Choquet fuzzy integral-based classifier ensemble technique for COVID-19 detection.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools for effective containment.
• Early detection of COVID-19 is crucial for mitigating community spread and reducing mortality.
• Traditional diagnostic methods can be time-consuming, highlighting the need for advanced computational approaches.

Purpose of the Study:

• To develop and evaluate a robust classifier ensemble for early COVID-19 detection using chest X-ray images.
• To leverage transfer learning with pre-trained convolutional neural network (CNN) models for improved feature extraction.
• To enhance classification accuracy by integrating base model predictions using Choquet fuzzy integral.

Main Methods:

• Utilized transfer learning with InceptionV3, DenseNet121, and VGG19 CNN models pre-trained on large datasets.
• Employed a classifier ensemble strategy combining predictions from multiple CNN models via Choquet fuzzy integral.
• Trained and validated the model on publicly available chest X-ray datasets (IEEE, Kaggle).
• Determined fuzzy-membership values based on individual classifier validation accuracy.

Main Results:

• Achieved high performance metrics: 99.00% average recall, 99.00% precision, 99.00% F-score, and 99.02% accuracy on combined datasets.
• Demonstrated strong generalization capabilities with 99.05% test accuracy on an unseen external dataset (CMSC-678-ML-Project GitHub).
• Outperformed existing ensemble methods and state-of-the-art approaches in COVID-19 classification.

Conclusions:

• The proposed Choquet fuzzy integral-based ensemble classifier offers a highly accurate and reliable method for COVID-19 detection from chest X-rays.
• This approach effectively addresses the challenge of limited COVID-19 data for training deep learning models.
• The method shows significant potential for clinical application in early pandemic detection and management.