MRI-Based Treatment Planning in Brain Radiation Therapy: Validation of An Atlas-Based Pseudo-CT Generation Method
D Pinnaduwage1*, A Sudhyadhom1, M Descovich1, K Hwang2, G Novak3,J Chen1, J Pouliot1, (1) Department of Radiation Oncology, University of California San Francisco, San Francisco, CA, (2) MR Applied Science Laboratory, GE Healthcare, Houston, TX, (3) Advantage Workstation Applications, GE Healthcare, Szeged, HungarySU-D-WAB-2 Sunday 2:05PM - 3:00PM Room: Wabash Ballroom
Purpose: To quantify the dosimetric accuracy of an atlas-based pseudo-CT (p-CT) generation method in MRI-based treatment planning, and to evaluate the influence of specific MRI sequences on p-CT quality.
Methods: For six brain IMRT patients, a p-CT was generated from an MR image set previously used for target delineation. A research version of Advantage SIM™ software (GE Healthcare) was used to produce the p-CT. The software uses body segmentation and bone enhancement on the input MRI, to produce a p-CT image with a contrast similar to that of a bone atlas generated offline. The IMRT plan including the isocenter was copied on to the p-CT and the dose was calculated in Pinnacle™. The two dose distributions were compared using (1) gamma analysis, and (2) differences between the average DVH for the target volume, and tissue contour.
Results: The p-CT visually co-registered well with both the MRI and conventional CT in four cases where Axial T1 Spin Echo or Fast Spoiled Gradient-Recalled Echo images were used. For these cases, the average Gamma Index 3%/3mm pass rates were 96.85%, 95.58% and 95.40%, in the axial, coronal and sagittal planes. On average, the absolute value difference between the average DVHs was 1.74 (max: 6.97) and 0.95 (max: 16.7) percentage points, for the tissue contour and target volume, respectively. Two of the MR image sequences (Axial T2 Flair and Axial T1 Fat Sat images) resulted in poor quality p-CTs and were not used for dose comparison.
Conclusion: A good agreement was seen between the p-CT and CT dose distributions in this preliminary study, showing the potential for using MRI for both target delineation and dose calculation for brain treatments, with the clinical integration of an automated p-CT generation method, and appropriate MRI sequences.