Department of Radiology - Medical Physics

Many fluorine based contrast agents have multiple resonance peaks in the spectrum and therefore exhibit prominently the so called chemical shift artefact. Several deconvolution techniques have been introduced for correcting these artefacts. However, so far it was never discussed how the image acquisition could potentially influence the possibility to correct the artefacts. Here we focus our interest on optimizing the imaging sequence such that the chemical shift artefact image can be deconvolved successfully without severe noise amplification.

We performed simulations of the chemical shift artefact caused by the compound 3M™ Fluorinert™ Electronic Liquid FC-84 (C7F16). Its spectrum is shown in Figure 1:

By implementing the signal equation for a FLASH sequence we studied the feasibility of correcting chemical shift artefact images. We investigated simulations with various echo times, readout bandwidths, matrix sizes and manipulations of the spectrum (such as saturating one or more lines). Since all these parameters influence the point spread function and therefore also the quality of the deconvolution, we try to find the best possible imaging parameters by exploring the parameter space.

The dark solid line in Figure 2 shows simulation results of the SNR of corrected chemical shift artefact images for different read out bandwidths. The simulations indicate that the deconvolution amplifies noise stronger than it gains signal by overlapping all existing resonances, since the SNR of the corrected image is always lower than the SNR of the original, uncorrected one. However, these simulations show some distinct minima, indicating that specific values of the readout bandwidth should be avoided. In order to confirm these results of the simulations we have perfomed phantom measurements with four different values of the readout bandwidth (see Figs. 2 and 3). The phantom measurements confirmed the simulation results.

Our simulations basically show that by direct deconvolution one cannot increase the SNR, unless special deconvolution techniques are implemented. However, the simulations also reveal that specific settings of acquisition parameters should be avoided, since they lead to a significant dropdown of the SNR of the corrected image.

Responsible scientists:

• Jochen Leupold
• Marco Reisert
• Dominik von Elverfeldt
• Jürgen Hennig
• Former Member: Mirko Meißner