| |
AIM 3: Segmentation of the MBL and extraction of quantitative traits
Implicit fiducialguided
spline alignment
Automatic
extraction of fiducials may fail. The fallback position is manual
delineation. From the experiment described above (see the feasibility
assesment section above), we will establish the number of fiducials
required to achieve adequate alignment accuracy. Assuming all 100 are
needed, it would take approximately 2.5 years to generate and align 1200
MBL brains. The desired positional accuracy of 36 mm (3.6 x the resolution
of the atlas) might be unattainable even with all 100 fiducials. If that
is the case, the MBL will be segmented as best as can be achieved, and
finer accuracy will demand manual adjustment of the anatomical templates
on a case-by-case basis.
While
we are aware that the positional exactness sought is ambitious, we believe
that given the homogeneity of the MBL, the algorithm will support this
degree of accuracy on most brains. The quality control test embedded in
the procedure will flag those brains in which performance is poorer. For
those we will manually refine the segmentation using NeuroTerrain.
|
|
