Cone-beam computed tomography for lung cancer - validation with CT and monitoring tumour response during chemo-radiation therapy

Introduction: Cone-beam computed tomography (CBCT) is a valuable image-guidance tool in radiation therapy (RT). This study was initiated to assess the accuracy of CBCT for quantifying non-small cell lung cancer (NSCLC) tumour volumes compared to the anatomical 'gold standard', CT. Tumour regression or progression on CBCT was also analysed. Methods: Patients with Stage I-III NSCLC, prescribed 60 Gy in 30 fractions RT with concurrent platinum-based chemotherapy, routine CBCT and enrolled in a prospective study of serial PET/CT (baseline, weeks two and four) were eligible. Time-matched CBCT and CT gross tumour volumes (GTVs) were manually delineated by a single observer on MIM software, and were analysed descriptively and using Pearson's correlation coefficient (r) and linear regression (R-2). Results: Of 94 CT/CBCT pairs, 30 patients were eligible for inclusion. The mean (+/- SD) CT GTV vs CBCT GTV on the four time-matched pairs were 95 (+/- 182) vs 98.8 (+/- 160.3), 73.6 (+/- 132.4) vs 70.7 (+/- 96.6), 54.7 (+/- 92.9) vs 61.0 (+/- 98.8) and 61.3 (+/- 53.3) vs 62.1 (+/- 47.9) respectively. Pearson's correlation coefficient (r) was 0.98 (95% CI 0.97-0.99, q < 0.001). The mean (+/- SD) CT/CBCT Dice's similarity coefficient was 0.66 (+/- 0.16). Of 289 CBCT scans, tumours in 27 (90%) patients regressed by a mean (+/- SD) rate of 1.5% (+/- 0.75) per fraction. The mean (+/- SD) GTV regression was 43.1% (+/- 23.1) from the first to final CBCT. Conclusion: Primary lung tumour volumes observed on CBCT and time-matched CT are highly correlated (although not identical), thereby validating observations of GTV regression on CBCT in NSCLC.