ESTRO 2025 - Abstract Book

S2981

Physics - Image acquisition and processing

ESTRO 2025

Conclusion: The proposed DL model demonstrated remarkable advancements in both accuracy and efficiency for EAN extraction, surpassing the performance of traditional analytical methods. Through comprehensive training on diverse material geometries, the model effectively captured and represented intricate EAN structures in head phantom images, as illustrated in Figure 2. These results highlight the potential of the DL approach to revolutionize precision in radiation treatment planning. Future efforts will prioritize the extraction of relative electron density, refinement of hyperparameters tuning, optimized material selection, and the application of this methodology to clinical CT imaging for real-world patient treatment optimization.

Keywords: PCD-CT, Effective Atomic Number, Deep Learning