Abstract:Background: Idiopathic Pulmonary Fibrosis (IPF) is a progressive fibrotic lung disease. However, the field of quantitative CT scan analysis in conjunction with pulmonary function test for IPF patients remains relatively understudied. In this study, we evaluated the diagnostic value of features derived high-resolution computed tomography (HRCT) for patients with IPF and correlated them with pulmonary function tests. Methods: We retrospectively analyzed the chest HRCT images and pulmonary function test results of 52 patients with IPF during the same period (1 week) and selected 52 healthy individuals, matched for sex, age, and body mass index (BMI) and with normal chest HRCT as controls. HRCT scans were performed using a Philips 256-row Brilliance iCT scanner with standardized parameters. Lung function tests were performed using a Jaeger volumetric tracer for forced vital capacity (FVC), total lung capacity (TLC), forced expiratory volume in first second (FEV1), FEV1/FVC, carbon monoxide diffusing capacity (DLCO), and maximum ventilation volume (MVV) metrics. CT quantitative analysis, including tissue segmentation and threshold-based quantification of lung abnormalities, was performed using 3D-Slicer software to calculate the percentage of normal lung areas (NL%), percentage of ground-glass opacity areas (GGO%), percentage of fibrotic area (F%) and abnormal lesion area percentage (AA%). Semi-quantitative analyses were performed by two experienced radiologists to assess disease progression. The aortic-to-sternal distance (ASD) was measured on axial images as a standardized parameter. Spearman or Pearson correlation analysis and multivariate stepwise linear regression were used to analyze the relationship between the data in each group, and the ROC curve was used to determine the optimal quantitative CT metrics for identifying IPF and controls. Results: ROC curve analysis showed that F% distinguished the IPF patient group from the control group with the largest area under the curve (AUC) of 0.962 (95% confidence interval: 0.85-0.96). Additionally, with F% = 4.05% as the threshold, the Youden's J statistic was 0.827, with a sensitivity of 92.3% and a specificity of 90.4%. The ASD was significantly lower in the late stage of progression than in the early stage (t = 5.691, P < 0.001), with a mean reduction of 2.45% per month. Quantitative CT indices correlated with all pulmonary function parameters except FEV1/FVC, with the highest correlation coefficients observed for F% and TLC%, FEV1%, FVC%, MVV% (r = - 0.571, - 0.520, - 0.521, - 0.555, respectively, all P-values < 0.001), and GGO% was significantly correlated with DLCO% (r = - 0.600, P < 0.001). Multiple stepwise linear regression analysis showed that F% was the best predictor of TLC%, FEV1%, FVC%, and MVV% (R2 = 0.301, 0.301, 0.300, and 0.302, respectively, all P-values < 0.001), and GGO% was the best predictor of DLCO% (R2 = 0.360, P < 0.001). Conclusions: Quantitative CT analysis can be used to diagnose IPF and assess lung function impairment. A decrease in the ASD may indicate disease progression.

Developing the Lung Graph-Based Machine Learning Model for Identification of Fibrotic Interstitial Lung Diseases

Application of HRCT-based radiomics to predict interstitial lung disease for juvenile dermatomyositis

Computer-aided identification of interstitial lung disease based on computed tomography.

Analysis of Validation Performance of a Machine Learning Classifier in Interstitial Lung Disease Cases Without Definite or Probable Usual Interstitial Pneumonia Pattern on CT Using Clinical and Pathology-Supported Diagnostic Labels

Identification of Idiopathic Pulmonary Fibrosis and Prediction of Disease Severity via Machine Learning Analysis of Comprehensive Metabolic Panel and Complete Blood Count Data

Severity-stratification of interstitial lung disease by deep learning enabled assessment and quantification of lesion indicators from HRCT images

Diagnosis of Idiopathic Pulmonary Fibrosis in High-Resolution Computed Tomography Scans Using a Combination of Handcrafted Radiomics and Deep Learning

Development and validation of a lung graph-based machine learning model to predict acute pulmonary thromboembolism on chest noncontrast computed tomography

Quantitative CT texture-based method to predict diagnosis and prognosis of fibrosing interstitial lung disease patterns

Development of a scoring system with multidimensional markers for fibrosing interstitial lung disease

Novel 3D-based deep learning for classification of acute exacerbation of idiopathic pulmonary fibrosis using high-resolution CT

Interstitial lung disease diagnosis and prognosis using an AI system integrating longitudinal data

Quantitative CT analysis of idiopathic pulmonary fibrosis and correlation with lung function study

Computed Tomography-Based Deep Learning Model for Assessing the Severity of Patients With Connective Tissue Disease-Associated Interstitial Lung Disease.

Quantitative Assessment of Interstitial Lung Disease Based on RDNet Convolutional Network.

Comprehensive Analysis of Fibroblast Activation Protein Expression in Interstitial Lung Diseases

Comparing radiomics models with different inputs for accurate diagnosis of significant fibrosis in chronic liver disease

Prognostic value of deep learning–based fibrosis quantification on chest CT in idiopathic pulmonary fibrosis

Comparison of deep-learning and radiomics-based machine-learning methods for the identification of chronic obstructive pulmonary disease on low-dose computed tomography images

Non-contrast computed tomography-based radiomics for staging of connective tissue disease-associated interstitial lung disease

Automatic Diagnosis of Significant Liver Fibrosis From Ultrasound B-Mode Images Using a Handcrafted-Feature-Assisted Deep Convolutional Neural Network