| Part Ⅰ Clinical research on assessment of collateral circulation in patients with the unilateral M1 MCA occlusion and its correlation with cerebral infarction based on multi-phase CTAObjective:To investigate the value of multi-phase computed tomographic angiography(MP-CTA)in accurate evaluation of collateral circulation in patients with the middle cerebral artery(MCA)M1 segment occlusion based on DSA,in order to find a fast,effective and safe evaluation tool of collateral circulation.To investigate the difference of collateral circulation opening in different subgroups patients with MCA occlusion and its correlation with stroke.Methods:The clinical and imaging data of 1986 patients who were treated in our hospital(The ethics board approved this study)from May 2015 to December 2020 and underwent one-stop whole brain CTP-CTA with Toshiba aquilion one 320 row dynamic volume CT were collected continuously.The patients were divided into symptomatic group(acute stroke in vascular occlusion responsible area confirmed by DWI examination)and asymptomatic group(clinical manifestation and DWI manifestation of no stroke in MCA occlusion side)based on the clinical manifestations and DWI manifestations.The original CT scan data were imported into the Vitrea FX after post-processing software reconstruction to obtain the regular nonenhanced CT images,and the single-phase CTA(SP-CTA)imgaes and MP-CTA images with the maximal intensity projection(MIP)reconstruction technique,and assessed the filling situation and number of collateral vessels on the affected side(referring to the healthy side vessel)and performed the Alberta Stroke Program Early CT Score(ASPECTS),to comparatively analyze,the lateralization phenomenon of anterior cerebral artery(ACA)and posterior cerebral artery(PCA)(thickening of main trunk,increase of branches and prolongation of terminals),and the difference of venous compensation between symptomatic group and asymptomatic group and its relationship with stroke.The value of MP-CTA and SP-CTA,lateralization and venous compensation in the evaluation of collateral circulation were compared based on DSA in 42 patients in the symptomatic group,combined with 90-day modified Rankin Scale(mRS)score to analyze the effect of MP-CTA collateral circulation score on the prognosis of symptomatic group.Kappa test was used to evaluate the consistency of collateral circulation scores of two doctors and different scoring methods including MP-CTA score,SP-CTA score,lateralization phenomenon and venous compensation based on DSA.Spearman’s correlation was used to analysis the relationship between collateral circulation classification and prognosis.The diagnostic efficacy of MP-CTA score in evaluating acute ischemic stroke(AIS)by using receiver operating characteristic(ROC),including the area under the curve(AUC),sensitivity,specificity.In addition,the relevant general clinical baseline data(such as gender,age,smoking history,history of hypertension,diabetes,hyperlipidemia,coronary heart disease and hyperhomocysteinemia)of all patients,and the admission National Institutes of Health Stroke Scale(NIHSS)scores of symptomatic groups were collected and analyzed statistically,and the clinical prognosis including good prognosis of mRS≤2 and poor prognosis of mRS>2 was recorded by telephone follow-up within 3 months based on the 90-day mRS score.Results:A total of 338 patients with unilateral middle cerebral artery occlusion were included in the present study,including 253 patients of symptomatic group[144 males and 109 females,aged 32-88 years old,with an mean age of 63.18 years± 11.31(standard deviation)]and 85 patients of asymptomatic group[48 males and 37 females,aged 22-67 years old,with an average age of 59.84 years±15.09(standard deviation)].1.Gender and the history of coronary heart disease,hyperlipidemia,smoking and diabetes between symptomatic group and asymptomatic group had not significant differences(all p>0.05).The age older,history of hypertension,increased C-reactive protein,hyperhomocysteinemia,increased erythrocyte sedimentation rate in the symptomatic group patients were higher than those in the asymptomatic group(all p<0.05).The stroke severity of patients in the symptom group at admission was significantly correlated with the clinical prognosis(p=0.001),the higher the NIHSS score,the worse prognosis(p=0.000).2.MP-CTA score of collateral circulation had the highest consistency with DSA(κ=0.821,p=0.000)compared with SP-CTA score and the lateralization of ACA and PCA,with an area under ROC curve of 0.842(95%confidence interval 0.707-0.976,p=0.001),diagnostic sensitivity of 93.55%,specificity of 90.91%and Youden index of 0.84.3.Most patients of the asymptomatic group(98.82%)had good collaterals,MP-CTA collateral score was mainly grade 3-4,the incidence of good collateral circulation was significantly higher than that in the symptomatic group with significant statistical difference(p<0.05).The number of distal veins and lateralization in the asymptomatic group was more higher than that in the symptomatic group(p<0.05),while the number of distal veins in the asymptomatic group was mainly increased or normal(90.59%).4.The evaluation score of collateral circulation with MP-CTA,SP-CTA and DSA was negatively correlated with the 90-day mRS score by using Pearson’s correlation analysis,with the correlation coefficients of-0.629,-0.274 and-0.686 respectively,and the statistical values were x2=16.593(p=0.000),x2=3.153(p=0.142)and x2=19.775(P=0.000).The quantitative score of collateral circulation with MP-CTA predicted clinical outcomes was significantly better than that of SP-CTA(p=0.000),with greater concordance with DSA evaluation.5.The kappa consistency test for the evaluation score of collateral circulation among MP-CTA,SP-CTA,lateralization and venous compensation by two imaging doctors showed that the data measured were dramatically consistent,with kappa values of 0.970,0.595,0.952 and 0.975 respectively,and thep values all equal to 0.000,while showing that the evaluation of SP-CTA was more subjective.Conclusions:1.Hyperlipidemia,coronary heart disease,hyperhomocysteinemia,increased erythrocyte sedimentation rate,increased C-reactive protein are independent risk factors for AIS in patients with unilateral M1 segment MCA occlusion,suggesting that active control of risk factors was of great significance to avoid the occurrence of AIS in such patients.2.The MP-CTA collateral circulation classification had greater consistency among observers,and it was reliable to evaluate the dynamic blood flow changes of brain tissue.Compared with the commonly used SP-CTA evaluation of collateral circulation in clinic,MP-CTA collateral evaluation has greater good consistency with DSA that is the gold standard in evaluating collateral circulation,which should be a simple,quick and ideal tool and should be widely available,and indicate that there is a risk of AIS when MP-CTA score is ≤2.5,a lower score is associated with a worse clinical outcomes.MP-CTA grade is an independent predictor of good prognosis.Part Ⅱ Clinical research on assessment of cerebral collaterals and cerebral infarction in patients with the unilateral M1 segment MCA occlusion based on 320-row CTPObjective:To investigate the value of computed tomographic perfusion(CTP)in evaluating collateral circulation in patients with M1 segment of the middle cerebral artery(MCA)occlusion,and to analyze the differences of cerebral perfusion in different subgroups patients with M1 segment MCA occlusion and their correlation with acute ischemic stroke(AIS).Methods:The clinical and imaging data of 1986 patients who were treated in our hospital from May 2015 to December 2020 and underwent one-stop whole brain CTP-CTA with Toshiba aquilion one 320 row dynamic volume CT were collected continuously.According to inclusion and exclusion criteria,a total of 338 patients with unilateral middle cerebral artery occlusion were recruited in the study,including 253 patients in the symptomatic group(CTP and DWI examination confirmed acute cerebral infarction in the responsibility area of M1 segment occlusion)and 85 patients in the asymptomatic group(no clinical manifestations and DWI manifestations of cerebral infarction on the MCA occlusive side).Patients in the symptom group(n=253)underwent CTP and CTA within one week after onset,and 42 patients in the symptom group underwent digital subtraction angiography(DSA).The original CT scan data were imported into Vitrea FX after post-processing software reconstruction to obtain the the regular non-enhanced CT images and CTP images.According to the diffusion restricted lesions displayed by Diffusion weighted imaging(DWI)or abnormal perfusion areas displayed by CTP images,the regions of interest were manually delineated on the corresponding CTP images.The values of perfusion parameters in each area of cerebral infarction(core infarction area,ischemic penumbra,benign insufficiency area)and its mirror area were measured and analyzed quantitatively,and the differences of quantitative perfusion parameters in benign hypoperfusion area of symptom group and affected area of asymptomatic group were compared and analyzed Area under the curve(AUC)of Receiver Operating characteristic curve(ROC)was used to assess the diagnostic efficacy of the quantitative parameter values of CTP for AIS,to evaluate the perfusion parameters with the highest specificity for the degree of cerebral ischemia,and to identify the diagnostic threshold for differentiating cerebral infarction core from ischemic penumbra.The differences of perfusion parameters and their relationship with cerebral infarction were compared,and the evaluation effect of main parameters on the prognosis of symptomatic group patients were analyzed in combination with 90-day modified Rankin Scale(mRS)score.In addition,All patients’NIHSS scores at admission and modified Rankin Scale scores(90d mRS scores)after 3 months later by telephone follow-up were also collected.Results:1.Compared with mirror healthy side,CT perfusion images of the core area of stroke showed a significant decrease in CBF and CBV(all p=0.000),shortening of MTT(p=0.000),and a significant prolongation of TTP and DT values(p=0.000);the CT perfusion map of ischemic penumbra showed that CBF decreased(P=0.000),CBV increased normally or slightly(p=0.000),and MTT,TTP and DT prolonged(all p=0.000).The CT perfusion map of benign hypoperfusion area showed that CBF decreased(p=0.000),CBV increased normally or slightly(p=0.008),and MTT,TTP and DT prolonged(all p=0.000).There was a decrease in the CBF of the diseased area in the asymptomatic group,but there was no significant difference between the two groups(p=0.094),and the CBV,MTT,TTP and DT in the diseased area were all higher than those in the normal control side(t values were 2.396,3.152,2.193 and 2.423 respectively;P values were 0.018,0.002,0.030 and 0.016 respectively),MTT of them was the most obvious.2.The CBF and CBV of benign hypoperfusion area of symptomatic group were lower than those in the lesion area of asymptomatic group,but there was no significant difference between them(t values were-1.882 and-0.613,p values were 0.061 and 0.540 respectively).MTT,TTP and DT in benign hypoperfusion area of symptomatic group were higher than those in the lesion area of asymptomatic group(t values were 6.709,3.513 and 3.391,p values were 0.000,0.001 and 0.001,respectively).The MTT,TTP and DT is more sensitive than CBF and CBV in reflecting the changes of cerebral perfusion.3.The omparison of cerebral perfusion parameters in infarct core area,penumbra area,benign hypoperfusion area in symptomatic group and the lesion area in asymptomatic group are as follows:the CBF and CBV increased gradually in four groups.CBF in infarct core of symptomatic group was lower than that in penumbra area,benign hypoperfusion area and lesion area of asymptomatic group,and CBF in ischemic penumbra area was lower than that in the lesion side of asymptomatic group with statistical differences(all p<0.05);CBF in ischemic penumbra area was lower than that in benign hypoperfusion area,and that in benign hypoperfusion area was lower than that in the lesion area of asymptomatic group,the difference was not statistically significant(all p>0.05);the CBV in infarct core area was lower than that in ischemic penumbra,benign hypoperfusion area and the lesion area of asymptomatic group(p<0.05);the CBV in ischemic penumbra was higher than that in benign hypoperfusion area and the lesion area of asymptomatic group without statistical differences(p>0.05).There was a progressive trend of decrease between TTP and DT,with statistically differences between the two pairwise comparisons of TTP and DT in the infarct core area,penumbra area and benign hypoperfusion area in symptomatic group(all p<0.05).The results of comparison among the four groups showed that the MTT in the core area of stroke was the lowest(3.93 s± 1.36 s),with significant statistical difference(p<0.01).The MTT in the penumbra area,benign hypoperfusion area and lesion area of asymptomatic group showed a gradually decreasing trend with statistically differences between the two pairwise comparisons(all p<0.05).4.The correlation coefficients of CBF,CBV,MTT,TTP and DT with prognosis were 0.638,0.686,0.370,-0.023 and-0.034 respectively,and the P values were 0.000,0.000,0.016,0.886 and 0.828 respectively through Pearson’s correlation analysis,which showed that CBV,CBF had the highest correlation with prognosis,showing a significant positive correlation,followed by MTT,and the correlation between TTP and DT was poor.Conclusions:1.CTP can quickly and quantitatively evaluate the status of collateral circulation in patients with unilateral middle cerebral artery occlusion.The CBV value<2.25ml/100g can be considered as the core area of stroke(the cut-off value for distinguishing the infarct core from the penumbra),and CBF value<22.5ml/(100g·min)indicates the risk of developing acute ischemic stroke.2.The CBF and CBV in the infarct core area,penumbra area,benign hypoperfusion area and lesion area of asymptomatic group show a gradual increase,a gradual decrease in TTP and DT.There is obvious compensatory blood flow in asymptomatic group.TTP,DT and MTT are more sensitive to the changes of cerebral hemodynamics.CBV and CBF are significantly positively correlated with prognosis,suggesting that early vasodilation or adequate collateral flow reperfusion is very important in patients with M1 MCA occlusion.Part Ⅲ The research on constructi on of multi-dimensional classification regression model for predicting cerebral infarction based on the characteristics of multi-phase CTA and CTPObjective:To investigate the multi-dimensional classification regression model for predicting stroke based on the characteristics of MP-CTA and CTP in predicting the risk of stroke in patients with unilateral middle cerebral artery(MCA)occlusion,so as to provide support for personalized precision diagnosis and therapy,scientific management of disease course and prognosis evaluation for the above patients.Methods:A multi-dimensional classification regression model for predicting stroke based on the data package of Python 3.7.9 and sklearn 0.23.2,was constructed by integrating common clinical risk factors(age,gender,hypertension,diabetes,hyperlipidemia,coronary heart disease,homocysteinemia,smoking history),laboratory indicators(including erythrocyte sedimentation rate,high sensitivity C-reactive protein,etc.)and multimodality CT features(including MP-CTA score,lateralization,distal venous compensation,CBV,CBF,etc.),and using random forest algorithm,the importance screening was carried out respectively.Data analysis and prediction model construction mainly includes the following four steps:data preprocessing,feature selection,data modeling and prediction model evaluation.Data preprocessing to eliminate irrelevant variables and extreme value processing;In order to avoid model overfitting,sample data were randomly sampled and randomly grouped according to 7:3.Randomly divided into training set and test set proportionally.Chi-square test and T-test was used for feature selection and relative important associated variables screening in feature selection to ensure the stability of results of training set and testing set and reduce the over-combination of prediction model to a certain extent.In the final prediction model construction,more than 60%of the selected characteristic variables were included into the Logistic regression model and the Random forest model to establish the fine cerebral infarction prediction model,and the two models were compared using ROC.The receiver operating characteristic curve(ROC)mainly including AUC and accuracy was drawn to assess the effectiveness of the model.Results:1.The random forest algorithm fused with the common clinical risk factors(age,hypertension,homocysteinemia),laboratory indicators(erythrocyte sedimentation rate,high sensitivity C-reactive protein)and multimodality CT features(MP-CTA score,lateralization,distal venous compensation,CBV,CBF),that are different between the symptomatic group and the asymptomatic group,constructed a multi-dimensional classification regression model for predicting stroke,the accuracy of training set and test set in distinguishing stroke from nonstroke was 98.46%and 96.92%respectively.There were significant differences in the proportion of the importance of each feature with the highest proportion of CBV,MP-CTA score and CBF(28.20%,21.20%and 16.00%respectively).The ROC curve was drawn to evaluate the effectiveness of the model with an accuracy of 0.96,the AUC=0.995,sensitivity of 0.97,specificity of 0.88,positive predictive value of 0.97 and negative predictive value of 0.92.The AUC and cut-off values for the evaluation of stroke by MP-CTA score,CBF and CBV were 0.889 and 2.5 points,0.875 and 22.5ml/(100g·min),0.995 and 2.25ml/100g,respectively.2.The common clinical risk factors(age,gender,hypertension,diabetes,hyperlipidemia,coronary heart disease,homocysteinemia,smoking history),laboratory indicators(mainly including erythrocyte sedimentation rate,high sensitivity C-reactive protein,etc.)and multimodality CT features(mainly including MP-CTA score,lateralization,distal venous compensation,CBV,CBF,etc.)between symptomatic and asymptomatic groups were fused using a random forest algorithm to construct a multi-dimensional classification regression model for predicting stroke,the accuracy of training set and test set in distinguishing stroke from non-stroke was 96.28%and 92.31%respectively.There were significant differences in the proportion of the importance of each feature,with the highest proportion of CBV,MP-CTA score and CBF(30.00%,18.70%and 17.30%respectively).The ROC curve was drawn to evaluate the effectiveness of the model,which yielded an accuracy of 0.92,the AUC=0.98,sensitivity of 0.94,specificity of 0.88,positive predictive value of 0.97 and negative predictive value of 0.78,showing that the diagnostic efficiency of the prediction model is also better,which was slightly lower than that of the multi-dimensional classification regression model for predicting stroke based on positive clinical risk factors,laboratory indicators and image characteristics for predicting stroke.3.Logistic regression function was used to construct a multidimensional classification model for predicting cerebral infarction based on fusing positive clinical risk factors(age,hypertension,and homocysteine levels),laboratory indicators(blood sedimentation,allergic C-reactive protein)and image features(MP-CTA,the asymmetry phenomenon,distal venous compensatory,CBV,CBF),the accuracy of distinguishing cerebral infarction from non-cerebral infarction in the training set and the testing set was 98.73%and 97.06%,respectively.The accuracy,AUC under the curve,sensitivity and specificity of ROC curve were 0.97,0.997,0.97 and 0.96,respectively.The negative predictive value(NPV)was 0.93 and the positive predictive value(PPV)was 0.99.Compared with the multi-dimensional classification model for predicting cerebral infarction based on positive clinical risk factors,laboratory indicators and image features constructed by random forest algorithm,the diagnostic performance of this prediction model is better.Calibration curve of testing sets indicated the prediction effects are good,and there was no difference between Random Forest prediction model and Logistic Regression prediction model.Conclusions:Constructing a multi-dimensional classification model fused with the common clinical risk factors(age,hypertension,homocysteinemia),laboratory indicators(ESR,high-sensitivity C-reactive protein)and multimodality CT features(MP-CTA,lateralization,distal venous compensation,CBV,CBF)with differences based on Random forest algorithm and Logistic regression function,could predict with high diagnostic efficiency the risk of stroke in patients with unilateral M1 MCA occlusion and can provide support for personalized accurate diagnosis and treatment,scientific management of disease course and prognosis assessment of above patients. |
PDF Full Text Request