Contrast Agent-Specific Parameter Optimization for T₁-Weighted Fast Spoiled Gradient Echo Imaging

:

Objectives:

The objective of this study is to derive a contrast agent-specific theoretical framework to optimize acquisition parameters for contrast-enhanced magnetic resonance imaging (MRI) based on the physiochemical properties of gadolinium-based contrast agents, focusing on fast spoiled gradient recalled echo sequences. The goal is to enhance the lesion-to-background contrast for improved diagnostic sensitivity in clinical applications.

Materials and Methods:

Signal equations for fast spoiled gradient recalled echo sequences were derived for nonenhancing and enhancing tissues using gadoterate meglumine and gadobutrol, characterized by distinct longitudinal (r₁) and transverse (r₂/r*₂) relaxivities. Simulations were conducted at 2 field strengths, 1.5T and 3.0T, and various scenarios were considered, including hypothetical lesions with T₁ ratios ranging from 1.1 to 1.8. The signal behavior was analyzed across a range of initial conditions, including different spin densities and field-strength dependent variations in tissue relaxation times. The optimal flip angle and repetition time combinations were determined to maximize contrast. In vivo validation was performed on 2 patients undergoing contrast-enhanced MRI of the brain, using the proposed acquisition parameters.

Results:

The modeling and simulations revealed that the flip angle that maximizes signal intensity for a contrast-enhancing lesion (Ernst angle) differs from the flip angle that maximizes T₁-dependent contrast between lesion and healthy tissue in unenhanced MRI (Pelc angle) and also differs from the flip angle that maximizes the same in contrast-enhanced MRI. The theoretical simulations indicated possible contrast gains of 24%–28% using optimized parameters. The in vivo acquisitions demonstrated contrast gains of 19%–44% for a diffuse enhancing lesion and 91% for a weakly enhancing focal lesion, when comparing the optimized acquisition parameters to manufacturer's default settings.

Conclusions:

Adjusted repetition time and flip angle values, derived using the proposed framework, improved the image contrast between healthy and diseased tissues, enhancing the visualization of abnormalities. This approach can be used to optimize routine clinical MRI protocols and balance scan time with contrast enhancement. This may translate to more precise lesion detection, potentially leading to earlier and more accurate diagnosis or treatment monitoring in clinical practice.

The use of magnetic resonance (MR) contrast agents, which exploit MR's sensitivity to paramagnetic effects, has significantly broadened MR imaging (MRI) applications in clinical practice.^1–3 These agents, typically injected into the bloodstream, enhance image contrast, improving visualization of anatomical structures and pathological changes. Gadolinium-based contrast agents (GBCAs), with a central gadolinium ion bound by a chelating agent, are particularly effective in highlighting tissue abnormalities and pathologies.^4–6

However, the appropriate choice of acquisition parameters (and optimization thereof) for contrast-enhanced MRI (CE-MRI) sequences remains, even today, a hot topic of discussion among MR practitioners in the contrast media space. For instance, during the meeting of the Gadolinium Research & Education Committee working group at the European Society of Magnetic Resonance in Medicine and Biology conference in 2024,⁷ there was an entire session focused on optimizing contrast-enhanced MRI, including 3 presentations specifically on parameter optimization and protocol optimization for contrast-enhanced MRI.

While the issue of signal optimization/maximization for spoiled gradient recalled echo (SPGR) sequences has been solved since the 60s⁸ as a function of flip angle for a given repetition time (TR), and since the early 90s as a function of flip angle for T₁ contrast optimization in unenhanced MRI,^9,10 neither of these 2 cases correspond to the optimal flip angle to obtain maximum contrast for T₁-weighted sequences in the presence of paramagnetic contrast media such as GBCAs. The optimization of T₁-dependent contrast (in the absence of contrast agents) has been studied extensively in the past,^11–15 but these efforts have primarily focused on optimizing the contrast between unenhanced gray and white matter. While this is important for unenhanced T₁-weighted imaging, it is not of primary importance for contrast-enhanced T₁-weighted imaging. Additional challenges faced by current practitioners include the availability of multiple field strengths¹⁶ (which affect both relaxation times and relaxivities of contrast agents), diverse imaging sequences and their various implementations, and differing relaxivity profiles of contrast agents. Considering these multiple variables, we propose a theoretical framework, grounded in MR physics, to guide the choice of MR sequence parameters in the context of contrast-enhanced T₁-weighted imaging. The specific outcome measure we sought to optimize was the image contrast between enhancing and nonenhancing tissue.

To illustrate the approach, the signal equations for fast spoiled gradient recalled echo (FSPGR) were modeled and simulated for healthy (background) and diseased tissues in the proximity of a paramagnetic contrast agent, characterized by its longitudinal (r₁) and transverse (r₂/r*₂) relaxivities, while at the same time considering field-strength dependent variations in key modeling parameters such as background tissue relaxation times, and baseline lesion relaxation times. Non-IR prepped FSPGR sequences were considered throughout this work.

As our site specifically employs 2 contrast agents for intracranial contrast-enhanced MRI—gadoterate meglumine (DOTAREM®, Guerbet, France, and CLARISCAN™, GE HealthCare, Oslo, Norway) and gadobutrol (GADOVIST®, Bayer, US)—these 2 molecules were selected as use cases for the modeling and simulation that could then be evaluated in vivo.

This approach aims to identify feasible and easily implementable pulse sequence reparametrizations to improve lesion visualization in the clinical setting.

MATERIALS AND METHODS

Hardware and Software

All computations carried out for this work were performed with an AMD Ryzen 5 5500 CPU equipped with 32GB of RAM on Windows 11. The programming language used throughout was Python (v.3.10), and the integrated development environment utilized was JupyterLab. Libraries used included numpy,¹⁷ matplotlib,¹⁸ and scipy.¹⁹ The Python scripts for the simulations will be made available upon reasonable request to the corresponding author.

Relaxation and Relaxivity

The expressions that relate longitudinal and transverse relaxivities r₁ and r₂ of a contrast agent, with the longitudinal and transverse relaxation rates R₁ and R₂* of a tissue (under the influence of said contrast agent) are given by:

$$R_1=R_{1,0}+r_{1}C$$ (1a)

$$R_2^{\ast }=R_{2,0}^{\ast }+r_{2,0}^{\ast }C$$ (1b)

Here, C stands for concentration dose, and is expressed in units of mM; R_1,0 and R*_2,0 are the baseline longitudinal and transverse relaxation rates, respectively, in units of s⁻¹; and r₁ and r₂ are the longitudinal and transverse relaxivities of the contrast agent, respectively, in units mM⁻¹·s⁻¹.

SPGR Signal Equation

We start our analysis from the signal equation of gradient-echo sequences (considering the case of an ideally spoiled gradient recalled echo sequence), summarized below:

$$S=k·\frac{{\rho }_H\left(1-e^{-\mathit{TR}·R_1}\right)e^{-\mathit{TE}·R_2^{\ast }}}{1-cos{\alpha }_F·e^{-\mathit{TR}·R_1}}sin{\alpha }_F$$ (2)

Where k is a scaling constant, ρ_H is the spin density term, TR is the repetition time, TE is the time to echo, and α_F is the flip angle.

Combining equation (2) with equations (1a) and (1b), and then re-arranging yields:

$$S=k·\frac{{\rho }_H\left(1-e^{-\mathit{TR}·\left(R_{1,0}+r_{1}C\right)}\right)e^{-\mathit{TE}·\left(R_{2,0}^{\ast }+r_2^{\ast }C\right)}}{1-cos{\alpha }_F·e^{-\mathit{TR}·\left(R_{1,0}+r_{1}C\right)}}sin{\alpha }_F$$ (3)

This is the signal equation for the spoiled gradient echo sequence that will be used as a base to model signal behavior as a function of TE, TR, and flip angle (in addition to R_1,0, R*_2,0, r₁, r₂, and C).

Assuming that ideal RF pulses are used, and with effective spoiling being applied to destroy any residual transverse magnetization prior to the actual data acquisition step, and assuming that TR < < T₁, T₂, then the optimal flip angle that maximizes T₁ contrast between 2 tissues (in the absence of the paramagnetic effects induced by the use of contrast media) can be found to be^9,10:

$${\alpha }_d=\frac{2E_1-1}{2-E_1}$$ (4)

With E₁ being equal to e ^{−TR · R ₁} .

Similarly, we also have the canonical Ernst angle⁸ that maximizes signal intensity from a spin species (including lesions) characterized by some T₁ value, and with the imaging being acquired at some TR value:

$${\alpha }_E=arccos\left(E_1\right)$$ (5)

With E₁ being equal to e ^{−TR · R ₁} . Equations 4 and 5 were used for visualization purposes on the same plots along with the corresponding contrast and signal intensity curves against either flip angle or TR.

Modeling and Simulations

Two contrast agents were considered during the modeling and simulations presented in the current study; gadoterate meglumine and gadobutrol whose relaxivity profiles are given by Rohrer et al²⁰ for both 1.5T and 3.0T and are described in Table 1.

TABLE 1.

Contrast Agent Relaxivity Profiles at 2 Field Strengths (1.5T and 3.0T)

	Gadoterate Meglumine	Gadobutrol
r 1 /r 2 (1.5T)	3.6/4.3 mM−1·s−1	5.2/6.1 mM−1·s−1
r 1 /r 2 (3.0T)	3.5/4.9 mM−1·s−1	5.1/7.1 mM−1·s−1

The contrast curves were simulated according to Equations (2) and (3) for each contrast agent, and each field strength considered for this study, and as defined as the absolute value of the difference in signal intensity between the lesion and the background brain tissue:

$$\mathit{\text{Contrast}}=\mid S{I}_{\mathit{\text{lesion}}}-S{I}_{\mathit{\text{background}}}\mid$$ (6)

For example, at 3.0T, and considering gadobutrol, for the lesion we would have:

$$S{I}_{\mathit{\text{lesion}}}\left(3.0\mathrm{T}\right)=k·{\rho }_H\frac{sin{\alpha }_F\left(1-e^{-\left(R_{1,0}^{\text{lesion}}+r_{1,\mathrm{GB}}C\right)·\mathrm{TR}}\right)}{\left(1-cos{\alpha }_F·e^{-\mathrm{TR}·R_{1,0}^{\text{lesion}}+r_{1,\mathrm{GB}}C)}\right)}{e}^{-\mathrm{TE}·\left(R_{2,0}^{\ast \text{lesion}}+r_{2,\mathrm{GB}}^{\ast }C\right)}$$ (7)

And for the background tissue, there is no uptake of gadolinium, which means we would have:

$$S{I}_{\mathit{\text{background}}}\left(3.0\mathrm{T}\right)=k·{\rho }_H\frac{sin{\alpha }_F\left(1-e^{-R_{1,0}^{\mathrm{bgr}}·\mathrm{TR}}\right)}{\left(1-cos{\alpha }_F·e^{-R_{1,0}^{\mathrm{bgr}}·\mathrm{TR}}\right)}{e}^{-\mathrm{TE}·R_{2,0}^{\ast \mathrm{bgr}}}$$ (8)

The absolute value of the difference between equations 7 and 8 would represent the contrast between a contrast-enhancing lesion and the nearby apparently healthy white matter that does not uptake contrast agent.

The main objective then was to evaluate the optimal acquisition parameter combinations (flip angle and TR) that can maximize the contrast for each agent, and an acquisition protocol for each of the 2 contrast agents considered throughout this study (eg, gadoterate meglumine and gadobutrol) will be proposed and presented.

When performing contrast simulations for contrast-enhancing lesions, 5 scenarios were considered, 4 scenarios pertaining to 1.5T and 1 scenario pertaining to 3.0T, including hypothetical lesions with T₁ ratios (i.e., T₁ of the lesion divided by T₁ of the background tissue) ranging from 1.1 to 1.8, while also taking into account different lesion spin densities.²¹ The fact that the lesions were modeled as having from 10% to 80% increased longitudinal relaxation times implies that these would appear as mildly to moderately hypointense with respect to the apparently healthy background. Detailed simulation parameters can be found in Table 2.

TABLE 2.

Simulation Scenarios

	Scenario 1	Scenario 2	Scenario 3	Scenario 4	Scenario 5
T 1 /T 2 (L)	1500/90 ms	930/90 ms	1500/90 ms	1500/120 ms	1700/135 ms
T 1 /T 2 (bgr)	800/80 ms	800/80 ms	800/80 ms	800/80 ms	1150/90 ms
ρ H (L)	0.72	0.72	0.88	1.00	0.88
ρ H (bgr)	0.72	0.72	0.72	0.72	0.72
k	100	100	100	100	100
Dose	0.1 mmol/kg	0.1 mmol/kg	0.1 mmol/kg	0.1 mmol/kg	0.1 mmol/kg

ρ_H, spin density/proton density; bgr, background tissue; k, scaling term; L, lesion.

Three imaging protocols were considered for simulations, whose acquisition parameters can be found in Table 3.

TABLE 3.

Acquisition Parameters for the Simulation Scenarios at 1.5T

	Protocol 1	Protocol 2	Protocol 3
TR (ms)	8.6	13.6	13.6
TE (ms)	3.2	3.2	3.2
Flip angle ( o )	12	13	15
NEX	1	1	1
PES	256	256	256
Acceleration	0	0	0
Scan time (s)	132	208	208

Protocol 1 corresponds to the default acquisition protocol. Protocol 2 corresponds to the protocol optimized for use with gadoterate meglumine, and Protocol 3 corresponds to the protocol optimized for use with gadobutrol.

NEX, number of averages; PES, phase encoding steps; TE, echo time; TR, repetition time.

In Vivo Validation—Convenience Sample From Clinical Practice

Following the modeling and simulation results, the proposed acquisition parameters of TR/FA = 13.6 ms/13^o (at 1.5T) and 13.6 ms/14^o (at 3.0T), for use with gadoterate meglumine were acquired prospectively on 2 patients independently referred for contrast-enhanced MRI of the brain. The aim of this convenience sample was to gather preliminary insights from testing the theoretical optimization results on a small but representative sample.

Additional exclusion criteria were as follows:

• (1)Patient not providing informed consent for the additional sequence to be acquired at the end of the prescribed scan, which would last approximately 3 minutes of scan time.
• (2)Lack of a contrast-enhancing lesion on the scan acquired using default parameters (eg, TR/FA = 8.6 ms/12^o). The 3.0T scanner was a GE Signa Premier (GE HealthCare, Milwaukee, WI), with a 48-channel phased-array AIR™ coil, software version 29.0, and the 1.5T scanner was a GE Signa HD28 (GE HealthCare, Milwaukee, WI), with a 16-channel phased-array head coil, software version 28.0.

For both subjects, as per the exclusion criterion (1), the proposed sequence was acquired after the acquisition of the default sequence, in line with the requirements of our site, where the standard-of-care imaging is prioritized over exploratory research. Representative precontrast T₁w images are shown next to the Protocol 1 and Protocol 2 postcontrast T₁w images in Figure 1.

FIGURE 1.

Representative images of unenhanced (left column), default protocol (middle column) and proposed protocol (right column) of the 2 subjects imaged in vivo. In the top row, a blue arrow is pointing in a weakly enhancing lesion. The top 2 rows showcase different slices from the same subject (the one imaged at 3.0T).

RESULTS

Modeling and Simulations

The dependence of lesion-to-background image contrast on flip angle is illustrated in Figure 2 for 2 representative TRs (8.6 ms and 13.6 ms), 2 combinations of our predefined scenarios and field strengths (Scenario 1 at 1.5T and Scenario 2 at 3.0T), and the 2 contrast agents used as illustrative examples. These results show that the flip angle required for optimal lesion-to-background contrast is different to both the Ernst angle and the Pelc angle and depends on the relaxivity characteristics of the contrast agent used.

FIGURE 2.

Contrast versus flip angle curves from modeling equation (6) with gadoterate meglumine as a contrast medium (dotted blue curves) or gadobutrol (dotted green curves) as if acquired with Protocol 1 (A and C), or with Protocol 2, for 2 different combinations of baseline relaxation times for the lesions (displayed on the figures). The orange and red circles correspond to the optimal flip angles yielding maximum T₁-dependent contrast in the contrast-enhanced setting for gadoterate meglumine or gadobutrol, respectively. The vertical gray line corresponds to the lesion's Ernst angle, and the vertical purple line corresponds to the lesion's Pelc angle.

In Figure 3, we show the dependence of image contrast on flip angle for a wider range of TR values, from 3.6 ms to 33.6 ms, for both example contrast agents, using Scenario 1 conditions at both 1.5T and 3.0T. These results show that increasing the TR also increases the flip angle at which the optimal image contrast (between contrast-enhancing lesion and background) is obtained but, importantly, also increases the achievable image contrast itself. Interestingly, we found minimal differences in optimal combination of TR and flip angle between field strengths of 1.5T and 3.0T, implying a stronger role for intrinsic characteristics of tissue relaxation times—this is evidenced by the strong similarity (in terms of maximum achievable lesion-to-background contrast) between the corresponding top and bottom row panels in Figure 3.

FIGURE 3.

Contrast versus flip angle curves for multiple TR values for gadoterate meglumine (2A and 2C) and gadobutrol (2B and 2D). Top row corresponds to Scenario 1, and bottom row corresponds to Scenario 1 but for 3.0T relaxivities.

The graphs in Figure 4 illustrate the differences between the Ernst angle and our proposed optimal postcontrast flip angle, as a function of TR, for our 2 illustrative contrast agents. Subfigures A, B, and C correspond to Simulation Scenarios 1, 2, and 5, respectively. The key takeaway message of this figure is 2-fold. First, across the range of simulation scenarios, the contrast agent-specific difference between the optimal flip angle (proposed by this work) minus the corresponding Ernst angle for the lesion, increases nonlinearly with TR. Second, the difference between the agent-specific curves (e.g., difference between gadoterate meglumine and gadobutrol) also increases nonlinearly with TR. Because scenarios 3 and 4 differ from Scenarios 1 and 2 respectively only in terms of the assumed spin densities, their δα_F curves are identical (because the lesions would have an identical Ernst angle) and are therefore not shown.

FIGURE 4.

Delta flip angle plot (δα_F) versus TR plot for gadoterate meglumine (blue dotted curve) and gadobutrol (solid orange curve). Subfigures A, B, and C, correspond to the simulation scenarios 1, 2, and 5. The horizontal array of angles on top of the x-axes of each subfigure (eg, 4.7^o, 5.9^o, 6.9^o, 7.8^o, etc) correspond to the lesion's Ernst angle, at each TR value seen directly below (eg, 5, 10, etc). H, healthy tissue; L, lesion.

In Figure 1, we see a panel of 3 columns and 3 rows, where the left-most column corresponds to the precontrast T₁w image (an MPRAGE sequence for 3.0T, and an IR-SPGR sequence for 1.5T), while the middle column corresponds to the default standard parametrization (TR = 8.6 msec and FA = 12 deg), and finally the right column corresponds to the optimized acquisition parameters proposed (ie, TR = 13.6 msec with FA = 13 deg at 1.5T and FA = 14 deg at 3.0T, for acquisition with gadoterate meglumine).

Table 4 showcases the simulated contrast gains across the 3 different acquisition protocols, for each of the 5 simulation scenarios described in Section 2.4. Protocol 1, representing the default acquisition parameters, was treated as the benchmark (e.g., at 100%) across all 5 simulation scenarios. The range of contrast improvements across protocols and scenarios was from +24% to +28%.

TABLE 4.

Contrast Improvements of Contrast Agent-Specific Optimized Acquisition Protocols Compared to Default Acquisition Protocols

	Scenario 1	Scenario 2	Scenario 3	Scenario 4	Scenario 5
Protocol 2 (Gd-DOTA)	+28%	+28%	+28%	+28%	+28%
Protocol 2 ( Gd-BT-DO3A )	+24%	+24%	+24%	+24%	+24%
Protocol 3 (Gd-DOTA)	+26%	+26%	+26%	+26%	+27%
Protocol 3 ( Gd-BT-DO3A )	+26%	+26%	+26%	+26%	+25%

The terms in parentheses represent the contrast agent being used; Gd-DOTA implies gadoterate meglumine has been used for that Protocol, and Gd-BT-DO3A implies that gadobutrol has been used. Simulation scenarios 1–5 are described in Table 2. Protocol 2 corresponds to the default, acquisition setup, and as such, its corresponding contrast was set to 100%. Protocols 2 and 3 correspond to the gadoterate-optimized protocol (TR/FA = 13.6 ms/13^o) and gadobutrol-optimized protocol (TR/FA = 13.6 ms/15^o), respectively.

In Vivo Validation

Five individuals, across both field strengths, consented to the study. Of these, 3 were excluded due to no contrast-enhancing lesion being present in the default scan (ie, the standard-of-care scan with parameters TR = 8.6 msec and FA = 12 degrees).

Overall, the in vivo images acquired for this study (using gadoterate meglumine) are illustrated for the default and optimized acquisition parameters in Figures 1, and 5–8, with Figures 5–8 showcasing the ROI placements and the corresponding signal intensity and contrast values, and with Figure 1 showcasing side-by-side representative precontrast, default-enhanced, and optimized images.

FIGURE 5.

In vivo contrast-enhanced T₁-weighted 3D FSPGR images from an example subject at 3.0T using gadoterate meglumine with (top left pair of images) default acquisition parameters TR/TE = 8.6/3.2 ms and flip angle = 12^o, and (top right pair of images) proposed acquisition parameters TR/TE = 13.6/3.2 ms and flip angle = 14^o. On the bottom row of the figure, we see bar charts corresponding to the ROI signal intensities, and contrast estimations. F1/F2/F3/F4 stands for foreground (e.g., lesion ROIs) and B1/B2/B3/B4 stand for background ROIs (e.g., white matter).

FIGURE 8.

In vivo contrast-enhanced T₁-weighted 3D FSPGR image from a subject at 3.0T using gadoterate meglumine with (top left image) default acquisition parameters TR/TE = 8.6/3.2 ms and flip angle = 12^o, and (bottom left image) proposed acquisition parameters TR/TE = 13.6/3.2 ms and flip angle = 14^o. On the right side of the figure, we see bar charts corresponding to the ROI signal intensities and contrast estimations. F1/F2 correspond to the lesion ROIs, and B1/B2 correspond to the background ROIs.

To quantify the image contrast for extended areas of enhancement, we defined 2 ROIs in foreground (enhancing) tissue and 2 ROIs in background tissue (nonenhancing white matter) on each slice in Figure 5. The ROIs were delineated on the default parameter image and then propagated onto corresponding locations on the co-registered image of the optimized protocol. The image contrast between enhancing and nonenhancing tissue was increased 25%–44% for the 4 combinations of foreground and background ROI in the first slice (Fig. 5, images on the left sides of the figure, where the foreground ROIs were symbolized with F1 and F2), and 19%–36% for the second slice (Fig. 5, images on the right sides of the figure, where the foreground ROIs were symbolized with F3 and F4). The image contrast gain for a focal, more weakly enhancing lesion was quantified for the small lesion indicated in Figure 6 as 91% (i.e., the lesion corresponding to the F1 foreground ROI) compared to the default acquisition.

FIGURE 6.

In vivo contrast-enhanced T₁-weighted 3D FSPGR images from an example subject at 3.0T using gadoterate meglumine with (top left pair of images) default acquisition parameters TR/TE = 8.6/3.2 ms and flip angle = 12^o, and (top right pair of images) proposed acquisition parameters TR/TE = 13.6/3.2 ms and flip angle = 14^o. On the bottom row of the figure, we see bar charts corresponding to the ROI signal intensities, and contrast estimations. F1 and F2 stands for foreground (e.g., lesion ROIs) and B1/B2/B3/B4 stand for background ROIs (e.g., white matter).

Contrast-enhanced images for the subject scanned at 1.5T are illustrated in Figure 7 showing a low-grade meningioma (confirmed by MR spectroscopy and MR perfusion—not shown). One ROI was defined in the foreground (enhancing) tissue, and 2 ROIs in background tissue (nonenhancing white matter). The ROIs were delineated on the default parameter image and then propagated onto the corresponding locations on the co-registered image of the optimized protocol. The image contrast between enhancing and nonenhancing tissue was increased by >50% for both combinations of background-foreground ROIs.

FIGURE 7.

In vivo contrast-enhanced T₁-weighted 3D FSPGR images from an example subject at 1.5T using gadoterate meglumine with (top left image) default acquisition parameters TR/TE = 8.6/3.2 ms and flip angle = 12^o, and (bottom left image) proposed acquisition parameters TR/TE = 13.6/3.2 ms and flip angle = 13^o. On the right side of the figure, we see bar charts corresponding to the ROI signal intensities, and contrast estimations. F1 stands for foreground (e.g., lesion) and B1/B2 stand for background ROIs (e.g., white matter).

Finally, in Figure 8, we show an additional slice from the same subject depicted in Figures 5 and 6, with the lesion appearing as ring-enhancing with a necrotic core. Two ROIs were used for the lesion foreground (i.e., F1 and F2), and 2 more ROIs were used for the background, where the ROIs were delineated on the default image (ie, Protocol 1) and then propagated to the proposed image. The image contrast between lesion and background was increased by at least +21%, and up to +89%, when comparing default parameters with proposed parameters.

DISCUSSION

The theoretical framework and results presented here show that the flip angle that maximizes image contrast between GBCA-enhancing and nonenhancing tissues in T₁ contrast-enhanced MRI is different than both the Ernst angle (that maximizes the signal intensity of a contrast-enhancing lesion), as previously shown,^9,10,12 and the Pelc angle⁹ (that maximizes T₁-dependent contrast between lesion and background healthy tissue in unenhanced MRI). The difference δα_F between the Ernst angle and optimal flip angle is a nonlinear function of TR (confirming prior results^12–14), and the difference between δα_F for the 2 contrast agents also increases nonlinearly with increasing TR. This framework allows SPGR acquisition parameters to be optimized for a specific contrast agent, improving enhanced versus unenhanced tissue image contrast. Even though a detailed and systematic literature review regarding contrast agent optimization was beyond the scope of the work presented herein, to our knowledge, this is the only study that attempts to propose an acquisition protocol that can be tailored to different types of contrast agents, and which increases the lesion-to-background tissue contrast using standard sequences and standard dosing.

For indicative tissue characteristic values (simulation scenarios), our framework predicted a theoretical boost of the order of 24%–28% in absolute contrast between contrast-enhancing lesion and nonenhancing background tissue. The in vivo results confirmed these image contrast gains using optimized sequence parameters, yielding increases of 19%–44% for diffuse enhancing tissue using gadoterate meglumine at 3.0T, and an even greater increase of 91% (i.e., approximately double) for a more focal, weakly enhancing lesion in the same subject.

For the in vivo case scanned at 1.5T with gadoterate meglumine (patient with a low-grade meningioma), similar trends were observed. The overall contrast increases that were evaluated on a slice-by-slice level, ranged between 11.7% and ~53% throughout the entire lesion, with an average increase of the order of 29%. This is not surprising, as it is known that lesions (and healthy tissues, too) do not always exhibit homogeneous native relaxation times nor is the blood-brain-barrier breakdown identical throughout the different lesion locations. These contrast gains could be particularly important for small and weekly enhancing lesions in clinical settings including preoperative treatment planning. Overall, across all foreground/background combinations in both subjects, in all slices, the average lesion-to-background contrast with the proposed parametrization was increased by +44%, compared to the default parametrization.

However, the proposed optimized parameter settings come at the expense of some additional scan time. Scan time depends on the number of phase encoding steps (PES), and the TR, assuming the number of averages (NEX) remains unchanged and set equal to 1. Specifically, the scan time is given by NEX×TR×PES. As a worked example, considering PES = 256 and TR = 8.6 msec (assuming no acceleration, and typical parameters), which yield a scan time of 2 minutes and 12 seconds. Changing the TR from 8.6 msec to 13.6 msec increases the scan time to 3 minutes and 28 seconds. One means of counteracting the scan time increase could be to use an acceleration factor; for example, a value of 0.5 would reduce the scan time to around 3 minutes (assuming slice thickness is unchanged), resulting in an overall decrease in SNR of around 10%–15% because of that.

More importantly, and in addition to the application of the proposed reparametrization, the modeling and simulation results (Figs. 2–4) can be used directly by practitioners to benchmark their current acquisition parameters. In our case and for our site, the objective (and motivator for this study) was to increase the contrast enhancement for all CE-MR scans of the brain, even at some expense of scan time. One other hypothetical site may (for example) already be using a larger TR than the one typically recommended by the manufacturers. We showcase, however, that a larger TR in itself does not necessarily guarantee an optimal increase in contrast enhancement unless it is also accompanied by the corresponding optimal flip angle (which is TR-dependent). Thus, if a site is acquiring its 3D FSPGR sequence for CE-MRI of the brain with acquisition parameters TR = 10 msec, and FA = 20 deg (same parameters for both agents), then the site can gain approximately +14.5% lesion-to-background contrast by simply reducing the FA from 20 deg to 10 deg (which is the optimal for acquisition with gadoterate meglumine at TR = 10 msec), or +12.5% by reducing the FA from 20 deg to 13 deg (which is the optimal for acquisition with gadobutrol at TR = 10 msec). Therefore, as a direct result of our theoretical framework, contrast gains can be achieved even without having to increase the TR (and thus, the scan time).

Conveniently, utilization of deep learning image reconstruction software²² (e.g., AIR Recon DL) may find useful applications in this domain too. By enabling the practitioner to acquire images with fractional NEX (e.g., 0.75), the scan time drops to around the same order of magnitude as if phase acceleration was turned on, without suffering from the SNR decrease associated with the acceleration and compensating for the SNR decrease from the fractional NEX. Nevertheless, even if some SNR is lost because of acceleration, this is not expected to influence the lesion-to-healthy tissue contrast (assuming that the overall SNR is still clinically sufficient), so from our point of view, the improvement in contrast may be worth the time investment even in busy clinics and diagnostic centers.

Specifically, simply increasing the NEX does not necessarily imply lesion-to-background contrast gains, unless the image is already very noisy. For example, and considering Protocol 1—the default protocol, if NEX was increased from 1 to 2, and assuming all other parameters were unchanged, the scan time would double, from 132 seconds to 264 seconds. From a signal point of view, the SNR would scale up by √NEX, which in this case, is ~41%. However, both the lesion, and the background, would benefit from these gains, and therefore even though the overall image quality would be significantly improved, and even though the scan time would have increased, this would not necessarily guarantee an increase in lesion-to-background contrast. To that extent, and specifically for these cases where the reporting radiologist wants to de facto increase the lesion-to-background contrast, our proposed reparametrization would achieve exactly that. Usually in contrast agent research, a particular sequence parametrization is selected based on best practices and consensus recommendations^23,24 at the time of study setup, and contrast agents can then be compared based on qualitative and semiquantitative visual assessment under identical acquisition conditions. To our knowledge, there have been no published studies addressing the topic of acquisition parameter optimization in the presence of GBCAs from the point of view that has been described herein.

Furthermore, while our results enable image acquisition to be optimized for a specific GBCA, a site may use more than 1 contrast agent and prefer to maintain a constant acquisition protocol independent of contrast agent used. In this case, our results suggest that for both field strengths considered, and regardless of contrast agent used and baseline relaxation times of the simulated tissues, the Pelc angle is closer (than the Ernst angle) to the optimal flip angle that maximizes T₁-dependent contrast for each 1 of the 2 contrast agents considered in this study. The Pelc angle may be a good compromise between the optimal flip angles for gadoterate meglumine and gadobutrol and for tissues with large baseline δT₁. Specifically, increasing the TR from 8.6 msec (which is the value typically proposed by manufacturers) to 13.6 msec, and changing the flip angle from 12^o (which is the value typically proposed by scanner manufacturers) to the corresponding Pelc angle, yields nearly optimal contrast between lesion and healthy tissue, across a wide range of baseline conditions. This parameter adjustment by itself should increase the contrast by 15%–20%, compared to what would have been obtained with the default TR/TE/FA = 8.6/3.2/12^o, regardless of which contrast agent used.

This study has limitations. First and foremost, studying the impact of these parameter adjustments across a broader range of contrast agents with varying relaxivity profiles would provide further insight into the generalizability of the proposed parameter adjustment. Second, only fast spoiled gradient echo sequences were considered, which do not have an inversion recovery module to prepare the magnetization (eg, BRAVO, MPRAGE, MP2RAGE, etc), neither 3D fast spin echo sequences of the saturation recovery style were considered. These sequences have gained popularity in recent years and are being used more frequently for CE-MR. Third, our in vivo assessment was preliminary and did not comprise a formal, prospective assessment of image contrast gains or qualitative radiologist assessment of the images. Instead, it was merely a short convenience study that was meant to provide some rough and preliminary evaluation of the theoretical results produced through the modeling and simulations.

Furthermore, the in vivo examples shown here suffer from an intrinsic limitation, due to the fact that the proposed sequence was acquired (in both cases) second to the default sequence. Given that the default sequence was roughly 2 minutes long, and the proposed sequence is roughly 3 minutes long, the time duration between start of default scan and end of proposed scan was approximately 5 minutes. This means that the proposed scan also benefitted from a (roughly) 5-minute delayed scan, which, according to the literature,²⁵ may also have contributed to the overall contrast gains realized. Future follow-up work should focus on a larger-scale prospective cross-vendor study to further assess repeatability and reproducibility of results obtained and presented herein. Additionally, both modeling & simulation, and in vivo components of the study presented herein, were focused exclusively on CE-MRI of the brain, and therefore future research should further extend the framework reported here to cover additional pulse sequences, contrast agents, and anatomies other than brain, and specifically cardiac and liver imaging.²⁶

Interestingly, the framework presented herein may be extended to provide modeling insights into optimal imaging window timings in the case of delayed enhancement scans. For instance, and still in the regime of MR neuroimaging, delayed T₂ FLAIR images are often acquired with a delay of usually around 15 minutes for the determination and characterization of active multiple sclerosis (MS) plaques, following the injection of the contrast medium.²⁷ In order to attempt to determine the optimal such window of delayed enhancement imaging with a FLAIR sequence, first thing that needs to be done is to redefine the signal equation from FSPGR to FLAIR. In our current methodological implementation, the time evolution of the target tissue concentration (and consequently the time evolution of the tissues' postcontrast relaxation times) was treated as a constant rather than a variable, this would need to be made explicit in the case where optimal timing parameters must be determined (like in the case of optimizing the delayed enhancement window). Therefore, instead of a standard/static amount of concentration C (corresponding to the peak concentration), its time dependence (i.e., as a function of elapsed time since administration) should be made explicit and be therefore modeled as C(t), implying consequently that the relaxation rates would be given by R₁(t) = R_1,0+r₁ × C(t), and where the C(t) will depend on several physiological and pharmacokinetic properties and parameters that are routinely evaluated through dynamic perfusion studies.

By obtaining the experimental values of these parameters from the literature,²⁸ we can then “backpropagate” them into the simulation models. Rather than repeating simulations with varying acquisition parameters and comparing the simulated contrast across different acquisition protocols (as was done in the current study), we could instead maintain constant acquisition parameters while varying the values of C(t)—corresponding to different time points after administration of each contrast agent. That way we can identify the optimal imaging windows specific to each contrast agent (and consequently assess the differences, if any, and plan accordingly).

Similarly, this method can be applied within the same theoretical framework to investigate optimal imaging timing windows in anatomical regions other than the brain, such as evaluating delayed myocardial enhancement typically performed with a phase-sensitive inversion recovery sequence.²⁹

Finally, the authors envision the methodological framework presented herein to find potential application in more advanced neuroimaging techniques such as neurofluid imaging,^30,31 which encompasses the identification of age-related differences in cerebrospinal fluid and interstitial fluid dynamics.

CONCLUSIONS

We have derived a theoretical framework to boost post-GBCA image contrast between contrast-enhancing and noncontrast-enhancing tissue for specific contrast agents via simple parameter adjustments in the TR and flip angle of a 3D T₁w FSPGR sequences. Modeling with typical parameters for brain imaging, and using gadoterate meglumine and gadobutrol as illustrative GBCAs, predicts an image increase of approximately 28%, at the expense of 76 seconds additional scan time. This excess scan time may easily be compensated through acceleration techniques, or a combination of deep learning denoising, fractional NEX, and phase acceleration. In vivo results (albeit limited) confirmed the theoretical lesion-to-background contrast gains. This sequence reparametrization may find immediate application in the detection of weakly enhancing intracranial lesions. Alternatively, our framework also informs “compromise” parameterizations, yielding similar imaging performance between 2 different GBCAs, despite their different relaxivity profiles, allowing comparable contrast between lesion and background tissue regardless of which 1 of the 2 contrast agents was utilized.

ORCID ID

Dimosthenis E. Gkotsis https://orcid.org/0000-0003-2813-0547