A medical diagnostic tool based on radial basis function classifiers and evolutionary simulated annealing.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Medical Informatics

Background:

• Medical records contain vast data valuable for decision support systems.
• Developing accurate, data-driven diagnostic tools is crucial for healthcare.

Purpose of the Study:

• To present a novel methodology for designing data-driven medical diagnostic tools using neural network classifiers.
• To adapt the Non-Symmetric Fuzzy Means (NSFM) algorithm for Radial Basis Function (RBF) classifiers and optimize them using Evolutionary Simulated Annealing (ESA).

Main Methods:

• Utilized Radial Basis Function (RBF) neural network architecture.
• Employed the Non-Symmetric Fuzzy Means (NSFM) training algorithm.
• Integrated the Evolutionary Simulated Annealing (ESA) technique for optimizing RBF models, enhancing the NSFM algorithm's ability to escape local minima.

Main Results:

• The ESA-NSFM algorithm demonstrated superior performance over standard RBF training and Support Vector Machines (SVMs) in accuracy and Matthews Correlation Coefficient (MCC) across nine benchmark datasets.
• The ESA-NSFM algorithm achieved statistically significant improvements in accuracy and MCC.
• The proposed approach exhibited faster computational times compared to SVMs.

Conclusions:

• The ESA-NSFM algorithm offers a robust, generic method for knowledge extraction from diverse medical records.
• This approach enhances diagnostic accuracy, outperforming existing classifier training methods.