[Mrtrix-discussion] MRtrix version 0.2.7 released

[Donald Tournier]

Dear MRtrix users,

I have just been alerted to the fact that the FA values produced by
MRtrix are underestimated by approximately 22.5%. This has now been
fixed in this new release. Any users that rely on MRtrix to produce FA
values are urged to regenerate their FA maps using this new version.

Note that this should NOT affect any CSD-based  tractography results,
since at no point are the FA values used for the tracking. Any users
wishing to report FA values should account for the bias in previous
versions of MRtrix. If MRtrix was used to produce all the FA maps,
then it should be a simple matter of scaling the values up by
sqrt(3/2) (i.e. multiply by 1.224744871). If the FA maps were
generated using a mixture of different packages, then the
MRtrix-generated FA maps should be regenerated, and any subsequent
analyses performed again.

For those interested, the problem was due to my code using "sqrt(3/2)"
to (correctly) scale the FA. Unfortunately, the compiler interprets
the operation "3/2" as an integer division, returning the integer "1"
rather than the expected floating-point number "1.5". Hence, the
multiplication by "sqrt(3/2)" had essentially no effect. This has now
been rectified by substituting "sqrt(1.5)". Thanks to Kerstin Pannek
(once again!) for reporting this bug.

Also included in this new release are some tools to ease integration
with other spatial normalisation packages, as well as a command-line
ROI statistics tool.

My apologies if this issue causes any inconvenience.
Cheers,

Donald.


-- 
Jacques-Donald Tournier (PhD)
Brain Research Institute, Melbourne, Australia
Tel: +61 (0)3 9496 4078