Click here to

  1. Home
  2. Meeting Program
  3. All Sessions
  4. Abstracts
Session: [Return to Session]

On the Utilization of a Pseudo-Inverse Consistency Metric to Evaluate Deformable Dose Accumulation Error for MRI-Guide Adaptive Radiotherapy

H Zhong*, J Garcia-Alvarez, K Kainz, X Li, Medical College of Wisconsin, Milwaukee, WI

Presentations

MO-E115-IePD-F2-2 (Monday, 7/11/2022) 1:15 PM - 1:45 PM [Eastern Time (GMT-4)]

Exhibit Hall | Forum 2

Purpose: Understanding uncertainty in deformable dose mapping (DDM) is essential, particularly in MRI-guided adaptive radiotherapy (MRgART) in abdomen where organ deformation is generally severe. The purpose of this study is to evaluate the efficiency of a pseudo-inverse dose reconstruction (PIDR) method in quantification and analysis of the DDM uncertainty for MRgART.

Methods: In PIDR, a pseudo-inverse deformation map was constructed for each DDM performed in MIM (MIM Software Inc.) and used in a forward dose mapping method to deform dose to the same image as the DDM. The DDM uncertainty was then estimated by calculating the difference of the dose distributions generated by DDM and PIDR. To verify the PIDR-based estimation, contours were delineated on 10 daily abdominal MRIs which were registered using both hybrid (H) and contour (C) -only registration algorithms, respectively. Mean doses before and after DDM were calculated on each OAR. The percentages of unique voxels were calculated to evaluate these registrations. With the PIDR doses taken as a reference, the 5mm and 5% gamma passing rates with a 10% threshold were calculated for each DDM and compared with changes in the mean doses and the percentages of unique voxels.

Results: The mean dose changes averaged over all registered OARs are 6.2±2.9% for C-DDM and 8.5±3.9% for H-DDM. The mean dose difference between C-DDM and C-PIDR is 10.8±3.1% and between H-DDM and H-PIDR is 12.6±3.4%. The gamma passing rates of the DDM-reconstructed doses are negatively correlated to the mean dose changes with Pearson coefficient R equal to -0.63 for H-DDM and -0.70 for C-DDM. Furthermore, H-PIDR is more strongly correlated to the percentage of the unique voxels than C-PIDR.

Conclusion: PIDR can be applied to DDMs for MRgART. The difference between DDM and PIDR may serve as a ballpark estimate of DDM error in individual voxels.

Keywords

Deformation, Dose, Registration

Taxonomy

TH- External Beam- Photons: adaptive therapy

Contact Email

Share:

Additional Links