Related Experiment Video
Updated: Jun 21, 2026
10:28
Open-Source Real-Time Closed-Loop Electrical Threshold Tracking for Translational Pain Research
Published on: April 21, 2023
1.3K
A comparative analysis of deep learning-based location-adaptive threshold method software against other commercially
Daebeom Park1, Eun-Ah Park2,3, Baren Jeong2
1Department of Clinical Medical Sciences, Seoul National University College of Medicine, Seoul, Korea.
The International Journal of Cardiovascular Imaging
|April 18, 2024
Summary
A new deep learning method (DL-LATM) accurately segments coronary arteries in computed tomography angiography (CCTA) images. This advanced software shows potential for improved coronary artery disease (CAD) evaluation by precisely identifying lumen and plaque areas.
Area of Science:
- Medical Imaging Analysis
- Cardiovascular Disease Diagnostics
- Artificial Intelligence in Healthcare
Background:
- Accurate segmentation of coronary arteries in coronary computed tomography angiography (CCTA) is crucial for evaluating coronary artery disease (CAD).
- Existing commercial software platforms may have limitations in precise lumen and plaque segmentation.
- Deep learning approaches offer potential for enhanced accuracy in medical image analysis.
Purpose of the Study:
- To evaluate the performance of a novel deep learning-based location-adaptive threshold method (DL-LATM) for coronary artery segmentation using CCTA.
- To compare the segmentation accuracy of DL-LATM against commercially available software platforms.
- To assess the utility of DL-LATM in quantifying lumen and plaque parameters, particularly in stenotic regions.
Main Methods:
- Coronary artery segmentation was performed using CCTA images with the DL-LATM software and commercial platforms.
- Intravascular ultrasound (IVUS) data from 26 vessel segments (19 patients) served as the gold standard for comparison.
- Statistical analyses, including Pearson correlation coefficient (PCC), intraclass correlation coefficient (ICC), and Bland-Altman plots, were used to compare lumen and plaque parameters.
Main Results:
- The DL-LATM software demonstrated segmentation performance closest to the IVUS gold standard for lumen volume and area, showing minimal bias and high PCC/ICC.
- In stenotic regions, DL-LATM exhibited superior accuracy in detecting lumen and plaque area compared to commercial software, with significantly higher correlation coefficients (p < 0.001).
- DL-LATM showed a bias closest to zero for lumen area (mean difference = -0.72 mm², 95% CI = -0.80 to -0.64 mm²) and plaque area (mean difference = 1.70 mm², 95% CI = 1.37 to 2.03 mm²).
Conclusions:
- The DL-LATM software platform shows high accuracy and reliability for coronary artery segmentation in CCTA images.
- DL-LATM significantly outperforms existing commercial software in quantifying lumen and plaque characteristics, especially in stenotic areas.
- This deep learning-based method has strong potential as an aiding tool for the comprehensive evaluation of coronary artery disease.
