Daniel Linneman Saenz MS, BS, Presenter: Nothing to Disclose

Venkata Chebrolu PhD, Abstract Co-Author: Nothing to Disclose

Ashley Anderson, Abstract Co-Author: Nothing to Disclose

Scott K. Nagle MD, PhD, Abstract Co-Author: Stockholder, General Electric Company

Sean B. Fain PhD, Abstract Co-Author: Research Grant, General Electric Company Research Consultant, Marvel Medtech, LLC

Bhudatt R. Paliwal PhD, Abstract Co-Author: Nothing to Disclose

Respiratory motion limits the precision of radiation therapy of lung cancer with range of motion often a centimeter or more. Respiratory gating and breath-holds mitigate motion but do not predict the tumor location during radiotherapy treatments. MRI-guidance is an emerging tool with soft tissue contrast superior to CT that enables quantitative imaging during treatment for adaptive, image guided radiotherapy. We have developed a robust, automatic segmentation algorithm well suited for contouring tumors using novel four-dimensional MRI during free breathing.

4D-MRI was acquired using a 3D radial trajectory with ultrashort echo time (UTE). The summed signal across the volume (reflecting the amplitude of the center of k-space) serves as a surrogate measure of lung volume allowing tracking of diaphragm and hence tumor position with a temporal resolution equal to the TR interval (~4.2 ms). This approach was used to bin the radial projections into different groups based both upon lung volume and on direction of motion (inspiration or expiration). The tumor could then be identified automatically using the morphological processing and successive localization (MPSL) segmentation. The MPSL segmentation algorithm first creates binary images through image intensity thresholds and then uses erosion and dilation to separate tumors from other structures to identify the tumor based on volume. In this work, the MPSL algorithm is tested on pre-existing patient data and with an MRI/CT compatible lung phantom with a moving target.

Automatic tracking quantified the motion and volume of a patient’s lung tumor. Motion was observed on the order of centimeters. The simulated motion in a phantom was also quantified with a magnitude of 34 mm with a volume of 20.6 cc. Furthermore, the center of k-space in a human subject has been studied and compared with a respiratory bellows data to find agreement in the waveforms and reconstructions.

The MPSL algorithm is capable of providing quantitative information about a target’s position, volume, and structure aided by 4D-MRI and suggests the ability for real-time guidance with k-space information.

To take advantage of image guided radiotherapy (IGRT), accurate and rapid segmentation of daily images is needed to quantify changes in tumor position and trajectory.

Saenz, D, Chebrolu, V, Anderson, A, Nagle, S, Fain, S, Paliwal, B, Robust Automatic Segmentation for Retrospectively Binned 4D-MRI Guided Radiotherapy.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12023361.html