Fast and Accurate Detection of COVID-19 Along With 14 Other Chest Pathologies Using a Multi-Level Classification:...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• The rapid spread of COVID-19 necessitates advanced diagnostic tools to support healthcare systems.
• Existing deep learning models for chest disease detection often lack the ability to identify multiple conditions simultaneously.

Purpose of the Study:

• To develop a fast and accurate diagnostic system for COVID-19 detection using X-ray images.
• To classify COVID-19 X-rays against normal cases and 14 other distinct chest diseases.

Main Methods:

• A novel, multilevel deep learning pipeline was designed for X-ray image classification.
• Transfer learning using ImageNet-pretrained models (ResNet50) facilitated rapid training.
• Image segmentation identified lungs and heart, feeding into a two-stage classification process.

Main Results:

• The pipeline achieved competitive accuracy for COVID-19 detection alongside 14 other chest diseases.
• The two-level classification achieved 96.04% training and 92.52% test accuracy for 3 classes (normal, COVID-19, other).
• The second level achieved 88.52% training and 66.63% test accuracy for 14 diseases, outperforming single-stage classification.

Conclusions:

• The proposed multilevel pipeline effectively detects COVID-19 and other chest diseases from X-rays with high accuracy.
• Dividing the classification task into sequential steps improves diagnostic performance compared to a single-stage approach.