Deep Learning-Based Fast Volumetric Image Generation for Image-Guided Proton Radiotherapy

Very fast imaging techniques can enhance the precision of image-guided radiation therapy, which can be useful for external beam radiation therapy. This work aims to develop a deep learning (DL)-based image-guide framework to enable fast volumetric image reconstruction for accurate target localization for treating lung cancer patients with gating, and it is presented in the context of FLASH which leverages ultrahigh dose-rate radiation to enhance the sparing of organs at risk without compromising tumor control probability. The proposed framework comprises four modules, including orthogonal kV X-ray projection acquisition, DL-based volumetric image generation, image quality analyses, and proton water equivalent thickness (WET) evaluation. We investigated volumetric image reconstruction using kV projection pairs with four different source angles. Thirty patients with lung targets were identified from an institutional database, each patient having a 4-D computed tomography (CT) dataset with ten respiratory phases. Considering all evaluation metrics, the kV projections with source angles of 135° and 225° yielded the optimal volumetric images. The patient-averaged mean absolute error, peak signal-to-noise ratio, structural similarity index measure, and WET error were <inline-formula> <tex-math notation="LaTeX">$75\pm 22$ </tex-math></inline-formula> hounsfield unit, <inline-formula> <tex-math notation="LaTeX">$19\pm 3$ </tex-math></inline-formula>.7 dB, <inline-formula> <tex-math notation="LaTeX">$0.938\pm 0.044$ </tex-math></inline-formula>, and −1.3%±4.1%. The proposed framework can rapidly deliver volumetric images to potentially guide proton FLASH treatment delivery systems.