目的:探讨CT征象联合影像组学在鉴别孤立性肺结核(TB)厚壁空洞和肺真菌(PM)厚壁空洞中的应用价值。方法:回顾性分析我院2010年1月—2024年5月确诊的表现为孤立性厚壁空洞66例TB患者和66例PM患者。首先依次采用单因素和多因素分析筛选出...目的:探讨CT征象联合影像组学在鉴别孤立性肺结核(TB)厚壁空洞和肺真菌(PM)厚壁空洞中的应用价值。方法:回顾性分析我院2010年1月—2024年5月确诊的表现为孤立性厚壁空洞66例TB患者和66例PM患者。首先依次采用单因素和多因素分析筛选出临床独立预测因子。然后利用汇医慧影(Radcloud3.0)沿着病灶肺窗平扫最大截面涂抹感兴趣区(ROI),从每个病灶中提取7类共1409个影像组学特征。按照8∶2比例分为训练集和测试集,通过方差阈值法(Variance Threshold)、最佳K折法(Select K Best)及最小绝对收缩选择算子(LASSO)进行特征降维和选择,筛选出最优的特征,采用逻辑回归(LR)分类器构建影像组学模型,最后进一步联合具有独立预测作用的临床因子进行联合模型构建,并使用列线图对联合模型进行了可视化。采用受试者工作特征曲线的曲线下面积(AUC)、准确度、灵敏度及特异度等来评价模型的诊断效能。结果:单因素和多因素分析结果显示病灶分布、洞内球征、卫星灶及胸膜凹陷征是鉴别TB/PM的独立影响因素。经过一系列特征降维选择,最终分别筛选出16个最优影像组学特征用于模型的构建。单独影像组学模型训练集的AUC为0.842,准确度、灵敏度及特异度分别是0.780、0.810和0.770,测试集的AUC为0.760,准确度、灵敏度及特异度分别是0.690、0.640和0.710;进一步联合病灶分布、洞内球征、卫星灶及胸膜凹陷征构建的联合模型训练集的AUC为0.944,准确度、灵敏度及特异度分别是0.870、0.870和0.870,测试集的AUC为0.781,准确度、灵敏度及特异度分别是0.690、0.640和0.710。联合模型的AUC值均高于单独影像组学模型的AUC。结论:影像组学对TB与PM孤立性厚壁空洞的鉴别具有一定的价值,CT征象联合影像组学特征的联合模型诊断效能优于单独影像组学模型。展开更多
Modern computer techniques have been in use for several years to generate three-dimensional visualizations of human anatomy. Very good 3-D computer models of the human body are now available and used routinely in anat...Modern computer techniques have been in use for several years to generate three-dimensional visualizations of human anatomy. Very good 3-D computer models of the human body are now available and used routinely in anatomy instruction. These techniques are subsumed under the heading “virtual anatomy” to distinguish them from the conventional study of anatomy entailing cadavers and anatomy textbooks. Moreover, other imaging procedures (X-ray, angiography, CT and MR) are also used in virtual anatomy instruction. A recently introduced three-dimensional post-processing technique named Cinematic Rendering now makes it possible to use the output of routine CT and MR examinations as the basis for highly photo-realistic 3-D depictions of human anatomy. We have installed Cinematic Rendering (enabled for stereoscopy) in a high-definition 8K 3-D projection space that accommodates an audience of 150. The space’s projection surface measures 16 × 9 meters;images can be projected on both the front wall and the floor. A game controller can be used to operate Cinematic Rendering software so that it can generate interactive real-time depictions of human anatomy on the basis of CT and MR data sets. This prototype installation was implemented without technical problems;in day-to-day, real-world use over a period of 22 months, there were no impairments of service due to software crashes or other technical problems. We are already employing this installation routinely for educational offerings open to the public, courses for students in the health professions, and (continuing) professional education units for medical interns, residents and specialists—in, so to speak, the dissecting theater of the future.展开更多
目的利用肺结节CT征象构建二元Logistic回归模型,预测浸润性腺癌的病理分级。方法回顾性分析2021年1月~2023年2月河南理工大学第一附属医院及新乡医学院第一附属医院术后病理为浸润性腺癌的磨玻璃结节303例的临床资料、病理分型及影像...目的利用肺结节CT征象构建二元Logistic回归模型,预测浸润性腺癌的病理分级。方法回顾性分析2021年1月~2023年2月河南理工大学第一附属医院及新乡医学院第一附属医院术后病理为浸润性腺癌的磨玻璃结节303例的临床资料、病理分型及影像学资料。根据病理结果将303个病灶分成两组即低级别组(以贴壁、腺泡或乳头状型为主型腺癌,没有或低于20%的高级别模式)262例和高级别组(任何大于20%高级别成分的腺癌)41例。两组间定量参数比较采用Mann-Whitney U检验,定性参数的比较采用χ^(2)检验。采用Logistic回归模型筛选独立预测因子,使用曲线下面积(area under the curve,AUC)值、校准曲线和决策分析曲线评价模型的区分度、校准度和临床使用价值。结果单因素分析结果显示,性别、空气支气管征、空泡征、血管集束征、胸膜凹陷征、长径、短径及CT强化净增值,差异有统计学意义(P<0.05),位置、毛刺征、实性成分比差异无统计学意义(P>0.05);Logistic回归分析显示,长径、CT强化净增值、血管集束征、胸膜凹陷征、空泡征是构建预测浸润性腺癌病理分级模型的独立预测因子。受试者工作特征(receiver operating characteristic,ROC)曲线分析结果显示,Logistic回归模型AUC值是0.846,敏感度为81.25%,特异性为86.52%。结论CT征象构建的模型预测浸润性腺癌病理分级有较好的预测能力及稳定性。展开更多
文摘目的:探讨CT征象联合影像组学在鉴别孤立性肺结核(TB)厚壁空洞和肺真菌(PM)厚壁空洞中的应用价值。方法:回顾性分析我院2010年1月—2024年5月确诊的表现为孤立性厚壁空洞66例TB患者和66例PM患者。首先依次采用单因素和多因素分析筛选出临床独立预测因子。然后利用汇医慧影(Radcloud3.0)沿着病灶肺窗平扫最大截面涂抹感兴趣区(ROI),从每个病灶中提取7类共1409个影像组学特征。按照8∶2比例分为训练集和测试集,通过方差阈值法(Variance Threshold)、最佳K折法(Select K Best)及最小绝对收缩选择算子(LASSO)进行特征降维和选择,筛选出最优的特征,采用逻辑回归(LR)分类器构建影像组学模型,最后进一步联合具有独立预测作用的临床因子进行联合模型构建,并使用列线图对联合模型进行了可视化。采用受试者工作特征曲线的曲线下面积(AUC)、准确度、灵敏度及特异度等来评价模型的诊断效能。结果:单因素和多因素分析结果显示病灶分布、洞内球征、卫星灶及胸膜凹陷征是鉴别TB/PM的独立影响因素。经过一系列特征降维选择,最终分别筛选出16个最优影像组学特征用于模型的构建。单独影像组学模型训练集的AUC为0.842,准确度、灵敏度及特异度分别是0.780、0.810和0.770,测试集的AUC为0.760,准确度、灵敏度及特异度分别是0.690、0.640和0.710;进一步联合病灶分布、洞内球征、卫星灶及胸膜凹陷征构建的联合模型训练集的AUC为0.944,准确度、灵敏度及特异度分别是0.870、0.870和0.870,测试集的AUC为0.781,准确度、灵敏度及特异度分别是0.690、0.640和0.710。联合模型的AUC值均高于单独影像组学模型的AUC。结论:影像组学对TB与PM孤立性厚壁空洞的鉴别具有一定的价值,CT征象联合影像组学特征的联合模型诊断效能优于单独影像组学模型。
文摘Modern computer techniques have been in use for several years to generate three-dimensional visualizations of human anatomy. Very good 3-D computer models of the human body are now available and used routinely in anatomy instruction. These techniques are subsumed under the heading “virtual anatomy” to distinguish them from the conventional study of anatomy entailing cadavers and anatomy textbooks. Moreover, other imaging procedures (X-ray, angiography, CT and MR) are also used in virtual anatomy instruction. A recently introduced three-dimensional post-processing technique named Cinematic Rendering now makes it possible to use the output of routine CT and MR examinations as the basis for highly photo-realistic 3-D depictions of human anatomy. We have installed Cinematic Rendering (enabled for stereoscopy) in a high-definition 8K 3-D projection space that accommodates an audience of 150. The space’s projection surface measures 16 × 9 meters;images can be projected on both the front wall and the floor. A game controller can be used to operate Cinematic Rendering software so that it can generate interactive real-time depictions of human anatomy on the basis of CT and MR data sets. This prototype installation was implemented without technical problems;in day-to-day, real-world use over a period of 22 months, there were no impairments of service due to software crashes or other technical problems. We are already employing this installation routinely for educational offerings open to the public, courses for students in the health professions, and (continuing) professional education units for medical interns, residents and specialists—in, so to speak, the dissecting theater of the future.
文摘目的利用肺结节CT征象构建二元Logistic回归模型,预测浸润性腺癌的病理分级。方法回顾性分析2021年1月~2023年2月河南理工大学第一附属医院及新乡医学院第一附属医院术后病理为浸润性腺癌的磨玻璃结节303例的临床资料、病理分型及影像学资料。根据病理结果将303个病灶分成两组即低级别组(以贴壁、腺泡或乳头状型为主型腺癌,没有或低于20%的高级别模式)262例和高级别组(任何大于20%高级别成分的腺癌)41例。两组间定量参数比较采用Mann-Whitney U检验,定性参数的比较采用χ^(2)检验。采用Logistic回归模型筛选独立预测因子,使用曲线下面积(area under the curve,AUC)值、校准曲线和决策分析曲线评价模型的区分度、校准度和临床使用价值。结果单因素分析结果显示,性别、空气支气管征、空泡征、血管集束征、胸膜凹陷征、长径、短径及CT强化净增值,差异有统计学意义(P<0.05),位置、毛刺征、实性成分比差异无统计学意义(P>0.05);Logistic回归分析显示,长径、CT强化净增值、血管集束征、胸膜凹陷征、空泡征是构建预测浸润性腺癌病理分级模型的独立预测因子。受试者工作特征(receiver operating characteristic,ROC)曲线分析结果显示,Logistic回归模型AUC值是0.846,敏感度为81.25%,特异性为86.52%。结论CT征象构建的模型预测浸润性腺癌病理分级有较好的预测能力及稳定性。
