Introduction

Aerobic glycolysis, a central metabolic feature of tumour, leads to elevated lactate production, setting lactate as a key marker in breast cancer¹. The non-invasive measurement of lactate in breast tumours using magnetic resonance spectroscopy (MRS) is challenging due to the dominant co-resonance lipid signal. Multiple quantum coherence (MQC), although suppresses lipid signals effectively, loses 50% of the target signal². Hence accurate measurement of lactate depends on optimal acquisition configuration to maximise signal to noise ratio (SNR) with minimal lipid contamination and additional SNR boost using phased-array coils. We therefore set out to systematically examine the role of MQC gradients and sequence timing on SNR and residual contamination signal, with subsequent signal combination algorithms^3-6 evaluated.

Methods

We conducted MQC optimisation in phantoms and evaluated phased-array combination algorithms on lactate spectra acquired from ex vivo breast tumours and patients.

Phantoms and Patients: MQC acquisition was optimised using a 10 mM lactate-embedded lard phantom and a porcine specimen in order to select the optimal MQC gradient. MQC gradients were systematically varied from 1 - 300 ms mT m^-1 (area of the first encoding gradient). Subsequently, TR optimisation was conducted using 10 mM and 25 mM lactate phantoms, a 10 mM lactate-embedded lard phantom, and a porcine specimen. The lactate spectra was acquired using MQC sequence with the optimised encoding gradient and varying TR of 1000 ms, 1250 ms, 1500ms, 1750 ms, 2000 ms and 2250 ms. The optimisation experiments were repeated five times. Sixteen breast tumours specimens (invasive carcinoma, 9 grade II and 7 grade III) freshly excised from patients (mean age, 59 years; age range, 39-78 years), and two female patients (age 49 and 58 years) with invasive carcinoma, grade III were enrolled in the study. The study was approved by the NHS Research Ethics Committee, and written informed consent was obtained from each patient prior to the study.

Data Acquisition: Single voxel MRS data were acquired on a 3 T MRI scanner (Achieva TX, Philips Healthcare, Best, Netherlands) using a 32-phased array coil for phantoms and ex vivo specimens and a 16-phased array breast coil for patients. Lactate spectra were acquired using single voxel MQC PRESS sequence² with repetition time (TR) of 1250 ms, echo time (TE) of 144 ms, spectral editing frequency at 4.1 ppm, 1024 data points. A reference spectrum without water suppression was acquired using single voxel PRESS sequence with TR/TE of 1250/144 ms. For the phantom study, an isotropic voxel of 3 cm was positioned to contain 50% lactate and 50% lard in lard phantom, and 50% muscle and 50% adipose tissue in porcine specimen. The patient study was conducted with the optimised MQC sequenceand lactate and reference spectra were collected from a single voxel snug-fit to the tumour.

Data Processing: All the algorithms were developed in MATLAB (MathWorks, Natick, MA, USA). Lipid suppression efficiency and lactate SNR were evaluated to determine the optimal gradient configuration. Lactate SNR per unit time (SNR/√TR) was evaluated to determine the optimal TR. Lipid suppression efficiency was quantified as the ratio of the area under the spectral peak within the range of [1.0, 1.6] ppm in the reference spectrum against lactate spectrum. SNR was calculated as the lactate spectral peak height divided by the standard deviation of the real part of the spectrum within the range of [-1, -2] ppm⁷. Time domain data from the optimised MQC sequence were combined using the algorithms of adaptively optimised combination (AOC)³, noise decorrelated combination (nd-comb)⁴, whitened singular value decomposition (WSVD)⁵, S/N², S/N, Signal, and Equal Weighting⁶. The amplitude and phase shift were determined from the dominant peak (either water at 4.7 ppm or lipid at 1.3 ppm) in the reference spectrum. The SNR improvement of lactate peak was used to evaluate the combination algorithms and was defined as percentage increase of SNR referenced to the SNR derived from Equal Weighting.

Results

In the lard phantom and porcine specimen, the lipid suppression efficiency reached superior level at MQC gradients greater than 10 ms mT m^-1 (Figure 1). The optimal MQC gradient at 50 ms mT m^-1 provided maximal retention of lactate SNR with complete suppression of lipid signals in both the lard phantom and porcine specimen (Figure 1). There were no significant differences in SNR per unit time in phantoms at different TR with highest SNR per unit time shown at TR of 1250 and 1500 ms (Table 1). In porcine specimen, significantly higher SNR per unit time was found at TR lower than 2000 ms. AOC yielded the highest SNR improvement with a gain of 49 ± 18% in excised tumours (Table 2, Figure 2a). In the two patients, AOC improved the SNR by 44% and 91% respectively following by nd-comb and WSVD (Table 2, Figure. 2b, Figure 3).

Discussion

MQC MRS showed superior suppression of lipid contamination signal with the optimal gradient identified at 50 ms mT m^-1 and optimal TR of 1250 or 1500 ms, to preserve maximal lactate signal. AOC algorithm provided more than 40% SNR improvement in specimens and patients on the optimised MQC configuration.

Conclusion

An optimised MQC MRS coupled with AOC significantly improves lactate measurement in patients with breast cancer.

Acknowledgements

The authors would like to thank Dr Matthew Clemence (Philips Healthcare Clinical Science, UK) for clinical scientist support, Ms Bolanle Brikinns, Ms Louisa Pirie, Ms Linda Lett, and Ms Kate Shaw for patient recruitment support, Ms Dawn Younie for logistic support, Mr Roger Bourne and Ms Mairi Fuller for providing access to the patients as well as Mrs Beverly MacLennan, Mrs Nichola Crouch, Mr Mike Hendry, and Ms Laura Reid for radiographer support.