Howdie Bruce,
At this lab (NRIMS in Cambridge, MA; www.nrims.harvard.edu), the machine provides a stack of images that, when reassembled, can replicate the entity (i.e., nucleus) being studied. We have been able to read these images into Slicer as TIFF files as well as in the raw data format produced by the machine.
After generating a 3D representation of the entity (presumably with FreeSurfer), we want to be able to segment areas of interest within this volume (comparable to focusing on the hypothalamus in the brain, but the boundaries may be user-defined and not anatomically-defined i.e. only part of the hypothalamus), and have the program automatically generate ROIs based on pixel count/size. The pixels contain quantitative information and thus their value cannot be tampered with.
Papers that you have published (Whole Brain Segmentation (2002); Automatically Parcellating (2004); Sequence Independent Segmentation (2004); Cortical Surface-Based Analysis I . (1999)) attest to the feasibility of this endeavor to automatically generate ROIs. We already have the ability to manually draw ROIs with a separate program. But, analogous to the example of MR technicians you referred to in several of these papers, this process is time-consuming. Knowing that the method you devised is as if not more accurate than those trained to do so is reassuring, especially since those here who would manually draw ROIs are not as specialized in this practice like MR technicians. But the details of the actual process referred to in these papers can be left for another day. Before that time comes, more basic problems have to be solved.
The online tutorial that incorporates Slicer and FreeSurfer (link on http://wiki.na-mic.org/Wiki/index.php/Slicer:Workshops:User_Training_101) shows how volumes/surfaces are generated in FreeSurfer from the 3 orthogonal planes in Slicer. Since we have been able to read our images into Slicer, we are now seeking to take the next step of generating volumes and surfaces from the 3 planes in Slicer.
However, there are several potential hindrances I foresee that could be deal-enders. I am hoping that you can either verify such a conclusion or provide a means of getting around these potential roadblocks.
First is the issue of file format. To maintain the quantitative data contained in the images, we have to work with either the raw data format that is generated by the machine or with 16-bit TIFF files that have been converted from this raw data. Other formats risk losing or will flat out lose the information that has to be preserved. Slicer has been able to read the TIFF and raw data formats, but in a previous correspondence with a user on 07-27-06, you noted that FreeSurfer does not support TIFF files. 11 months later, is this still the case or has the program been modified to accept files in TIFF format?
In your reply to this user, you said that FreeSurfer supports MRI volumes as input. The online demo for incorporating FreeSurfer referred to above has screen shots from Slicer. So if we can read our data as TIFF files into Slicer and save the collection of these TIFF files as a volume file, would FreeSurfer be content to read it as a volume file rather than as a collection of TIFF images?
Second, it seems like the aseg file is crucial for the generation of 3D surfaces and volumes. Is this file provided by the MRI machine? (If so, we likely have been stopped in our tracks.) Or is there a way I can create an aseg file from the cross-sectional images I have? (My impression is that this aseg file provides the boundaries for the surfaces/volumes to be generated.)
Thanks again.
Greg
_________________________________________________________________ Need a break? Find your escape route with Live Search Maps. http://maps.live.com/default.aspx?ss=Restaurants~Hotels~Amusement%20Park&...