We are doing some QA exams with one subject scanned on different 3T scanners (with comparable/nearly identical imaging parameters). Afterwards, we looked into aseg values to compare them with the standard recon-all procedure. And we came across a strange result. Between two scanners the raw volumes are roughly the same, but the eICV's differ a lot (almost 30%). We looked into the data and found that even the mask volumes are alike (they differ less than 5% between the machines). the only outlier value is the eICV of one of the scanners. As this value is estimated, can this be a marker of some difficulties in normalization to Talairach upon the data of the scanner? Should it be alarming when an eICV is 30% lower than the mask volume? 

Best regards,

Patrycja