T ri V I V

Fig. 7.3a-e. A combination of T1-weighted and T2*-weighted dynamic contrast enhanced acquisitions was used in the imaging protocol for DCE-MRI studied in a patient with a posterior falcine meningioma (Zhu et al. 2000). The pulse sequence consisted of a series of three-dimensional radiofrequency spoiled T1-weighted gradient echo acquisitions. Calculated maps are (a) and (b): Pre- and postcontrast images from T1-weighted dynamic series. There is diffuse enhancement of leptomeninges on the postcontrast T1-weighted images. c Parametric map of rCBV from T2*-weighted data. d Map of extravascular uptake from T1-weighted data. e Map of leakage space from T1-weighted data. [Reprinted with permission from Zhu et al. (2000)]

c d that a precontrast measurement of T1 should be performed before injection the contrast agent. One important factor contributing to erroneous measurements of T1(0) in blood is flow. The coherent movement of flowing fluid can alter T1 of the signal arising from spins therein (Lee et al. 1999). Placing the central-encoding lines within the low blood flow window can, to some extent, minimize the problems of the inflow effect (Zheng et al. 1999). Triggering difficulties, saturation effects, or problems related to the mathematical conversion of signal intensity into relaxation rates may further complicate the measurement of the T1(0). Another problem may arise when the signal saturates less than what is predicted theoretically, originating from the assumption that the slice profile exhibits a rectangular slice profile. In reality, the excitation flip angle will vary across the slice, concomitantly implying an incorrect measurement of T1(t), since the flip angle is included in the theoretical signal equation. Very fast T1 mapping is one possibility to eliminate this problem (McKenziE et al. 1999; Freeman et al. 1998), but their practical applications in DCE-MRI studies are yet to be investigated. The value of T1(0) for blood is therefore difficult to measure in vivo, since the experimenter may not stop the blood flow. Consequently, the T1 of the arterial blood has in many studies not actually been measured, but has been assumed from previously measured T1 values in humans (Noseworthy et al. 1999) or has been assumed to be the same as the value measured ex vivo (Strouse et al. 1996).

Since tumor contrast kinetics is often sampled over several minutes, the potential for patient motion is substantial. T1-weighted DCE-MRI is therefore vulnerable to partial volume errors (also present in the T2 and T2* DSC-MRI), both in AIF(t) and c(t), and since tumors are known to be heterogeneously distributed, the result may be a heterogeneous enhancement. From simulation studies, it has been demonstrated that the partial volume effect gives rise to a non-linear relationship of 1/T1-1/T1(0) with respect to the contrast agent concentration (van Osch et al. 2001). Another approach to the problem could be to separate the signal from a voxel into contributions from large vessels and capillaries by independent component analysis (Carroll et al. 2002). By this method, a pure large-vessel signal can be obtained from a voxel, with no partial volume effect.

The signal as a function of TE, TR and flip angle can usually be calculated on the basis of the Ernst formula (Zhu et al. 2000). Different pulse sequences result in different mathematical expressions, and the T1 weighting may therefore strongly depend on the pulse sequence parameters applied. Moderate T1 weighting can easily be performed using standard spoiled GE techniques, either as the standard or as the EPI mode. However, pre-pulses such as saturation or inversion pulses allow a significantly heavier T1 weighting. Fritz-Hansen and coworkers (1996), for example, demonstrated that the arterial input function could be accurately measured using an inversion-prepared spoiled GE sequence. Henderson et al. (2000) used a similar approach with the use of a saturation prepared spoiled GE sequence, which was related to T1 through the complex equation,

where TI is the time from the saturation pulse to the first excitation pulse in the GE sequence, and n is the number of phase-encoding lines before acquisition of the center of the Fourier space. Based on this methodology, it was demonstrated that simultaneous measurements of the tissue blood flow and the permeability surface product could be accurately performed within minutes after administration of contrast agent. Sequences equipped with prepulses, especially inversion prepared pulses, suffer from their longer scan times, and a compromise must be found between sufficient temporal resolution, multiple slices and the extent of T1 weighting.

As seen from Eq. 4, the mathematical conversion of signal intensities into molar units depends on the specific relaxivity (r1) of the gadolinium complex in the tissue of interest; the ability of a contrast agent to enhance proton relaxation is defined as the relaxivity. This conversion is occasionally based on the assumption that the relaxivity is the same in plasma and tissue. Ex vivo measurements of the relaxivity in the myocardium of frogs (Donahue et al. 1994) and in vivo measurements of the relaxivity in the rat kidney (Pedersen et al. 2000), however, suggest that there is a distinct difference between plasma and tissue. To our knowledge, relaxivities of brain tumors have not been fully investigated, although steps have been taken to quantify such cell cultures ex vivo (Schmalbrock et al. 2001).

Correspondingly, accurate determination of c(t) based on measured signal intensities and precon-trast T1 is not trivial. Alternatively, conversions of signal intensities into quantitative measures have been performed from empirical in vitro calibrations (Lombardi et al. 1997). Although it is not strictly correct to do so, such methods will often allow a representative determination of the tissue contrast concen tration time curve and the qualitative arterial input function, but one must be aware of the limitations in such a process, and usually the same calibration profile cannot be used if pulse sequence parameters are different from the in vivo experiment.

The extravascular uptake occurs more slowly than compared with changes in AIF(t). The temporal resolution is therefore determined by AIF(t) more than c(t). The temporal requirements for sampling AIF(t) have been considered in detail by Henderson et al. (2000). They concluded that an accurate representation of AIF(t), requires 1 s of temporal resolution. The temporal requirements are furthermore closely linked to the pharmacokinetic model used. It should be emphasized that different pharmacokinetic models exist to handle Tl-weighted DCE-MRI data (Tofts 1997; Fenstermacher et al. 1981; Parker and Tofts 1999; St Lawrence and Lee 1998), allowing calculations of tissue blood flow, permeability surface area, mean capillary transit time, fractional volume of the extracellular extravascular space, fractional blood volume, extraction ratio, and more. These parameters are of great interest as indicators of tumor angiogen-esis, and essential component of growth of tumors. On the other hand, since more parameters are modeled, it is likely that there is a reduction in the precision with which each parameter can be measured. The subsequent need for high signal-to-noise ratio may therefore be more than realistically obtainable, and the alleviation of this problem may be going from parametric pixel-by-pixel analyses to region-of-inter-est analyses (Henderson et al. 2000). Another way is to reduce the number of model parameters, such as in the flow-limited models or in the permeability surface-limited models, where it is supposed that flow or extravascular uptake is the dominating factor. Characterization of the transcapillary transport of contrast agent in brain tumors have been extensively investigated by Brix and coworkers (1991) and by Gowland and coworkers (1992), to whom we refer for more information about T1-weighted DCE-MRI modeling.

0 0

Post a comment