ESTRO 2024 - Abstract Book

S3809

Physics - Image acquisition and processing

ESTRO 2024

excellent reconstruction metrics, outperforming significantly the state-of-the-art 2D/3D deformable registration and bi-planar projection-based 3D reconstruction methods, as shown in Table 1.

Method

PSNR (dB)

SSIM

Backprojection

10.29(±0.46) 27.80 (±1.4) 28.35 (±1.86) 28.24 (±1.78)

0.225 (± 0.013) 0.884 (±0.026) 0.908 (±0.029) 0.919 (±0.023)

3D Reconstruction w/o planning CT [2]

2D/3D Deformable Registration [3]

3D Reconstruction and final deformation only Ours (Iterative 3D Reconstruction and Deformation)

32.20 (±1.50)

0.951 (±0.015)

Table 1. Metrics of reconstruction for our approach and previous methods.

Conclusion:

In this work, we tackle a pivotal challenge in radiation therapy of generating highly precise 3D CT reconstructions from exclusively biplanar projections. By integrating data from planning CT scans and simultaneously reconstructing and registering the two signals, we paved the way for daily 3D precision, potentially transforming IGRT. This innovative technique can usher in a transformative era, reducing treatment implementation toxicity, redefining patient positioning accuracy, elevating the quality of image fusion, and potentially reshaping the future of adaptive radiotherapy. We aim to refine and integrate our reconstruction into existing medical systems and explore dose