Advance Warning Methodologies for COVID-19 Using Chest X-Ray Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Coronavirus disease 2019 (COVID-19) emerged in December 2019, creating an urgent need for early diagnostic tools.
• Early-stage COVID-19 detection from chest X-rays is challenging as infection signs are often subtle.

Purpose of the Study:

• To evaluate state-of-the-art Machine Learning (ML) techniques for early COVID-19 detection using chest X-ray images.
• To propose and assess a novel compact classifier, Convolutional Support Estimator Network (CSEN), for scarce-data classification tasks.
• To introduce the Early-QaTa-COV19 dataset for benchmarking early-stage COVID-19 pneumonia detection.

Main Methods:

• Evaluation of compact classifiers and deep learning approaches for COVID-19 detection.
• Implementation and testing of the Convolutional Support Estimator Network (CSEN) on the Early-QaTa-COV19 dataset.
• Comparison of CSEN performance against other deep learning models like DenseNet-121.

Main Results:

• The CSEN model achieved high performance with over 97% sensitivity and 95.5% specificity.
• DenseNet-121 demonstrated strong results among deep networks, yielding 95% sensitivity and 99.74% specificity.
• The Early-QaTa-COV19 dataset comprises 1065 early-stage COVID-19 pneumonia samples and 12544 control samples.

Conclusions:

• Machine learning, particularly CSEN, shows significant potential for the early detection of COVID-19 from chest X-rays.
• The developed Early-QaTa-COV19 dataset provides a valuable resource for training and evaluating early-stage COVID-19 detection models.
• Accurate early diagnosis systems are crucial for managing the global health impact of COVID-19.