Room: Exhibit Hall
Purpose: By integrating knowledge-based model prediction and Auto-Planning module, we established a well-developed automatic planning station to generate radiation plans, thereby improving plan quality and consistency.
Methods: 12 previous patients with cervical cancer were enrolled in this study as test group. Another 30 patients were selected into establishing a knowledge-based model by linear fitting the equivalent uniform dose and equivalent uniform distance. By combining this knowledge-based model and Auto-Planning module of Pinnacle platform, an automatic planning station was developed based on Pinnacle scripts and python codes. PAP-plans of patients from test group were generated by Auto-Planning module and predicted EUD objectives of bladder and rectum from knowledge-based model. On the contrary, a corresponding plan, AP-plans were generated by Auto-Planning module and default objectives from template. By comparing dose/volume indices and EUD-EUL metric between AP-plans and PAP-plans, the efficiency and accuracy of this automatic planning station could be evaluated on the basis of improvement of plan quality and consistency.
Results: All dose/volume indices of bladder and rectum from PAP-Plans were lower than corresponding indices value from AP-Plans (Bladder: V20: 72% vs 79%, V30: 50% vs 59%, V40: 29% vs 35%, V50: 14% vs 16%, V60: 3% vs 4% and Dmean: 31 Gy vs 34 Gy; Rectum: V20: 82% vs 89%, V30: 51% vs 59%, V40: 28% vs 33%, V50: 10% vs 13%, V60: 0.8% vs 1% and Dmean: 32 Gy vs 34 Gy). On the other hand, the coverage and homogeneity of PTV from PAP-Plans were similar with AP-plans. According to EUD-EUL evaluation, bladder and rectum of PAP-Plans received lower doses than AP-Plans.
Conclusion: This automatic planning station generated clinical accepted plans which dramatically sparing the bladder and rectum. Furthermore, the application of predicted objectives into Auto-planning process could significantly improve the consistency of plans.