Songan Mao, Presenter: Nothing to Disclose

Huanmei Wu PhD, Abstract Co-Author: PerkinElmer Inc Varain Inc

George Sandison, Abstract Co-Author: Nothing to Disclose

Shiaofen Fang, Abstract Co-Author: Nothing to Disclose

Generating 3D volume of the tumor and critical organs in real-time during image guided radiation treatment will be of great importance for precise dose delivery and accurate dose calculation. This project will develop an algorithm to generate 3D volume of the tumor and critical organs in real-time for image guided cancer radiation treatment.

The markers had been implanted before the 4DCT images were acquired. For any interested time instance, only the 3D positions of the markers were acquired (target phase). The 3D volume were reconstructed based on the following procedure. First, based on the motion sequence similarity using the combined amplitude and phase information, the most appropriate phase of the 4DCT is chosen as the source phase. The corresponding displacements of markers between the source and target phases are calculated. For each marker, the distance based weight coefficient displacement vectors are computed based on the marker displacement. The smaller Euclidean distance, the larger weight coefficient will be. For each voxel, the weighted marker displacement vector is applied to predict its new positions in the target phase. Iteratively predicting all the voxels will generate the 3DCT at the target phase.

The simulation framework prototype have been implemented. For the validation, a 4DCT phase is randomly chosen as the source image. A non-linear artificial deformation function, which has considered the effect of target movement, rotation, and volume deformation, has been applied to the source 3DCT to generate the ground truth at the target phase. The marker positions at the target phase can also be computed from the deformation function. With the target and source marker positions and the source 3DCT, the iterative morphing approach will predict the 3DCT at the target phase. The voxel level difference between the ground truth and predicted 3DCT is used to assess the simulation accuracy. In one simulation, the average voxel coordinate differences averaged over 5443 voxels on the directions of the x, y, z-axis are 0.31, 1.10 and 0.19 pixels, respectively. The average Euclidean distance between the corresponding voxels is 1.3 pixels.

The proposed algorithm can simulate and accurately predict 3D volume information in real-time situation, which is potentially useful for image guided cancer radiation treatment.

NA

Mao, S, Wu, H, Sandison, G, Fang, S, Real-time Profile Based 3D Tumor Volume Generation in Image Guided Cancer Radiation Treatment.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14013831.html