A Controlled Generative Model for Segmentation of Liver Tumors
Nasim Nasiri, Amir Hossein Foruzan, Yen‐Wei Chen · 2019
In this paper, we introduce a generative model representing the formation of liver CT images and employ it for segmentation of hepatic tumors. We extend the graphical model used in label fusion techniques for the segmentation of multi-modality Magnetic Resonance brain images. To resemble atlas data, we segment the first slice by a physician and propagate it for processing of subsequent slices. We include a knowledge-based constraint in the model that is considered as a key difference with classical generative models. We compared our method with conventional Bayesian techniques and state-of-the-art hepatic tumor segmentation algorithms. We achieved a measure of 0.86 ± 0.06 and 0.76 ± 0.09 regarding Dice and Jaccard metrics respectively that is better than recent researches.