Breast MRI Diagnosis Assistant (BMRIDA)

Project aim:
To investigate the potential of image processing algorithms for assisting the diagnosis of breast cancer on the basis of contrast-enhanced magnetic resonance imaging (CE-MRI).

Project description:
The radiopaqueness of dense breast tissue makes the search for a primary tumour in x-ray mammograms difficult, if not impossible. Contrast-enhanced MRI of the breast has shown promising results and is currently under investigation as a screening tool for young women at genetic risk of breast cancer in the UK (see MARIBS). Radiologists evaluate the images on the basis of a standardised protocol, which includes the analysis of contrast agent (Gd-DTPA) uptake and washout curves as well as morphological features of suspicious regions.

Registration of CE MR breast images:
The visual assessment of contrast enhancement during one visit is based on subtracting the pre-contrast image from the post-contrast images, since contrast agent and fat appear both bright in T1-weighted MR images. Motion artifacts, however, can prevent a reliable diagnosis. The non-rigid registration algorithm developed by Daniel Rueckert [1,2,4], has been shown to significantly reduce the effects of movement artifacts in subtracted CE breast MR images [3]. Its effectiveness is illustrated in Figure 1. Click here for another example.

(a) (b) (c)
Figure 1: Maximum intensity projections of difference image (post-contrast - pre-contrast image) after (a) no registration, (b) rigid registration and (c) non-rigid registration.

Volume Change due to Registration:
Generally, non-rigid registration algorithms can change the volume of structures seen in images. While this may be necessary for inter-subject registration, such volume changes are, however, highly unlikely during a CE MR mammography acquisition since no external forces are applied to the breast and the gap between image acquisitions is short. In [5] we evaluated the volume change associated with non-rigid registration of 15 contrast enhancing breast lesions and found volume shrinkage or expansion of up to 20%. Click here for more information.

Volume changes can be reduced either by imposing a locally rigid transformation [5] or by the introduction of a volume preserving regularization term to the registration's optimization scheme. The question remains, however, how to measure the residual registration error since no ground truth is available.

Validation of Registration:
Registration algorithms for CE MR mammography have been developed since 1995. Evaluation of the quantitative performance of these algorithms has, to date, been inadequate. The only plausible method so far has been visual assessment, but this is insufficient in regions of contrast enhancement. Other assessments of performance have relied on simulated deformations that are based on the same transformation model as the registration algorithm, which is likely to bias results. Analysis of the similarity measure itself does not constitute a validation.

• Method: We have presented a novel validation method for non-rigid image registration in [6,7]. It is based on applying the registration to misaligned images, generated from plausible deformations simulated by biomechanical models using finite element methods. Our method allows the estimation of registration accuracy at each point within the volume of interest, in this case the breast. Click here for details.
• Assessment of Biomechanical Breast Model: The ability of the biomechanical breast model to predict breast deformations was assessed in [8] for a range of material models and boundary conditions. Click here for more information.
• Results: Initial validation studies have been conducted to compare the single- and multi-level registration strategies [9] and to assess the usefulness of a volume-preservation constraint [10]. Click here for more information.

This project is a collaboration between:

Prof. David Hawkes, Dr. Julia Schnabel, Dr. Luke Sonoda and Christine Tanner
Computational Imaging Science Group
Division of Radiological Sciences and Medical Engineering
The Guy's, King's and St. Thomas' Schools of Medicine and Dentistry
Guy's Hospital, London SE1 9RT, UK

Dr. Andreas Degenhard, Dr. Carmel Hayes, Prof. Martin Leach
Section of Magnetic Resonance
Institute of Cancer Research
The Royal Marsden NHS Trust
Sutton SM2 5PT, UK

Publications:

[10] C. Tanner, J. A. Schnabel, A. Degenhard, A. D. Castellano-Smith, C. Hayes, M. O. Leach, D. R. Hose, D. L. G. Hill, D. J. Hawkes. Validation of volume-preserving non-rigid registration: Application to contrast-Enhanced MR-mammography. In Proc. Medical Image Computing and Computer-Assisted Interventions (MICCAI 2002), Tokyo, Japan, September 2002, vol. 2489 of Lecture Notes in Computer Science, pp. 307-314, Springer Verlag, 2002.

[9] J. A. Schnabel, C. Tanner, A. D. Castellano-Smith, A. Degenhard, C. Hayes, M. O. Leach, D. R. Hose, D. L. G. Hill, D. J. Hawkes. Finite element based validation of non-rigid registration using single- and multi-level free-form deformations: Application to contrast-enhanced MR mammography. In M. Sonka and J. M. Fitzpatrick, editors, Proc. SPIE Medical Imaging 2002: Image Processing, San Diego, CA, pp. 550-581, 2002.

[8] C. Tanner, A. Degenhard, J. A. Schnabel, C. Hayes, L. I. Sonoda, M. O. Leach, D. R. Hose, D. L. G. Hill, D. J. Hawkes. A comparison of biomechanical breast models: a case study. In Proc. SPIE Medical Imaging 2002: Image Processing, volume 4683, pages 1807-1818. SPIE, 2002.

[7] J. A. Schnabel, C. Tanner, A. D. Castellano-Smith, A. Degenhard, M. O. Leach, D. R. Hose, D. L. G. Hill, D. J. Hawkes. Validation of non-rigid image registration using finite element methods: Application to breast MR images. IEEE Transactions on Medical Imaging, vol. 22(1), 2003. In press.

[6] J. A. Schnabel, C. Tanner, A. Castellano-Smith, M. O. Leach, C. Hayes, A. Degenhard, R Hose, D. L. G. Hill, D. J. Hawkes. Validation of non-rigid registration using finite element methods. In Proc. Information Processing in Medical Imaging (IPMI'01), University of California at Davis, 2001, 18-22 June 2001, vol. 2082 of Lecture Notes in Computer Science, pp. 344-357, Springer Verlag, 2001.

[5] C. Tanner, J. A. Schnabel, D. Chung, M. J. Clarkson, D. Rueckert, D. L. G. Hill and D. J. Hawkes. Volume and shape preservation of enhancing lesions when applying non-rigid registration to a time series of contrast enhancing MR breast images. In Third Int. Conf. on Medical Image Computing and Computer-Assisted Intervention (MICCAI'00), Volume 1935 of Lecture Notes in Computer Science, Springer Verlag, pages 327-337, Pittsburgh, PA, 2000.

[4] D. Rueckert, L. I. Sonoda, C. Hayes, D. L. G. Hill, M. O. Leach, and D. J. Hawkes. Non-rigid registration using free-form deformations: Application to breast MR images. IEEE Transactions on Medical Imaging, 18(8):712-721, 1999.

[3] E. R. E. Denton, L. I. Sonoda, D. Rueckert, S. C. Rankin, C. Hayes, M. Leach, D. L. G. Hill, and D. J. Hawkes. Comparison and evaluation of rigid and non-rigid registration of breast MR images. Journal of Computer Assisted Tomography, 5:800-805, May 1999.

[2] D. Rueckert, L. I. Sonoda, E. Denton, S. Rankin, C. Hayes, D. L. G. Hill, M. Leach, and D. J. Hawkes. Comparison and evaluation of rigid and non-rigid registration of breast MR images. In Proc. SPIE Medical Imaging 1999: Image Processing, pages 78-88, San Diego, CA, February 1999.

[1] D. Rueckert, C. Hayes, C. Studholme, P. Summers, M. Leach, and D. J. Hawkes. Non-rigid registration of breast MR images using mutual information. In First Int. Conf. on Medical Image Computing and Computer-Assisted Intervention (MICCAI'98), Lecture Notes in Computer Science, Springer-Verlag, pages 1144-1152, Cambridge, MA, 1998.

This work was funded by:

The Engineering and Physical Sciences Research Council