ESTRO 2024 - Abstract Book

S3104

Physics - Autosegmentation

ESTRO 2024

[3] F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, and K. H. Maier-Hein, “nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation,” Nat Methods, vol. 18, no. 2, pp. 203–211, 2021, doi: 10.1038/s41592 020-01008-z. [4] J. Caldas-Magalhaes et al., “Validation of Imaging With Pathology in Laryngeal Cancer: Accuracy of the Registration Methodology,” International Journal of Radiation Oncology*Biology*Physics, vol. 82, no. 2, pp. e289–e298, 2012, doi: https://doi.org/10.1016/j.ijrobp.2011.05.004.

2457

Digital Poster

Comparison of PET based semi-automatic delineation methods with histopathology as reference

Maryam Zarei 1 , Elin Wallsten 1 , Josefine Grefve 1 , Karin Söderkvist 2 , Kristina Sandgren 1 , Joakim Jonsson 1 , Angsana Keeratijarut Lindberg 1 , Erik Nilsson 1 , Anders Bergh 3 , Camilla Thellenberg Karlsson 2 , Anders Widmark 2 , Katrine Riklund 4 , Jan Axelsson 1 , Sara Strandberg 4 , Tufve Nyholm 1 1 Umeå University, Department of Radiation Sciences, Radiation Physics, Umeå, Sweden. 2 Umeå University, Department of Radiation Sciences, Oncology, Umeå, Sweden. 3 Umeå University, Department of Medical Biosciences, Pathology, Umeå, Sweden. 4 Umeå University, Department of Radiation Sciences, Diagnostic Radiology, Umeå, Sweden

Purpose/Objective:

Prostate cancer (PC) is the most commonly diagnosed malignancy among men in Sweden. External beam radiotherapy has made effective contributions to the management of localized prostate cancer. It has been indicated that recurrences following conventional radiation therapy frequently occur at the site of the dominant lesions [1, 2]. Accordingly, focal intraprostatic boost has been suggested, requiring accurate delineation of intraprostatic lesions. Prostate-specific membrane antigen positron-emission tomography (PSMA-PET) is considered a useful clinical tool for imaging of PC. The goal of this study is to compare different PET-based semi-automatic segmentation approaches to histopathology.

Material/Methods:

The research sample included fifteen high-risk patients with Gleason Score 4+4 or higher from the PAMP dataset [3]. [68Ga]-PSMA PET/MR examination with an average injection-to-imaging interval of 61 minutes was done prior to prostatectomy. Prostate lesions identified in hematoxylin-eosin stained whole-mount histopathology were transferred to the coordinate system of the in-vivo PET/MR examination according to the workflow described by Sandgren et al [3]. To compare various semi-automatic segmentation methods with histopathology data, the PET images were trimmed to include only the prostate with a two-pixel margin. Then, manual corrections were applied to eliminate portions that included the bladder zone. To segment intraprostatic lesions in fifteen patients, different approaches and their parameters were varied over a set of values. These approaches included absolute threshold of standardized uptake value (SUV) in the range of 2 to 4.5, relative threshold of 20-50% of intraprostatic SUV max, adaptive threshold, and multi-level Otsu with two (Otsu2) and three threshold (Otsu3) values. The registration process and the implementation of PET-based semi-automatic segmentation methods were performed in Hero software