Retrospective correlation of radiation pneumonitis with dose distributions computed with different algorithms for locally advanced lung cancer patients.

Document Type

Conference Proceeding

Publication Date

2017

Publication Title

Med Phys

Abstract

Purpose: To correlate radiation pneumonitis (RP) with dose distributions, computed with different algorithms for 51 patients with locally advanced lung cancer. Methods: We retrospectively investigated dose distributions for 51 lung cancer patients treated with 60-66 Gy in 2 Gy fractions. The majority of tumors were centrally located. A radiation oncologist graded the RP based on the CTCAE scheme, defining clinical symptoms according to severity, grades 1-4. Nineteen patients experienced RP. Treatment plans were computed using AAA in Eclipse with IMRT or 3D conformal techniques. These plans were recomputed using the AcurosXB and pencil beam (PB) algorithms with identical plan parameters. Dosimetric parameters for normal lung (total lung-CTV) such as mean lung dose (MLD), V5, and V20 were correlated with RP. Results: Overall RP grade increases with MLD, V20, and V5. MLD showed the strongest association with RP grade (R2 = 0.74). As a function of RP grade, AAA, AcurosXB, and PB algorithms showed similar curves and trends, though the curves for AAA and AcurosXB were in better agreement relative to PB. V5 was found to be noticeably lower for PB algorithm relative to AAA and AcurosXB. Percent difference of mean V5 for RP grade 1, 2, and 4 are -5.2±0%, -3.4±0.9%, and -2.1±1.9%, respectively, relative to AAA. This is due to lack of inability to account for the lateral electron spreading using the PB algorithm. Conclusion: For the purposes of outcome modeling, AAA, AcurosXB and PB algorithms showed similar correlations with radiation-induced pneumonitis, however, AAA and AcurosXB curve shapes were in better agreement. These findings need to be validated with a large dataset.

Volume

44

Issue

6

First Page

2846

This document is currently not available here.

Share

COinS