http://loni.ucla.edu/Software/rss.php?package=BrainParser LONI Brain Parser software segments 56 anatomical brain structures (both cortical and sub-cortical based on the LPBA40 atlas, Shattuck et al. 2007) by using a learning-based approach and a pre-trained model targeted at common structures of interest. The LONI Brain Parser will expect skull-stripped, registered and normalized data before segmenting (for more information, see the documentation) and this data will subsequently need to be warped back to its original space. Alternatively, a LONI Pipeline Workflow containing LONI Brain Parser is available for download or online usage with these pre- and post- processing steps already set up. The automated segmentation portion usually takes 30 minutes while the whole workflow takes about an hour depending on the complexity of the data and computing power available. Although LONI Brain Parser comes pre-trained with an ideal model, the software also has the ability to train the model to segment the user\'s own regions of interest. The user will need a set of at least 10 TI MRI volumes, hand-segmented, ground truth labels and will need to supervise the training process. The quality of the MRI volumes and the labels will affect the success of the auto-segmentation and must be in correspondence. There is no limitation to how few or how many structures the software can be trained to segment. Mon, 23 Nov 2009 23:29:00 +0000 5 0