Enhancing the Diagnosis of Behçet's Disease Using Machine Learning: A Comparative Study on Clinical Data From Saudi...

Area of Science:

• Rheumatology
• Computational Biology
• Medical Informatics

Background:

• Behçet's disease (BD) presents diagnostic challenges due to its rarity and diverse symptoms within rheumatic immune diseases.
• Current understanding of BD's pathophysiology and distinguishing features remains limited, hindering accurate and timely diagnosis.

Purpose of the Study:

• To explore clinical and laboratory features that differentiate Behçet's disease from other rheumatic conditions.
• To investigate the utility of machine learning algorithms in identifying key diagnostic markers for BD.

Main Methods:

• Collected clinical data from 148 patients (76 BD, 72 rheumatoid arthritis).
• Applied machine learning algorithms including Random Forest (RF), XGBoost, and Support Vector Machines (SVMs).
• Utilized feature importance analysis including permutation-based methods and SHAP values.

Main Results:

• Random Forest model achieved 96.7% accuracy and an Area Under the Curve (AUC) of 1.0; XGBoost achieved an AUC of 0.9985.
• Feature importance analysis highlighted oral ulcers as the most critical distinguishing feature, aligning partially with established diagnostic criteria.
• Machine learning models demonstrated strong performance in distinguishing BD from other rheumatic diseases.

Conclusions:

• This study provides an exploratory framework for understanding BD's unique characteristics and feature interactions.
• Machine learning approaches show promise for developing future diagnostic support systems for Behçet's disease.
• Identifying key diagnostic features like oral ulcers can aid in earlier recognition and management of BD.