radiomics features were further combined with three IOS indices and two QCT parameters to construct a more competent fusion Model 6, which achieved the best AUC (training set: AUC =0.959, testing set: 0.960, external validation set: AUC =0.949) compared to a single feature model, as well as the CT imaging model (QCT and radiomics features). Through multi-dimensional feature synergy, the Model 6 achieves efficient discrimination between COPD and asthma. It not only overcomes the limitations of single-parameter diagnostics but also offers a new technological pathway for the precise classification of chronic respiratory diseases.
There are several limitations in our study that merit mention. First, our study was performed at a single center and the sample size was relatively small, especially in the external validation cohort, which may introduce bias and limit the robustness and transportability of our model. Hence, future studies should include a larger sample size of patients from multiple centers for further analysis. Second, discrepancies may exist in QCT and IOS parameters between patients with severe asthma and those with mild-to-moderate asthma. In this study, the majority of asthma patients were outpatients with mild symptoms, and no comprehensive analysis was conducted on patients across different severity levels; this may limit the generalizability of the model. In addition, patients with COPD are older than those with asthma, and there are differences in gender composition between the two groups. In future studies, we will recruit a more diverse and balanced cohort of participants to develop a more comprehensive predictive model. Third, it is unlikely to explore the long-term outcomes of the patients in the present study; thus, further stratified analysis and longitudinal follow-up studies are need to demonstrate the ability of the fusion features model, including IOS indices combined with QCT parameters and radiomics features.
Conclusions
Our study has demonstrated that significant differences in PFTs, IOS, and QCT parameters were identified between the COPD group and the asthma group. Furthermore, the model constructed by integrating IOS, QCT, and radiomics features s has been demonstrated to be able to distinguish COPD from asthma, which has the potential to serve as an alternative and/or a complement tool for COPD and asthma patients who cannot tolerate routine PFTs in clinical practice.
Supplementary
The article’s supplementary files as
Acknowledgments
The authors would like to express gratitude to the Department of Radiology and Lung Function Laboratory teams for their support of this study.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study conformed with the principles outlined in the Declaration of Helsinki and its subsequent amendments. This study was approved by the ethics committee of The First Affiliated Hospital of Zhengzhou University (No. 2024-KY-1413-001) and individual consent for this retrospective analysis was waived.
Footnotes
Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1-2558/rc
Funding: This study was funded by the Science and Technology Key Program of Henan Province (grant No. 242102311069).
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1-2558/coif). W.Z. and X.L. are employees of Neusoft Medical Systems Co., Ltd., a for-profit company. However, this affiliation did not influence the study design, data analysis, or manuscript preparation. The other authors have no conflicts of interest to declare.
Data Sharing Statement
Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1-2558/dss
References
- 1.Safiri S, Carson-Chahhoud K, Noori M, et al. Burden of chronic obstructive pulmonary disease and its attributable risk factors in 204 countries and territories, 1990-2019: results from the Global Burden of Disease Study 2019. BMJ 2022;378:e069679. 10.1136/bmj-2021-069679 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Global, regional, and national deaths, prevalence, disability-adjusted life years, and years lived with disability for chronic obstructive pulmonary disease and asthma, 1990-2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet Respir Med 2017;5:691-706. 10.1016/S2213-2600(17)30293-X [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ford ES, Murphy LB, Khavjou O, et al. Total and state-specific medical and absenteeism costs of COPD among adults aged ≥ 18 years in the United States for 2010 and projections through 2020. Chest 2015;147:31-45. 10.1378/chest.14-0972 [DOI] [PubMed] [Google Scholar]
- 4.Nurmagambetov T, Kuwahara R, Garbe P. The Economic Burden of Asthma in the United States, 2008-2013. Ann Am Thorac Soc 2018;15:348-56. 10.1513/AnnalsATS.201703-259OC [DOI] [PubMed] [Google Scholar]
- 5.Yawn BP. Differential assessment and management of asthma vs chronic obstructive pulmonary disease. Medscape J Med 2009;11:20. [PMC free article] [PubMed] [Google Scholar]
- 6.Gibson PG, Simpson JL. The overlap syndrome of asthma and COPD: what are its features and how important is it? Thorax 2009;64:728-35. 10.1136/thx.2008.108027 [DOI] [PubMed] [Google Scholar]
- 7.Global Initiative for Chronic Obstructive Lung Disease. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease, 2024. Available online: www.goldcopd.org
- 8.Hikichi M, Hashimoto S, Gon Y. Asthma and COPD overlap pathophysiology of ACO. Allergol Int 2018;67:179-86. 10.1016/j.alit.2018.01.001 [DOI] [PubMed] [Google Scholar]
- 9.Ruppel GL, Enright PL. Pulmonary function testing. Respir Care 2012;57:165-75. 10.4187/respcare.01640 [DOI] [PubMed] [Google Scholar]
- 10.Culver BH, Graham BL, Coates AL, et al. Recommendations for a Standardized Pulmonary Function Report. An Official American Thoracic Society Technical Statement. Am J Respir Crit Care Med 2017;196:1463-72. 10.1164/rccm.201710-1981ST [DOI] [PubMed] [Google Scholar]
- 11.King GG, Bates J, Berger KI, et al. Technical standards for respiratory oscillometry. Eur Respir J 2020;55:1900753. 10.1183/13993003.00753-2019 [DOI] [PubMed] [Google Scholar]
- 12.Zaidan MF, Reddy AP, Duarte A. Impedance Oscillometry: Emerging Role in the Management of Chronic Respiratory Disease. Curr Allergy Asthma Rep 2018;18:3. 10.1007/s11882-018-0757-y [DOI] [PubMed] [Google Scholar]
- 13.Li LY, Yan TS, Yang J, et al. Impulse oscillometry for detection of small airway dysfunction in subjects with chronic respiratory symptoms and preserved pulmonary function. Respir Res 2021;22:68. 10.1186/s12931-021-01662-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Chen KY, Hung MH, Kao KC. Impulse oscillometry and its independent role in the diagnosis of chronic obstructive pulmonary disease. Heliyon 2024;10:e23627. 10.1016/j.heliyon.2023.e23627 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Sugawara H, Saito A, Yokoyama S, et al. A retrospective analysis of usefulness of impulse oscillometry system in the treatment of asthma. Respir Res 2020;21:226. 10.1186/s12931-020-01494-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Xu J, Sun X, Zhu H, et al. Long-term variability of impulse oscillometry and spirometry in stable COPD and asthma. Respir Res 2022;23:262. 10.1186/s12931-022-02185-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.McDonough JE, Yuan R, Suzuki M, et al. Small-airway obstruction and emphysema in chronic obstructive pulmonary disease. N Engl J Med 2011;365:1567-75. 10.1056/NEJMoa1106955 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Herth FJF, Kirby M, Sieren J, et al. The Modern Art of Reading Computed Tomography Images of the Lungs: Quantitative CT. Respiration 2018;95:8-17. 10.1159/000480435 [DOI] [PubMed] [Google Scholar]
- 19.Kirby M, Smith BM, Quantitative CT. Scan Imaging of the Airways for Diagnosis and Management of Lung Disease. Chest 2023;164:1150-8. 10.1016/j.chest.2023.02.044 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Dai Q, Zhu X, Zhang J, et al. The utility of quantitative computed tomography in cohort studies of chronic obstructive pulmonary disease: a narrative review. J Thorac Dis 2023;15:5784-800. 10.21037/jtd-23-1421 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Fan L, Seo JB, Ohno Y, et al. Chest CT in the Evaluation of COPD: Recommendations of Asian Society of Thoracic Radiology. Br J Radiol 2025;tqaf111. 10.1093/bjr/tqaf111 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Luzzi S, Pianigiani T, Dilroba A, et al. Computed tomography in severe asthma assessment: a systematic review. J Asthma 2025;62:919-28. 10.1080/02770903.2025.2460549 [DOI] [PubMed] [Google Scholar]
- 23.Hartley RA, Barker BL, Newby C, et al. Relationship between lung function and quantitative computed tomographic parameters of airway remodeling, air trapping, and emphysema in patients with asthma and chronic obstructive pulmonary disease: A single-center study. J Allergy Clin Immunol 2016;137:1413-1422.e12. 10.1016/j.jaci.2016.02.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Choi S, Haghighi B, Choi J, et al. Differentiation of quantitative CT imaging phenotypes in asthma versus COPD. BMJ Open Respir Res 2017;4:e000252. 10.1136/bmjresp-2017-000252 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Guiot J, Vaidyanathan A, Deprez L, et al. A review in radiomics: Making personalized medicine a reality via routine imaging. Med Res Rev 2022;42:426-40. 10.1002/med.21846 [DOI] [PubMed] [Google Scholar]
- 26.Zhou T, Guan Y, Lin X, et al. CT-based whole lung radiomics nomogram for identification of PRISm from non-COPD subjects. Respir Res 2024;25:329. 10.1186/s12931-024-02964-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zhou TH, Zhou XX, Ni J, et al. CT whole lung radiomic nomogram: a potential biomarker for lung function evaluation and identification of COPD. Mil Med Res 2024;11:14. 10.1186/s40779-024-00516-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Feng S, Zhang W, Zhang R, et al. The Identification and Severity Staging of Chronic Obstructive Pulmonary Disease Using Quantitative CT Parameters, Radiomics Features, and Deep Learning Features. Respiration 2026;105:259-71. 10.1159/000548595 [DOI] [PubMed] [Google Scholar]
- 29.Yang Y, Zeng N, Chen Z, et al. Multi-Layer Perceptron Classifier with the Proposed Combined Feature Vector of 3D CNN Features and Lung Radiomics Features for COPD Stage Classification. J Healthc Eng 2023;2023:3715603. 10.1155/2023/3715603 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Zhou X, Ma Y, Zhou T, et al. A computed tomography-based lung radiomics nomogram to identify acute exacerbation of chronic obstructive pulmonary disease: a multi-institutional validation study. J Thorac Dis 2025;17:7762-77. 10.21037/jtd-2025-972 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Chen X, Wang X, Huang S, et al. Study on Predicting Clinical Stage of Patients with Bronchial Asthma Based on CT Radiomics. J Asthma Allergy 2024;17:291-303. 10.2147/JAA.S448064 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Graham BL, Steenbruggen I, Miller MR, et al. Standardization of Spirometry 2019 Update. An Official American Thoracic Society and European Respiratory Society Technical Statement. Am J Respir Crit Care Med 2019;200:e70-88. 10.1164/rccm.201908-1590ST [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Zhang W, Zhao Y, Tian Y, et al. Early Diagnosis of High-Risk Chronic Obstructive Pulmonary Disease Based on Quantitative High-Resolution Computed Tomography Measurements. Int J Chron Obstruct Pulmon Dis 2023;18:3099-114. 10.2147/COPD.S436803 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Wu Y, Kang K, Han C, et al. A blind randomized validated convolutional neural network for auto-segmentation of clinical target volume in rectal cancer patients receiving neoadjuvant radiotherapy. Cancer Med 2022;11:166-75. 10.1002/cam4.4441 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Mishima M. Physiological differences and similarities in asthma and COPD--based on respiratory function testing. Allergol Int 2009;58:333-40. 10.2332/allergolint.09-RAI-0131 [DOI] [PubMed] [Google Scholar]
- 36.Kanda S, Fujimoto K, Komatsu Y, et al. Evaluation of respiratory impedance in asthma and COPD by an impulse oscillation system. Intern Med 2010;49:23-30. 10.2169/internalmedicine.49.2191 [DOI] [PubMed] [Google Scholar]
- 37.Kim SR, Park KH, Son NH, et al. Application of Impulse Oscillometry in Adult Asthmatic Patients With Preserved Lung Function. Allergy Asthma Immunol Res 2020;12:832-43. 10.4168/aair.2020.12.5.832 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Hogg JC, McDonough JE, Suzuki M. Small airway obstruction in COPD: new insights based on micro-CT imaging and MRI imaging. Chest 2013;143:1436-43. 10.1378/chest.12-1766 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Kitaguchi Y, Yasuo M, Hanaoka M. Comparison of pulmonary function in patients with COPD, asthma-COPD overlap syndrome, and asthma with airflow limitation. Int J Chron Obstruct Pulmon Dis 2016;11:991-7. 10.2147/COPD.S105988 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Huang CH, Chou KT, Perng DW, et al. Using Machine Learning with Impulse Oscillometry Data to Develop a Predictive Model for Chronic Obstructive Pulmonary Disease and Asthma. J Pers Med 2024;14:398. 10.3390/jpm14040398 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Verleden SE, Hendriks JMH, Snoeckx A, et al. Small Airway Disease in Pre-Chronic Obstructive Pulmonary Disease with Emphysema: A Cross-Sectional Study. Am J Respir Crit Care Med 2024;209:683-92. 10.1164/rccm.202301-0132OC [DOI] [PubMed] [Google Scholar]
- 42.Moslemi A, Kontogianni K, Brock J, et al. Differentiating COPD and asthma using quantitative CT imaging and machine learning. Eur Respir J 2022;60:2103078. 10.1183/13993003.03078-2021 [DOI] [PubMed] [Google Scholar]
- 43.Cui S, Shu Z, Ma Y, et al. A novel computed tomography radiomic nomogram for early evaluation of small airway dysfunction development. Front Med (Lausanne) 2022;9:944294. 10.3389/fmed.2022.944294 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Zhu Z, Zhao S, Li J, et al. Development and application of a deep learning-based comprehensive early diagnostic model for chronic obstructive pulmonary disease. Respir Res 2024;25:167. 10.1186/s12931-024-02793-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Chen M, Copley SJ, Viola P, et al. Radiomics and artificial intelligence for precision medicine in lung cancer treatment. Semin Cancer Biol 2023;93:97-113. 10.1016/j.semcancer.2023.05.004 [DOI] [PubMed] [Google Scholar]
- 46.Amudala Puchakayala PR, Sthanam VL, Nakhmani A, et al. Radiomics for Improved Detection of Chronic Obstructive Pulmonary Disease in Low-Dose and Standard-Dose Chest CT Scans. Radiology 2023;307:e222998. 10.1148/radiol.222998 [DOI] [PMC free article] [PubMed] [Google Scholar]