Hi,
I've been using PNM without problem for the last few years. At some point over the last 2 months it seems to have stopped working on the systems I have been using. This first occurred on a SLED 11.4 64 bit linux system, but I get the same behaviour on a Fedora 23 64 bit system.
I suspect the problem may be related to a unix update, as I didn't update fsl over this period, but I am not sure which update may have caused this.
The problem seems to be in the production of the .js file by the first stage of PNM:
This file used to be produced (on my systems at least) in a standard decimal format such as "-0.5,-4.14002e-11,0,0.444928,0\n" + ... etc.
Now (on both the SLED and fedora systems I've tried) it is produced in a c style hexadecimal format i.e. "0x1p-1,0x1.be37e48001149p-33,0x0p+0,0x1.814c74266cb87p-1,0x0p+0\n"
The hex numbers appear to be correct (i.e. when I convert them and plot them I am getting the appropriate cardiac and respiratory traces and the peak detection works), but when I attemtpt to run the second stage of PNM it generates regressors (i.e. the nifti ones) which are all NaNs.
Any idea about how this can be sorted?
thanks
Mike
|