COVID-19 Recognition Using Ensemble-CNNs in Two New Chest X-ray Databases.

Area of Science:

• Medical Imaging
• Computer Vision
• Machine Learning

Background:

• COVID-19 detection from X-rays is challenging due to limited labeled data and fragmented research.
• Existing studies lack unified datasets and standardized evaluation protocols for COVID-19 X-ray analysis.

Purpose of the Study:

• To address data scarcity and lack of standardization in COVID-19 X-ray recognition.
• To develop and evaluate deep learning models for accurate COVID-19 detection using X-ray images.
• To create and release benchmark datasets for COVID-19 X-ray analysis.

Main Methods:

• Creation of two new COVID-19 X-ray datasets: a three-class and a five-class dataset.
• Evaluation of various deep learning architectures on the proposed datasets.
• Development and implementation of an Ensemble of Convolutional Neural Networks (Ensemble-CNNs) approach.

Main Results:

• The proposed Ensemble-CNNs model achieved high recognition accuracies of 100% and 98.1% in the three-class and five-class scenarios, respectively.
• Overall recognition accuracies for the Ensemble-CNNs were 75.23% (three-class) and 81.0% (five-class).
• The developed datasets are made publicly available for benchmarking.

Conclusions:

• The proposed Ensemble-CNNs approach demonstrates superior performance in COVID-19 X-ray recognition compared to individual deep learning architectures.
• The publicly available datasets and findings provide a valuable resource for advancing research in AI-driven medical diagnostics for COVID-19.
• This work facilitates standardized evaluation and encourages further development in the field of computer-aided detection of COVID-19 from X-rays.