Introduction

1. S Distortion (mm) Distortion (mm) 10 0 +1 0 -1 Phase Direction Phase direction Read Direction Read Direction Chang corrected LTO gradient induced distortion (read) forward read gradient (fwdgrad) Distortion (mm) 10 0 Phase Direction Phase Direction Read Direction reverse read gradient (revgrad) gradient induced distortion (phase) reverse read gradient (revgrad) CT of LTO L. Moore1, S.J. Doran2 and M.O. Leach1 2 Department of Physics, University of Surrey, Guildford, GU2 7XH, UK 1 CRC Clinical Magnetic Resonance Research Group, Institute of Cancer Research, Sutton, SM2 5PT, UK Pelvic MR scans for radiotherapy planning : correction of system and patient induced distortions Introduction Magnetic resonance (MR) imaging is recognised as offering potential benefits in the delineation of target volumes for radiotherapy (RT). For many soft tissue sites MR is able to provide significantly enhanced visualisation of the lesion, thereby improving target definition1. A major concern regarding the use of MR for RT planning, however, is that the images are prone to both geometrical and intensity distortions. These distortions are a consequence both of system distortion (arising from main magnetic field inhomogeneity and non-linearities in the applied magnetic field gradients) and of effects arising from the patient being imaged, namely susceptibility and chemical shift variations. Gradient performance is known to vary with time and in particular can be altered during servicing 2. This abstract describes a method for removing both system and patient-induced distortions from MR scans to provide MR scans of the pelvis suitable for RT planning. Methods Quantification of System Distortion: A phantom system described previously3 comprising an interlocking flat-topped couch insert, a reference frame and a linearity test object (LTO) is scanned using an acquisition sequence employing forward and reverse gradients. Susceptibility induced distortions arising from the test object itself can therefore be corrected using a simplified form of Chang and Fitzpatrick’s method4. The phantom system then undergoes a computed tomography (CT) scan. Any differences between spot positions as visualised on MR images corrected for object induced distortion and those seen on an equivalently positioned CT image are associated with non-linearities of the gradients. The discrete set of measured distortions is fitted with a 2-D polynomial leading to an estimated distortion for every pixel in the image volume. Preliminary results indicate that gradient induced distortions may be as large as 20 mm towards the edge of the field of view. The reference frame and couch insert contain rows of water filled marker tubes which, when imaged, give a pattern of spots the locations of which uniquely define the position of the slice. Any significant change of these patterns between imaging the phantom and imaging the patient indicates a change in the factors contributing to MR distortion and therefore the previously determined gradient induced corrections cannot be employed. Results Results obtained using each stage of the method described above on a patient volunteer are depicted below. Maps, detailing the difference between Chang corrected images and the images acquired using forward and reverse gradients illustrate the contribution to distortion from magnetic field inhomogenities and chemical shift effects. The final difference map shows the degree of distortion arising from gradient non-linearities on a MR image previously corrected for object induced distortion. To demonstrate the impact of such non-linearities we show difference maps from two different parts of the body, the legs (situated a distance of about 15 cm from the isocentre) and the prostate (situated near the isocentre). Note that all images acquired using a reversed gradient have undergone an initial left-right reversal such that they can be subtracted from the Chang corrected image. Discussion and further work To date in vivo results obtained using our distortion correction system compare well with what is expected. The largest patient induced distortions are apparent at the fat/air interface and the largest gradient induced distortion appear towards the edge of the field of view. The accuracy of our system will be tested further by comparing fully corrected MR images with equivalent CT images; the patient having been reproducibly positioned in both examinations. The next version of the software will perform a full 3-D analysis, correcting for distortions in slice profiles. References 1. Khoo VS et al.[1997] Radiother. Oncology. 42 : 1-5 2. Lemieux L, Barker G [1998] Med. Phys. 25 (6) : 1049-1054 3. Finnigan DF [2000] ISMRM 2000 : 1688 4. Chang H, Fitzpatrick JM [1992] IEEE Trans. Med. Imaging. 11 : 319-329 Acknowledgement The authors gratefully acknowledge the financial support of the Cancer Research Campaign. Chang - fwdgrad largest differences at fat/air interface (Difference map) Chang - revgrad differences at interfaces in opposite directions forward read gradient (fwdgrad) Chang - fwdgrad (difference map) forward read gradient (fwdgrad) Chang - revgrad (difference map) reverse read gradient (revgrad) Chang corrected image reverse read gradient (revgrad) Chang corrected Chang image now corrected for gradient induced distortion* Chang image now corrected for gradient induced distortion* 20 Distortion (mm) Distortion (mm) 0 +1 0 -1 -20 Phase Direction Phase encode Frequency Direction Frequency Encode Chang- corrected LTO gradient induced distortion (read) forward read gradient (fwdgrad) Effects of gradient of induced distortion - note large differences towards the edge of the FoV (Difference Map) 15 10 0 -10 Distortion (mm) Effects of gradient of induced distortion (not that great near isocentre) (Difference map) Phase Direction Frequency Direction gradient induced distortion (phase) * At present, we correct only inside the portion of the image spanned by the regular grid points of the LTO. CT of LTO