Spatial interference effects in lactate PRESS spectra

A problem with lactate PRESS spectroscopy (or that of any J-coupled metabolite) is chemical shift displacement. In particular, the PRESS box that corresponds to the CH3 protons of lactate (at 1.33 ppm) is shifted relative to the PRESS box associated with the CH protons (at 4.11 ppm), as illustrated below:

This drawing is meant to show the displacement due to the two refocusing pulses in the sequence. (There is also a smaller displacement perpendicular to the page associated with the excitation pulse, but it doesn’t give rise to signal cancellation.) B is the refocusing pulse bandwidth (specifically the FWHM) in Hz, and S is the chemical shift difference between the CH and CH3 protons in Hz. S increases linearly with field strength, which makes for a challenge at 3T and beyond.

The crux of the problem is this: the evolution of the CH3 spins depends on the pulses experienced by the CH spins to which they are J-coupled. Yablonskiy et al. (Magn. Reson. Med. 39, 169-178, 1998) have described this dependence in terms of a very simple four-compartment model for the CH3 PRESS box:

1.  In volume V1, the PRESS box intersection, the CH protons experience both of the refocusing pulses. As a result, the corresponding CH3 doublet inverts at TE 144, TE 432, etc.

2.  In volume V2, the CH protons experience the first refocusing pulse but not the second one.

3.  In volume V3, the CH protons experience the second refocusing pulse but not the first one.

4.  In volume V4, CH protons are untouched by refocusing pulses. As such, the CH3 protons contribute to (positive) spin echoes for any value of TE – just like uncoupled spins.

Yablonskiy’s paper gives general expressions for the signal sk± from each of the volumes:

s1± = r V1 exp[±i p J TE]

s2± = r V2 exp[±i p J TE1]

s3± = r V3 exp[±i p J (TE1 – TE)]

s4± = r V4

where “±” refers to the two doublet components, r is the spin density, and TE1 is the first echo time in the PRESS sequence.

For TE = 1/J = 144 ms, these expressions become

s1± = -r V1

s2± = r V2 exp[±i p J TE1]

s3± = -r V3 exp[±i p J TE1]

s4± = r V4

Note that s2± and s3± exactly cancel and that s1± and s4± have opposite sign. With respect to the total available signal stot± = r V = r (V1 + V2 + V3 + V4),

s1± / stot± = V1 / V = -(B – S)2 / B2

s4± / stot± = V4 / V = S2 / B2

and their sum sact± = -[1 – (2S/B)]

(S/B à 0 gives rise to a perfectly inverted doublet.)

To assess deviations from perfection, the Excel spreadsheet that accompanies this note contains a plethora of B values for different coils, field strengths, and pulse choices (current “normal” and “sharp” options, plus the long-TE and short-TE choices from the old PRESS sequence). Even with the new pulses, the 3T lactate signal at TE 144 is reduced: only ~59-62% of maximum inversion using the T/R head coil, and only ~40-44% of maximum inversion using the body coil. To improve the numbers, I examined two other possibilities:

1.  Define a new “extreme” pulse mode that uses the fm_ref07 pulse. [Despite its ungainly nature and long duration (16.8 ms for B1max = 13.5 mT), the pulse does work – Peter Barker and Richard Edden at Johns Hopkins have used it for 2D CSI.] With this extreme pulse in place, the 3T TE 144 signal level rises to 78% of maximum inversion with the T/R head coil and 67% of maximum inversion with the body coil.

2.  Do as Greg Metzger and others have done in the prostate: over-prescribe the PRESS box and use very high bandwidth power-level-3 REST slabs (or slightly crisper fm_sat01 pulses) with a negative gap to define the effective boundaries of the box. Assuming perfect saturation of unwanted spins, the 3T TE 144 signal level soars to 92% of maximum inversion with the T/R head coil and 89% of maximum inversion with the body coil.

Two additional results of interest from the Yablonskiy paper …

(1) At TE = 2/J = 288 ms, the signal expressions become

s1± = r V1

s2± = r V2 exp[±i p J TE1]

s3± = r V3 exp[±i p J TE1]

s4± = r V4

and sact± / stot± = 1 - 2 (S/B)(1 – S/B)[ 1 – cos(p J TE1) ± i sin(p J TE1) ]

= 1 - 2 (S/B)(1 – S/B)[ 1 – cos(2p TE1 / TE) ± i sin(2p TE1 / TE) ]

Signal loss due to spatial interference thus occurs at TE 288 as well, even though the contributions from volumes V1 and V4 are in phase. The loss is minimized, however, if the first echo time is made as small as possible.

(2) For TE = 4/J = 576 ms and TE1 = TE/2 = 288 ms (i.e., a symmetric echo pattern), there is NO spatial interference from any of the four volumes and sact± = stot±.

Jim Murdoch

Philips Medical Systems, Cleveland OH

December 9, 2005