Development of Rheumatoid Arthritis Classification from Electronic Image Sensor Using Ensemble Method.

Area of Science:

• Medical informatics
• Machine learning
• Rheumatology

Background:

• Rheumatoid arthritis (RA) is a chronic autoimmune disease causing progressive musculoskeletal damage, fatigue, and reduced physical function.
• Wearable sensors and imaging technologies are increasingly used for RA data acquisition.
• Accurate classification of RA and orthopaedic conditions is crucial for effective patient management.

Purpose of the Study:

• To evaluate the effectiveness of ensemble machine learning methods for classifying Rheumatoid Arthritis (RA) and orthopaedic datasets.
• To compare the performance of different ensemble algorithms (bagging, Adaboost, random subspace) using k-NN and Random Forest as base learners.
• To assess classification accuracy using various performance metrics and cross-validation techniques.

Main Methods:

• Utilized RA and orthopaedic datasets for classification tasks.
• Implemented ensemble classifiers: bagging, Adaboost, and random subspace.
• Employed k-NN (K-nearest neighbours) and Random Forest (RF) as base learners.
• Evaluated performance using holdout and 10-fold cross-validation, with metrics including precision, recall, F-measure, and ROC curve.

Main Results:

• The random subspace classifier with k-NN achieved 97.50% accuracy on the RA dataset (Dataset 1).
• The bagging classifier with RF achieved 94.84% accuracy on the orthopaedic dataset (Dataset 2).
• Ensemble methods demonstrated substantial improvements in classification accuracy compared to base learners.

Conclusions:

• Ensemble machine learning techniques offer a powerful approach for classifying complex medical conditions like RA and orthopaedic disorders.
• The choice of ensemble method and base learner significantly impacts classification performance.
• Further research into ensemble methods can enhance diagnostic accuracy and patient outcomes in rheumatology and orthopaedics.