Room: Exhibit Hall | Forum 3
Purpose: Ray-tracing dose calculation algorithms can perform poorly when calculating narrow fields in heterogeneous tissues. In the treatment of lung cancer with stereotactic body radiation therapy (SBRT), this can result in delivered doses to the target volume that are considerably lower than prescribed. We investigate relationships between these differences and patterns of local failure in patients with progression after SBRT.
Methods: Thirty-four of 217 patients (16%) showed evidence of local progression after receiving SBRT on an Accuray CyberKnife at a single institution within 2007-2016. Patients were treated with 3-5 fractions to prescribed doses of 40-50Gy, as calculated using a clinical RayTrace (RT) algorithm to a planning target volume (PTV) generated with 5mm and 8mm margins in the axial and superior-inferior directions. For patients with local progression, diagnostic scans acquired following tumor recurrence were rigidly registered to the original planning CT based on local lung anatomy adjacent to the recurrence. Recurrences were categorized as in-field, marginal, or involved-lobe, based on the relative position and overlap of the recurrence with respect to the PTV. Plans were recalculated using the clinically-commissioned Multiplan Monte Carlo (MC) algorithm for a subset of 21 locally-progressed patients. For both the PTV and the overlap of PTV and recurrence, we compared the volume receiving at least prescription dose (VRx) and the doses covering 95% (D95%), evaluating the dosimetric difference between algorithms.
Results: Five of the 21 recurrences were categorized as in-field, 9 marginal, and 7 involved-lobe. The average VRx[PTV] was 96±1.4% for RT-calculated plans and 44.6±25.7% for MC-calculated plans. Similarly, D95%[PTV] was 48.4±2.4Gy and 37.7±5.9Gy for the respective RT and MC-plans. The VRx[overlap] was 98±3.5% amongst the RT-plans and 58.1±34.1% for the MC-plans.
Conclusion: For N=21 recurrent patients, SBRT treatment plans generated with RT demonstrated lower than prescribed average tumor coverage when recalculated with using MC algorithm (p<0.01).