ESTRO 2021 Abstract Book

S1562

ESTRO 2021

evaluate if the ACE algorithm accounts adequately for this interface. To evaluate the accuracy of ACE for HDR skin brachytherapy two separate end-to-end tests were carried out using thermo-luminescent dosimeters (TLDs) to measure delivered dose. Materials and Methods Two patient masks with wax stand-off from previous clinical treatments were planned and prescribed with the appropriate dose under local protocol. A scalp and a nose treatment were selected as these represent the most common ‘large’ and ‘small’ clinical sizes and treatment sites at this local centre. Source positions were manually selected and two optimisations for the dwell times were performed for each site. For each site, the loading dwell times were first optimised and delivered under TG-43, then fixed to recalculate in ACE. The second plan was then optimised and delivered using ACE and then recalculated using TG-43 dose with the dwell times fixed. For all ACE calculations the patient contour, Perspex immobilisation cast and wax stand-off was delineated and assigned a uniform density of water. Any voxel outside these external contours was assumed to be air. Each plan was delivered with a Flexitron HDR Ir-192 afterloader (Elekta, Sweden) using bolus bags underneath the mask as a patient surrogate and to simulate patient scatter. Dose at the ‘skin’ surface was measured in each delivery with a calibrated TLD placed at the center of the treatment area. Results For both scalps plans, TG-43 overestimated the dose from the measured TLD by +12.6% and +9.7%. Whereas for ACE, the difference between calculated and measured was +3.4% and +2.7%. For the nose, the difference between TG-43 and TLD results were +0.7% and +0.6%, whereas ACE underestimated the dose by -2.0% and -2.1%. Conclusion For the nose treatment, both TG-43 and ACE were within a clinically acceptable tolerance of the measured doses. For the scalp TG-43 significantly overestimated the ‘skin’ dose in both plans, while the ACE calculated plans were in closer agreement. This could be a result of the ‘missing’ backscatter during treatment under TG-43, and the higher scatter/primary ratio in the lager treatment geometry. The results suggest that in all cases ACE is able to successfully compensate for the presence of air in the calculation volume. PO-1833 Evaluating the ability of Sun Nuclear’s DoseCHECK software to detect clinically significant errors A. Starke 1 , J. Poxon 1 , N. MacDougall 1 1 Barts Health NHS Trust, Radiotherapy Physics, London, United Kingdom Purpose or Objective To establish if DoseCHECK, the independent 3D dose calculation module of Per Fraction, is sufficiently sensitive to detect treatment plan errors when using clinically relevant analysis parameters. Materials and Methods Deliberate errors were written into treatment DICOM plan files to simulate potential corruption in the software planning process. The error mode chosen was MLC positional errors on one leaf bank (B) as these would have significant impact on the patient plan accuracy. Ten previously treated pelvis VMAT patient plans (two prostate, five prostate and nodes and three gynae) were exported from the Eclipse TPS in DICOM format. Each plan file was then modified in Matlab, introducing deliberate MLC errors. Six error plans were created for each patient: the central five MLC leaves were retracted by 1mm and 2mm respectively; the central ten MLC leaves were retracted by 1mm and 2mm; all bank B MLC leaves were pulled out by 1mm and 2mm. These modified plan files were sent to Per Fraction, along with the dose files, structure set and CT image from the original plan. Per Fraction therefore compared its dose calculation on the corrupted plan file(s) with the treatment planning system’s calculation of the original plan. Plans were reported as a pass or fail using the following local gamma analysis parameters: 3% dose, 2mm DTA, 30% threshold, with a 95% pass rate. Results Table 1 shows if Per Fraction gave a pass or a fail for each of the error tests:

Original plan 5 leaves 1mm 10 leaves 1mm Bank B 1mm 5 leaves 2mm 10 leaves 2mm Bank B 2mm

Prostate 1

✓

✗

✗

✗

✗

✗

✗

Prostate 2

✓

✗

✗

✗

✗

✗

✗

Prostate & nodes 1 ✓

✓

✗

✗

✗

✗

✗

Prostate & nodes 2 ✓

✓

✓

✓

✓

✗

✗

Prostate & nodes 3 ✓

✓

✓

✗

✗

✗

✗

Prostate & nodes 4 ✓

✓

✓

✗

✗

✗

✗

Prostate & nodes 5 ✓

✓

✓

✗

✗

✗

✗

Gynae 1

✓

✓

✓

✓

✓

✗

✗

Gynae 2

✓

✓

✓

✗

✗

✗

✗

Gynae 3 ✗ Table 1: Results of error testing on ten pelvis VMAT patients; ✗ represents a fail and ✓ represents a pass. For all ten patients Per Fraction produced a failure when the entire bank of MLC leaves was retracted by 2mm and when the central ten leaves were moved 2mm. It can also be seen that three of the patient plans passed when all MLCs were moved by 1mm; two of these were gynae plans. The two prostate plans failed when five central MLCs were moved by 1mm. Given these two plans had the smallest target volume it is unsurprising they were most sensitive to deliberate MLC errors. Overall the results show good sensitivity to detecting MLC positional errors when analysing plans using clinically relevant parameters. Conclusion The results show that the independent dose calculation module of Per Fraction has a high degree of sensitivity for detecting MLC positional errors, a necessary quality of any software used for independent plan checking in radiotherapy. ✓ ✓ ✓ ✓ ✓ ✗