Analysis of automatic match results for cone-beam computed tomography localization of conventionally fractionated lung tumors.

PURPOSE To evaluate the dependence of an automatic match process on the size of the user-defined region of interest (ROI), the structure volume of interest (VOI), and changes in tumor volume when using cone-beam computed tomography (CBCT) for tumor localization and to compare these results with a gold standard defined by a physician's manual match. METHODS AND MATERIALS Daily CBCT images for 11 patients with lung cancer treated with conventionally fractionated radiation therapy were retrospectively matched to a reference CT image using the Varian On Board Imager software (Varian, Palo Alto, CA) and a 3-step automatic matching protocol. Matches were performed with 3 ROI sizes (small, medium, large), with and without a structure VOI (internal target volume [ITV] or planning target volume [PTV]) used in the last step. Additionally, matches were performed using an intensity range that isolated the bony anatomy of the spinal column. All automatic matches were compared with a manual match made by a physician. RESULTS The CBCT images from 109 fractions were analyzed. Automatic match results depend on ROI size and the structure VOI. Compared with the physician's manual match, automatic matches using the PTV as the structure VOI and a small ROI resulted in differences ≥ 5 mm in 1.8% of comparisons. Automatic matches using no VOI and a large ROI differed by ≥ 5 mm in 30.3% of comparisons. Differences between manual and automatic matches using the ITV as the structure VOI increased as tumor size decreased during the treatment course. CONCLUSIONS Users of automatic matching techniques should carefully consider how user-defined parameters affect tumor localization. Automatic matches using the PTV as the structure VOI and a small ROI were most consistent with a physician's manual match, and were independent of volumetric tumor changes.