目的:结合机器学习与影像组学特征构建预测急性缺血性脑卒中(acute inschemic strohe,AIS)机械取栓治疗后预后的模型并进行验证.方法:回顾性分析在南京市第一医院就诊的AIS患者,按随机数字表法分为训练集(n=105)和测试集(n=50),另收集在南京医科大学附属常州市第二人民医院就诊的AIS患者(n=45)作为外部验证.采用A.K.软件提取弥散加权成像(diffusion weighted imaging,DWI)和灌注加权成像(perfusion weighted imaging,PWI)病变区的影像特征,应用最低绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)回归模型筛选最佳影像组学特征,基于所选特征通过支持向量机(support vector machine,SVM)分类器建立预测急性脑卒中预后预测模型,使用受试者操作特征(receiver operating character-istic,ROC)曲线评价模型的预测效能,并应用验证集对模型进行外部验证.结果:每例患者DWI和PWI图像各提取1 316个影像组学特征,降维后筛选出40个与卒中预后高度相关的特征.ROC曲线分析显示联合DWI+PWI的模型预测训练集和测试集的曲线下面积(area under curve,AUC)(训练集:0.981;测试集:0.891)均高于单序列模型(DWI或PWI),其准确度分别达0.943、0.900.外部验证结果显示基于DWI+PWI的模型同样优于单序列(DWI或PWI)的预测模型,灵敏度和特异度分别为0.864、0.783,其准确度可达0.822.结论:结合机器学习与影像组学特征构建的模型可预测AIS机械取栓治疗预后,并具有较好的泛化能力.

A new objective fabric pilling detection method is proposed for monochrome fabric without pattern. Firstly, through wavelet multi-resolution decomposition and reconstruction, remove the high-frequency noise such as fluff which is significantly different from the size of pills and low-frequency noise such as uneven illumination. After wavelet multi-resolution decomposition of fabric image, through analyzing the energy change law between each wavelet decomposition layer, and then the fabric is judged whether it is rough or fine texture. Then, according to the different color characteristics of fine texture and rough texture fabrics, the even symmetry two-dimensional Gabor filter with smoothing effect and the odd symmetry two-dimensional Gabor filter with edge extraction ability in different directions are selected for texture removal. Finally, fabric pilling parameters are extracted to realize fabric pilling grade evaluation. The main contribution of this method is: on the premise of removing the noise such as uneven illumination and fluff, on the basis of considering the different color change laws of fine texture and rough texture, using the directionality of fabric texture and the randomness of pills direction to effectively remove the texture noise, so as to improve the accuracy of fabric pilling grade evaluation.

目的 探讨卒中机械取栓治疗前后扩散加权成像(DWI)影像组学的机器学习预测预后的效果.资料与方法 回顾性分析2017年1月—2020年9月在南京市第一医院接受机械取栓治疗的卒中患者,其中训练集157例,测试集68例.采用A.K.软件分别提取治疗前后DWI梗死区影像组学特征,并应用最低绝对收缩和选择算子回归模型筛选最佳影像组学特征,基于所选特征通过支持向量机分类器建立卒中机械取栓后预后的预测模型,利用受试者工作特征曲线评价模型的预测效能.收集常州市第二人民医院卒中患者(验证集75例)对模型进行外部验证.结果 每例患者治疗前后DWI图像共提取792个影像组学特征,降维后筛选出20个与预后高度相关的特征.受试者工作特征曲线分析显示支持向量机分离器建立的模型预测训练集患者预后的曲线下面积(AUC)为0.984,准确度达0.974;预测测试集患者预后的AUC为0.960,准确度达0.928;预测验证集患者预后的AUC为0.901,准确度达0.898;并具有较高的外部验证一致性(P>0.05).结论 基于治疗前后DWI的影像组学特征构建的模型对卒中机械取栓后预后预测具有较高的效能和较好的泛化能力.

目的:评价基于CT双期增强图像的不同深度迁移学习(DTL)模型对甲状腺良恶性结节的分类效能.方法:采用相同程序架构和相同数据集对3种DTL模型(VGG19、ResNet50和DenseNet201)的分类诊断效能进行测试和评估.以不同模型在训练集和测试集中的最高预测符合率和在验证集中的符合率、召回率、F1评分和受试者工作特性曲线(ROC)下面积作为评估模型效能的指标.结果:DenseNet201模型获得了最好的训练和测试结果,在训练集和测试集中的最高预测符合率分别为1.00和0.98;VGG19模型用时最长,其在训练集和测试集中的预测符合率分别为0.99和0.98,较DenseNet201略差;ResNet50模型用时最短,但测试结果最差,在训练集和测试集中的最高符合率分别为0.93和0.92.VGG19、ResNet50和DenseNet201模型在验证集中的平均符合率为0.96、0.92和0.98),召回率分别为0.96、0.91和0.98,F1评分分别为0.96、0.91和0.98.DenseNet201模型的ROC曲线下面积为0.98,高于VGG19模型(0.95)和ResNet50模型(0.91).结论:基于DenseNet201的DTL模型对甲状腺CT良恶性结节具有较高的分类效能,有助于提高影像诊断准确性.

Objective To investigate the altered spontaneous cerebral activity in patients with type 2 diabetic retinopathy (T2DR). Methods Twenty-one patients with T2DR and sixteen healthy control subject underwent rs-fMRI scans,and the data were analyzed statistically using regional homogeneity(ReHo)method to observe the change of ReHo value.Results Compared to the control group,the T2DR group showed significantly increased ReHo value in the right occipital gyrus,occipital gyrus,inferior occipital gyrus and lingual gyrus regions (t=5.30,P＜0.05,voxel＞30,AlphaSim corrected),and significantly decreased ReHo value in the left posterior cingulate,margin lobe,right inferior parietal lobule,superior temporal gyrus and hippocampus (t=-4.01,-4.86,P＜0.05,voxel＞30, AlphaSim corrected).Conclusion The patients with T2DR showed significantly increased ReHo values in the brain visual cortex and visual pathway that were associated with the injury of brain function regions.It is of important value to evaluate brain dysfunction in patients with T2DR using ReHo method of rs-fMRI.

Diabetes is often associated with impairments in brain functioning. However, the injury of specific functioning areas of the brain is not clear. To address this problem, the present study was designed to investigate possible brain functioning change in specific brain areas, particularly in areas associated with vision function, in patients with proliferative and nonproliferative diabetic retinopathy (PDR and NPDR) using the diffusion-weighted imaging technology. Conventional MRI was performed in 45 diabetic patients, 30 of whom had diabetic retinopathy (DR) involvement (half PDR, and half NPDR) and 15 of whom were diabetic patients without retinopathy and with normal ophthalmologic examination. The apparent diffusion coefficient (ADC) values were calculated in the orbitofrontal cortex (OFC), cingulated gyrus, thalamus, dorsomedial and dorsolateral frontal cortex, and corona radiate. The ADC values of the OFC, cingulated gyrus, and visual cortex were significantly increased in patients with PDR and NPDR compared with both patients without retinopathy and the control group (P<0.01). The ADC values of the OFC, cingulated gyrus, and visual cortex were significantly increased in patients with PDR compared with NPDR. The duration of disease and values of hemoglobin A1c were significantly correlated with ADC values of the OFC, cingulated gyrus, and visual cortex, respectively (P<0.01 or <0.05). We observed significantly increased ADC values of the visual center (OFC, cingulated gyrus, and visual cortex), supporting the association between DR and impairment in brain functioning. Diffusion-weighted imaging may serve to assess subclinical neurological involvement in DR, even when brain structural changes are absent.

Objective To evaluate the role of magnetic resonance diffusion-weighted imaging(DWI)in assessing brain injury in type 2 diabetic patients with diabetic retinopathy.Methods Fifty-four patients with type 2 diabetis mellitus were divided into three groups of A(with proliferative diabetic retinopathy),B(with non-proliferative diabetic retinopathy)and C(without diabetic retinopathy)with 18 cases each.Eighteen healthy volunteers were taken as the controls(group D).The apparent diffusion coefficient(ADC)of seven brain-related functional regions was measured by DWI of 3.0 T magnetic resonance imaging system.Results The ADC values of orbitofrontal cortex,cingulated gyrus and visual cortex were higher in groups of A and B than those in groups of C and D(P<0.01),which were higher in group A than those in group B(P<0.01).The course of disease was longer and HbA1c level was higher significantly in group A than those in group C(P<0.01).The course of disease and HbA1c level in diabetic patients were positively correlated to ADC values of the orbitofrontal cortex(rs=0.408 and 0.592,P<0.01),cingulated gyrus(rs=0.384 and 0.514,P<0.01)and visual cortex(rs=0.292 and 0.506,P<0.05).Diabetic retinopathy was positively correlated with ADC value of the visual cortex(rs=0.632,P<0.01).Conclusion An increase of ADC in the visual center indicates that diabetic retinopathy has a certain relation with brain function damage.DWI technique is valuable in assessing brain-related functional impairments of diabetic retinopathy in subclinical stage.

Objective To investigate the possible brain diffusion changes in different sites of the brain,particularly in areas associated with vision function,of patients with proliferative and nonproliferative retinopathy by using diffusion-weighted imaging (DWI).Methods Forty-five type 2 diabetic patients,admitted to our hospital from March 2012 to March 2013,were chosen in our study;according to the fundus examination,they were divided into group of proliferative retinopathy (n=15) group of non-proliferative retinopathy (n=15) and group of no retinopathy (n=15); another 15 healthy controls were chosen.T2WI and DWI with 3.0T magnetic resonance imaging (MRI) were performed; the apparent diffusion coefficient (ADC) values in the orbitofrontal cortex,cingulated gyrus,thalamus,dorsomedial and dorsolateral frontal cortex,and corona radiate were calculated; and then,Spearman's correlation analysis was performed between ADC values and both course of disease and glycosylated hemoglobin level.Results The ADC values of orbitofrontal cortex,cingulated gyrus and visual cortex in groups of proliferative and non-proliferative retinopathy were significantly increased as compared with group of no retinopathy and control group (F=95.268,P＜0.05); the ADC values of orbitofrontal cortex,cingulated gyrus and visual cortex in group of proliferative retinopathy were significantly increased as compared with those in group of non-proliferative retinopathy (P=0.004,0.001 and 0.009,respectively);no significant difference of them was noted between group of no retinopathy and control group (P＞0.05).The mean ADC values of orbitofiontal cortex,cingulated gyrus and visual cortex was positively correlated to the duration of disease (r=0.567 and P=0.001; r=0.491 and P=0.01; r=0.428 and P=0.003) and glycosylated hemoglobin level (r=0.336 and P=0.014; r=0.296 and P=0.049; r=0.370 and P=0.012).Conclusion Increased ADC values of visual center (orbitofrontal cortex,cingulated gyrus and visual cortex) conform the association between diabetic retinopathy and brain functioning impairment; DWI may serve to assess subclinical neurogical involvement in diabetic retinopathy,even when brain structural changes are absent.