Encrypted login | home

Program Information

Deformation Mapping and Shape Prediction with 3D Tumor Volume Morphing


S Mao

S Mao1*, H Wu1 , S Fang1 , M Lu2 , (1) Indiana University-Purdue University, Indianapolis, IN, (2) PerkinElmer Medical Imaging, Santa Clara, CA

Presentations

SU-E-I-21 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

Purpose: Tumor deformation occurs with patient respiratory motion and the mapping of the 3DCT images will be of great help. This project will develop an innovative approach to define the iso-surface feature points and 3D volume mapping.

Methods: Previous acquired 4DCT images are retrospectively used with the following steps. First, one source 3DCT is selected based on motion pattern similarity. Second, the minimized bounding boxes of the tumor on both the source and target phases are derived based on tumor iso-surface. The third step is to select the feature points (landmarks) on the source tumor iso-surface using the bounding box. Fourth, the nearest tumor iso-surface intensity and position of both the source and target phases are projected to the six 2D planes, and then the corresponding landmarks are mapped from the source phase to the target phase based on image template matching algorithm. Landmarks alignment was applied with a preset displacement. Last, the entire tumor volume are mapped from the source phase to the target phase with modified Shepard morphing method.

Results: A prototype has been developed and preliminary experiments have been performed. The mapping results are evaluated with the similarity of the image intensity histogram and the displacement of the tumor volume. For example, one set of simulated used phase 0 of 4DCT as the source phase and try to volume mapping for the phase 7, 8 and 9. The average landmark intensity differences of the predicted and actual 3D volume are 1.09, 1.36, and 1.08 for the phase 7, 8, and 9, respectively, with the average stdev of 0.82. The average volume displacement is between the calculated and actual 3D is 1.41 (unit).

Conclusion: The proposed approach allows predicting the tumor shape and volume, which is potentially useful for image guided cancer radiation treatment.


Contact Email: