The Use of Inverse Phase Fourier Image to Accommodate Intensity Inhomogeneities in Medical Image Registration

Medical image registration is generally faced with the confounding effect of spatially dependent intensity variations. This can be the case when images have been acquired using the same imaging modality, for example, in magnetic resonance imaging and also when using various histology and staining processes. We propose the application of an intensity-invariant dense feature extraction method through the use of phase Fourier transforms. The approach allows medical images containing intensity in homogeneities to be aligned and warped as part of a feature-based registration technique. Registration performance was evaluated on mono-modality and multi-modality data, namely magnetic resonance and histology images. Qualitative and quantitative validation was conducted with respect to two established image intensity correction methods.