192
the abnormal area belongs to which part of the brain. Based on above analysis, our method
is expected to detect all kinds of abnormalities with rich textural information in our brain just
like Erus’s algorithm (2014), but our feature dimensionality is much less than Erus’s, which
may improve computing efficiency. Our technique may identify more complex pathological
changes than VBM, because the technique is based on the region, while VBM is based on
voxels. In our algorithm, we use textural and multiple atlas information to distinguish abnor-
malities without the help of doctors, while Aoki (2012) has to rely on radiologists to do the
similar job. In addition, our technique does not require any training in advance.
4 CONCLUSION
We discuss ...