A brainstem stereotactic atlas in a three-dimensional magnetic resonance imaging navigation system: first experiences with atlas-to-patient registration
K. Niemann et al., A brainstem stereotactic atlas in a three-dimensional magnetic resonance imaging navigation system: first experiences with atlas-to-patient registration, J NEUROSURG, 90(5), 1999, pp. 891-901
Object. The authors describe a computer-resident digital representation of
a stereotactic atlas of the human brainstem, its semiautomated registration
to sagittal fast low-angle shot three-dimensional (3-D) magnetic resonance
(MR) imaging data sets in 27 healthy volunteers and 24 neurosurgical patie
nts, and an analysis of the subsequent transforms needed to refine the init
ial registration.
Methods. Contour drawings from the atlas, which offer the 70th percentile o
f variation of anatomical structures, were interpolated into an isotropic 3
-D representation. Initial atlas-to-patient registration was based on the f
astigium/ventricular floor plane reference system. The quality of the fit w
as evaluated using superimposition of the atlas and MR images. If necessary
, the atlas was tailored to the individual anatomy by using additional tran
sforms. On average, the atlas had to be stretched by 2 to 6% in the three d
irections of space. Scale factors varied over a broad range from -8 to +19%
and the benefit of visual interactive control of the atlas-to-patient regi
stration was evident. Analysis of distances within the pens measured in the
midsagittal MR imaging slices and the required scale factors revealed sign
ificant correlations that may be used to reduce the amount of user interact
ion in the coregistration substantially. In 70.6% of the cases, the atlas h
ad to be shifted in a cranial direction along the brainstem axis (in 25.5%
of cases 3-4 mm, in 45.1% of cases 1-2 mm). This was due to a more caudal p
osition of the fastigium cerebelli on the MR images compared with the atlas
.
Conclusions. This observation, in conjunction with the variability of the h
eight of the fourth ventricle in our MR imaging data (range 6.1-15.2 mm, me
an 10.1 mm, standard deviation 1.8 mm) calls into question the role of the
fastigium cerebelli as an anatomical landmark for localization within the b
rainstem.