1114 - Incorporating Predicted Weekly Anatomies into IMPT Robust Optimization in Nasopharyngeal Cancers

M. K. H. Chan¹, and Y. Zhang²; ¹Department of Radiation Oncology, University Nebraska Medical Center, Omaha, NE, ²Department of Medical Physics and Biomedical Engineering, University College London, London, United Kingdom

Purpose/Objective(s):

Conventional robust optimization (RO) based on a single planning CT, per se, cannot account for the anatomical change of anatomy in intensity modulated proton therapy (IMPT), which is common in nasopharyngeal cancers (NPC). We propose to incorporate multiple predicted anatomies derived from a population averaged anatomy model (AM) in robust optimization and hypothesize that such strategy can improve the plan robustness in terms of tumor coverage. Materials/

Methods: Five NPC patients who had weekly validation CT (vCT) from fraction 5 to 35 with 5 as interval were chosen. Predicted anatomical CTs are generated by a predictive anatomical model that captures the average deformation of population at each time interval. High- and low-risk CTVs received 70Gy and 54.25Gy (CTV1 and CTV2), respectively. Two RO strategies with the same setup error (SE) and proton range error (RE) of 3mm and 3% were studied:

1. conventional RO (cRO) based on one single planning CT (pCT).
2. anatomical RO (aRO) incorporating predicted anatomies at the 5^th , 15^th , and 30^th fraction in addition to pCT.

Plan acceptance criteria were 98% of CTVs achieving 94% of the prescription doses (i.e., V94% >98%) in the voxel-wise minimum (voxmin) dose distribution under 3mm SE / 3% RE on pCT. These plans were subsequently evaluated on each vCT with and without 1mm SE and 3% RE, producing the voxmin 1mm3% CTV V94%. The accumulated doses from all vCTs were also analyzed.
Results:

All patients achieved voxmin(3mm3%) V94% = 99.5% in CTV1&2 with cRO and aRO on pCT. For CTV1, voxmin (1mm3%) V94% was < 98% with cRO on 4, 2 and 1 vCTs in 3 patients and with aRO3mm3% on 3 and 1 vCTs in 2 of the same 3 patients as with cRO. For CTV2, V94% was < 98% on one vCT with cRO and aRO in one different patient. All patients showed accumulated voxmin(1mm3%) V94% >98% for CTV1&2 with both cRO and aRO. A comparison of the nominal planning and accumulated doses of the OARs are shown in Table 1.
Conclusion: RO incorporating multiple weekly predicted anatomies improves interfractional dose coverage to high-risk CTV without significant trade-off for the OAR doses. Table 1. Comparison of the planning and accumulated OAR doses with cRO and aRO. Plan robustness in the OAR doses is quantified b the ratio of the accumulated to the planning doses.

Nominal dose		Ipsilateral parotid	Contralateral parotid	Submandibular glands	Oral cavity	Superior pharyngeal constrictor muscles (PCM)	Middle PCM	Inferior PCM
Planning	cRO	30.7	26.9	38.4	18.2	59.6	33.6	21.4
	aRO	26.3	27.0	37.1	18.7	60.1	33.7	21.2
Accumulated	cRO	35.2	30.8	41.4	21.8	61.1	38.7	24.7
	aRO	30.9	31.3	40.4	22.2	61.5	38.8	24.2
Ratio of planning / accumulated	cRO	1.15	1.14	1.08	1.20	1.02	1.15	1.15
	aRO	1.17	1.16	1.09	1.19	1.02	1.15	1.14