Automated Estimation of Lung Tumor Locations for Tumor-Based Patient Setup Using MV-CBCT Images in Stereotactic Body Radiotherapy
S Yoshidome*1, H Arimura2, K Nakamura3, Y Shioyama3, K Atsumi4, H Yoshikawa1, K Nishikawa1, H Hirata2, (1) Department of Clinical Radiology, Kyushu University Beppu Hospital (2) Department of Health Sciences, Faculty of Medical Sciences, Kyushu University (3) Department of Clinical Radiology, Graduate School of Medical Sciences, Kyushu University (4) Department of Radiology, Kyushu University Beppu HospitalSU-E-J-40 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: Patient setup procedure in stereotactic radiation therapy should be performed based on a tumor region, not bone structures. The goal of this study was to develop an automated method of estimation of lung tumor locations for tumor-based patient setup using megavoltage cone-beam computed tomography (MV-CBCT) images in stereotactic body radiotherapy (SBRT).
Methods: Planning CT, treatment MV-CBCT images (4.125 MV-CBCT images/patient), and DICOM-RT structure sets for 8 patients were employed for this study. The patients had solitary lung tumors smaller than 25 mm (range of effective diameter: 23 - 8.8 mm and median: 17.7 mm) and received SBRT. In the proposed method, the lung tumor locations were estimated in MV-CBCT images by using tumor templates obtained from corresponding planning CT images. First, a MV-CBCT image was globally aligned with a planning CT image by finding the location with the maximum cross-correlation coefficient, and then a gross target volume (GTV) region in the structure set was placed in the planning CT and MV-CBCT images. Second, a tumor template was produced by cropping the dilated GTV region in the planning CT image. Finally, a tumor location was estimated within the dilated GTV region in the MV-CBCT image by using the tumor template matching and calculating the centroid of the dilated GTV region. Gold standards of tumor locations were determined by a radiation oncologist and two radiological technologists in the clinical practice.
Results: A mean error between the gold standard and the tumor location estimated by the proposed method was 4.1 and standard deviation was 2.3 mm for 8 patients..
Conclusion: The results suggest that the proposed method using the MV-CBCT may be one of useful tools for tumor-based patient setup in SBRT.