Image Co-Registration (Image Fusion)

1. Defining image fusion
   1. Patient is scanned with CT, MRI, and/or SPECT
      1. CT - spatial resolution is between 5 - 10 mm
      2. MRI - spatial resolution is approximately 1 mm
      3. SPECT - spatial resolution is greater than 10 mm
      4. Dedicated PET - spatial resolution 5 - 10 mm
   2. Accurate image registries (for each type of modality mentioned) with appropriate spatial resolution combines 2 or more of the above data sets to give anatomical and physiological information on the patient
   3. Information gives increased accuracy for the surgeon, radiation therapies, and interventional radiologist
2. Four steps are required in order to fuse a set of images
   1. All data must be transferred to a single computer system
   2. Registration of the 3-D or tomographic data sets must all have a common coordinate system (DICOM compatible as an example)
   3. Transform each set of data to a common coordinate system with appropriate image registries
      1. All must have same coordinates on x, y, and z axis
      2. Spatial resolution must be the same for all sets of images
   4. Display and evaluate images from the combined registered data sets
3. Data transfer using a DICOM compatible software is essential. Many computer/camera manufactures have converted to DICOM compatible images
   1. Hence a Siemen head CT images and an SMV SPECT brain should both be DICOM compatible, allowing image fusion to occur
   2. DICOM images are also be used to send images off site via the internet for review of acquired data
4. Image registration (three methods)
   1. Visual method
      1. The viewer lines up the images for each modality acquired
      2. Error is usually within 4 mm or 3 degrees
   2. Interactive point or landmarks for each modality acquired
      1. At least three points are identified
      2. Raw data and/or slices are then lined up to correspond to each modalities data set
   3. Automated registration
      1. Used to improve the accuracy of the fusion process
      2. Corresponding surfaces are identified for each data set acquired
      3. Computer automatically lines up the images
5. Transformation of data
   1. Usually done via a matrix multiplication of image coordinates
   2. Done in a matrix format
   3. This will include changes in image intensities, gray scales and/or color for each data set
6. Visualization and interpretation
   1. Controlled by the observer
   2. Different sets of images maybe fused together to give anatomical and physiological data
7. There is a site on the internet entitled The Whole Brain Atlas
   1. Link to http://www.med.harvard.edu/AANLIB/home.html
   2. Normal and abnormal brain images can be observed in an image fusion format
   
   

   Return to the Table of Contents

   2/16