Question

Question Posted 09/03/24:
Dear expert
I want to ask is there any recommendation for what slice from the .dcm file i can take to convert it to an image. I want to train the image instead of the whole scan because it will take resource .I'm looking for slices that have the information for parts that is most affected by alzheimer. May i know which slice may be best for each plane (saggital, axial and coronal). Thank you very much
Response posted 09/04/24 by Rob Reid:
Hello,
I understand that many neural networks are setup for 2D images instead of 3D, and that 3D necessarily uses more resources than 2D, but which slice is best out of a 3D stack has no predetermined answer. The placement of people's heads varies from scan to scan.

If you cannot use a network which accepts 3D input, you could try making "find the best slice in a stack" as a problem for a separate network. It would not really be efficient though. Since "best" is defined here by "most affected by Alzheimer's", you would end up having to train many instances of the networks to find the answer.
Response posted 09/05/24 by ADNI MRI Core:
Thank you for your question to the ADNI MRI Core.

Unfortunately ADNI MRI doesn't have a specific answer we can provide. However, If you are looking for one slice, perhaps you could focus on the High Resolution Hippocampus sequence that has been acquired in ADNI3 and ADNI4. These are oblique coronal sequences focusing on the Hippocampus which has proven to play a critical role in learning and memory.

Good luck!

ADNI MRI
Go back to list of topics >
2024 Alzheimer’s Disease Neuroimaging Initiative
This website is funded by the Alzheimer’s Disease Neuroimaging Initiative