Quantification and Minimization of Uncertainties of Internal Target Volume for Stereotactic Body Radiation Therapy of Lung Cancer

Purpose: To quantify uncertainties in delineating an internal target volume (ITV) and to understand how these uncertainties may be individually minimized for stereotactic body radiation therapy (SBRT) of early stage non-small cell lung cancer (NSCLC). Methods and Materials: Twenty patients with NSCLC who were undergoing SBRT were imaged with free-breathing 3-dimensional computed tomography (3DCT) and 10-phase 4-dimensional CT (4DCT) for delineating gross tumor volume (GTV)3D and ITV10Phase (ITV3). The maximum intensity projection (MIP) CT was also calculated from 10-phase 4DCT for contouring ITVMIP (ITV1). Then, ITVCOMB (ITV2), ITV10Phase+GTV3D (ITV4), and ITV10Phase+ITVCOMB (ITV5) were generated by combining ITVMIP and GTV3D, ITV10phase and GTV3D, and ITV10phase and ITVCOMB, respectively. All 6 volumes (GTV3D and ITV1 to ITV5) were delineated in the same lung window by the same radiation oncologist. The percentage of volume difference (PVD) between any 2 different volumes was determined and was correlated to effective tumor diameter (ETD), tumor motion ranges, R3D, and the amplitude variability of the recorded breathing signal (v) to assess their volume variations. Results: The mean (range) tumor motion (RSI, RAP, RML, and R3D) and breathing variability (v) were 7.6 mm (2-18 mm), 4.0 mm (2-8 mm), 3.3 mm (0-7.5 mm), 9.9 mm (4.1-18.7 mm), and 0.17 (0.07-0.37), respectively. The trend of volume variation was GTV3D <ITV1 <ITV2 ≈ ITV3 < ITV4 < ITV5. The means ± SDs of these volumes were 11.1 ± 9.3 cc, 13.2 ± 10.5 cc, 14.9 ± 11.0 cc, 14.7 ± 11.4 cc, 15.9 ± 11.7 cc, and 16.4 ± 11.8 cc, respectively. All comparisons between the target volumes showed statistical significance (P≤.001), except for ITV2 and ITV3 (P=.594). The PVDs for all volume pairs correlated negatively with ETD (r≤−0.658, P≤.006) and positively with R3D (r≥0.503, P≤.047). The PVDs for pairs of ITV2 vs ITV5 and ITV5 vs ITV4 negatively correlated with ETD (r=0.502, −0.626; P=.047, .010). No other correlation was found. Conclusion: Uncertainties in individualized ITVs for SBRT of early stage NSCLC could effectively be minimized by combining information from 3DCT, 4DCT, and MIP. If these images cannot be efficiently contoured, a combination of ITVMIP and GTV3D could be an effective alternative.