Dolph et al.

From CADDementia
Jump to: navigation, search

Classification of Alzheimer's disease using structural MRI

C.V. Dolph, M.D.Samad, K.M. Iftekharuddin (Vision Lab, Old Dominion University, VA, USA)


Summary

Alzheimer's disease (AD) typically progresses with the atrophy of white matter along with expansion of cerebrospinal fluid. Our algorithm classifies AD using ratios of WM to CSF in horizontal slices of structural MRI.

Stepwise Explanation

There are 5 major steps to our algorithm: skull stripping, registration, segmentation, feature selection, and classification.

1. Skull stripping

Brainsuite [1] was used to extract the brain. Normally, the default diffusion parameters are used. Each extracted brain is inspected for excessive or inadequate stripping. The diffusion parameters are adjusted if necessary.

2. Registration

Deformable Registration via Attribute Matching and Mutual-Saliency Weighting (DRAMMS) [2] tool is used to register the brain to the Alzheimer’s disease atlas [3] acquired from the LONI at USC.

3. Segmentation

Brainsuite is used again for the segmentation into WM, GM, and CSF

4. Feature Extraction

Using an in iterative process, the best features were found to be the WM to CSF ratios from horizontal slices 25 to 70.

5. Classification

A SVM-RBF classifier is used to determine the class of each subject MRI.

References

  • Shattuck DW and Leahy RM (2002) BrainSuite: An Automated Cortical Surface Identification Tool Medical Image Analysis 8(2):129-142.
  • DRAMMS: Deformable registration via attribute matching and mutual-saliency weighting. Yangming Ou, Aristeidis Sotiras, Nikos Paragios, Christos Davatzikos. Medical Image Analysis, 15(4): 622-639, 2011.
  • Shattuck DW, Sandor-Leahy SR, Schaper KA, Rottenberg DA, and Leahy RM (2001) Magnetic Resonance Image Tissue Classification Using a Partial Volume Model NeuroImage 13(5):856-876.

Contact

If you use this methodology, please cite:

CV Dolph, MD Samad, KM Iftekharuddin (Vision Lab, Old Dominion University, VA, USA) Classication of Alzheimer's disease using structural MRI

For questions on the method, please contact Chester Dolph: cdolp001@odu.edu