Evaluation of the Aggregated Time Savings in Adopting Fast Brain MRI Techniques for Outpatient Brain MRI

Abstract

Introduction

Clinical validation studies have demonstrated the ability of accelerated MRI sequences to decrease acquisition time and motion artifact while preserving image quality. The operational benefits, however, have been less explored. Here, we report our initial clinical experience in implementing fast MRI techniques for outpatient brain imaging during the COVID-19 pandemic.

Methods

Aggregate acquisition times were extracted from the medical record on consecutive imaging examinations performed during matched pre-implementation (7/1/2019–12/31/2019) and post-implementation periods (7/1/2020–12/31/2020). Expected acquisition time reduction for each MRI protocol was calculated through manual collection of acquisition times for the conventional and accelerated sequences performed during the pre- and post-implementation periods. Aggregate and expected acquisition times were compared for five of the most frequently performed brain MRI protocols: brain without contrast (BR−), brain with and without contrast (BR+), multiple sclerosis (MS), memory loss (MML), and epilepsy (EPL).

Results

The expected time reductions for BR−, BR+, MS, MML, and EPL protocols were 6.6 min, 11.9 min, 14 min, 10.8 min, and 14.1 min, respectively. The overall median aggregate acquisition time was 31 [25, 36] min for the pre-implementation period and 18 [15, 22] min for the post-implementation period, with a difference of 13 min (42%). The median acquisition time was reduced by 4 min (25%) for BR−, 14.0 min (44%) for BR+, 14 min (38%) for MS, 11 min (52%) for MML, and 16 min (35%) for EPL.

Conclusion

The implementation of fast brain MRI sequences significantly reduced the acquisition times for outpatient brain MRI protocols.

Introduction

Radiology departments increasingly face difficult trade-offs between the demand for medical imaging services, access to imaging resources, and length of exams in an era of declining reimbursements.¹ Practical strategies for expanding and expediting patient access to high fixed cost imaging modalities such as magnetic resonance imaging (MRI) promise to improve patient care and satisfaction.² Among these strategies, minimizing protocol length represents a major opportunity for providing substantial time savings and increasing MRI capacity.

Previous clinical validation studies of accelerated MRI sequences have established fast MRI techniques as key tools for decreasing acquisition time and motion artifacts while preserving image quality.^3–5 The operational benefits following from the time savings brought about by accelerating individual sequences are relatively less well explored. Initial analyses have revealed potentially dramatic time savings with concomitant increases in patient throughput and revenue. For example, analysis of accelerated musculoskeletal MRI examinations performed at a large academic institution suggested two-fold or greater acceleration in exam time without sacrifice of diagnostic quality.^6,7 Algorithms to mitigate motion artifact in MRI may result in large cost savings, estimated to be on the order of at least tens of thousands of dollars per scanner per year.^8,9

The purpose of this study was to assess the aggregated time savings upon incorporating accelerated sequences into the most commonly performed brain MRI protocols at our institution, as part of a larger organizational effort to expedite imaging examinations during the COVID-19 pandemic. The COVID-19 pandemic heightened awareness of the need for efficient imaging protocols to address a multitude of issues, including decreasing the backlog of delayed patient examinations brought about by stay-at-home orders and providing for added time between patients to enforce enhanced infection control measures. We hypothesized that the aggregated time savings from concatenating multiple accelerated sequences for the most commonly used MRI sequences (e.g., diffusion-weighted imaging: DWI, susceptibility weighted imaging: SWI, and fluid attenuated inversion recovery: FLAIR) would result in meaningful reductions in acquisition time across a variety of brain MRI examinations.

Methods

This retrospective study was approved by the Institutional Review Board of our institution with a waiver of informed consent. Patient privacy was ensured in compliance with the Healthcare Information Portability and Accountability Act.

From January through June of 2020, our tertiary-care academic medical center implemented previously optimized accelerated sequences into the clinical brain MRI protocols on three outpatient MRI scanners. Consecutive imaging examinations performed on these scanners were included for analysis of acquisition times during matched pre-implementation (7/1/2019–12/30/2019) and post-implementation periods (7/1/2020–12/30/2020).

In the following sections, the study MR scanners and protocols are described first below, followed by descriptions of how the acquisition times (total versus expected) were determined for each protocol during the pre- and post-implementation periods.

MR Scanners and Protocols

Previously validated accelerated MRI sequences incorporating fast imaging techniques, namely, simultaneous multislice imaging (SMS)¹⁰ and Wave-CAIPI (controlled aliasing in parallel imaging) for highly accelerated 3D imaging^11,12, were implemented on two 3T scanners and one 1.5T scanner (MAGNETOM Vida, Prisma-Fit, and Avanto-Fit, respectively; Siemens Healthineers, Germany). Specifically, SMS was used to accelerate DWI and Wave-CAIPI was used to accelerated all 3D sequences including SWI, Sampling Perfection with Application optimized Contrasts using different flip angle Evolution (SPACE) T2-weighted imaging, SPACE-FLAIR, and Magnetization Prepared Rapid Gradient Echo (MPRAGE). Fast low angle shot (FLASH) gradient echo T1 sequences replaced the conventional Turbo spin echo (TSE) T1 sequence. Acceleration factors were increased for TSE FLAIR, and for all 3D sequences. Detailed acquisition parameters of the sequences are provided in Table 1.

Table 1.

Acquisition parameters of conventional MR sequences.

	Conventional MR sequences
Parameters	DWI	T1	2D TSE FLAIR	SWI	SPACE T2	SPACE FLAIR	MiRAGE	2D TSE T2 (no accelerated version)
Slice thickness (mm)	5	5	5	2	1	1	1	5
FoV	220 × 220	220 × 220	178 × 220	165 × 220	256 × 256	256 × 256	256 × 256	192 × 220
Matrix	128 × 128	256 × 256	396 × 512	182 × 256	256 × 256	386 × 384	256 × 256	448 × 512
TR (ms)	3200	2150	9000	30	6000	5000	2530	7060
TE (ms)	83	54	104	20	397	355	2	85
Flip angle	90	150	150	120	120	120	13	120
B-values (sec/mm2)	1000

Acquisition times were compared for the five most commonly performed brain MRI protocols during the pre- and post-implementation periods, consisting of: brain without contrast (BR−), brain with and without contrast (BR+), multiple sclerosis (MS), memory loss (MML), and epilepsy (EPL) protocols. The five most frequently performed protocols encompass the majority of outpatient indications for brain MRI at our institution. For example, the brain with and without protocol (BR+) is used for the evaluation of brain tumors including metastatic disease at our institution. The brain without contrast (BR−) protocol is used for stroke evaluation. Neuroinflammatory diseases other than multiple sclerosis are evaluated using the MS protocol. All protocols featured accelerated DWI using in-plane and slice acceleration and SWI sequences. 2D TSE T2 sequence was not optimized and did not have an accelerated counterpart; the sequence was included in all MR brain protocols, except for EPL. Specific accelerated MRI sequences included in the MR protocols are listed in Table 2. Representative images of the conventional and fast MRI sequences are shown in Figures 1 and 2.

Table 2.

Acquisition parameters of fast MR sequences.

Parameters	Fast MR sequences
	SMS DWI	T1 FLASH	2D TSE FLAIR	Wave-CAIPI SWI	Wave-CAIPI SPACE T2	Wave-CAIPI SPACE FLAIR	Wave-CAIPI MPRAGE
Slice thickness (mm)	5	5	5	2	0.9	1	1
FoV	240 × 240	225 × 240	195 × 240	224 × 256	250 × 250	240 × 240	260 × 260
Matrix	150 × 150	384 × 512	195 × 320	168 × 256	256 × 256	256 × 256	240 × 240
TR (ms)	2300	240	3000	41	3200	5000	2300
TE (ms)	114	5	123	23	410	424	3
Flip angle	90	70	150	15	120	120	8
B-values	1000

Figure 1.

Representative fast 2D MR sequences: A) SMS DWI (trace & ADC), B) pre-contrast T1 FLASH, C) post-contrast T1 FLASH, and D) accelerated TSE FLAIR. Representative conventional 2D MR sequences: E) DWI (trace & ADC), F) pre-contrast TSE T1, G) post-contrast TSE T1, and H) TSE FLAIR. The acquisition time for the representative sequence images are listed.

Figure 2.

Representative fast 3D MR sequences: A) pre-contrast Wave-CAIPI MPRAGE, B) post-contrast Wave-CAIPI MPRAGE, C) Wave-CAIPI SWI, D) Wave-CAIPI SPACE T2, and E) Wave-CAIPI SPACE FLAIR. Representative conventional 3D MR sequences: F) pre-contrast MPRAGE, G) post-contrast MPRAGE, H) SWI, I) SPACE T2, and J) SPACE FLAIR. The acquisition time for the representative sequence images are listed.

Acquisition Times

Both expected and total post-implementation acquisition times were reported, as defined below, to provide predicted time savings attributable to individual sequences and those found in the protocols in aggregate.

Expected Acquisition Time

The expected acquisition time reduction for each MR sequence was calculated by subtracting the acquisition time of the accelerated sequences from those of the corresponding pre-implementation conventional MRI sequences; this was achieved by manual collection of acquisition times for five representative studies for each sequence.

Total Acquisition Time

The total acquisition time was calculated as the difference between the start and end times entered by the MRI technologists into the Radiology Information System (RIS), which was rounded to the nearest minute. The total acquisition time reflected the actual scan time, i.e., the time from the beginning of the first MR sequence to the end of the last MR sequence.

Due to the coding format of multi-body part examinations, separating acquisition times between body parts was not possible. Therefore, examinations with multiple accession numbers were excluded and contributed to the relatively low apparent exam volumes. Examinations with incomplete acquisition time data were also excluded.

Improvement of acquisition data documentation by the technicians resulted in fewer studies excluded during the post-implementation period. Furthermore, our PACS provider was changed between the pre- and post-implementation periods and acquisition data prior to the PACS change (i.e. pre-implementation period) were retrospectively uploaded to the new data system, which resulted in loss of some information.

Statistical Analysis

The acquisition times for each protocol were not normally distributed and were compared using the nonparametric Wilcoxon rank sum test. Statistical analysis was performed using the Prism software (GraphPad, San Diego, CA). Statistical significance was set at P less than 0.05.

Results

A total of 1008 examinations were performed during the pre-implementation period. 174 examinations were excluded for being a part of a multi-accession examination and 393 examinations were excluded due to incomplete acquisition time data.

A total of 1088 examinations were performed during the post-implementation period. 271 examinations were excluded for being a part of a multi-accession examination and 79 examinations were excluded due to incomplete acquisition time data.

The overall study volumes (% of total volume) for the BR−, BR+, MS, MML, and EPL protocols were 47 (11%), 368 (83%), 6 (1%), 8 (2%), and 12 (3%), respectively, during the pre-implementation period, and 97 (12%), 597 (75%), 31 (4%), 27 (3%), and 40 (5%), respectively, during the post-implementation period.

Expected Acquisition Time

The expected acquisition times obtained from representative examinations for conventional 2D sequences were 3.0 min for DWI, 2.8 min for TSE T1, and 3.6 min for TSE FLAIR (R = 2; Table 3). The expected acquisition times for the accelerated brain 2D MR sequence counterparts were 1.1 min for SMS DWI, 0.7 min for T1 FLASH, and 1.8 min for TSE FLAIR (R = 3; Table 3). The average time reduction for the three 2D sequences was 1.9 min per sequence.

Table 3.

MRI sequences implemented in different brain MRI protocols.

Brain MR protocol	Sagittal T1 (unaccelerated)	2D TSE T2 (unaccelerated)	SMS DWI	2D T1 FLASH	2D TSE FLAIR R = 3	Wave-CAIPI SWI	Wave-CAIPI SPACE T2	Wave-CAIPI 3D SPACE FLAIR	Pre-contrast Wave-CAIPI 3D T1 MPRAGE	Post-contrast Wave-CAIPI 3D T1 MPRAGE
BR−	x	x	x		x	x
BR+	x	x	x	x	x	x				x
MS		x	x			x		x	x	x
MML		x	x			x		x	x
EPL			x	x		x	x	x		x

TSE-turbo spin-echo, FLASH=Fast Low Angle SHot, SWI=susceptibility-weighted imaging, DWI=diffusion-weighted imaging, FLAIR=Fluid Attenuated Inversion Recovery, SPACE= Sampling Perfection with Application optimized Contrasts using different flip angle Evolution, MPRAGE=Magnetization Prepared Rapid Gradient-Echo, CAIPI=Controlled Aliasing in Parallel Imaging, SMS=simultaneous multi-slice

The expected acquisition times from representative examinations for the conventional 3D sequences were 5.1 min for SWI, 6.4 min for SPACE FLAIR, and 5.6 min for MPRAGE (Table 3). The expected acquisition times for the accelerated brain 3D MR sequence counterparts were 2.9 min for Wave-CAIPI SWI, 3.6 min for Wave-CAIPI SPACE FLAIR, and 2.4 min for Wave-CAIPI MPRAGE (pre- and post-contrast; Table 3). The average acquisition time reduction for the three 3D sequences was approximately 3 min per sequence.

The expected protocol acquisition time reductions from implementing the accelerated sequences were calculated by subtracting the acquisition times of the accelerated sequences from those of the corresponding pre-implementation conventional MRI sequences. Through this method, the expected time reductions for BR−, BR+, MS, MML, and EPL protocols were 6.6 min, 11.9 min, 14 min, 10.8 min, and 14.1 min, respectively.

Total Acquisition Time

The overall median acquisition time was 31 [25, 36] min for the pre-implementation period and 18 [15, 22] min for the post-implementation period (Table 4), with a difference of 13 min (42%). The median acquisition time was significantly reduced in all protocols during the post- versus pre-implementation period (Table 4). The median acquisition time was reduced by 4 min (25%) for BR−, 14 min (44%) for BR+, 14 min (38%) for MS, 11 min (52%) for MML, and 16 min (35%) for EPL (Table 5). The absolute differences between median acquisition time reduction and the expected time reductions for BR−, BR+, MS, MML, and EPL were 2.6 min, 2.1 min, 0.0 min, 0.2 min, and 1.9 min, respectively.

Table 4.

Expected acquisition times of conventional versus fast brain MRI sequences.

Conventional 2D sequences	acquisition time (min)	Fast 2D sequences	acquisition time (min)	Time reduction (min)
DWI (R=2)	3.0	SMS DWI (R=2, MB=2)	1.1	1.9
TSE T1	2.8	T1 FLASH	0.7	2.1
TSE FLAIR (R=2)	3.6	TSE FLAIR (R=3)	1.8	1.8
TSE T2	0.67	No accelerated counterpart	n/a	0
Sagittal T1	1.7	No accelerated counterpart	n/a	0
Conventional 3D sequences	acquisition time (min)	Fast 3D sequences	acquisition time (min)	Time reduction (min)
SWI (R=2)	5.1	Wave-CAIPI SWI (R=4)	2.2	2.9
SPACE T2 (R=6)	4.2	Wave-CAIPI SPACE T2 (R=9)	3	1.2
SPACE FLAIR (R=2)	6.4	Wave-CAIPI SPACE FLAIR (R=4)	3.6	2.8
MPRAGE (R=2)	5.6	Wave-CAIPI MPRAGE (R=4)	2.4	3.2

R=acceleration factor, MB=multiband acceleration factor, TSE-turbo spin-echo, FLASH=Fast Low Angle SHot, SWI=susceptibility-weighted imaging, DWI=diffusion-weighted imaging, FLAIR=Fluid Attenuated Inversion Recovery, SPACE= Sampling Perfection with Application optimized Contrasts using different flip angle Evolution, MPRAGE=Magnetization Prepared Rapid Gradient-Echo, CAIPI=Controlled Aliasing in Parallel Imaging, SMS=simultaneous multi-slice.

Table 5.

Comparison of total acquisition times during the pre-implementation period versus post- implementation period.

Brain MR protocol	Pre-implementation acquisition time in minutes (median [25th percentile, 75th percentile])	n	Post-implementation acquisition time in minutes (median [25th percentile, 75th percentile])	n	Reduction in median acquisition time (min [%])	P value
BR−	16 [15, 23]	47	12 [9, 18]	97	4 [25]	<0.001
BR+	32 [27, 36]	368	18 [16, 21]	597	14 [44]	<0.001
MS	37 [34, 63]	6	23 [17, 32]	31	14 [37]	0.005
MML	21 [11, 24]	8	10 [7, 12]	27	11 [52]	0.006
EPL	46 [41, 49]	12	30 [26, 33]	40	16 [35]	<0.001

Discussion

Accelerated MRI sequences have been previously evaluated in clinical settings with a primary focus on diagnostic image quality compared to their conventional counterparts.^{3–5,13–15} The operational benefits of accelerated MRI sequences, including their synergistic time savings, have not been studied as fully. Our results indicate that the implementation of accelerated MRI sequences in the most performed brain MRI protocols significantly decreased the acquisition times on both 1.5T and 3T scanners, offering up to 52% reduction in scan time for the longest examinations and an overall 42% reduction in median acquisition time.

Reducing scan time is not only beneficial from an operational standpoint but is also of importance for improving patient care and access. Longer scan times increase the likelihood that patients will move during the scan, which can degrade image quality and decrease diagnostic accuracy.^16,17 Remaining still for prolonged periods of time is especially difficult for certain patient populations, including those who are acutely ill, in pain or claustrophobic, as well as pediatric patients, resulting in diminished diagnostic quality.^18–20 Implementation of accelerated MR sequences can increase access for patients that would otherwise not tolerate the long scan times of standard MR protocols, prevent repeat scans due to motion degradation, and improve patient comfort and satisfaction.²¹ Future investigations are needed to better evaluate whether the more motion-degradation resistant fast MR sequences are able to reduce callback rates and number of repeat sequences.

Previous studies have demonstrated that accelerated MR acquisitions using techniques such as Wave-CAIPI and SMS achieve similar diagnostic performance compared to standard MR sequences for detecting a range of intracranial pathology, including intracranial hemorrhage, brain metastases, and ischemic, inflammatory, and demyelinating disease.^4,12,22–27 Furthermore, 3D brain MR sequences are increasingly used clinically to improve the volumetric evaluation and detection of brain metastases and demyelinating lesion burden, in which the localization of small lesions requires a greater degree of anatomic precision.^28,29 3D MR sequences, however, require longer acquisition times than their two-dimensional sequence counterparts and are more susceptible to motion artifact due to the volumetric encoding. The MR acceleration techniques used in our study resulted in greater acquisition time reduction for 3D sequences than 2D sequences, demonstrating the importance of 3D sequence optimization and the relatively greater degree of operational benefit, as well as the potential to acquire more information per unit time.

The implementation of the accelerated MR techniques in the selected protocols highlights the efficacy of using accelerated sequences to improve workflow during the COVID-19 pandemic. To reduce the risk of viral transmission, our hospital adopted rigorous sanitation protocols between examinations, including equipment disinfection, floor disinfection, donning of protective clothing for healthcare workers, and appropriate disposal of medical waste.³⁰ The time saved by implementing the fast MR protocols accommodated the heightened infection control procedures without requiring longer booking slots and helped our department maintain daily throughput, which was particularly important in addressing the backlog of examinations that had accumulated during the state-wide stay-at-home order.

The use of diagnostic imaging has continued to increase in the United States during the past decade.^31,32 Contributing factors include increased availability of MRI scanners, increased demand by clinicians and patients, and increased surveillance of patients with diseases such as cancer.^31,33,34 Several strategies have been proposed to address the rising demand for imaging. One such method is for imaging centers and hospitals to expand the hours of scanner operation to accommodate the increase in number of ordered examinations. Healthcare systems could also acquire more scanners or contract with additional outpatient imaging centers to address the need. These methods, however, require initial financial buy-in for hiring additional personnel and acquiring new space and equipment. Optimization of MRI protocols improves scanner productivity without the need for additional financial investment and mitigates downstream costs.³⁵ With an approximate 42% decrease in median acquisition times, implementation of fast MR sequences has the potential to increase operational efficiency substantially.

The time saved per examination may enable the eventual shortening of examination time slots and result in increased patient throughput. The implementation of fast MR protocols shortened the median scan time by 13 min; assuming a dedicated outpatient brain MR scanner is open for 15 hours a day and each outpatient brain MR examination is scheduled for a median time of 45 minutes, the median examination time could be shortened to approximately 30 minute sessions. This time savings would translate to 10 additional examinations that could be performed each day per scanner. If brain MRI reimbursement is $500 (based on Medicare data)³⁶, then this would yield an additional $5,000 per day. Extrapolating further, assuming the scanner is open all year except for holidays (355 days), the increased revenue would total $1,775,000 per scanner per year.

A limitation of this study was the lack of operational and non-image acquisition related time assessment, such as time spent on room set up, patient preparation, and exams performed per day. The optimization of image acquisition time through shortened protocols was prioritized by departmental leadership as an effective way to improve patient throughput systematically. The extrapolation of the potential downstream benefits to patient throughput was made on the assumption that non-image acquisition related aspects of the examinations remained unchanged. Second, only three outpatient MR scanners were included in this study as the first scanners upon which the optimized protocols were implemented. These scanners were thus selected for preliminary assessment of the fast MR protocols, which have since been implemented on multiple additional scanners, including on both inpatient and outpatient scanners across the enterprise. Future investigation with the inclusion of more scanners and a separate analysis of inpatient examination times is warranted to elucidate the impact of such optimized protocols on imaging volume throughput, scanner productivity, and patient access. The sample size was further limited by coding format that prevented extraction of acquisition times between body parts in multi-accession examinations.

Conclusion

The real-world implementation of fast brain MRI sequences significantly reduced acquisition times for the most performed brain MRI protocols. The projected benefits to the clinical workflow and patient care are substantial, and include increased imaging volume throughput, improved patient access for valuable MRI resources, increased departmental revenue, and improved patient and healthcare worker safety during the COVID-19 pandemic.