ESTRO 2024 - Abstract Book

S3154

Physics - Autosegmentation

ESTRO 2024

Conclusion:

Low DSC and high volume disagreement for the peripheral zone segmentations indicate more training samples are needed for a quality prostate zone segmentation. At least one segmentation quality metric was statistically different for each slice thickness in 4 mm to 8 mm range, this highlights the need to use models trained on images of a similar characteristic to the test dataset and the potential to miss structures when imaging at a large slice thickness. These observations are in line with reported behaviours for autosegmentation quality varying CT slice thickness [3]. Future work will include examining feasibility of training a model at larger slice thicknesses to match available diagnostic scans, with the aim to quantify more rigorously the need to use a standard imaging protocol for pelvic MRI scans.

Keywords: Prostate, autosegmentation

References:

[1] Gibala S, Obuchowicz R, Lasek J, Schneider Z, Piorkowski A, Pociask E, Nurzynska K. Textural Features of MR Images Correlate with an Increased Risk of Clinically Significant Cancer in Patients with High PSA Levels. J Clin Med. 2023 Apr 12;12(8):2836. doi: 10.3390/jcm12082836. PMID: 37109173; PMCID: PMC10146387. [2] Diaz-Pinto, Andres & Alle, Sachidanand & Ihsani, Alvin & Asad, Muhammad & Nath, Vishwesh & Pérez-García, Fernando & Mehta, Pritesh & Li, Wenqi & Roth, Holger & Vercauteren, Tom & Xu, Daguang & Dogra, Prerna & Ourselin, Sebastien & Feng, Andrew & Cardoso, Manuel Jorge. (2022). MONAI Label: A framework for AI-assisted Interactive Labeling of 3D Medical Images. [3] Huang K, Rhee DJ, Ger R, Layman R, Yang J, Cardenas CE, Court LE. Impact of slice thickness, pixel size, and CT dose on the performance of automatic contouring algorithms. J Appl Clin Med Phys. 2021 May;22(5):168-174. doi: 10.1002/acm2.13207. Epub 2021 Mar 29. PMID: 33779037; PMCID: PMC8130223.

3085

Proffered Paper

To curate or not to curate: simple auto-curation for deep learning-based parotid gland segmentation

Victor I.J. Strijbis 1,2 , Oliver J. Gurney-Champion 3,4 , Berend J. Slotman 1,2 , Wilko F.A.R. Verbakel 1,2

1 Amsterdam UMC location Vrije Universiteit, Radiation Oncology, Amsterdam, Netherlands. 2 Cancer Center Amsterdam, Cancer Treatment and Quality of Life, Amsterdam, Netherlands. 3 Amsterdam UMC location Univeristy of Amsterdam, Radiology and Nuclear Medicine, Amsterdam, Netherlands. 4 Cancer Center Amsterdam, Imaging and Biomarkers, Amsterdam, Netherlands

Purpose/Objective: