Evaluation of three automatic brain vessel segmentation methods for stereotactical trajectory planning.

Area of Science:

• Neurosurgery
• Medical Imaging
• Image Analysis

Background:

• Stereotactic procedures demand precise trajectory planning to avoid critical structures like blood vessels.
• Advancements in imaging and image recognition offer potential for automated vessel detection, enhancing patient safety during neurosurgical planning.

Purpose of the Study:

• To evaluate the feasibility of a blood vessel detection and warning system for stereotactic planning.
• To assess the efficacy of current imaging and vessel segmentation techniques for this application.

Main Methods:

• Acquired 3T/7T MRI data (T1CE, MRA) from nine subjects.
• Applied vessel segmentation using Vesselness, FastMarching, and LevelSet methods on 45 stereotactic trajectories.
• Compared automated segmentation results against manual segmentation by two experts.

Main Results:

• The LevelSet method demonstrated the highest performance (ICC 0.76), outperforming FastMarching (ICC 0.70) and Vesselness (ICC 0.56).
• All methods showed high negative predictive values (>97%), with positive predictive values ranging from 65-90%.
• The system could provide 2-3 warnings per trajectory with approximately 50% positive predictive value in a clinical setting.

Conclusions:

• Integrating a clinically relevant vessel detection and collision warning system into stereotactic planning software is feasible.
• Both T1CE and MRA imaging sequences are suitable for this automated application.