Related Experiment Video
Updated: May 1, 2026
08:40
Quantitation of Protein Expression and Co-localization Using Multiplexed Immuno-histochemical Staining and Multispectral Imaging
Published on: April 8, 2016
12.4K
Low-complexity atlas-based prostate segmentation by combining global, regional, and local metrics
1The Department of Radiation Oncology, University of California Los Angeles, California 90095.
Medical Physics
|April 4, 2014
Summary
This study introduces an efficient atlas-based segmentation method that reduces nonrigid registrations while maintaining accuracy. The novel approach significantly improves prostate MRI segmentation compared to state-of-the-art methods.
Area of Science:
- Medical Image Analysis
- Computational Anatomy
- Biomedical Engineering
Background:
- Accurate automatic structure segmentation is crucial in medical image processing.
- Atlas-based methods are state-of-the-art but computationally intensive due to numerous nonrigid registrations.
- Deformable anatomical structures necessitate efficient segmentation techniques.
Purpose of the Study:
- To enhance the efficiency of atlas-based segmentation without compromising accuracy.
- To demonstrate the method's validity in MRI-based prostate segmentation.
- To reduce the computational cost of nonrigid registrations in atlas-based segmentation.
Main Methods:
- A hybrid approach combining global, regional, and local metrics for improved accuracy and reduced nonrigid registrations.
- Initial affine registration using global mean squared error (gMSE) for coarse alignment.
- Target-specific regional MSE (rMSE) to select a relevant atlas subset for subsequent nonrigid registration.
- Label fusion based on a weighted combination of rMSE and local MSE (lMSE) with spatial regularization.
Main Results:
- The proposed method achieved a median/mean Dice Similarity Coefficient (DSC) of over 0.87/0.86 on 30 prostate MR images.
- Outperformed state-of-the-art atlas-based segmentation (median/mean DSC 0.84/0.82) using only eight nonrigid registrations.
- Demonstrated significant improvement in computational efficiency compared to existing methods.
Conclusions:
- The method offers desirable scalability, requiring a fixed number of nonrigid registrations independent of atlas size.
- Achieved superior performance and computational efficiency for prostate segmentation compared to current atlas-based approaches.
- The metric-based rationale is extensible to other similarity metrics like correlation or mutual information.
