Signal Intensity Change

The choice of pulse sequence to monitor contrast agent kinetics must satisfy the above criteria but also provide acceptable spatial resolution and tissue coverage. As T2-weighted sequences tend to take more time to collect and the effect of contrast agent on signal intensity is negative [i.e. signal decreases with increasing contrast agent concentration hence reducing the signal to noise ratio (SNR) of the experiment], Trweighted sequences offer many advantages. Previously groups have used spin echo sequences to monitor contrast agent accumulation (Larsson et al. 1990; Tofts and KermodE 1991). Larsson et al. (1990) demonstrated a linear relationship between signal intensity and contrast agent concentration when the TR was reduced to 500 ms or less, though their temporal resolution for a single slice was only 68 s. More recently groups have used faster sequences including EPI (Gowland et al. 1992) and turboFLASH (Boetes et al. 1994) to monitor contrast agent accumulation with a temporal resolution of a few seconds. The relationship between the contrast agent concentration and the relative increase in signal intensity can be derived from the Bloch equations (Haase et al. 1986) for any imaging sequence. The signal intensity

This relationship remains approximately valid across a range of values for TR/T: and a. The constant of proportionality is a function of TR, g, p, and, as the flip angle decreases, a. The difficulty in comparing this constant between studies is the sensitive nature of g. The loading of the coil, receiver settings at the MR console and image reconstruction parameters alter the intensity of the signal in the image. Hence, it is necessary to relate the signal intensity to an internal standard. Other groups have used samples with known characteristics (Shames et al. 1993; Vaxiee et al. 2003) at a fixed location within the field-of-view or the signal from fat located close to the region of interest (Heywang et al. 1989). Placing a sample within the field-of-view in a clinical imaging study is often problematic, complicating patient positioning. The signal from fat (for example in a breast imaging study) is often very variable (Pedevilla et al. 1995) and the values obtained for contrast agent concentration are therefore not reproducible. An alternative is to relate the signal intensity post-contrast to that pre-contrast. This has the advantage of maintaining the position of the standard in relation to the enhancing structure and requires no prior positioning. However, the use of pre-contrast signal intensity also introduces the pre-contrast T of the structure into the analysis. If we assume signal intensity is proportional to 1/T (Eq. 5.4), then:

rpTR [T1gj T10) lL J

where S0 and SGd, and T10 and T1Gd are the signal intensities and spin-lattice relaxation times before and following administration of contrast agent respectively and r1 is the relaxivity of contrast agent. Dividing by the pre-contrast signal we obtain:

500'

benign malignant a benign malignant a

Consequently the relative increase of signal intensity following administration of contrast agent is related to both the spin-lattice relaxivity of contrast agent and the pre-contrast T1 of the tissue. This difficulty is highlighted in Fig. 1, which shows data from a study of patients with tumours of the breast (Mussurakis et al. 1997). The enhancement measured using relative changes in signal (Fig. 5.1a) suggests that benign lesions enhance more than malignant lesions. In fact, the benign lesions tend to have a longer native T1 and actually show less uptake of contrast agent (Fig. 5.1b).

T1 Measurement

The linear relationship between signal intensity and contrast agent concentration is only approximately true over a limited range of contrast agent concentration. The linear relationship between 1/T1 and contrast agent concentration has been shown to hold over a much wider range of Gd concentrations by a number of groups (Rosen et al. 1990; Donahue et al. 1994; Judd et al. 1995). Unfortunately, the measurement of T1 in vivo is a non-trivial problem and, of particular significance, accurate measures of T1 are often time consuming to obtain. A considerable body of literature has developed on this subject and an exhaustive description of the existing techniques is not attempted here, merely an overview of the more common approaches.

The methods of measuring T1 using MR images fall broadly into two categories:

1. Inversion/saturation recovery prepared imaging sequences, and

2. Variable saturation techniques.

benign malignant

Fig. 5.1a,b. Box and whisker plots showing the distribution of measures of maximum enhancement seen in a study of 58 breast tumours (MussurAkis et al. 1997). The use of relative signal increase as a measure of contrast agent uptake indicates that benign lesions enhance more significantly (a). Conversely, the assessment of changes in 1/T1 [by the method of Hittmair et al. (1994)] as a measure of contrast agent uptake indicates that malignant lesions actually take up more contrast agent (b)

benign malignant b

Fig. 5.1a,b. Box and whisker plots showing the distribution of measures of maximum enhancement seen in a study of 58 breast tumours (MussurAkis et al. 1997). The use of relative signal increase as a measure of contrast agent uptake indicates that benign lesions enhance more significantly (a). Conversely, the assessment of changes in 1/T1 [by the method of Hittmair et al. (1994)] as a measure of contrast agent uptake indicates that malignant lesions actually take up more contrast agent (b)

Each technique may be accomplished using a number of imaging sequences including; spin echo, EPI, or gradient-echo imaging with the appropriate additional pulses and subsequent processing algorithms.

Inversion or Saturation Recovery Techniques

An inversion pulse (Bluml et al. 1993) or a series of saturation pulses (Parker et al. 2000) provides T1-weighted preparation of the signal dependent upon the subsequent delay prior to acquisition of a normal imaging sequence. Using a series of images,

ro 300

each obtained with a different delay, the T: of the sample may be estimated (Bluml et al. 1993). While this is perhaps the most precise method of obtaining an estimate of T1, and the precision increases with the number of delay times used, it can be very time consuming. For example, an inversion recovery spin echo image with 128 phase encoding steps collected at four different delay (TI) times between 50 and 950 ms (maximum delay <TR) requires an imaging time of 1.0x128x4=8 min 32 s. In fact, this sequence would only be suitable for estimates of T1 values up to around 700 ms. Precise estimates of longer T1 values requires the use of a longer maximum TI and consequently, TR. The time required for such measurements can be reduced significantly by the use of snapshot-FLASH (Bluml et al. 1993) or EPI-based approaches (Gowland and Mansfield 1993) to sample an entire image at each TI, although these methods can be degraded by severe point spread function artefacts (Parker et al. 2000). Further time savings can be made using Look-Locker techniques (Freeman et al. 1994), albeit with limitation on the number of slices that may be acquired per unit time.

Variable Saturation Techniques

T1 may be estimated using the ratio of two spin echo images collected with different TRs, but again imaging times can become prohibitive. A similar approach is to use gradient echo images with variable flip angles (Fram et al. 1987). The signal intensity obtained using a FLASH sequence has been described above (Eq. 5.3). Rearranging this equation yields:

1-exp

,exp

r. Hence a plot of Y

against X for a range of flip angles will result in a straight line, and T may be calculated from the slope. Wang et al. (1987) have described the optimal sequence parameters for minimisation of the error in the calculated T1 when only two flip angles are used. With a given T1 two flip angles can be chosen which provide a greater precision in T1 estimate than a comparable spin echo pair. It may, however, be difficult to choose an appropriate pair of flip angles if the sample contains an unknown or large range of T1 values. Here the use of a number of equally spaced flip angles may be employed. However, in this case it may be advantageous to use a more computationally intensive non-linear fit using the original FLASH equation (Eq. 5.3) if the precision is to be improved (Wang et al. 1987).

In the field of dynamic contrast-enhanced MRI it is common to measure T1 using variable flip angle gradient echo acquisitions, usually while keeping TR constant. Such techniques require only relatively short acquisition times, which allows good temporal resolution, and may be used in multi-slice (Hittmair et al. 1994) or 3D volume modes to provide tissue coverage (BRookes et al. 1999) (Zhu et al. 2000). A simple protocol for a quantitative DCE-MRI study utilises a single heavily proton density-weighted (PD-weighted) acquisition, acquired prior to contrast agent administration, followed by numerous T1-weighted acquisitions over time (see for example Parker et al. 1997; Evelhoch 1999; Li et al. 2000). This could be achieved by using a low flip angle for the PD-weighted acquisition, and a higher flip angle for each of the dynamic T1-weighted acquisitions, whilst keeping TR short to maintain temporal resolution. Note that the PD-weighted acquisition is obtained only once; T1 is always estimated by comparing the signal intensity of the T1-weighted acquisitions (before or after contrast agent administration) with this single PD-weighted acquisition. Such a strategy allows T1 to be estimated rapidly throughout the time course of signal enhancement.

Factors Affecting Measurement Accuracy

The accuracy of T measurement may be compromised by a number of factors: machine non-linearities (in the main field, gradients, or radio-frequency (RF) amplifier), which are not accounted for in the calculation, unpredicted sample artefacts (e.g. susceptibility artefacts caused by ferromagnetic objects), sequence dependent errors, partial volume effects or flow and motion. These factors may affect all of the techniques available for T1 measurement, but certain sequence dependent errors are of particular relevance for the variable saturation techniques described.

Slice imperfections. Slice selective RF-pulses used to excite the imaging slice are never perfectly rectangular and therefore the sample receives a range of different flip angles through the slice. These imperfections (usually manifest in "peaking" at the slice edges) are often magnified in the estimate of T resulting in loss of accuracy. These errors may be corrected via careful calibration (Parker et al. 1997; Brookes et al. 1999) or via modification of the calculation procedure if the true pulse profile is known (Parker et al. 2001). IR techniques employing a non-selective inversion pulse tend to remain largely immune to these problems as the T1 estimate is principally determined by the TI time and not the read-out sequence. Even in the case of a slice selective IR measurement the effect of an imperfect inversion slice profile can effectively be factored out as a contributing factor to inversion inefficiency.

RF power. Calculations of T1 rely on a predicted signal behaviour following excitation pulses with controlled flip angles. Often the pulse transmitted may not achieve the expected amplitude due to RF-transmit-ter coil inhomogeneities or improper calibration of RF amplitude. This will clearly cause a problem in the multiple flip angle technique of T1 estimation since there will be a tendency to underestimate T1 if the RF power is too low. These problems may be particularly evident when coils with a non-linear response are used for RF transmission. However, once again, if the RF profile of a given coil is known it is possible to correct for spatially varying RF transmission fields (Parker et al. 2001).

A number of approaches have been developed to minimise the errors associated with slice imperfections (Parker et al. 2000, 2001) or RF power miscali-bration (Cron et al. 1999) requiring additional preparation or calculation. If T1 is to be measured in the clinical setting, then it is desirable that the measurement technique has a minor effect upon the normal imaging protocol. The multiple flip angle approach of Wang et al. (1987), though limited in precision, provides a rapid, and easily implemented methodology and has been employed in 3D mode where slice imperfections are minimised, particularly in the central sections of the 3D block (Brookes et al. 1999; Zhu et al. 2000).

0 0

Post a comment