[Neuroimaging] Should we prefer nii.gz or nii?

Matthew Brett matthew.brett at gmail.com
Mon Aug 31 09:59:46 CEST 2015


Hi Abraham,

On Mon, Aug 31, 2015 at 8:38 AM, Alexandre ABRAHAM
<abraham.alexandre at gmail.com> wrote:
> Hi Michael,
>
>> Do others have thoughts about the relative merits of compressed vs.
>> uncompressed files?
>
>
> This is a choice that depends heavily on the usage and the underlying
> hardware. As an example, nii is faster on my laptop, since I have a SSD. On
> my desktop, niigz tends to be faster but, for some tasks, I use nii because
> I want the data to be memmapped. On our cluster, I made some benchs and nii
> is the winner. That also depends if your processing is IO-bound or
> CPU-bound.
>
> What would be very useful though is to be able to read niigz files by
> chuncks. With nii, it's easy: the data is memapped so I don't need special
> processing. But if an uncompressed niigz file does not fit in memory, it
> would be good to still be able to process it online. I have code for that
> but it is very specific to my usage.

Are you using something like the slicing functionality in nibabel?

http://nipy.org/nibabel/images_and_memory.html#saving-time-and-memory

Cheers,

Matthew


More information about the Neuroimaging mailing list