Retrospective evaluation of decision-making for pancreatic stereotactic MR-guided adaptive radiotherapy.

Background/purpose: Stereotactic-magnetic-resonance-guided-online-adaptive-radiotherapy (SMART) is a promising tool for pancreas stereotactic-body-radiotherapy. Our online-adaptive-radiotherapy (On-ART) process relies on daily image overview by the managing radiation-oncologist, who determines the need for creating a predicted plan if significant interfractional anatomical changes are noted. Predicted plans are achieved through applying the baseline plan on deformed and manually adjusted contours based on daily imaging. If the dose to the target volume or organs-at-risk (OARs) violate constraints, an adapted plan is generated and delivered for treatment. In-depth review of daily images and deformed contours is limited by time and inter-observer variations. This study evaluates the reliability of our On-ART decision-making process. All fractions retrospectively underwent a predicted plan for off-line decision-making to adapt (Off-ART). Decisions to adapt were compared using On-ART and Off-ART approaches.

Material/methods: Thirty-five sets of daily images were analyzed from seven patients who underwent five fractions of SMART. Each OAR was fully re-contoured off-line by the same physician for each fraction. Off-ART decision was re-evaluated for each fraction.

Results: N = 14/35 fractions were adapted based on On-ART decision-making versus N = 25/35 with Off-ART. The concordance between On-ART and Off-ART decision was 87.5% for the 16 fractions using a predicted plan online and 42% for the 19 fractions using only visual image review for On-ART decision-making.

Conclusions: Daily-image visual review is not reliable to determine benefit or not for adaptive radiation-therapy. Online predicted plan, based on deformed and manually adjusted contours, should be generated for every fraction that is delivered using SMART in order to reliably optimize treatment plans daily.