Advancing thyroid care: An accurate trustworthy diagnostics system with interpretable AI and hybrid machine learning...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Endocrinology

Background:

• Thyroid disease prevalence is increasing globally, contributing to mortality.
• Machine learning (ML) offers potential for early detection and treatment of thyroid conditions.
• Interpretable predictions are crucial for patient and stakeholder trust in medical AI.

Purpose of the Study:

• To develop an interpretable thyroid classification model using explainable AI (XAI).
• To investigate the contribution of predictive features in thyroid disease classification.
• To enhance clinical decision-making and patient care through accurate and understandable AI predictions.

Main Methods:

• Data preprocessing and balancing using SMOTE-ENN.
• Hybrid classifiers (RDKVT, RDKST) trained on selected features (Univariate, Information Gain).
• Shapley Additive Explanation (SHAP) for feature importance analysis.

Main Results:

• The RDKST classifier achieved 98.98% accuracy using Information Gain features.
• Key features influencing outcomes included T3, TT4, TSH, FTI, and T3_measured.
• The model effectively balanced classification performance with outcome interpretability.

Conclusions:

• The developed interpretable ML model shows high accuracy in thyroid disease classification.
• SHAP analysis provides insights into feature contributions, enhancing model transparency.
• This approach can improve clinical decision support and patient management for thyroid disorders.