Tumor Extraction for Brain Magnetic Resonance Imaging Using Modified Gaussian Distribution
Magnetic Resonance Imaging (MRI) is extensively used in the study of brain. Segmentation of MR brain images is necessary for a number of clinical investigations of various complexity, change detection, cortical labeling, and visualization in surgical planning. The volume of enhancing lesions, fol...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2006
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/6109/1/Fk_2006_19.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Magnetic Resonance Imaging (MRI) is extensively used in the study of brain.
Segmentation of MR brain images is necessary for a number of clinical
investigations of various complexity, change detection, cortical labeling, and
visualization in surgical planning. The volume of enhancing lesions, following the
administration of paramagnetic contrast agent is an important indicator of pathology
in multiple sclerosis (MS). Manual estimation of enhancing lesion volumes
introduces significant errors, and operator bias, besides being time consuming and
subjective. Therefore, there is a need for automatic identification and estimation of
volumes of the present MS lesions specially by dealing with a large number of
images that are typically acquired in multi-center clinical trials.
In the developed techniques, 150 T1- and T2-weighted spin echo images were taken
from the routine scans of Kuala Lumpur General Hospital.Multiple sclerosis lesions visualized by morphological MRI are classified through a
feature map technique on T1 weighted MRI tissue. Gray level morphology methods
are used to make tissue types in the images more homogenous and minimize
difficulties with connections to outside tissue. A method for hzzy connectedness
and combinations of the different segmentation techniques were experimented. A
gain-based correction method; probability density function model are used to cluster
white and gray matters, cerebrospinal fluid, and meninges. Results of segmentation
have been validated by a group of neuro-radiologists.
3D visualization has been implemented for the segmented regions as well as brain
lesion. The visualization of the segmented structures uses a combination of volume
rendering and surface rendering.
The mutual information algorithms used in this work has been developed and
experimented in the system and has proven to yield more accurate and stable results
than other algorithms.
Currently testing the validation of the proposed segmentation in a validation study
that compares resulting MS lesion as well as gray and white matter tissue structures
with Neural Network expert segmentation system. The proposed method versus
Neural Network rater validation showed an average validation score of overlap ratio
of >85% for gray and white matters tissue segmentation and for MS lesion the rater
validation showed an average overlap ratio of > 87%. |
|---|