Related Experiment Video
Updated: May 27, 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
14.1K
Identifying COVID-19-Infected Segments in Lung CT Scan Through Two Innovative Artificial Intelligence-Based
Zeinab Momeni Pour1, Ali Asghar Beheshti Shirazi1
1Department of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran.
Archives of Academic Emergency Medicine
|February 17, 2025
Summary
This study introduces two novel Artificial Intelligence (AI) models, SE-UNETR and SE-HQRSTNet, for improved identification of COVID-19 infection regions in lung CT scans. The developed AI models demonstrate enhanced performance over existing methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Artificial intelligence (AI) algorithms have significantly advanced medical diagnostics.
- Accurate identification of COVID-19 infected regions in lung CT scans is crucial for patient management.
- Existing AI models for medical image analysis require further improvement in performance and efficiency.
Purpose of the Study:
- To introduce two novel AI models, SE-UNETR and SE-HQRSTNet, for the segmentation of COVID-19 infected regions in lung CT scans.
- To evaluate the performance of these models against existing methodologies using established datasets.
- To demonstrate the superior efficiency and accuracy of the proposed AI-driven approach.
Main Methods:
- Developed two 3D segmentation networks: SE-UNETR (Squeeze and Excitation UNet TRansformers) and SE-HQRSTNet (Squeeze and Excitation High-Quality Resolution Swin Transformer Network).
- SE-UNETR utilizes a 3D UNet architecture with a Vision Transformer (ViT) encoder for sequential volumetric data representation.
- SE-HQRSTNet integrates High-Resolution Networks (HRNet), Swin Transformer modules, and Squeeze and Excitation (SE) blocks with Multi-Resolution Feature Fusion (MRFF).
- Evaluated models using 5-fold cross-validation and data augmentation on the COVID-19-CT-Seg and MosMed datasets.
Main Results:
- The SE-UNETR and SE-HQRSTNet models achieved improved Dice values for infection masks on the COVID-19-CT-Seg dataset by 3.81% and 4.84%, respectively, compared to prior work.
- On the MosMed dataset, the Dice values increased from 66.8% to 69.35% (SE-UNETR) and 70.89% (SE-HQRSTNet).
- These results indicate significant performance gains in identifying COVID-19 affected areas.
Conclusions:
- The proposed SE-UNETR and SE-HQRSTNet models demonstrate superior performance and efficiency in segmenting COVID-19 infected regions in lung CT scans.
- These AI models represent a significant advancement over existing methodologies for medical image analysis in the context of COVID-19 detection.
- The developed networks offer a promising tool for enhancing the accuracy and speed of COVID-19 diagnosis through automated CT scan analysis.
