Segmentation of MR Brain Images
for Delineation of Tumors

(Funded by Ohio Aerospace Institute)


This work involves design and implementation of an automatic method for segmentation and extraction of brain tumors in brain magnetic resonance (MR) images. The method being developed will be automatic. However, in cases where image quality is poor, or the tumors are not well defined, the software requires the user to specify the approximate position of a tumor by the use of a mouse input device. This system allows the user to view an extracted tumor, determine its volume and shape, and modify it if desired. The developed software is planned to be included in the software package currently in use at Cleveland Clinic.

The segmentation is carried out in two steps. In the first step, by intensity thresholding, the approximate position and shape of a tumor is determined. Then in the second step, the extracted tumor is refined using edge information in the image. An active surface model is used to get from the initial segmentation to the final one. The surface model keeps the topology of a tumor closed and stops the tumor from being merged with other nearby regions. The accuracy of the method is determined using brain MR images that have already been manually segmented by a neuroradiologist.

Images from real patients are used in this study to evaluate the accuracy of the method and its ease of use. The images being used have been obtained from Cleveland Clinic. No new images are acquired solely for the purpose of this study.

P.I. Ardy Goshtasby, WSU

Co-P.I. Jim Leonard, Wright Lab.

Collaborator: Charles Steiner, Cleveland Clinic

Student Participants: Marcel Jackowski, Thisath Kularatna


Segmentation Results

I. Slice-by-slice segmentation (2-D)

II. Direct volume segmentation (3-D)


[WSU Home Page] [CSE Department Home Page] [Intelligent Systems Lab Page]

For more information contact: A. Goshtasby (ardeshir@cs.wright.edu).

Last modified: 4/25/97.