Anatomical Localization of Intracranial Electrodes Using Synthetic MRI Generated from Computed Tomography

Area of Science:

• Neurosurgery
• Medical Imaging
• Artificial Intelligence

Background:

• Deep brain stimulation (DBS) relies on accurate localization of intracranial electrodes, typically using MRI and CT.
• CT lacks soft tissue contrast, posing challenges when MRI is inaccessible or contraindicated.
• SynthSR, a deep learning tool, generates synthetic T1-weighted MRI from CT, potentially overcoming MRI limitations.

Purpose of the Study:

• To evaluate the accuracy of DBS electrode localization using synthetic MRI generated by SynthSR from CT.
• To compare SynthSR-based localization with standard MRI-CT fusion methods.
• To assess the feasibility of a CT-only approach for DBS electrode localization.

Main Methods:

• Analysis of CT, MRI, and postoperative CT scans from 44 Parkinson's disease patients undergoing STN-DBS.
• Generation of synthetic T1 MRI scans from preoperative CT using SynthSR.
• Co-registration of postoperative CT to synthetic and real MRIs for electrode localization.
• Comparison of localization accuracy using 3D Euclidean and radial distances and anatomical mapping.

Main Results:

• SynthSR-based localization showed mean radial error of 1.21 mm and Euclidean distance of 1.67 mm, comparable to electrode diameter.
• Anatomical localization concordance was 95.5% for overall STN targeting and 81.8% within STN subdivisions.
• Localization differences varied across cases, indicating areas for further improvement.

Conclusions:

• SynthSR demonstrates preliminary feasibility for DBS electrode localization, offering a viable CT-only alternative.
• This approach can facilitate postoperative evaluation when MRI is unavailable or contraindicated.
• Continued improvement in synthetic image generation may broaden access to imaging-based DBS analysis.