 |
 |
Hi, I think we need to realize how MRIConvert deals with converted data. Below is the description of file formats which MRIConvert produces. (I took this from fileformats.txt which comes with MRIConvert program.) ---Begin quote from fileformats.txt--- Analyze 7.5 Analyze 7.5 files are output in the LAS coordinate system. The image data is flipped in the column direction before writing. Currently, the orient field is always set to the value 0, indicating transverse unflipped data, regardless of whether the data represents transverse, sagital, coronal, or oblique slices. Most Analyze viewers plot image data starting from the lower left corner of the view. For "LA" axial images, this results in a "radiological" view identical to the original DICOM view. Note the "LAS" is a left-handed coordinate system. Left-handed data ordering is sometimes refered to as "radiological" ordering, and right-handed ordering as "neurological". Those terms can be misleading and should probably be restricted to their original context: describing the appearance of two-dimensional transverse brain images. SPM99/Analyze The SPM99/Analyze output of MRIConvert is similar to the Analyze 7.5 output, with a few important distinctions. In the header, the "originator" field, left blank for Analyze 7.5, is now defined as the "origin" field. MRIConvert sets this to the middle of the volume. More importantly, the image data is rotated 180 degrees before writing. This means that transverse slices are stored RAS. NIfTI NIfTI output is based on the most straightforward conversion of DICOM to NIfTI according to the NIfTI standard. Image data is not rewritten, and the image orientation relative to NIfTI's standard RAS axes is given by the qform fields. You can expect axial images to be stored LPS. How they are displayed and interpreted will (as always) depend entirely on the software you use to read them, its assumption about data order, and its support (or lack thereof) for the qform information. Any software fully compliant with the NIfTI standard should be able to read these files and their orientations. FSL/NIfTI This is an output format specifically designed for use with FSL. It should be emphasized that this is a fully compliant NIfTI file, and not an FSL-specific variant. However, data order and some header fields have been modified make use with FSL a bit smoother. Data is flipped top bottom, as in the Analyze 7.5 output, to avoid problems with FSL's coordinate system. This will also result in a more intuitive view in FSLView and other FSL-generated reports. All transformations are accounted for in the qform fields. Also, the data type for unsigned 16 bit data has been changed to signed 16 bit, as FSL does not yet support the unsigned 16 bit data type. This should not be a problem for MRI data, as it uses only the lower 12 bits. Also, for Siemen's diffusion images, bvals and bvecs files will be created automically. The values in these files will reflect any rotation of the image volume relative to the gradient coordinates, and will be correspond to the order of the data on disk. ---End quote--- From this description, if we want to use "SPM99/Analyze" format .hdr/.img images, we need to set defaults.analyze.flip = 0, as Dorian points out. However, not so many people use SPM99 anymore, so I would suggest the following; 1) if you use SPM5 and later, convert image with "NIFTI" or "FSL NIFTI" format. (The difference between these two formats is described above.) 2) if you need to use SPM2, convert image with "Analyze7.5" format, and set defaults.analyze.flip=1. Regards, Kiyotaka On Thu, 26 Feb 2009 12:37:26 +0100 Dorian P <[log in to unmask]> wrote: > Hello Steven, Artur and Hakan > > As I had the same problem I have to tell that the problem didn't arise > from MRIconvert, but from SPM. MRIconvert outputs correct files. Then > there is a setting at the spm_defaults.m: > defaults.analyze.flip = 1; % <<= Very important. Relates to L/R > > If you leave this to "1" SPM will flip the images again and all your > results are inverted. I had to set it to "0" and run the analysis > again to obtain correct results. > > Regards. > Dorian > > 2009/2/26 Hakan Kutucu <[log in to unmask]>: > > Hi, > > > > Artur is absolutely right. Left and right sides are changed by MRIConverter. > > I used it and my calculations are wrong now. > > > > > > ----- Original Message ----- From: "Artur Pilacinski" > > <[log in to unmask]> > > To: <[log in to unmask]> > > Sent: Wednesday, February 25, 2009 2:43 PM > > Subject: Re: [SPM] How to Use MRIConverter 2.0 to Convert multiple volumes > > fMRI DICOM files to SPM Analyze images: still no solution > > > > > > Hi, > > > > And why don't you use the SPM built-in converter? > > > > > > If you use MRI Converter 2.0, be aware that it causes a few more problems > > with converting images from GE Signa scanners (one of them is swapping of L > > & R sides of image, depending on sequence properties). I also attempted to > > use this software (seemed quite usable at first glance), but it turned out > > to > > cause too much of trouble. > > > > > > Best, > > > > Artur > > ----------------------------- Kiyotaka Nemoto, M.D. Department of Psychiatry, Ikeda Hospital 3690-2 Kaiharaduka Ryugasaki, Ibraki 301-0856 Japan E-mail: [log in to unmask]