MRI image classification for disease diagnosis

I am working on MRI images for image classification. My aim is Feature extraction and then classifier. Any ideas that I could use behind these steps?

I have done like (i) Segmentation and extracted the segmented voxels. Now the segmented voxels are of different sizes for each image. So no idea how to input this to a classifier? tried with mean and other statistics features. accuracy is less.

(ii)planning to take the histogram of the segmented region. But the voxels are in int16 format and if I change to uint8 format is it meaningful?

Thanks for all your help!

[Image Analyst]

"So no idea how to input this to a classifier? " <== You pass them in through an input argument list of a function.

If you change from uint16 to uint8 you will lose intensity resolution. I would not do that unless you are really low on memory.

[Jab]

Thanks for your prompt reply!