Cone beam computed tomography (CBCT) is routinely used in image-guided radiation therapy (IGRT) for patient positioning purpose. Frequent and accurate CBCT is desired to ensure setup reproducibility, which is critical for planning target volume margin reduction and hence reducing dose to normal tissue and permitting dose escalations. CBCT also plays a key role in adaptive radiotherapy (ART) by supporting the up-to-date patient geometry for treatment replanning. Two major issues hinder wide applications of CBCT in IGRT and ART. 1) Excessive imaging dose. The imaging dose up to 100~300 cGy per treatment course elevates risks of secondary cancer and genetic defects. 2) Low image quality mainly due to scatter. Scattered x-ray photons at the image detector reduce CBCT contrasts and introduce image errors up to 350 HU. Besides degrading patient setup accuracy, scatter artifacts prevent CBCT from quantitative applications in ART, e.g. CT-to-CBCT deformable registration and dose calculations. Over the years, these two problems have been separately addressed. Iterative algorithms, particularly of compressed sensing (CS) type, show promise on reconstructing CBCT with undersampled data. It alone, however, cannot address the scatter problem. Although measurement-based methods using a beam blocker directly probe scatter accurately and robustly, partially missing primary data causes large reconstruction errors. In this project, we propose to solve the two problems in a unified framework. Specifically, we will use an innovatively designed rotating beam blocker to randomly block x-ray measurements within each projection. The generated shadow area allows scatter measurement and removal. The random undersampling yields a projection matrix with favorable mathematical properties that permit high-quality CBCT reconstruction under CS framework. Preliminary studies have demonstrated the feasibility and effectiveness of this approach. The goal of this project is to develop and optimize the proposed system and to demonstrate its advantages in dose reduction (more than 90% compared to the clinical standard scan) and quality improvement (<20 HU error), as well as its clinical impacts on IGRT and ART. We will pursue three Specific Aims (SAs). SA1. Develop a software system to support the proposed workflow. SA2. Research, design and optimize the beam blocker via Monte Carlo simulations and phantom experiments. SA3. Evaluate clinical impacts in patient studies. In contrast to conventional approaches that separately addressed the imaging dose and scatter problems, our method will solve them in a unified framework with seamlessly combined strengths of two state-of-the-art techniques while eliminating their drawbacks. It will empower IGRT and ART with a novel and practical CBCT approach with substantially improved image quality and reduced imaging dose. Although focusing on CBCT in radiotherapy, the method can evolve into a standard solution for other volumetric CT systems. For non-cancer patients, it is more crucial to reduce imaging dose while maintaining image quality. As such, our work is expected to benefit almost all radiology patients.
This project will develop and validate a novel cone beam CT (CBCT) data acquisition and reconstruction scheme for image-guided radiation therapy (IGRT) to solve the high imaging dose and scatter contamination problems in a unified framework. Randomly sampling x-ray projection data achieved through a rotating beam blocker allows dose reduction and scatter measurement and a compressed-sensing based iterative reconstruction algorithm is used to reconstruct CBCT from the scatter-corrected undersampled data. The high- quality CBCT with reduced radiation dose will facilitate patient setup in IGRT and treatment adaptations in adaptive radiotherapy.
| Shi, Linxi; Vedantham, Srinivasan; Karellas, Andrew et al. (2018) The role of off-focus radiation in scatter correction for dedicated cone beam breast CT. Med Phys 45:191-201 |
| Shen, Chenyang; Gonzalez, Yesenia; Chen, Liyuan et al. (2018) Intelligent Parameter Tuning in Optimization-Based Iterative CT Reconstruction via Deep Reinforcement Learning. IEEE Trans Med Imaging 37:1430-1439 |
| Gao, Hewei; Zhu, Lei; Fahrig, Rebecca (2017) Virtual scatter modulation for X-ray CT scatter correction using primary modulator. J Xray Sci Technol 25:869-885 |
| Shi, Linxi; Vedantham, Srinivasan; Karellas, Andrew et al. (2017) X-ray scatter correction for dedicated cone beam breast CT using a forward-projection model. Med Phys 44:2312-2320 |
| Zhu, Lei (2016) Local filtration based scatter correction for cone-beam CT using primary modulation. Med Phys 43:6199 |
