MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Predicting Neurovascular Sparing in Function-Preserving Stereotactic Ablative Radiotherapy (SAbR) of Prostate Cancer: An Innovative Implementation of Deep Learning Technique

N Hassan Rezaeian*, Y Gonzalez , A Leiker , A Laine , R Hannan , M Folkert , R Timmerman , N Desai , X Jia , The University of Texas Southwestern Medical Ctr, Dallas, TX

Presentations

(Sunday, 7/29/2018) 4:30 PM - 5:00 PM

Room: Exhibit Hall | Forum 7

Purpose: The most common quality-of-life (QOL) affected by radical therapy for primary prostate cancer is erectile dysfunction. In contrast to surgical prostatectomy, implementation of rapidly expanding SAbR paradigm in combination with recent advances in radiotherapy techniques for target delineation, using MR/CT, and precise patient setup brings a potential for sparing neurovascular-elements (NEs) and keeping erectile function after radiotherapy. At our institution, we initiated a multi-center randomized phase II trial (POTEN-C) to evaluate the impact of NE-sparing SAbR on a patient-reported sexual QOL endpoint. Predicting a best dose painting that spare unilateral NEs with SAbR while maintaining target coverage and normal tissue constraints for adjacent OARs is essential. In this study we propose to use a deep learning technique to predict the best achievable dose metrics for the OARs and the target.

Methods: Our retrospective study uses 12 patients with low to intermediate risk without high-risk MR lesions within 5mm. Rectal hydrogel-spacer is implanted. A clinically acceptable plan for each patient is generated to satisfy constrains at our institution. We collected a series of geometrical parameters combined with dose metrics in these plans. We used 10/12 of these data sets to train a 3-layer fully connected convolutional-neural-network (CNN). We examined the accuracy of our trained network with the remaining 2 data sets to predict the dose metrics and reconstruct 3D-dose painting.

Results: Plans with unilaterally spared NEs are successfully generated to meet constraints. Our CNN method is able to predict maximum dose to rectum, bladder wall, and NE with differences of <5%,<2%,<5%, respectively. The CNN can also predict volume and circumference based dosimetric data. On average, our CNN method predicts all dose metrics with an accuracy of <7%.

Conclusion: The CNN-based dose predication method is effective to establish a guideline for treatment planning of the complex NE-sparing SAbR of prostate cancer.

Keywords

Not Applicable / None Entered.

Taxonomy

TH- External beam- photons: treatment planning/virtual clinical studies

Contact Email