A computer aided diagnosis system for thyroid disease using extreme learning machine.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Computational Biology

Background:

• Thyroid disease diagnosis requires accurate and efficient tools.
• Existing computer-aided diagnosis (CAD) systems face challenges in performance and speed.
• Feature selection and model optimization are critical for enhancing diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate an effective and efficient CAD system for thyroid disease diagnosis.
• To integrate Principle Component Analysis (PCA) for feature reduction and Extreme Learning Machine (ELM) for classification.
• To optimize ELM parameters, specifically the number of hidden neurons, for improved diagnostic performance.

Main Methods:

• A three-stage CAD system was designed, starting with PCA for dimension reduction and discriminative feature extraction.
• An ELM classifier was employed for predictive model construction, with an experimental method to determine the optimal number of hidden neurons.
• The PCA-ELM system was evaluated on a thyroid disease dataset from the UCI machine learning repository using 10-fold cross-validation.

Main Results:

• The proposed PCA-ELM system achieved a mean accuracy of 97.73% and a maximum accuracy of 98.1% in thyroid disease diagnosis.
• The system demonstrated superior performance compared to other related methods.
• PCA-ELM exhibited significantly faster processing times than Support Vector Machine (SVM) based CAD systems.

Conclusions:

• The developed PCA-ELM system is a powerful and efficient tool for thyroid disease diagnosis.
• The integration of PCA and ELM offers excellent diagnostic performance and speed.
• This approach holds promise for improving clinical decision-making in thyroid disease management.