A machine learning-based classification model for interstitial lung disease in rheumatoid arthritis.
Mingyao Li1, Qiaoli Wang2, Junfeng He1
1Department of Rheumatology and Immunology, Deyang People's Hospital, Deyang, Sichuan, China.
Frontiers in Medicine
|June 1, 2026
Summary
Machine learning models effectively classify rheumatoid arthritis-associated interstitial lung disease (RA-ILD) using routine clinical data. The CatBoost and decision tree models show promise for risk stratification in primary care settings.
Area of Science:
- Rheumatology
- Pulmonology
- Medical Informatics
Background:
- Rheumatoid arthritis-associated interstitial lung disease (RA-ILD) poses a significant clinical challenge.
- Early and accurate classification of RA-ILD is crucial for timely intervention and improved patient outcomes.
- Developing predictive models using accessible parameters can aid in primary hospital settings.
Purpose of the Study:
- To develop and validate a preliminary classification and diagnostic model for RA-ILD.
- To utilize routine clinical and laboratory parameters for risk assessment.
- To construct a practical tool for use in primary hospital settings through machine learning.
Main Methods:
- Retrospective collection of clinical data from rheumatoid arthritis (RA) patients.
- Division of patients into RA and RA-ILD groups, followed by random splitting into training and validation sets.
- Application of LASSO regression to identify significant features, followed by construction of five machine learning models (CatBoost, logistic regression, SVM, decision tree, random forest).
- Performance evaluation using AUC, accuracy, precision, recall, and F1 score; SHAP framework for feature importance analysis.
Main Results:
- Out of 410 RA patients, 100 (24.39%) had RA-ILD.
- Seven key features were identified: age, smoking history, LYMPH, LDH, RF, CA125, and CA199.
- The CatBoost model achieved the highest AUC (0.784) and lowest Brier score (0.158) in the validation set.
- The decision tree model showed comparable AUC (0.783) and the highest recall (0.653) and F1-score (0.603).
- SHAP analysis highlighted CA199, CA125, and age as primary predictors in the CatBoost model.
Conclusions:
- CatBoost and decision tree models demonstrate comparable and favorable performance for RA-ILD classification.
- These models show potential for clinical application in RA-ILD risk stratification.
- Further external validation is recommended to confirm the generalizability of the findings.
Related Concept Videos
01:26Asthma-II: Pathophysiology and Classification
Asthma is a prevalent chronic respiratory condition marked by inflammation and hyperresponsiveness of the airways. Its pathophysiology involves complex interactions among inflammatory pathways, immune responses, and neural mechanisms.
Additionally, environmental and genetic factors play crucial roles in determining an individual's susceptibility to asthma and the severity of their condition.
Critical processes in asthma pathophysiology include:
Additionally, environmental and genetic factors play crucial roles in determining an individual's susceptibility to asthma and the severity of their condition.
Critical processes in asthma pathophysiology include:
01:17Classification of Illness
The meaning of illness is individualized to each person who experiences an alteration in health. In contrast, disease is a medical term indicating a pathological change in the structure and function of the body or mind. It is a condition that has specific symptoms and boundaries.
An illness is a response to a disease in which the person's level of functioning is changed compared with a previous level. The general classification of illness includes acute and chronic.
Acute illness is severe and...
An illness is a response to a disease in which the person's level of functioning is changed compared with a previous level. The general classification of illness includes acute and chronic.
Acute illness is severe and...