Study On ⁶⁰Co And ¹⁹²Ir Sources In High-dose-rate Intracavity-interstitial Brachytherapy Of Cervical Cancer

High-dose-rate（HDR）intracavity-interstitial（IC-IS）brachytherapy（BT）is an essential treatment for cervical cance.Currently ¹⁹²Ir occupied most of the market share,after continuous technical improvements,the geometry of miniaturized ⁶⁰Co sources has become comparable to that of ¹⁹²Ir sources.⁶⁰Co has significantly greater economic advantages,but the difference between ⁶⁰Co and ¹⁹²Ir in HDR IC-ISBT has yet to be tested.In addition,when designing the HDR IC-ISBT plan,how to shorten the implantation needles reconstruction time and improve the reconstruction accuracy is an urgent issue to be solved.Purpose:We designed HDR IC-ISBT plans for ⁶⁰Co and ¹⁹²Ir,and then compared the dosimetric characteristics and clinical applications of the two sources with the published literature to verify the feasibility of ⁶⁰Co instead of ¹⁹²Ir for HDR IC-ISBT.We developed a deep learning method based on 3D U-Net for automatic segmentation of implantation needles in CT images.This study could provide a good basis for the automation of brachytherapy treatment planning and provide a reference for workers in related fields.Material and methods:A total of 140 patients with cervical cancer receiving HDR IC-ISBT were included in this study.We made ⁶⁰Co and ¹⁹²Ir HDR IC-ISBT plans for 30 of these patients,respectively.The dose parameters of clinical target volume（CTV）and organs at risk（OARs）were extracted from TPS.The equivalent dose in2 Gy/f（EQD2）,equivalent uniform biological effective dose（EUBED）,equivalent uniform dose（EUD）,tumor control probability（TCP）,and normal tissue complication probability（NTCP）were calculated using radiobiological models.The differences between the ⁶⁰Co and ¹⁹²Ir IC-ISBT plans were compared using the paired t-test.The remaining 110 patients were divided into training data,validation data and test data in the ratio of 80:20:10.We used a deep learning method based on3D U-Net for implantation needles image segmentation.The model was evaluated using the Dice similarity coefficient（DSC）.Results:At D_2cc and D_1cc,for ⁶⁰Co versus ¹⁹²Ir,there was a statistically significant difference in the physical dose and the EQD₂ for each OAR（p<0.05）,with the mean value of ⁶⁰Co lower than ¹⁹²Ir.When ⁶⁰Co≥1 Ci and ¹⁹²Ir≥4 Ci,or⁶⁰Co=0.5 Ci and ¹⁹²Ir=2 Ci,the EUBED ratio（60Co/192Ir）at the CTV was mostly close to 1,and the difference between them was not significant;at the OARs,the mean value of ⁶⁰Co was lower than that of ¹⁹²Ir.There was no statistical difference between ⁶⁰Co and ¹⁹²Ir in the EUD（93.93 versus 93.92 Gy）and TCP（97.07%versus 97.08%）of the tumors.The mean NTCP value of ⁶⁰Co was lower than that of¹⁹²Ir.We proposed a 3D U-Net-based deep learning method for automatic segmentation and reconstruction of implantation needles.Because of the early stopping,the model finished training after 32 epochs.The average DSC was 0.50 for the training data and the validation data.In the future,we will make further improvements based on the current results.Conclusions:Considering the CTV and OARs,the dosimetric parameters of60Co and 192Ir are comparable.Compared with 192Ir,the use of 60Co for HDR IC-ISBT can ensure a similar tumor control probability while providing better protection to the OARs.In addition,60Co has obvious economic advantages and can be promoted as a good alternative to 192Ir.In this study,we proposed a deep learning model based on 3D U-Net.In the future,the model will be able to perform implantation needles segmentation with better accuracy and shorten the reconstruction time,and provide the foundation for the automation of brachytherapy treatment planning.