Nasopharyngeal cancer adaptive radiotherapy with CBCT-derived synthetic CT: deep learning-based auto-segmentation precision and dose calculation consistency on a C-Arm linac.
Authors
Affiliations (4)
Affiliations (4)
- Department of Radiation Oncology, Second Affiliated Hospital of Xuzhou Medical University, Xuzhou, China. [email protected].
- Department of Radiation Oncology, Second Affiliated Hospital of Xuzhou Medical University, Xuzhou, China.
- Department of Radiation Oncology, Xuzhou First People's Hospital, Xuzhou, China.
- Technology Development Department, Anhui Wisdom Technology Co.,Ltd, Hefei, China.
Abstract
To evaluate the precision of automated segmentation facilitated by deep learning (DL) and dose calculation in adaptive radiotherapy (ART) for nasopharyngeal cancer (NPC), leveraging synthetic CT (sCT) images derived from cone-beam CT (CBCT) scans on a conventional C-arm linac. Sixteen NPC patients undergoing a two-phase offline ART were analyzed retrospectively. The initial (pCT<sub>1</sub>) and adaptive (pCT<sub>2</sub>) CT scans served as gold standard alongside weekly acquired CBCT scans. Patient data, including manually delineated contours and dose information, were imported into ArcherQA. Using a cycle-consistent generative adversarial network (cycle-GAN) trained on an independent dataset, sCT images (sCT<sub>1</sub>, sCT<sub>4</sub>, sCT<sub>4</sub><sup>*</sup>) were generated from weekly CBCT scans (CBCT<sub>1</sub>, CBCT<sub>4</sub>, CBCT<sub>4</sub>) paired with corresponding planning CTs (pCT<sub>1</sub>, pCT<sub>1</sub>, pCT<sub>2</sub>). Auto-segmentation was performed on sCTs, followed by GPU-accelerated Monte Carlo dose recalculation. Auto-segmentation accuracy was assessed via Dice similarity coefficient (DSC) and 95th percentile Hausdorff distance (HD<sub>95</sub>). Dose calculation fidelity on sCTs was evaluated using dose-volume parameters. Dosimetric consistency between recalculated sCT and pCT plans was analyzed via Spearman's correlation, while volumetric changes were concurrently evaluated to quantify anatomical variations. Most anatomical structures demonstrated high pCT-sCT agreement, with mean values of DSC > 0.85 and HD<sub>95</sub> < 5.10 mm. Notable exceptions included the primary Gross Tumor Volume (GTVp) in the pCT<sub>2</sub>-sCT<sub>4</sub> comparison (DSC: 0.75, HD<sub>95</sub>: 6.03 mm), involved lymph node (GTVn) showing lower agreement (DSC: 0.43, HD<sub>95</sub>: 16.42 mm), and submandibular glands with moderate agreement (DSC: 0.64-0.73, HD<sub>95</sub>: 4.45-5.66 mm). Dosimetric analysis revealed the largest mean differences in GTVn D<sub>99</sub>: -1.44 Gy (95% CI: [-3.01, 0.13] Gy) and right parotid mean dose: -1.94 Gy (95% CI: [-3.33, -0.55] Gy, p < 0.05). Anatomical variations, quantified via sCTs measurements, correlated significantly with offline adaptive plan adjustments in ART. This correlation was strong for parotid glands (ρ > 0.72, p < 0.001), a result that aligned with sCT-derived dose discrepancy analysis (ρ > 0.57, p < 0.05). The proposed method exhibited minor variations in volumetric and dosimetric parameters compared to prior treatment data, suggesting potential efficiency improvements for ART in NPC through reduced human dependency.