Matthew Aquilina Repeatability of Radiomics Features Extracted from Dynamic FET PET Logan Derived

Final Masters Project Presentation Repeatability of Radiomics Features Extracted from Dynamic FET PET Logan-Derived Surrogate VT Maps By: Mr. Matthew Aquilina Abstract Dynamic O-(2-[18F]fluoroethyl)-L-tyrosine ([18F]FET) positron emission tomography (PET) offers a route to richer quantitative imaging measurements in glioma. This thesis evaluates the repeatability of radiomics features, defined here as quantitative descriptors of image intensity, geometry, and texture, extracted from Logan-derived surrogate VT maps within one documented workflow, with reported tables and figures regenerated from the saved Logan-map, tumour region-of-interest, radiomics-feature, and provenance files used for the thesis analysis. The analysis is based on an audited test–retest subset of eight participants (16 scans) drawn from a parent cohort of 24 participants. The default analysis used Logan-derived surrogate VT maps with binWidth = 0.04. The keeper definition combined ICC(1,1)≥ 0.85 with non-ICC exclusions for ROI-volume coupling and proportional bias. This yielded 67 of 106 baseline features above the ICC threshold, 56 after the ROI-volume confound screen, and 51 final keepers. Sensitivity analyses then tested whether those repeatability and screening conclusions persisted under changes to gray-level discretization and image-filter family: 44 features remained keepers across all tested bin widths, and because that all-bin-width set was fully contained within the 51-feature crossfamily overlap, the final consensus set contained 44 canonical features comprising a 39-feature intensity-derived core plus 5 shape/geometry-derived features. The thesis therefore argues for a pre-modelling, stability-first workflow in which downstream modelling begins from this consensus-supported shortlist, treats the shape/geometry-derived features separately, and keeps uncertainty, agreement, and robustness checks visible. ========================= For more information about UWA Medical Physics vist our: Website: https://www.uwamedicalphysics.org Weblog: http://www.uwamedicalphysics.com Facebook: / groups Instagram: / uwa_medical_physics LinkedIn: / medical-physics-uwa-11979b379 Twitter: @MedicalUwa YouTube Channel: / @medicalphysicsuwa =========================