A multi-scale CNN with atrous spatial pyramid pooling for enhanced chest-based disease detection.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Deep Learning

Background:

• Early detection of COVID-19 and pneumonia is crucial for effective treatment.
• Chest X-ray analysis is a common diagnostic tool, but accuracy can vary.
• Existing deep learning models may struggle with capturing diverse features in X-ray images.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for the early detection of COVID-19 and pneumonia.
• To enhance the model's feature extraction capabilities for improved diagnostic accuracy.
• To validate the model's performance against existing state-of-the-art approaches.

Main Methods:

• A convolutional neural network (CNN) integrated with atrous spatial pyramid pooling (ASPP) was developed.
• Transfer learning and data augmentation techniques were employed to optimize performance and address data scarcity.
• The model's multi-branch architecture was designed for adaptable disease prediction.

Main Results:

• The ASPP-enhanced CNN achieved a validation accuracy of 98.66% for COVID-19 detection.
• The model achieved a validation accuracy of 83.75% for pneumonia detection.
• Performance metrics including accuracy, precision, F1-score, recall, specificity, and AUC surpassed other state-of-the-art methods.

Conclusions:

• The developed deep learning model demonstrates high accuracy and robustness for early COVID-19 and pneumonia detection.
• The integration of ASPP and transfer learning significantly improves diagnostic performance on chest X-ray images.
• This approach holds potential for faster, more reliable clinical diagnoses of respiratory conditions.