Awake pediatric brain-MRI: a fast multi-sequence scan augmented with motion compensated single-shot 2D acquisitions

Abstract

Background

Magnetic resonance imaging (MRI) is susceptible to motion artifacts, a particular challenge in pediatric imaging. External motion tracking devices and navigator techniques have been previously explored, but challenges persist necessitating sedation.

Purpose

To establish a new awake brain-MRI exam tolerated by children with diagnostic quality images.

Materials and Methods

Participants were prospectively recruited and investigated using an in-house developed multi-sequence scan called NeuroMix that produces T1-weighted, T2-weighted, T2*-weighted, T2-FLAIR, and diffusion-weighted images in under 3 minutes. Additionally, a self-created motion tracking device was attached to participants' foreheads to perform prospective motion correction (PMC) on 2D single-shot sequences that produce higher resolution images of the same contrasts as NeuroMix.

Three neuroradiologists scored the completed series for artifacts. The effects of age group (<5 vs ≥5 years) and sequence type (NeuroMix vs PMC) were evaluated with a Chi²-test.

Results

Of the 64 participants recruited (mean age 6.7 years [2.7 standard deviation]) 58 completed their examination. Head motion recorded during PMC sequences revealed prevalent superior-inferior displacements [25% (67/293) exceeding 13.2 mm], and chin-up/down rotations [25% (67/293) exceeding 13.7°]. Sequence redundancy through NeuroMix and PMC scans resulted in 93% (54/58) of completed examinations having all series essential for producing an MRI-report rated as artifact-free, and therefore a report of high confidence in 84% (54/64) of participants. 22% (13/58) of completed exam reports could have been written using NeuroMix alone, the remaining required PMC- T2-weighted or T2-FLAIR sequences.

Conclusion

This protocol reliably provided diagnostic quality images and reports with high radiologist confidence and could reduce the use of procedure sedation in children.

Abbreviations

PMC = prospective motion correction, WRAD = wireless radio-frequency triggered acquisition device, FLAIR = fLuid attenuated inversion recovery, PROPELLER = periodically rotated overlapping parallel lines with enhanced reconstruction

Summary

A 25-minute brain MRI protocol, featuring a quick multi-sequence scan and 2D single-shot acquisitions guided by a forehead-mounted motion tracker, can reduce the need for sedation in imaging young patients.

Key Results

• In a prospective study of 64 pediatric participants (mean age 6.7 years) undergoing their first awake brain-MRI, 58 completed their examination.

• Median examination duration was 22 minutes [16-26 IQR], with one-quarter of series acquired during a range exceeding 13.7° chin-up/down rotation.

• In 93% (54/58) of completed examinations, all series deemed important for reporting by the radiologist were artifact-free.

• Use of sedation in imaging young patients could be substantially reduced.

Introduction

Magnetic Resonance Imaging (MRI) is sensitive to motion because each sequentially sampled data point contributes to every pixel of the final image. Inconsistencies result in complex artifacts such as ringing, ghosting, or blurring.

Various methods have been proposed for detecting and correcting head motion. These include taking advantage of redundancy in the acquired data,^1–3 interleaving fast low-resolution images (navigators),^4–8 and adding external motion tracking hardware.^9–11 However, even accurate motion measurement faces challenges with large head movement due to interactions with the static magnetic field.¹² This interaction results in pose-dependent phase variations in the brain that are further exacerbated by the head's movement relative to the radiofrequency receive coil array. These secondary effects introduce irregularities in k-space, complicating measurement, and reconstruction, and ultimately limit the extent of correctable motion.

It has therefore become necessary to combine clinical MRI scans with sedatives such as intranasal dexmedetomidine¹³ or general anesthesia for imaging pediatric patients, who are less cooperative and more likely to move. At our institution, all patients younger than 6 years are scanned under sedation or general anesthesia. Older patients are scanned awake, of which approximately 30%-40% require rescans under sedation. This prolongs the examination session and requires additional staff, equipment, and training where delayed examinations can result in substantial logistical challenges. This additional overhead has been shown to increase the cost of an examination by 3- and 9-fold for sedation and anesthesia-based workflows respectively.¹⁴ Furthermore, it adds some risk to an otherwise safe non-ionizing diagnostic tool.

The current study aims to determine the level of motion compensation necessary to match the success rates of sedation-based brain-MRI workflows. Additionally, the diagnostic performance of our newly developed 25-minute brain imaging protocol is assessed. The protocol is structured into two layers. The first layer operates under the assumption that the patient is not always moving, utilizing a single-scan brain exam called NeuroMix¹⁵ to capture all the information required for the radiological report as quickly as possible. The second layer captures slower, higher-resolution images using a self-made external tracking device to measure and update the scan geometry to follow the moving head, while simultaneously taking advantage of single-shot 2D imaging to capture partitions of the data. This mitigates the secondary effects of motion as the field state only needs to be constant for each partition.¹⁶

Materials and methods

Cohort

The study received approval from the Swedish Ethical Review Authority (2020-04878). Pediatric participants were recruited prospectively at a single center from the regular clinical workflow and returned for the study scan if the clinical investigation was unsuccessful. All participants were undergoing their first awake scan session without the use of contrast. Patients lacking verbal communication were excluded.

Motion robust pulse sequences

All imaging was conducted on a GE SIGNA Premier 3 T system employing a 48-channel head coil. A motion-robust pediatric brain MRI protocol was devised for the study (Table S1). All pulse sequences used in the protocol were developed by the authors at Karolinska University Hospital using the KS Foundation (https://ksfoundationepic.org/) framework.

Layer 1—NeuroMix

The initial sequence for all participants was NeuroMix,¹⁵ comprising motion-robust 2D single-shot modules and multi-shot 3D echo planar imaging (EPI) modules; Sag 3D-T1w (21 seconds) and Ax 3D-SWI (29 seconds). For the first seventeen exams, NeuroMix used a T1-fluid attenuated inversion recovery (FLAIR) module with an EPI readout. This was modified to periodically overlapping parallel lines with an enhanced reconstruction (PROPELLER) module for the rest of the study at the cost of 17 seconds of additional scan time.

Layer 2—Prospective motion correction

For the remainder of the protocol, 2D prospectively motion corrected (PMC) sequences with greater anatomical detail were acquired (referred to as PMC scans for the remainder of the text). This included T2w-, T1-FLAIR-, and T2-FLAIR-PROPELLER as well as sequential susceptibility weighted EPI¹⁷ and diffusion weighted EPI¹⁸ scans. The 2-mm T2w-PROPELLER protocols were preferred over the 4 mm, particularly when the referring question was related to seizure focus. These sequences are however twice the duration and were substituted for the 4-mm acquisitions when not specifically requested. For exams with time remaining, a PMC-enabled Cartesian 3D T1w spoiled gradient echo (SPGR) sequence was included. This was to serve as a reference in investigating the efficacy of PMC in 3D versus 2D single-shot encoding schemes. In-plane retrospective motion correction without blade rejection was performed for the PROPELLER image reconstructions.

A Wireless Radiofrequency-triggered Acquisition Device (WRAD),¹⁹ designed and fabricated by the authors, was employed to monitor head pose (Figure 1). The WRAD, about the size of a USB flash drive, was attached to a small sled and mounted on each participant's forehead via bio-compatible double-sided tape (Video S2). The WRAD enables real-time monitoring of head movement during scanning by sensing the gradient magnetic fields during 2-ms-long sinusoidal navigators. The WRAD behavior was orchestrated by messages encoded in tiny (0.1° flip angle) radiofrequency pulses preceding each navigator.²⁰

Figure 1.

(A) The WRAD device is shown with and without its protective casing. The three orthogonal pickup coils, that detect the gradient spatial encoding and radiofrequency pulses, are enclosed in the head of a character-shaped casing. The device is compatible with a universal serial bus type-A connector, allowing it to charge the onboard battery that gives 5 hours of continuous operation. (B) The device slides into a small sled with a curved surface to comfortably fit the participant's forehead (centre). The mount adheres to the skin using bio-compatible double-sided tape. (C) The device transmits the participants' pose out of the scanner using a low-power 2.4 GHz radio transceiver. The messages are decoded by a wireless receiver and forwarded to the scanner control computer within 3 ms of the end of each navigator. The messages are received by the scanner control computer to update the gradient, RF, and oscillator settings (4-8 ms) so that the field of view follows the patient as they move. WRAD = Wireless Radiofrequency-triggered Acquisition Device; RF = radiofrequency.

Navigators were inserted before each excitation pulse and inversion pulse for PMC sequences. Pose measurements were computed by the WRAD and transmitted out of the bore via a low-power 2.4 GHz radio link (Figure 1). The scanner control computer received updates through an ethernet port within 3 ms of each navigator's end. Subsequent pulse sequence waveform modifications took 4-8 ms (depending on sequence complexity) to ensure a latency of under 11 ms. Volunteers were imaged to demonstrate the utility of PMC (Figure 2).

Figure 2.

Example images (Coronal T2w-PROP 2 mm) showing the effect of PMC on image quality by reproducing participant motion in a healthy volunteer scan. (A) participant scan with a moderate amount of motion and PMC motion updates applied to the FOV (turned on). (B) reference volunteer image without motion. (C) image where the volunteer intentionally moved like the participant, with PMC turned off. (D) same as (C) but in this case PMC was turned on. The PMC series (D) is of a similar quality to the no-motion reference (B). PMC = prospective motion correction, FOV = field of view.

Scanning preparation and procedure

While play therapy²¹ and mock MRI scanners²² have been employed for pediatric preparation, this study aimed to assess protocol success using standard adult patient resources. Deviations from standard workflow included allowing all enrolled participants to watch a film of their choice during the examination and decorating the scanner room. Aside from headphones, no motion-restrictive padding was used. In the latter part of the study, two preparation videos were created to familiarize children with the process, from hospital arrival to the end of the exam, offering a first-person view of entering the tunnel (available at: bit.ly/wradfilm). For anxious participants, acoustic noise reduction was used, at the expense of increased scan time and geometric distortions in EPI-based sequences. Parents were permitted to accompany participants into the scanner bore if needed.

Data analysis

Three radiologists at Karolinska University Hospital rated the completed series and assessed their confidence in writing the exam report:

• Neuroradiologist, 12 years of experience.

• Neuroradiologist, 8 years of experience.

• Pediatric Neuroradiologist, 12 years of experience.

Examination outcome assessment

Examinations were categorized based on the neuroradiologists’ assessments of whether repeating the protocol with sedation would yield the same or improved report. The importance of a tissue contrast was based on the diagnostic question and the neuroradiologist’s opinion:

• not needed—no artifacts in all series.

• probably not needed—no artifacts in the series important for the report.

• probably needed—some motion artifacts in important series.

• needed—strong artifacts in important series.

• incomplete—important series are missing due to participant refusal to enter the scanner or aborting the scan.

Examinations where the report was unaffected by artifacts “probably not needed” or “not needed” were considered a success. The other categories were treated as failures.

Sequence utility assessment

PMC-sequences that in addition to NeuroMix were critical for writing the report were recorded to measure the diagnostic performance of NeuroMix alone. The potential to shorten the protocol while maintaining report quality and diagnostic confidence was also recorded.

Series ratings assessment

Individual series were rated by diagnostic value relative to an expected artifact-free version of the same sequence:

• equal—none or minor artifacts that do not impact the interpretability of the series.

• reduced—some motion artifacts that result in reduced interpretability of the series.

• not diagnostic—strong artifacts.

Readers rated all series over the period of one week.

Participant motion

All participants were placed in the scanner head-first supine. It is expected that they mostly roll their head on the pillow when looking around in this placement, rotating about a vector $\stackrel{\rightharpoonup }{r}=\left(r_x, r_y,r_z\right)$ about a center located at the posterior side of the skull. A FOV located near the center of the head is therefore anterior to the center of rotation with a positive y-displacement $\stackrel{\rightharpoonup }{d}=\left(0, y, 0\right)$. As the head rolls, the required change in position of the FOV $\stackrel{\rightharpoonup }{p}$can be estimated through the vector cross product:

$$\overrightarrow{p}\mathrm{ }\approx \overrightarrow{r}\mathrm{ }\times \mathrm{ }\overrightarrow{d}=(-{yr}_z,\mathrm{ }0,\mathrm{ }y{r}_x)$$

A negative correlation between x-displacement ($p_x)$ and z-rotation ($r_z)$ and a positive correlation between z-displacement ($p_z)$ and x-rotation ($r_x)$ are therefore expected for rolling motion.

All motion profiles (available at: http://brainmrimotion.org/, consolidated under Adam vanNiekerk@CNS/KI) were tagged with the median rating for each series as well as summary metrics, such as z-position range (amount of superior-inferior motion), x-rotation range (head rotation as if moving your chin up and down), and motion fraction; defined as the proportion of the scan spent with an x-angular velocity exceeding one standard deviation from the mean computed overall series (1.71° per second).

Statistics

The effect of participant age on examination outcome was evaluated using a Chi-squared test. Inter-rater reliability was assessed using Cohen's kappa statistic. A median rating was computed for all rated series to resolve disagreements between readers. Series not rated by all three readers were excluded. Contingency tables compared the effect of PMC versus NeuroMix, and age category on the median series ratings. Both comparisons were evaluated using a Chi-squared test. The Pearson correlation coefficient was used to measure the strength of the linear correlation between rotations and positions to assess the level of rolling motion.

Cohort and exam description

Sixty-four pediatric participants, with an average age of 6.7 years (±2.7 standard deviation) were recruited between June 2021 and July 2023. The study included patients aged 3-9 years requiring a brain MRI scan. Also included were select patients aged 9-15 years expected to be unable to remain still. Clinical indications included acquired injury (n = 16), malformation (n = 10), seizure focus (n = 21), tumor (n = 8), and mixed pathology (n = 9).

Examination outcomes

Notably, three participants agreed to enter the scanner bore only with their guardians, all resulting in successful exams (Table 1). Of the ten failed exam outcomes, four completed their exam, but still had some artifacts in important image series (6%: probably needed), three refused to enter the scanner and the remainder aborted the scan before the series relevant for the clinical question could be acquired (9%: incomplete exam). None of the completed exams were rated as sedation being “needed.” The study success rate was therefore 84% (54/64). The protocol was completed in 58 participants resulting in a success rate of 93% (54/58) for the underlying technique. Pathology was identified in 17 participants. The median exam time was 22 minutes, with an interquartile range of 19-26 minutes.

Table 1.

Participant cohort.

Participants	5+	<5	All (%)
Acceptance
Refused to enter scanner	1	2	3 (4.7 %)
Aborted scan, incomplete	1	2	3 (4.7%)
Completed with parent	1	2	3 (4.7%)
Completed (self)	44	11	55 (85.9%)
Exam indication
Acquired Injury	11	5	16 (25.0%)
Malformation	6	5	11 (17.2%)
Seizure focus	19	2	21 (32.8%)
Tumor	7	1	8 (12.5%)
Multiple	4	4	8 (12.5%)
All	47(73%)	17(17%)	64 (100%)
Participant comfort/preparation
Watched preparation video	8	3	11 (17.2%)
Acoustic Noise Reduction	7	8	15 (23.4%)
Found pathology
Yes	14	4	18 (28%)
No	30	10	40 (62.5%)

For participants aged 5 years and older, the exam success rate was 89% (42/47), and 71% (12/17) for the younger group. However, the age category did not significantly impact series ratings (Table 2). The reduced success rate was primarily due to incomplete exams.

Table 2.

Examination ratings and the effect of participant age on outcome.

	5+	<5	All	% [95% CI]
Examination rating a
Incomplete exam	2	4	6	9%, [0.027, 0.161]
Needed	0	0	0	0% [0, 0]
Probably needed	3	1	4	6%, [0, 0.128]
Probably not needed	12	2	14	22%, [0.136, 0.302]
Not Needed	30	10	40	63%, [0.526, 0.724]
Examination outcome (P = .15)
Success	42	12	54	84% [0.768, 0.919]
Failure	5	5	10	16% [0.080, 0.231]
All	47	17	64

^aAssessed on whether a repeat exam with sedation would be necessary to address the exam indication.

Clinical utility of NeuroMix and PMC

A 3-minute-long exam would have been possible in 22% (13/58) of the exams (Table 3) as NeuroMix sufficed for writing the report. Most of these occurred when the referring question related to tumor detection or malformation. For the “seizure focus” and “acquired injury” referring questions (n = 34), the higher resolution PMC scans were deemed critical in all except one case, particularly the 2 mm T2w-PROPELLER (Table S3). The Coronal T2-FLAIR-PROPELLER sequence was considered critical for 43% (22/51) of occasions scanned. The T1w, diffusion, and susceptibility weighted images in NeuroMix were sufficient in almost all exams as the PMC sequences of the same pulse sequence were seldom regarded as critical. Further protocol optimization based on the referring question would therefore have been possible.

Table 3.

Clinical indications for completed reports that could have been written from NeuroMix alone.

Report required:	Referring question
	Acquired injury	Malformation	Multiple	Seizure focus	Tumor	All	% [95% CI]
NeuroMix Only	1	4	2	0	6	13	22% [0.094, 0.354]
NeuroMix and PMC	15	5	5	19	1	45	78% [0.646, 0.906]
All	16	9	7	19	7	58

Series ratings of image quality

The three radiologists assessed a total of 678, 688, and 693 series for image quality, of which 666 were rated by all three (Table 4). The pair-wise inter-rater agreement was moderate with Kappa = 0.49 between readers A and C, Kappa = 0.49 between readers B and C, and Kappa = 0.43 between readers A and C. The median reviewer yielded 82.9% (522/666) rated with none or minor artifacts (equal).

Table 4.

Series ratings and the effect of sequence type and participant age.

Diagnostic value a	Equal	Reduced	Not Diagnostic	All (%)
Series ratings by reader n (percent) Cohen’s kappa statistic; κΑΒ = 0.49, κBC = 0.49 and κΑC = 0.43
Reader A	593	85	10	688 (99%)
Reader B	548	80	50	678 (98%)
Reader C	514	153	26	693 (100%)
Effect of sequence type (P = .08)
NeuroMix	322	57	18	397 (60%)
PMC	230	35	4	269 (40%)
Effect of participant age category (P = .54)
5+ years	424	66	16	506 (76%)
<5 years	128	26	6	160 (24%)
Reader Median	552 (82.9%)	92 (13.8%)	22 (3.3%)	666

^aCompared to a hypothetical motion-free examination.

PMC sequences outperformed NeuroMix in terms of series rating (p = .08). This was mainly due to fewer PMC scans being rated “not diagnostic” (Table 4). The proportion of series rated as “equal” was comparable.

For cooperative subjects (Figure 3, exam A) the T2w PMC acquisitions were mostly beneficial for their alternative slice orientations and higher resolution. For participants who moved more often (Figure 3, exam B) the rapid 3D NeuroMix volumes would show subtle artifacts.

Figure 3.

Example images (all with series ratings of “equal”) from Exam A (left frontal, cavernoma, arrow) and Exam B (right frontal, old contusion, arrow). The motion summary metrics are displayed below each PMC-enabled series. For Exam A, the participant was cooperative, resulting in high-quality images for all series. In this case, the alternative slice orientations of the PMC series were important for visualizing the pathology. Exam B shows images from a participant who moved for each PMC series, but still had high-quality images throughout. PMC = prospective motion correction, FLAIR = fluid attenuated inversion recovery.

Series ratings grouped by sequence and sorted by the proportion of series rated “equal” can be found in Table S4. The 3D T1w-SPGR was the worst performing sequence rated “equal” 26% (8/31) of scans, highlighting the challenges caused by the secondary effects of motion. For 2D sequences, fast motion during the echo-train caused signal dropouts which also affected the PMC sequences, because no field of view updates were applied during the echo-train. Sequences with shorter echo-trains (SWI, DWI, and T1-FLAIR, 60-125 ms) therefore performed better on average. The T2w 2-mm PROPELLER sequences never received a rating worse than “reduced” and were rated “equal” 85% (22/26) and 84% (16/19) for coronal and axial planes, respectively. The 3D NeuroMix modules, although more sensitive to motion than the other sequences, performed well—likely due to their short scan duration.

Participant motion

Scatterplots depict the various updated field-of-view states across all PMC sequences (Figure 4). Chin up/down motion was common, visualized on the sagittal plane, when participants shifted attention between watching a film and tilting their heads to see their guardian. A Pearson correlation coefficient of 0.47 between x-rotation and z-translation and −0.27 between z-rotation and x-displacement suggests rolling motion. Lifting of the head anteriorly was observed when participants moved significantly away from the coil array (Figure 4 -Sagittal plane). All PMC series rated “not diagnostic” had excessive motion in the inferior direction relative to where the technologist had positioned them at the exam start (>30 mm).

Figure 4.

Updated field of view locations relative to the start of each series for all PMC-enable scans. The left-hand side shows the field of view position projected onto the sagittal plane with the marker color encoding x-rotation (chin up/down) whereas the right-hand side shows the axial plane projection with marker color encoding z-rotation (chin left/right).

Figure 5 illustrates the motion summary metrics computed for each PMC scan and how the level of participant motion impacted the series rating. Notably, 25% (67/269) of series were acquired during motion exceeding superior-inferior motion of 13.2 mm, chin up/down nodding of 13.7°, and 11.2% of the scan duration spent moving.

Figure 5.

Quartiles of motion summary metrics are computed over each PMC (prospective motion correction) scan. Quality rated in comparison to hypothetical motion-free scan. The distribution of the metrics is illustrated through 4 rectangles of equal area. Each one represents one quartile (67 series). The Pie charts illustrate the ratings within each quartile. At the fourth quartile for each metric, where the range of motion is broadest, image quality is degraded, and the proportion of series rated as equal is reduced as PMC is unable to maintain image quality over such large/frequent motion.

Sequence redundancy was important for achieving a high exam success rate. As the level of motion increased first the 3D acquisitions began to fail, then the single-shot NeuroMix sequence-modules followed by artifacts in the 2D PMC sequences. Some participants moved throughout their exam making the acquisition of artifact-free 3D volumes challenging (Figure 6, exam C). In these exams the 2D single-shot NeuroMix and PMC series were critical. On the few occasions where all the NeuroMix series failed, the exam success depended solely on the PMC-enabled series (Figure 6, exam D).

Figure 6.

Example images (from series with varied ratings) from Exam C (left parietal, gliosis after previous asphyxia, arrow) and Exam D (slight volume reduction and signal change in ventral thalamus, solid arrow, and posterior lentiform nucleus, dashed arrow, bilaterally). For Exam C, the T2w and T2-FLAIR series was important for visualizing pathology. In each case, the PMC sequences produced high-quality images despite the large amount of motion. In Exam D, the NeuroMix series were all degraded (in T2-FLAIR-SSFSE the inversion pulse missed the slice resulting in a T2w image). However, the PMC series resulted in a successful exam outcome. PMC = prospective motion correction, FLAIR = fluid attenuated inversion recovery, SSFSE = single shot fast spin echo.

Discussion

Here we described a sedation-free method for pediatric brain-MRI, leading to high-confidence radiological reports for 84% (54/64) of participants. Failures were predominantly due to incomplete exams or reluctance to enter the tunnel, primarily among participants under 5 years of age. Among the participants who completed their examination, the newly developed two-layered 25-minute examination achieved a 93% (54/58) success rate.

The first layer involved a rapid multi-sequence scan called NeuroMix, generating eight image series with full brain coverage in 3 minutes. It produced all the images necessary for formulating a report for 22% (13/58) of the exams, especially those performed for tumor evaluation. The high diagnostic value of NeuroMix is promising, given its hardware independence, and simplicity. It suits exploratory exams where lower resolution suffices. NeuroMix's speed positions it as a valuable trauma imaging tool.

The second layer utilized a wearable motion tracker to measure head pose and promptly update scans within 11 ms, to ensure accurate alignment of the slice stack for 2D magnetization-prepared pulse sequences—that performed equally well even though each slice is sampled seconds after the inversion pulse.

Attachment of the wearable motion tracker minimally affected workflow and had high levels of acceptance from participants. It enabled high-resolution T2w and T2-FLAIR imaging crucial for complementing the NeuroMix image series. While motion-corrected diffusion-weighted, susceptibility-weighted, and T1-FLAIR images were successfully acquired, these series—in addition to NeuroMix—were rarely pivotal for report writing, allowing room for protocol optimization and shorter exams.

Participants moved predominantly in the superior-inferior direction in a way that was correlated to a nodding “chin up/down” rotation. This type of motion is similar to other studies,^23,24 however, the upper limit of the motion we observed appears larger. This could be because child-parent interactions were not discouraged. Moreover, participants underwent little preparation, besides the Supplementary Material provided.

A noteworthy advantage of a successful awake MRI scan is its potential as a learning opportunity for the patient, likely leading to improved cooperation in future exams. Consequently, this approach is well-suited for patients requiring frequent scanning. Additionally, maintaining a record of patient motion history could prove advantageous in designing protocols for subsequent exams, for instance, 3D Cartesian volumes would be plausible in the 25% of participants who moved very little.

The increased workload on technologists with awake pediatric patients is a significant consideration that must be posed against the additional staff, training, and resources required for sedation-based workflows. The presented protocol could find use in a clinical setting as an express-brain alternative. Toward the study's conclusion, we successfully scheduled participants in consecutive 30-minute slots, a more than twofold improvement over the present sedation-based workflows at our institution.

In pediatric neuroradiology, the need for sedation and its increased resources is a substantial clinical problem, especially in younger children, who are sedated more often.^14,25 The success rate of the presented exam can thereby increase imaging availability, by reserving the time and resource-demanding sedation procedures for those that cannot be imaged awake.

Author contributions

Adam van Niekerk (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing), Jan Svoboda (Conceptualization, Investigation, Project administration, Supervision), Åsa Aspelin (Formal analysis, Investigation, Project administration), Tim Sprenger (Conceptualization, Investigation, Methodology), Henric Rydén (Conceptualization, Methodology, Software, Writing—review & editing), Ola Norbeck (Conceptualization, Investigation, Methodology, Software), Enrico Avventi (Conceptualization, Investigation, Methodology, Software), Katarina Johansson (Investigation, Methodology, Project administration), Natalie Rygaard (Investigation, Methodology, Project administration), Peter Lindberg (Investigation, Methodology, Project administration), Ronny Wickström (Conceptualization, Investigation, Project administration, Resources), Anna Falk-Delgado (Conceptualization, Data curation, Formal analysis, Investigation, Writing—review & editing), and Stefan Skare (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—review & editing)

Supplementary material

Supplementary material is available at Radiology Advances online.

Funding

This work was funded by Barncancerfonden (MT2020-0006). S.S. receives research support from General Electric Healthcare (A04).

Conflicts of interest

Please see ICMJE form(s) for author conflicts of interest. These have been provided as supplementary materials. T.S. is an employee of General Electric Healthcare.

Data availability

All motion profiles recorded during this study are available at: http://brainmrimotion.org/ (consolidated under Adam vanNiekerk@CNS/KI), each dataset is stamped with artifact level in the corresponding image-series. Patient preparation videos are available at: bit.ly/wradfilm. All pulse sequences used in this work were developed using the KSFoundation framework, available at: https://ksfoundationepic.org/. The NeuroMix sequence is available upon filling out the request form here: https://ksfoundationepic.org/getthecode. The volunteer reproduction of patient motion was performed utilizing the motion visualization tool available here: https://gitlab.com/neuromr_karolinska/mocovisualisation.git. Data containing patient information is not available.