Related Experiment Videos
Improved quality of re-sliced MR images using re-normalized sinc interpolation
N A Thacker1, A Jackson, D Moriarty
1Imaging Science and Biomedical Engineering, The Medical School, University of Manchester, Manchester, UK. nat@sv1.smb.man.ac.uk
Journal of Magnetic Resonance Imaging : JMRI
|October 3, 1999
Summary
Re-normalizing sinc interpolation kernels significantly improves performance in magnetic resonance (MR) imaging. A small re-normalized kernel achieves results comparable to large kernels, reducing computation time by up to 30x.
Area of Science:
- Medical Imaging
- Image Processing
- Computational Science
Background:
- Sinc interpolation is a common method for resampling medical images.
- Standard sinc interpolation can be computationally intensive and may introduce errors.
- Optimizing interpolation kernels is crucial for efficient and accurate image processing.
Purpose of the Study:
- To evaluate the performance improvement of re-normalized sinc interpolation kernels.
- To compare the accuracy and computational efficiency of standard versus re-normalized kernels.
- To assess the potential for clinical application of optimized MR volume re-slicing.
Main Methods:
- Compared standard and re-normalized sinc kernels of various sizes.
- Used data from four magnetic resonance (MR) imaging sequences.
- Evaluated interpolation error using systematic pixel intensity offset and residual variance.
- Compared computation time with standard sinc interpolation and the AIR program.
Main Results:
- A small (5x5x5) re-normalized kernel yielded errors comparable to a large (13x13x13) 'gold standard' kernel.
- Re-normalized kernels achieved up to 30 times faster computation compared to standard sinc interpolation.
- The re-normalization method demonstrated significant performance gains.
Conclusions:
- Re-normalizing the interpolation kernel offers substantial performance improvements for sinc interpolation in MR imaging.
- This optimization allows for significantly reduced computation time without compromising accuracy.
- The findings bring MR volume re-slicing closer to practical clinical implementation.