Hi everybody!
I hope that any of you could help me. I'm currently performing a volumetric research of the Hippocampus, involving VBM, WFU_PickAtlas and REX ( ROI Extraction).
In order to analyze the volume of the Hippocampus I have been using REX and the script get_totals (both in Matlab) in parallel as I would like verify the correctnesses of the volume out. But to my concern the outputs differentiate largely from each other, even though I'm using the same inputs. Can anyone explain to me why this is happening?
Thank you in advance!
Best regards; Cecilio.
In case of small anatomical structures consider the partial volume effect which plague almost all clinical tomographic imaging techniques.
In many digital atlases, the atlas itself suffers to some extent from the same effect, so it is impossible to avoid large effects unless the volumes under study are sufficiently large. This fenomenum cannot be ignored when comparing different atlases.
Also do not forget that probably different algorithms/techniques have been used for different atlasses to delineate the contours of the various structures. Small but systematic differences translate in relative large errors in small structures and in structures where the surface to volume ratio is relatively large.
For example the first few pages of the book: "In vivo atlas of deep brain structures"by Lucerna, Salpietro, Alafaci and Tomasello Springer press 2001 ISBN 3-540-42561-6 describes in detail how the imagery for such atlases are obtained. Given the limitations of the source data, the effects can be quite large for structures with a high surface to volume ratio. In case of such high S/V ratio's tiny differences in such algorithms tend to result in significant differences in registered volume.
The larger the S/V ratio is, the more sensitive any volume definition approach is to these effects.
The smaller V is, the more sensitive.
The smaller the voxelsize of the obtained images the more sensitive.
Even differences in post processing of CT / MRI / PET / SPECT data will affect perceived volumes, within the modality.
Judocus, I assume you meant "The greater the voxelsize of the obtained images the more sensitive."
@ Lambert Zip: The greater the voxelsize and the smaller the volumes-of-interest, the more sensitive. Sorry.
Yes, I get that. But you wrote: "The smaller the voxelsize of the obtained images the more sensitive."
Thank you so much for your recommendations and inputs. It helped me a lot!
Best regards; Cecilio.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Human-Computer Interaction