TU‐A‐BRA‐03: Knowledge‐Based and Patient‐Geometry Specific IMRT Treatment Planning

Purpose: To create an IMRT treatment planning approach by which planners can access the knowledge of prior clinical plans to control the quality of new plans. Method and Materials: Overlap volume histogram (OVH) is used to access the knowledge of prior clinical plans. It allows planners to identify a reference group of prior clinical plans that contains geometric information “similar” to a new patient. The best plan in terms of OAR sparing in the reference group will then provide input to IMRT optimization for the new patient. In a retrospective OVH‐assisted planning demonstration, 15 patients were randomly selected from a database of 91 prior head‐and‐neck patients with a three‐level prescription: 58.1 Gy, 63 Gy and 70 Gy. A leave‐one‐out methodology was applied to generate the DVHs for each OVH‐assisted plan (OP). The database‐generated DVHs were then used by a planner who had no knowledge of the clinical plans (CPs). To evaluate the effectiveness of our methodology, the dosimetric results for three sets of plans: CPs, OPs after the first‐round optimization and final OPs were compared by the Wilcoxon p test. Results: Averages of optimization rounds required for completing CPs and OPs were 27.6 and 1.9 (p<0.00001); three OPs were completed in a single optimization round. For both OPs, averages of standard deviation to the PTV⁶³ deceased by ∼0.5 Gy (p<0.02); averages of D_{0.1 cc} to the cord+4mm decreased by ∼6.5 Gy (p<0.0001); averages of D_{0.1 cc} to the brainstem decreased by ∼7.5 Gy (p<0.005); averages of V(30 Gy) to the contra‐lateral parotid decreased by ∼8% (p<0.0001). Additionally, both OPs were comparable with or better than the CPs in PTV uniformity, conformity and other OAR sparing. Conclusion: The method offers a way of predicting clinically achievable doses ahead of planning, making IMRT planning no longer a trial and error process.