Accurate patient alignment without unnecessary imaging using patient-specific 3D CT images synthesized from 2D kV images

Abstract

In radiotherapy, 2D orthogonally projected kV images are used for patient alignment when 3D-on-board imaging (OBI) is unavailable. However, tumor visibility is constrained due to the projection of patient’s anatomy onto a 2D plane, potentially leading to substantial setup errors. In treatment room with 3D-OBI such as cone beam CT (CBCT), the field of view (FOV) of CBCT is limited with unnecessarily high imaging dose. A solution to this dilemma is to reconstruct 3D CT from kV images obtained at the treatment position. We propose a dual-models framework built with hierarchical ViT blocks. Unlike a proof-of-concept approach, our framework considers kV images acquired by 2D imaging devices in the treatment room as the solo input and can synthesize accurate, full-size 3D CT within milliseconds. We demonstrate the feasibility of the proposed approach on 10 patients with head and neck (H&N) cancer using image quality (MAE: < 45HU), dosimetric accuracy (Gamma passing rate ((2%/2 mm/10%): > 97%) and patient position uncertainty (shift error: < 0.4 mm). The proposed framework can generate accurate 3D CT faithfully mirroring patient position effectively, thus substantially improving patient setup accuracy, keeping imaging dose minimal, and maintaining treatment veracity. Effective and accurate imaging guidance is critical for precise patient alignment, accurate tumor tracking, accurate delivery of radiation therapy and to protect organs that should not be irradiated. However, high-quality imaging guidance usually can only be provided following detailed imaging using a large amount of radiation. We propose a computational method that can generate the full size 3D images required as image guidance from X-Ray images. We demonstrated its utility using data from 10 people with head and neck cancer. Our proposed approach can be used by existing treatment machines to improve the accuracy of patient alignment and hence ensure more accurate treatment of patients. Ding et al. propose a deep learning-based model for fast and accurate 3D CT reconstruction given 2D kV (X-Ray) images as the solo inputs. The experimental results and analysis indicate that the proposed framework can be used for accurate and robust patient alignment with minimum imaging dose.