Magnetic Resonance Imaging

Magnetic resonance imaging (MRI) offers a 3D dataset, arbitrary imaging planes, and unparalleled soft-tissue contrast, making it the modality of choice for imaging the vast majority of soft-tissue tumors. Because of its greater soft-tissue contrast, MRI has been shown to provide more consistent target delineation than CT for a variety of sites (48-50) and integration of magnetic resonance (MR) images has been shown to reduce interphysician variation, resulting in more reproducible plans (51). Most treatment planning systems permit "fusion" (i.e., deformable registration) of images acquired in MRI scanners with the CT planning images to aid in target delineation. This step has the potential for registration errors and is time consuming.

Performing radiation treatment planning based directly on MRI images, or MRI simulation, could circumvent this step. MRI scanners could be adapted for simulation by adding triangulation lasers and modifying the patient table to be flat. However, there are a number of challenges that must be overcome to perform radiation treatment planning based solely on MRI images. These challenges are the subject of active research and include, (i) the generation of DRRs for treatment verification, (ii) the correction of spatial distortions due to nonuniform magnetic field gradients, (iii) an adaptation of the planning software for tissue inhomogeneity correction, (iv) reproducible immobilization within the constraints of magnet bore size, (v) an imaging field of view which encompasses the entire surface of the patient, and (vi) addressing motion artifacts due to the long scan times.

As previously described, DRRs are required to verify patient setup relative to bony landmarks, on the treatment table. Various approaches have been proposed to generate DRRs from MRI images including automated techniques using a defined range of intensity values on Tl-weighted images and manual segmentation of the bones (52). Because the MR dataset does not contain density information, the material within the specified upper/lower threshold can be assigned pseudodensities. This allows dosimetrists to generate MR-based DRRs of the skull. Further research and development are required to adapt current treatment planning software to this technique and extend it to other body sites.

Spatial distortions on MR images are chiefly the result of nonlinear magnetic field gradients (which, ideally should produce magnetic fields that vary linearly with position). In principle, this deviation from linearity is knowable for each scanner architecture and can be accurately corrected. All commercial systems allow users to apply 2D geometric correction algorithms to the image sets (Fig. 6). Correction for error in the third dimension (slice select), is achievable but not readily available on most MRI scanners at this time (53). Note that the magnitude of spatial distortion is directly proportional to the distance from the scanner isocenter, and if not corrected, may introduce a significant error in the spatial allocation of a radiation isocenter, based on surface references which are always located at a radial distance

Figure 6 Axial MR simulation images. The isocenter is placed at the level of the prostate gland, and fiducial markers (arrow) are placed on the skin at the point of reflection of the triangulation lasers. (A) Uncorrected and (B) corrected images for gradient nonlinearity show a small difference in the projected location of the isocenter (dashed line intersection) and in skin contour measurements.

Figure 6 Axial MR simulation images. The isocenter is placed at the level of the prostate gland, and fiducial markers (arrow) are placed on the skin at the point of reflection of the triangulation lasers. (A) Uncorrected and (B) corrected images for gradient nonlinearity show a small difference in the projected location of the isocenter (dashed line intersection) and in skin contour measurements.

from the scanner isocenter. This is less of a problem for brachytherapy planning as catheters and tumors are positioned at the scanner isocenter.

Minor patient-induced distortions, such as susceptibility and chemical shift distortions, are not easily corrected. However, early studies have confirmed that the magnitude of radiation dosimetric error introduced by spatial distortion after geometric corrections are applied to MR images, is minimal and comparable to those obtained with CT planning (54,55). Again, more studies are needed to confirm the dosimetry accuracy of MRI simulation and to identify those body sites especially prone to susceptibility and chemical shift distortions.

Unlike CT Hounsfield units, MR images do not contain information related to the electron density of tissues. Attenuation correction for inhomogeneities in tissue electron density is not easily applied to treatment plans generated solely from MR images. However, the electron density of soft tissues is almost identical to that of water, and there is only minimal change in the attenuation of high-energy X rays through bone. The major sources of dosimetric error related to beam attenuation are the lungs and air cavities. Segmenting air cavities and the lungs on MR images, and manually assigning them densities, can easily circumvent majority of the problems (52). In the balance, one must judge whether improvements gained in target delineation accuracy with MRI supersede the small dosimetric error introduced by a lack of attenuation correction for electron density.

In the vast majority of cylindrical high-field MRI systems, patients are placed inside a 60-cm diameter bore. This severely constrains patient position and immobilization, which must be identical between simulation and treatment. Open magnets offer more freedom of positioning and patient size at the cost of much lower magnetic field strengths, and consequently, a lower image quality. Moreover, immobilization devices must be MRI compatible. With careful planning and adaptation, these constrains are easily surmountable for the majority of patients and treatment sites.

The problem of motion artefact is not unique to MRI, and is the subject of much research in radiation delivery as detailed previously. MRI scan times often closely resemble radiation treatment times and may well be better suited for treatment planning than CT "snapshots in time.'' Moreover, similar or identical solutions to target motion can be applied to both image acquisition and radiation delivery, such as respiratory gating (56). For abdominal images, drugs like glucagon can be administered to temporarily inhibit peristalsis and improve image quality (57). Finally, anatomical movies that can measure the magnitude of organ or target motion for a given patient over time can be acquired with MRI, allowing for a smaller and more accurate margin allocation around the CTV (58).

Beyond its evident advantages in anatomical imaging, MRI promises to be a leading imaging modality in the field of biological imaging. It is beyond the scope of this chapter to review all the emerging techniques in this field. However, a few concepts are especially germane to radiation treatment planning and should be emphasized. First, any MRI technique, from anatomical to molecular, that aids in improving tumor definition stands to significantly improve the quality of radiation treatment. Promising advances in tumor delineation within the prostate gland with magnetic resonance spectroscopic imaging (MRSI) (59) is a good example, whereby clinicians may now preferentially target a predominant tumor nodule for delivery of higher doses (60). Second, imaging techniques, which help to characterize physiological or biological properties of tumors and their subvolumes, may be used to generate a radiosensitivity map, and as such the dose required for tumor cure. For example, this may include maps of hypoxia (61), tumor cell proliferation (62), and angiogenesis (63). Third, MRI examinations can be safely repeated at various intervals throughout a course of radiotherapy and may play a key role in the dynamic adaptation of radiation treatment plans as a tumor responds to therapy.

Figure 7 (Caption on facing page)
0 0

Post a comment