Related Experiment Video
Updated: Oct 24, 2025
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.4K
Deep Transfer Learning Based Unified Framework for COVID19 Classification and Infection Detection from Chest X-Ray
Sankar Ganesh Sundaram1, Saleh Abdullah Aloyuni2, Raed Abdullah Alharbi2
1Department of Artificial Intelligence and Data Science, KPR Institute of Engineering and Technology, Coimbatore, 641407 Tamil Nadu India.
Arabian Journal for Science and Engineering
|August 16, 2021
Summary
This study introduces a deep transfer learning framework for detecting COVID-19 from chest X-rays. The novel model achieves high accuracy in classifying and segmenting infections, aiding in pandemic management.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- COVID-19 pandemic necessitates rapid diagnostic tools.
- Chest X-rays are vital for COVID-19 detection and severity assessment.
- Automated solutions are crucial for screening and management.
Purpose of the Study:
- To develop a deep transfer learning framework for COVID-19 detection and infection segmentation from chest X-rays.
- To create a two-stage cascaded model integrating classification and segmentation subnetworks.
- To evaluate the framework's performance on public datasets.
Main Methods:
- A novel two-stage cascaded deep transfer learning framework was designed.
- A fine-tuned residual SqueezeNet served as the classifier.
- A fine-tuned SegNet semantic segmentation network, enhanced with Gaussian Mixture Model-based super pixel segmentation, was used for infection segmentation.
Main Results:
- The framework achieved 99.69% accuracy for binary classification and 99.48% for three-class classification.
- Mean accuracy for infection segmentation was 83.437%.
- Experimental results demonstrated the model's superiority over existing methods.
Conclusions:
- The proposed unified model shows high efficacy in COVID-19 detection and infection segmentation from chest X-rays.
- The framework offers potential for automated screening, long-term monitoring, and severity evaluation.
- Future extensions could focus on biomarker definition and quantitative severity assessment.
