JMIR Publications

ABSTRACT

Background:

Coal workers’ pneumoconiosis (CWP) is the most prevalent occupational disease that causes irreversible lung damage. Early prediction of CWP is the key to blocking the irreversible process of pulmonary fibrosis. The prediction of CWP based on imaging data and biomarker detection is constrained due to high cost and poor convenience.

Objective:

We utilize easily detectable clinical data to construct a prediction model for CWP through machine learning (ML) methods.

Methods:

A ML prediction model based on small samples and easily accessible clinical data was developed in this study. Firstly, a three-dimensional feature space based on occupational exposure history, lung function parameters, and blood indicators was constructed, and data imbalance processing and performance comparison analysis were conducted using six algorithms including LightGBM, RF, XGBoost, CatBoost, SVM, and LR. Then, through the optimization strategy of Grid Search and Optuna algorithm, the best performing model was selected.

Results:

The results showed that the LightGBM model optimized by Optuna performed the best in comprehensive evaluation indicators, with accuracy, recall, F1 score, and AUC value increased by 0.38%, 3.7%, 1.34%, and 0.34%, respectively, compared to the corresponding indicators of LightGBM. Finally, the SHAP interpretable method was used to analyze the contributions of job types, lung function indicators, and other blood indicators to the model.

Conclusions:

The research results have confirmed the potential of combining simple multidimensional features with ML algorithms for predicting CWP, and provided new ideas for early diagnosis and intervention of CWP patients.