Skip to main content
  • Development and Validation of a Simple Landmark Placement Protocol for Establishing Correspondence Between Brain Images

    Final Number:

    Jonathan C. Lau MD; Andrew G. Parrent MD; John Demarco; Ali R. Khan PhD; Terry M. Peters PhD

    Study Design:

    Subject Category:
    Image Guided Applications and Brain Mapping

    Meeting: 2018 ASSFN Biennial Meeting

    Introduction: Template and atlas-guidance are fundamental aspects of stereotactic neurosurgery. Accurate spatial correspondence between the template and patient images is a crucial step in being able to use templates to assist with surgical implantation. In the absence of a robust quantitative approach, we sought to propose and validate a set of point landmarks that could be quickly, accurately, and reliably placed on brain images.

    Methods: A series of neuroanatomical landmarks were identified in consultation with an experienced neurosurgeon. Consensus was achieved on a set of 32 landmarks (Figure 1). Over a series of neuroanatomy tutorials, novice participants (N=8) were trained to perform the protocol on 3 publicly available brain templates: Colin27 [1], MNI2009b [2], and Agile12v1.0 [3]. For each template, our participants placed the landmarks four times (384 landmarks/participant). Fiducial localization error (FLE) was calculated to establish reliability. We performed K-means clustering on the principal components of the landmark-specific point clouds.

    Results: Intra- and inter-rater reliability were 1.24 +/- 0.17 mm and 1.24 +/- 1.94 mm respectively. Out of 3258 landmarks placed, there were 24 (0.74%) outliers more than 10 mm from the group mean, which we classified as mislabeled and thus discarded. Significant differences in FLE were identified between templates (Colin27: 1.11 +/- 1.05 mm; MNI2009b: 0.95 +/- 0.82 mm; Agile12v1.0: 1.02 +/- 0.94 mm). K-means clustering of the principal components identified three clusters (Figure 2). Landmark placement time was estimated at 30 minutes.

    Conclusions: Our landmarks provide an intuitive and anatomically-driven framework for establishing quality of registration between brain images. While overall FLE was within an acceptable range, point cloud distributions were heterogeneous. The proposed protocol is reproducible, less manually intensive, and more sensitive to local errors than segmentation-based or qualitative evaluation of correspondence. This may hold value for a broad number of applications including template-to-patient registration and teaching neuroanatomy.

    Patient Care: Template and atlas-guidance are fundamental aspects of stereotactic neurosurgery. Better metrics for establishing spatial correspondence will enable an improved understanding of the limits of template-to-subject alignment as relevant to stereotaxy.

    Learning Objectives: By the conclusion of this session, participants should be able to: 1. Understand how image registration can be prone to alignment biases that should be accounted for when using templates for stereotactic assistance. 2. Describe how point landmarks provide local information in millimeters about the quality of spatial alignment between images. 3. Consider point landmarks as complementary to segmentation-based and qualitative assessment of brain image registration and hold value for teaching neuroanatomy.

    References: 1. Holmes, C. J. et al. Enhancement of MR Images Using Registration for Signal Averaging. J. Comput. Assist. Tomogr. 22, 324–333 (1998). 2. Fonov, V. et al. Unbiased nonlinear average age-appropriate brain templates from birth to adulthood. Neuroimage 47, S102–S102 (2009). 3. Lau, J. C. et al. Ultra-High Field Template-Assisted Target Selection for Deep Brain Stimulation Surgery. World Neurosurg. 103, 531–537 (2017).

We use cookies to improve the performance of our site, to analyze the traffic to our site, and to personalize your experience of the site. You can control cookies through your browser settings. Please find more information on the cookies used on our site. Privacy Policy