Clinical Application of Different Image Segmentation Software for the Delineation of Organs
at Risk in Thoracic Tumor Radiotherapy
GAO Wenchao1, WANG Junliang1, ZHENG Qianqian2, CAO Jingli1, LI Jianxiong1
1. Department of Radiation Oncology, Fifth Medical Center of PLA General Hospital, Beijing 100071, China;
2. Department of Radiation Oncology, Beijing Shijingshan Hospital, Beijing 100043, China
Abstract：Objective To compare the delineation effect between three automatic delineation software (rtStation, AccuContour, RTMind)
and manual delineation on thoracic organs at risk (OARs). Methods A total of 15 patients with thoracic tumor radiotherapy
in the Department of Radiotherapy of Fifth Medical Center of PLA General Hospital were selected. OARs were manually sketched
by the radiotherapy oncologists. After confirmation, the sketching results were used as the reference standard, and then OARs
were sketched by three kinds of image segmentation software. The time, Dice similarity coefficient (DSC) and Hausdorff distance
(HD) of manual and automatic sketching were compared. Results The speed of the three automatic contouring software was
AccuContour>RT-Mind>rtStation, and the contouring time was all completed within 3 min, while manual contouring took about
30 min. The three kinds of automatic contour software had good delineation effect for heart, left lung, right lung and spinal cord,
with DSC above 0.85, but poor delineation effect for esophagus and trachea, and the DSC of AccuContour and RT-Mind software
in delineating heart and spinal cord was significantly higher than that of rtStation software (P<0.05). AccuContour software was
significantly superior to rtStation and RT-Mind software in DSC mapping of left and right lung (P<0.05). The spinal cord HD values
of the three automatic contour software were the lowest, and the tracheal HD value was the highest, but the spinal cord HD values of
AccuContour and RT-Mind software were significantly lower than those of rtStation software (P<0.05). The cardiac HD of rtStation
software was significantly higher than that of AccuContour and RT-Mind software (P<0.05). The esophageal HD of the three kinds
of automatic contour software was lower than 2 cm, and AccuContour software was significantly better than rtStation and RT-Mind
software (P<0.05). For left and right lung, RT-Mind software HD exceeded 2 cm, which was significantly higher than rtStation and
AccuContour (P<0.05), and the contour effect was poor. Conclusion When the three kinds of image segmentation software were
used to sketch the chest OARs automatically, the auto-segmentation results of esophagus and trachea are basically satisfactory,
but need to be manually modified by radiation oncologist, which needs to be further improved. All three image segmentation can
satisfactorily contour the heart, lungs and spinal cord, which can improve the work efficiency for radiation oncologist.
高文超1，王军良1，郑倩倩2，曹井丽1，李建雄1. 不同图像分割软件用于胸部危及器官自动勾画的临床应用[J]. 中国医疗设备, 2023, 38(7): 50-54.
GAO Wenchao1, WANG Junliang1, ZHENG Qianqian2, CAO Jingli1, LI Jianxiong1. Clinical Application of Different Image Segmentation Software for the Delineation of Organs
at Risk in Thoracic Tumor Radiotherapy. China Medical Devices, 2023, 38(7): 50-54.
