Regular VOI atlases in PMOD are based on a label image which specifies for each pixel to which VOI it belongs. The situation is different for a probabilistic atlas. There, each pixel has a probability value assigned for belonging to a specific VOI. Therefore, probabilistic atlases require a real-valued image for every VOI in the atlas. In PMOD, the probability images are arranged as frames in a dynamic series. In the label list file the third column therefore corresponds to the frame encoding a VOI's probability, rather than its label number in a static image.
Note: Probabilistic atlases can only be applied to already normalized images. The normalization functionality is not supported.
When applying a probabilistic atlas for statistics, calculation takes the probabilities into account as describe for the different measures. For some of them a threshold criterion is required, which can be entered in the Probability Threshold field. The example below illustrates the functionality using a probabilistic atlas with seven Thalamus sub-regions.
The conversion of probabilistic VOIs into contour VOIs requires specification of a discretization threshold in the Set threshold of probability field of the outlining window.
Note that with multiple selected VOIs the outlining procedure constructs non-intersecting VOIs by assigning a pixel to the VOI with the highest p-value.