iSpine - Spine Image Fusion

iSpine


Spinal Imaging and Image Analysis

Mission


To provide and validate a method that fuses CT and MR details into a single image.

Challenges


  1. Assessment of CT and MR images must be mentally aligned and fused together.
  2. Detection of changes on unregistered, uncombined images is an error prone task.

Research


Spine Image Fusion via Graph Cuts


Spine Image Fusion: Spine CT and MRI Spine Image Fusion: Spine CT and MRI

We propose a novel CT/MR spine image fusion algorithm based on graph cuts. The algorithm allows physicians to visually interpret corresponding soft tissue and bony detail within a single image, excluding the need for mental alignment and correlation needed when both CT and MR images are required for diagnosis. We state image fusion as a discrete multi-label optimization problem, which can be solved efficiently with graph cuts by alpha-expansion. The proposed energy function balances the contributions of three competing terms: two squared error terms, corresponding to input images with preference to strong MR and CT edges, and a prior favouring smooth solutions by encouraging neighbouring pixels to have similar fused-image values. We further introduce a transparency-labeling formulation, which significantly reduces the computational load. The proposed graph cut fusion guarantees nearly global solutions, while avoiding the pixelation artifacts that affect standard wavelet based methods.[1]

Four existing multi-modality image fusion methods were used for comparison with our graph cuts fusion method. The figure below shows sample input images, including the registered CT and MR inputs and the results of the five fusion methods.

Spine Image Fusion: Spine CT and MRI


Fig. 5. Sample fused Images: top left - input CT, top right input MR, middle fused images: A) averaging method, B) discrete wavelet transform, C) contourlet, D) Piella’s variational method, E) our proposed graph cut method, bottom row: magnified images within the region of interest indicated in the images above.[1]

Approach




Graph Cuts Fusion Method

mr/ct spine image fusion process


Fig. 3. Flow chart of the image fusion process: The MR is registered to the CT image. The CT threshold and its histogram is adjusted to match the MR. Finally the images are fused.[1]

Validation


Several quantitative comparisons over 40 pairs of CT/MR images, acquired from 20 patients, were reported. The 3D MR/CT images were then registered, pre-processed, and fused as sets of 2D images. Validation was completed initially based on the clarity of the detail presented in the fused image, and secondly on a statistical comparison of the bone structure and soft tissue detail transferred to the fused images. Four clinical cases were used to exhibit the benefits of the fusion system.[1]

Collaborators





st. joseph's hospital
lawson health research institute
robarts research
london health sciences centre
general electric
western university
digital imaging group of london