Related Experiment Video
Updated: Nov 28, 2025
03:40
Three-Dimensional Printing of a Complex Aortic Anomaly
Published on: November 1, 2018
6.9K
A Three-Dimensional Deep Convolutional Neural Network for Automatic Segmentation and Diameter Measurement of Type B
Yitong Yu1, Yang Gao1, Jianyong Wei2
1Department of Radiology, Fuwai Hospital, Peking Union Medical College & Chinese Academy of Medical Sciences; State Key Lab and National Center for Cardiovascular Diseases, Beijng, China.
Korean Journal of Radiology
|November 25, 2020
Summary
A novel deep learning method accurately segments and measures Type B aortic dissection (TBAD) dimensions, outperforming manual measurements in speed and consistency. This 3D CNN approach offers a promising tool for automated aortic morphology evaluation.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Medicine
- Cardiovascular Imaging
Background:
- Type B aortic dissection (TBAD) requires precise segmentation and diameter measurement for effective management.
- Current manual measurement methods are time-consuming and prone to inter-observer variability.
Purpose of the Study:
- To develop and validate an automatic deep learning method for segmenting and measuring TBAD.
- To compare the accuracy, stability, and efficiency of the deep learning method against manual measurements.
Main Methods:
- A three-dimensional (3D) deep convolutional neural network (CNN) was implemented for automatic segmentation and measurement.
- The method was applied to aortic computed tomography angiographic images from 139 patients with TBAD.
- Performance was evaluated by comparing Dice scores, measurement errors, Bland-Altman analysis, and measurement time against manual methods.
Main Results:
- High Dice coefficients (0.958-0.961) were achieved for entire aorta, true lumen, and false lumen segmentation.
- The deep learning method demonstrated significantly lower average measurement errors (1.64%-2.50%) compared to manual methods (4.13%-11.67%).
- The deep learning approach was substantially faster (21.7 min vs. 82.5 min) and showed no intra- or inter-observer variability, unlike manual measurements.
Conclusions:
- The 3D deep CNN method provides accurate and stable segmentation and diameter measurement for TBAD.
- This automated approach shows potential for efficient evaluation of aortic morphology, reducing radiologist workload.
