Hi Brandon,
You could also try decreasing the resampling resolution, but I'm not sure what an acceptable lower bound would be on this... Things certainly start looking a little "chunky" around 4mm (at least compared to 2mm), so I personally don't like going beyond this.
We're also hitting memory issues with large group ICAs, so it would be nice to know other tricks to get around it. I think we're limited to ~32 GBs of RAM, and we're generally seeing memory issues when we have over 200 inputs/subjects with an image size of 61x73x61x192 (x,y,z,t; 3mm resampling).
Cheers,
David
On Sep 12, 2011, at 10:38 AM, Brandon Pye wrote:
> Thank you for your response.
>
> We have ~36.5 GB of uncompressed data. Is the best/only option for us to use a computer/server with 36 GB of RAM? Is there any way for us to modify how it processes/compiles the program?
>
> Thanks,
> Brandon
|