A Hermite-Based Method for Bone SPECT/CT Image Fusion with Prior Segmentation

Abstract

Multimodal medical image fusion is an interesting application that has become essential in many tasks of medical diagnosis. The combination of medical images acquired from different modalities might significantly contribute to improve the diagnosis and detection process of several diseases. SPECT (Single Photon Emission Computed Tomography) and CT (Computed Tomography) are frequently fused for illness detection tasks. In this work, we have developed a novel scheme for bone SPECT/CT image fusion based on the Hermite transform (HT). It consists of a powerful tool which is able to decompose an image into a set of coefficients defined in the space of the Hermite polynomials. Three main stages are performed for the fusion process: (1) Input images are decomposed using the HT, (2) A fusion rule is applied to combine the resulting coefficients, and (3) The inverse HT is computed to obtain the fused image. Since we are interested in providing a technique focused on evaluating the bone structure, in the fusion rule we introduce prior information based on a previous segmentation obtained from the CT data. Several studies were used for performance assessment.