Simultaneous Mapping of rCBV and Krans from Drcemri

Simultaneous mapping of rCBV and Ktrans from DRCE-MRI is complicated by the synergistic effects of intravascular and extravascular contrast agent on measured signal intensity. Li et al. (2000b) (Li and Jackson 2003) have described a novel method which uses an initial data decomposition step to derive separate concentration time course data for intravascular and extravascular contrast media. This data decomposition is performed using prior knowledge of the shape of the intravascular contrast concentration time course curve, which is assumed to be identical to that observed in large vessels and the predicted shape of the contrast concentration time course curve in the extravascular extracellular space, known as the leakage profile. The shape of the leakage profile can be shown to be the integral of the intravascular concentration time course curve and the ratio between this integral and the observed data represents the Ktrans, which is known as Kfp (first pass) for this technique. The estimation of the leakage profile is based on an assumption that, since the data collection period is very short, there is no significant contrast in the EES through the measurement period. This assumption breaks down where extraction fractions are high and leads to systematic underestimation of high values of Ktrans (Li and Jackson 2003).

This approach has a number of significant advantages over conventional techniques. The generation of both intravascular and extravascular contrast time course data allows separate analysis for CBV and Ktrans. This removes the tendency of simplistic analyses to overestimate Ktrans in the region of large blood vessels so that pixel rejection rates due to inappropriately high Ktrans estimations are reduced from as high as 50% in gliomas to zero. The separation of the curve fitting analyses for intravascular and extra-vascular data reduces the number of free parameters and makes the technique more robust to images with poor signal-to-noise ratio (Li and Jackson 2003).

The technique is described in more detail in Chap. 6. Briefly it is based on a tumour leakage profile (LP), which could be defined in a two-compartment kinetic model:

where Cv(t) denoted the intravascular component of C(t); and rt --11-f)

If it is assumed that the backflow of the CA from extravascular space to the blood space during the first passage of CA was negligible, LP reduces to:

LP is calculated from the time dependent plasma-contrast concentration function, Cp(t), in 3D T1W dynamic studies. Dynamic MRI was performed using 3D T1W radio-frequency (rf) spoiled gradient-echo images (T1W-GRE) of large volume (acquisition matrix size = 128x128x25) and high time resolution (At < 5.1 seconds). Substantial improvements have been made in the implementation of this approach since its first description (Jackson et al. 2002a; Li et al. 2003).

The performance of the method was evaluated by comparing results to those obtained from more conventional methods in patients with primary brain neoplasms. The technique produced maps of Ktrans that appeared to be free of any contribution from intra-vascular contrast agent. Maps of vp showed close correlation with maps of blood volume calculated from independently acquired dynamic susceptibility weighted MRI examinations with no evidence of residual permeability effects (Haroon et al. 2002a; Li et al. 2000b, 2003; Zhu et al. 2000a). The novel feature of the first pass leakage profile (FPLP) method is that it uses only data collected during the first passage of the contrast bolus through the target tissue so that data acquisition is extremely fast compared to conventional method for measuring permeability (Brix et al. 1991; Larsson et al. 1990; Tofts and Kermode 1991).

Monte Carlo simulations have been performed to assess the accuracy and precision of the FPLP method (Li and Zhu 2002). These show that FPLP method produces accurate measurements of fractional plasma volume and of transfer constant where the leakage rate is not high (Ktrans < ~0.2 min-1); Measurements of Ktrans and vp were highly reproducible and were less affected by low SNR than conventional curve fitting approaches. However, FPLP will underestimate Ktrans when the backflow of the CA from extravascular space to the plasma during the first passage of CA is not negligible. A combined use of the FPLP and conventional curve fitting methods has shown the potential to overcome this problem and provide optimal accuracy and precision in quantification of Ktrans and vp (Li and Zhu 2002).

Figure 9.7 shows images from the central slice of the tumour volume from a patient with a glioma. The T1-CBV map calculated using conventional methods (Bruening et al. 1996; Hacklander et al. 1996b) has mixed contribution from both perfusion and contrast leakage (Fig. 9.7a). The geometrical distribution of high values (yellow and red) conforms poorly to that seen on either T2*-CBV (Fig. 9.7e) or Ktrans maps calculated using a modified version of the standard Tofts' analysis (Zhu et al. 2000c) (Fig. 9.7c). There is close concordance between maps of T1-CBVcorrected (Fig. 9.7d) and T2*-CBV (Fig. 9.7e). The distribution of high permeability values on Kfp maps (Fig. 9.7b) is similar to that on Ktrans maps (c) but Ktrans maps again show higher levels of noise and considerable residual contributions from first pass effects in vessels. Figure 9.7f compares the histograms of Kfp and Ktrans from the whole tumour. The tumour region of interest was manually drawn and summed from 12 slices. There is a drop in the number of pixels seen on Ktrans (denoted as k in Fig 9.7f) maps at low levels (<0.1 min-1). This is due to the high rate of fitting failures when the tri-exponential model is applied to the data points with low C(t). The Ktrans histogram is also skewed at the high end with a long thin tail due to the inclusion of mea

Map Ktrans Tofts

Fig. 9.7a-f. Colour-rendered parametric maps of T1-rCBV (a), Kfp (calculated using the FPLP model) (b), Ktrans [calculated using a modified version of the standard Tofts' analysis (Zhu et al. (2000c)] (c), 3D T1-CBVcorrected (d), and T2*-rCBV (e) from a patient with a glioma. The distribution of "hot" (yellow and red) areas in the tumour on the uncorrected T1-rCBV map does not agree with either the k or T2*-rCBV map. In contrast, the distribution of high values on the Kfp map is similar to that in the Ktrans map but with less noise. 3D T1-rCBVcorrected agrees closely with the T2*-CBV map. The histograms of permeability surface product values of whole tumour of the patient, calculated using the new (solid line) and conventional methods (dashed line), are compared in (f)

Fig. 9.7a-f. Colour-rendered parametric maps of T1-rCBV (a), Kfp (calculated using the FPLP model) (b), Ktrans [calculated using a modified version of the standard Tofts' analysis (Zhu et al. (2000c)] (c), 3D T1-CBVcorrected (d), and T2*-rCBV (e) from a patient with a glioma. The distribution of "hot" (yellow and red) areas in the tumour on the uncorrected T1-rCBV map does not agree with either the k or T2*-rCBV map. In contrast, the distribution of high values on the Kfp map is similar to that in the Ktrans map but with less noise. 3D T1-rCBVcorrected agrees closely with the T2*-CBV map. The histograms of permeability surface product values of whole tumour of the patient, calculated using the new (solid line) and conventional methods (dashed line), are compared in (f)

surements from voxels with large vessel components. The two histograms correspond reasonably well where k values lie in the mid range (0.1-0.4 min-1).

Clinical Applications of Advanced Parametric Analysis Techniques

A small number of studies have recently appeared describing results from combined studies of both tumour blood volume and vessel leakage.

Combined T1W and T2W Image Acquisition

Zhu et al. (2000c) used sequential T1W DRCE-MRI and T2*-weighted DSCE-MRI to produce estimates of Ktrans and rCBV. The pre-enhancement method worked well for all cases in this study whether con trast leakage was large or small. Production of parametric maps of rCBV, Ktrans and ve calculated from T1W and T2W dynamic MRI of 15 patients with brain tumours (five glioma, five meningioma, five acoustic neuroma) (Zhu et al. 2000c) allowed comparison of parameters on a pixel by pixel basis. This comparison demonstrated strong correlation between rCBV and Ktrans in 11 of 15 patients. However, decoupling between pixel-wise rCBV and Ktrans was found in four patients who had lesions with moderate Ktrans and ve elevation but no increase of rCBV. Figure 9.8 shows an example of one of these cases with low rCBV. The rCBV map in this case with meningioma demonstrates a heterogeneous tumour with low values between normal grey and white matter. Both Ktrans and ve maps delineate the tumour clearly against the background of non-enhancing brain tissues.

Apparently, in some tumours areas of high contrast leakage are not associated with increases of tumour blood volume. Such decoupling between permeability and blood flow may be of immediate significance, not only indicating inefficient blood supply (Jain and

Gerlowski 1984), but also reflecting the difference of time scales involved in the different angiogenic processes. Anti-angiogenic treatment (VEGF inhibition) has been shown to reduce Ktrans in a period of hours whilst CBV remains unaltered (Amoroso et al. 1997; Hawighorst et al. 1998; Jensen 1998; Lund et al. 1998), reflecting the continuous modulation of VEGF activity according to the metabolic demands in tumours (Hawighorst et al. 1998). It may be postulated therefore that a loss of co-location of Ktrans and rCBV associated with low tumour blood volume described here will be one initial marker of successful inhibition of angiogenic drive. Evidence that this type of de-coupling does occur is also seen in the case illustrated in Fig. 9.9, where a meningioma is associated with extensive increases in Ktrans in the

0 0

Post a comment