2971 - Feasibility of Target Delineation Using Low-Energy Level Monoenergetic Images Derived from Photon-Counting Computed Tomography for Pelvic Radiotherapy Planning

Purpose/Objective(s): CT simulation is essential to routine radiotherapy planning and dose calculation. However, conventional CT images lack sufficient soft tissue contrast, making accurate soft tissue tumor delineation difficult or even impossible. An additional MRI simulation for target delineation is often required, but MRI introduces several potential sources of error, including geometric distortion, artifacts, MRI–CT registration errors, and even misalignment due to physiological changes between scans. Photon-counting Computed Tomography (PCCT) acquires the entire spectrum of the incident X-ray beam, enabling the reconstruction of Virtual Monoenergetic Images (VMI). Low-energy VMIs provide significantly higher soft tissue contrast than conventional CT images. This study aims to determine the optimal VMI and to evaluate the feasibility of using low-energy VMI for target delineation in comparison with MRI. Quantitative analysis was performed on images from a patient with a pelvic tumor.

Materials/Methods: A retrospective analysis was conducted on a patient with CA anus who underwent external beam radiotherapy. Both contrast-enhanced PCCT and MRI simulations were performed prior to radiotherapy planning. To identify the optimal energy level, VMIs were reconstructed from 40 to 80 keV in 10 keV increments. These images were quantitatively evaluated using signal-to-noise ratio (SNR), lesion contrast (LC), and contrast-to-noise ratio (CNR). The optimal VMIs were then used to perform 2D target delineation with active contouring. The accuracy of the resulting contours was assessed by comparing with the oncologist-approved MRI contours, using the Dice similarity coefficient (DSC) and the 95th percentile Hausdorff distance (HD95).

Results: The 40 keV VMIs demonstrated the highest SNR, LC, and CNR among the evaluated energy levels, indicating the most suitable VMI for target delineation in this case. Tumor was delineated on 5 slices using the 40 keV images. Compared with the oncologist-approved MRI contours, all DSC were = 0.86, and all HD95 were = 4 mm. These results indicate high similarity and good spatial agreement between the VMI-based and MRI-based delineations.

Conclusion: The high DSC and low HD95 between 40 keV contours and MRI contours suggest that pelvic tumor delineation using low-energy VMIs with contrast enhancement is feasible and shows promising accuracy. This approach could potentially serve as an alternative or complement to MRI in external beam radiotherapy planning. It may be especially valuable in gynecological brachytherapy, where steep dose gradients make registration errors particularly influential. Before broad clinical implementation, additional case studies and larger cohort analyses are needed to confirm the robustness and generalizability of these findings.

VMI energy level (keV)	SNR	LC	CNR
40	9.22	70.75	3.98
50	8.57	41.92	2.89
60	7.89	24.34	1.94
70	7.25	13.33	1.17
80	6.30	6.24	0.55

Slice	DSC	HD95 (mm)
#1	0.91	2.79
#2	0.93	2.79
#3	0.88	3.55
#4	0.92	2.95
#5	0.86	3.94