Inter-modality non-rigid breast image registration using finite-element method

Abstract

We have investigated a new approach for co-registration of functional PET images with anatomical MR images for breast-cancer detection. This registration is required for multimodality PET-MRI breast image fusion, an approach that could prove useful as an alternative to surgical breast biopsy, following equivocal or difficult-to-interpret mammograms. Our method uses fiducial skin markers placed on the breast that can be observed in PET and MRI. The intermodal geometrical displacements (i.e. differences between PET and MRI marker locations) are estimated from tomographic images. A finite-element-method (FEM) model of breast tissue, based upon an analogy between the orthogonal components of the displacement field and a steady-state heat transfer (SSHT) problem was implemented. This equivalence is valid in that the displacement distribution problem is analogous to the SSHT in solids. The material properties of the breast appear here as coefficients of thermal conductivity. Our model considers the observed marker displacements as FEM temperature loads and distributes them linearly (stepwise) over the entire breast volume. To test the performance of our method, an elastic breast phantom with simulated internal lesions and external markers placed on four meridians was imaged with PET and MRI. We estimated fiducial-registration errors and target-registration errors in the PET-MRI co-registration. The effects of the number, location, and distribution of the markers on the accuracy of the procedure were also investigated. We have established that multimodality breast-image co-registration using our SSHT FEM approach with external fiducial markers is accurate to within /spl sim/5 mm.