Hello all,

I have been running into a curious problem with a couple brains (8 out of 150 seem to have this problem) running through the recon-all pipeline. Upon loading brainmask.mgz, T1.mgz, and corresponding white matter and pial surfaces in Freeview, my colleague and I received an error that T1.mgz failed to load. Segmentation and parcellation seemed to run smoothly and accurately, however, the size of the T1.mgz file was 0. Assuming the file had been corrupted somehow (our server through which we run recon-all and store our data has tons of available memory), we reran the brains affected. Upon loading up the same re-run surfaces, T1.mgz loaded into the GUI as white, washed out voxels, displaying no brain at all. The size of this file is large, as well as all other surface and volume files. Curiously, the brainmask file seems normalized/stripped fine and surfaces displayed over the T1 look accurately segmented. I attached a screenshot displaying the T1.mgz file and surfaces that I'm referring to. 

I have a few concerns and questions. Because T1.mgz doesn't seem to be necessary for downstream manual editing and processing, can we proceed as normal, only displaying and using the brainmask file to make edits? Will our surface and volume metrics export properly without a usable T1 file? Lastly, any clues to how this might have occurred are welcomed. It seems like the mri_normalize step is the culprit, although the nu.mgz files corresponding to these brains seem normally sized.

Best,