Robust organ segmentation is a prerequisitefor computer-aided diagnosis, quantitative imaging analysis,pathology detection, and surgical assistance. For organs with highanatomical variability (e.g., the pancreas), previous segmentationapproaches report low accuracies, compared with well-studied organs, such as the liver or heart. We present an automatedbottom-up approach for pancreas segmentation in abdominalcomputed tomography (CT) scans. The method generates ahierarchical cascade of information propagation by classifyingimage patches at different resolutions and cascading (segments)superpixels. The system contains four steps: 1) decomposition ofCT slice images into a set of disjoint boundary-preserving super-pixels; 2) computation of pancreas class probability maps via dense patch labeling; 3) superpixel classification by pooling bothintensity and probability features to form empirical statistics incascaded random forest frameworks; and 4) simple connectivitybased post-processing. Dense image patch labeling is conductedusing two methods: efficient random forest classification on image histogram, location and texture features; and more expensive (but more accurate) deep convolutional neural network classification,on larger image windows (i.e., with more spatial contexts).Over-segmented 2 ? D CT slices by the simple linear iterative clustering approach are adopted through model/parameter cali-bration and labeled at the superpixel level for positive (pancreas)or negative (non-pancreas or background) classes. The proposed method is evaluated on a data setof80manuallysegmented CT volumes, using six-fold cross-validation.