Related Experiment Video
Updated: Nov 7, 2025
A Magnetic Resonance Imaging Protocol for Stroke Onset Time Estimation in Permanent Cerebral Ischemia
Published on: September 16, 2017
Algorithms for segmenting cerebral time-of-flight magnetic resonance angiograms from volunteers and anemic patients
Alexander Saunders1,2,3, Kevin S King4, Stefan Blüml1,2
1Children's Hospital Los Angeles, Department of Radiology, Los Angeles, California, United States.
Insights
Evaluating six cerebral arterial segmentation algorithms revealed the local minimum cost path method as the fastest and most accurate. This automated approach significantly outperforms manual segmentation for analyzing brain vasculature in various patient groups.
Area of Science:
- Medical Imaging
- Computational Anatomy
- Neuroscience
Background:
- Cerebral arterial segmentation is crucial for diagnosing and managing cerebrovascular diseases.
- Existing segmentation algorithms vary in accuracy and speed, necessitating performance evaluation.
- Hemodynamic characteristics in conditions like sickle cell disease can challenge segmentation accuracy.
Purpose of the Study:
- To assess the real-world performance of six distinct cerebral arterial segmentation algorithms.
- To compare automated segmentation methods against manual segmentation and inter-observer variability.
- To evaluate algorithm performance across patient groups with diverse hemodynamic profiles.
Main Methods:
- Time-of-flight magnetic resonance angiograms from 33 subjects (normal, sickle cell disease, non-sickle anemia) were analyzed.
- Six algorithms (Otsu's, K-means, region growing, active contours, minimum cost path, U-net) were tested using global and local region-selection methods.
- Manual segmentation by two readers served as the ground truth, with Matthew's correlation coefficient (MCC) measuring agreement.
Main Results:
- Automated segmentation times were significantly faster (0.1–172.9s) compared to manual segmentation (10h).
- All algorithms showed lower performance than inter-observer agreement, with the local minimum cost path method performing best.
- Performance varied significantly between algorithms and patient groups, with higher accuracy in non-sickle anemia and lower in sickle cell disease compared to controls.
Conclusions:
- The local minimum cost path algorithm offers a substantial speed advantage (3 orders of magnitude) over manual segmentation.
- While slightly less accurate than inter-observer agreement, this automated method is non-inferior for sickle cell disease and superior for non-sickle anemia.
- Automated cerebral arterial segmentation shows promise for efficient and effective clinical application, particularly in diverse patient populations.
Abstract:
Purpose: To evaluate six cerebral arterial segmentation algorithms in a set of patients with a wide range of hemodynamic characteristics to determine real-world performance. Approach: Time-of-flight magnetic resonance angiograms were acquired from 33 subjects: normal controls ( ), sickle cell disease ( ), and non-sickle anemia ( ) using a 3 Tesla Philips Achieva scanner. Six segmentation algorithms were tested: (1) Otsu's method, (2) K-means, (3) region growing, (4) active contours, (5) minimum cost path, and (6) U-net machine learning. Segmentation algorithms were tested with two region-selection methods: global, which selects the entire volume; and local, which iteratively tracks the arteries. Five slices were manually segmented from each patient by two readers. Agreement between manual and automatic segmentation was measured using Matthew's correlation coefficient (MCC). Results: Median algorithm segmentation times ranged from 0.1 to 172.9 s for a single angiogram versus 10 h for manual segmentation. Algorithms had inferior performance to inter-observer vessel-based ( , ) and voxel-based ( , ) measurements. There were significant differences between algorithms ( ) and between patients ( ). Post-hoc analyses indicated (1) local minimum cost path performed best with vessel-based ( , ) and voxel-based ( , ) analyses; and (2) higher vessel-based performance in non-sickle anemia ( ) and lower voxel-based performance in sickle cell ( ) compared with normal controls. All reported MCCs are medians. Conclusions: The best-performing algorithm (local minimum cost path, voxel-based) had 9.59% worse performance than inter-observer agreement but was 3 orders of magnitude faster. Automatic segmentation was non-inferior in patients with sickle cell disease and superior in non-sickle anemia.
