Retracted in 2025 — Article related to the retraction: https://www.thetransmitter.org/retraction/authors-retract-science-paper-on-controversial-fmri-method
Reference: Toi PT, Jang HJ, Min K, et al. In vivo direct imaging of neuronal activity at high temporospatial resolution. Science. 2022;378(6616):160-168. doi:https://doi.org/10.1126/science.abh4340 (Preprint)
fMRI pioneer and director of the NIH fMRI facility, Peter Bandettini: "It is a paper that did cause a huge stir. I had a bunch of neuroscientists give me a call or just email me saying "what's up with this paper, should I be doing this?". Also, we have been trying to do this in our group". (OHBM Neurosalience podcast - https://www.youtube.com/watch?v=vGhf9utIXwM&t=1208s)
The article had an editorial expression of concern four months ago. The following news article explains that there had been two efforts that failed to replicate the results: https://www.the-scientist.com/news/is-diana-fmri-data-real-71285
The issues may be related to selecting specific experiments/trials for inclusion into the dataset.
One of the authors of the study mentions: "it was better to average over earlier trials when the neuronal response is “fresh” and to take breaks between trials. He also encouraged averaging across animals to minimize baseline fluctuations that he believes come from spontaneous neuroactivity." He adds that he has replicated the technique in both mice and humans and that he expects to publish that work soon.
“This is healthy,” he says. “We need this thinking, rethinking. And eventually—that part I’m optimistic— we’ll find the paths. We’ll really be able to move the field forward.”
Reference: Roth BJ. Can MRI Be Used as a Sensor to Record Neural Activity? Sensors. 2023;23(3):1337-1337. doi: https://doi.org/10.3390/s23031337
Spins are like small magnets or compasses and can be compared to clocks whose hands move in 360 degrees.
The electric activity of the brain creates a changing magnetic field.
That magnetic field will push the spins: it will be as if pushing the hands of clocks and changing their phase in the 360 degree circle.
If we can measure the phase shift, we can calculate the magnetic field. Is the phase shift measurable?
Related excerpts, mainly to illustrate the applicable reasoning:
"If the cardiac magnetic field in the heart produces a 10 nT field lasting 0.5 ms, the phase shift will be about 0.001 radians, or roughly a tenth of a degree."
Manbir Singh "concluded that measurement of action currents would require detecting a phase shift of about a third of a degree".
"Bodurka and Bandettini performed similar experiments and concluded that a magnetic field as small as 0.2 nT lasting for 40 ms could be detected using MRI".
This article also cites the advantages of using low magnetic fields as opposed to high magnetic fields.
Cardiac MRI has long been able to measure spin phase shifts arising from motion, blood flow, and—critically—electromagnetic fields generated by cardiac bioelectric currents. The brain’s neural currents are ~1000× weaker, making direct detection far more challenging. Unlike the brain, where neural currents are extremely weak and spatially heterogeneous, the heart produces large, coherent depolarization waves during each beat. These electrical currents generate magnetic fields strong enough to induce detectable phase perturbations in the MR signal, enabling forms of direct electrophysiological MRI.
This feasibility has been demonstrated in multiple domains:
MR‑electrophysiology has shown that MRI can guide electrophysiological procedures, visualize arrhythmogenic substrates, and track current flow patterns during activation [1,2].
MR‑mapping of cardiac currents has been explored in real‑time MRI–guided electrophysiology, where the magnetic fields associated with cardiac activation influence MR phase and can be used to support ablation planning and substrate characterization [3].
The underlying reason is fundamentally biophysical:
cardiac action potentials involve large, synchronized transmembrane currents, producing magnetic fields several orders of magnitude stronger than those generated by cortical pyramidal neurons. Neural currents in the brain are approximately 10³ times weaker, spatially less coherent, and rapidly attenuated by surrounding tissue [4]. As a result, while cardiac MRI can exploit these electromagnetic effects to detect or map electrical activation, direct magnetic resonance detection of neural currents in the brain remains below the sensitivity threshold of current MR physics.
In short, the heart occupies a unique regime where bioelectric fields are sufficiently strong, coherent, and temporally structured to produce measurable MR phase shifts—making it the only organ where direct MR detection of bioelectric currents has been repeatedly demonstrated in vivo.
References
[1] Weiss, S. et al. Real‑time MRI‑guided electrophysiology: Current status and future perspectives.
[2] Nazarian, S. et al. Magnetic resonance imaging–guided cardiac electrophysiology procedures.
[3] Schaeffter, T. & Razavi, R. Electrophysiology in the MRI environment: Techniques and applications.
[4] Hämäläinen, M. et al. Magnetoencephalography—theory, instrumentation, and applications to noninvasive studies of the working human brain.