Evaluation on Pixel Value Conversion Methods for Dose Calculation with Kilo-Voltage Cone Beam CT Images

The purpose of our study is to evaluate the accuracy of pixel-value conversion methods for dose calculation with kilo-voltage cone-beam computed tomography (CBCT) images in radiotherapy. We investigated the relationship between CT numbers reconstructed with a CBCT and a conventional multi- slice CT (MSCT). To convert CBCT numbers to MSCT numbers, we used three conversion methods. One was the method using an electron-density CT-phantom (EP method). The second was the method using mean values in arbitrary defined several specific areas in CBCT and MSCT images (SA method). The last was the method using segmented regions (SR method), in which we extracted lung, bone and soft-tissue regions in CBCT and MSCT images, and replaced the pixel value of the CBCT image with the median (or mean, mode) of the MSCT numbers in each corresponding region. And after the conversion of CBCT numbers to MSCT numbers, we calculated the dose distribution using this image. The accuracy of dose distribution was verified with a heterogeneous phantom and clinical lung images. We compared the results with initial plans using a pass rate. The differences in distance-toagreement (DTA) 2mm pass rates between the dose distribution calculated with a converted CBCT image and that of the initial plan of the phantom were 98.5%, 100%, 100% and 20.4% for the EP, SA, SR method and the case without any conversion, respectively. And in the results of the γ analysis (2mm, 2%), the pass rates were 100%, 100%, 100% and 30.5% for the EP, SA, SR method and the case without any conversion, respectively. In the SR method (median) the accuracy was much improved. As for the clinical lung images, the same tendency was observed as that of the phantom. The SR method could improve the dose calculation accuracy using a converted CBCT image.