Tip:
Highlight text to annotate it
X
This is a demonstration of Ezys image registration program
It was recorded on a laptop with NVIDIA GeForce 445M graphics card
At first, I will demonstrate Ezys' by doing an inter-subject registration
Two human brain images (a source and a target) of different subjects are loaded
The source image is being deformed so that it would match the target image
Window on the right shows the progress in real time
The top-left quarter of the window shows superposition of source ant target images.
If the registration is successful, superposing green source and red target images should create a yellowish image
The top-right quarter shows estimated local volume differences between two brain images
Green areas indicate which regions were bigger in the original source image, and red areas indicate which regions were smaller.
By moving a pointer over brain images one can see numerical values indicating local volume differences displayed on "visualization" tab.
By changing "Jacobian transparency" we can see how well detected local volume differences match with the underlying target image
Transformation matching the source and the target images can also be animated.
Changing Jacobian transparency a few times to see whether detected volume differences match with the underlying brain structures
Changing Jacobian contrast
Now I will demonstrate how Ezys can estimate volume changes in a brain over time (intra-subject registration)
Let's load two brain images taken two years apart of a person who has semantic dementia.
Also select a symmetric image registration scheme to get a better accuracy
And click on "Start". It takes some time to get the images loaded...
Both images are loaded and the registration begins
At first images are rigidly aligned and then the source image is elastically deformed.
We can see the image match being gradually improved
And the detected volume changes are also displayed in red and green
Registration A to B is done. Now Ezys will register B to A and average the results to get a better accuracy.
It takes some time, as this video was recorded on a laptop
And finally the registration has finished
We should increase Jacobian contrast as to make the subtle changes visible
By moving a pointer it is possible to read numerical values of detected volume changes at every point
Note a significant atrophy of temporal lobes (consistent with semantic dementia)
We can also animate the transformation as before to see the changes which happened over 2 years.
Going through the slices again...
Thank you for watching!