Objective:To develop a deep learning-based approach for predicting the dose distribution of intensity-modulated radiotherapy (IMRT) for breast cancer patients, and evaluate the feasibility of applying the predicted dose distribution in the automatic treatment planning.Methods:A total of 240 patients with left breast cancer admitted to Fudan University Shanghai Cancer Center were enrolled in this study: 200 cases in the training dataset, 20 cases in the validation dataset and 20 cases in the testing dataset. A modified deep residual neural network was trained to establish the relationship between CT image, the contouring images of target area and organs at risk (OARs) and the dose distribution, aiming to predict the dose distribution. The predicted dose distribution was utilized as the optimization objective function to optimize and generate a high-quality plan.Results:Compared with the dose distribution of clinical treatment plan, the predicted dose distribution for target areas and OARs showed no statistical significance except for a simultaneous boost target PTV 48Gy. And the treatment plan generated based on the predicted dose distribution was basically consistent with the predicted outcomes. Conclusion:Our results demonstrate that the deep learning-based approach for predicting the dose distribution of IMRT for breast cancer contributes to further achieving the goal of automatic treatment planning.

Objective:To find out the key factors that affect the clinical outcome of euploid single blastocyst transfer.Methods:This retrospective study included 956 patients' clinical data of frozen euploid single blastocyst transfer cycles for preimplantation genetic testing (PGT) in Shanghai Ji Ai Genetics and IVF Institute from January 2014 to January 2018. According to the outcome of transplantation, the patients were divided into pregnancy group ( n=509) and non-pregnancy group ( n=447), and the clinical data between the two groups were compared to screen out the key factors affecting the pregnancy rate of transplantation; at the same time, the pregnancy group was further divided into live birth subgroup and miscarriage subgroup, and the clinical data between the two subgroups were compared and the significant factors affecting the abortion were screened by logistic regression. Results:The endometrial thickness of pregnancy group was significantly thicker than that of non-pregnancy group [9(8,10) mm vs. 8(8,10) mm] ( P=0.006); the clinical pregnancy rate and the live birth rate of blastocyst with good quality were significantly higher than those with poor quality (59.2% vs. 46.3%, 50.2% vs. 37.7%) ( P=0.001). In addition, compared with miscarriage subgroup, blastocyst development time was a risk factor of abortion, and the OR value was 0.536 (95% CI=0.309-0.929). Conclusion:In the frozen embryo transfer cycles of single euploid blastocyst, the thicker endometrium, the faster development and the higher score of blastocyst are the key factors for the success of the implantation. In the process of maintaining the late pregnancy, the faster development of blastocyst (D5) has a higher live birth rate and a lower miscarriage rate.

Objective:To find out the key factors that affect the clinical outcome of euploid single blastocyst transfer.Methods:This retrospective study included 956 patients' clinical data of frozen euploid single blastocyst transfer cycles for preimplantation genetic testing (PGT) in Shanghai Ji Ai Genetics and IVF Institute from January 2014 to January 2018. According to the outcome of transplantation, the patients were divided into pregnancy group ( n=509) and non-pregnancy group ( n=447), and the clinical data between the two groups were compared to screen out the key factors affecting the pregnancy rate of transplantation; at the same time, the pregnancy group was further divided into live birth subgroup and miscarriage subgroup, and the clinical data between the two subgroups were compared and the significant factors affecting the abortion were screened by logistic regression. Results:The endometrial thickness of pregnancy group was significantly thicker than that of non-pregnancy group [9(8,10) mm vs. 8(8,10) mm] ( P=0.006); the clinical pregnancy rate and the live birth rate of blastocyst with good quality were significantly higher than those with poor quality (59.2% vs. 46.3%, 50.2% vs. 37.7%) ( P=0.001). In addition, compared with miscarriage subgroup, blastocyst development time was a risk factor of abortion, and the OR value was 0.536 (95% CI=0.309-0.929). Conclusion:In the frozen embryo transfer cycles of single euploid blastocyst, the thicker endometrium, the faster development and the higher score of blastocyst are the key factors for the success of the implantation. In the process of maintaining the late pregnancy, the faster development of blastocyst (D5) has a higher live birth rate and a lower miscarriage rate.

Objective: To study the clinical application of three-dimensional (3D) visualization technology in liver resection of complicated liver cancer. Methods: A retrospective analysis of 28 patients with complicated liver cancer treated from June 2017 to June 2018 in the Department of Hepatobiliary Surgery, the Affiliated Tumor Hospital of Guangxi Medical University. There were 26 males and 2 females, aged (46±10) years old. A treatment plan on how to perform liver resection for these patients was developed under the guidance of 3D visualization technology. The actual surgical procedures, operation time, intraoperative blood loss, and postoperative complications were documented. The virtual resected liver volume was compared with the actual resected liver volume. The virtual surgical resection margin was also compared with the actual surgical resection margin. Results: All the 28 patients with complicated liver cancer completed the 3D visualization analysis with the location, shape and quantity of tumor being clearly shown. Of the 27 patients who underwent liver resection, 13 underwent anatomical hepatectomy, and 14 underwent non-anatomical hepatectomy. The operation time ranged from 145 to 350 min (median 240 min). The intraoperative blood loss ranged from 100 to 1 500 ml (median 300 ml). The incisional wound healed slowly in 4 patients, pleural effusion developed in 8 patients, and ascites in 2 patients. There were no significant differences in the virtual resected liver volume compared with the actual resected liver volume (P>0.05). There was an excellent positive correlation between the patient's virtual resected liver volume and the actual resected liver volume (r=0.986, P<0.05). There was no significant difference between the virtual surgical resection margin and the actual surgical resection margin (P>0.05). There was an excellent positive correlation between virtual surgical resection margin and the actual surgical resection margin (r=0.983, P<0.05). Conclusion Three-dimensional visualization technology accurately assessed the liver status, optimized surgical procedures, and played an important role in liver resection of complex liver cancer.

Objective To establish and evaluate a novel and non-invasive diagnostic model using routine laboratory serological indexes in cirrhotic patients.Methods A retrospective study was conducted on 1044 consecutive patients with hepatocellular carcinoma (HCC) treated by hepatectomy in the Affiliated Tumor Hospital of Guangxi Medical University from September 2013 to December 2016.These patients were divided into a training cohort (n =783) and a validation cohort (n =261) using the 3 ∶ 1 matching principle.Logistic regression analysis was used to identify independent risk factors related to occurrence of cirrhosis in the training cohort,and then a PPH score was established.The accuracy of the model in predicting cirrhosis in two groups was evaluated respectively by the area under the receiver operating characteristic curve (AUC) and goodness of fit,and compared with the following commonly used predictive systems:the model for endstage liver disease (MELD) score,fibrosis index based on 4 factor score (FIB-4),Forns score and aspartate aminotransferase to platelet ratio index score (APRI).Results Univariate and multivariate Logistic regression analysis in the training cohort showed prothrombin time,platelet count and hepatitis B surface antigen positivity were closely related to occurrence of cirrhosis.The accuracy of the PPH score (AUC =0.705) in diagnosing cirrhosis in the training cohort was significantly better than the MELD score (AUC =0.557),APRI score (AUC =0.598),FIB-4 score (AUC =0.597) and Forns score (AUC =0.665).Similar results were obtained in the validation cohort (AUC:0.702 vs 0.554 vs 0.624 vs 0.634 vs 0.580).The goodness of fit indicated that there was no significant difference between the actual and predicted values of cirrhosis in the two cohorts,and the model was in good agreement.Conclusions A novel and non-invasive model for the diagnosis of cirrhosis was successfully established.The accuracy of this model in diagnosing cirrhosis was better than the MELD,APRI,Fib-4 and Forns scores.This model has significance in guiding clinical treatment decision in HCC patients with cirrhosis.

Objective Multiple software applications and systems and complex hardware are utilized in the process of radiotherapy. In this paper,Xenapp-based visualized system was adopted to integrate each system during the radiotherapy procedures. Methods Windows2008r2 operating system,Citrix Xenapp 6. 5 sp1 enterprise edition,sqlserverr2008 express database were utilized. The Xenapp server was installed and the commonly used software applications were installed and published on this server. CitrixReceiver was installed on the computers in the local network in Department of Radiotherapy. A web browser was utilized to log in the Xenapp server to access all the applications published on the server. Results The application of Citrix Xenapp-based virtualized system significantly reduced the cost of computer hardware by more than 50％,decreased the maintenance cost of technical personnel. Besides,it made the process of Department of Radiation Oncology seamless and improved the efficiency of the staff in the Department of Radiation Oncology. Conclusions The Xenapp-based virtualized system can save cost and improve efficiency in the process of radiotherapy,which is worthy of widespread application in clinical practice.

Objective To study the impact of anatomical resection (AR) for hepatocellular carcinoma with microvascular invasion on early tumor recurrence.Methods 178 consecutive patients with solitary hepatocellular carcinoma who underwent liver resection at the Affiliated Tumor Hospital of Guangxi Medical University from January 2012 to December 2015 were retrospectively studied.These patients were divided into four groups according the types of resection (anatomical resection AR or non-anatomical resection NAR) and the surgical margins (narrow or wide margin).All the patients were followed up until the tumor recurred or two years after surgery.The recurrence-free survival and the types of recurrence for the 4 different groups were compared.Results The 6 months,1-,2-year disease-free survival rates in the AR (n =55) and NAR groups (n =123) were 87.0％,79.2％,74.5％ and 78.5％,61.3％,45.7 ％,respectively,(P ＜ 0.05).Through pair-wise comparisons of the four groups,there were no significant differences in early recurrence between the narrow surgical margin group and the wide surgical margin group,regardless of the types of surgery.The early recurrence rates of the AR groups were significantly lower than that of the NAR groups,regardless of the widths of the surgical margins.Multivariate analysis showed that AR was a protective factor of early recurrence (HR =0.417,95％ CI 0.229 ～ 0.761).Further analysis of the recurrence patterns of AR and NAR showed that the recurrence types were mainly sohtary for AR (solitary,61.5％;multiple 30.8％) and multiple recurrence for NAR (solitary,32.2％;multiple 61.0％).Conclusion AR improved early recurrence-free survival of patients with microvascular invasion.

Objective To study the value of serum prealbumin-bilirubin score (PALBI) in predicting posthepatectomy liver failure (PHLF) for patients with HBV related hepatocellular carcinoma (HCC).Methods A retrospective study was conducted on 919 HBV-related HCC patients who underwent hepatectomy from September 2013 to December 2016 at the Affiliated Tumor Hospital of Guangxi Medical University.These patients were divided into a training cohort (n =689) and a validation cohort (n =230) using the 3 ∶ 1 matching principle.The training cohort was divided into the control group (n=546) and the PHLF group (n=143) according to whether PHLF occurred.The multivariate logistic regression model was used to analyze the factors related to PHLF in the training cohort,and then the PALBI score was established.The ability of the PALBI score to predict PHLF was evaluated by the area under the receiver operating characteristic curve (AUC) and compared with the Child-Pugh,model for end-stage liver disease (MELD),and albumin-bilirubin (ALBI) scores.Results Univariate and multivariate logistic regression analyses showed the factors including HBV-DNA≥ 103 IU/ml,total bilirubin,prealbumin,platelet count,AST,prothrombin time,intraoperative blood loss ≥400 ml and major liver resection were closely related to PHLF.The ability of the PALBI score (AUC =0.733) to predict PHLF preoperatively was superior to the ChildPugh score (AUC =0.562),the MELD score (AUC =0.652) and the ALBI score (AUC =0.683) in the entire training cohort.Similar results were obtained in the entire validation cohort (AUC:0.752 vs.0.599 vs.0.641 vs.0.678).To eliminate the effect of a small residual liver volume on PHLF,the ability of each of these scores in the training and validation cohorts to predict PHLF was calculated respectively in these 2 cohorts of patients who underwent only minor liver resection,and similar results were obtained.Conclusion The PALBI score was significantly superior to the Child-Pugh,MELD and ALBI scores in predicting PHLF in patients with HBV-related HCC who underwent liver resection.The PALBI score is a simple,non-invasive and reliable novel model in predicting PHLF.

Objective To determine whether Auto-Planning-based volumetric modulated radiotherapy(Auto-VMAT)planning can improve planning efficiency without compromising plan quality compared with current manual trial-and-error-based volumetric modulated arc therapy(Manual-VMAT) planning for patients with rectal cancer. Methods Ten patients with stage Ⅱ-Ⅲ rectal cancer who underwent Dixon surgery were enrolled as subjects. The Pinnacle 9.10 planning system was used to design Manual-VMAT and Auto-VMAT plans. Dose distribution,homogeneity index(HI),conformity index(CI), D meanvalues of different organs at risk or dose-volume histogram of regions of interest,total planning time, and manual planning time were compared between the two plans. The differences were analyzed by paired t test. Results Dosimetric prescriptions were achieved in both plans. There were no significant differences in HI or CI between the Auto-VMAT plans and the Manual-VMAT plans(0.058 vs. 0.058, P=0.972;0.921 vs. 0.940,P=0.115). Compared with the Manual-VMAT plans,the V 40,D mean,and D 50%of the bladder were significantly reduced by 25.6%, 11.5%, and 8.9%, respectively, in the Auto-VMAT plans(P=0.004,0.016,0.001);the V 40,D mean,and D 50%of the small intestine were also significantly reduced by 12.1%,5.4%,and 6.8%,respectively,in the Auto-VMAT plans(P=0.023,0.001,0.001);the V 30, D mean,and D 50%of the left and right femoral heads were slightly reduced in the Auto-VMAT plans. The Auto-VMAT plans had significantly longer total planning time but significantly shorter manual planning time than the Manual-VMAT plans(50.38 vs. 36.81 min, P= 0.000;4.47 vs. 16.94 min, P= 0.000). Conclusions Compared with the Manual-VMAT plans, the Auto-VMAT plans have substantially shorter manual planning time and improved planning efficiency.

Background and purpose:Radiomics is an emerging field that generates large amounts of valuable clinical information through extracting quantitative imaging features. The purpose of this study was to use the radiomics approach to assess the value of features captured from PET and CT in predicting the therapeutic effect in stageⅠ non-small cell lung cancer (NSCLC) after stereotactic ablative radiotherapy (SABR).Methods:Patients with stageⅠ NSCLC conifrmed by pathology and treated with SABR were included retrospectively. The gross tumor volume (GTV) was deifned by two radiologists. PET and CT scan images were collected, and radiomic features were further extracted and analyzed. Non-negative matrix factorization was used to distinguish patients with or without local control.Results:Sixteen patients were eligible for analysis. This study identiifed two PET features (LL_GLCM_Maximal_Correlation_Coeffcient and HL_GLRMS_LRE) captured from PET/CT as having signiifcance in classifying patients with or without disease development. This study not ifnd similar results in CT scans.Conclusion:It seems feasible to use radiomics information effects from PET/CT to predict therapeutic effects of SABR in stageⅠ NSCLC. Further investigation is needed.