1. a. Department of Medical Equipment丨 b. Radiation Therapy Center丨 c. Department of Medical Image, Nanjing First Hospital,
Nanjing Medical University, Nanjing Jiangsu 210006, China丨 2. Medical Physics Research Center, Nanjing Medical University,
Nanjing Jiangsu 210029, China丨 3. Nanjing Emergency Medical Center, Nanjing Jiangsu 210003, China
Abstract：Objective To compare the imaging differences of bladder structure in CT and Cone Beam CT (CBCT) images of patients
with cervical cancer after radiotherapy, and quantitatively evaluate the accuracy of three automatic delineation software in these
images, and to explore the feasibility of existing models for automatically delineate CBCT images, to achieve rapid assessment
based on bladder filling state of CBCT images provides research foundation. Methods The location CT images and CBCT images
of 20 patients with cervical cancer by radiotherapy in our hospital were retrospectively analyzed. The imaging differences of bladder
structure were analyzed by mean values of HU (HUmean), entropy (ENT) and inverse different moment (IDM). The accuracy of
automatically delineated bladder structures in CT and CBCT images was quantitatively evaluated by volume difference (ΔV%) and
dice similarity coefficient (DSC). Results The ENT, IDM, HUmean of bladder structure in CT images were 2.79±0.40, 0.72±0.03,
7.64±15.06, and 5.12±0.59, 0.54±0.06, 26.17±24.19 in CBCT images, respectively. The differences among all indicators were
statistically significant (P<0.01). The DSC values of the automatically delineated bladder structures from three softwares were all
greater than 0.7 in CT and less than 0.7 in CBCT. The difference of volume between manual and automatic structures in CT was less
than that of CBCT. Conclusion The fluctuation of CT value and noise of bladder structure in CBCT images are significantly higher
than those in CT images. The three kinds of automatic delineation software based on deep learning can accurately delineate bladder
structures in CT images, but can not be used in CBCT images yet, due to the influence of original training data and inherent imaging
differences between images.
赵紫婷1a，时飞跃1b, 2，王敏1b，周蕾蕾1c，吴倩倩1a，蒋红兵1a,3. 膀胱结构在CT与CBCT图像及自动勾画软件间的比较[J]. 中国医疗设备, 2021, 36(4): 5-8.
ZHAO Ziting1a，SHI Feiyue1b, 2，WANG Min1b，ZHOU Leilei1c，WU Qianqian1a，JIANG Hongbing1a,3. Comparison of Bladder Structures between Planning CT and CBCT Images and in Different
Automatic Delineation Softwares. China Medical Devices, 2021, 36(4): 5-8.
