Contents

• Abstract
• Acknowledgments
• 1 Introduction
  • 1.1 Motivation
    • 1.1.1 Multiple Sclerosis Lesion Segmentation
    • 1.1.2 Automatic Registration
    • 1.1.3 Image Compression
  • 1.2 Goals
  • 1.3 Research Methodology
  • 1.4 Thesis Overview and Scope
• 2 MR Imaging and Data Characteristics
  • 2.1 Overview
  • 2.2 Basic Principles of MRI
  • 2.3 RF Inhomogeneity
    • Tissue Intensity Profiles and Variances
  • 2.4 Partial Volume Effect
  • 2.5 Background Noise Characteristics
  • 2.6 Brain Tissue Characteristics
  • 2.7 Summary
• 3 Previous Work
  • 3.1 Overview
  • 3.2 Intracranial Boundary Detection
    • 3.2.1 Automatic Thresholding
    • 3.2.2 Region Growing and Boundary Detection
    • 3.2.3 Statistical Segmentation
    • 3.2.4 Active Contour Models
  • 3.3 Radio Frequency Correction
    • 3.3.1 ``Phantom'' Correction
    • 3.3.2 Approximate ``Phantom'' Correction
    • 3.3.3 Surface Fitting
    • 3.3.4 Implicit Correction
    • 3.3.5 Homomorphic Filtering
  • 3.4 Summary
• 4 Nonlinear Anisotropic Diffusion Filtering
  • 4.1 Overview
  • 4.2 Definition
  • 4.3 Discrete Implementation
    • 4.3.1 1D Filtering
    • 4.3.2 2D Filtering
    • 4.3.3 3D Filtering
    • 4.3.4 Stability Analysis
  • 4.4 Experiments
    • 4.4.1 Edge Tracking
    • 4.4.2 Noise Reduction
    • 4.4.3 Intracranial Boundary Detection
      • 1D Filter
      • 2D Filter
      • 3D Filter
    • 4.4.4 Performance
      • 1D Filter
      • 2D Filter
      • 3D Filter
  • 4.5 Summary
• 5 Active Contour Models (``Snakes'')
  • 5.1 Overview
  • 5.2 Energy Formulation
    • 5.2.1 Internal Energy
      • Continuity Energy
      • Balloon Force
    • 5.2.2 External Energy
      • Image Intensity Energy
      • Image Gradient Energy
    • 5.2.3 Regularization
      • Continuity Energy
      • Balloon Energy
      • Intensity Energy
      • Gradient Energy
  • 5.3 Experiments
    • 5.3.1 Intracranial Boundary Detection
    • 5.3.2 Contribution of Diffusion Filtering
    • 5.3.3 Contribution of Gradient Direction
    • 5.3.4 Contribution of Balloon Force
    • 5.3.5 Effect of the Parameters
      • Continuity Energy Weighting ()
      • Balloon Energy Weighting ()
      • Intensity Energy Weighting ()
      • Intensity Energy Sensitivity ()
      • Gradient Energy Weighting ()
      • Gradient Energy Sensitivity ()
      • Maximum Iterations ()
      • Number of Contour Point Moves ()
      • Deformable Contour Size
      • Neighborhood Search Space Size ()
    • 5.3.6 Performance
  • 5.4 Summary
• 6 Homomorphic Filtering
  • 6.1 Overview
  • 6.2 Spectral Domain Homomorphic Filtering
    • Poor Performance
    • Circular Convolution
  • 6.3 Spatial Domain Homomorphic Filtering
  • 6.4 Homomorphic Filtering using a Low Pass Filter
  • 6.5 Experiments
    • Results
    • Performance
  • 6.6 Summary
• 7 Intracranial Boundary Detection and RF Correction
  • 7.1 Overview
  • 7.2 Segment Head
    • 7.2.1 Histogram Image Volume
    • 7.2.2 Determine Background Threshold Level
    • 7.2.3 Remove Noise
  • 7.3 Generate Initial Brain Mask
    • 7.3.1 Normalize Image Volume
    • 7.3.2 Nonlinear Diffusion Filter
    • 7.3.3 Automatic Threshold
    • 7.3.4 Eliminate Misclassified and Non-brain Regions
  • 7.4 Generate Final Brain Mask
    • 7.4.1 Convert the Initial Mask to a Set of Contours
    • 7.4.2 Compute the Image Gradient
    • 7.4.3 Deform the Contours
  • 7.5 Correct Intensity
    • 7.5.1 Apply Brain Mask
    • 7.5.2 Homomorphic Filter
    • 7.5.3 Normalize Result
• 8 Results
  • 8.1 Overview
  • 8.2 MRI Data Set Descriptions
    • 8.2.1 Data Set 1
    • 8.2.2 Data Sets 2 and 3
    • 8.2.3 Data Sets 4 and 5
  • 8.3 Head Mask
    • 8.3.1 Data Set 1
    • 8.3.2 Data Set 2
    • 8.3.3 Data Set 3
    • 8.3.4 Data Set 4
    • 8.3.5 Data Set 5
    • 8.3.6 Performance
    • 8.3.7 Discussion
  • 8.4 Initial Brain Mask
    • 8.4.1 Data Set 1
    • 8.4.2 Data Set 2
    • 8.4.3 Data Set 3
    • 8.4.4 Data Set 4
    • 8.4.5 Data Set 5
    • 8.4.6 Performance
    • 8.4.7 Discussion
  • 8.5 Final Brain Mask
    • 8.5.1 Data Set 1
    • 8.5.2 Data Set 2
    • 8.5.3 Data Set 3
    • 8.5.4 Data Set 4
    • 8.5.5 Data Set 5
    • 8.5.6 Performance
    • 8.5.7 Discussion
  • 8.6 Intensity Correction
    • 8.6.1 Data Set 1
    • 8.6.2 Data Set 2
    • 8.6.3 Data Set 3
    • 8.6.4 Performance
    • 8.6.5 Discussion
  • 8.7 Summary
• 9 Summary
  • 9.1 Review
  • 9.2 Conclusions
    • 9.2.1 Intracranial Boundary Detection
    • 9.2.2 RF Correction
  • 9.3 Future Work
    • 9.3.1 Intracranial Boundary Detection
    • 9.3.2 RF Correction
• References
• About this document ...