Fundamental and Clinical Studies for Effectiveness of Zero-filling Interpolation on k-space for Improvement of Sharpness in Magnetic Resonance Imaging Poster No.: C-0709 Congress: ECR 2014 Type: Scientific Exhibit Authors: E. Ebihara 1, H. Nagashima 2, N. Hayashi 2, M. Okawara 1, M. Kudo 1, R. Kitazaki 1, S. Imai 3, T. Ogura 2, K. Doi 2 ; 1 Maebashi-city/JP, 2 Maebashi/JP, 3 Maebashi City/JP Keywords: DOI: Osteoporosis, Physics, MR, MR physics 10.1594/ecr2014/C-0709 Any information contained in this pdf file is automatically generated from digital material submitted to EPOS by third parties in the form of scientific presentations. References to any names, marks, products, or services of third parties or hypertext links to thirdparty sites or information are provided solely as a convenience to you and do not in any way constitute or imply ECR's endorsement, sponsorship or recommendation of the third party, information, product or service. ECR is not responsible for the content of these pages and does not make any representations regarding the content or accuracy of material in this file. As per copyright regulations, any unauthorised use of the material or parts thereof as well as commercial reproduction or multiple distribution by any traditional or electronically based reproduction/publication method ist strictly prohibited. You agree to defend, indemnify, and hold ECR harmless from and against any and all claims, damages, costs, and expenses, including attorneys' fees, arising from or related to your use of these pages. Please note: Links to movies, ppt slideshows and any other multimedia files are not available in the pdf version of presentations. www.myesr.org Page 1 of 14
Aims and objectives The zero-filling interpolation (ZIP) technique in magnetic resonance imaging (MRI) is a method that can expand the original matrix size without increasing the scanning time by placing the acquired data in the central regions on k-space, and by filling the data of zero in the outer regions at high spatial frequencies. Figure 1 illustrates ZIP processing in k- space, which has been reported to improve visual sharpness of MR images [1]. On the other hand, it has been reported also that detailed objects were not confirmed clearly by applying ZIP processing [2]. Therefore, we investigated the characteristics of ZIP processing by performing physical and visual evaluations, and examined the usefulness of ZIP. Images for this section: Fig. 1: Illustration of zero-filling interpolation (ZIP) technique in k-space. Page 2 of 14
Methods and materials A slit image with matrix size of 512 512 pixels was prepared by the width of one pixel in the central portion of the image. The shaded slit image was also created by applying a Gaussian filter with a kernel size of 9 9 pixels to simulated slit image. Figure 2 shows the line spread function (LSF) for slit image without and with Gaussian filter. These slit images were reconstructed to square images of 1024 and 2048 pixels as side length by using our ZIP software. Figure 3 shows the overall scheme for the automated application of ZIP processing in MR images. The LSF was determined from these slit images, and the modulation transfer function (MTF) was then obtained by applying the Fourier transform. Furthermore, the uniform images and pin pattern images were obtained by scanning an MRI phantom with use of 1.5 T-MRI scanners (Ingenia, PHILIPS). For the scanning conditions, the fast-spin echo sequence was chosen, and the scan parameters were set as follows: repetition time = 400 ms, echo time = 16 ms, number of signal averages = 1, slice thickness = 1.7 mm, field of view (FOV) = 128 128 and 256 256 mm, and matrix size = 256 256 pixels. These two images were reconstructed to square images of 512, 1024, and 2048 pixels as side length. For comparison between the scanned original images and reconstructed images, the signal-to-noise ratio (SNR) was evaluated by using these uniform images. Moreover, the shape reproducibility and the resolution for the pin pattern images were evaluated visually by using Scheffe's method of paired comparison with ten observers. As a visual evaluation method, various images obtained with and without ZIP processing were displayed side by side on a liquid-crystal display (LCD) monitor. Two different images were then evaluated on a discrete rating scale. Images for this section: Page 3 of 14
Fig. 2: Slit images and signal profile curves: (a) without and, (b) with Gaussian filtering with a kernel size of 9 9 pixels. Page 4 of 14
Fig. 3: Overall computerized scheme for automated application of ZIP processing in MR images. Page 5 of 14
Results Figure 4 shows the signal characteristic of ZIP processing in simulated slit images. On the LSF with ZIP processing, the findings of undershoots and noise appeared. However, the amplitude of these findings was reduced in the LSF of slit image with Gaussian filtering. Figures 5 and 6 show the MTF determined from two different slit images with and without ZIP processing. In Figure 5, the MTF in reconstructed image by ZIP processing without filtering was over 1.0 at low frequencies, because of undershoots appeared on the LSF. As a measure of the spatial resolution, we determined the spatial frequency at the MTF value of 0.50 in figure 6, and the increase in spatial frequency was observed by applying the ZIP processing. These results seem to suggest that ZIP processing can increase slightly sharpness in the MR image. Figure 7 shows the SNRs in original images acquired with different matrix size (pixel size) and reconstructed images with ZIP processing. It is apparent that the SNRs in all conditions were comparable. We conclude that the noise appeared by the ZIP processing was very similar to the original noise in image acquired by MRI scanner probably because the noise with ZIP processing was embedded in the acquired image. Figure 8 shows results of the visual evaluation using Scheffe's method of paired comparisons. In Figures 8(a) and (b), a significant difference was observed between the original images and the reconstructed images with ZIP processing (P < 0.001). Furthermore, for the original images acquired with FOV of 256 256 mm in Fig. 8(b), there were significant differences among the images with ZIP processing for different matrix size (P < 0.001). Figure 9 shows the pin pattern images acquired with FOV of 128 128 mm and three different images reconstructed with ZIP processing. In comparing images of 0.5 mm as a pixel size with and without the ZIP processing, the reproducibility of the object shape among images appeared different visually and improved noticeably by applying ZIP processing. Figure 10 shows the pin pattern images acquired with FOV of 256 256 mm and reconstructed images with ZIP processing. During the visual evaluation using original images at the pixel size of 1.0 mm, all of signals for pin size of 1.0 mm or more could be identified with ZIP processing. This result shows that it is possible to separate the connected, distorted pin signals by using ZIP processing. However, signals less than 1.0 mm showed no improvement. Therefore, the ZIP processing might be able to separate the connected signals such as adjacent vessels in a clinical image when the size of signals may be larger than the pixel size in scanning. Thus, it seems that ZIP processing would be useful for imaging with a wide FOV, and for improving in sharpness of small objects when the scanning time is reduced. Images for this section: Page 6 of 14
Fig. 4: Signal profile curves: (a), (b) without Gaussian filtering; (c), (d) with Gaussian filtering; (a), (c) without ZIP processing; and (b), (d) with ZIP processing at matrix size of 1024 1024 pixels. Page 7 of 14
Fig. 5: Modulation transfer function (MTF) measured from slit images without Gaussian filtering. Page 8 of 14
Fig. 6: MTF measured from slit images with Gaussian filtering. Page 9 of 14
Fig. 7: Signal-to-noise ratio (SNR) in original images acquired with different matrix size (pixel size) and reconstructed images with ZIP processing. Page 10 of 14
Fig. 8: Results of visual evaluation by Scheffe's method of paired comparisons using original images and reconstructed ZIP images: (a) original image acquired with 128 128 mm FOV and 256 256 pixels matrix size, (b) original image acquired with 256 256 mm FOV and 256 256 pixels matrix size. Page 11 of 14
Fig. 9: Pin pattern images: (a) original image acquired with 128 128 mm FOV and 256 256 pixels matrix size, (b); (c); and (d) reconstructed images of matrix size of 512 512, 1024 1024, and 2048 2048 pixels, respectively, with ZIP processing. Page 12 of 14
Fig. 10: Pin pattern images: (a) original image acquired with 256 256 mm FOV and 256 256 pixels matrix size, (b); (c); and (d) reconstructed images of matrix size of 512 512, 1024 1024, and 2048 2048 pixels, respectively, with ZIP processing. Page 13 of 14
Conclusion Our results indicated that the application of ZIP processing can improve the sharpness of small objects in clinical MR images. However, further studies for the verification of clinical usefulness with ZIP processing will be required in the future. Personal information References 1. T. Nakamura, S. Naganawa, T. Koshikawa, et al: High-spatial-resolution MR cisternography of the cerebellopontine angle in 90 seconds with a zero-fill interpolated fast recovery 3D fast asymmetric spin-echo sequence, AJNR Am J Neuroradiol, Vol.23, No.8, pp.1407-1412, 2002. 2. A. Ogura, F. Maeda, A. Miyai, et al: The property of zero-fill interpolation on slice plane of magnetic resonance imaging, Japanese Journal of Clinical Radiology, Vol.52, No.6, pp.799-806, 2007. Page 14 of 14