Clinical Use of Bedside Portable Low-field Brain Magnetic Resonance Imaging in Patients on ECMO: The Results from Multicenter SAFE MRI ECMO Study

Purpose: Early detection of acute brain injury (ABI) is critical for improving survival for patients with extracorporeal membrane oxygenation (ECMO) support. We aimed to evaluate the safety of ultra-low-field portable MRI (ULF-pMRI) and the frequency and types of ABI observed during ECMO support. Methods: We conducted a multicenter prospective observational study (NCT05469139) at two academic tertiary centers (August 2022-November 2023). Primary outcomes were safety and validation of ULF-pMRI in ECMO, defined as exam completion without adverse events (AEs); secondary outcomes were ABI frequency and type. Results: ULF-pMRI was performed in 50 patients with 34 (68%) on venoarterial (VA)-ECMO (11 central; 23 peripheral) and 16 (32%) with venovenous (VV)-ECMO (9 single lumen; 7 double lumen). All patients were imaged successfully with ULF-pMRI, demonstrating discernible intracranial pathologies with good quality. AEs occurred in 3 (6%) patients (2 minor; 1 serious) without causing significant clinical issues. ABI was observed in ULF-pMRI scans for 22 patients (44%): ischemic stroke (36%), intracranial hemorrhage (6%), and hypoxic-ischemic brain injury (4%). Of 18 patients with both ULF-pMRI and head CT (HCT) within 24 hours, ABI was observed in 9 patients with 10 events: 8 ischemic (8 observed on ULF-oMRI, 4 on HCT) and 2 hemorrhagic (1 observed on ULF-pMRI, 2 on HCT). Conclusions: ULF-pMRI was shown to be safe and valid in ECMO patients across different ECMO cannulation strategies. The incidence of ABI was high, and ULF-pMRI may more sensitive to ischemic ABI than HCT. ULF-pMRI may benefit both clinical care and future studies of ECMO-associated ABI.


Introduction
The use of extracorporeal membrane oxygenation (ECMO) has dramatically increased in the past decade for patients with refractory cardiopulmonary failure. 1 However, ECMO conveys an elevated risk of acute brain injury (ABI), such as ischemic stroke, intracranial hemorrhage (ICH), and hypoxic-ischemic brain injury. 2Early diagnosis of ABI appears important as it can guide clinicians regarding timely cessation and judicious resumption of anticoagulation therapy, mitigating the dire consequences of ECMO-associated ABI which is known to increase mortality by a factor of 2-3. 3,4urrently, the timing and the true prevalence of ABI are unknown because assessing neurological status in ECMO patients is challenging and often delayed owing to the patients' critical illness complicated by the routine use of sedation, paralytics, and the lack of standardized neurological examination and assessment.Even if standardized neurological monitoring improves the detection of ABI, 2,5 their persistent cardiopulmonary instability makes the timely diagnosis and management of ABI in these patients di cult. 5More than 30% of patients do not receive head CT (HCT) scans during ECMO support due to an inability to safely transport patients and the lack of readily available trained transport personnel. 5Even where possible, head CT has limited sensitivity for detecting hyperacute to acute ischemic ABI. 6The gold standard for diagnosing ABI, conventional magnetic resonance imaging (MRI), relies on strong magnetic elds (1.5-3T) which are incompatible with extracorporeal life support circuits and equipment due to safety concerns (heating, migration, and malfunction).In summary, an inability to obtain timely neuroimaging remains a signi cant barrier to effectively detecting and treating ABI in patients on ECMO.
Ultra-low-eld (< 0.1T) portable MRI (ULF-pMRI) technology speci cally addresses this issue, enabling clinically meaningful, sensitive imaging in the presence of ferromagnetic materials in complex clinical care settings, such as intensive care units (ICUs), without causing signi cant adverse events. 7By virtue of its markedly smaller magnetic eld footprint compared to conventional MRI, this technology allows for the use of equipment that is not MR-compatible, even as close as a few feet of the system. 7ULF-pMRI has a reduced speci c absorption rate which diminishes the heating of conductive devices and implants, and eliminates pulse sequence parameter constraints. 8Our phantom and animal studies have shown the safety and compatibility of pMRI with the ECMO circuit components, without deleterious sequelae of the magnetic forces or heating, 9 a safety and e cacy supported by its use in a case series in adults supported by ECMO. 10 Our current study aimed to determine the safety of ULF-pMRI and validate its diagnostic capability in ECMO patients in a multicenter prospective observational study.Furthermore, we aimed to investigate the frequency and types of ABI occurring during ECMO support, an evaluation unhindered by the logistics and danger of transporting these patients to off-unit imaging.

Methods
The study was conducted following the Declaration of Helsinki and approved by the Institutional Review Boards of The Johns Hopkins Medicine (IRB00285716 approved on 08/10/2021) and the University of Texas-Houston (HSC-MS-22-0608 approved on 11/07/2022).Informed consent was obtained from a legally authorized representative as ECMO patients were unable to provide consent.This study, conducted from August 2022 to November 2023, was a prospective observational study (SAFE MRI ECMO study: NCT05469139) performed at these two academic tertiary referral centers (Johns Hopkins Hospital and the University of Texas-Houston) with high ECMO volumes.

Patient Population
Patients enrolled in the study were adults (≥ 18 years) supported with either venoarterial (VA) or venovenous (VV) extracorporeal membrane oxygenation (ECMO).Exclusion criteria were the well-known contraindications for conventional MRI (Supplemental Table 1).

Study Procedure
Point-of-care MR exams were performed with a 64mT Swoop® MR imaging system (Hyper ne, Inc., Guilford, CT; hardware versions 1.6 and 1.8 with software versions 8.6.1 and rc8.7).The MR system was moved into the patient's room by a trained operator.All ICU equipment remained outside the 5-gauss line.
Once the patient's bed was aligned with the MR system's transfer bridge, 4-6 trained people (physician leader, post-doctoral fellow or coordinator, respiratory therapist, perfusionist, and nurses) used the patient's bed sheets to move the patient into the scanner using a lift-and-slide maneuver.Detailed explanations and safety protocols in patient and equipment positioning are described in our prior publication. 10T1-weighted, T2-weighted, uid-attenuated inversion recovery (FLAIR), and diffusionweighted imaging (DWI) with automatically calculated apparent diffusion coe cient (ADC) map sequences were acquired.All MR images were read by a board-certi ed neuroradiologist in each center (H.I.S.; J.G-H.), blinded to patients' clinical information and neurological status.
When possible, pMRI was obtained within the rst 72 hours of ECMO cannulation allowing comparison to a conventional head CT (HCT), the latter ordered as part of a standardized neuromonitoring protocol, 5 performed within 24 hours of the pMRI exam.The patient's vital signs were monitored continuously and recorded every minute during the MR exam/ image acquisition.In addition, ECMO ow, cannula position, and endotracheal tube position were monitored during the scan.Patient demographics, pre-ECMO comorbidities, severity of critical illness including Glasgow Coma Scale (GCS), ECMO cannulation strategy, ECMO duration, and survival to hospital discharge (mortality) were also recorded.

Adverse Events and Stopping Rules
Adverse events (AEs) were monitored during the pMRI scan and de ned as follows: i) a change in mean arterial pressure (MAP) of ± 20%; ii) a decrease in ECMO ow of > 10%; or iii) a decrease in peripheral oxygen saturation (SpO 2 ) of > 10%.AEs were further categorized by severity, as minor or serious, and by relation to the pMRI exam, related or unrelated, as assessed by the study team (S.-M.C., A.G., G.J.W., and K.D.).In the case of a serious AE, MR exams were paused to assess the cause and safety of resuming the exam.

Outcomes
Primary outcomes were the safety of using ULF-pMRI in patients during ECMO support, de ned as completion of pMRI exam without serious adverse events (AEs), and validation of useful, legible imaging acquisition.Secondary outcomes included the frequency and types of ECMO-associated ABI and a comparison of its detection by MR images compared to HCT images.

Statistical Analysis
Descriptive statistical analysis was performed.Continuous variables were expressed as medians with interquartile range (IQR).Categorical variables were expressed as frequencies with percentages.Wilcoxon rank-sum and Pearson's chi-squared tests were used to compare continuous and categorical variables, respectively.Demographic and clinical variables including ECMO variables in subjects with and without ABI were compared.P values < 0.05 were considered statistically signi cant.All analyses were carried out in STATA 17 (StataCorp, LLP, College Station, TX).

Patient Characteristics
Based on the inclusion criteria, 53 patients were eligible for a scan and 3 patients could not be scanned due to facial/scalp edema and large head size that did not t with the head coil of the ULF-pMRI scanner.Thus, ULF-pMRI was performed in 50 patients (median = 58 years; 52% males), 34 (68%) with VA-ECMO and 16 (32%) with VV-ECMO.Of 34 VA-ECMO patients, 11 (32% of 50) were centrally cannulated and 23 (68%) were peripherally cannulated.In 16 VV-ECMO patients, 9 (56%) had single lumen cannulation and 7 (44%) had double lumen cannulation.The studied patients had a median GCS of 6 and Sequential Organ Failure Assessment (SOFA) of 13 at the time of ULF-pMRI.Baseline and clinical characteristics, including ECMO variables, of the cohort are summarized in Table 1.

Safety of ULF-pMRI
All patients were imaged successfully with ULF-pMRI, achieving the primary outcomes in 2 centers.AEs occurred in 3 (6%) patients (Table 2, Supplemental Table 2) with 2 minor AEs (4%) and 1 serious AE (2%).An ECMO suction event (minor AE, related) occurred in one patient (2%) due to insu cient support of the thoracolumbar spine leading to exion at the groin.Repositioning the body, avoiding exion of the spine between the bed and the scanning platform, resolved this.The ULF-pMRI scan was resumed without further issues.The second minor AE event was a transient decrease in ECMO ow from 6.0 to 0.9 L/min, accompanied by 4% decrease in MAP for several minutes.The patient recovered spontaneously.This AE was deemed unrelated to pMRI as the patient had experienced a similar event earlier in the day, well before the scan.One (2%) serious AE deemed related to the scan when, 6 minutes into the scan, the patient's intra-aortic balloon pump (IABP) stopped functioning.The scan was paused.The IABP had been set to ECG trigger, and we suspected that the magnetic eld generated by the pMRI system was affecting the patient's ECG leads, causing signi cant artifacts.The IABP trigger was then switched to internal trigger at a set rate of 75 beats per minute, and the scan was resumed and completed successfully without any hemodynamic issues (Supplemental Fig. 1).

Validation of ULF-pMRI
All patients received ULF-pMRI exams at the patient's bedside with T1-weighted, T2-weighted, FLAIR, and DWI with ADC map sequences.The average time from ECMO cannulation to pMRI was 3 days (Table 2).Overall, pMR images demonstrated good quality images to discern intracranial pathologies with good quality, assessed by the board-certi ed neuroradiologists.Eight patients (16%) had imaging artifacts (Supplemental Table 3).The artifacts were mostly regional and did not in uence the ability to diagnose ABI, except for one case (2%) that had a signi cant artifact which impacted the accurate assessment of intracranial pathology.This artifact was noted only on DWI sequence at the skull base and the imaging resolution was suboptimal, with an uncertainty if the observed abnormality represented infarct vs. artifact since it encompassed different vascular territories (Fig. 2).

Discussion
Our study demonstrated the safety and validity of an ULF-pMRI system in patients with ECMO support, which is an external validation of our rst case series of human subjects. 10ULF-pMRI was rst shown to be bene cial in neurocritical care ICU patients as an innovative neuroimaging system at the bedside with good sensitivity and speci city for diagnosing ABI 7 ; ischemic stroke and ICH detection for pMRI were 100% 11 and 80% 12 in non-ECMO patients.We previously demonstrated the safety and compatibility of ULF-pMRI in ECMO without deleterious magnetic force (no displacement of the cannula) or heating of the ECMO circuit or its components in a phantom model and 3 human subjects. 9Now based on the results of our multicenter prospective cohort study, we externally validated the safety of ULF-pMRI in 50 patients on ECMO support.This data is important not only for ECMO patients but also for those with other cardiac devices in whom ≥ 1.5T MRI is contraindicated where ULF-pMRI has the potential to change our neurological diagnostic approach.As timely and accurate ABI diagnosis is crucial in improving the overall ECMO outcomes, 13 ULF-pMRI can eliminate the need to transport unstable ECMO patients, which has the potential to universally reshape clinical care and outcomes and ECMO clinical trials with the promise of accurately diagnosing ABI promptly.

Adverse Events
We reported 2 minor AEs (ECMO suction event and a transient decrease in ECMO ow) in our prior publication in 3 patients, 10 which was likely related to the initial learning curve in scanning ECMO patients with pMRI.After optimizing our work ow and the logistics of performing pMRI in ECMO patients (i.e., optimal patient positioning: positioning the patient's head and body as at as possible, preventing exion of the spine between the bed and the scanning platform 10 ), we only had 1 AE in 47 patients between 2 centers.One (2%) serious AE (related) occurred when the IABP stopped functioning as a result of being set to ECG trigger with the magnetic eld causing signi cant ECG artifacts on telemetry.Although IABP trigger was switched to an internal set rate (internal-trigger) in our case, in patients who are pulsatile, pressure-trigger would be preferred as an internal set rate could in ate the balloon during systole and adversely affect left ventricular unloading.

Incidence and Type of ABI
We previously demonstrated that the application of the standardized noninvasive neuromonitoring protocol including HCT led to an expected increase in early ABI detection with 33% ABI frequency. 5,13herefore, our study con rmed our hypothesis that ULF-pMRI detects ABI (44%) better than the conventional approach utilizing HCT.Notably, ULF-pMRI was able to detect even small-to moderate-sized ischemic infarcts, which otherwise would have not been possible to detect in HCT (Fig. 1).Also, it's important to note that ULF-pMRI likely is not able to detect tiny "punctate" infarcts that often are visualized in conventional MRI.When comparing ULF-pMRI to HCT in 18 patients who had both, ULF-pMR outperformed HCT for the detection of ABI, especially ischemic injury, and may underperform for hemorrhagic injury (Table 4, Supplemental Fig. 2).suboptimal sensitivity to ICH, although a small sample size, may improve with modi cation of the imaging acquisition technology.Of note, the standardized neuromonitoring protocol was associated with a signi cant improvement in favorable neurological outcomes (modi ed Rankin Scale ≤ 3) in ECMO survivors compared to the era before the standardized neuromonitoring protocol (33% vs. 14%). 5,13This highlights the importance of adding ULF-pMRI to the standardized neuromonitoring approach, as early accurate ABI detection and early interventions are critical for mitigating ABI (i.e., anticoagulation management) and worsening neurological outcomes.

Potential Bene ts in ECMO Clinical Trials
Recent randomized controlled trials (RCTs) on extracorporeal cardiopulmonary resuscitation (ECPR) patients (ARREST, 13 PRAGUE, 14,15 and INCEPTION 16 ) and cardiogenic shock patients (ECLS-SHOCK 17 ; albeit 77% of included patients had pre-ECMO cardiac arrest) trials demonstrated that ABI is the leading cause of death and ECMO discontinuation in VA-ECMO patients, which most likely contributed to the neutral trial results in survival outcomes.The unfavorable neurological prognosis (43%) was the leading cause of ECMO discontinuation in the INCEPTION trial, 16 and 25% of deaths were due to brain death in the PRAGUE trial in the ECPR group. 14,15Although ABI is a major factor, if not the leading cause, for poor survival outcomes in the ECMO RCTs, there is signi cant heterogeneity and a lack of standardization in ABI de nitions and post-ECMO neurological approach and neurocritical care.Our study suggests that ULF-pMRI, as a standard neuroimaging diagnostic tool, can be used in future clinical trials (standardization, Fig. 3).

Limitations
Our study has limitations.First, the timing of ULF-pMRI and HCT were not mandated in the study.
Although we provide information on the comparison between two neuroimaging modalities, only 36% (n = 18) had both imaging studies within 24 hours.Furthermore, ongoing injuries can occur within several hours between the scans, which cannot be captured in our study.However, we took a pragmatic approach as moving "unstable" ECMO patients for another neuroimaging scan always carries a risk during transport.Second, we were not able to provide information on "symptomatic" vs. "asymptomatic" brain lesions as most of these patients are heavily sedated early in the ECMO course and, thus, it is di cult to draw a meaningful conclusion whether all observed ABIs were clinically signi cant.However, it is also important to highlight the fact that any "silent" ABI is highly likely associated with long-term outcomes such as cognitive function. 18Because neurological exams are often not feasible or sensitive in the presence of critical illness and sedative use in ECMO, the results of this study provide an insight as to how to improve future ECMO clinical trial design.Lastly, other cardiac devices such as percutaneous ventricular assist devices or left ventricular assist devices are often placed along with ECMO and these patients were excluded from this trial as the safety was not proven.Pre-clinical and clinical safety were demonstrated in IABP and pMRI. 19nclusions ULF-pMRI is shown to be safe in ECMO patients across different ECMO cannulation strategies.The incidence of ABI was high with 44% in ULF-pMRI, with ischemic stroke being the most common type.Our study, by introducing and testing an alternative form of neuroimaging, will signi cantly lower this barrier to diagnosing and treating ECMO-associated ABI.This innovative approach can inform the design of subsequent clinical trials and validate strategies to mitigate ECMO-associated ABI.

Declarations
Source of Funding This research was funded by Hyper ne, Inc.Dr. Cho is supported by NIH (1K23HL157610).Dr. Keller is funded by NHLBI 5K08HL14332.Dr. Gusdon is supported by the NINDS (K23NS121628).

Table 1
Baseline and clinical characteristics of ECMO patients.

Table 2
Time from cannulation to ULF-pMRI and the frequency of adverse events during the scan.

Table 3
Incidence and type of acut brain injury (ABI) in ULF-pMRI.

Table 4
In 18 patients in whom both a HCT and ULF-pMRI were obtained, the breakdown of MRI and HCT diagnosis in the 10 patients with ABI.