2V-CBCT: Two-Orthogonal-Projection Based CBCT Reconstruction and Dose Calculation for Radiation Therapy Using Real...

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Image Reconstruction

Background:

• On-board cone-beam computed tomography (CBCT) is crucial for radiation therapy (RT) quality assurance and adaptive treatments.
• CBCT is not always acquired for every treatment fraction, limiting its routine use.
• Two orthogonal projections are consistently available, presenting an opportunity for alternative imaging methods.

Purpose of the Study:

• To investigate the feasibility of two-orthogonal-projection-based CBCT (2V-CBCT) for image reconstruction and dose calculation in RT.
• To assess the accuracy of 2V-CBCT dose calculations compared to full-view CBCT using real projection data.

Main Methods:

• A novel coarse-to-fine deep learning strategy was developed for 2V-CBCT reconstruction.
• A 3D deep neural network was used for initial volume prediction, exploiting inter- and intra-slice information.
• A 2D deep neural network with perceptual loss refined the reconstruction slice-by-slice.

Main Results:

• The study successfully demonstrated the feasibility of 2V-CBCT reconstruction and dose calculation using real projection data.
• Dose calculation accuracy using 2V-CBCT was found to be comparable to full-view CBCT for both photon and proton RT.
• This marks the first feasibility study of 2V-CBCT with real projection data for RT applications.

Conclusions:

• 2V-CBCT is a feasible technique for radiation therapy dose calculation, even with limited projection data.
• The proposed deep learning approach effectively addresses the ill-posed nature of 2V-CBCT reconstruction.
• 2V-CBCT holds significant potential for improving RT quality assurance and adaptive therapy workflows.