MENU

Click here to

Ă—

Are you sure ?

Yes, do it No, cancel

A Deep Learning-Based Interactive Software Tool to Assist Physicians Revising CTV Contours to Achieve Balanced Tumor Coverage and Organ Sparing

A Balagopal*, D Nguyen, m mashayekhi, M Lin, A Garant, N Desai, R Hannan, Y Weng, X Gu, S Jiang, UT Southwestern Medical Center, Dallas, TX

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: radiotherapy, planning tumor volume(PTV) encloses the clinical target volume(CTV) with margins to account for possible uncertainties in beam alignment, patient positioning and organ motion. Ideally, the margin should be determined solely by the magnitudes of the uncertainties involved. In practice, physicians usually consider doses to nearby healthy tissues when deciding on the size of margin. This results in a smaller CTV being treated which is not re-analyzed by physicians to ensure all adverse areas are included. Since treatment planning is time consuming, a real-time feedback does not exist for physicians to estimate the optimal CTV that provides the desired balance between tumor coverage and toxicity and includes all the disease. We propose a deep learning based tool that performs real-time dose prediction and estimates the dose to tumor and organs at the CTV segmentation and revision stage thus helping physicians’ make a decision on the optimal CTV for a particular patient.
Methods: post-operative prostate cancer patients treated with VMAT were used for this study. 40 patients were set aside for testing. A 3D-UNet with Residual connection was used as the model. Inputs to the model are organ distance-maps and PTV masks. In the organ distance-maps, the value of a voxel is its minimum distance to PTV boundary. Value is positive for organ voxels that lie outside and negative for those inside the PTV.
Results: model is capable of predicting the dose distribution in <0.7s for 1 patient. Averaging across all organs-at-risk, our model is capable of predicting the organs-at-risk max dose and mean dose within 1% and 1.3% respectively of the prescription dose on the test data.
Conclusion: real time dose prediction tool capable of predicting optimal dose prediction for a particular CTV with a specific margin was developed which enables physicians to ensure an optimal CTV is treated.

Funding Support, Disclosures, and Conflict of Interest: Funding Support: VARIAN Research Grant

Keywords

Dosimetry, Radiation Therapy, Segmentation

Taxonomy

Not Applicable / None Entered.

Contact Email