MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All
  4. Evaluating the Performance of a Generative Adversarial Network-Based (GAN-Based) Architecture for Automatic Detection of Fiducial Markers in Prostate MRI-Only Radiotherapy Images

Evaluating the Performance of a Generative Adversarial Network-Based (GAN-Based) Architecture for Automatic Detection of Fiducial Markers in Prostate MRI-Only Radiotherapy Images

K Singhrao1*, J Fu1, D Ruan2, A Mikaeilian2, N Parikh2, A Kishan2, J Lewis3,(1) David Geffen School of Medicine at UCLA, Los Angeles, CA, (2) Dept. of Radiation Oncology, UCLA, Los Angeles, CA, (3) Cedars-Sinai Medical Center, Beverly Hills, CA

Presentations

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

Room: AAPM ePoster Library

Purpose: markers (FMs) appear as small signal voids in MRI and are often difficult to discern. Existing clinical solutions require multiple MR sequences and/or manual interaction and specialized expertise. Here, we develop a robust method for automatic FM detection in prostate cancer MRI-only simulation images.


Methods: prostate cancer patients who received marker implantation prior to MRI and CT simulation imaging were selected for this study. Multiple-Echo T1-VIBE MRI images were acquired, and patients were stratified based on the presence of intraprostatic calcifications. Ground truth (GT) contours were defined by an expert on MRI using CT images. Training was performed using the pix2pix Generative Adversarial Network (GAN) image-to-image translation package and model testing was performed with five-fold cross validation. An experienced medical dosimetrist and a medical physicist each manually contoured FMs in MRI images. The DICE scores and false discovery rates (FDR) are reported for the GAN-classifier and human observers. Using physician defined contours on CBCT images, the target registration error (TRE) was quantified for the GAN-classifier and human observers.


Results: of implanted FMs were correctly identified using the GAN classifier. Two expert raters correctly identified 97% and 96% of FMs, respectively. The main source of false discoveries was intraprostatic calcifications. The FDR was 8.7%, 2.6% and 1.7% for the GAN classifier, rater-1 and rater-2, respectively. The mean TRE differences between alignments from GAN and human detected markers, and GT were less than 0.8mm±6.1mm. T-tests with a 95% confidence intervals demonstrated that there was no statistical difference between alignments performed using both the human and automatically detected markers and the CT-based clinical standard.


Conclusion: have developed a GAN-based approach to automatically detect and classify FMs in a clinically representative patient cohort. The automatic GAN-model predicted markers can allow for patient alignment with similar accuracy to the current CT-based clinical standard.

Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by Varian Medical Systems, Inc.

Keywords

MRI, Simulation, Image-guided Therapy

Taxonomy

IM/TH- MRI in Radiation Therapy: MRI for treatment planning

Contact Email

Additional Links