光子数え込みコンピューテッドトモグラフィーにおける光子線量計算の精度—膵臓症例を対象とした仮想単色画像と電子密度（Rho）画像に基づく比較研究

Photon-counting computed tomography (PCCT) offers versatile anatomic information because of better energy discrimination and higher spatial resolution than conventional energy-integrating computed tomography (CT). With its rapid applications in diagnostic imaging, its potential within radiation oncology remains largely unexplored. Successful radiation therapy (RT) relies on both high-quality images to delineate tumor volume and accurate physical information of the anatomy for dose calculation. Specific to RT, dose calculation stands as a pivotal component in the process of treatment planning. This proof-of-concept study aimed to assess the accuracy of dose calculation and build the benchmark using PCCT images from a clinically operating NAEOTOM Alpha PCCT scanner from Siemens Healthineers.

A total of 29 patients receiving abdominal contrast-enhanced PCCT scans were included in this study with institutional review board approval. The following 2 sets of reconstructed images were selected for this study: (1) virtual monoenergetic images (VMIs) of 70 keV, which resembles the conventional CT image at 120 kVp, and (2) electron density (Rho) images derived from PCCT, which provides direct physical information for photon dose calculation. In addition to the default Hounsfield look-up table (HLUT) implemented in a treatment planning system, which is suitable for conventional CT images, a specific HLUT converting 70 keV VMI CT numbers to relative electron density was established following the consensus guidelines. Patient's organs were contoured using a deep-learning-based autosegmentation model in syngo.via RT Image Suite from Siemens Healthineers. An elective and a gross tumor volume were simulated in the pancreatic region for treatment planning within this study. All patients were planned with a 9-field intensity modulated RT with simultaneous integral boost regimen 25/33 Gy using an in-house deep-learning-based autoplanning model. Planning guidelines followed the institutional pancreas stereotactic body RT protocol. Eleven dose-volume histogram (DVH) metrics were included for comparison of 3-dose calculation approaches (conventional, 70 keV VMI and Rho).

The results revealed minimal dosimetry differences between 70 keV VMI and Rho images, with DVH metric percentage differences predominantly within 1% range (max: -1.60%) and 3-dimensional gamma analysis (1 mm/1%) pass rate >98% for most cases. The patient with the largest differences showed an average DVH metric difference of -1.32% and 3-dimensional gamma pass rate of 94.82%. Overall isodose distribution remained similar between 2 images for each patient. Dose calculation using the default HLUT in Eclipse showed DVH differences within a 1.5% range.

The results demonstrated good agreement in dose calculation between 70 keV VMI and Rho images. For regions with iodine contrast, Rho images provided by PCCT can suppress the contrast enhancement, thereby mitigating unnecessary uncertainties. This benchmark performance shows that PCCT can be safely used for dose calculation with certain advantages, making it a valuable alternative for RT treatment planning.