Motivation: This study addresses the need for accurate and efficient hybrid CT and MR image guidance. Current methods require extensive scans from both machines, which can be costly and time-consuming. Goal(s): The aim is to develop a novel imaging framework that can reconstruct CT and MR images from sparse X-ray projections while maintaining image accuracy. Approach: We used a generalizable neural representation model with an anatomy-adaptive layer that learns deformation vectors from a single prior scan, updating its weights based on few X-ray projections for on-treatment reconstruction. Results: The proposed framework outperforms conventional methods in reconstructing CT and MR images with only 10 projections. Impact: The proposed framework enables efficient, low-dose imaging with exceptional tissue contrast, which could transform fields like radiation therapy by reducing scan requirements without sacrificing image quality. This can potentially enable treatment guidance (MR) with simultaneous dose calculation (CT).
Pastor‐Serrano et al. (Tue,) studied this question.