Related Experiment Video
Updated: Aug 11, 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
14.3K
COVID-19 Lung CT image segmentation using localization and enhancement methods with U-Net
Ahmet Ilhan1,2, Kezban Alpan3,2, Boran Sekeroglu3,2
1Department of Computer Engineering, Near East University, Nicosia, 99138, Cyprus, Mersin 10 Turkey.
Procedia Computer Science
|February 6, 2023
Summary
This study introduces a new AI system for segmenting COVID-19 pneumonia lesions in lung CT scans. The method enhances image contrast, significantly improving lesion detection accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Accurate segmentation of pneumonia lesions in lung CT images is crucial for diagnosing COVID-19 and assessing disease severity.
- AI-based systems face challenges with low-contrast abnormal regions in CT scans, hindering precise segmentation.
- Image preprocessing techniques are being explored to overcome these limitations and improve AI model performance.
Purpose of the Study:
- To develop and evaluate an AI system for accurate segmentation of COVID-19 related pneumonia lesions in lung CT images.
- To investigate the efficacy of histogram-based non-parametric region localization and enhancement (LE) as a preprocessing step for deep learning segmentation.
- To improve the discriminative features of infected lung regions for more accurate AI-driven segmentation.
Main Methods:
- A novel COVID-19 Lung-CT segmentation system combining histogram-based LE methods with the U-Net architecture was proposed.
- The LE method was applied to COVID-19 infected lung CT images to detect and enhance abnormal regions.
- The enhanced images were then used to train the U-Net model for segmenting COVID-19 affected areas.
Main Results:
- The proposed system achieved high performance metrics: 97.75% accuracy, 0.85 dice score, and 0.74 Jaccard index.
- Preprocessing lung CT images with LE methods significantly improved the U-Net model's segmentation ability, increasing the dice score by 0.21.
- The enhanced feature extraction and segmentation capabilities demonstrated the effectiveness of the LE preprocessing step.
Conclusions:
- The integration of LE methods as a preprocessing step substantially enhances the performance of AI-based segmentation for COVID-19 lung CT images.
- The proposed system offers a promising approach for accurate and reliable segmentation of pneumonia lesions, aiding in clinical diagnosis and management.
- The findings suggest potential broader applications of the LE method in segmenting various types of medical images.
