Deep learning-based medical image segmentation of the aorta using XR-MSF-U-Net

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Cardiovascular Disease Analysis

Background:

• Manual segmentation of cardiac aorta in CT and MRI is subjective and time-consuming.
• Accurate segmentation is crucial for diagnosing cardiovascular diseases.
• Existing methods may lack efficiency and precision.

Purpose of the Study:

• To develop an automated segmentation technology for cardiac aorta using CT and MRI.
• To improve the analysis of patient conditions and reduce cardiovascular disease misdiagnosis and mortality.
• To overcome the subjectivity and unrepeatability of manual segmentation.

Main Methods:

• Implementation of the X ResNet (XR) convolution module for efficient feature extraction.
• Integration of a Multi-scale features fusion module (MSF) with attention for enhanced network details.
• Utilizing the XR-MSF-U-Net model for cardiac aorta segmentation in CT and MRI images.

Main Results:

• The XR-MSF-U-Net model demonstrated superior segmentation performance on CT and MRI datasets compared to the benchmark U-Net model.
• Achieved a 7.99% improvement in DSC and an 11.01 mm reduction in HD on CT datasets.
• Achieved a 10.19% improvement in DSC and a 6.86 mm reduction in HD on MRI datasets, outperforming similar models.

Conclusions:

• The XR-MSF-U-Net model offers a novel and effective approach for cardiac aorta segmentation in CT and MRI.
• This technology enhances diagnostic accuracy and efficiency, potentially aiding in cardiovascular disease management.
• The study provides a valuable tool for the segmentation of aortic CT and MRI, offering substantial clinical benefits.