Exploiting Multiple Optimizers with Transfer Learning Techniques for the Identification of COVID-19 Patients

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Chest X-rays are a viable method for COVID-19 detection.
• Existing methods often require large datasets and struggle with limited image boundaries.

Purpose of the Study:

• To develop a robust and efficient transfer learning method for early COVID-19 detection.
• To identify normal versus COVID-19 positive patients using limited training data.
• To improve detection accuracy in medical image analysis.

Main Methods:

• Employed transfer learning techniques with data augmentation.
• Evaluated five state-of-the-art models (AlexNet, MobileNetv2, ShuffleNet, SqueezeNet, Xception) with three optimizers (Adam, SGDM, RMSProp).
• Utilized a 10-fold cross-validation on publicly available datasets.

Main Results:

• MobileNetv2 with Adam optimizer at a 3e-4 learning rate achieved the highest performance.
• Achieved average accuracy (97%), recall (96.5%), precision (97.5%), and F-score (97%).
• Demonstrated superior results compared to other model-optimizer combinations.

Conclusions:

• The proposed transfer learning method is effective for early COVID-19 detection.
• The approach offers a time-efficient and accurate solution, especially with limited data.
• This method shows competitive performance against existing literature and potential for clinical application.