Methods

Following a family history for language impairment, 10 families were selected through linguistic testings where at least one of the children tested impaired on Dysphasia Screening Test (DST). All participants of the study underwent neurological testing and MRI scans were performed on the members over 8 years of age. The scans were corrected for intensity non-uniformity [2] and were automatically registered into stereotaxic space [3]. The images were automatically classified into grey, white and CSF using a 3D artificial neural network classifier [4],

=11in fig1.eps

Tranformation of MRI volume. A. MRI volume in native space B. Volume registered into stereotaxic space and tissue classified. Skull, Brainstem, cerebellum, ventricular csf and deep nuclei were removed from the tissue classified volume using a manually segmented brain mask.

In order to measure 'cortical' atrophy only, all structures outside the cerebral parenchyma as well as unwanted brain structures (including ventricular CSF, subarachnoid CSF surrounding the subcortical structures, deep nuclei, brain stem and cerebellum) were masked out using a brain mask generated from the average of 53 brains from our control subject database. The resulting volumes included only the cortical gray matter, cortical white matter and subarachnoid CSF strictly surrounding the cerebral hemispheres. Volume occupied by gray matter, white matter and CSF was calculated in control and test groups. A database of CSF/gray ratio was created for the control subjects and FLI subjects were tested against this database.