Application Of MRI-based Cerebral Oxygen Metabolism Imaging And Perfusion Imaging In Ischemic Stroke

Objective:Cerebral oxygen metabolism and cerebral perfusion are closely related to brain function and cognitive activity,which play important roles in aging and ischemic stroke.Firstly,this study aimed to investigate the trajectories of resting-state cerebral metabolic rate of oxygen（CMRO₂）,oxygen extraction fraction（OEF）,and cerebral blood flow（CBF）in aging and ischemic stroke with cluster analysis of time evolution（CAT）for quantitative susceptibility mapping（QSM）and quantitative blood oxygen level dependent（q BOLD）imaging-based OEF and CMRO₂ mapping（CAT-QQ）and three-dimensional pseudo-continuous arterial spin labeling（3D-p CASL）.Secondly,the factors associated with the prognosis of ischemic stroke were analysed.Finally,the feasibility of 3D-p CASL for deep learning-based automatic detection of cerebrovascular disease was initially explored.Methods:In Part I of this study,108 healthy subjects were prospectively collected（age:20-70 years;42 males）.Enhanced gradient echo T₂ star weighted angiography（ESWAN）and3D-p CASL were used for reconstruction of resting-state voxel-wise CMRO₂,OEF,and CBF maps.Values of the three parameters were extracted in global gray matter（GGM）,global white matter（GWM）,6 lobes and 4 deep gray matter（DGM）regions.Comparisons of CMRO₂,OEF,and CBF between GGM and GWM and between males and females were performed.Correlation of the three parameters with age were also analyzed.In Part II of this study,115 patients in different ischemic stroke phases were retrospectively collected for horizontal comparison of OEF in the infarcted lesion and longitudinal comparison of OEF,CBF,and CMRO₂ in different lesion areas between two scans.The correlations of OEF,CBF,and CMRO₂ in the infarcted lesion in the first scan with clinical outcomes were analyzed where the primary outcome was the National Institutes of Health Stroke Scale（NIHSS）score at the second scan and the second outcome was modified Ranking Scale score at 3-month follow-up（3m-m RS）.In Part III of this study,55 subacute ischemic stroke patients with large artery atherosclerosis were retrospectively collected.Arterial transit artifact（ATA）on 3D-p CASL and good circulation（GC）on magnetic resonance angiography（MRA）were ranked as markers of leptomeningeal collaterals and fast collaterals,respectively.Univariate analyses and multivariable logistic regression analyses were conducted to identify the independent predictors of functional independence（3m-m RS≤2）.In Part IV of this study,a deep learning algorithm called two stream suppression Vision Transformer（Vi T）fusion network was proposed for large vessel severe stenosis and occlusion（LVSSO）detection.Three hundred and eighty patients with anterior circulation ischemic stroke or transient ischemic attack were retrospectively enrolled as the internal cohort and 34 patients were prospectively collected as the external cohort.The performances of three models that used DWI+3D-p CASL images or individual DWI or 3D-p CASL images were compared by conducting receiver operating characteristic（ROC）curve analysis.The algorithm was validated in the external cohort using the model with the best performance.Results:In healthy subjects,CMRO₂ and CBF were higher while OEF was lower in GGM when compared to those in GWM（P<0.05）.Females showed higher CMRO₂ and CBF in the GGM and GWM than males（P<0.05）.The difference of OEF between different genders did not reach statistical significance.After adjusting the effect of gender,there were weak negative correlations of CMRO₂ and CBF in the GGM with age（P<0.05）while no significant correlations were found between OEF in the GGM,or three parameters in the GWM as well as in the pallidum and age.CBF in the insula,CMRO₂ in the putamen and three parameters in the caudate were significantly different between age groups.In ischemic stroke patients,OEF in the infarcted lesion was lower than that in the contralateral mirror area regardless of stroke phases（P<0.05）and showed a declining tendency from acute to chronic phase（P=0.022）.Five of the 11 patients with longitudinal scans showed reversal（RR）of the lesion.Relative OEF（r OEF,compared to the contralateral mirror area）in the RR increased while r CBF decreased from the first to the second scan（P<0.05）.r CMRO₂was relatively stable.Two patients showed mismatch according to the enlarged lesion volume.The r OEF of the mismatch area fluctuated around 1.0 at the first scan and then decreased while CBF and CMRO₂ decreased continuously.In the first scan,the higher OEF and the lower CBF and CMRO₂ were correlated with lesion expansion,higher NIHSS score at the second scan and higher 3m-m RS.In the correlation analysis of collateral circulation and prognosis of ischemic stroke,ATA（adjusted Odds Ratio[OR]:13.785,95%confidence interval[CI]:2.608-72.870,P=0.002）and GC（adjusted OR:8.317,95%CI:1.629-42.454,P=0.011）were independent predictors of functional independence and they shared a positive synergistic effect.In the comparison of the three models for automatic detection of LVSSO,the area under the ROC curves（AUCs）were 0.922 for DWI,0.958 for 3D-p CASL,and 0.963 for DWI+3D-p CASL.The AUCs were not significantly different among the three models（P>0.05）.In the external cohort,the AUC for DWI+3D-p CASL was 0.877.Conclusions:CAT-QQ together with 3D-p CASL are of feasibility for accurate quantification of cerebral oxygen metabolism and cerebral perfusion in normal aging and ischemic stroke.Cerebral oxygen metabolism,cerebral perfusion and collateral circulation are closely related to the prognosis of ischemic stroke patients.Deep learning-based non-angiographic MRI is valuable to automatic detection of cerebrovascular disease.Future work will continue to focus on the MRI-based cerebral oxygen metabolism and perfusion imaging in ischemic stroke and other brain diseases,improving image reconstruction quality and speed,providing multi-dimensional pathophysiological parameters of the lesions,predicting disease development and prognosis with precision,and guiding scientific and effective therapy.