R2* estimation with CSI: A Pilot and Feasibility Study
Eamon K Doyle, MS1, Jonathan M Chia, MS2, Krishna S Nayak, PhD3,4, and John C Wood, MD, PhD4,5

1Biomedical Engineering, University of Southern California, Sierra Madre, CA, United States, 2Philips Healthcare, Cleveland, OH, United States, 3Electrical Engineering, University of Southern California, Los Angeles, CA, United States, 4Biomedical Engineering, University of Southern California, Los Angeles, CA, United States, 5Cardiology, Children's Hospital of Los Angeles, Los Angeles, CA, United States


R2 (1/T2) and R2* (1/T2*) are useful metrics related to tissue iron loads. Clinical iron estimation at 3T is limited by the achievable echo times in Cartesian spin echo and gradient echo imaging. We present a modified CSI spectroscopy sequence robust to motion for high SNR liver R2* estimation with the potential for simultaneous R2 estimation.


In clinical tissue iron estimation, R2* (1/T2*) provides important information about iron load.[1] However, the dynamic range of iron quantitation is limited by the duration of the readout. The effective dynamic range of iron concentrations is up to 40 mg/g dry weight at 1.5T, and up to 20 mg/g at 3T. Spectroscopic sequences such as CSI may overcome these limitations by eliminating frequency encoding and allowing complete sampling of the R2* envelope from a spin echo. We propose a modified CSI sequence that provides high SNR data for R2* measurement. The proposed protocol uses 3 free-breathing scans, which improves patient comfort, decreases scan time, and increases dynamic range relative to standard gradient echo-based R2* imaging.


A CSI spectroscopy sequence with phase encoding in 2 directions with no readout gradient (Figure 1) was modified to use an excitation pulse based on a Dolph-Chebyshev window function to reduce echo time while maintaining slice selectivity. Scanning was performed in iron-loaded human research subjects as part of an IRB-approved study. The following parameters were used in all acquisitions: voxel = 20x20x15mm, 1024 samples, 32/2kHz sampling/spectral BW, 90° slice-selective excitation, 180° non-selective inversion, matrix=8x7, and TE/TR=[3.1,7.0,13]/1000ms. Reference liver iron levels were determined from clinically indicated MRI examinations for iron overload at 1.5 Tesla. Studies were performed on a 1.5T Achieva (SW R3.2.2, Philips) with a 16 element SENSE Torso XL coil. A 3-slice, single gradient echo scan was acquired with the following parameters: 16 echoes linearly spaced echoes with TE/TR=0.96-11.47/50ms, FA=30°, matrix=72x67, voxel=1.25x1.25x10mm, BW=4409Hz/px.

Voxelwise R2* was estimated from CSI data by fitting free induction decays (FID) with a mono-exponential plus constant model with nonnegativity constraints on S0 (signal intensity at t=0), R2*, and noise bias using nonnegative least squares in MATLAB (Mathworks, Natick, MA). R2* values for the clinical images was estimated using a truncated exponential model. CSI-based R2* estimates were compared to clinical R2* estimates to assess accuracy.


R2* estimates did not vary systematically across the three spin echoes, allowing them to be averaged to increase SNR. Averaged CSI R2* estimates were plotted against reference 1.5T results in four patients (Figure 2). The solid line represents the 3T-1.5T calibration determined empirically and by Monte Carlo modeling [2]. Overall agreement was within expected measurement error but we will need to study patients with even greater iron burden to evaluate the full dynamic range of the CSI sequence. Figure 3 contains a representative example of an acquired FID and resulting fit output. All reviewed FIDs reached the noise floor in within the first 10 ms.


These pilot data indicate that a modified CSI sequence may provide robust R2* estimates in heavily iron loaded tissues. These studies were performed with the patient breathing freely; however, breath-hold imaging could be achieved using SENSE factors of 2 in two directions.

The CSI-R2* measurement in the patient with the highest iron burden was slightly below the predicted agreement with 1.5T. Further, R2* estimates did not vary systematically with spin-echo time. We anticipated some prolongation of R2* with increase echo time because of a length-bias effect i.e. proton species with shorter R2 would have greater weighting with longer echo time; a similar effect in CPGM-estimated R2 has previous been described theoretically[3] and shown in simulation.[4] Both the underestimation of high iron loads and the lack of apparent R2 weighting may have been caused by a failure to acquire the fasted decay signals with high off-resonance. We believe the post-acquisition filter bandwidth of 2000 Hz was insufficient to characterize the rapid decay at high iron burden in spite of sufficient sampling rate. Performing the spin echo CSI sequence at several spin echo times may allow us to simultaneously characterize R2 and R2* and potentially probe cellular iron distribution characteristics.

In subsequent studies, we hope to achieve robust R2 estimates by adding a longer echo time to better capture signal decay in patients with low-to-moderate iron overload. Further, we will increase readout bandwidth to increase the dynamic range of R2* sensitivity. These changes will increase the dynamic range of the sequence to measure R2* values over 4000 Hz, or 50 mg FE/g dry tissue, the upper limit of clinical liver iron loads. Implementation on a 1.5T magnet would reach iron loads up to 100 mg/g. Additional patient and phantom studies are ongoing as part of an NIH-funded research study.


R2* estimation using a modified CSI sequence with a shortened TE shows promise as a robust, fast, comfortable method of performing spatially localized iron load assessment.


This work is supported by the National Institute of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Grant R01-DK097115. Clinical science and research support is provided by Philips Healthcare in kind.


[1] J. C. Wood, P. Zhang, H. Rienhoff, W. Abi-Saab, and E. J. Neufeld, “Liver MRI is more precise than liver biopsy for assessing total body iron balance: a comparison of MRI relaxometry with simulated liver biopsy results,” Magn. Reson. Imaging, vol. 33, no. 6, pp. 761–767, Jul. 2015.

[2] N. R. Ghugre, E. K. Doyle, P. Storey, and J. C. Wood, “Relaxivity-iron calibration in hepatic iron overload: Predictions of a Monte Carlo model,” Magn. Reson. Med., vol. 74, no. 3, pp. 879–883, Sep. 2014.

[3] J. H. Jensen and R. Chandra, “Theory of nonexponential NMR signal decay in liver with iron overload or superparamagnetic iron oxide particles,” Magn. Reson. Med., vol. 47, no. 6, pp. 1131–1138, Jun. 2002.

[4] N. R. Ghugre, “Calibration of iron-mediated MRI relaxation by Monte Carlo modeling,” Ph.D. Thesis, University of Southern California, Los Angeles, 2008.


Figure 1 – Pulse sequence diagram demonstrating the modified CSI sequence. Excitation occurs in 0.5 ms, allowing the echo time to be reduced to 3.1 ms. Further sequence optimizations and reduced excitation angles will further reduce the TE in future revisions.

Figure 2 – Patient results demonstrate good agreement with the theoretical 3T-1.5T R2* calibration for iron loaded tissues.

Figure 3 – Example of acquired FID and exponential fit. FID truncated to 50 samples.

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