Chest X-ray Classification for the Detection of COVID-19 Using Deep Learning Techniques.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• The COVID-19 pandemic strained global health systems, necessitating rapid and accurate diagnostic tools.
• Chest X-rays offer a safe alternative for diagnosing COVID-19 due to its high contagion rate.
• Deep learning techniques show promise for disease classification and detection in the medical field.

Purpose of the Study:

• To develop and evaluate a deep learning-based technique for classifying COVID-19 infections from other non-COVID-19 conditions using chest X-rays.
• To compare the performance of three pre-trained deep learning models: EfficientNetB1, NasNetMobile, and MobileNetV2.
• To optimize deep learning models through fine-tuning and regularization for improved classification accuracy.

Main Methods:

• Utilized three pre-trained deep learning models: EfficientNetB1, NasNetMobile, and MobileNetV2.
• Employed an augmented dataset for training the deep learning models.
• Implemented two distinct training strategies, including model fine-tuning, hyperparameter optimization, and classification head regularization.

Main Results:

• The EfficientNetB1 model, with a regularized classification head, demonstrated superior performance compared to NasNetMobile and MobileNetV2.
• The proposed technique achieved an accuracy of 96.13% in classifying four distinct classes: COVID-19, viral pneumonia, lung opacity, and normal.
• The optimized deep learning approach significantly improved classification performance.

Conclusions:

• Deep learning models, particularly EfficientNetB1 with regularization, are effective for classifying COVID-19 from chest X-rays.
• The proposed technique offers a highly accurate and potentially rapid method for COVID-19 diagnosis.
• This approach shows superiority in accuracy compared to existing methods in the literature.