MR exploration has become a gold standard for diagnosing several neurological disorders due to its superior tissue contrast that increases the fidelity of the diagnosis. In addition to having a pivotal role in the early diagnosis of some neurological and neurodegenerative diseases, MR is also crucial for longitudinal or post-therapeutic follow-up during the patient’s lifespan.
In some cases, several improvements to MR acquisitions have been adapted to specific pathologies, providing more accurate and specialized visualization of lesions and abnormalities.
For example, neurological disorders and conditions like epilepsy are often related to specific brain areas and functions, such as the cerebral cortex. Focal cortical dysplasia (FCD) is a condition where the brain’s cortex doesn’t form properly during fetal development, leading to abnormal organization of brain cells. FCD is a common lesional cause of epilepsy and may be difficult to detect using conventional MR acquisitions.
FCD studies in MR are very sensitive to contrast change of the white matter (WM) at the junction with the cortical ribbon. Traditional sequences such as T2 Cube FLAIR and T1 MP-RAGE provide sufficient information about these anatomical structures and their morphological aspects. Anomalies are often well depicted with a change of contrast in areas where a normal contrast is expected (Figure 1A and 1B).
However, in some cases with subtle changes between gray matter (GM) and WM, it can be difficult to define borderline lesions. A novel imaging approach, 3D Edge-Enhancing Gradient Echo (3D EDGE) has been shown to increase the depiction of the borders or boundaries between WM and GM using an inversion recovery technique, providing additional ease for image reading and interpretation (Figure 1C).1
Figure 1.
(A) Coronal reformat from sagittal T2 Cube FLAIR, 1 x 1 x 1 mm, 4:02 min. (B) Coronal reformat from sagittal 3D T1 MP-RAGE, 0.9 x 0.9 x 1 mm, 3:09 min. (C) Coronal reformat from sagittal 3D T1 EDGE, 1 x 1 x 1 mm, 5:30 min. Yellow circles show areas of contrast change. 3D EDGE clearly enhances the depiction of the abnormal tissue. Images acquired on SIGNA™ Premier 3.0T are courtesy of Centre Hospitalier Sainte-Anne, Paris, France.
Materials and method
The EDGE sequence is a 3D gradient echo-based acquisition with a fundamental T1 inversion recovery (IR) contrast. While different pulse sequences can be adapted for this technique, for GE HealthCare protocols, we used 3D T1 MP-RAGE for 3.0T and 1.5T systems and 3D BRAVO for SIGNA™ 7.0T.
MR voxels may contain entirely GM, WM or a combination of both. Voxels shared by WM and GM are sensitive to the bounce point phenomenon (or artifact)2 (Figure 2). Selection of an inversion time nulling these partially shared volumes (or voxels) helps provide the typical contrast needed for 3D EDGE (Figure 3).
Figure 2.
Bounce point occurs in magnitude-reconstructed IR images at intermediate time (Ti) when two substances sharing the same voxels cross the zero-axis, resulting in a black outline of the structures contained within these voxels.
Figure 3.
(A) Coronal reformat from sagittal 3D T1 MP-RAGE; squares represent voxels within GM, WM and the intermediate area between the two structures. (B) Coronal reformat from sagittal 3D T1 EDGE; 3D EDGE contrast shows WM/GM boundary as a negative signal.
For the SIGNA 7.0T MR system, we used the 3D BRAVO sequence as a baseline. Multiple scans were conducted on subjects across a wide age range, from young to elderly. The trials involved systematic adjustments to three key parameters:
- Flip angle (FA): Modified to optimize the overall SNR.
- Recovery time: Adjusted to enhance contrast.
- Inversion time (TI): Also referred to as preparation time (prep time), this parameter was fine-tuned to improve the bounce point (EDGE) effect.
Scan parameters were changed to match optimized results. Increasing FA from 6° to 10° gives a less grainy aspect for the image, well perceived on WM. Increasing the recovery time from 1.6 seconds to 2 seconds changed the contrast of both cerebral spinal fluid (CSF) and WM, with CSF being more bright and WM more dark (Figure 4). TI of 750 msec has shown good results for nulling the voxels containing similar volumes of GM and WM.
Figure 4.
(A) Axial 3D BRAVO, 0.8 x 0.8 x 0.8 mm, 6:54 min., TI=750 ms, recovery time=1,600 ms, FA=6°. (B) Axial 3D BRAVO, 0.8 x 0.8 x 0.8 mm, 7:51 min., TI=750 ms, recovery time=2,000 ms, FA=10°. (C) Zoomed portion from A and (D) zoomed portion from B; note how the SNR is better on WM and the contrast of the CSF is brighter on 4D. Images acquired on SIGNA 7.0T.
After a few more scans, we determined a more equilibrated contrast between GM and WM was needed to optimize the appearance of the black outline. By keeping FA at 10° and slightly reducing the TI to 740 msec, we set the recovery time to 1.5 seconds. These parameters showed consistent imaging results on subjects of different ages (Figure 5).
Figure 5.
Sagittal reformat from axial 3D EDGE, 0.8 x 0.8 x 0.8 mm, 6:14 min., TI=740 ms, recovery time=1500 ms and FA=10°. (A) Elder subject and (B) young subject. Images acquired on SIGNA 7.0T.
To ensure optimal image quality, AIR™ Recon DL was set to high, enabling acquisitions with high spatial resolution (0.8 x 0.8 x 0.8 mm). To maintain a reasonable scan time, the parallel imaging technique, ARC, was applied to accelerate acquisition in both the phase and slice directions (Figure 6).
Figure 6.
The user interface showing acquisition parameters for 3D EDGE on SIGNA 7.0T. Yellow stars show the most relevant parameters to set.
Figure 7.
The user interface showing acquisition parameters for 3D EDGE on SIGNA Premier 3.0T. Yellow stars show the most relevant parameters to set.
A similar approach was applied to both 3.0T and 1.5T systems. However, only the 3D T1 MP-RAGE pulse sequence was considered for generating EDGE contrast. The sequence parameters were inspired by published literature,¹ and trials were conducted at customer sites. The resulting images were reviewed by end users for final approval (Figure 8).
Figure 8.
(A) Coronal and (B) axial reformats from (C) sagittal 3D T1 EDGE, 1 x 1 x 1 mm, 5:30 min. Images acquired on SIGNA™ Premier 3.0T are courtesy of Centre Hospitalier Sainte-Anne, Paris, France.
The standard acceleration technique ARC was used for 3.0T in addition to HyperSense (Figure 7). Only ARC was applied for acceleration at 1.5T (Figure 9).
Figure 9.
The user interface showing acquisition parameters for 3D EDGE on SIGNA™ Artist 1.5T. Yellow stars show the most relevant parameters to set.
To support the adoption of 3D EDGE for brain imaging protocols, it was essential to optimize the sequence acquisition time. This required several modifications to the original acquisition settings.
The initial approach was focused on reducing scan time for 3.0T systems. Using the previously proposed acquisition (Figures 7 and 8) as a benchmark for the desired EDGE contrast, we carefully adjusted key parameters, including field of view, slab oversampling, matrix size,
receive bandwidth, flip angle, recovery time, inversion time, and ARC and HyperSense factors. This parameter tuning was the first critical step toward scan time optimization. Next, multiple scans (up to eight per subject) were performed by varying the TI. Signal values were then extracted from consistent regions of interest (ROIs) and placed on WM and GM (Figure 10). These values were plotted on fitted models to visualize the evolution of MR signal intensity as a function of TI. The optimal TI was identified as the point of intersection—or the closest approximation—between the WM and GM signal curves (Figure 11).1
Figure 10.
(A) Several EDGE acquisitions acquired by varying TI times, (B) ROIs placed on WM and GM. (C, D) Zoomed images of the two ROIs. Images acquired on SIGNA Premier 3.0T.
Figure 11.
Fitted models to visualize the evolution of MR signal intensity as a function of TI for GM and WM. (A) and (B) are two different subjects scanned on SIGNA Premier 3.0T.
During additional trials, a TI value of 508 ms consistently produced reliable EDGE contrast (Figures 12 and 13). However, the results are still pending final review and approval by end users.
Figure 12.
(A) Axial and (B) coronal reformats from the new proposed sagittal 3D T1 EDGE, 1 x 1 x 1 mm, 3:34 min. Images acquired on SIGNA Premier 3.0T.
Figure 13.
The user interface showing acquisition parameters for the new proposed 3D EDGE on SIGNA Premier 3.0T. Yellow stars show the most relevant parameters to set.
Sonic DL 3D
A new alternative to achieve shorter scan times for 3D EDGE is now available with the recent expansion of Sonic DL™— GE HealthCare’s innovative deep-learning (DL)-based acceleration technique—being applied to 3D acquisitions. Specifically, the examples originally scanned on the SIGNA 7.0T using ARC (Figures 4 to 6) demonstrated an impressive 42% reduction in scan time with Sonic DL (Figures 14 to 15). This advancement supports improved standardization and seamless integration of 3D EDGE into conventional brain imaging protocols.
Figure 14.
Axial 3D EDGE, 0.8 x 0.8 x 0.8 mm, TI=740 ms, recovery time=1500 ms, FA=10°. Same subject was scanned with (A) ARC factors of 2 x 2 in 5:53 min. and (B) Sonic DL factor of 4.5 in 3:26 min. Images acquired on SIGNA 7.0T.
Figure 15.
The user interface showing how to set acceleration factor for Sonic DL on SIGNA 7.0T.
Discussion
The 3D EDGE acquisition is a valuable complementary sequence for brain imaging protocols, particularly in the evaluation of epilepsy cases such as FCD. Enhanced by the advanced capabilities of AIR Recon DL and Sonic DL, this sequence delivers exceptional image quality with shorter scan times on GE HealthCare MR systems.
The authors acknowledge additional contributions from GE HealthCare employees:
Julie Poujol, PhD,
MR Clinical Scientist
Jean-Sébastien Louis, PhD,
MR Clinical Scientist
Alan Canoville,
MR Clinical Education Specialist
Adrien Launay,
MR Clinical Education Specialist
References
- Middlebrooks, Erik H., Chen Lin, Erin Westerhold, et al. 2020. “Improved detection of focal cortical dysplasia using a novel 3D imaging sequence: Edge-Enhancing Gradient Echo (3D-EDGE) MRI.” NeuroImage: Clinical 28:102449.
- Questions and Answers in MRI. 2025 “IR Bounce Point Artifact.” Accessed September 19, 2025: https://mriquestions.com/ir-bounce-point.html.
- Next
- →
