MRI sees exactly where the ultrasound is working, in real time, while it works
MRgFUS begins entirely inside the bore. The patient lies on an MR-compatible table with an integrated ultrasound transducer built into the tabletop or helmet, and the same magnet that will later confirm the treatment outcome first acquires the high-resolution anatomical images used to plan it. No incision, no separate imaging suite, no registration error between "where we planned" and "where we treat" — planning and delivery share one coordinate system from the first image onward.
A conventional image-guided procedure typically registers a pre-operative scan (CT or MRI) to the patient's position in a separate treatment room — a process that introduces registration error every time the patient moves, breathes, or shifts on the table. MRgFUS eliminates this problem structurally: the diagnostic-quality MRI used for planning is acquired on the exact same magnet, with the patient in the exact same position, that will later steer and confirm the ultrasound treatment.
High-resolution T1- and T2-weighted sequences delineate the target with sub-millimeter precision — the ventral intermediate (Vim) nucleus of the thalamus for essential tremor, a fibroid's boundary and vascular supply for uterine fibroids, or a prostate lobe for focal ablation. The operator overlays the treatment trajectory directly onto these images.
Because the ultrasound beam converges through tissue from outside the body, the planning step must ensure that every one of the transducer's hundreds of individual elements has a clear acoustic path to the focal point — avoiding bone (which absorbs and scatters ultrasound energy unpredictably), major vessels (which can carry heat away or risk hemorrhage), and structures where even minor off-target heating would be unacceptable, such as the internal capsule or optic tract near a thalamic target.
For transcranial treatments, a low-dose CT is fused with the planning MRI specifically to model skull density and thickness, since the skull is the single largest source of aberration and near-field heating in the entire treatment chain — the phased-array transducer uses this map to individually delay and correct each element's pulse before treatment begins.
ExAblate Neuro received FDA approval for unilateral thalamotomy in essential tremor in 2016, and ExAblate for uterine fibroid ablation in 2004 — both built on the same principle: plan, treat, and verify without ever moving the patient out of the magnet.
Before any therapeutic-level energy reaches tissue, MRgFUS systems fire a deliberately weak "test sonication" — enough energy to produce a small, measurable, fully reversible temperature rise, but far below the threshold for any tissue damage. This single low-risk pulse answers the most important question in the entire procedure: is the focal spot exactly where the plan says it should be?
Even with perfect planning images, the acoustic focus can land slightly off the intended anatomical target — tissue heterogeneity, small patient motion, or (for transcranial work) imperfect skull-density correction can all shift the true focal spot by a millimeter or more. Firing directly at therapeutic power without checking this would risk ablating the wrong tissue.
Instead, the operator fires a single low-power sonication — enough to produce a small, transient, fully reversible temperature rise of roughly 3–6°C — and lets MR thermometry directly visualize where that heat actually appeared. Because the heating is sub-ablative, no tissue damage occurs regardless of whether the spot needs correction.
The thermometry map from the test pulse is compared, pixel by pixel, against the planned target coordinates. Any offset — typically well under a millimeter with modern phased-array correction, but occasionally a few millimeters for challenging transcranial geometries — is fed back into the beam-steering electronics, which electronically re-aim the focus without physically moving the transducer or the patient.
This alignment-verify-correct cycle can be repeated as many times as needed, at essentially no cost to the patient beyond a few minutes, until the test sonication lands precisely on target. Only then does the procedure move to full-power therapeutic sonications.
Because the transducer contains hundreds of individually phased elements, "aiming" the beam is an electronic correction — recalculating time delays across the array — rather than a mechanical repositioning, so re-verification between sonications costs seconds, not minutes.
The defining capability of MRgFUS is that the same scanner delivering images can also measure temperature, in real tissue, in real time, during treatment. Proton resonance frequency (PRF) shift thermometry turns the MRI signal itself into a thermometer, generating a color-coded temperature map that updates every 1–3 seconds and is overlaid directly on the anatomical image the operator is already watching.
PRF thermometry exploits a subtle but highly reliable physical effect: as tissue water heats up, the hydrogen bonding network around water protons weakens slightly, changing the local electron shielding around each proton and shifting its resonance frequency by a small, remarkably temperature-independent-of-tissue-type coefficient — roughly −0.01 ppm per °C.
By acquiring a baseline phase image before heating and comparing it, voxel by voxel, to phase images acquired during sonication, the scanner converts a tiny phase shift into a temperature-change map. Because the coefficient is nearly constant across most soft tissues (fat is a notable exception), this method works without needing tissue-specific calibration — a major reason it became the clinical standard for thermal therapy monitoring.
A fresh temperature map every 1–3 seconds means the operator is effectively watching the lesion form in real time, sonication by sonication, rather than inferring it indirectly. The color overlay — typically a blue-to-yellow-to-red gradient — is displayed directly on the anatomical image, so the operator sees both where the anatomy is and how hot it is getting at the same moment.
This feeds directly into thermal dose calculations. Tissue damage is not just a function of peak temperature but of temperature sustained over time, commonly expressed as cumulative equivalent minutes at 43°C (CEM43). Real-time thermometry lets the system integrate this dose live and stop a sonication automatically the instant the target dose is reached — something impossible with any imaging modality that only shows results after the fact.
Because PRF thermometry is essentially insensitive to tissue type (except fat), a single calibration works across brain, uterine, and prostate tissue alike — one of the key reasons real-time MR thermometry became the universal safety layer across MRgFUS applications rather than a technique specific to one organ.
Open-loop thermal therapy fires a pre-planned dose and hopes the anatomy holds still and behaves as modeled. MRgFUS instead closes the loop: every sonication's thermometry result directly informs the parameters of the next one. If the focal spot has drifted, or tissue is heating faster or slower than expected, power, position, and duration are all adjustable in real time — before the next pulse, not after the treatment is finished.
Between every sonication, the control software (and the operator) compares the actual thermometry result to the predicted one along three axes:
• Spatial drift — has the hot spot appeared a millimeter or more from the intended coordinates, due to patient motion, respiratory shift, or tissue property changes from prior heating? • Rate of heating — is temperature rising faster than modeled (risking overshoot and unintended margin) or slower (risking an underdosed, ineffective sonication)? • Near-field heating — is energy accumulating in tissue the beam passes through on the way to the focus, such as the scalp or skull, faster than at the target itself?
Any of these findings can trigger an adjustment before the next pulse: re-steering the electronic focus, reducing or increasing acoustic power, shortening or lengthening sonication duration, or inserting a cooling interval to let near-field tissue dissipate heat.
Every other image-guided ablative technology — stereotactic radiosurgery, radiofrequency ablation guided by pre-procedure imaging, even earlier generations of ultrasound therapy — delivers a fixed, pre-planned dose and confirms the result only afterward, when nothing can be corrected mid-treatment. MRgFUS instead treats each sonication as a small, low-risk experiment: start conservatively, measure the true tissue response, then titrate.
This stepwise, verified escalation is the reason MRgFUS thalamotomy protocols typically begin several sonications at deliberately sub-therapeutic power — confirming focal accuracy and thermal response — before ramping to the ablative dose that produces the permanent lesion, all within a single continuous session.
A typical essential tremor treatment involves 10–20 sonications, with early pulses used purely for alignment and dose-response confirmation and only the final several delivered at full ablative power — each one informed by the thermometry from the sonication before it.
The final advantage of performing an entire procedure inside one imaging system is the ending: the moment the last sonication completes, the same MRI that planned and monitored the treatment immediately confirms its result — lesion size, location, and margins — while the patient is still on the table. There is no separate follow-up scan, no days of uncertainty, and no need to bring the patient back for a different imaging study to learn whether the treatment worked.
Immediately after the final therapeutic sonication, standard T2-weighted and contrast-enhanced MRI sequences are acquired on the same table, in the same position, within the same magnet. These sequences show the thermal lesion directly — its size, its location relative to the planned target, and critically, its margins relative to nearby structures that needed to be avoided.
Because this confirmation happens in the same coordinate frame as planning and treatment, there is no registration uncertainty between "where we think the lesion is" and "where it actually is" — the same problem that made pre-MRgFUS image-guided therapies dependent on separate, delayed follow-up imaging to know whether treatment succeeded.
The pivotal trial for MRgFUS thalamotomy in essential tremor (Elias et al., NEJM 2016) demonstrated roughly 50–60% improvement in hand tremor scores, sustained at multi-year follow-up, with the most common adverse effects — gait unsteadiness and paresthesia — being transient in the majority of patients. Uterine fibroid ablation and prostate tissue ablation followed similar paths: immediate, same-session confirmation of treatment extent, entirely without incision or ionizing radiation.
Beyond thermal ablation, the same real-time MRI-guidance principle is now being extended to non-thermal applications: mechanical histotripsy, sonodynamic therapy (activating a systemically administered sensitizer only within the focal volume), transient blood-brain barrier opening for targeted drug delivery, and neuromodulation — all inheriting the same core advantage of exact, real-time, verified targeting.
Unlike radiofrequency ablation or stereotactic radiosurgery, which confirm treatment effect with delayed, separately-scheduled imaging days to weeks later, MRgFUS confirms the lesion in the same session, on the same table — turning "did it work?" from a days-long wait into an immediate answer.