Machine Learning

Dose Prediction with U-net: A Feasibility Study for Predicting Dose Distributions from Contours using Deep Learning on Prostate IMRT Patients

Tagged: , ,

This topic contains 0 replies, has 1 voice, and was last updated by  arXiv 1 year, 10 months ago.


  • arXiv
    5 pts

    Dose Prediction with U-net: A Feasibility Study for Predicting Dose Distributions from Contours using Deep Learning on Prostate IMRT Patients

    With the advancement of treatment modalities in radiation therapy, outcomes haves greatly improved, but at the cost of increased treatment plan complexity and planning time. The accurate prediction of dose distributions would alleviate this issue by guiding clinical plan optimization to save time and maintain high quality plans. We have developed a novel application of the fully convolutional deep network model, U-net, for predicting dose from patient contours. We show that with this model, we are able to accurately predict the dose of prostate cancer patients, where the average dice similarity coefficient is well over 0.9 when comparing the predicted vs. true isodose volumes between 0% and 100% of the prescription dose. The average differences in mean and max dose for all structures were within 2.3% of the prescription dose.

    Dose Prediction with U-net: A Feasibility Study for Predicting Dose Distributions from Contours using Deep Learning on Prostate IMRT Patients
    by Dan Nguyen, Troy Long, Xun Jia, Weiguo Lu, Xuejun Gu, Zohaib Iqbal, Steve Jiang
    https://arxiv.org/pdf/1709.09233v1.pdf

You must be logged in to reply to this topic.