三期相CT图像分割模型的鼻咽癌肿瘤靶区自动勾画研究

目的 探究使用基于3D U-Net结合三期相CT图像的分割模型对鼻咽癌肿瘤原发灶(GTVnx)和转移的区域淋巴结(GTVnd)自动勾画的有效性和可行性。方法 回顾性收集215例鼻咽癌病例的电子计算机体层扫描(CT),包括平扫期(CT)、增强期(CTC)和延迟期(CTD)3个期相共计645组图像。采用随机数字表法,将数据集划分为172例训练集和43例的测试集。设置了包括三期相CT图像模型及期相微调模型共计6个实验组:三期相CT图像模型即仅使用平扫期(CT)A1组、仅使用增强期(CTC)A2组、仅使用延迟期(CTD)A3组和同时使用三期相(All)A4组。期相微调模型:CTC微调B1组和CTD微调B2组。使用Dice相似性系数(DSC)和95%豪斯多夫距离(HD95)作为定量评价指标。结果 使用三期相CT(A4)进行GTVnd靶区自动勾画相比于仅使用单期相CT(A1、A2、A3)获得更好的勾画效果(DSC:0.67 vs. 0.61、0.64、0.64, t=7.48、3.27、4.84,P<0.01; HD95: 36.45 mm vs. 79.23、59.55、65.17 mm,t=5.24、2.99、3.89,P<0.01),差异有统计学意义。使用三期相CT(A4)对于GTVnx的自动勾画效果相比于仅使用单期相(A1、A2、A3)无明显提升(DSC: 0.73 vs. 0.74、0.74、0.74;HD95: 14.17 mm vs. 8.06、8.11、8.10 mm),差异无统计学意义(P>0.05)。在GTVnd的自动勾画中,B2、B3 vs. A1模型具有更好的自动勾画精度(DSC:0.63、0.63 vs. 0.61, t=4.10、3.03, P<0.01;HD95:58.11、50.31 mm vs. 79.23 mm,t=2.75、3.10, P<0.01)。结论 使用三期CT扫描对于鼻咽癌GTVnd靶区具有更好的自动勾画效果。通过期相微调模型,可以提升平扫CT图像上GTVnd靶区的自动勾画精度。     

基于图谱库的自动轮廓勾画软件在宫颈癌自适应放疗中的应用

目的 评估基于图谱库的自动轮廓勾画软件(ABAS)在宫颈癌自适应放疗中的应用.方法 选取2014年1月至3月收治的21例已行第1程调强放疗的宫颈癌患者,将其已勾画器官的CT图像及第2程未勾画器官的定位CT图像传输至ABAS软件系统,以第1程图像为模板图像,第2程定位图像作为目标图像.在第2程定位图像上手工勾画出靶区和危及器官,将ABAS软件自动勾画图像和医师手工勾画的靶区、危及器官图像传输至飞利浦Pinnacle计划系统,对两组结果进行评估.比较相似性指数(DSC)和勾画体积.结果 ABAS自动勾画与医师手工勾画的DSC平均值均大于0.7,其中靶区的DSC最大为肿瘤临床靶区(CTV, 0.89±0.08),最低为肿瘤区(GTV, 0.72±0.16).对于危及器官,DSC最高的为右股骨头(0.88±0.05),最低为直肠(0.73±0.07).左右髂骨自动勾画体积较手工勾画小,且差异具有统计学意义(t=3.37、2.74, P<0.05),其他轮廓体积差异无统计学意义.结论 宫颈癌放疗过程中,基于图谱库的ABAS勾画软件,节省了临床器官勾画工作时间,加强自动勾画后的轮廓修改,并建立患者模板数据库,得到满意度更高的重合结果,也为开展自适应放疗提供了强有力的支持.     

基于磁共振图像手工勾画宫颈癌危及器官稳定性分析

目的 以宫颈癌MR图像为基础,分析不同勾画者、同一勾画者不同时间手工勾画危及器官（OARs）的稳定性,为放疗计划设计的序列选择进行初步探索。方法 回顾性分析中山大学肿瘤防治中心放疗科2016—2018年经MR-sim扫描的30例宫颈癌患者,选取共有的T1WI、T1dixonc和T2WI 3序列MR图像导入Monaco计划系统,由两位临床放疗医师独立在每个患者3个序列上分别勾画膀胱、直肠、肛管、左/右股骨头。其中一位医师在完成首次勾画工作后的一个月再次完成T1WI序列及各OARs的勾画。统计分析各OARs的相似性系数（DSC）、豪斯多夫距离（HD）和位置差异（Δx、Δy、Δz）。结果 不同勾画者在T1WI、T1dixonc、T2WI序列和同一个勾画者在T1WI序列上勾画5个OARs的HD值均<2 mm;位置差异均<5 mm。不同勾画者和同一个勾画者不同时间勾画的DSC、HD及位置差异与OARs体积呈正相关性（R=0.178~0.582,P<0.05）。因肛管体积较小（7.385±1.555）cm³,DSC值均<0.7表现稍差外,其余OARs平均DSC值均>0.82。通过两两比较3个序列上不同勾画者勾画OARs的DSC、HD发现,T1WI序列直肠、左/右股骨头的DSC值、膀胱、左/右股骨头、直肠的HD值以及肛管、右股骨头Δz轴差异均优于T1dixonc,差异有统计学意义（t=-3.116~3.604,P<0.05）;T1WI序列直肠DSC值和肛管HD值较T2WI序列好（t=2.934、3.677,P<0.05）;T1dixonc序列直肠DSC、肛管HD差异稍优于T2WI（t=6.806、2.130,P<0.05）;T2WI序列勾画骨组织（左/右股骨头）稳定性优于T1WI、T1dixonc,且差异均具有统计学意义（t=-6.580~6.542,P<0.05）。结论 基于MR图像的不同勾画者和同一勾画者不同时间勾画膀胱、直肠、股骨头稳定性较好,肛管次之。且T1WI序列OARs勾画稳定性较优于T1dixonc、T2WI。     

定位CT图像和CT/MR扩散加权成像融合图像对食管癌放疗GTV勾画的比较

目的 比较定位CT图像和CT与磁共振扩散加权成像(MR DWI)融合图像(CT/MR DWI)对食管癌放疗GTV勾画的差异.方法 收集经细胞学或组织病理学证实的20例行根治性放疗的食管鳞癌患者,由6名放疗科医师分别在Pinnacle工作站的定位CT图像上和CT/MR DWI融合图像上勾画食管癌原发灶GTV,不包括转移淋巴结.计算GTV体积的均数、标准差、变异系数(CV=标准差/均数),最大值与最小值的比(Ratio=最大值/最小值).比较两组间GTV体积的变异系数CV和最大值与最小值的比值(Ratio)的差异.结果 CT图像上勾画的GTV体积最大差值为55.71 cm³,CT/MR DWI融合图像上勾画的GTV体积最大差值为13.89 cm³(F=12.80,P<0.05),两组GTV体积的变异系数分别为0.30±0.08和0.11±0.04,两组GTV体积最大值与最小值的比值分别为2.38±0.62和1.34±0.13,差异有统计学意义(Z=-3.92、-3.92,P<0.05).结论 和定位CT图像相比,CT/MR DWI图像融合显示食管癌GTV较为直观,能够在一定程度上提高判断病变范围的一致性,从而减少不同医师间勾画靶区差异性.     

两种自动勾画头颈部危及器官方法的比较研究

目的 设计一种基于深度学习的自动勾画模型,用于勾画头颈部危及器官（OARs）,并与基于图谱方法的Smart segmentation勾画软件进行比较。方法 自动勾画模型由基于深度学习神经网络的分类模型和勾画模型组成。分类模型将CT图像从头脚方向分为6个分类,将每个OARs对应分类的CT图像输入勾画模型进行分割勾画。自动勾画模型使用150例病例训练模型,Smart segmentation使用相同的150例病例组成图谱库,两者同时对20例测试集进行勾画。使用相似度系数（DSC）和豪斯多夫距离（HD）评估2种方法勾画准确性,同时记录两种方法勾画花费时间。根据数据是否满足正态分布,分别使用配对t检验和Wilcoxon符号秩和检验。结果 自动勾画模型的DSC和HD结果如下：脑干为0.88和4.41 mm、左眼球为0.89和2.00 mm、右眼球为0.89和2.12 mm、左视神经为0.70和3.00 mm、右视神经为0.80和2.24 mm、左颞叶为0.81和7.98 mm、右颞叶为0.84和8.82 mm、下颌骨为0.89和5.57 mm、左腮腺为0.70和11.92 mm和右腮腺为0.77和11.27 mm。除腮腺外,自动勾画模型勾画结果均优于Smart segmentation,差异有统计学意义（t=3.115~7.915,Z=-1.352~-3.921,P<0.05）。同时,自动勾画模型速度比Smart segmentation提高了51.28%。结论 利用深度学习方法建立了自动勾画头颈部OARs的模型,得到较准确结果,勾画精度和速度均优于Smart segmentation软件。     

局部晚期鼻咽癌诱导化疗后原发病灶靶区勾画的探讨

目的 探讨局部晚期鼻咽癌诱导化疗后原发病灶(GTVnx)靶区勾画。方法 选择2012-2013年收治的52例局部晚期鼻咽癌患者,诱导化疗2~3周期后行CT定位、标记及图像采集;同期采取相同体位行鼻咽部MRI平扫及增强扫描,采集T₁W₁增强图像;分别在CT图像及MRI图像进行GTV_nx勾画;转移淋巴结、CTV1、CTV2及正常组织均在CT图像进行勾画;通过放疗计划系统进行MRI/CT图像GTV_nx靶区融合;两套靶区给予相同处方剂量及正常组织限量,物理师进行调强放疗计划设计。比较不同图像下诱导化疗后GTV_nx、各靶区照射体积及剂量、正常组织受量变化。结果 在局部晚期鼻咽癌诱导化疗后GTV_nx勾画中,MRI图像勾画靶体积大于CT图像[(43.14±28.40)、(40.09±27.04)cm³,t=3.791,P<0.001];MRI图像勾画靶体积与诱导化疗前原发病灶体积差值[(27.90±11.86)cm³]小于CT图像勾画体积差值[(30.64±11.86)cm³](t=3.948,P<0.001)。两套计划原发病灶靶区照射体积比较,融合靶区计划(41.71±26.86)cm³大于CT图像计划[(38.65±25.66)cm³](t=4.098,P<0.001),但靶区剂量及正常组织受量差异无统计学意义。结论 采用MRI图像进行局部晚期鼻咽癌诱导化疗后原发病灶靶区勾画、MRI/CT靶区图像融合进行放疗计划设计,增加原发灶靶区体积及照射体积,可能减少诱导化疗后放射治疗靶区勾画漏靶发生。     

磁共振模拟定位及磁共振诊断影像与定位CT融合的精准性比较

目的 与磁共振成像（MRI）诊断影像（MRIdiag）相比较,评估MRI模拟定位影像（MRIsim）与模拟CT融合的精准性,为MRIsim的进一步应用提供参考。方法 选择行MRIsim同时又有MRIdiag的患者24例,其中脑胶质瘤、鼻咽癌和前列腺癌各8例。将每位患者的MRIsim、MRIdiag影像分别与模拟CT融合,在3种图像上分别勾画危及器官（OAR）,在CT与MRIsim融合影像（F_CTMsim）、CT与MRIdiag融合影像（F_CTMdiag）上分别勾画靶区。评估基于MRIsim、MRIdiag与CT之间OAR的一致性指数（CI）、形似指数（DSC）、图像相似性指数（S）。基于F_CTMsim的靶区和CT的OAR设计IMRT计划,评估靶区和OAR的剂量学差异。结果 除了鼻咽癌患者的全脑和前列腺癌患者的盆骨外,其余OAR基于3种图像的体积值无明显差异（P>0.05）;MRIsim的所有OAR与CT比较的CI和DSC值均高于MRIdiag,其中50%的OAR差异有统计学意义（t=2.58~5.47,P<0.05）。MRIsim、MRIdiag与CT比较,S值分别为0.89、0.83（t=5.77,P<0.05）,相对于MRIdiag,MRIsim改善了S值10%（2%~56%）。OAR和靶区剂量学差异无统计学意义（P>0.05）。结论 相对于MRIdiag,将MRIsim引入放疗与模拟CT融合可以明显改善图像匹配的精准性。但两者基于刚性匹配肿瘤区域结合手动调整方法所勾画的靶区之间没有剂量学差异。     

扫描视野对乳腺癌调强放射治疗剂量学影响的研究

目的 研究不同扫描视野(FOV)CT图像对乳腺癌根治术后放射治疗中危及器官自动勾画及剂量计算精度的影响。方法 使用相同扫描条件在患者模拟定位CT等中心处及扩展扫描射野(eFOV)处建立50、60、70和80 cm FOV的电子密度转换曲线并比较其差异;扫描已知体积的标准模体,比较模体在不同FOV重建图像上自动勾画的差异。简单随机抽样选取2020年1月至2022年6月广东省第二人民医院乳腺癌患者30例,获取不同FOV模拟定位CT图像进行危及器官自动勾画,并与医师的勾画进行比较;基于FOV₅₀图像设计治疗计划,将计划移植到不同FOV重建图像进行剂量计算,比较剂量计算结果的差异。结果 以不同FOV重建CT图像建立的电子密度转换曲线基本一致。在等中心处,标准模体随FOV增大,模体勾画体积与实际体积差异增大,最大为6 cm³(4.8%);在自动勾画中,脊髓、气管、食管、甲状腺、健侧乳腺和皮肤的勾画精度随FOV增大而减小(t= -28.43~8.23,P＜0.05), 基于不同FOV图像的剂量计算比较中,锁骨上淋巴结区域靶区V₉₅、最大剂量和平均剂量,危及器官剂量学的差异无统计学意义(P>0.05); 计划靶区覆盖度随FOV增大而减小(最大差异为4.06%)。结论 乳腺癌根治术后放疗中危及器官自动勾画应选择FOV₅₀重建图像,电子密度转换曲线应依据eFOV区域电子密度模体影像建立,首选eFOV₈₀的重建图像进行剂量计算。     

扇形束CT引导的在线自适应放疗在 宫颈癌中的临床实践

目的 探讨人工智能(AI)辅助联合低剂量扇形束计算机断层扫描(FBCT)引导的在线适应性放疗(OART)治疗宫颈癌的可行性及安全性。方法 在联影uCT-ART平台予11名宫颈癌(10名术后辅助,1名根治性放疗)患者行高年资放疗医师主导触发的OART。分析AI辅助低剂量FBCT引导的OART治疗宫颈癌的可行性,包括评估自动分割轮廓质量、自动放射治疗计划、OART在线剂量学分析、OART流程时长;以及11名宫颈癌患者的放疗相关不良反应分析。结果 在297个分次治疗中,经高年资放疗医师判断,共启动81次OART,人均启动OART 7.4次。OART流程平均总时长为18.97 min OART调整平均时长为15.87 min。11例患者在定位CT上经AI辅助勾画工具自动分割感兴趣区域(ROI),得到ROI_auto经高年资放疗医师修改及审核后得到ROI_edit,其中临床靶区(CTV)的Dice相似系数为0.85 ± 0.04,不劣于前期模型0.89 ± 0.02(P＞0.05),95% 豪斯多夫距离为(5.64 ± 1.60)mm,优于前期模型构建的(6.28 ± 2.31)mm(t=-2.34,P＜0.05)。OART启动后轮廓勾画策略为优先采用CTV刚性拷贝/OAR自动分割+高年资放疗医师修改。OART计划的计划靶区(PTV)剂量分布更为紧凑,且剂量整体更接近处方剂量;OART计划中OAR剂量控制更贴近临床要求,OAR受照剂量显著低于影像引导放疗(IGRT)计划;适形指数CI与均匀性指数HI优于手动计划。在OART过程中,分次间靶区体积变化范围主要集中在±5%的范围内,而OAR体积变化范围波动较大,且无明显规律。消化系统、泌尿系统、造血系统发生急性不良反应均为PRO-CTCAE 1~2级,未见3级及以上的不良反应发生。结论 本研究描述了基于uCT-ART的OART系统在宫颈癌放疗的成功实施,证实了OART临床应用的可行性与安全性。     

基于CT/MRI/PET-CT进行鼻咽癌大体肿瘤区及淋巴结勾画的对照研究

目的 比较基于计算机体层成像（CT）、磁共振成像（MRI）、¹⁸氟-脱氧葡萄糖正电子发射计算机断层显像（¹⁸F-FDG PET-CT）3种图像勾画鼻咽癌（NPC）大体肿瘤靶区（GTV）及淋巴结的差异,并以MRI为参考研究基于PET自动勾画NPC GTV的最佳标准摄取值（SUV）。方法 选取拟行放疗的鼻咽癌患者53例,依次获得CT模拟定位、MRI模拟定位及PET图像,在3种图像上分别勾画GTV与阳性淋巴结,分别命名为GTV_MRI、GTV_CT、GTV_PET2.5（SUV=2.5）、Lymph_MRI、Lymph_CT、Lymph_PET2.5,比较不同图像确定的GTV及淋巴结差异。将MRI与PET-CT配准后,得到相交区域GTV_∩2.5,依据SUV=4.0、4.5、5.0、5.6为阈值分别在PET上自动勾画GTV_PET4.0、GTV_PET4.5、GTV_PET5.0、GTV_PET5.6,比较不同GTV之间体积、相似系数（DSC）的差异。结果 在3种图像中,GTV_MRI分别较GTV_CT增加了1.73%（P>0.05）,较GTV_PET2.5减小了21.34%（t=-3.52,P<0.05）;Lymph_PET2.5的体积分别为Lymph_MRI、Lymph_CT的1.61、1.87倍（t=-4.12、-5.18,P<0.05）。PET上表现为高摄取的淋巴结体积平均为无或低摄取的4.07倍（t=5.50,P<0.05）。GTV_PET4.0与GTV_MRI的DSC为0.78±0.27,低于GTV_PET2.5与GTV_MRI的DSC （0.84±0.18）,但GTV_PET4.0与GTV_∩2.5体积基本相当（P>0.05）。结论 基于CT、MRI、¹⁸F-FDG PET-CT勾画鼻咽癌GTV及淋巴结时,MRI可更好显示肿瘤边界;而基于¹⁸F-FDG PET-CT自动勾画GTV时,推荐SUV=4.0作为阈值。     

Clinical target volume segmentation based on gross tumor volume using deep learning for head and neck cancer treatment

Accurate clinical target volume (CTV) delineation is important for head and neck intensity-modulated radiation therapy. However, delineation is time-consuming and susceptible to interobserver variability (IOV). Based on a manual contouring process commonly used in clinical practice, we developed a deep learning (DL)-based method to delineate a low-risk CTV with computed tomography (CT) and gross tumor volume (GTV) input and compared it with a CT-only input. A total of 310 patients with oropharynx cancer were randomly divided into the training set (250) and test set (60). The low-risk CTV and primary GTV contours were used to generate label data for the input and ground truth. A 3D U-Net with a two-channel input of CT and GTV (U-Net_GTV) was proposed and its performance was compared with a U-Net with only CT input (U-Net_CT). The Dice similarity coefficient (DSC) and average Hausdorff distance (AHD) were evaluated. The time required to predict the CTV was 0.86 s per patient. U-Net_GTV showed a significantly higher mean DSC value than U-Net_CT (0.80 ± 0.03 and 0.76 ± 0.05) and a significantly lower mean AHD value (3.0 ± 0.5 mm vs 3.5 ± 0.7 mm). Compared to the existing DL method with only CT input, the proposed GTV-based segmentation using DL showed a more precise low-risk CTV segmentation for head and neck cancer. Our findings suggest that the proposed method could reduce the contouring time of a low-risk CTV, allowing the standardization of target delineations for head and neck cancer.     

Automatic segmentation of prostate magnetic resonance imaging using generative adversarial networks

BackgroundAutomatic and detailed segmentation of the prostate using magnetic resonance imaging (MRI) plays an essential role in prostate imaging diagnosis. Traditionally, prostate gland was manually delineated by the clinician in a time-consuming process that requires professional experience of the observer. Thus, we proposed an automatic prostate segmentation method, called SegDGAN, which is based on a classic generative adversarial network model.Material and methodsThe proposed method comprises a fully convolutional generation network of densely con- nected blocks and a critic network with multi-scale feature extraction. In these computations, the objective function is optimized using mean absolute error and the Dice coefficient, leading to improved accuracy of segmentation results and correspondence with the ground truth. The common and similar medical image segmentation networks U-Net, FCN, and SegAN were selected for qualitative and quantitative comparisons with SegDGAN using a 220-patient dataset and the public datasets. The commonly used segmentation evaluation metrics DSC, VOE, ASD, and HD were used to compare the accuracy of segmentation between these methods.ResultsSegDGAN achieved the highest DSC value of 91.66%, the lowest VOE value of 15.28%, the lowest ASD values of 0.51 mm and the lowest HD value of 11.58 mm with the clinical dataset. In addition, the highest DSC value, and the lowest VOE, ASD and HD values obtained with the public data set PROMISE12 were 86.24%, 23.60%, 1.02 mm and 7.57 mm, respectively.ConclusionsOur experimental results show that the SegDGAN model have the potential to improve the accuracy of MRI-based prostate gland segmentation.Code has been made available at: https://github.com/w3user/SegDGAN     

CT图像重建视野大小对放射治疗计划剂量计算及体积评估的影响

目的 探讨CT图像重建视野(FOV)大小对放射治疗计划剂量计算及体积评估可能存在的影响。方法 对16例鼻咽癌患者的CT原始扫描数据分别行45 cm常规FOV和65 cm扩展视野(EFOV)重建并传输至放射治疗计划系统,所有病例均在常规FOV重建的CT图像上勾画肿瘤体积(GTV)、临床靶区(CTV)及脑干、晶体、腮腺、脊髓等危及器官,并制定7野等角动态调强放射治疗计划(GTV处方剂量70 Gy)。两种重建方法图像按照医学数字影像通信3.0标准(DICOM 3.0)坐标方式融合后,拷贝常规FOV图像上的靶区及危及器官至EFOV图像,并将治疗计划移植至EFOV图像,治疗计划中心为两种重建方法图像的同一DICOM坐标,利用剂量体积直方图(DVH)工具计算两种重建方法图像上GTV、CTV和脑干、晶体、腮腺、脊髓的体积、最大剂量(D_max)、平均剂量(D_mean)及最小剂量(D_min)。将入组病例的每个治疗计划7野分别导入常规45 cm FOV和65 cm EFOV重建的二维通量图验证设备Mapchek 1175的模体,距离通过协议(DTA)分析5 cm深度平面绝对剂量的计算和实测结果通过率。结果 两种重建方法图像上的靶区和危及器官的体积差异具有统计学意义,所有入组病例靶区和危及器官在常规FOV图像上的体积均大于EFOV图像上的体积。较小体积的晶体最大剂量D_max常规FOV与EFOV图像之间差异有统计学意义(t =-3.14, P<0.007),其余靶区及危及器官的最大剂量D_max差异无统计学意义。CTV和GTV平均剂量D_mean在EFOV图像上大于FOV图像,差异有统计学意义(t=-6.45、-5.65, P< 0.001),危及器官的平均剂量D_mean和靶区及危及器官最小剂量D_min差异均无统计学意义。两种重建方法图像上治疗计划的7野通过率之间差异无统计学意义。结论 在放射治疗CT模拟定位过程中图像重建FOV的大小对于靶区及部分危及器官的体积及剂量计算结果和治疗计划的评价存在影响;观察和验证二维通量图通过率,两者之间的差异并不显著。     

A Prostate Fossa Contouring Instructional Module: Implementation and Evaluation

Purpose/ObjectiveRadiation oncology trainees frequently learn to contour through clinical experience and lectures. A hands-on contouring module was developed to teach delineation of the postoperative prostate clinical target volume (CTV) and improve contouring accuracy.MethodsMedical students independently contoured a prostate fossa CTV before and after receiving educational materials and live instruction detailing the RTOG approach to contouring this CTV. Metrics for volume overlap and surface distance (Dice similarity coefficient, Hausdorff distance (HD), and mean distance) determined discordance between student and consensus contours. An evaluation assessed perception of session efficacy (1 = “not at all” to 5 = “extremely”; reported as median[interquartile range]). Non-parametric statistical tests were used.ResultsTwenty-four students at two institutions completed the module, and 21 completed the evaluation (88% response). The content was rated as “quite” important (4[3.5-5]).The module improved comfort contouring a prostate fossa (pre 1[1-2] vs. post 4[3-4], p<.01), ability to find references (pre 2[1-3] vs. post 4[3.5-4], p<0.01), knowledge of CT prostate/pelvis anatomy (pre 2[1.5-3] vs. post 3[3-4], p<.01), and ability to use contouring software tools (pre 2[2-3.5] vs. post 3[3-4], p=.01). After intervention, mean DSC increased (0.29 to 0.68, p<0.01) and HD and mean distance both decreased, respectively (42.8 to 30.0, p<.01; 11.5 to 1.9, p<.01).ConclusionsA hands-on module to teach CTV delineation to medical students was developed and implemented. Student and expert contours exhibited near “excellent agreement” (as defined in the literature) after intervention. Additional modules to teach target delineation to all educational levels can be developed using this model.     

Targeting the GTV in medically inoperable endometrial cancer using brachytherapy

PURPOSEWe aimed to determine the relationship between gross tumor volume (GTV) dose and tumor control in women with medically inoperable endometrial cancer, and to demonstrate the feasibility of targeting a GTV-focused volume using imaged-guided brachytherapy.METHODS AND MATERIALSAn endometrial cancer database was used to identify patients. Treatment plans were reviewed to determine doses to GTV, clinical target volume (CTV), and OARs. Uterine recurrence-free survival was evaluated as a function of CTV and GTV doses. Brachytherapy was replanned with a goal of GTV D98 EQD₂ ≥ 80 Gy, without regard for coverage of the uninvolved uterus and while respecting OAR dose constraints.RESULTSFifty-four patients were identified. In the delivered plans, GTV D90 EQD₂ ≥ 80 Gy was achieved in 36 (81.8%) patients. Uterine recurrence-free survival was 100% in patients with GTV D90 EQD₂ ≥ 80 Gy and 66.7% in patients with EQD₂ < 80 Gy (p = 0.001). On GTV-only replans, GTV D98 EQD₂ ≥ 80 Gy was achieved in 39 (88.6%) patients. Mean D_2cc was lower for bladder (47.1 Gy vs. 73.0 Gy, p < 0.001), and sigmoid (47.0 Gy vs. 58.0 Gy, p = 0.007) on GTV-only replans compared to delivered plans. Bladder D_2cc was ≥ 80 Gy in 11 (25.0%) delivered plans and four (9.1%) GTV-only replans (p = 0.043). Sigmoid D_2cc was ≥ 65 Gy in 20 (45.4%) delivered plans and 10 (22.7%) GTV-only replans (p = 0.021).CONCLUSIONSOAR dose constraints should be prioritized over CTV coverage if GTV coverage is sufficient. Prospective evaluation of image-guided brachytherapy to a reduced, GTV-focused volume is warranted.     

Automated Lung Segmentation on Chest Computed Tomography Images with Extensive Lung Parenchymal Abnormalities Using a Deep Neural Network

ObjectiveWe aimed to develop a deep neural network for segmenting lung parenchyma with extensive pathological conditions on non-contrast chest computed tomography (CT) images.Materials and MethodsThin-section non-contrast chest CT images from 203 patients (115 males, 88 females; age range, 31–89 years) between January 2017 and May 2017 were included in the study, of which 150 cases had extensive lung parenchymal disease involving more than 40% of the parenchymal area. Parenchymal diseases included interstitial lung disease (ILD), emphysema, nontuberculous mycobacterial lung disease, tuberculous destroyed lung, pneumonia, lung cancer, and other diseases. Five experienced radiologists manually drew the margin of the lungs, slice by slice, on CT images. The dataset used to develop the network consisted of 157 cases for training, 20 cases for development, and 26 cases for internal validation. Two-dimensional (2D) U-Net and three-dimensional (3D) U-Net models were used for the task. The network was trained to segment the lung parenchyma as a whole and segment the right and left lung separately. The University Hospitals of Geneva ILD dataset, which contained high-resolution CT images of ILD, was used for external validation.ResultsThe Dice similarity coefficients for internal validation were 99.6 ± 0.3% (2D U-Net whole lung model), 99.5 ± 0.3% (2D U-Net separate lung model), 99.4 ± 0.5% (3D U-Net whole lung model), and 99.4 ± 0.5% (3D U-Net separate lung model). The Dice similarity coefficients for the external validation dataset were 98.4 ± 1.0% (2D U-Net whole lung model) and 98.4 ± 1.0% (2D U-Net separate lung model). In 31 cases, where the extent of ILD was larger than 75% of the lung parenchymal area, the Dice similarity coefficients were 97.9 ± 1.3% (2D U-Net whole lung model) and 98.0 ± 1.2% (2D U-Net separate lung model).ConclusionThe deep neural network achieved excellent performance in automatically delineating the boundaries of lung parenchyma with extensive pathological conditions on non-contrast chest CT images.