Regression Forest-Based Atlas Localization and Direction Specific Atlas Generation for Pancreas Segmentation

Abstract

This paper proposes a fully automated atlas-based pancreas segmentation method from CT volumes utilizing atlas localization by regression forest and atlas generation using blood vessel information. Previous probabilistic atlas-based pancreas segmentation methods cannot deal with spatial variations that are commonly found in the pancreas well. Also, shape variations are not represented by an averaged atlas. We propose a fully automated pancreas segmentation method that deals with two types of variations mentioned above. The position and size of the pancreas is estimated using a regression forest technique. After localization, a patient-specific probabilistic atlas is generated based on a new image similarity that reflects the blood vessel position and direction information around the pancreas. We segment it using the EM algorithm with the atlas as prior followed by the graph-cut. In evaluation results using 147 CT volumes, the Jaccard index and the Dice overlap of the proposed method were 62.1% and 75.1%, respectively. Although we automated all of the segmentation processes, segmentation results were superior to the other state-of-the-art methods in the Dice overlap.

3D FCN Feature Driven Regression Forest-Based Pancreas Localization and Segmentation

“…First-order difference feature f 1 (v) and second-order difference feature f 2 (v) are represented as The regression forest is trained using a training database that includes CT volumes and their corresponding manually segmented pancreas regions. A regression tree is constructed using patches obtained from the training database, as previously explained [8,11].…”

“…These methods depend on other organ segmentation results or require manual interactions. Recently, pancreas segmentation methods have been proposed with automated localization techniques [7,8]. Roth et al [7] used deep convolutional neural networks with holistically-nested networks for localization.…”

“…Roth et al [7] used deep convolutional neural networks with holistically-nested networks for localization. Oda et al [8] used the regression forest technique to localize the pancreas atlas based on global feature values. Other methods [7,8] showed higher pancreas segmentation accuracies than the previous methods.…”

“…The 3D information of pancreas surroundings, which has not adequately considered, is important for more accurate localization. Oda et al [8] used global feature values for localization. A global feature value is the firstorder difference of the CT values sampled at many local regions in CT volumes.…”

“…Global feature values, which only capture the edge information from CT values, do not utilize the more useful information contained in the CT volumes for such localizations as texture or appearance. In these methods [7,8], the tissue or organ information surrounding the pancreas is not well utilized for localization.…”