Abstract
The purpose of this work is to evaluate the practicality of a treatment planning method based only on magnetic resonance imaging (MRI) for radiotherapy. Many investigators have demonstrated that dose escalation with three-dimensional conformal radiation therapy (3DCRT) and intensity-modulated radiation therapy (IMRT) potentially increases the tumor control rate while keeping complication risk at a reasonable level. As dose levels are increased, the precise information of target location and size and the accuracy of dose delivery become crucial. Magnetic resonance imaging (MRI) provides superior image quality for soft-tissue delineation over computed tomography (CT) and is widely used for target and organ delineation in radiotherapy for treatment planning. The main drawback of this modality for treatment planning is the lack of electron density information in the MR images. In this study, we segment two sets MRI images of brain by FCM (fuzzy c-means clustering) method as a target image. To overcome the limitation of MRI in dose calculation, we assigned electron density values to typical anatomical structures. We use the non-linear convert method (Artificial Neural Networks) to transform MRI to Homemade CT (HCT) images. Our results show that the dose differences between HCT images and real CT images are within ±2% in different depths, photon energy, and field sizes. Compared with the homogeneous images, HCT-based treatment planning revealed the more accurate dose and dose distribution. In addition, HCT images can also provide DRR (digitally reconstructed radiograph) for radiography simulation.