(Tuesday, 7/31/2018) 1:45 PM - 3:45 PM
Room: Room 207
Purpose: Pediatric internal dosimetry plays a crucial role in modern personalized medicine. The present study combines advanced Monte Carlo (MC), Machine Learning (ML) techniques, and pediatric computational models to assess absorbed doses per organ in clinical practice. The goal is to create a CT dosimetry database which will help clinicians to predict absorbed doses during pediatric CT acquisitions.
Methods: A case study is presented for validating our methodology. The GATE MC toolkit was used for modeling a multislice helical CT scanner. The scanner was validated with the standard CTDI phantoms. A series of pediatric computational models in the 5–14 years age range were used to simulate realistic helical chest, abdomen/pelvis and head CT protocols to calculate the absorbed doses per organ and the dose variations due to varying anatomies. Specific characteristics of each phantom are collected creating a dataset that will be used to feed a supervised ML algorithm for matching each patient to the most similar phantom.
Results: The CT scanner model was validated against experimental CTDI measurements with statistical differences being lower than ~11%. First, absolute (mGy) and normalized-to-CTDIvol absorbed doses (mGy/mGy) were correlated with patient weight and effective diameter for the fully irradiated organs using linear regression. Contribution to effective dose was estimated for all organs of interest of each model and each protocol. Dose variations for different anatomical models under the same CT protocol are presented.
Conclusion: Anatomical differences, in children in similar age and/or weight, result in differences of up to 100% in specific organs. The proposed approach optimizes the dosimetry assessment of a pediatric patient who is going to undergo a clinical CT exam. The effective use of the proposed study requires the extension of the dosimetry database, incorporating more CT scanner models, and a large dataset of pediatric models, that will be simulated for different protocols.
Authors:
Theodora Kostou, Panagiotis Papadimitroulas, Dimitris Mihailidis, Ioannis Kopsinis, George C. Kagadis
Funding Support, Disclosures, and Conflict of Interest: This study is part of a project that has received funding from the European Union Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 691203.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
