We took a trip out to the MD Andersen Cancer Center this morning to talk to Dr. Bud Wendt (a former professor of Image Processing at Rice) to get a brief introduction to Nuclear Medicine and Single-Photon Emission Computed Tomography (SPECT). We viewed a few of the machines which use tomographic data acquisition - a gamma camera, an MRI scanner, and a CAT scanner.
The gamma camera data is the most feasible one for our project. They are used to detect the amount of radiation inside of a person's body after they have the radiation injected. After the radiation spreads out, one can see the different concentrations of gamma radiation which can be detected by a massive camera. There are two cameras located on either side of the patient which rotate around the body to acquire data.
The gamma camera is made up of three layers (diagrams will be attached soon) - the collimator (which collimates the gamma rays entering the camera), the crystal, and the detectors. The collimation of the rays faciliates the reconstruction since we will be dealing with data that comes in perpendicularly. The gamma rays will all be parallel.
The raw data will be a vector of the COUNTS of gamma rays. These counts are usually calibrated with a source of known value to make sure that everything is in order.
To test our reconstruction code, we will need test data. In SPECT, the test data is referred to as "phantom" data. It is basically a circle with shadings of gray - the simplest cross-section.
Back Projection is the method of choice for our reconstruction. We will be given a vector of gamma ray counts and we will map them onto the 2-D plane. Then take another vector from a different angle and map them onto that same plane. And continue until they are all mapped. Then whereever the most amount of overlap is, that is where the highest concentration of gamma radiation is. Remember the PBJ analogy! (Sorry inside joke!) We will need a RAMP FILTER to intensify the peaks and de-intensify the valleys so that the layers are more distinct.
For our purposes, we will be ignoring gamma ray attenuation due to spreading of the rays. In general radiation which is further will be dimmer and lower in resolution in the reconstruction.
The camera takes planar pictures, so when the hospital reconstructs the images, they are 3-D. We will be only dealing with a slice of the data, so our reconstruction will be 2-D.
The Center of Rotation of the machine is not exactly reliable. We would like for the center of the body to be projected into the same column every time, but since the equipment is so heavy, there is really no way to avoid sagging of the instrument or the ground itself. We will ignore this for now as well.
Single-Photon Emission Computed Tomography - A Primer
by Robert J English, CNMT
Image Reconstruction from Porjection - The Fundamentals of Computerized
Tomography
by Gabor T. Herman