Main Session
Sep
28
PQA 03 - Digital Health Innovation and Informatics, Patient Safety & Quality, and Radiation and Cancer Biology
2561 - Evaluation and Comparison of Volumetric and Geometric Accuracy of Synthetic CT Created from MRI with Conventional CT for Prostate Radiotherapy Planning
Presenter(s)
Venkatesan Karuppusamy, MS, CMD, BS - Siemens Healthineers, Princeton NJ, Princeton, NJ
V. Nappady Joy1, V. Karuppusamy2, S. Beriwal3, C. Eichner4, and A. Amjad5; 1Siemens Healthineers, Forchcheim, Bayern, Germany, 2Siemens Healthineers, Princeton, NJ, 3Allegheny Health Network Cancer Institute, Pittsburgh, PA, 4Siemens Healthineers AG, Forchcheim, Germany, 5Department of Radiation Oncology, Medical College of Wisconsin, Milwaukee, WI
Purpose/Objective(s):
Synthetic CT (sCT) is a technique using AI/deep learning to generate CT-like images (with Hounsfield Units) from MRI scans, eliminating the need for separate CT scans in radiation therapy planning, reducing radiation and saving time. In MRI only workflow, it can also be used to match for daily target localization. There are no published studies comparing volumetric accuracy of sCT to planning CT scans. The goal of this study is to assess the volumetric accuracy of sCT relative to conventional CT (CT) for prostate radiotherapy planning and daily localization for prostate cancer.Materials/Methods:
In and opposed phases of MRI Dixon sequence from Siemens Magnetom Vida (3 Tesla) Scanner were converted to sCT using VC10 syngo.via version of Radiation Therapy Imaging Suite (RTIS) post-processing software. Two independent users, one dosimetrist and one Physician contoured prostate on eleven paired scans. Prostate volumes and geometric displacement metrics (AP/PA and lateral) were extracted at the apex, mid-gland, and base. Statistical comparisons included paired t-tests, Pearson correlation, and Bland–Altman agreement analysis. Results: Mean prostate volume measured 50.3 cc (IQR 32.15-65.43) on sCT and 49.72 cc (IQR 33.27-64.81) on CT, with mean difference of .58 cc (p = .62). A strong correlation was observed between modalities (r = 0.955). Bland–Altman analysis demonstrated a small positive bias with narrow limits of agreement. Positional metrics showed small differences across all dimensions with no systematic shift toward either modality (table 1). Variability in both AP/PA and lateral directions remained low, indicating geometric consistency.Conclusion:
Synthetic CT demonstrated excellent agreement with planning CT for prostate volume and dimensions estimation. The minimal volumetric bias and dimensions differences support the clinical feasibility of sCT as a substitute for CT in prostate radiotherapy workflows including planning and daily localization for treatment. Expanded multicenter validation may further solidify its integration into routine treatment planning. Table-1: All the volumes are in milliliters (ml) and all the geometric measurements are in centimeters(cm).| Synth CT mean | Synth CT IQR | CT mean | CT IQR | Mean Diff | p-value | |
| Volume(ml) | 50.30 | 32.15-65.43 | 49.72 | 33.27-64.81 | 0.58 | 0.62 |
| Apex AP/PA | 3.44 | 3.00-4.00 | 3.24 | 2.80-3.60 | 0.21 | 0.05 |
| Apex R/L | 3.99 | 3.80-4.40 | 3.79 | 3.40-4.30 | 0.20 | 0.11 |
| Mid AP/PA | 4.08 | 3.86-4.30 | 3.90 | 3.50-4.40 | 0.18 | 0.05 |
| Mid R/L | 4.70 | 4.10-5.20 | 4.56 | 4.10-5.10 | 0.14 | 0.06 |
| Base AP/PA | 3.89 | 3.50-4.50 | 4.01 | 3.50-4.40 | -0.13 | 0.41 |
| Base R/L | 4.41 | 3.90-4.90 | 4.60 | 4.14-5.20 | -0.18 | 0.06 |