Vertebral Column Pathology Diagnosis Using Ensemble Strategies Based on Supervised Machine Learning Techniques.

Area of Science:

• Bioinformatics and machine learning applications in medical diagnostics.
• Development of advanced computational methods for healthcare.

Background:

• Automatic medical diagnosis using machine learning (ML) requires rigorous performance evaluation.
• Existing ML methods for classifying biomedical characteristics face challenges in accuracy and reliability.

Purpose of the Study:

• To propose and evaluate novel Voting and Stacking (VC and SC) ensemble strategies for enhanced automatic diagnosis of vertebral column orthopedic illnesses.
• To improve the efficacy of traditional baseline classifiers through auto-tuning and ensemble techniques.

Main Methods:

• Developed ensemble strategies (VC and SC) using diverse auto-tuned supervised machine learning techniques.
• Selected top-performing classifiers (kNN, NB, LR, LDA, QDA, SVM, ANN, DT) using grid-search K-Fold cross-validation for hyperparameter auto-tuning.
• Compared ensemble strategy performance against individual auto-tuned baseline classifiers using accuracy, precision, recall, F1-score, and ROC-ACU metrics.

Main Results:

• The VC ensemble strategy showed performance comparable to the best individual baseline classifier (kNN).
• The SC ensemble strategy, incorporating all baseline classifiers, achieved over 95% in all evaluated metrics.
• Analysis identified specific misclassified disease elements, highlighting classifier reliability.

Conclusions:

• Ensemble strategies, particularly the SC method, offer a significant improvement for the automatic diagnosis of vertebral column diseases.
• The proposed SC ensemble strategy is highly suitable for clinical application due to its superior performance and reliability.