Clinical Evaluation of a 2-Minute Ultrafast Brain MR Protocol for Evaluation of Acute Pathology in the Emergency and Inpatient Settings

DISCUSSION

There has been growing interest in accelerated brain MR imaging techniques in recent years, especially for vulnerable patient populations, such as pediatric patients and those presenting with acute symptoms in the emergency and inpatient settings. These patient populations are often motion prone and are in settings where timely diagnosis is critical to guiding treatment, management, and disposition on expedited time scales. In our study, we found that the 2-minute ultrafast ms-EPI brain MR protocol, which was 8 minutes (80%) faster than the reference protocol, provided similar overall diagnostic image quality and was concordant with the reference brain MR imaging protocol for the main clinical diagnosis in all but 1 case (98.5%).

Other brain MR acceleration techniques, including single-shot EPI and previously reported variations, such as multicontrast EPI (EPIMix), have been reported.^9,10 However, at high acceleration factors (R > 2), these techniques can suffer from increased noise, geometric distortion, and residual aliasing artifacts.^18,19 A major limitation of single-shot echo-planar FLAIR imaging is the poor tissue contrast between WM and GM. EPIMix, on the other hand, is an ultrafast brain imaging technique by using single-shot EPI to rapidly acquire multiple contrasts.^11,12,20 ms-EPI uses an interleaved EPI method that utilizes multiple excitations, resulting in significantly reduced geometric distortion and higher SNR than single-shot-EPI.^14⇓-16 The proposed multishot EPI approach employed in this paper requires slightly greater scan time (because of the acquisition of multiple shots) with the expected trade-off of improved image quality in the form of increased SNR and reduced image distortion compared with a single-shot EPI approach, including EPIMix. Single-shot EPI may still have a valuable role in extremely motion-prone patients, or when extreme scan speed is desired.

The ultrafast brain protocol used in this study is based on the ms-EPI technique with a machine-learning assisted reconstruction framework that was previously described by Clifford et al.¹⁷ Despite its inherently increased sensitivity to susceptibility-induced variations of the magnetic field, EPI has been previously suggested as an alternative to TSE-based T2-weighted imaging in the clinical setting.^16,21 There are some known limitations to the technique, including signal loss of the extracranial soft tissues and mild geometric distortion in the temporal regions, which are inherent to EPI and are partly mitigated by the multishot acquisition. As expected, the reviewing neuroradiologists rated overall image quality, image noise and geometric distortion to be significantly better in the reference protocol images compared with the ultrafast protocol images. More importantly, however, the overall diagnostic quality of the ultrafast protocol was not significantly worse and was considered to be equivalent compared with the reference standard. In the acute clinical settings, such as the emergency department and the inpatient wards, time to diagnosis, treatment, and management directly correlate with clinical outcomes, including mortality and morbidities, for multiple neurologic conditions.²² In addition to patient outcomes, a delay in diagnosis and management also adds to the burden of the hospital systems as it may increase length of hospitalization, the need for increased hospital capacity, and the need for additional health care services.²³ Thus, with equivalent overall clinical diagnostic utility, the ultrafast brain MR imaging may be an appropriate first imaging examination of choice to exclude life-threatening and emergent neurologic pathology.

This study was focused on patients with acute neurologic symptoms and adds to a growing body of literature evaluating ultrafast MR imaging protocols in the brain. A recently published study evaluating patients with acute stroke imaged with a deep-learning enhanced 2-minute multicontrast EPI examination found that characterization of patients with stroke by using this protocol was equivalent to the reference sequences.²⁴ Another recent study showed that a deep-learning assisted ultrafast multishot EPI examination implemented at 1.5T was effective in detecting acute intracranial pathology.²⁵ Patients evaluated in the acute setting may have underlying chronic or nonemergent intracranial pathology that would require additional follow-up and consultation. Further evaluation is needed to determine the efficacy of the ultrafast brain MR protocol for detecting nonacute intracranial findings. Furthermore, by greatly accelerating MR acquisition across different contrasts, there is a trade-off of limited evaluation of extracranial structures. This needs to be taken into account when evaluating ms-EPI and clinical suspicion for extracranial processes will require further evaluation with standard MR protocols.

In our study, there was only 1 case (of 66) where there was a finding (punctate focus of cortical restricted diffusion) that was not as well visualized on the ultrafast brain protocol but was seen on the reference protocol. Upon further review, the finding may have been due to differences in section positioning and section thickness (4 mm for the reference DWI and 5 mm for the ultrafast DWI), leading to partial volume contamination. This does raise the question of the sensitivity of the ultrafast MR protocol at picking up small findings and whether the acquisition acceleration to exclude emergent pathology outweighs the slightly lower sensitivity at detecting small findings, which most likely have little impact on clinical care. Future studies with a larger patient population with a wide range of imaging findings is needed to provide a more detailed assessment of the ultrafast brain protocols.

The accelerated acquisition time also benefits from reduced sensitivity to patient motion as the overall acquisition time is shorter. In 11 cases (16.7%), there was less motion on the ultrafast brain protocol images than on the reference protocol images. In patients with altered mental status who do not have the ability to hold still, the ultrafast brain MR protocol may provide superior evaluation compared with the reference protocol, which may be rendered nondiagnostic from motion artifacts. A limitation of the reduced sampling of the ultrafast brain protocol ms-EPI sequences is that, if motion were to occur in between shots, motion artifacts may be inadvertently exacerbated from intershot phase error. Motion-correction techniques, such as navigator-based prospective motion correction or scout accelerated motion estimation and reduction techniques, may be helpful to address intershot phase errors with minimal impact on overall acquisition time.^26,27

Pediatric patients are an important population for whom motion during MR imaging is a concern, especially infants and very young children who may have difficulty remaining still for extended periods. General anesthesia may be required for challenging patients, but the use of anesthetics in pediatric patients has known potential negative effects, including nausea, vomiting, and disorientation, or more serious adverse events, such as cardiorespiratory depression.²⁸ In the acute clinical setting, institutions may implement abbreviated MR protocols to avoid sedation that includes only a single contrast, usually T2-weighted imaging.^29⇓-31 The sensitivity of abbreviated MR protocols, however, is limited with 1 study reporting 14% of cases had undetected findings on a fast-brain MR imaging protocol.³² An ultrafast 1-minute pediatric brain protocol has been previously proposed, which uses optimized faster versions of commercially available sequences.³³ While the preliminary image quality was good in a previously proposed ultrafast MR protocol, there was no direct evaluation of the final clinical diagnosis. Accelerated protocols, however, have a limited role in sedated patients as the patient is already under anesthesia and standard imaging will provide superior image quality.

There are several limitations to this study. First, although the patient sample size is adequate with 66 patients, most did not exhibit any acute intracranial findings. Future investigations need to include a larger sample size comprising patients with acute findings to further determine sensitivity and specificity of the ultrafast brain MR protocol. It will be important to determine the optimal balance between reducing scan time and potential slight loss in sensitivity in detecting abnormal findings. Second, this study only included adult patients in the emergency and inpatient settings. Pediatric patients, as mentioned, may greatly benefit from accelerated MR examinations; therefore, it will be important to further assess the efficacy of the ultrafast brain MR protocol in this population. Third, this study was conducted exclusively in acute clinical settings. The ultrafast brain MR protocol may also benefit patients who are claustrophobic or unable to tolerate long MR examinations for various reasons. Fourth, given the inherent differences in image characteristics of the reference and ultrafast MR imaging protocols, true blinding was difficult to achieve for side-by-side comparisons. This is unavoidable but does represent real-life situations when new sequences are introduced clinically and radiologists are asked to evaluate the quality of the new sequence compared with the reference. Last, the neuroradiologist reviewers assessed the ultrafast brain MR protocol and the reference brain MR protocol in succession rather than by randomized evaluation. This method is more stringent on the ultrafast brain MR protocol as the reference brain MR protocol is reviewed after the ultrafast brain MR protocol to identify any missed findings.