Unit:
Κατεύθυνση Ιατρική Φυσική-ΑκτινοφυσικήLibrary of the School of Health Sciences
Supervisors info:
Δαμηλάκης Ιωάννης, Καθηγητής, Ιατρική Σχολή, Πανεπιστήμιο Κρήτης
Περισυνάκης Κωνσταντίνος, Καθηγητής, Ιατρική Σχολή, Πανεπιστήμιο Κρήτης
Μαρής Θωμάς, Καθηγητής, Ιατρική Σχολή, Πανεπιστήμιο Κρήτης
Original Title:
Πρόβλεψη δόσης ακτινοβολίας ασθενών που υποβλήθηκαν σε υπολογιστική τομογραφία κωνικής δέσμης με χρήση τεχνητής νοημοσύνης
Translated title:
Radiation dose prediction of patients that underwent cone beam computed tomography using artificial intelligence
Summary:
Cone beam computed tomography (CBCT) developed soon after the development of computed
tomography address the acquiring tomographic projections with only one rotation of the gantry.
Early attempts for the development of a CBCT system started in the decade of 1970. Currently,
use of CBCT is widely spread, mostly for dental examinations, and radiotherapy.
In conventional radiotherapy systems, patient’s positioning is done mostly using laser pointers
and the monitoring of his position is achieved using fan beam CT, which does not provide real
time treatment monitoring for position improvement. These problems are solved with built-in
CBCT systems.
The disadvantage of this technique is that the patient dose is increased, something that makes
the need for dosimetry more than imperative. A number of methods have been developed to
calculate dose based on measurements or computations. Computed Tomography Dose Index
(CTDI) is the most prominent dose calculation method, which is based on standardized exposure
protocols and homogeneous cylindrical phantoms to produce an estimate of absorbed dose. This
method can lead to a deviation between the real dose and the calculated, due to anatomical and
elemental composition differences between human body and phantom.
Another method of dosimetry is Monte Carlo simulations, which allows for more personalized
dose estimation. Examination images and scan parameters are inserted in a software and the
dose is calculated pixel by pixel.
For the above reasons, the need for personalized and real time dosimetry is implied that considers the specific patient body characteristics, the exposure settings and the system used. The advert of artificial intelligence in the medical physics field can be exploited towards for this purpose.
In this dissertation radiographic and CBCT thorax and abdomen examinations were simulated
with the Monte Carlo technique, in order to calculate the dose that the patient’s organs of interest have absorbed. The results for chest CBCT examinations, in combination with other dose predictors, were used for the development and training of an artificial intelligence model. The simulations were done using Impact MC software and examinations that were occurred in diagnostic and radiotherapy CT in University General Hospital of Heraklion in 2020.
The aim of this algorithm was to predict in real time the dose of each patient in specific organs
of interest.
Main subject category:
Health Sciences
Keywords:
Dose, Radiation, Cone beam computed tomography, Dosimetry, Artificial intelligence
