Target Visualization Comparison for Lung Cancer Treatment In Radiation Therapy: 4DCT Vs 5DCT
Abstract
Purpose
4DCT is widely considered the standard for lung cancer radiation therapy treatment planning (TP) despite susceptibility to breathing-related motion artifacts that often render images unusable. Five-dimensional CT (5DCT), a model-based CT reconstruction method, has been proposed as an alternative that can produce artifact-free images. This work evaluates performance differences between 4DCT and 5DCT in treatment planning (TP).
Methods
Twenty patients were prospectively enrolled (IRB-11-0620) to consecutively collect 4DCT and 5DCT images during CT-simulation. 5DCT employs a motion model derived from 25 deformably registered free-breathing CTs and a bellows-based respiratory signal to model tumor motion across arbitrary breathing amplitudes. Three blinded physicians contoured internal target volumes (ITV) for both techniques in two sessions separated by 30-days to prevent recall bias. Motion-compensated images were subsequently rated on image quality (1:Good, 2:Flawed,3:Unusable). ITVs were compared using volume percent difference (VPD), Hausdorff Distance (HD), mean distance to agreement (MDA), Dice Similarity Coefficient (DSC), and Jaccard Index (J). Statistical significance in scoring difference was evaluated using Wilcox signed-rank testing alongside intraclass correlation coefficients (ICC) to assess interobserver reliability across contours. Accuracy and precision of 5DCT image reconstruction were reported as residual model error(mm).
Results
4DCT scores averaged1.5±0.68, with 5DCT averaging 1.12±0.32. Scoring differences were statistically significant (p<0.05), indicating meaningful image quality improvement with 5DCT. Three images were rated unusable, all being 4DCTs. All metrics showed significant interobserver agreement (p<0.05), ranging from moderate (VPD, ICC=0.60) to good-excellent (DSC, J, HD, MDA, ICC=0.86-0.95). VPD, HD, MDA, DSC, and J averaged 37.91±25.93%, 8.67±5.53mm, 1.79±1.08mm, 0.69±0.12, and 0.54±0.14, respectively. 5D model residual errors were 2.02±0.66mm (range:1.70-2.60mm).
Conclusion
Results demonstrate significant differences between modalities. 5DCT offers an improved alternative via artifact-free quantitative reconstruction. Further investigation on the implication of potential geographical target misses by 4DCT is warranted.