(Sunday, 7/29/2018) 3:00 PM - 6:00 PM
Room: Exhibit Hall
Purpose: The conventional radiation treatment planning (RTP) system uses a dose calculation algorithm. Normally, it takes long time for dose calculation in the clinical field. The purpose of this study is to develop a prototype algorithm applied the deep learning technique to achieve more fast radiation treatment plan.
Methods: For a design of an artificial neural network, we establish the RTP method by using multi-layer neural network model with some modifications. When a new input (patient data) data has been uploaded, it assigns a proper treatment plan through a supervised learning technique from a database. This mechanism leads to faster and more correct treatment plan creation than the conventional dose planning based on the either algorithm or Monte Carlo. This database was constructed by using a lot of results from Monte Carlo simulation operation. And it was interlocked with the deep learning algorithm.
Results: We designed the algorithm that apply deep learning technique. The algorithm is composed of two sections which is made up of four layers. And we acquired dose volume histogram (DVH) by using deep learning algorithm. The dose of planning target volume (PTV) was sufficiently deposited and the dose of organ at risk (OAR) was less deposited. In comparison with each area, the area under the PTV curve was increased as 14.1% compared to the area under the initial PTV curve and the area under the OAR curve was decreased as 51.5%. In addition, the acquisition time of each treatment plan was averagely 3 – 8 second.
Conclusion: In this study, we successfully created the treatment plan by accessing the dose distribution database. Although we confirmed fast speed for acquiring the treatment planning, the accuracy evaluation should be progressed by the additional studies. In terms of time, we succeed greatly fast acquisition of the proton therapy plan.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
