Purpose: This study was undertaken to answer the following questions in breast irradiation: (a) How many calculation planes are sufficient for three-dimensional (3-D) treatment planning? (b) Is pseudo-3-D planning system sufficiently accurate for 3-D treatment planning of a breast?
Methods and materials: We carried out dose calculations and differential dose-volume analysis on three representative patients covering the range of breast size encountered in a clinic. The breast volumes were reconstructed from computed tomography (CT) scans using three slices, five slices and the full CT scan respectively. An established 3-D dose algorithm and two pseudo-3-D commercial systems were used in the calculations. Comparison of isodose distributions were made between the central axis plane, a cephalic and a caudal plane 6 cm above or below the central axis respectively.
Results: When comparing isodose distributions generated with conventional two-dimensional treatment planning with 3-D dose calculations, the former underestimated the size and magnitude of the hot spots in the medial and the lateral subcutaneous (SC) regions. When comparing the three-slice with the full CT model, while the three-slice model was found to be adequate for the "small" and the "medium" size patients, the full CT model provided a more accurate representation of dose distributions for the "large" patient. Comparison of a true 3-D algorithm with pseudo-3-D algorithms showed that while the latter systems were adequate for the "small" and the "medium" patients, significant differences were noted between the true 3-D and the pseudo-3-D algorithms for the "large" patient.
Conclusion: For patients whose breast contours vary slowly within the tangential fields, a three-slice CT scan as well as a pseudo-3-D approach appears to be adequate for clinical decision. However, for patients with large variation of contours within the tangential fields, a full scale CT scan with a true 3-D dose algorithm is more accurate than either the three-slice or the five-slice model.