Accurate IMRT fluence verification for prostate cancer patients using ‘in-vivo’ measured EPID images and in-room acquired kilovoltage cone-beam CT scans

Abstract

Background

To investigate for prostate cancer patients the comparison of ‘in-vivo’ measured portal dose images (PDIs) with predictions based on a kilovoltage cone-beam CT scan (CBCT), acquired during the same treatment fraction, as an alternative for pre-treatment verification. For evaluation purposes, predictions were also performed using the patients’ planning CTs (pCT).

Methods

To get reliable CBCT electron densities for PDI predictions, Hounsfield units from the pCT were mapped onto the CBCT, while accounting for non-rigidity in patient anatomy in an approximate way. PDI prediction accuracy was first validated for an anatomical phantom, using IMRT treatment plans of ten prostate cancer patients. Clinical performance was studied using data acquired for 50 prostate cancer patients. For each patient, 4–5 CBCTs were available, resulting in a total of 1413 evaluated images. Measured and predicted PDIs were compared using γ-analyses with 3% global dose difference and 3 mm distance to agreement as reference criteria. Moreover, the pass rate for automated PDI comparison was assessed. To quantify improvements in IMRT fluence verification accuracy results from multiple fractions were combined by generating a γ-image with values halfway the minimum and median γ values, pixel by pixel.

Results

For patients, CBCT-based PDI predictions showed a high agreement with measurements, with an average percentage of rejected pixels of 1.41% only. In spite of possible intra-fraction motion and anatomy changes, this was only slightly larger than for phantom measurements (0.86%). For pCT-based predictions, the agreement deteriorated (average percentage of rejected pixels 2.98%), due to an enhanced impact of anatomy variations. For predictions based on CBCT, combination of the first 2 fractions yielded gamma results in close agreement with pre-treatment analyses (average percentage of rejected pixels 0.63% versus 0.35%, percentage of rejected beams 0.6% versus 0%). For the pCT-based approach, only combination of the first 5 fractions resulted in acceptable agreement with pre-treatment results.

Conclusion

In-room acquired CBCT scans can be used for high accuracy IMRT fluence verification based on in-vivo measured EPID images. Combination of γ results for the first 2 fractions can largely compensate for small accuracy reductions, with respect to pre-treatment verification, related to intra-fraction motion and anatomy changes.