Haitao Zhu, Presenter: Nothing to Disclose

Kazuyuki Demachi, Abstract Co-Author: Nothing to Disclose

To develop a method to correct chemical shift misregistration by two measurements and simple magnitude operation.

For an object composed only by water and lipid, Fourier domain data can be expressed by the sum of the signals. F(k)=Fw(k)+Fl(k)exp(-i2πΔf n(k)/BW) where Δf is off-resonance frequency; n(k) is the sample point number relative to the 0 point; BW is the bandwidth. Another set of data are acquired with different bandwidth BW'. Since acquisition time is much shorter compared with relaxation times, it is assumed that Fw(k) and Fl(k) are not affected by bandwidth. Therefore： F'(k)=Fw(k)+Fl(k)exp(-i2πΔf n(k)/BW') Fw(k) and Fl(k) can be calculated by solving the two equations above. The image without misregistration can be obtained by inverse Fourier transform of F0(k)=Fw(k)+Fl(k) Phantom was made by a smaller tube filled with bean oil sealed in a bigger one filled with water. Experiment was implemented on a Varian 4.7T animal MRI system using spin echo pulse sequence. TR=500ms, TE=30ms; bandwidths are 21.3, 42.5, 85.0 kHz respectively; FOV=50mm*50mm; Matrix=256*256. Fourier domain data are generated by magnitude data.

Significant misregistration can be noticed in the images without correction. Spatial shifts of the lipid components are 8, 4 and 2 pixels for 21.3, 42.5, 85.0 kHz bandwidths respectively. Misregistrations are removed after correction using any combination of two images above. The results are compared with the method using opposite readout gradients and show better correction performance. Noise is analyzed by using "energy" of the image which is the summation of the absolute square of each pixel. Correction using two different bandwidths has much lower energy which means high SNR. Contrast to the method using 4 measurements to avoid the singularity, this work utilizes 2 measurements and uses their average as the approximation at the singularity points. Results show no deterioration in image quality, but the scanning time is reduced to half.

This work provides an approach to correct chemical shift misregistration by two arbitrary bandwidths without the need of phase data. It also performs well when the spatial shift is much larger than one pixel.

The two-bandwidth-correction method demonstrates high SNR performance and good time efficiency, which can be a supplementary method to other lipid misregistration correction techniques.

Zhu, H, Demachi, K, Chemical Shift Misregistration Correction Utilizing Images at Two Different Bandwidths.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8001401.html