Tumor segmentation from magnetic resonance (MR) images may aid in tumor tre
atment by tracking the progress of tumor growth and/or shrinkage. In this p
aper we present the first automatic segmentation method which separates non
-enhancing brain tumors from healthy tissues in MR images to aid in the tas
k of tracking tumor size over time. The MR feature images used for the segm
entation consist of three weighted images (T1, T2 and proton density (PD))
for each axial slice through the head. An initial segmentation is computed
using an unsupervised fuzzy clustering algorithm. Then, integrated domain k
nowledge and image processing techniques contribute to the final tumor segm
entation. They are applied under the control of a knowledge-based system. T
he system knowledge was acquired by training on two patient volumes (14 ima
ges). Testing has shown successful tumor segmentations on four patient volu
mes (31 images). Our results show that we detected all six non-enhancing br
ain tumors, located tumor tissue in 35 of the 36 ground truth (radiologist
labeled) slices containing tumor and successfully separated tumor regions f
rom physically connected CSF regions in all the nine slices. Quantitative m
easurements are promising as correspondence ratios between ground truth and
segmented tumor regions ranged between 0.368 and 0.871 per volume, with pe
rcent match ranging between 0.530 and 0.909 per volume. (C) 2001 Elsevier S
cience B,V. All rights reserved.