Click here to


Are you sure ?

Yes, do it No, cancel

Predicting Photon Dosimetry of EBT3 Gafchromic Films to Improve Accurate Dosimetry Using Machine Learning Method

YU-SHEN Lin*, E-DA hospital, Kaohsiung, Taiwan (R.O.C.)


(Sunday, 7/29/2018) 3:00 PM - 6:00 PM

Room: Exhibit Hall

Purpose: GAFChromic EBT-3 �lms ( International Specialty Products, Wayne, NJ, USA) is widely used for verification of radiation dose. EBT-3 has great spatial resolution, tissue equivalent and no energy dependence for high energy. The calibration curve of EBT-3 film should be established. Because the various dose-response is observed in the different lot numbers of the film and radiation facilities.In this study, we use machine learning method to build the calibration curve of radiochromic EBT-3 film, and then compare with the calibration curve of FilmQA pro (Ashland, US, 2016).

Methods: Machine learning: Tensorflow database was used in this study to build calibration model. Three pixel values were selected by the multi-dimensional tensor features as follows red, blue and green. In addition, optical density, field size and resolution were evaluated.Measurement: Radiochromic EBT-3 film was placed at 5 cm depth in a solid water phantom, and field size was 5×5. The films were irradiated with a 6 MV photon beam using Varian Clinac® iX System linear accelerator. The films were exposed to radiation dose of 50, 100, 200, 300, 400, 500, 700, 1000 cGy . Epson® Expression® 11000XL scanner was used to detect the density of all films using three spatial resolution of 48, 72, 144 dpi .Comparison:The calibration curve of EBT-3 film was analyzed using FilmQA pro software, and that was compared with the prediction curve of machine learning method.

Results: In this study, EBT-3 film dosimetry was calculated by FilmQA pro software and that was evaluated and analyzed using machine learning method .

Conclusion: The calibration curves were similar between FilmQA pro software and machine learning method analyses. The features of EBT-3 film dosimetry system in machine learning method were successfully verified. Therefore, training data and feature were improved the accuracy of the model in the future.


Absolute Dosimetry, Dose Response


TH- External beam- photons: Quality Assurance - Linear accelerator

Contact Email