PET segmentation challenge using a data management and processing infrastructure


The methods will be assessed with voxel-by-voxel ground-truth from the numerical simulation in the case of simulated images, from associated high resolution images / known characteristics of objects in the case of physical phantoms, and from either histopathological measurements or statistical consensus of several manual delineations by experts for clinical images. Performance of the methods will be assessed according to the two metrics sensitivity and positive predictive value. Methods will be ranked according to both metrics separately, as well as according to the mean of both. Although the ranking according to the mean of both metrics will constitute the main ranking, ranking according to each of the metric separately will also be provided, as it may be of interest to consider them for specific applications (i.e., sensitivity could be considered more important for target delineation in radiotherapy, whereas specificity could be considered more important for quantitative parameters extraction for diagnosis or therapy monitoring).