Radiation Physics

PV QA 3 - Poster Viewing Q&A 3

TU_20_3312 - Design and Evaluation of a Semi-Automated Algorithm for Segmentation of Anti-[18F]FACBC (Fluciclovine F18) PET Images for Post-Prostatectomy Radiation Therapy.

Tuesday, October 23
1:00 PM - 2:30 PM
Location: Innovation Hub, Exhibit Hall 3

Design and Evaluation of a Semi-Automated Algorithm for Segmentation of Anti-[18F]FACBC (Fluciclovine F18) PET Images for Post-Prostatectomy Radiation Therapy.
E. Schreibmann1, D. M. Schuster2, P. R. Patel1, J. W. Shelton3, S. Cooper2, H. Raghuveer2, and A. Jani1; 1Department of Radiation Oncology, Winship Cancer Institute of Emory University, Atlanta, GA, 2Emory University, Atlanta, GA, 3Department of Radiation Oncology, Winship Cancer Institute at Emory University, Atlanta, GA

Purpose/Objective(s): Anti-[18F]FACBC (Fluciclovine F18) PET-CT, or FACBC, is an FDA-approved radiotracer used for imaging local and extra-prostatic disease in recurrent prostate cancer. Since most abnormalities imaged with this technique are irregular and adjacent to enhancing normal physiology, manual target delineation is tedious and time-consuming. In the following, we present a novel semi-automated segmentation algorithm tailored to the particularities observed in these datasets.

Materials/Methods: An algorithm was developed and tested on a 59-lesion set imaged with FACBC as part of an ongoing randomized trial at our institution. There was wide heterogeneity across patients in lesion size, location, appearance, and pattern of enhancement that hindered use of segmentation based on SUV thresholding. In the standard approach, time-consuming manual edits were needed to erase normal–appearing anatomy around the lesion, followed by a selection of a suitable iso-SUV level upon review by an expert physician. Our algorithm eliminates these drawbacks by first using the Otsu algorithm [a thresholding method that converts a grayscale image to a binary one] to automatically detect the SUV intensity that isolates the lesion from the surrounding background anatomy. This employed a fast marching algorithm to evolve a level set from a user-defined seed point, with the Otsu partitioning used as a speed map for the level set evolution. From a practical point of view, the user simply clicks on the desired enhancement, and the algorithm automatically detects the optimal threshold and isolates it from similar enhancements. Our algorithm was quantitatively validated by comparing the segmentations obtained with the proposed and standard approaches using well-accepted evaluation measures.

Results: The mean discrepancy between compared segmentations measured with the Hausdorff distance was 1.42 mm (standard deviation of 1.57mm), while their similarities were measured at 0.53 (0.20) with the Dice coefficient and 0.38 (0.19) with the Jaccard Index. The segmentations overlapped at 0.53 (0.2) and the volume similarity was -0.54 (0.76). Given that most lesions in the dataset are less than 5 mm in diameter, these results document good agreement between the manual and semi-automated approaches. Additionally, times are reduced from 10-20 minutes of editing to 2-3 seconds with our algorithm.

Conclusion: The developed algorithm simplifies the integration of FACBC imaging into the clinical workflow, providing an efficient segmentation solution that is able to isolate cancer-related lesions despite the characteristic tumor heterogeneities observed across these datasets. Our algorithm may be generalizable to other PET radiotracers.

Author Disclosure: E. Schreibmann: Research Grant; Varian Medical Systems. Honoraria; Varian Medical Systems. D.M. Schuster: None. P.R. Patel: None. J.W. Shelton: None. S. Cooper: None. A. Jani: Advisory Board; Blue Earth Diagnostics. Travel Expenses; Blue Earth Diagnostics.

Send Email for Eduard Schreibmann


Assets

TU_20_3312 - Design and Evaluation of a Semi-Automated Algorithm for Segmentation of Anti-[18F]FACBC (Fluciclovine F18) PET Images for Post-Prostatectomy Radiation Therapy.



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Design and Evaluation of a Semi-Automated Algorithm for Segmentation of Anti-[18F]FACBC (Fluciclovine F18) PET Images for Post-Prostatectomy Radiation Therapy.