Medical Volume Reconstruction


Ken Chidlow

We are looking at ways to speed up SPECT data reconstruction based on the highly accurate EM algorithm using texture mapping hardware. The SPECT data is captured by a gamma camera that needs to sit in one position for about 3 minutes to gather enough photons comming from the radioactive subject for a single image. This means that when reconstructing SPECT images, we have less camera angles to work with then we would if we had MRI or CT data. MRI and CT datasets are reconstructed using the popular filtered backprojection(FBP) method. We cannot use FBP method when reconstruction SPECT as FBP needs many camera images for accurate reconstruction. Iterative algorithms can accurately reconstruct data with fewer images, but are very computationally expensive. We are looking at implementing an iterative algorithm using graphics hardware to speed up this reconstruction.