Deep-learning method for fully automatic segmentation of the abdominal aortic aneurysm from computed tomography...

Area of Science:

• Medical Imaging
• Cardiovascular Surgery
• Artificial Intelligence

Background:

• Abdominal aortic aneurysm (AAA) is a major global cause of mortality, often asymptomatic until critical stages.
• Accurate quantification of AAA geometric properties and growth is crucial for predicting clinical outcomes and surgical planning.
• Current manual segmentation methods for AAA in computed tomography (CT) scans are time-consuming and prone to significant variability.

Purpose of the Study:

• To develop and validate an automated deep learning-based model for segmenting abdominal aortic aneurysm (AAA) tissues.
• To overcome the limitations of manual segmentation in terms of accuracy, efficiency, and inter-operator variability.
• To enable precise quantification of AAA geometric properties and monitor disease progression.

Main Methods:

• A three-step deep learning model was developed for automated AAA tissue segmentation.
• The model first extracts the aorta and iliac arteries.
• Subsequently, it detects the aneurysm lumen and other relevant AAA tissues.

Main Results:

• The automated segmentation model demonstrated high accuracy in identifying AAA tissues.
• Results showed very good agreement between the automated segmentation and expert manual segmentation.
• The deep learning approach significantly reduces the time and variability associated with manual measurements.

Conclusions:

• The proposed deep learning model offers a reliable and efficient automated solution for AAA segmentation from CT images.
• This technology can enhance the clinical prognosis of AAA by enabling accurate geometric quantification and growth monitoring.
• Automated segmentation holds the potential to standardize AAA assessment and improve patient management strategies.