Related Experiment Video
Updated: Nov 18, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.1K
COVID-19 lung CT image segmentation using deep learning methods: U-Net versus SegNet.
1Mechatronics Program for the Distinguished, Tishreen University, Distinction and Creativity Agency, Latakia, Syria.
BMC Medical Imaging
|February 9, 2021
Summary
Deep learning models SegNet and U-NET efficiently detect COVID-19 infection in CT lung images. SegNet excels at binary classification, while U-NET is better for multi-class segmentation, aiding rapid diagnosis and treatment prioritization.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer-Aided Diagnosis
Background:
- Urgent need for efficient tools to diagnose COVID-19 patients.
- Focus on detecting and labeling infected tissues on CT lung images.
- Utilizing deep learning for medical image analysis.
Purpose of the Study:
- Investigate SegNet and U-NET for semantic segmentation of infected lung tissue in CT scans.
- Evaluate their performance as binary and multi-class segmentors.
- Provide reliable computer-based techniques for COVID-19 detection.
Main Methods:
- Employed SegNet and U-NET deep learning networks for image tissue classification.
- Trained, validated, and tested networks on CT lung images.
- Calculated statistical scores to compare segmentation performance.
Main Results:
- SegNet achieved superior binary classification accuracy (0.95 mean accuracy).
- U-NET demonstrated better multi-class segmentation performance (0.91 mean accuracy).
- Both models proved effective in discriminating infected lung tissue.
Conclusions:
- Semantic segmentation of CT scans aids COVID-19 diagnosis and severity quantification.
- Proposed computer-based techniques reliably detect infected lung tissue.
- Automated detection can expedite global COVID-19 treatment efforts.
