Rapid high-resolution sodium relaxometry in human breast

Glen Morrell¹, Josh Kaggie², Matthew Stein¹, Scott Parker¹, and Neal Bangerter³

¹Radiology, University of Utah, Salt Lake City, UT, United States, ²Radiology, University of Cambridge, Cambridge, United Kingdom, ³Electrical and Computer Engineering, Brigham Young University, Provo, UT, United States

Synopsis

We have performed rapid high-resolution breast sodium MRI relaxometry using a custom sodium breast phased array coil. Clear delineation of short- and long-T2* components of the sodium signal is possible with a spatial resolution of 3.75 x 3.75 x 4mm over the entire breast with a total imaging time of under 10 min. This method will allow the investigation of the potential of sodium relaxometry to improve the specificity of breast MRI for the detection of breast cancer.

Purpose

To investigate the T2* of sodium in human breast tissue with a rapid high-resolution sodium MR relaxometry protocol.

Introduction

MRI is the most sensitive imaging modality for the detection of breast cancer, but its intermediate specificity limits its use as a screening examination in the general population. Sodium breast MRI has the potential to improve the specificity of breast MRI. Elevated sodium MRI signal has been demonstrated in malignant breast lesions compared to benign lesions (1). Sodium signal demonstrates a biexponential decay with long and short T2* components. The elevated sodium signal seen in malignant lesions may represent a shift in the population of sodium ions into a longer T2* population, possibly related to association with cell membranes. Thus measurement of the short and long T2* components of the sodium signal may help discriminate between benign and malignant breast lesions. Sodium relaxometry of human breast has not been previously performed due to constraints on SNR and imaging time. Our recently developed sodium phased-array breast coil (2) along with rapid k-space acquisition techniques allows unprecedented SNR for sodium breast MRI which we have used to perform 3D sodium relaxometry of the whole breast at a resolution of 3.75 x 3.75 x 4.0 mm in under ten minutes.

Methods

In this IRB-approved HIPAA-compliant study, five subjects were studied after giving written informed consent. Each subject underwent a sodium MRI relaxometry examination on a 3.0T Siemens Trio system which consisted of 5-7 complete 3D sodium images of the breast performed at varying echo times, each requiring under 2 minutes. The typical protocol included echo times of 0.27ms, 0.45ms, 1.1 ms, 4 ms, and 14 ms with total imaging time less than 10 min. A few of the subjects were imaged with up to two additional echo times within this range. T2* estimation was performed on a set of valid image pixels selected by a magnitude threshold applied to the 0.27ms echo time image. Both mono-exponential and bi-exponential T2* fitting was performed using Matlab.

Results/Discussion

Figure 1 shows a typical sagittal slice from a full 3D single-breast examination. A signal intensity curve from a typical single point in this slice is shown with the corresponding mono- and bi-exponential curve fits. Similar curve fitting was performed for every pixel in the 3D sodium data set for each of the five subjects. Mean T2* values across the entire breast and standard deviation for the T2* estimates are given in the table in Figure 2 for both mono- and bi-exponential curve fits for each of the five subjects. Mono-exponential curve fitting yields average breast sodium T2* of 14.22 ms. Bi-exponential T2* fitting yields an average short T2* component of 4.15ms and average long T2* component of 116.3ms. The histogram in Figure 3 shows that the long T2* component has a mode at about 17ms. The much longer average T2* of 116.3ms given in the table of Figure 2 is caused by a long tail in the distribution of long T2* values, likely secondary to the relatively short longest echo time of 14ms, which leads to some inaccuracy in measurement of the long T2* component. In all five subjects, bi-exponential T2* fitting yields much lower residual mean squared error than mono-exponential fitting, supporting the observation that sodium T2* decay in the breast is bi-exponential. This is illustrated for a single slice of one study in Figure 4. Figure 5 shows that the relative magnitude of the signal from the short and long T2* components shows some variability. Differences in relative signal from short and long T2* components could become a useful discriminating feature of benign vs. malignant lesions. Further study with subjects with breast cancer is necessary to investigate this possibility.

Conclusion

We have implemented sodium T2* relaxometry in human breast with a spatial resolution of 3.75 x 3.75 x 4mm covering the entire breast in a total image time of under 10 min. Our results in five subjects without malignancy show a definite bi-exponential decay of the sodium signal with robustly identifiable short- and long-T2* components of about 4 ms and 17 ms respectively. Future work will investigate the utility of relaxometry in the discrimination of benign from malignant breast lesions.

Acknowledgements

This work was supported by NIH grant 5 K08 CA112449 and by a seed grant from the University of Utah Vice President's Office.

References

1. Ouwerkerk R, Jacobs MA, Macura KJ, Wolff AC, Stearns V, Mezban SD, Khouri NF, Bluemke DA, Bottomley PA. Elevated tissue sodium concentration in malignant breast lesions detected with non-invasive 23Na MRI. Breast Cancer Res Treat 2007;106(2):151-160.

2. Kaggie JD, Hadley JR, Badal J, Campbell JR, Park DJ, Parker DL, Morrell G, Newbould RD, Wood AF, Bangerter NK. A 3 T sodium and proton composite array breast coil. Magn Reson Med 2014;71(6):2231-2242.

Figures

Figure 1: Example of sodium T2* fit with mono- and bi-exponential decay. Single sagittal slice from a 3D full-breast data set. T2* fit for mono-exponential and bi-exponential decay at a selected pixel (black circle) shows almost exact fit to the data (red circles) with the bi-exponential fit. Over the entire breast, the bi-exponential fit results in about 1/3 the mean squared residual error as a mono-exponential fit, showing that robust bi-exponential T2* estimation is possible.

Figure 2: Table showing whole-breast averaged values of sodium T2* from mono-exponential fit and short and long T2* values for bi-exponential fit. Results are similar across the five subjects. Mean squared residual error of bi-exponential fit is much lower than for mono-exponential fit.

Figure 3: Whole-breast sodium T2* histogram from Subject 1. Histogram of T2* based on mono-exponential fit (blue) and short and long T2* components from bi-exponential fit (red and green) are superimposed. A long tail of higher T2* values for the long T2* component is not fully included. Histograms of the other four subjects are similar.

Figure 4: Squared residual error of sodium T2* fit in one sagittal slice of a 3D full-breast data set in Subject 1. Units are arbitrary but the same for both error maps. Mean squared error of T2* fit is much lower for bi-exponential T2* fit.

Figure 5:Short and long sodium T2* components in a single sagittal slice of a 3D full-breast data set in Subject 1. The magnitude of the short- and long-T2* components are shown in the top panels. Some difference in the distribution of short- and long-T2* components is evident, which may form a basis for discrimination of tissue types (e.g. benign vs. malignant). T2* values for the short and long components are shown in the bottom images.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)

0404