A Novel Machine Learning Approach to Assist Early Diagnosis of Diffuse Panbronchiolitis.
Authors
Affiliations (10)
Affiliations (10)
- Department of Internal Medicine and Research Center for Pulmonary Disorders, Jeonbuk National University Medical School, Jeonju, Korea.
- Research Institute of Clinical Medicine of Jeonbuk National University-Biomedical Research Institute of Jeonbuk National University Hospital, Jeonju, Korea.
- Department of Internal Medicine, Sinchon Severance Hospital, Seoul, Korea.
- Department of Otorhinolaryngology-Head and Neck Surgery, Jeonbuk National University Medical School, Jeonju, Korea.
- Department of Medical Informatics, Jeonbuk National University Medical School, Jeonju, Korea.
- Department of Radiology, Jeonbuk National University Medical School, Jeonju, Korea.
- Respiratory Drug Development Research Institute, Jeonbuk National University Medical School, Jeonju, Korea. [email protected].
- Respiratory Drug Development Research Institute, Jeonbuk National University Medical School, Jeonju, Korea.
- Laboratory of Respiratory Immunology and Infectious Diseases, Korea Zoonosis Research Institute, Jeonbuk National University, Iksan, Korea. [email protected].
- Respiratory Drug Development Research Institute, Jeonbuk National University Medical School, Jeonju, Korea. [email protected].
Abstract
Diffuse panbronchiolitis (DPB) is a rare and progressive inflammatory lung disease affecting the small airways; however, it is often misdiagnosed as other respiratory conditions, such as nontuberculous mycobacterial infection or bronchiectasis. This study aimed to apply machine learning (ML) algorithms to improve early diagnostic accuracy for DPB. ML algorithms were applied using clinical, laboratory, and radiological data from 99 patients with suspected DPB. Patients were categorized into two groups based on established diagnostic criteria and major diagnostic criteria for DPB without impaired lung function. Seven ML models were evaluated. The least absolute shrinkage and selection operator regression model demonstrated the highest predictive accuracy. The analysis identified two key diagnostic factors, allergic rhinitis and the presence of macronodules on computed tomography scans, both of which were strongly associated with DPB. These results highlight the first application of ML in diagnosing DPB and underscore the significance of allergic rhinitis and macronodules as critical indicators for early detection. Incorporating ML techniques into clinical practice could improve the diagnostic accuracy and efficiency for rare diseases such as DPB. Further research involving larger patient datasets is recommended to validate these results and refine the diagnostic criteria for DPB.