Cardiff symposium talks and Jiscmail group

Two main requests that came out of the symposium were for access to talks and an online discussion forum.

  

We are implementing access to talks bit-by-bit as speakers opt in.  Take a look at the program at :  

https://sites.google.com/site/gabamrssymposium/meeting-schedule

in which titles will either link to downloadable talk or the submitted abstract if speakers have allowed it.  Some speakers are delaying providing talks, but will do so at a later date, so do keep checking (or perhaps even contact speakers directly if you have a special interest).

Following Gareth Barker's excellent suggestion, we have started the GABAMRS jiscmail group - this has the advantage of thread being searchable and archived by a UK-based university service.  It is accessible to all.  To subscribe to the list: 

Send an email to listserv@jiscmail.ac.uk Subject: BLANK Message: SUBSCRIBE GABAMRS Firstname Lastname

Hopefully, this will become a discussion forum for any issues of data acquisition, processing and interpretation, and participation will be as open as it was during the symposium.

Frequency correction revisited

In Gannet, frequency and phase correction of individual dynamics is performed by fitting the Cr signal in the frequency domain to estimate the frequency and phase shift (as per Waddell et al. MRI 2007).  This approach has at least two limitations.  Firstly, there is some similarity (over a small frequency range) between a small frequency shift and a small phase shift, leading to ambiguity in the fitting parameters.  Secondly, the approach is SNR-restrictive because it focuses in on one peak (of many) to figure out the frequency and phase.  For this reason, the correction will fall over below a certain voxel size (separate from the question of whether the final GABA signal is of good SNR through averaging).

Therefore we are in the process of implementing a new time-domain algorithm discussed by Jamie Near at ISMRM 2013, which estimates frequency and phase shifts by using all of the available signal.  This approach seems to result in substantially improved correction, as typified by the example below.

Comparison of GannetLoad output using the old frequency correction (left) and the new Spectral Registration algorithm (right).  The post-correction spectrum has substanitally smaller subtraction artefacts using Spectral Registration, and the frequency-correction output below shows much improved consistency between spectra post-alignment.

Algorithm courtesy of Jamie Near; Figure courtesy of Ashley Harris.