Abstract Book

ESTRO 37

S386

PO-0750 Association between lung density changes and radiation pneumonitis after IMRT for lung cancer A. I Ghanem 1,2 , W. Qixue 3 , K. Al Feghali 1 , R. Khalil 1 , C. Liu 3 , M. Ajlouni 1 , B. Movsas 1 , I. Chetty 3 1 Henry Ford Health System, Radiation Oncology, Detroit, USA 2 Alexandria University, Clinical Oncology department, Alexandria, Egypt 3 Henry Ford Health System, Medical Physics, Detroit, USA Purpose or Objective Definitive chemoradiotherapy (CRT) is potentially curative for locally advanced bronchogenic carcinoma and non-surgical candidates. However, radiation induced pneumonitis (RP) constitutes a major threat to the quality of life. The aim of our work is to investigate the impact of lung density changes as well as other risk factors in predicting RP in patients with lung cancer undergoing CRT using IMRT technique. Material and Methods After IRB approval we queried lung cancer cases treated definitively using CRT with available follow up CT chest datasets. We excluded all patients with prior chest RT, incomplete treatment and cases lacking post-RT images. All patients underwent 4D-CT simulation and the treatment plan was prescribed to cover PTV with 60-66 Gy in 2 Gy fractions using IMRT. Planning CT images with dose distribution were mapped to follow up CT using deformable registration to assess Hounsfield Unit (HU) changes after RT as portrayed in figure (1) excluding any tumor specific changes. Radiation oncologists graded RP by reviewing surveillance encounters using CTCAE v4.03 scale and were blinded to HU changes. Spearman’s correlation was plotted for post-RT HU changes against all RP grades. Chi-square and Fisher’s exact test were used to study factors associated with RP ≥2.

baseline metrics were balanced among RP groups with no significant differences, except that symptomatic RP requiring medical intervention (G2 & 3) had significantly more chronic obstructive lung disease (COPD) (93% vs 67% for G2+ vs G0-1; p=0.006 ).

Conclusion Deformable registration and quantification of HU changes between baseline and follow up CT is strongly predictive for RP and can be integrated in regular follow up protocols after validation in a prospective trial setting. More robust dose constraints need to be tailored for cases with baseline COPD as they are at increased risk for RP. PO-0751 Radiomic feature changes in small cell lung cancer IGRT relate significantly to overall survival G. Price 1 , A. McWilliam 1 , J. O'Connor 1 , M. Van Herk 1 , C. Faivre-Finn 1 1 The University of Manchester, Radiation Related Research, Manchester, United Kingdom Purpose or Objective There is unmet need to improve survival in Small Cell Lung Cancer (SCLC). Developments in radiotherapy and integration with systemic therapy have improved survival. However, current median survival is still only 16-30 months in limited stage disease. Non-invasive prognostic biomarkers are needed for SCLC. In this study we examine the potential of routine IGRT data in predicting clinical outcome in SCLC patients treated with radical radiotherapy. Material and Methods IGRT images from 46 patients were analysed to determine how the radiomic features of their tumour varied during treatment. Patients had received once daily or twice daily concurrent chemoradiotherapy in the CONVERT trial. The GTVs defined on planning CT scans were propagated to cone beam CT (CBCT) images using deformable registration (ADMIRE, Elekta). Radiomic features (92 features) were calculated within the GTV on the unfiltered CBCT data using the open source reference standard toolkit PyRadiomics. The change in features during treatment were approximated as linear functions of time. The relationship between each feature’s rate of change and overall survival was evaluated by univariate Cox regression. Features with a significant univariate hazard ratios were assessed together with clinical factors in a multivariate Cox model to control for multiple comparison. Results Univariate analysis shows the rate of change of 8 radiomic features to have significant hazard ratios. Multivariate analysis shows only one feature, the rate of change in spherical disproportion, to be associated with overall survival (p=0.002). Neither clinical factors nor changes in tumour volume are shown to have prognostic value (Table 1). Spherical disproportion, the ratio of the tumour surface area to that of a sphere of equal volume (the inverse of sphericity), is a measure of tumour spiculation. Figure 1 shows that patients may be stratified into risk groups based upon the rate of change of the feature. More rapidly reducing spiculation is associated with worse survival.

Results After a median follow up of 17 months (2-57) we were able to identify 118 patients who met our inclusion criteria. All had a smoking history with mean pack years (46±29), 51% were males and median age was 70 (45-88). Stage III were the majority (64%) and adenocarcinoma was present in 52.5% followed by squamous and small cell carcinoma (40 & 7.5%). Carboplatin/taxol was the most utilized regimen (54%) followed by platinum/pemetrexed or eteposide (26% & 15%).Grade 1 (G1), a radiological finding; RP was detected in 42 cases (35.6%); grade 2 (G2) & 3 (G3) in 22(18.6%) and 7(5.9%) respectively. The mean corresponding density increases for both lungs - CTV were 10, 26.7, 29.3 & 58.1 HU for G0-3 RP respectively that resulted in a significant Spearman’s correlation (r=0.43, p<0.0001 ) as shown in figure (2). Similar outcomes were depicted for ipsilateral lung - CTV as well as the V20 volumes ( p<0.001 ). Another significant correlation was observed between increasing mean lung dose (MLD) (r=0.26, p =0.01 ) and V20 (r=0.22, p=0.02 ) with RP grades, although both were within normal constraints (V20: 21.3±3.9 % and MLD: 13.1±3.9 Gy). All