Background and purpose: Radiation therapy has entered the era of precise radiotherapy. Setup error becomes important factor affecting the effects of radiotherapy. The aim of this study was to analyze the setup accuracy of the in-house developed technique of body plate with vacuum cushion and thermoplastic mask (Group A) and the conventional technique of arm support (Group B) in thoracic tumor radiotherapy. Methods: A total of 19 patients with thoracic tumor were enrolled in this study and randomly separated into Group A and Group B. The patients of Group A underwent the secondary set-up: align the marker on vacuum cushion and patient’s body to the room laser

cover the patient’s body with thermoplastic mask and align the marker on the mask to the room laser. The patients of Group B were directly setup by aligning the marker on the patient’s body to the room laser. The kilo-voltage cone beam computed tomography (KV-CBCT) was performed on each patient to collect the pre- and post-treatment CBCT images. The CBCT images were registered to the planning CT to analyze the translational error of Group A and Group B. Results: The pre-treatment set-up errors of Group A vs Group B were (1.06±0.58) vs (1.82±0.82) mm in left and right (LR) direction

(1.31±0.40) vs (2.18±1.20) mm in superior and inferior (SI) direction

and (1.28±0.66) vs (2.94±1.81) mm in anterior and posterior (AP) direction. The post-treatment set-up errors of Group A vs Group B were (0.86±0.54) vs (1.29±0.58) mm in LR direction

(1.07±0.58) vs (1.08±0.45) mm in SI direction

and (0.98±0.53) vs (1.56±0.63) mm in AP direction.Conclusion: The in-house developed immobilization technique of body plate with vacuum cushion and thermoplastic mask was more accurate and reproducible than the conventional immobilization technique of arm support in thoracic tumor radiotherapy.