Click here to

Session: Multi-Disciplinary General ePoster Viewing [Return to Session]

Deep-Learning-Based Lesion Segmentation On 18F-Fluciclovine PET/CT

T Wang*, Y Lei, E Schreibmann, J Roper, D Schuster, T Liu, A Jani, X Yang, Emory University, Atlanta, GA


PO-GePV-M-332 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: Novel PET radiotracers have been shown to significantly impact radiotherapy decision making, target delineation, and disease-free survival in patients with recurrent prostate cancer after prostatectomy. Treatments are mostly limited to academic centers. We propose a deep learning-based method to automatically segment pelvic node and prostate bed lesions on 18F-fluciclovine (anti-1-amino-3-[18F] fluorocyclobutane-1-carboxylic acid, FACBC) PET/CT images for salvage post-prostatectomy radiotherapy.

Methods: Our proposed method, named hierarchical activation network, consists of three main subnetworks: a fully convolutional one-stage object detection (FCOS) network, a hierarchical module, and a mask segmentation network. The FCOS is employed to detect the volumes-of-interests (VOIs) of the prostate bed and pelvic nodal lesions. The hierarchical module is used to derive an activation map to improve the classification accuracy of the lesion boundary. The mask segmentation network utilizes the detected VOIs obtained from the FCOS network and the activation map obtained from the hierarchical module to perform binary segmentation of lesions within the detected VOIs. To evaluate the proposed method, we retrospectively investigated 106 lesions from 59 prostate cancer patients who had 18F-fluciclovine PET/CT studies. Each dataset has lesions contoured by physicians that served as the ground truth and training targets. The proposed method was trained and evaluated by a five-fold cross-validation strategy.

Results: Across all lesions, segmentation accuracy results are as follows: Hausdorff distance 95th percentile (4.84±3.87 mm); centroid distance (2.08±1.69 mm); volume difference (1.93±8.53 cc); and Dice similarity coefficient (0.68±0.14). Quantitative evaluations show the accuracy of the proposed method and confirm the feasibility of the proposed segmentation method.

Conclusion: The accuracy of the proposed method implies that it has great potential to improve efficiency while mitigating the observer dependence in pelvic and prostate bed lesion contouring of 18F-fluciclovine PET/CT images, thereby potentially making a superior treatment option more widely accessible for patients with recurrent prostate cancer.

Funding Support, Disclosures, and Conflict of Interest: Emory University gets royalties from sale of fluciclovine and they have sponsored projects at Emory


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email