目的建立基于CT影像组学的机器学习模型应用于术前预测肺磨玻璃结节(ground-glass nodules,GGNs)浸润程度。方法回顾性收集2021年3月—7月于上海市肺科医院胸外科及2019年1月—2022年5月于兰州大学第一医院胸外科就诊的结节实性直径<...目的建立基于CT影像组学的机器学习模型应用于术前预测肺磨玻璃结节(ground-glass nodules,GGNs)浸润程度。方法回顾性收集2021年3月—7月于上海市肺科医院胸外科及2019年1月—2022年5月于兰州大学第一医院胸外科就诊的结节实性直径<3 cm肺GGNs患者的临床资料。根据患者术后病理结果分为非浸润性病变和浸润性病变,按7∶3比例随机分为训练集和测试集。从每例患者的CT图像中提取影像组学特征(1317个),采用最小冗余最大相关性(max-relevance and min-redundancy,mRMR)筛选和分类类别最相关的前100个特征,最小绝对收缩与选择算子(least absolute shrinkage and selection operator,LASSO)进行影像组学特征选择,支持向量机(support vector machine,SVM)分类器建立预测模型,计算曲线下面积(area under the curve,AUC)、敏感度、特异度、准确率、阴性预测值、阳性预测值指标评估该模型的性能,绘制预测模型的校准曲线及决策曲线来评估模型的准确度和临床获益,并在测试集及不同结节直径亚组中进行性能分析;将该模型的预测性能与Mayo以及Brock模型进行对比。2名初级胸外科医师评估GGNs的浸润程度以调查该模型的临床效用。结果共纳入400例患者,其中女267例、男133例,平均年龄(52.4±12.7)岁。训练集280例,测试集120例。从训练集数据最终筛选出8个影像组学特征建立SVM模型,该模型在训练集及测试集中AUC值、敏感度、特异度分别为0.91、0.89、0.75,0.86、0.92、0.60。并且在训练集0~10 mm、10~20 mm以及测试集0~10 mm、10~20 mm亚组中均表现出较好的预测性能,AUC值分别为0.82、0.88,0.84、0.72。SVM模型明显优于Mayo模型(0.73)和Brock模型(0.73)。在该模型辅助下,医师A和B区分浸润性腺癌的AUC值、敏感度、特异度、准确率均明显提高。结论该基于影像组学的SVM模型有助于区分非浸润性病变和浸润性病变,针对不同大小的GGNs也有较稳定的预测性能,其相比Mayo及Brock模型预测性能更佳。可协助临床医师更为准确判断GGNs浸润程度,并制定更合适的诊疗决策,实现精准化治疗。展开更多
Autotoxicity is one of the major factors that impede continuous cropping.It is defined as the toxic influence of chemicals released from one plant species on the germination and growth of individuals of the same speci...Autotoxicity is one of the major factors that impede continuous cropping.It is defined as the toxic influence of chemicals released from one plant species on the germination and growth of individuals of the same species.Here, in order to exam the autotoxicity of tobacco root exudates, root exudates were collected from tobacco plants grown both in cultural solution and on natural soil.Using ultraperformance liquid chromatography coupled with triple-quadrupole tandem mass spectrometry, main autotoxic chemical substances in the root exudates were identified.The autotoxic effects of suspected autotoxins on seed germination(including germination rate,germination potential, germination index, and vigor index) and seedling growth were analyzed.Dibutyl phthalate(or diisobutyl phthalate), dioctyl phthalate, and diisooctyl phthalate were identified in tobacco root exudates.It was observed that high concentrations(greater than 0.5 mmol L^(-1)) of each identified phthalate ester caused significant(P < 0.05) inhibition of tobacco seed germination and seedling growth.It can be concluded that phthalate esters such as dibutyl phthalate, diisobutyl phthalate, and diisooctyl phthalate in tobacco root exudates may play an important role in tobacco autotoxicity.展开更多
文摘目的建立基于CT影像组学的机器学习模型应用于术前预测肺磨玻璃结节(ground-glass nodules,GGNs)浸润程度。方法回顾性收集2021年3月—7月于上海市肺科医院胸外科及2019年1月—2022年5月于兰州大学第一医院胸外科就诊的结节实性直径<3 cm肺GGNs患者的临床资料。根据患者术后病理结果分为非浸润性病变和浸润性病变,按7∶3比例随机分为训练集和测试集。从每例患者的CT图像中提取影像组学特征(1317个),采用最小冗余最大相关性(max-relevance and min-redundancy,mRMR)筛选和分类类别最相关的前100个特征,最小绝对收缩与选择算子(least absolute shrinkage and selection operator,LASSO)进行影像组学特征选择,支持向量机(support vector machine,SVM)分类器建立预测模型,计算曲线下面积(area under the curve,AUC)、敏感度、特异度、准确率、阴性预测值、阳性预测值指标评估该模型的性能,绘制预测模型的校准曲线及决策曲线来评估模型的准确度和临床获益,并在测试集及不同结节直径亚组中进行性能分析;将该模型的预测性能与Mayo以及Brock模型进行对比。2名初级胸外科医师评估GGNs的浸润程度以调查该模型的临床效用。结果共纳入400例患者,其中女267例、男133例,平均年龄(52.4±12.7)岁。训练集280例,测试集120例。从训练集数据最终筛选出8个影像组学特征建立SVM模型,该模型在训练集及测试集中AUC值、敏感度、特异度分别为0.91、0.89、0.75,0.86、0.92、0.60。并且在训练集0~10 mm、10~20 mm以及测试集0~10 mm、10~20 mm亚组中均表现出较好的预测性能,AUC值分别为0.82、0.88,0.84、0.72。SVM模型明显优于Mayo模型(0.73)和Brock模型(0.73)。在该模型辅助下,医师A和B区分浸润性腺癌的AUC值、敏感度、特异度、准确率均明显提高。结论该基于影像组学的SVM模型有助于区分非浸润性病变和浸润性病变,针对不同大小的GGNs也有较稳定的预测性能,其相比Mayo及Brock模型预测性能更佳。可协助临床医师更为准确判断GGNs浸润程度,并制定更合适的诊疗决策,实现精准化治疗。
基金supported by the Key Laboratory Project of CNTC(No.110201603010)the Scientific and Technological Project of Zhengzhou Tobacco Research Institute of CNTC(No.112011CZ0580)
文摘Autotoxicity is one of the major factors that impede continuous cropping.It is defined as the toxic influence of chemicals released from one plant species on the germination and growth of individuals of the same species.Here, in order to exam the autotoxicity of tobacco root exudates, root exudates were collected from tobacco plants grown both in cultural solution and on natural soil.Using ultraperformance liquid chromatography coupled with triple-quadrupole tandem mass spectrometry, main autotoxic chemical substances in the root exudates were identified.The autotoxic effects of suspected autotoxins on seed germination(including germination rate,germination potential, germination index, and vigor index) and seedling growth were analyzed.Dibutyl phthalate(or diisobutyl phthalate), dioctyl phthalate, and diisooctyl phthalate were identified in tobacco root exudates.It was observed that high concentrations(greater than 0.5 mmol L^(-1)) of each identified phthalate ester caused significant(P < 0.05) inhibition of tobacco seed germination and seedling growth.It can be concluded that phthalate esters such as dibutyl phthalate, diisobutyl phthalate, and diisooctyl phthalate in tobacco root exudates may play an important role in tobacco autotoxicity.
