Segmentation of Subcortical Sructures in MR Images

1
Segmentation of Subcortical Sructures in MR Images

• VPA

2
Outline

• Introduction
• MRI Segmentation.
• Subcortical structures.
• Challenging Problems.
• Low Contrast
• Shape Topological Complexity
• Database
• Tools
• MRI visualization tool.
• Make traning set
• Generate ground truth.
• Segmentation Issue
• Our Task
• Results
• Validation

3
Introduction Subcortical Structures (BG)

• Neurodegenerative diseases, such as Alzheimer and
  Parkinson.
• Early detection requires thorough understanding
  of the chemical and anatomical changes in the
  brain.

• Automatic segmentation methods to detect
  changes
• Determine shape abnormalities and help
  radiologists to find out the functionals of
  different organs in neurological diseases

4
Introduction Subcortical Structures (BG)
VPA
4
5
Challenging Problems Shape Topological
Complexity, Low Contrast

• Low Contrast
• Very similar intensity values for different
  tissues
• Very closed positions of organs.
• Multi pieced structured organs.

VPA
5
6
Database ASM Siemens (Avanto)

• T2 and PD sequence, dicom.
• 18 normal patient
• 1. 5 T
• 512 x 448 dimension
• 3 mm slice thickness

• T1 sequence, dicom
• 3 Patient
• 1. 5 T
• 512 x 448, dim.
• 1 mm slice thickness

7
Tools Visualization
VPA
7
8
Tools Generate Training
VPA
8
9
Segmentation IssueOur Task

• Use of curve evolution based segmentation methods
• Region Based (Chan Vese)
• The use of shape priors
• Embedding non parametric joint shape model into
  segmentation process.
• The use of information, obtained from spatial
  position relation within the organs.
• Embedding relative pose prior information of
  neighboring structures.

10
Segmentation Results

• Putamen Caudate

Without Shape Knowledge
With Joint Shape Knowledge
With Shape Knowledge
11
Segmentation Validation
VPA
11