Publications Repository - Helmholtz-Zentrum Dresden-Rossendorf

Background and aims
Daily Prompt-Gamma-Imaging (PGI) provides benefits for treatment verification (TV) in proton therapy by enabling the detection of anatomical changes during treatment and allowing a potential margin reduction. An additional benefit of PGI-TV emerges in the context of online-adaptive proton therapy (OAPT). There, PGI-TV could not only detect intra-fractional anatomical changes, but also act as a safety net for the adapted treatment within the OAPT feedback loop. Subsequently, we outline translational achievements and next steps towards PGI-TV in OAPT.

Methods
As depicted in Figure 1, (I) we acquired clinical data with a PGI-slit-camera prototype for different tumor locations. For every PGI-monitored fraction, in-room control CT (cCT) with dual-energy scans in treatment position were available. Subsets of these data have already been used to retrospectively evaluate (II-A) the capability of PGI to detect anatomical changes and (II-B) the potential of PGI TV to reduce range uncertainty margins. (III) Furthermore, we are developing a fully automated online PGI evaluation. For OAPT, this requires processing cone-beam CT (CBCT) scans as input for calculating the PGI reference, which is tested in a first simulation study using different phantom and patient datasets. (IV) Finally, we are preparing an interventional PGI trial for prostate-cancer proton therapy.

Results
According to Figure 2, (II-A) PGI classification models for detecting anatomical changes showed an accuracy of 77% for measured PGI data (prostate) and 93% for realistic PGI simulations (H&N). (II-B) With PGI-TV, combined with DirectSPR-based range prediction, the range uncertainty margin can be reduced from 7mm to 3mm in prostate-cancer treatments. (III) We have reduced the PGI processing time by ~64% and demonstrated in phantoms that the PGI reference simulation can be calculated from CBCTs without relevant limitations. (IV) In our funding-approved interventional trial with daily PGI monitoring and reduced range uncertainty (cf. III), PGI will trigger cCT (with an intended false-positive trigger rate <20%) directly after treatment with adaptions applied on the next treatment day
.
Conclusions
Clear progress towards an interventional application of PGI-TV has been made. Its application in OAPT workflows seems particularly beneficial and promising. The first prospective application within an interventional trial is in preparation.