Deformable anatomic templates embed knowledge into patient's brain images: Part 1. Construction and display.

OBJECTIVE: This paper describes the methods used to create annotated deformable anatomic templates (DATs) and display them in a patient's axial 2-dimensional and reformatted volume brain images. METHODS: A senior neuroradiologist annotated and manually segmented 1185 color-coded structures on axial magnetic resonance images of a normal template brain using domain knowledge from multiple medical specialties. Besides the visible structures, detailed pathways for vision, speech, cognition, and movement were charted. This was done by systematically joining visible anatomic anchor points and selecting the best fit based on comparisons with cadaver dissections and the constraints defined on the companion 2-dimensional images. RESULTS: The DAT is commercially available for use on a picture archiving and communication system or as a standalone workstation. CONCLUSIONS: The DAT can quickly embed extensive, clinically useful functional neuroanatomic knowledge into the patient's brain images. Besides labeling visible structures, DAT displays clinically important, previously uncharted subdivisions of the fiber tracts.

Deformable anatomic templates embed knowledge into brain images: part 2. Validation using functional magnetic resonance imaging of the motor hand.

OBJECTIVE: This study evaluated the concordance between the Deformable Anatomic Template (DAT)-identified origin of motor hand fibers and localization of the motor cortex of the hand by functional magnetic resonance imaging (fMRI). METHODS: Preoperative fMRI during hand motor tasks was performed on 36 hemispheres in 26 patients with gliomas in or near eloquent areas. Reformatted volume-rendered surface images were labeled with the DAT's hand motor fibers and fMRI data. Five reviewers assessed the data for concordance. RESULTS: Available fMRI data were diagnostically usable in 92% (33/36 analyzed hemispheres), with DAT anatomic accuracy in the remaining cases. The DAT prediction and fMRI findings were concordant in all 9 normal hemispheres and in 20 (83%) of 24 glioma-bearing hemispheres. The 4 discordant cases resulted from substantial mass effect by large frontal tumors. CONCLUSIONS: This study validated DAT's anatomic atlas and alignment process for the expected position of the motor cortex of the hand.