AC magnetic sensing using the SIRS effect combined with bSSFP at ultra-low field
Bragi Sveinsson1,2,3, Neha Koonjoo1,2,3, Bo Zhu1,2,3, Thomas Witzel1,2, and Matthew Rosen1,2,3

1Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 2Harvard Medical School, Boston, MA, United States, 3Physics, Harvard University, Cambridge, MA, United States


Direct detection of neuronal currents has long been a goal within MRI, with the aim of improving upon the spatial and temporal resolution of BOLD fMRI. So far, good results have been shown in phantoms but detection in vivo has proven difficult. A promising current detection technique is Stimulus-Induced Rotary Saturation (SIRS), but the BOLD signal can contaminate SIRS measurements, possibly explaining inconclusive in vivo results so far. A new sequence was developed and tested in an ultra-low-field (ULF) regime (6.5 mT) where paramagnetic effects such as BOLD are reduced and is more suited for SIRS measurements in vivo.


To detect AC magnetic fields in an ultra-low-field (ULF) MRI system by creating a balanced Steady-State Free Precession (bSSFP) imaging technique that uses Stimulus-Induced Rotary Saturation (SIRS).


The SIRS sequence (Fig. 1) is a powerful method for measuring low-frequency magnetic stimuli with MRI1,2, but physiological effects such as BOLD make in vivo validation challenging. With the goal of investigating SIRS imaging at a low magnetic field with no BOLD contamination, a ULF 6.5 mT scanner with a maximum gradient amplitude of 1 mT/m was used. With these parameters, a normal SIRS scan would be slow and the small SIRS contrast hard to detect. The SNR efficiency of bSSFP results in a strong signal in reasonable scan times, particularly due to the near unity value of T2/T1 at such low magnetic fields, making it well suited for the ULF scanner3. Therefore, a new sequence, combining bSSFP with SIRS, was developed as shown in Fig. 2. The magnetization is flipped to the y-axis and a spin-lock pulse is applied, sensitizing the magnetization to the frequency γBSL. The magnetization will rotate by α=γBstimTSL away from the spin-lock axis in the doubly rotating frame, while also precessing by φ=γBSLTSL around BSL, thus lying on the cone shown in Fig. 2c. The magnetization is then rotated back to the normal rotating frame to enable signal readout and relaxation, before being flipped back in the next TR. Effectively, this will be a bSSFP sequence with a flip angle of α. While this angle will be small, the steady-state signal can be made large or small, since the bSSFP signal as a function of off-resonance or RF phase cycling gives high spikes spaced by 2π4.

Bloch simulations of the sequence were performed in MATLAB (Fig. 3). The parameters were T1=630ms, T2=625ms, Bstim=5nT, fstim=40Hz, γBSL=40Hz, TSL=25ms, TR= 50ms. The parameters fstim, TR–TSL, TSL, and Bstim were then varied. The process was then repeated for T1=127ms and T2=76ms, the measured values for white matter at 6.5mT.

Then, a wire loop with a 4cm diameter and 10 turns was inserted into a phantom filled with CuSO4 solution (Fig. 4). A sinusoidal current was generated and passed through the loop. Experiments were performed with T1= 630ms, T2=625ms, Bstim=17nT, γBSL=100Hz, TSL=41ms, TR=55ms, and fstim= [90,95,97,100,103,105,110] Hz with a 32×32×19 imaging matrix, FOV= 28cm×21cm×20cm and 120 signal averages with a scan time of 33 min.


The simulations showed the magnetization tracing a cone along the SL axis until settling on a circle in the steady-state. Varying the scan parameters showed quite similar behavior to bSSFP, such as a spike-like signal as a function of TR-TSL at small angles. This is because TR-TSL determines the relative phase difference between the stimulus wave and the magnetization between SL pulses, much like RF phase cycling in normal bSSFP. The simulations showed a roughly sinc shaped frequency response, as expected from the (rectangular) SL pulse. Since the linearly polarized stimulus field will have equal right-hand and left-hand circularly polarized components, it has a positive and negative frequency component in the doubly-rotating frame, whose responses add up.

The phantom scans (Fig. 5) revealed a steady-state signal even without stimulus. This is likely due to off-resonance effects and was found to be minimized when the RF pulse was applied about 25 Hz off resonance. When the stimulus wave was turned on, it dominated this background signal and resulted in a net signal change of about 40%.


Our observed signal change of 40% is quite large for a 17nT stimulus field, and carries promise of detecting much weaker stimuli in this system. This could eventually lead to the detection of fields from currents in neuronal bundles in the brain, expected to be around 0.1-1 nT5. The dark signal likely means we are between spikes in Fig. 3c. Future experiments will investigate other points along this curve.

The method described here not only allows SIRS imaging at our 6.5mT magnet, but also leverages several advantages of that system in addition to negligible BOLD contamination. Due to the operation at ULF, the repeated SL pulses throughout the scan do not raise SAR concerns. Furthermore, the scanner is very homogenous in absolute (Hz) terms, so achieving the desired point on the bSSFP curve (Fig. 3c) is more feasible than on standard clinical scanners.


A new bSSFP sequence using SIRS was developed, simulated, and used in phantom scans at a 6.5mT scanner. The simulations showed similar behavior to bSSFP and the phantom scans showed a 40% signal change with a stimulus field of 17nT.


DARPA 2016D006054


1: Witzel et al. Neuroimage 2008;42:1357-1365. 2: Jiang et al. MRM 2016;75:519–526. 3: Sarracanie et al. Nature Scientific Reports 2015;5:15177. 4: Carr. Phys Rev 1958;112:1693–1701. 5: Halpern-Manners et al. PNAS 2010;107:8519-8524.


a) A standard SIRS sequence. A spin-lock (SL) preparation module rotates the magnetization vector in the doubly rotating frame, reducing its longitudinal component. A spoiler gradient then spoils any transverse magnetization. The magnetization is then quickly read with an EPI readout. b) A normal bSSFP image of the phantom used. A loop for inducing magnetic stimuli is visible, but no current is applied. c) A 3D 32×32×19 SIRS sequence with a GRE readout. A strong stimulus current with Bstim= 1μT was applied before the excitation and readout of each k-space line. This is a strong stimulus, for illustrative purposes only.

A combination of bSSFP and SIRS. a)The magnetization (M) starts fully relaxed. b)The magnetization gets flipped to the transverse plane. c)A SL pulse sensitizes M to an external stimulus magnetic field, with amplitude Bstim and frequency ωstim=γBSL. The stimulus tilts the magnetization by α=γBstimTSL. d)M is flipped back from the transverse plane, is read out and experiences relaxation. e)M is then flipped down again in the next TR. This is effectively a bSSFP sequence with a flip angle α induced in the doubly rotating frame. In regions without stimulus, the bSSFP flip angle is zero creating no steady-state signal.

a)Bloch simulations showing how the magnetization gets flipped between the longitudinal axis and the SL axis, forming a cone during the transient phase. b)The frequency response of the sequence, a sum of two sincs. c)The signal over a range of TR-TSL (the non-SL portion of TR). The spikes occur when the timing makes the stimulus waveform continue to be aligned with the magnetization. This is similar to the behavior of standard bSSFP at small angles, on resonance, with no RF phase cycling. d)-e)Signal as a function of TSL and Bstim. f)-j)Same simulations but for white matter parameters at ULF field.

a) A 10-turn circular wire loop with a diameter of 4 cm, for carrying stimulus current. b) The loop in panel a was inserted in a 3D printed coil filled with a CuSO4 solution. An image of the loop inside the coil can be seen in Fig. 1b.

a)The phantom in Fig. 1b, imaged with the presented method without a stimulus. b)The same acquisition as in a) but with a stimulus of Bstim= 17nT applied at fstim= 100Hz. The SL pulse was sensitized to a frequency of 100Hz. c)A GRE SIRS scan like the one in Fig. 1c with the 17nT stimulus. Clearly, very little contrast is generated. d)Signal change as a function of stimulus frequency. The results show a roughly 40% signal change around 100Hz, which is quite pronounced. For illustrative purposes, a)-c) were acquired with double the scan time as the curve points to improve SNR.

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)