Related Experiment Video
Updated: Aug 7, 2025
A Rat Carotid Artery Pressure-Controlled Segmental Balloon Injury with Periadventitial Therapeutic Application
Published on: July 9, 2020
NAG-Net: Nested attention-guided learning for segmentation of carotid lumen-intima interface and media-adventitia
Qinghua Huang1, Liangrun Zhao1, Guanqing Ren2
1School of Artificial Intelligence, OPtics and ElectroNics (iOPEN), Northwestern Polytechnical University, Xi'an, 710072, Shaanxi, China; School of Mechanical Engineering, Northwestern Polytechnical University, Xi'an, 710072, Shaanxi, China.
Insights
A new deep learning model, NAG-Net, accurately segments carotid interfaces for intima-media thickness measurement, aiding early cardiovascular disease detection. This method integrates clinical knowledge and simplifies post-processing for precise results.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Cardiology
Background:
- Cardiovascular diseases (CVD) are a leading global cause of mortality.
- Accurate measurement of intima-media thickness (IMT) is crucial for early CVD screening and prevention.
- Existing methods for segmenting carotid Lumen-Intima Interface (LII) and Media-Adventitia Interface (MAI) lack clinical domain integration and require complex post-processing.
Purpose of the Study:
- To propose a novel nested attention-guided deep learning model (NAG-Net) for accurate LII and MAI segmentation.
- To incorporate clinical domain knowledge into the segmentation process.
- To develop a method that yields fine contours with simplified post-processing.
Main Methods:
- Development of NAG-Net, comprising two nested sub-networks: Intima-Media Region Segmentation Network (IMRSN) and LII and MAI Segmentation Network (LII-MAISN).
- Integration of clinical domain knowledge via attention maps generated by IMRSN.
- Application of transfer learning with pre-trained VGG-16 weights to enhance feature extraction and address data scarcity.
- Design of a channel attention-based encoder feature fusion block (EFFB-ATT) for efficient feature representation.
Main Results:
- NAG-Net achieved accurate segmentation of LII and MAI, producing fine contours with minimal post-processing.
- The model successfully incorporated clinical domain knowledge, improving segmentation focus.
- Transfer learning and EFFB-ATT enhanced feature extraction capabilities.
- Extensive experiments confirmed NAG-Net's superior performance over state-of-the-art methods across all evaluation metrics.
Conclusions:
- NAG-Net offers a significant advancement in automated carotid interface segmentation for IMT measurement.
- The model's ability to integrate clinical knowledge and simplify post-processing makes it a valuable tool for CVD early detection.
- NAG-Net demonstrates high potential for clinical application in cardiovascular risk assessment.
Abstract:
Cardiovascular diseases (CVD), as the leading cause of death in the world, poses a serious threat to human health. The segmentation of carotid Lumen-intima interface (LII) and Media-adventitia interface (MAI) is a prerequisite for measuring intima-media thickness (IMT), which is of great significance for early screening and prevention of CVD. Despite recent advances, existing methods still fail to incorporate task-related clinical domain knowledge and require complex post-processing steps to obtain fine contours of LII and MAI. In this paper, a nested attention-guided deep learning model (named NAG-Net) is proposed for accurate segmentation of LII and MAI. The NAG-Net consists of two nested sub-networks, the Intima-Media Region Segmentation Network (IMRSN) and the LII and MAI Segmentation Network (LII-MAISN). It innovatively incorporates task-related clinical domain knowledge through the visual attention map generated by IMRSN, enabling LII-MAISN to focus more on the clinician's visual focus region under the same task during segmentation. Moreover, the segmentation results can directly obtain fine contours of LII and MAI through simple refinement without complicated post-processing steps. To further improve the feature extraction ability of the model and reduce the impact of data scarcity, the strategy of transfer learning is also adopted to apply the pretrained weights of VGG-16. In addition, a channel attention-based encoder feature fusion block (EFFB-ATT) is specially designed to achieve efficient representation of useful features extracted by two parallel encoders in LII-MAISN. Extensive experimental results have demonstrated that our proposed NAG-Net outperformed other state-of-the-art methods and achieved the highest performance on all evaluation metrics.
More Related Videos
04:48Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
11:00A Model of Disturbed Flow-Induced Atherosclerosis in Mouse Carotid Artery by Partial Ligation and a Simple Method of RNA Isolation from Carotid Endothelium
Published on: June 22, 2010