Tuning parameters in main.m to sequence with slightly different parameters.

[jberrebi]

Hi,
We have an agreement with the university of Illinois for using the sequence uiuc_speechPSF_fast. The version that is currently installed on our Siemens Prisma allows us to acquire speech imaging at resolution 1.9mm x 1.9mm x8.0mm and using partial separability functions and random phase encoding. There is one navigator for each sampled k-space line taking each 4,94ms to be sampled so that the sampling rate is 101,2Hz. Although the sequence can sample in 3D, it is used in 2D (one slice only).

I have acquired 5 times 40 full volumes (about 50 sec) for test and I have adapted the main.m script in the LowRankSpeechRecon toolbox accordingly to the parameters we have in our sequence, i.e.:

• line 33: Nn=1, instead of Nn=4; (one k-space line per navigator)
• line 69: Ipat=1 instead of Ipat=2;

I had to create two versions of Data2mat, but apart from that the program runs until it comes to the PSLR Recon "Reg_pwls_pcg"
Reg_pwls_pcg: 164 of 200 1593.5
Reg_pwls_pcg: 165 of 200 1596.5
Warn: Reg_pwls_pcg 152: wrong direction at iter=165; try using stop_grad_tol?
Reg_pwls_pcg 154: ratio = 80.4475

If I gently force the program to run to the end, I get a series of noisy pictures looking like follows:

Here is what I wonder:

• Do the toolbox (2023) and the sequence (2020) belong to the same generation, or are they compatible?
• There are two options on the sequence regarding random 2D sampling: "No_Nav_256" or "Cart_nav". The protocol had the value "No_nav_256" but should it be changed?
• I did not change the model order of 50 as it seems to be reasonable.
• I haven't changed the regularization parameter. Should I make it vary?
• Should I acquire speech for longer time?

I would be very thankful for any help.

Best,
Jonathan Berrebi