Meeting Banner
Abstract #1746

A New Adaptive Markov Random Field Model in a Coupled Level Set Framework for Bladder Wall Segmentation in MR Images

Hao Han1, Lihong Li2, Chaijie Duan3, Hao Zhang1, Zhengrong Liang1

1Dept. of Radiology, Stony Brook University, Stony Brook, NY, United States; 2Dept. of Engineering Science & Physics, College of Staten Island of the City University of New York, Staten Island, NY, United States; 3Dept. of Biomedical Engineering, Tsinghua University, Shenzhen, Guangdong, China

Bladder cancer is the fifth leading cause of cancer related death, especially for aged males in the United States. Early detection of bladder lesion is so crucial that advanced techniques are required to precisely and safely differentiate tumors from the normal bladder wall. In this work, we developed a novel approach to precisely segment the inner border and outer border of the bladder wall, and it was shown that our new approach outperforms the previous approach developed in our group. The bladder lesion can be visualized through a 3-D rendering model on wall thickness calculated from the segmented bladder wall.

Abstract PDF Poster PDF

Keywords

abnormal abnormalities achieve adaptive advantages asked become biomedical bladder border borders brook cancer carcinoma cause certain china city class classes classification clinical college colors common computer conducted consistent converged cost coupled crucial cure dataset deaths denotes dept detect detection diagnosed diagnosis digital dilation disease early effective effectively electric engineering evaluate evaluated evaluation eventually evolution evolved evolving expanding experts field fifth finalized firstly fourth framework full functions furthermore generated goal heated identify illustrates improvement initializing inner inside integrate integration intensity island iterative label labels layers lesions lumen measured model modified morphological nature needed neighboring noise novel original outer outside panel path patient phantom phantoms physics posteriori precise preliminary preserving previously primary prior probability process promising proposed radiology random recurrence reduce related rendering reported requires resection robustness safe satisfactory scale science score scores secondly segmentation segmented segmenting simulated since soft stony structure studies surface takes terms thicknesses trans true tumors view volume wall