Automated Basilar Artery Lumen Segmentation for High Resolution in Black Blood MRI.

Area of Science:

• Medical Imaging
• Neuroscience
• Artificial Intelligence

Background:

• The basilar artery (BA) is crucial for posterior brain circulation and prone to diseases like atherosclerosis.
• Accurate assessment of BA lumen and wall is vital for diagnosing and managing cerebrovascular conditions.

Purpose of the Study:

• To develop an automated image segmentation technique for detecting lumen and wall boundaries in BA black blood MR vessel wall images.
• To improve the accuracy and efficiency of BA structure evaluation in clinical practice.

Main Methods:

• A two-component method involving manual centerline tracking for 2D cross-sections and a deep learning model (Detectron2/Mask RCNN) for automated lumen segmentation.
• Training and testing of two models: arterial wall detector (AWD) and lumen detector (LD) on 26 MRI scans (20 training, 6 testing).

Main Results:

• High true positive detection rates achieved: 95.7% for arterial wall (AWD), 91.4% for lumen (LD), and 97.6% for combined AWD+LD.
• Low mean Hausdorff distances to ground truth: 1.19 ± 1.55 mm for AWD and 0.55 ± 0.78 mm for LD.
• Effective and accurate labeling of thin vascular structures (BA lumen and wall) using transfer learning.

Conclusions:

• The developed automated segmentation method accurately identifies basilar artery lumen and wall boundaries.
• This technique offers a valuable tool for physicians to enhance the evaluation of basilar artery in cerebrovascular disease treatment.