Attention U-net approach in predicting Intensity Modulated Radiation Therapy dose distribution in brain glioma tumor

Today, intensity-modulated radiation therapy (IMRT) is one of the methods used to treat brain tumors. In conventional treatment planning methods, after identifying planning target volume (PTV), and organs at risk (OARs), and determining the limitations for them to receive radiation, the dose distribution is performed based on optimization algorithms, which is usually a time-consuming method. In this article, artificial intelligence is used to acquire the knowledge used in the treatment planning of past patients and to plan for new patients to speed up the process of treatment planning and determination of the appropriate dose distribution. In this paper, using deep learning algorithms, two different approaches are studied to predict dose distribution and compared with actual dose distributions. In the first method, only the images containing PTV and the distribution of the corresponding doses are used to train the convolutional neural network, but in the second one, in addition to PTV, the contours of four OARs are also used to introduce the network. The results show that the performance of both methods on test patients have high accuracy and in comparison with each other almost have the same results and high speed to design the dose distribution. Because the Only-PTV method does not have the process of OARs identifying, applying it in designing the dose distribution will be much faster than using the PTV-OARs method in the whole of treatment planning.