Artificial Intelligence (AI)-Based Segmentations vs Manually Adjusted AI-Based Segmentations as a Pre-Processing Step in Whole-Body PET Dosimetry Calculations

Vries, Pleun de (2023) Artificial Intelligence (AI)-Based Segmentations vs Manually Adjusted AI-Based Segmentations as a Pre-Processing Step in Whole-Body PET Dosimetry Calculations. Bachelor's Thesis, Life Science and Technology.

Preview	Text bLST_2023_VriesPde.pdf Download (999kB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (141kB)

Abstract

Positron emission tomography (PET) with radiolabelled monoclonal antibodies, known as immunoPET, is a promising method for non-invasive tumour detection. The evaluation of newly developed immunoPET tracers, like 89ZED88082A, is necessary to assess radiation doses and ensure safety. Currently, organ segmentation is carried out manually, which is not only time-consuming but also subject to variability. The current study aims to compare AI-based segmentations to manually adjusted AI-based segmentations, aiming to reduce analysis time and variability in dosimetry assessments. Organ volumes and estimated effective doses were obtained using the inhouse developed Biodistribution tool, Residence Time Calculator and OLINDA/EXM. The impact of segmented organ volumes on estimated effective doses was evaluated, showing a mean absolute percentage error of 2.51% for organ volumes and 0.80% for estimated effective doses, indicating slight variations between the two methods. The analysis indicated that smaller organs exhibited relatively higher errors, which was supported by the average Jaccard indexes calculated per organ. These findings highlight the challenges and potential of AI-based segmentation methods in clinical settings, where accurate organ segmentation is crucial for dose estimation and radiation protection.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Tsoumpas, C. and Sluis, J. van
Degree programme:	Life Science and Technology
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	06 Jul 2023 09:08
Last Modified:	06 Jul 2023 09:08
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/30194

Actions (login required)

View Item