![]() |
![]() |
| Korean J Med > Volume 100(5); 2025 > Article |
|
FUNDING
JWS was supported by the Bio & Medical Technology Development Program (NRF-2022M3A9E4082647) of the National Research Foundation of Korea (NRF) funded by the Ministry of Science & ICT, Republic of Korea, and was also supported by the Korea National Institute of Health research project (2024ER090500) and the Korea Environment Industry & Technology Institute through Core Technology Development Project for Environmental Diseases Prevention and Management Program funded by the Korea Ministry of Environment (RS-2022-KE02197), Republic of Korea.
AUTHOR CONTRIBUTIONS
Ju Hyun Oh contributed to the conceptualization, literature search, drafting, and revision of the manuscript. Jin Woo Song supervised the study, provided conceptual guidance, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript.
| Study | ILD type | Purpose | Methods | Result |
|---|---|---|---|---|
| Kim et al. [17] (2018) | Various (healthy subjects, OP, UIP, AIP) | Improve ILD pattern classification accuracy with deeper CNN | CNN with six layers | Classification accuracy improved from 81.27% to 95.12% |
| Transfer learning to enhance performance | ||||
| Choe et al. [16] (2022) | Four ILD subtype (UIP, NSIP, OP, CHP) | Assist radiologists using CBIR-based DL tool | DL-based CBIR model | CBIR-assisted DL model improved reader accuracy |
| Before CBIR, 46.1%; after CBIR, 60.9%; p < 0.001 | ||||
| Walsh et al. [27] (2018) | IPF | Classify IPF based on ATS/ERS criteria | DL trained on IPF criteria from radiologists’ labels | DL accuracy 73.3%, outperforming average radiologist (70.7%) |
| Salisbury et al. [28] (2017) | IPF | Predict FVC decline and prognosis | Adaptive multi-feature method and Bayesian classifier | Fibrotic volume correlated with FVC decline |
| Bayesian classifier AUC not reported | ||||
| Chassagnon et al. [23] (2020) | SSc-ILD | Predict FVC decline using DL and lung shrinkage detection | Elastic registration and DL classifier | Elastic registration + DL predicted FVC decline with 83% accuracy |
| Jacob et al. [34] (2016) | IPF, CTD-ILD | Predict prognosis from radiomics features using ML | CALIPER with radiomics + ML algorithm | CALIPER vessel volume associated with FVC decline |
| Jacob et al. [49] (2018) | IPF, CTD-ILD, | Predict prognosis from radiomics features using ML | CALIPER with radiomics + ML algorithm | CALIPER vessel volume associated with FVC decline |
| Ungprasert et al. [13] (2017) | IIM-ILD | Correlation between CALIPER-derived quantitative parameters and pulmonary function | CALIPER software system applied to CT imaging and correlated with baseline and 1-year pulmonary function | CALIPER fibrosis extent correlated with DLCO %pred (r = -0.51, p = 0.002), TLC, and O2 saturation |
| Humphries et al. [50] (2024) | Various ILDs | To classify UIP patterns and to evaluate its prognostic significance | Using a 3D CNN trained on expert-labeled HRCT images to classify UIP vs. non-UIP and validated prognostic relevance | The model achieved an AUC of 0.87 and pre-dicted UIP scores correlated with mortality (HR, 2.3; p < 0.001) |
| Jacob et al. [19] (2017) | HP | Predict outcome using automated CT stratification in HP | Automated CT stratification (radiomics + classifier) | Automated CT score predicted survival (HR, 2.7; p < 0.01) |
| Maldonado et al. [20] (2014) | IPF | Automated radiologic pattern quantification for survival prediction | Automated pattern quantification using CALIPER software system | Pattern quantification stratified survival |
| Huang et al. [21] (2024) | IPF | Classify acute exacerbation of IPF using 3D deep learning | 3D CNN-based deep learning model with HRCT | 3D CNN achieved AUROC 0.92 in classifying IPF acute exacerbation |
AI, artificial intelligence; CT, computed tomography; ILD, interstitial lung disease; OP, organizing pneumonia; UIP, usual interstitial pneumonia; AIP, acute interstitial pneumonia; CNN, convolutional neural network; NSIP, nonspecific interstitial pneumonia; CHP, chronic hypersensitivity pneumonitis; CBIR, content-based image retrieval; DL, deep learning; IPF, idiopathic pulmonary fibrosis; ATS/ERS, American Thoracic Society/European Respiratory Society; FVC, forced vital capacity; AUC, area under the curve; SSc-ILD, systemic sclerosis-associated interstitial lung disease; CTD-ILD, connective tissue disease-associated interstitial lung disease; ML, machine learning; CALIPER, Computer-Aided Lung Informatics for Pathology Evaluation and Rating; IIM, idiopathic inflammatory myopathy; DLCO, diffusing capacity for carbon monoxide; TLC, total lung capacity; HRCT, high-resolution computed tomography; HR, hazard ratio; HP, hypersensitivity pneumonitis; AUROC, area under the receiver operating characteristic curve.

![]() |
![]() |