MO-F-BRB-01: Extracting Material Information from the CT Numbers by Artificial Neural Network for Use in the Monte
S Sina1,2,3, R Faghihi1,2,3, A Meigooni1,2,3
1School of Mechanical Engineering, Shiraz university, Shiraz, Iran.
Purpose:
The Artificial Neural Networks (ANNs) are useful in solving nonlinear processes, without the need to mathematical models of the parameters. Since the relationship between the CT numbers and material compositions is not linear, we can use the AANs for tissue density calibration. The aim of this study is to obtain the composition and mass density of different tissues which are necessary in Monte Carlo simulation of different tissues in brachytherapy treatment planning using ANNs.
Methods:
The ANNs were used for mass density calibration. First, the density and composition of several tissues of the body, along with their corresponding CT numbers are used as the training samples. After the network is trained, it would give us the material information, i.e. mass density, and material composition corresponding to each CT numbers. The tissue compositions and densities predicted by the ANN for each CT number, were compared with the real values of such parameters. The tissue parameters predicted by the ANN were used as the phantom materials for obtaining the dose at different distances from Pd-103, and Cs-137 brachytherapy sources. Finally the dose at different distances of the real phantoms were compared with dose around the phantoms predicted by ANN.
Results:
The ANN used in this study, can predict the material compositions of different tissues precisely. For example, it can give the mass densities of bone, water, and muscle with the percentage differences of 0.62%, - 1.1%, and 0.33% respectively. Comparing the dose distribution inside the water phantom predicted by Artificial Neural Networks and the real water phantom, shows the percentage difference of less than 0.7% and 2% for Cs-137 and Pd-103 respectively.
Conclusions:
The ANNs are applicable in determination of tissue parameters from the CT images data, and the material compositions and density obtained by this methods can be used for material definition in Monte Carlo simulations.
More Related Videos
10:24Neutron Radiography and Computed Tomography of Biological Systems at the Oak Ridge National Laboratory's High Flux Isotope Reactor
Published on: May 7, 2021
04:09Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
Published on: October 10, 2018