Randomized Explainable Machine Learning Models for Efficient Medical Diagnosis.

Area of Science:

• Medical Diagnostics
• Artificial Intelligence
• Machine Learning

Background:

• Deep learning models excel in medical diagnostics but face challenges due to high computational costs and lack of transparency ('black-box' problem).
• These limitations hinder adoption in time-critical and resource-constrained healthcare environments.

Purpose of the Study:

• To investigate the efficacy of randomized machine learning models, specifically Extreme Learning Machines (ELMs) and Random Vector Functional Link (RVFL) networks, for medical diagnostics.
• To integrate Explainable AI (XAI) techniques, including Local Interpretable Model-agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP), to enhance model interpretability.

Main Methods:

• Implementation of ELMs and RVFL networks, incorporating stochasticity for reduced computational complexity and training time.
• Application of LIME and SHAP to elucidate the decision-making processes of the randomized models.
• Performance evaluation using datasets for genitourinary cancers and coronary artery disease.

Main Results:

• RVFL models demonstrated superior performance compared to traditional deep learning approaches.
• Achieved 88.29% accuracy with 6.22s computation for genitourinary cancers and 81.64% accuracy with 0.0308s computation for coronary artery disease.
• The integrated XAI techniques successfully provided interpretable insights into model predictions.

Conclusions:

• Randomized machine learning models, particularly RVFL, offer a promising avenue for efficient and transparent medical diagnostics.
• These models present a viable solution for overcoming the computational and interpretability challenges of deep learning in healthcare.
• The study advocates for the adoption of accessible and interpretable AI in medical diagnosis to improve treatment outcomes.