In-vitro Gd-DTPA Relaxometry Studies in Oxygenated Venous Human Blood and Aqueous Solution at 3 and 7T

Abstract

In-vitro T₁ and T₂^* relaxivities (r₁ and r₂^*) of Gd-DTPA (GaD) in oxygenated human venous blood (OVB) and aqueous solution (AS) at 3T and 7T were calculated. GaD concentrations ([GaD]) in OVB and AS were prepared in the range 0–5 mM. All measurements were acquired at 37±2 °C. At both 3T and 7T, a linear relationship was observed between [GaD] and R₁ in both AS and OVB. At 7T, r₁ in AS decreased by 7.5% (p = 0.045) while there was a negligible change in OVB. With respect to R₂^*, a linear relationship with [GaD] was only observed in AS, while a more complex relationship was observed in OVB; quadratic below and linear above 2 mM at both field strengths. There was a significant increase of over four-fold in r₂^* with GaD in OVB at 7T (for [GaD] above 2mM, p ≪0.01) as compared to 3T. Furthermore, in comparison to r₁, r₂^* in AS was less than two-fold higher at both field strengths while in OVB it was ~twenty-fold and ~ninety-fold higher at 3T and 7T, respectively. This observation emphasizes the importance of r₂^* knowledge at high magnetic fields, ≥3T. The comparison between r₁ and r₂^* presented in this work is crucial in the design and optimization of high field MRI studies making use of paramagnetic contrast agents. This is especially true in multiple compartment systems such as blood where r₂^* dramatically increases while r₁ remains relatively constant with increasing magnetic field strength.

INTRODUCTION

The paramagnetic effectiveness of a contrast agent (CA) is measured by its relaxivity which is defined as the rate by which the relaxation rate changes per unit molar CA concentration. Typically the effect of clinically used paramagnetic gadolinium based contrast agents on the longitudinal relaxation rate constant R₁≡1/T₁ is of most interest. However, understanding CA based changes in R₂^*≡1/T₂^*, a function of both spin-spin interactions and local magnetic field inhomogeneities, is becoming more important as clinical and research endeavors expand into ever increasing magnetic fields. The increasing R₂^* effect with gadolinium based CAs at higher fields compete with the T₁-weighted signal increase relied upon in post-contrast anatomic imaging, contrast enhanced angiography studies and dynamic contrast enhanced MRI (DCE-MRI). Increasing concentrations of CA result in higher R₁ values and increased signal on T₁-weighted acquisitions while also producing higher R₂^* which disproportionately decreases signal at higher concentrations. Based on the results determined in this study, sub-millisecond T₂^* relaxation times at 7T are possible with Gd-DTPA concentrations ([GaD]) of 5 mM in the blood. Such concentrations can easily be reached in-vivo when high injection rates of 3 ml/s are used (1) such as in DCE-MRI studies of the prostate. Previous studies have shown that, without correcting for T₂^* signal attenuation, subject dependent determination of the arterial input function was impossible for such studies (2). Similar effects are expected to negatively impact other contrast enhanced studies when performed at 7T such as first pass perfusion and angiography exams relying on T₁-weighted enhancement (3). Understanding the field dependent relationship between the two relaxation rate constants with respect to CA concentration in blood can be undertaken by defining both R₁ and R₂^* relaxivities, r₁ and r₂^* respectively, and will be necessary for understanding the tradeoffs of using gadolinium based CAs at high fields and in the optimization of both acquisition parameters and injection paradigms. While there is a wealth of information investigating the relaxivity of gadolinium based CA in the literature, there are also an equal number of varying acquisition methods and experimental conditions which can greatly vary the results. Therefore, the goal of this work was to use consistent experimental methods at both 3T and 7T to determine r₁ and r₂^*. For the paramagnetic CA, the relaxivity is a function of the motional correlation times, which includes several dynamic components characterizing electron–proton dipolar coupling (4) and involving the magnetic moment of the metal ion. As the solvent of the CA has a tremendous effect on the relaxation properties, we investigated the relaxivity in both an aqueous solution (AS) and in fully oxygenated venous blood (OVB), both at physiological temperatures of 37±2 °C. Because of its prominence in the literature and clinically use in our institution and elsewhere, Gadolinium-DTPA (Magnevist™, Bayer, Germany) (GaD) was the CA used in this study.

THEORY

The addition of a paramagnetic CA like GaD to a solution increases both relaxation rate constants R₁ and R₂. The observed R_1,2(obs) after addition of GaD is the sum of the native relaxation rate of the solution R_1,2(N) plus any contribution from GaD defined as R_1,2(GaD). Thus R_1,2(obs) = R_1,2(N) + R_1,2(GaD) (assuming independence of the relaxation pathways). In dilute solutions of GaD the observed relaxation rate constant R₁ is assumed to be linearly dependent on [GaD]. The slope of the dependence is the relaxivity and the y-intercept is the native relaxation rate of the sample prior to the addition of GaD i.e. R_1,2(obs) = r_1,2(N)×[GaD] + R_1,2(N), where r_1,2(N) is the relaxivity defined as the incremental increase in the relaxation rate per unit concentration of GaD. The apparent transverse relaxation rate constant R₂^* is defined as R₂^* = R₂ + R₂’, where R₂^* is the observed transverse relaxation rate. R₂ is the spin–spin relaxation rate constant and characterizes fluctuating magnetic field inhomogeneity effects, and R₂’ (R₂’≡1/T₂’) is that induced by the local magnetic field inhomogeneities within the voxel (5).

MATERIALS AND METHODS

Commercially available formulation of GaD, having a concentration of 0.5 M, was mixed with OVB and AS to obtain final concentrations in the range 0–5.18 mM(OVB) and 0–4.83 mM(AS), respectively.

OVB Sample Preparation

Venous human blood for this study was collected in a Vacuette heparin vacutainer (Greiner, Monroe, NC) through venipuncture using a standard 23-gauge butterfly needle from a healthy donor. A high gauge needle was used to minimize hemolysis during blood collection. The tube was then gently inverted six times to thoroughly mix the blood in the vacutainer with heparin. This was done to avoid blood clot formation which would have rendered the specimen unacceptable for use. The blood in each vacutainer was then collected in a 1L side arm conical flask by pouring it slowly along the wall. The blood was then exposed to moist O₂ and air using an in-house oxygenation apparatus. In this setup, dry O₂ from the cylinder was passed through a conical flask containing distilled water (Kandiyohi Bottled Water Co., Willmar, MN). The resulting moist O₂ was then passed through a pressure gauge before interacting with the venous blood in the conical flask. The moist O₂ pressure was maintained at 5–10 SLPM (standard liters per minute) during oxygenation. The conical flask was stirred gently for 1.5 hours until the sO₂ reached 100%. An iStat (Abbott Point of Care Inc., Princeton, NJ) portable clinical analyzer was used to measure hematocrit (Hct), sO₂ and pO₂. A blood gas analyzer (Rapidlab-248, Siemens, Deerfield, IL) was used to measure pO₂. The final pO₂ (293.8 mm Hg) was much higher than that of arterial blood (95 mm Hg)(6).

A pO₂ decrease in plastic tube stored blood samples due to metabolism has been reported (7). This drop in pO₂ can be attributed to oxygen consumption by leucocytes in blood and diffusion of oxygen through the walls of the plastic tube. While T₁ is known to have a slight linear dependency on pO₂ (6), a high pO₂ (hyperbaric O₂) was chosen to ensure that the blood sO₂ stayed above arterial blood sO₂ levels (>94%) through the duration of the MRI measurements.

GaD Solution in AS and OVB Preparation

A 0.1 M HEPES buffer solution was used to maintain the pH of distilled water at 7. GaD samples in AS were made by adding [GaD] to this buffer solution. OVB obtained earlier was used as the solvent for making the GaD arterial blood solutions. A Corning (Corning Inc., Corning, NY) 15mL clear polypropylene centrifuge tube was placed on a Mettler AM100 electronic balance (Mettler-Toledo, Inc., Columbus, OH). Microscopic volumes of GaD were added to the tube using a micropipette and the weight was recorded. This was followed by adding OVB or 0.1M (pH-7) HEPES solution to the tube and the final weight of the solution was recorded. The concentration accuracy of the GaD samples was calculated based on the weight measurements. Using the recorded weights and densities of GaD, human blood and distilled water at 22°C the final [GaD] obtained were, GaD in OVB: 0.00, 0.58, 1.12, 1.79, 1.98, 2.36, 2.88, 3.70, 3.90, 4.38, 5.18 mM. GaD in AS: 0.0, 0.43, 1.00, 1.48, 1.93, 2.42, 3.05, 3.42, 3.90, 4.42, 4.83 mM.

Phantom Setup

For both OVB and AS studies a plastic holder was used to house the 5 ml polystyrene round bottom test tubes containing the various [GaD] solutions. The holder was placed in a perforated container and surrounded by water bath tubing. A larger container then housed the whole setup and was filled with 0.45% saline solution to facilitate B₀ shimming over the phantom, to load the transmit coil and provide efficient heat transfer from the water bath hose. All studies were performed at a temperature of 37±2 °C. This temperature was maintained by pumping heated water through the water bath tubing (Thermo Scientific Neslab RTE- 7 Digital One). MR images of the phantom setup are shown in Figure 1. The temperature was monitored using a thermistor (YSI 400 series) positioned in the water bath but away from the sample tubes and a DigiSense (Cole-Parmer, Chicago, IL) Temperature Controller. Blood oxygenation (sO₂, pO₂) and Hct were measured before and after each set of T₁, T₂^* experiments. Each blood phantom sample was gently turned upside down three times to ensure mixing before placement inside the scanner.

Figure 1.

a) Axial and b) sagittal scout images of the phantom setup. Legend: 1- Water bath tubing. 2- Outer plastic container. 3- 0.45% Saline water. 4- Vial. 5- Vial holder. 6-Inner plastic container. Coronal view of the OVB phantom from the first echo of the multiecho acquisition used for calculating R₂^* for (c) 3T and (d) 7T. Coronal view of the OVB phantom from the TI = 20 ms, IR-TFE acquisition for (e) 3T and (f) 7T. The location of circular ROIs used for data analysis within each acquisition for the zero GaD vial are shown in (c) through (f).

MR Instrumentation

All 3T measurements were performed on a Siemens Magnetom Trio scanner. Signal reception was achieved using a 12 channel head coil. RF transmission was performed using the scanner body coil. All 7T measurements were done on a Magnex 7T, 90 cm bore magnet with Siemens console and head gradients using a 16 channel transceive head coil.

T₁ Measurement

Multiple inversion recovery turbo flash (8) (IR-TFL) acquisitions were used to measure T₁ by varying the inversion delay. The inversion pre-pulse was accomplished with a non-selective adiabatic inversion pulse. Imaging parameters for the IR-TFL sequence include: TR, 10 s; TE, 1.33 ms (3T), 1.05 ms (7T); nominal flip angle, 6°; acquisition matrix, 128²; field of view, 190 mm; slice thickness, 3.5–5.0 mm; number of excitations NEX= 2; bandwidth per pixel, 1000 Hz (3T), 1395 Hz (7T). The inversion time (TI) was varied from 20 to 9000 ms (20–30 TIs). A centrically ordered phase encode ordering allowed the short TI times to be achieved at the expense of spatial blurring especially for samples with higher [GaD]. Examples of the TI=20 ms for both 3T and 7T are shown in Figure 1e and 1f, respectively. The T₁ measurement using the IR-TFL method was validated by using an inversion recovery spin echo acquisition with the AS phantom at 3T (IR-SE; TR, 3 s; TE, 11 ms; flip angle, 90°; acquisition matrix, 256²; field of view, 300 mm²; slice thickness, 5mm; NEX= 1; TI, 23 ms).

The average MR signal intensity (SI) and standard deviation for each region of interest (ROI) within each vial were obtained from the TFL images well within the vials to avoid partial volume effects with the wall of the container (Figure 1e,f). For the inversion recovery sequence, the data was reconstructed in the real mode to allow negative signal intensities for fitting the T₁ recovery curve. T₂^* effects resulting from global field inhomogeneities were minimized by performing B₀ shimming within a localized coronal slab which included all the vials used in the analysis. The T₁ values were obtained by performing an instrumental fit on the ROI SIs versus their respective TIs. The equation used to fit the data was SI(ROI) = A[0] × (1–2×exp(−A[1]×TI)) + A[2], where A[0], A[1] and A[2] are fit parameters such that T₁ = 1/A[1]. The equation was fit to the data using the LMFit function within IDL (ITT, Visual Information Solutions, Boulder, CO) which employs a Levenberg-Marquardt algorithm to find a solution to the fitted parameters. A linear relationship between R₁ and [GaD] was assumed to calculate r₁. The equation used to fit the data was R₁= r₁ · [GaD] + Intercept where r₁ is the slope. The R₁ vs. [GaD] data was fit using a linear instrumental fit function in Origin 8.1(OriginLab Corp., Northampton, MA).

T₂^* Measurement

Multi-echo gradient echo acquisitions were performed to acquire the signal intensity data used to calculate T₂^*. Solvent and field strength dependent acquisition parameters are given in Table 1. Varying TE ranges and the gaps between successive echo times (ΔTE) were chosen to measure the relaxation characteristics of each sample. At 3T, measurements on both AS and OVB were acquired with two concatenated acquisitions while for OVB at 7T, 3 interleaved acquisitions were acquired with a minimum effective ΔTE (ΔTE_eff) of 0.81 ms to capture the potentially rapidly decaying signals for high [GaD]. Acquisition parameters common to all acquisitions included a 10 degree flip angle and 280 mm field of view. Images from the first TE of the multi-echo gradient echo acquisition and representative ROIs used for analysis are shown from the OVB phantom for both field strengths in Figure 1c and 1d, respectively. T₂^* values were obtained by performing a monoexponential instrumental fit of the ROI SIs versus their respective TEs using the function: SI(ROI) = A[0]×(exp(−A[1]×TE))+A[2], where A[0], A[1] and A[2] are fit parameters and A[1] = R₂^*. The optimization routine LMFIT was used within IDL to fit the equation to the SI amplitudes. Parameter A[2] was set by identifying the SI mean magnitude of the background noise. The mean noise was also used as a threshold to limit the range of included echo times for R₂* estimation.

Table 1.

Acquisition parameters for R₂^* calculations.

B0	Sample	# Acquisitions	# Echoes per Acquisition	# Echoes Total	Minimum TE (ms)	ΔTE per Acquisition (ms)	Minimum ΔEeff (ms)*	TE Range (ms)	Matrix Size
3T	AS	2†	12	24	1.80	2.38	2.38	1.8–56.54	256
3T	OVB	2†	12	24	1.80	2.38	2.38	1.8–56.54	256
7T	AS	1	12	12	2.01	2.80	2.80	2.01–32.81	320
7T	OVB	3‡	4	28	2.36	2.44	0.81	2.36–11.29	320

^†Acquisitions were concatenated.

^‡Acquisitions were interleaved to produce a minimum effective ΔTE shorter than the ΔTE per acquisition.

^*The ΔTE after combining the data from all acquisitions.

To determine r₂^* relaxivity, the R₂^* versus [GaD] data was characterized by both a quadratic and a combination of both quadratic and linear functions. First, the full range of [GaD] was fit solely with the quadratic function (R₂^* = A×[GaD]₂ + B×[GaD] + C). Second, only lower [GaD] concentration were fit with the quadratic function while higher concentrations were fit with a linear relationship (R₂^* = A×[GaD] + B). The R₂^* vs. [GaD] data was fit using the instrumental fit functions in Origin 8.1.

The errors reported take into consideration the error propagation from volumetric analysis, nonlinear fit (T₁ and T₂^* calculation) and linear/nonlinear fits (r₁ and r₂^* calculation). The standard errors of the fitting parameters were calculated from the square root of a main diagonal value of the variance-covariance matrix.

RESULTS

The native R₁ and R₂^* values for OVB and AS have been summarized in Table 2. The curve fits used to obtain r₁ and r₂^* relaxivities for [GaD] in AS and OVB are shown in Figure 2. The r₁ relaxivities for GaD in AS and OVB are shown in Table 3. For an increase in field strength from 3T to 7T there was a small (7.5 %), but statistically significant (p = 0.045), decrease of GaD r₁ in AS with no observable change in OVB r₁ with an increase in field strength.

Table 2.

R₁ and R₂^* values for Oxygenated Venous Blood (OVB) and Aqueous Solution (AS) at 3T and 7T at 37 °C.

Sample	R1 (sec−1)	R2* (sec−1)
3T OVB	0.56±0.00	9.94±0.56
7T OVB	0.49±0.00	28.50±1.15
3T AS	0.24±0.00	1.38±0.81
7T AS	0.24±0.00	0.94±1.03

Figure 2.

Plots of R₁ and R₂^* vs. [GaD] for Oxygenated Venous Blood (OVB) and Aqueous Solution (AS) at 3 and 7T. Plots of relaxivity modeled with linear (red) and quadratic (blue) functions are shown. A linear model was used to calculate R₁ relaxivity (r₁) for 3T and 7T in both AS (a,b) and OVB (c,d). A linear model was also used to calculate R₂^* relaxivity (r₂^*) in AS (e,f). To determine r₂^* in OVB, a quadratic function for [GaD] ≤ 2 mM and linear function for [GaD] ≥ 2 mM was found to characterize the data better than a quadratic fit alone over the whole range of [GaD] (g,h).

Table 3.

T₁ relaxivities (r₁) of Gd-DTPA in Oxygenated Venous Blood (OVB) and Aqueous Solution (AS) at 37 °C for both 3 and 7T along with respective pO₂, sO₂ and Hematocrit (Hct) ranges used for the measurements.

B0	OVB				AS
	pO2	sO2	Hct	r1	r1
T	mm Hg	%	% PCV	mM−1s−1	mM−1s−1
3	114.5– 166.85	97–98	44	3.20±0.06	3.44±0.08
7	293.8	100	44	3.20±0.06	3.18±0.09

The relationship between R₂^* and [GaD] for OVB and AS at 3T and 7T is shown in Table 4. For an increase in field strength from 3T to 7T there was a 3 percent (p = 0.2) increase in GaD r₂^* relaxivity in AS (Table 4). In blood, the data was best described by a combined quadratic-linear relationship where a quadratic function better described the experimental data up to 1.98 mM while a linear function best described the R₂^* dependence on higher [GaD]. The most significant change in r₂^* was observed in OVB which increased over four-fold (for [GaD] beyond 2mM, p ≪0.01).

Table 4.

Relationship of Gd-DTPA concentration [GaD] with R₂^* in Aqueous Solution (AS) and Oxygenated Venous Blood (OVB) at 3T and 7T.

B0	AS				OVB
	[GaD] : 0 to 4.83 mM				[GaD] : 0 to 1.98 mM						[GaD] : 1.98 to 5.18 mM
	R2* = A×[GaD] + B				R2* = A×[GaD]2 + B×[GaD] + C						R2* = A×[GaD] + B
	A	Erræ	B	Erræ	A†	Erræ	B§	Erræ	C‡	Erræ	A⌂	Erræ	B	Erræ
T	sec−1mM−1		sec−1		sec−1mM−2		sec−1mM−1		sec−1		sec−1mM−1		sec−1
3 7	4.21 4.34	0.04 0.10	1.21 1.40	0.11 0.29	5.76 42.94	4.77 5.38	26.61 40.88	8.08 7.02	8.99 28.25	2.60 1.84	68.90 290.81	6.85 14.63	−41.97 −270.20	15.35 33.62

^†Coefficient “A” controls the speed of increase of R₂^* from the parabola vertex.

^æFit Error.

^§B together with A controls the axis of symmetry of the parabola which is at [GaD] = −B/2A.

^‡Coefficient ‘C” gives the R₂^* value of OVB (i.e. [GaD] = 0).

^⌂Slope gives r₂^*.

DISCUSSION

The r₁ and r₂^* relaxivities of GaD depend on R₁ and R₂^* which in turn depend on many factors such as macromolecular content, temperature, pH, magnetic field strength, acquisition parameters and oxygenation. To put the values obtained in this study in the context of previous work, r₁ of GaD in water and blood obtained under similar experimental conditions are included in Tables 5 and 6, respectively.

Table 5.

A review of T₁ relaxivities (r₁) of Gd-DTPA (GaD) in water published in literature. ^â per kg of H₂O; W-Water; DW-Deionized Water; TF-Turbo Flash; ST-Siemens Trio; M7T-Magnex7T; SPIRFL-Single Point Inversion recovery Flash; SM63SP-Siemens Magnetom 63SP; TPIR-Two Pulse Inversion Recovery; BPC20-Bruker PC20 NMR Analyzer; SM-Siemens Magnetom; SMT-Siemens Magnetom Trio; VI-Varian Inova; WB-Water Baseline; HBNMRS-Home Built NMR Spectrometer.

Ref. No.	Year. YYYY	Sol. type	Temp °C	B0 T	r1 mM−1s−1	[GaD] mM	Scan Type	Scanner	pH	TE ms	TR ms	TI ms
Our Data	2013	AS	37±2	3	3.44±0.08	0–4.83	TF	ST	7	1.3	10000	20–9000
Our Data	2013	AS	37±2	7	3.18±0.09	0–4.83	TF	M7T	7	1.1	10000	20–9000
(15)	1997	DW	38	1.5	2.95	0.01–5.93	SPIRFL	SM63SP	-	20	1000	25–6000
(16)	2005	W	37	0.47	3.4±0.2	0.25,0.5	TPIR	BPC20	-	-	-	-
(16)	2005	W	37	1.5	3.3±0.2	0.25,0.5	TPIR	SM	-	-	-	-
(16)	2005	W	37	3	3.1±0.2	0.25,0.5	TPIR	SMT	-	-	-	-
(16)	2005	W	37	4.7	3.2±0.2	0.25,0.5	TPIR	VI	-	-	-	-
(17)	1999	W	37	6.3	3.00±0.56	-	IR	HBNMRS	-	-	-	-

Table 6.

A review of T₁ relaxivities (r₁) of Gd-DTPA (GaD) in blood published in literature ; CB-Canine Blood; HB-HumanBlood; TF-Turbo Flash; RFSSGE-RF Spoiled Segmented Gradient Echo; VFR-Variable Field Relaxometer; ST-Siemens Trio; M7T-Magnex7T; SM-Siemens Magnetom; TPIR-Two Pulse Inversion Recovery..

Ref No.	Year YYYY	Sol. type	Temp °C	B0 T	r1 mM−1s−1	[GaD] mM	Scan Type	Scanner Type	TE ms	TR ms	TI ms
Our Data	2013	OVB	37±2	3	3.20±0.06	0–5.18	TF	ST	1.3	10000	20–9000
Our Data	2013	OVB	37±2	7	3.20±0.06	0–5.18	TF	M7T	1.1	10000	20–9000
(18)	1999	HB	37	1.5	3.92	0–5	RFSSGE	VFR	-	-	-
(16)	2005	CB	37	1.5	4.3±0.3	0.2–0.5	TPIR	SM	-	-	-

While previous studies investigating r₂^* of GaD have not been performed to date, native R₂^* in blood has been of great interest. Values at 1.5 T have shown arterial blood to have values ranging from 3.94±0.50(9) to 5.02±0.2(10) s⁻¹. In the current study, large increases in native blood R₂^* were measured at 3T and 7T; 9.94±0.56 and 28.50±1.15 s⁻¹, respectively.

Dependence of R₂^* on Oxygenation and GaD in Blood

The characteristic linear relationship between R₂^* and [GaD] observed in AS was not observed in OVB. Blood has three major distinguishing characteristics from a homogenous sample such as AS. First, compared to saline, blood experiences longer motional correlation times (slow motion) resulting in increased proton-electron dipolar coupling. Second, it is multi-compartmental, consisting of the plasma and red blood cells (RBCs), where GaD (only present in the plasma) affects magnetic susceptibility differences between RBC and plasma. Third, it contains hemoglobin, the oxygen saturation of which also impacts the susceptibility mismatch between the two compartments. In the case of fully oxygenated blood, the multiple compartments, and the dynamic processes between them, most likely results in the complex relationships observed between R₂^* and [GaD].

In the full range of physiological conditions, the oxygen dependence of blood R₂^* is usually attributed to oxygenation based susceptibility of red blood cells (RBCs) where oxyhemoglobin is diamagnetic and de-oxyhemoglobin is paramagnetic, relative to blood plasma. In OVB, microscopic magnetic field gradients are set up between RBCs and plasma resulting in static and dynamic dephasing of spins that can result in an increase in R₂^* (11).

In deoxygenated blood, it was shown in a study by Blockley et al. that a quadratic behavior existed between R₂^* and [CA] (ProHance) (11). In this previous study a parabolic curve was fit to the data with the inflection point occurring at increasing concentrations with increasing field strength (1.0 mM and 1.3 mM at 3.0T and 7.0T, respectively). This relationship was described as originating from an initial susceptibility difference between the paramagnetic RBCs and the diamagnetic plasma in deoxygenated blood. As the [CA] increases there is likely an increase in the net paramagnetism of blood plasma which could lead to an averaging of the RBC versus plasma susceptibility variations. The inflection point in the R₂^* curve represented the point at which the CA increases the paramagnetism of the plasma to match that of the RBCs. Beyond this point, the susceptibility difference between plasma and blood again increased as did R₂^*.

In the current study, oxygenated blood is investigated where the RBC are diamagnetic with respect to the blood plasma. While there is no inflection point observed in this data there is still an observed quadratic relationship between R₂^* and [Gad] at lower concentrations from 0–2 mM, however a quadratic function did not sufficiently describe the data above this range. The observed quadratic versus linear contributions was significantly different between 3T and 7T. This field dependence could be a result of varying contributions of susceptibility induced versus dipolar relaxation pathways further influenced by varying exchange and diffusion regimes with changes in [GaD].

While this study was not explicitly designed to elucidate the relaxation mechanisms involved, a brief discussion of factors that could contribute to the observed behavior follows. First, in the presence of GaD, which affects the relaxation rate constant of blood plasma predominantly, the regime of isothermal dynamics is shifted towards the intermediate motional regime (IMR) from the fast motional regime (FMR). In this study, because the apparent rate constants were altered by changing the concentration of GaD, and since GaD is only present in the plasma (site A) and not the RBC (site B), the “shutter speed”, i.e., |R_A−R_B|, of dynamic processes is varied. The dynamic processes which include exchange and diffusion which collectively result in dynamic averaging (DA) and are additionally characterized by the rate constant k_DA (12). In this particular case the regime of exchange or diffusion in local susceptibility gradients or intracellular-extracellular water transport are varied with the increase of the concentration of GaD (13). It appears that the presence of increasing GaD in plasma shifts the regime of the dynamic process by modifying the intrinsic relaxation rate constants at specific magnetic sites A or B undergoing exchange or diffusion, as well as the apparent populations of sites A (i.e., plasma water protons interacting with GaD) or B (RBC water protons). Moving in this direction takes us from a quadratic to a linear relationship with increasing [GaD] where |R_A − R_B| ≫ k_DA.

Effect of hyperoxygenation

Based on initial in-house experiments, the OVB samples were necessarily hyperoxygenated in order to maintain the blood at arterial sO₂ levels (sO₂ > 94%) over the complete MR scan period. At our highest pO₂ of 294 mm Hg, Othee et al. have shown that the fraction of oxygen in the blood that is freely dissolved and not bound to hemoglobin is approximately 7% compared to 4% when exposed to room air (pO₂ = 138 mm Hg)(14). At 8.4T, the increase in dissolved oxygen from 4% to7% would result in a minimal increase of 6% in blood T1 when measured at room temperature (14). In another study, Blockley et al. reported R₁ of whole blood versus (1−Y) at 3T and 7T, where ‘Y’ represents the fractional sO₂ (11). In our 3T studies where Y = 0.97 and 7T studies where Y = 1, the predicted longitudinal relaxation times following Blockley’s relation would be 1730 ms and 2041 ms, respectively. These values are in excellent agreement to our hyper-oxygenated OVB T₁ values of 1788.09±11.78 ms at 3T and 2055.43±14.00 ms at 7T.

The effect of hyperoxygenation on transverse relaxation is arguably much more complex. Molecular oxygen affects relaxation through paramagnetic interactions and, in addition, could influence magnetic susceptibility variations in the sample thus leading to different conditions of DA. These mechanisms are challenging to characterize and beyond the scope of the presented work. While a potential limitation of the current study, hyperoxygenation was required to maintain arterial sO₂ levels with the chosen experimental methods.

Acquisition Methods

Alternative strategies, both in terms of the experimental setup and data acquisition were also considered for this study. However, the desire to obtain a wide distribution of concentrations at physiological temperature and sO₂ led to the use of the IR-TFL sequence and rapidly imaging samples in an experimental setup with a concentration range from 0–5 mM. Determination of T₁ values using the IR-TFL method was validated with those obtained using the IR-SE sequence with the final values only differing by 3%.

As the Hct in the blood settles over the duration of the study, there is a change in the Hct content in the scanned MRI slice. An increase in Hct will lead to a higher value of r₁. Using the IR-TFL sequence allowed the data to be acquired rapidly in comparison to the IR-SE method which decreased the total data acquisition time from 65 to 12 minutes. Secondly, the vials were positioned vertically to lengthen the distance over which settling would occur. The scan plane was also positioned through the center of the vials to minimize the pile up of Hct at the bottom or the absence of Hct at the top of the vile. Based on repeat data acquisitions, a 0.4% decrease in r₁ per minute and a 0.3% increase in the slope of R₂^* vs. [GaD] per minute was found. Thus, the effect due to settling was minimal in this study.

The challenges with the chosen acquisition methods were T₂^* blurring of the IR-TFL acquisition and the potential for a spatially varying B₀ impacting primarily R₁ and R₂^* calculations, respectively. To minimize the effect of T₂^* blurring, which increases with [GaD], ROIs well within each sample vile were used for R₁ determination. Despite this strategy, slight variations from the linear relaxivity curve are still observed at the highest [GaD], Figure 2a–d. To address B₀ homogeneity, a local B₀ shim was used. While a majority of the signal used for B₀ shimming originated from the water bath, field perturbations from within and immediately around the samples had the potential of biasing the global shim. For the AS samples, the expected linear relationship between R₂^* and [Gad] is observed for both 3T and 7T indicating that a reasonably homogeneous B₀ field was obtained with the chosen shimming methods.

This study provides important information on the susceptibility effects on R₂^* relaxation rate constants and their dependence on external static magnetic field in the presence of GaD. Importantly, because the susceptibility induced relaxation channel is one which significantly contributes to the free precession R₂ through its dependence on DA, our study indirectly provides insight into the free precession transverse relaxation of the blood in the presence of GaD as well. It should be emphasized that for accurate estimation of the intrinsic relaxation parameters from T₂ measurements, the measurements of R₂ relaxation dispersion using a Carr-Purcell-Meiboom-Gill (CPMG) acquisition is necessary.

CONCLUSION

The r₁ of AS and OVB remained similar both between sample types and field strengths with only minor differences observed. The R₂^* relaxivity of AS was also similar at both 3T and 7T with only a 3% increase at the higher field strength. Furthermore, in AS, r₂^* values at 3T and 7T were only 22% and 36% higher than the respective r₁ values. There was, on the other hand, a large effect of the multi-compartmental nature of blood which resulted in an apparent nonlinear relationship between R₂^* and [GaD] and tremendous increases in r₂^* not observed the homogeneous AS phantom. OVB r₂^* exhibited a large field dependence with an approximate four-fold increase in R₂^* relaxivity at 7T compared the 3T. In addition, OVB r₂^* was twenty-fold higher at 3T and ninety-fold higher at 7T compared to their respective r₁ values.

Knowledge of r₁ and r₂^* relaxivities are paramount to using contrast agents at any field strength. This study demonstrates sharply increasing r₂^* in the multi-compartment blood samples greatly outpaces the relatively static r₁ characteristics with increasing field strengths. Unless accounted for, studies relying on the typical T₁ based contrast enhancement typically afforded by paramagnetic contrast administration at lower field strengths may obtain erroneous results.

Abbreviations used

GaD

Gadolinium-DTPA

[GaD]

Gd-DTPA concentration

r₁

T₁ relaxivity

r₂^*

T₂^* relaxivity

CA

Contrast Agent

OVB

Oxygenated Venous Blood

AS

Aqueous Solution

SLPM

Standard Liters Per Minute

Hct

Hematocrit

IR-TFL

Inversion Recovery - Turbo Flash

DA

Dynamic Averaging