[Quantitative Evaluation of Average Intensity Projection Images in Radiotherapy Planning for Primary Lung Cancer]

Average intensity projection (AIP) is an image generation method obtained by averaging the images of each phase of 4D-CT. This study aimed to evaluate tumor respiratory motion and, using the union of contours from all 4D-CT phases (4D_Total) as the reference, to compare contour agreement among contours derived from 3D-CT, maximum intensity projection (MIP), and AIP images. In addition, dose metrics in treatment planning were evaluated to explore the dosimetric impact of contour differences among the image sets.

Eight patients who underwent lung stereotactic body radiotherapy (SBRT) at our institution were included. Tumor motion was first characterized from centroid positions across ten 4D-CT phases. Next, 4D_Total was generated by taking the union of the contours extracted from each phase. Finally, contours from 4D-CT AIP, 4D-CT MIP, and slow-scan 3D-CT (3D_Slow) were compared with 4D_Total using volume difference, centroid difference, Dice similarity coefficient (DSC), Jaccard index, 95th percentile Hausdorff distance (HD95), mean surface distance (MSD), and surface Dice similarity coefficient with a 2-mm tolerance. In addition, treatment plans were generated using each contour set, and Dmax (cGy), lung V5 (%), lung V20 (%), conformity index (CI), and homogeneity index (HI) were compared.

Tumor motion was greatest in the inferior-superior direction and was generally multi-directional. For AIP, centroid difference was ≤1 mm in each direction (LR/AP/IS), and the median values of DSC, Jaccard index, HD95, MSD, and surface DSC (2 mm) were 0.831, 0.711, 1.707 mm, 0.451 mm, and 0.981, respectively, showing good geometric agreement with 4D_Total. No significant differences were observed between AIP and MIP, whereas both AIP and MIP showed better agreement than 3D_Slow. In the dose evaluation, no significant overall differences were observed for any metric among the four contour sets. In the exploratory analysis, paired comparisons using 4D_Total as the reference showed a directional decrease in lung V20 for AIP and a directional increase in CI for MIP. These findings should not be interpreted as definitive conclusions, but rather as results of an exploratory analysis based on a limited number of cases.

The AIP method showed good geometric approximation to 4D_Total and may represent the time-dependent displacement of tumor position within a single dataset. In this limited exploratory analysis, no major disadvantage in treatment planning was observed. However, the observed changes in dose metrics, including lung V20, should not be interpreted as evidence of dosimetric superiority and require further validation. Because of uncertainty in the reference contour and the limited number of cases, further investigation is needed to clarify its clinical significance.