Non-parametric geodesic active regions: method and evaluation for cerebral aneurysms segmentation in 3DRA and CTA

Area of Science:

• Medical imaging analysis
• Image segmentation algorithms
• Computational anatomy

Background:

• Accurate segmentation of vascular structures is crucial for diagnosing and treating various medical conditions.
• Existing segmentation methods often struggle with complex vascular geometries and image noise.

Purpose of the Study:

• To develop and evaluate a novel, generic algorithm for segmenting vascular structures.
• To specifically apply and assess the algorithm's performance in segmenting cerebral aneurysms from 3D Rotational Angiography (3DRA) and Computed Tomography Angiography (CTA) data.

Main Methods:

• A geometric deformable model with an energy functional incorporating high-order multiscale features.
• A non-parametric statistical framework for robust segmentation.
• Application to cerebral aneurysm segmentation in 3DRA and CTA datasets.

Main Results:

• High overlap index with ground-truth segmentations (91.13% for 3DRA, 73.31% for CTA).
• Mean error distance close to in-plane resolution (0.40 mm for 3DRA, 0.38 mm for CTA).
• Favorable comparison against alternative deformable model techniques (geodesic active regions, active contours without edges).

Conclusions:

• The proposed algorithm provides accurate and robust segmentation of vascular structures, particularly cerebral aneurysms.
• The method demonstrates superior performance compared to existing deformable model techniques.
• This technique holds significant potential for improving clinical diagnosis and treatment planning involving vascular abnormalities.