Study On The Evaluation Of Collateral Circulation In Patients With Acute Ischemic Stroke Based On Whole-brain 4D-CTA/CTP Technolog

Objective: To develop a quantitative assessment method of collateral circulation based on the whole-brain 4D-CTA technique,combined with first and second level collateral circulation,and to investigate its value in predicting the prognosis of patients with acute ischemic stroke(AIS).Materials and methods: AIS patients with large vessel occlusion(LVO)in the anterior circulation who were hospitalized in the Department of Neurology at our hospital from July 2020 to April 2021 were collected and their demographic and baseline clinical characteristics were recorded.The AIS patients were divided into a good prognosis group(0-2 points)and a poor prognosis group(3-6 points)according to the modified Rankin Scale(m RS).Baseline four-dimensional computed tomography angiography(4D-CTA)data of the patients were analyzed,and the first-and second-level collateral circulation were evaluated separately.The first collateral circulation score(FCCS)was obtained by assessing the circle of Willis on raw images in the arterial phase of 4D-CTA.The arterial peak phase image of the perfusion time density curve was selected as the arterial phase.The image post-processing techniques,including multiplanar reformation(MPR)and maximum intensity projection(MIP),were used to analyze the nine arteries that formed the circle of Wills,respectively.Evaluation and scoring were performed,and the sum of the scores of the nine arteries was the patient’s FCCS.The second collateral circulation score(SCCS)was assessed using the modified ASITN/SIR(American Society of Interventional and Therapeutic Neuroradiology/Society of Interventional Radiology)collateral grading scale based on dynamic multi-period CTA in the coronal 4D-CTA reconstruction images.Principal component analysis was used to develop a first principal component model to investigate the contribution of FCCS and SCCS to the total collateral circulation.The sum of FCCS and SCCS was defined as the comprehensive collateral circulation score(CCCS).Differences in baseline clinical characteristics of patients were compared between groups of good and poor prognosis.Logistic regression was used to analyze the relationship between FCCS/SCCS/CCCS and prognosis.The ROC curves of the three collateral circulation assessment methods were drawn,and the area under the curve(AUC)was calculated to evaluate the efficacy of the three methods for predicting prognosis.Results: ⑴ Subjects studied: A total of 71 patients with AIS were included,of whom 49 were male and 22 were female,aged 19-86 years,with a mean of 65.03 ± 14.4 years.There were 35 cases(49%)in the good prognosis group and 36 cases(51%)in the poor prognosis group,with a significant difference in age between the two groups(p<0.05).⑵Relationship of FCCS,SCCS and CCCS: The eigenvector values of FCCS and SCCS obtained from the principal component model were both 0.707,so the two contribute similarly to the total collateral circulation.They both showed no statistical difference between the poor and good prognosis groups(p>0.05).However,CCCS showed a statistical difference between the good and poor prognosis groups(p=0.003).⑶ Multivariate logistic regression analysis showed only CCCS(OR:0.632;95%CI:0.457-0.873;p=0.005),and NIHSS(OR:1.164;95%CI:1.059-1.279;p=0.002)were independent of prognosis predictors.⑷ Among the ROC curves of FCCS,SCCS,and CCCS,the AUC of CCCS was the largest(0.688,p=0.007).Conclusion: The compensatory value of FCCS in acute ischemic stroke was similar to that of SCCS.CCCS,combined the integrity assessment of the Willis circle with second-level collateral circulation based on dynamic CTA,could be used as a new imaging indicator to quantitatively assess collateral circulation and predict prognosis.Objective: The whole-brain 4D-CTA technique was used to analyze the cortical venous velocity changes in acute ischemic stroke with large vessel occlusion(LVO-AIS)in the anterior circulation,and to investigate the value of delayed cortical venous visualization on the affected side of AIS patients in assessing collateral circulation and prognosis.Methods: The whole-brain one-stop 4D-CTA/CTP data,baseline clinical data,and 90-day m RS of AIS patients with internal carotid artery or middle cerebral artery M1/M2 segment occlusion were collected from June2020 to October 2021.Whole-brain one-stop 4D-CTA/CTP data were also collected from a group of subjects with normal outcomes(healthy controls).The venous inflow time(VIT),the venous peak time(VPT),and the venous outflow time of the cerebral cortex(VOT)on both sides of the patients and the healthy control group were analyzed and recorded,respectively.Statistical description and comparison of VIT,VPT,and VOT were performed on the affected side and the healthy side of the patient group,and the healthy control group.Second,the patient groups were divided into subgroups according to different prognosis,collateral circulation score,time of onset,responsible vessel,and TOAST typing,and VIT,VPT,and VOT were statistically described and compared for each subgroup.Third,Spearman correlation analysis and binary multivariate logistic regression analysis were used to explore the relationship between the three venous time and the main clinical indicators such as m RS,arterial collateral score,and NIHSS.Finally,receiver operating characteristic(ROC)curves were drawn to assess the value of cortical vein imaging delay in predicting prognosis.result:(1)Subjects studied: 149 LVO-AIS were collected.The baseline clinical characteristics of the patients were as follows:(1)Age 19-92 years,mean 66.81±13.75 years,97 cases in men and 52 cases in women.(2)Onset time: <6 hours 52 cases,6-24 hours 43 cases,>24 hours 54 cases.(3)Responsible vessels: 96 cases of MCA occlusion,26 cases of ICA occlusion,22 cases of ICA combined with MCA occlusion,3 cases of ICA combined with MCA and ACA occlusion,and 2 cases of MCA combined with ACA occlusion.(4)The median of NIHSS was 8(2)points.(5)The median of modified dynamic CTA ASITN/SIR collateral score was 3(1)points.(6)TOAST classification: 91 cases of LAA,50 cases of CE,and 8cases of SOE.(7)The treatment methods: 38 cases of endovascular interventional therapy,20 cases of thrombolytic treatment,8 cases of thrombolytic bridge endovascular interventional therapy,and 83 cases of conservative treatment.(8)Thirty-five cases followed hemorrhagic transformation.A total of 73 cases of whole brain one-stop 4D-CTA/CTP data were collected from the healthy control group.(2)Cortical venous velocity changes in the patient group: The median of VIT and VPT were delayed on the affected side and the healthy side in the patient group compared with normal controls.There was no noticeable delay in the median of VOT.There were statistically significant differences in VIT,VPT and VOT between the affected side and the healthy side in the patient group and between the affected side in the patient group and the control group(p<0.05).There were statistically significant differences in VIT and VPT between the healthy side in the patient group and the control group(p<0.05).(3)Inter-subgroup comparisons:(1)VIT and VPT on the affected side of the patients were significantly different between the groups with good and poor collateral circulation(p<0.05).(2)There were significant differences in VIT and VPT of patient’s affected side between the groups with good and poor prognoses(p<0.05).(3)The patients were divided into three subgroups according to the onset time:<6h,6-24 h,and >24h.VIT and VPT were significantly different among the three subgroups.To further differences: VIT and VPT in group <6h and group 6-24 h had significant differences from those in group >24h(p<0.05),there were no significant differences in VIT and VPT between group <6h and group 6-24h(p>0.05).(4)According to TOAST classification,all patients were divided into three subgroups: LAA,CE,and SOE.There were significant differences among the three subgroups in VIT and VPT(p<0.05).VIT and VPT showed significant differences between the LAA group/SOE group and the CE group(all p<0.05).There was no significant difference in VIT and VPT between LAA and SOE groups(p>0.05).(4)Correlation and regression analysis: The patient’s VIT and VPT on the affected side were positively correlated with m RS and NIHSS(p<0.05),and VIT and VPT on the affected side were negatively correlated with arterial collateral score and onset time(p<0.05).VIT and VPT on the affected side were not associated with hemorrhagic transformation(p>0.05).Univariate logistic regression analysis found that age,VIT/VPT/VOT on the affected side,NIHSS,onset time,and arterial collateral score were associated with prognosis(p<0.1).All the significant factors(p<0.1)in the univariate logistic regression analysis were included in the binary multivariate logistic regression model,and the conditional forward stepwise regression was used to analyze,and it was found that VPT on the affected side,NIHSS,and arterial collateral score were independent predictors of poor prognosis,the model predicted a poor prognosis with a correct percentage of 79.6%,after limiting the onset time of patients within 24 hours,VPT on the affected side and NIHSS were independent predictors of poor prognosis,and the correct percentage of predicting poor prognosis rose to 91.5%.(5)ROC curve: the AUC of NIHSS(0.786)was the larger between the three independent predictors of poor prognosis,the AUC of VPT(0.697)on the affected side were lower.Analysis of patients of onset <24h: the AUC of the NIHSS decreased(0.753),the AUC of VPT(0.711)on the affected side rose.Conclusions: VPT on the affected side of 4D-CTA in patients with LVO-AIS can be a reliable imaging indicator of poorly established collateral circulation and poor prognosis.Altered cortical venous velocity in patients may be useful for the understanding of pathophysiological changes in stroke,determination of disease stage,and etiological typing.Objective: To establish an objective,fast and accurate model for lateral branch circulation assessment based on deep learning algorithms using whole-brain dynamic CTA images.Methods: We collected LVO-AIS patients in the anterior circulation who were hospitalized in the neurology department of our hospital from June 2020 to August 2021.Their demographic and baseline clinical characteristics were recorded.The baseline one-stop whole-brain four-dimensional computed tomography angiography(4D-CTA)/CT perfusion(CTP)data of all AIS patients were analyzed.The images of the arterial phase,arteriovenous phase,venous phase,and late venous phase were extracted from 4D-CTA according to the perfusion time density curve.Then the non-contrast CT scan(NCCT)was subtracted respectively from the four phases images to obtain the subtraction images of the arterial phase,arteriovenous phase,venous phase,and late venous phase.Each patient was marked with good or poor collateral circulation according to the modified ASITN/SIR(American Society of Interventional and Therapeutic Neuroradiology/Society of Interventional Radiology)collateral grading scale based on dynamic multi-phase CTA.Based on the Res Net34 classification network,we developed a single-image input processing network and a multi-image input network for binary classification evaluation of the pros and cons of collateral circulation.The training and test sets were 65 and 27 cases,respectively,and Monte Carlo cross-validation was used for five iterations.The network performance was evaluated by accuracy,precision,recall,F1 score,and area under the receiver operating characteristic curve(ROC).Results:(1)Patient features: 92 LVO-AIS patients in anterior circulation were included.The baseline clinical characteristics of the patients were as follows:(1)Average age was 66.24±13.64 years,including61 males and 31 females.(2)MCA occlusion(with or without ACA occlusion)in 64 cases,ICA occlusion in 14 cases,ICA and MCA occlusion in 13 cases,ICA,MCA and ACA occlusion in 1case.(3)The median NIHSS on admission was 8.50(12)points.(4)The median modified ASITN/SIR score based on dynamic multi-phase CTA was 3(1)points.(5)TOAST classification included: 63 cases of large-artery atherosclerosis(LAA),24 cases of cardio embolism(CE),and 5 cases of stroke of other determined etiology(SOE).(6)Treatment methods included: 22 cases of endovascular interventional therapy,8 cases of thrombolytic treatment,3 cases of thrombolytic bridge endovascular interventional therapy,and 59 cases of conservative treatment.(6)19 cases followed hemorrhagic transformation.(2)Good collateral circulation group versus poor collateral circulation group: age,NIHSS,occluded artery and hemorrhagic transformation were statistically different between groups(p<0.05).There was no significant difference in gender,onset time,treatment method,and TOAST classification between the two groups(p>0.05).(3)Model performance:(1)The accuracy of the single-image input processing network was 0.852±0.045,the precision was 0.932±0.034,the recall rate was 0.827±0.076,the FI score was 0.860±0.044,and the AUC was 0.89±0.05.(2)the accuracy of the multi-image input processing network was The accuracy was 0.822±0.017,the precision was0.571±0.081,the recall was 0.813±0.056,the FI score was 0.836±0.008,and the AUC was 0.47±0.12.Conclusion: Based on the ResNet34 classification network,the single-image input processing network that used the axial maximum density projection images of the four-phase subtraction CTA for splicing input could classify the AIS collateral circulation better.This automated collateral assessment method can help streamline clinical workflows,aid clinical decision-making,and screen patients for reperfusion therapy.