Biomedical Engineering Reference
In-Depth Information
Table 2.
Performance of the Classification Approach Based on
analysis of the CC, when Applied to Savant Datasets
Confidence Rate
Autistic
Normal
Postmortem Savant
Postmortem Savant
85%
-
10/15
-
29/30
90%
-
9/15
-
29/30
95%
-
7/15
-
25/30
Note that this approach was not tested on the postmortem datasets because of
geometric distortions.
5.
DISCUSSION
In this study, we proposed two neuroimaging-based approaches to discrim-
inate between autistic and typically developing brain. These approaches exploit
the reported anomalies of the white matter and the deficit in the area of the corpus
callosum in the autistic brain in order to extract geometrically based features that
can be used as accurate discriminating measures between the two groups. Unlike
traditional approaches, which mainly depend on volumetric descriptions of differ-
ent brain structures, the proposed classification techniques are less sensitive to the
selection of ages as well as to segmentation methods.
Research on the pathology of autism lacks consistency between findings. This
may be explained by the unavailability of larger sample sizes and closely matched
control groups, along with the failure to account for such confounding factors as
age, sex, IQ, socioeconomic status, mental retardation, and handedness. The use
of the distribution of the distance map inside the white matter, and the anatomical
features that the corpus callosumreveals through nonrigid registration, as explained
in this work, may provide for a biomarker in autism that allows to evaluate the
validity among varied studies.
Different brain structures will be investigated in order to follow the develop-
ment and variations of the autistic brain over time. This investigation will not be
limited to the white matter region but will also include gray matter studies. The
approach will be tested on postmortem datasets for more results and validation.
Eventually, a study needs to be established in which patients are scanned
according to a defined protocol. Typical T1- and T2-weighted MRI scans as well
as DTI need to be obtained. In addition, fMRI, MEG, as well as EEG may be used
in our future work.
