Integration of a knowledge-based constraint into generative models with applications in semi-automatic segmentation of liver tumors

Amir Hossein Foruzan | Nasim Nasiri | Yen-Wei Chen

---

URL :   http://research.shahed.ac.ir/WSR/WebPages/Report/PaperView.aspx?PaperID=158973
Date :  2020/02/06
Publish in :    Biomedical Signal Processing and Control
DOI :  https://doi.org/10.1016/j.bspc.2019.101725
Link :  http://dx.doi.org/10.1016/j.bspc.2019.101725
Keywords :Generative modelsLiver tumorsLiver CT imagesGraphical modelsKullback-Leibler divergenceBayesian segmentationa

Abstract :

---

tAccurate delineation of liver tumors in medical images is a vital step in diagnosis, treatment planning, andmonitoring. In this paper, we utilize a generative model for segmentation of abnormal liver regions. Afterpreprocessing of an input image, the ROI of the tumor is determined, and the boundary of the abnormalregion in a single slice is specified. Then, the remaining slices are processed by a generative model that isenhanced by the integration of a constraint. We search for the boundary of the tumor by a probabilisticapproach and obtain the solution using the Bayesian inference. The Kullback-Leibler divergence is usedto measure the consistency of the results to the model’s constraint. We evaluated the proposed methodusing synthetic and clinical data. In the public dataset, we achieved a Dice measure of 0.84 ± 0.06, whichoutperforms state-of-the-art hepatic tumor segmentation algorithms. Concerning all available clinicalimages, the Dice index of the proposed method is 0.90 ± 0.03.