Evaluation of proton tomography measurements with neural networks

Dr. Papp Gábor (ELTE), Bíró Gábor (Wigner FK), Feiyi Liu (ELTE), Xiangna Chen (CCNU, Wuhan), Dudás Bence (ELTE), Misur Patricia (ELTE) (2021.06.01-08.31)

Abstract: One effective way to kill localized cancerous tumors that are not accessible by surgery is radiation treatment with a proton (or heavier ions as He or C, respectively). In the process, one treatment is usually sufficient compared to conventional radiation therapy, since the proton is very well focused, (with an accuracy of 1 mm, heavier ions with even greater accuracy). However, because the in matter penetration profiles of proton and gamma rays are different, CT tomography does not calibrate the proton beam and does not allow accurate device alignment, resulting in practice in treatment that is far less accurate than the theoretical limit. Greater accuracy can be achieved by proton tomography using a proton beam used for treatment at a higher energy. To detect particles passing through the patient, we developed a detector system based on ALPIDE chips and CERN technology in the framework of the international pCT collaboration (https://wiki.uib.no/pct/index.php/Main_Page). Because processing of the detector signals is a time-consuming process, we want to speed it up by using a neural network: the goal is to develop and train a neural network that can tell the direction and energy of protons leaving the body based on detector signals. By measuring these at several angles, a tomographic image of the examined area can be obtained and the data required for the treatment can be calculated.

Next Post Previous Post