CBCT-Guided Adaptive Photon and Proton Radiotherapy

Application

Generating synthetic MRI images from CBCT images with deep learning.

Key Benefits

Generate high contrast MRI images from cone-beam CT images.
High quality, high contrast images to enable precise radiation treatment.

Market Summary

Radiation therapy (RT) is one of the most common treatment modalities for cancer and is used in over half of all treatment plans. Over the past decade, the combination of 3D imaging and RT has enabled the precise delivery of radiation to the tumor while sparing healthy tissues (e.g., lung, intestines, and spinal cord). Computer-aided tomography (CT) and X-rays create a 3D map of the tumor and healthy tissues, allowing the oncologist to deliver radiation to the cancer. Cone-beam Computed Tomography (CBCT) is an imaging method used to obtain 3D images of tissues during daily patient treatment. CBCT is inexpensive, accurate, and uses a lower radiation dose than traditional CT but produces lower contrast images. Hence, new technologies are needed to improve the CBCT contrast, leading to superior treatment outcomes and less radiation-induced toxicity.

Technical Summary

Researchers have developed a method to enhance the quality of CBCT (cone-beamed computed tomography) images by producing synthetic MRI images (sMRI) via machine learning. The method consists of training a deep learning algorithm 3D cycleGAN to translate CBCT images into MRI images then capture features on the CBCT images using dense blocks and attention gates. When the system is trained the CBCT images can be transmitted into the model, which will generate patches of sMRI. These patches can then be fused together to reconstruct a whole sMRI image.