Presentation on theme: "Treatment Planning of HIFU: Rigid Registration of MRI to Ultrasound Kidney Images Tara Yates 1, Penny Probert Smith 1, J. Alison Noble 1, Tom Leslie 2,"— Presentation transcript:
Treatment Planning of HIFU: Rigid Registration of MRI to Ultrasound Kidney Images Tara Yates 1, Penny Probert Smith 1, J. Alison Noble 1, Tom Leslie 2, Anthony McIntyre 3, Rachel Phillips 3 1 Dept. of Engineering Science, Parks Road, University of Oxford, Oxford, UK 2 HIFU Unit, Churchill Hospital, Headington, Oxford,UK 3 Department of Radiology, Churchill Hospital,Oxford, Uk ACKNOWLEDGEMENTS: Thanks to EPSRC for funding, EP/C00633X. MRI Segmentation Segmentation of MR images used a statistical intensity based method. This method applied Ostu’s thresholding method  to separate the image into distinct classes. The segmentation was then further refined using a Markov random field Expectation- Maximization approach as described in . Figure 1: The application of the Sticks Algorithm to reduce speckle in an ultrasound image of the kidney a) original ultrasound image b) result of the sticks algorithm c) the manually selected points Method To register one image to another a similarity between the two must be established. The method presented here uses shape as a basis for registration as it is robust to noise. In each image regions within and outside the kidney were determined (segmentation), and points on the kidney identified. Then ellipses are fitted to these points, and the geometric transform between ellipses calculated (registration). Ultrasound Segmentation Manual segmentation was used for the ultrasound images. Around 35 points were selected as shown in Fig 1c). The “sticks” filter proposed by Czerwinski  was used to reduce speckle prior to point selection. Figure 2: Segmentation of MR image for one slice through the kidney It is assumed that the shape of the kidney undergoes little deformation in subsequent scans and is independent of scan position. The MR slice with the ellipse found to have the closest major-minor axis ratio to the ellipse fitted to the ultrasound data was selected for registration from the kidney volume. Rigid registration of the two images was achieved using the parameters of the ellipse, namely the position of the ellipse centre, the angle between the major axis and the x-axis and the size of the major and minor axis, to provide the transformation parameters, transformation (Tx,Ty), rotation (Rx,Ry) and scale (Sx,Sy). Introduction For HIFU to be successful in a clinical setting accurate knowledge of the position of the focus in relation to the tumour boundaries is required. In the planning stages of HIFU therapies at the Churchill Hospital, Oxford, both pre-operative MRI and US images are acquired. The registration of the pre-operative MRI volume to the intra-operative real-time 2D US images would provide extra structural information to assist treatment. Registration To avoid a full 3D-2D registration, a 2D oblique slice from the MR volume is selected. Slice selection and ultimately the image registration is achieved through the use of 2D shape matching. The segmentation steps identify boundary points for the kidney in both imaging modalities. An ellipse is fitted through each set of boundary points using a least squares method and the major and minor axes are calculated. An example of an ellipse fitted through the boundary points of the segmented kidney from both ultrasound and MR images is shown in Fig. 3. Figure 3: Ellipse fitted to a) US and b) MR data Results and validation The results for a one patient study are shown in Fig. 4. Validation The calyceal system appears in both images as a contrasting intensity central region. It can be used as an independent measure to validate the method of slice selection and registration. Fig. 5 shows the overlay of the registered MR outline of the selected slice (fig. 5b) and two alternative slices on the ultrasound image. The calyceal system is most closely matched in the selected slice Registration time (excluding segmentation) is approximately 1 second. Conclusion With further validation the speed of the registration and the ability to process the majority of the MR data prior to surgery would realistically allow the application of this method to the real-time ultrasound images acquired during HIFU treatment. References 1.R.N. Czerwinski, D.L Jones and W.D. O’Brien, “Detection of Lines and Boundaries in Speckle Images – Application to Medical Ultrasound”, IEEE Transactions on Medical Imaging, 18(2), pp , February N. A. Ostu, “Threshold Selection Method from Grey-Level Histograms”, IEEE Trans. Syst. Man. Cybern., SMC-9, (1), pp 62-66, K.Van Leemput, F.Maes, D. Vandermeulen, and P. Suetens, “Automated Model-based Tissue Classification of MR Images of the Brain”, IEEE Transactions on Medical Imaging, 18(10), pp , October 1999 Figure 4: Registration of an MR image to an ultrasound image of the kidney. The boundary segmented from the MR image is shown as an overlay on the ultrasound image. Figure 5: The registration of different slices from the MR volume. The shape of the central calyceal system is most closely matched in the selected slice 5b).