Validation of a fiducial-based atlas localization method for deep brain stimulation contacts in the area of the subthalamic nucleus

Abstract

Differences in the location of active contacts with respect to the subthalamic nucleus (STN) may account for much variability in motor, psychiatric and cognitive responses to deep brain stimulation (DBS) in Parkinson disease (PD) patients. Because localization of STN based on hypointensity in T2-weighted MR images is unreliable and further limited by artifacts from the metal electrodes, we developed and validated a method to transform brain images into stereotactic space [Mai JK, Assheuer J, Paxinos G. Atlas of the Human Brain, 2nd ed. San Diego: Elsevier Academic; 2004] using reliably-identified anatomic fiducials identified in high-resolution T2-weighted pre-operative MR images. Average intraclass correlation between two raters for 29 PD patients was 0.93 for those fiducials used to define the atlas. Accuracy of the registration was tested by comparing the rater-identified centers of the red nuclei with their predicted locations from the fiducial-based atlas transformation. Mean discrepancies were 0.1, 0.9, and 0.0mm (x, y, z) with standard deviations of 0.9, 0.7 and 1.1mm, respectively. Because post-operative determination of contact location with respect to the STN is necessary due to possible shifting of electrodes during surgical placement, we identified active contacts on post-operative CT images and transformed their locations into stereotactic space. This method provides an accurate and reliable means for STN DBS contact localization.

Fig 1

Crosshairs and horizontal line identify the locations of all fiducials. For AC, PC, OX, OT, and RN, the left figure is a transverse section and the right figure is a coronal section of an interpolated T2-weighted MR image. A transverse section is shown for Pu and coronal sections are shown for BE and Tilt.

Fig 2

A illustrates the electrode artifact seen when viewing CT images scaled to intensities where brain tissue is visible. B is the same CT image as A but scaled to the peak intensity of the electrodes. C is 2 sections below the section illustrated in A & B and is scaled to the peak intensity of the right-hand electrode, only part of which penetrates this plane. D is a coronal section of a CT scan of the DBS electrode with a 0.2 mm wire positioned immediately adjacent to its tip. A, B and C are 2-mm transverse sections; D acquired in 0.6 mm sections.

Fig 3

Peak density of the DBS electrode tip in Hounsfield units as a function of the distance it extends into a 2-mm thick section of a CT image.

Fig 4

Composite mean MR images of 29 STN DBS patients. A corresponds to section 39 (22.6 mm posterior to AP) of Mai et al. (2004). B corresponds to section 34 (16.0 mm posterior to AP). Both sections show the structural boundaries of Mai overlaid with the RN outlined in white in A and the STN outlined in white in B.