Related Experiment Videos
Application of the extremum stack to neurological MRI
A Simmons1, S R Arridge, P S Tofts
1Department of Clinical Neurosciences, Institute of Psychiatry, London, UK. a.simmons@iop.bpmf.ac.uk
IEEE Transactions on Medical Imaging
|September 15, 1998
Summary
The extremum stack method for image segmentation shows promise but struggles with elongated objects. Adaptive diffusion techniques can improve its performance by preventing premature merging of anatomical regions in medical imaging.
Area of Science:
- Computer Vision
- Medical Imaging Analysis
- Image Segmentation
Background:
- The extremum stack is a multiresolution image description and segmentation method.
- It analyzes intensity extrema across scale space for data-driven segmentation.
- Its applicability to neurological magnetic resonance imaging (MRI) data is explored.
Purpose of the Study:
- To evaluate the performance of the extremum stack for image segmentation.
- To identify limitations and propose improvements for the extremum stack method.
- To assess its utility in analyzing neurological MRI data.
Main Methods:
- The extremum stack method was applied to neurological MRI data.
- Experimental evaluation of shift-, scale-, and rotation-invariance was performed.
- Modifications using variable conductance diffusion were investigated.
Main Results:
- The extremum stack demonstrated shift-, scale-, and rotation-invariance.
- It produced natural segmentations for compact anatomical regions.
- Elongated objects were poorly handled, with premature merging of regions observed.
Conclusions:
- The extremum stack is a potentially useful image segmentation tool, particularly for compact structures.
- Adaptive variable conductance diffusion effectively addresses premature merging issues.
- Improvements enhance the robustness of the extremum stack for medical image analysis.