Dual-energy CT-based proton treatment planning to assess patient-specific range uncertainties

Purpose/Objective:

To reduce range uncertainties in particle therapy arising from a generic heuristic conversion (HLUT) of CT numbers in stopping-power ratios (SPRs), an accurate patient-specific SPR prediction is desirable. Treatment planning based on dual-energy CT (DECT) can account for tissue diversity and potentially contribute to shrink clinical safety margins. Consequently, in this study dose distributions derived from both a clinical HLUT and a patient-specific DECT-based SPR prediction are compared and range deviations are quantified for two different treatment sites.

Material/methods:

Based on a database of more than 1000 clinical DECT scans acquired with a single-source DECT scanner (Siemens Somatom Definition AS), 10 prostate cancer and 52 head tumor patients were selected to assess intra- and interpatient tissue diversity and its impact on SPR prediction. To evaluate age- and sex- dependent variability, the head tumor cohort was divided in children, women and men. DECT scans were converted in 79 keV pseudo-monoenergetic CT scans (MonoCTs) and SPR datasets derived by voxelwise calculations of electron density and effective atomic number using syngo.via (Siemens Healthineers). In XiO (Elekta) clinical proton treatment plans were recalculated (a) on MonoCTs using the clinical HLUT and (b) on SPR datasets to quantify range-dependent dose differences.

Results:

The voxelwise correlation of SPR and CT number is similar for men and women, but differs considerably between adults and children in bony tissue, likely due to the amount of calcium embedded in bones, which increases with age. Based on voxelwise SPR comparisons, the clinical HLUT predicts on average (2.2 ± 0.6) % larger SPRs for head tumor patients and (1.7 ± 0.3) % larger SPRs for prostate cases. The impact of both approaches on dose distributions is shown in Fig. 1 and 2 for an exemplary head tumor and prostate cancer patient. In the head case, the HLUT predicts a 1.7 % shorter range (2.4 mm) resulting from a 0.7 mm range underestimation in water-filled ventricles (not precisely predicted by the HLUT) and different SPR predictions for brain. A range deviation of up to 3.0 % (7.1 mm) is obtained in the prostate case, which is mainly caused by different SPR predictions for bone marrow and muscle. These range differences in single beams are not compensated in the overall treatment plan.

Conclusion:

In contrast to a generic HLUT, a DECT-based SPR prediction can individually consider age- and sex- dependent tissue variability in proton treatment planning. This diversity information can also provide suggestions for subgroup-specific improvements of the heuristic CT calibration. The assessment of relative SPR and dose differences underlines the clinical potential of DECT, which now needs to be confirmed against a ground truth. Further investigations of patients’ DECT scans enable comprehensive SPR evaluations to quantify CT-related range uncertainties and to assess clinical safety margins.