A new study shows that when it comes to analyzing brain scans, the results can be dramatically affected by everything from software version to platform. The study looked at 30 MRI scans of brains using an imaging tool called FreeSurfer, which runs on either Linux or Mac OS X. Depending on what version of the software was used (versions 4.3.1, 4.5, and 5.0 were all tested) the size and thickness of parts of the brain varied by anywhere from five to 15 percent. There were also differences depending on whether or not you used a Mac or an HP computer running CentOS, and even which version of OS X was used — for the most part the differences were in the two to five percent range, but in some areas of the brain they reached as high as 15 percent.
While sound methodology is key in science, the key takeaway from the authors of the study is to be cautious when upgrading the tools you use — especially in the middle of a study. Design differences like whether software uses a Unix-based Mac shell or a Linux-based one, for instance, can end up affecting the output in ways that might not be immediately obvious. It's also important to know just how good the tools actually are. As the authors explain so diplomatically, "given the large and significant differences between the latest version v5.0.0 and earlier versions, it is concluded that an assessment of the accuracy of FreeSurfer is desirable."
Update: We've changed the text to clarify that while the Mac was compared to a PC made by HP, it was running a Linux-based OS.