Objective To observe the performance of ConvNeXt architecture model(SC2-Net)integrated with feedforward attention(FA)for segmentation of pancreas from abdominal CT images.Methods 3D abdominal CT images of 80 healthy adults(Dataset 1)and 68 patients with pancreatic lesions(Dataset 2)were included.ConvNeXt network model was established and enhanced by introducing a FA mechanism,a scalable convolution block(SCB)and a feature gating(FG)module into the encoder section.The performance of the model for segmenting pancreas were comparatively evaluated with other models(Swin UNETR,nnFormer,UNETR,TransBTS models based on Transformer and 3D UX-NET model based on ConvNeXt),while conduct ablation experiments were performed on the added modules.Results SC2-Net accurately segmented pancreas from abdominal CT images,with Dice similarity coefficient(DSC),95％Hausdorff distance(HD95)and the mean surface distance(MSD)of 0.92±0.01,(1.08±0.05)mm and(2.12±0.01)mm in Dataset 1,respectively.The DSC and HD95 of SC2-Net segmentation of pancreas were both superior to those of other models.In Dataset 2,SC 2-Net achieved DSC,HD95 and MSD of 0.82±0.03,(3.35±0.36)mm and(0.87±0.15)mm,respectively,surpassing all other models.SC2-Net achieved complete pancreas segmentation in both datasets,whereas other models demonstrated under-segmentation or mis-segmentation.FA module significantly improved segmentation performance when integrated into the baseline network.Conclusion SC2-Net could improve segmentation of pancreas from abdominal CT images.

Objective To investigate the feasibility and accuracy of convolutional neural networks for automatically delineating the pancreatic cancer gross target volume(GTV)in pancreatic enhanced CT.Methods The localizable enhanced CT images of 114 patients with pancreatic cancer were retrospectively selected,in which the GTV was manually delineated using AccuContour.The imaging data were then import to AccuLearning and randomly divided as the training set,validation set and test set at a ratio of 8:1:1.Flex and Segresnet were used to train the automatic segmentation model,with each network structure trained continuously 3 times using fixed training parameters.The model was evaluated in terms of Dice similarity coefficient(DSC),95%Hausdorff distance(HD95),average symmetric surface distance(ASSD)and relative volume difference(RVD).Results In the model training phase,Flex-3 test results in Flex group were the worst,with a minimum DSC of 0.14%and an average DSC of 56.30%,while Flex-1 performed well,achieving a minimum DSC of 47.90%and an average DSC of 67.35%.Meanwhile,Segresnet-2 in Segresnet group had the worst test results,with a minimum DSC of 0.00%and an average DSC of 42.46%,while Segresnet-3 test results were better,with a minimum DSC of 42.65%and an average DSC of 63.28%.In the fixed testing phase,the best results among all were as follows:average DSC and RVD values of 63.88%and 29.41%in Segresnet-3 group,average ASSD value of 4.43 mm in Segresnet-2 group,and average HD95 value of 12.87 mm in Segresnet-1 group.Conclusion Both Flex and Segresnet architectures of convolutional neural network can be used for the automatic pancreatic tumor GTV segmentation training,with Segresnet performing better in comprehensive evaluation.

Objective To investigate the feasibility and accuracy of convolutional neural networks for automatically delineating the pancreatic cancer gross target volume(GTV)in pancreatic enhanced CT.Methods The localizable enhanced CT images of 114 patients with pancreatic cancer were retrospectively selected,in which the GTV was manually delineated using AccuContour.The imaging data were then import to AccuLearning and randomly divided as the training set,validation set and test set at a ratio of 8:1:1.Flex and Segresnet were used to train the automatic segmentation model,with each network structure trained continuously 3 times using fixed training parameters.The model was evaluated in terms of Dice similarity coefficient(DSC),95%Hausdorff distance(HD95),average symmetric surface distance(ASSD)and relative volume difference(RVD).Results In the model training phase,Flex-3 test results in Flex group were the worst,with a minimum DSC of 0.14%and an average DSC of 56.30%,while Flex-1 performed well,achieving a minimum DSC of 47.90%and an average DSC of 67.35%.Meanwhile,Segresnet-2 in Segresnet group had the worst test results,with a minimum DSC of 0.00%and an average DSC of 42.46%,while Segresnet-3 test results were better,with a minimum DSC of 42.65%and an average DSC of 63.28%.In the fixed testing phase,the best results among all were as follows:average DSC and RVD values of 63.88%and 29.41%in Segresnet-3 group,average ASSD value of 4.43 mm in Segresnet-2 group,and average HD95 value of 12.87 mm in Segresnet-1 group.Conclusion Both Flex and Segresnet architectures of convolutional neural network can be used for the automatic pancreatic tumor GTV segmentation training,with Segresnet performing better in comprehensive evaluation.

Objective To observe the performance of ConvNeXt architecture model(SC2-Net)integrated with feedforward attention(FA)for segmentation of pancreas from abdominal CT images.Methods 3D abdominal CT images of 80 healthy adults(Dataset 1)and 68 patients with pancreatic lesions(Dataset 2)were included.ConvNeXt network model was established and enhanced by introducing a FA mechanism,a scalable convolution block(SCB)and a feature gating(FG)module into the encoder section.The performance of the model for segmenting pancreas were comparatively evaluated with other models(Swin UNETR,nnFormer,UNETR,TransBTS models based on Transformer and 3D UX-NET model based on ConvNeXt),while conduct ablation experiments were performed on the added modules.Results SC2-Net accurately segmented pancreas from abdominal CT images,with Dice similarity coefficient(DSC),95％Hausdorff distance(HD95)and the mean surface distance(MSD)of 0.92±0.01,(1.08±0.05)mm and(2.12±0.01)mm in Dataset 1,respectively.The DSC and HD95 of SC2-Net segmentation of pancreas were both superior to those of other models.In Dataset 2,SC 2-Net achieved DSC,HD95 and MSD of 0.82±0.03,(3.35±0.36)mm and(0.87±0.15)mm,respectively,surpassing all other models.SC2-Net achieved complete pancreas segmentation in both datasets,whereas other models demonstrated under-segmentation or mis-segmentation.FA module significantly improved segmentation performance when integrated into the baseline network.Conclusion SC2-Net could improve segmentation of pancreas from abdominal CT images.

Objective:To compare dose distributions of hypofractionated radiotherapy for pancreatic cancer between IMPT and VMAT.Methods:Ten pancreatic cancer cases were included in this retrospective study. Photon (Edge) and proton (Proteus?PLUS) plans were designed by Eclipse and RayStation TPS, respectively. All plans were transferred to MIM system for extraction of parameters, which included Dmin, Dmean and Dmax of PTV, conformity index (CI), new conformity index (nCI), homogeneity index (HI), gradient index (GI), coverage, Dmax and dose-volume of the organs at risk (OARs). Results:There was no significant difference in CI between the two groups. The higher PTV Dmin, Dmean, Dmax, D98%, D2%, HI, coverage and the better GI, D2 cmwere found in VMAT ( t/ Z=-4.63-5.32, P<0.05). The lower 10%_PD was found in IMPT ( t=-7.47, P<0.05). Regarding the OARs, Dmax of the intestine, stomach, and duodenum and Dmean of the left kidney were similar between two groups without significant difference ( P>0.05). The D5 cm 3 of the intestine, D10 cm 3 of the stomach, D5 cm 3 and D10 cm 3 of the duodenum, D2/3 of the left kidney, Dmean and D2/3 of the right kidney were lower in IMPT than those in VMAT ( t/ Z=-8.12--2.60, P<0.05). However, the Dmax and D0.35 cm 3 of the spinal cord were higher in IMPT than those in VMAT ( t=7.30, 6.77, P<0.05). Conclusions:Both of hypofractionated radiotherapy plans of pancreatic cancer designed by VMAT and IMPT could meet clinical needs. No significant difference was found in Dmax of the adjacent gastrointestinal tracts between the two groups. While IMPT had the advantage over VMAT in the case of lower dose-volumes of the gastrointestinal tracts. Nevertheless, less protections of the OARs in front of the tumor volume could be provided by IMPT compared with VMAT.

Objective:To compare the dose distribution among CyberKnife, Tomotherapy, Edge, Triology and γ-knife in stereotactic body radiation therapy (SBRT) for pancreatic cancer.Methods:Clinical data of 10 panreatic cancer patients receiving CyberKinife treatment were retrospectively analyzed. The treatment plans were designed by five apparatuses from five centers according to the uniform requirement. All plans were transferred to MIM system for the extraction of parameters, which mainly included D min, D mean and D max of PTV, conformity index (CI), new conformity index (nCI), homogeneity index (HI), gradient index (GI), coverage, D max and dose-volume of the stomach and bowel. Results:The best CI and nCI were obtained in Triology ( P<0.001), and the worst HI was found in γ-knife ( P<0.001). The best GI was found in CyberKnife, followed by γ-knife and Tomotherapy, and Edge showed the worst GI ( P<0.001). The highest D min of PTV was found in both Edge and Triology, while lower D min of PTV was found in CyberKnife and Tomotherapy ( P<0.001). Additionally, γ-knife provided the highest D mean and D max of PTV ( P<0.001). Regarding the organs at risk, the lowest D max and D 5cm 3 of the bowel ( P<0.001), D max of the stomach ( P=0.003), D max( P=0.001), D 5cm 3 ( P<0.001) and D 10cm 3 ( P=0.005) of the duodenum, D max( P<0.001) and D 0.35cm 3 ( P<0.001) of the spinal cord were found in CyberKnife. The highest D max of the bowel was found in γ-knife. Furthermore, the highest D 5cm 3 of the duodenum was demonstrated in Edge ( P<0.001) and Tomotherapy provided the highest D max( P<0.001) and D 0.35cm 3 of the spinal cord ( P<0.001). Conclusions:All five radiotherapy apparatuses can meet the requirement of SBRT for pancreatic cancer. More rapid dose fall-off could be obtained via CyberKnife and γ-knife. Triology and Edge provide better target conformity. CyberKnife can better protect the gastrointestinal tract.

Objective The aim of this study was to investigate the dosimetric advantages of Gating in the treatment of prima-ry hepatic cancer with large segmentation. Methods A retrospective analysis of 10 patients with primary liver cancer from August 2017 to November 2018 after interventional therapy was performed using three consecutive phases of end-tidal phase to achieve pa-tient-controlled large-segment radiotherapy. Ten patients underwent 4DCT localization scan,and 10 respiratory phase sequences were reconstructed by respiratory wave-form,and the images were transmitted to the MIM6. 7. 6 workstation. In the MIM workstation, full-time phase maximum density projection(MIP-10),full-time phase average density projection(Mean-10),end-expiration 3 phase maximum density projection(MIP-3) and end-expiration 3 phase average density projection( Mean-3) were generated re-spectively,where MIP was used for target delineation and Mean for dose calculation. The radiotherapy doctor delineated IGTV-10 and IGTV-3 on the MIM workstation,and released CTV-10,CTV-3,PTV-10 and PTV-3 to compare the volume differences of the target area. After the target area was drawn,the image was transmitted from the MIM workstation to the Eclipse treatment planning sys-tem,and the full-time phase plan(Plan-10)with the same conditions and three consecutive phase-phase gating plans(Plan-3) were prepared. The prescriptive dosage was given at 50 Gy/10 f/2weeks. Comparing the HI and CI of the target area,the comparison of organs at risk included: the average dose of liver Dmean,the irradiation volume of liver less than 15Gy,the Dmax of small intestine, the Dmax of colon, the Dmax of stomach, the average dose of the kidney Dmean, the heart Dmax, and the spinal cord Dmax. Results The volume of the target area delineated at the end of expiratory phase was less than that of the target area outlined by the full-time phase in IGTV,CTV and PTV,and the difference was statistically significant(P<0. 05). In the two groups of seven field coplanar lage-segment radiotherapy plans,the 3-phase respiratory gating plan significantly reduced the dose of the organs at risk, and the difference was statistically significant(P<0. 05). At the same time,there was no statistically difference in the HI and CI be-tween of the two groups(P>0. 05). Conclusion The gated target area delineation and planning design of the three consecutive pha-ses of end-tidal phase reduce the volume of IGTV,CTV and PTV target regions compared with the selection of full-time phase,and have obvious advantages in the planned dosimetry. The irradiation dose that threatens the organs is worthy of being promoted and ap-plied in the large-scale radiotherapy of liver cancer.

Objective To propose the method of dose distribution calculated by one-step optimization with 7 shells (Cao method) and compare with that by three-step optimization with 4 shells (Blanck method) and CyberKnife treatment plans for pancreatic cancer. Methods 20 cases of pancreatic cancer who underwent CyberKnife treatment were retrospectively analyzed,and CT was performed to localize and delineate the target area and endangering organs. Dosage was optimized and evaluated with Blanck method and Cao method. The planning target volume (PTV) conformity index (CI), new conformity index (nCI), homogeneity index (HI),gradient index (GI), coverage, dose-volume and doses to organs at risk were compared. Results Compared with Blanck method, CI (1.11 ± 0.05 vs 1.15 ± 0.05), nCI (1.20 ± 0.06 vs 1.23 ± 0.06), coverage [(92.48 ± 1.85)% vs (93.53 ± 2.15)%], volumes encompassed by 100% and 30% prescription dose line (36.46 ± 16.64 vs 38.19 ± 17.68; 286.19 ± 126.52 vs 320.93 ± 154.82) and monitor unit (56 369 ± 20 019 vs 57 814 ± 20 531) were significantly decreased,while GI was increased (3.22 ± 0.19 vs 3.11 ± 0.19), and all the differences were statistically significant (P<0.05). Additionally, Dmax of the intestine (21.17 ± 2.90 vs 20.63 ± 3.13), D10cc of the stomach (12.78 ± 2.57 vs 13.11 ± 2.43), D5ccof the duodenum (11.01 ± 3.45 vs 11.50 ± 3.25), D10ccof the duodenum (9.30 ± 3.31 vs 9.78 ± 3.07) and D0.35ccof the spinal cord (6.09 ± 0.98 vs 6.59 ± 0.92) were all significantly decreased (P<0.05). No significant differences were found on other parameters. Conclusions Better dose distributions are accessible by one-step optimization with 7 shells in CyberKnife treatment plans for pancreatic cancer.

Objective To assess the cumulative doses and side effects after two courses of stereotactic body radiation therapy (SBRT) for pancreatic cancer. Methods Twenty-four pancreatic cancer patients who received two courses of SBRT were enrolled. Organ endangering dose accumulations were calculated by rigid and non-rigid registration. All doses were recalculated to an equivalent dose of 2 Gy per fraction. Results The median of accumulated maximal dosage (Dmax) and dosage per 1cc(D1cc) of the stomach,duodenum and the bowel were 43.87 and 35.28 Gy 3,35.53 and 26.59 Gy3,45.08 and 36.18 Gy3; and the median volume under the dosage of 10Gy (V10) was 107.40,23.98 and 169.26cc, respectively. The median accumulated Dmaxand the dosage of 35% volume(D0.35) of the spinal cord was 8.42 and 7.83Gy3. The median cumulative Dmeanand D2/3of the left and right kidney were 5.18 and 3.65 Gy3, 3.50 and 2.57 Gy3, respectively. The median cumulative Dmeanand D50%of the liver was 5.18 and 3.64Gy3,respectively. The median summed dose to the overlapping radiation field of the two courses was 93.38 Gy3. No grade 3-4 toxicity occurred. Conclusions The cumulative doses to organs at risk as dose constraints were safe and acceptable,which could be used as a reference to evaluate whether a second SBRT could be done after initial SBRT for pancreatic cancer.

Objective To investigate the safety and efficacy of re-irradiation with stereotactic body radiotherapy(SBRT) for treating locally recurrent advanced pancreatic cancer.Methods From 2014 to 2017,7 patients with stage Ⅲ pancreatic cancer were treated by re-irradiated with SBRT at Shanghai Changhai Hospital.SBRT was delivered via the G4 type cyberknife robotic stereotactic radiosurgery system in all the patients.The median dose of the first SBRT was 35Gy/5-7 fx,and the median dose of re-irradiation with SBRT was 31 Gy/5-8 fx.6 patients had undergone sequential chemotherapy either with gemcitabine or S-1 based therapy except one patient who refused the chemotherapy.Results There were 5 male and 2 female patients.The median overall survival (OS) of 7 patients was 30 months.Patients were re-irradiated with SBRT after a median interval of 10 months after the first SBRT.Median OS and locally relapse-free survival (LFRS) from re-irradiation were 13 months and 11 months,respectively.Three months after re-irradiation,3(42.9％) patients had partial remission and 4 patients had stable disease.Pain disappeared in 4 patients at the end of reirradiation and significant pain was alleviated in 2 patients 1 month after re-irradiation.There were no toxicities of grade 3 or higher grade during two courses of SBRT.Conclusions For patient with locally recurrent advanced pancreatic cancer,SBRT re irradiation regimen was associated with acceptable toxicity,which can effectively alleviate the pain,prolong the survival and improve the life quality.