Improving thyroid disorder diagnosis via innovative stacking ensemble learning model

Area of Science:

• Computational biology and bioinformatics
• Medical informatics and artificial intelligence

Background:

• Thyroid disorders, including hypothyroidism, hyperthyroidism, and nodules, are globally prevalent, affecting millions.
• Untreated thyroid conditions can lead to severe health complications, underscoring the need for accurate diagnosis.
• Timely and precise diagnosis is essential for effective management and treatment of thyroid diseases.

Purpose of the Study:

• To develop a comprehensive machine learning (ML) technique for the accurate diagnosis of thyroid disorders.
• To evaluate the efficacy of various ML algorithms and an ensemble approach for thyroid condition detection.

Main Methods:

• Data preprocessing included handling missing values, encoding categorical features, and feature selection.
• The synthetic minority over-sampling technique (SMOTE) addressed class imbalance.
• Five ML algorithms (logistic regression, SVM, decision tree, random forest, gradient boosting) and a stacking ensemble method were employed.

Main Results:

• A 10-fold cross-validation was used for robust model evaluation and to prevent overfitting.
• The proposed stacking ensemble model achieved a diagnostic accuracy of 99.86%.
• The ensemble approach significantly outperformed individual ML models in diagnostic performance.

Conclusions:

• Machine learning, particularly ensemble methods, demonstrates high capability in diagnosing thyroid disorders.
• The study highlights the potential of ML for enhancing the accuracy and timeliness of thyroid disease diagnosis.
• This approach can aid clinicians in better managing patients with various thyroid conditions.