Probability maps classify ischemic stroke regions more accurately than CT perfusion summary maps

Abstract

Objectives To compare single parameter thresholding with multivariable probabilistic classification of ischemic stroke regions in the analysis of computed tomography perfusion (CTP) parameter maps. Methods Patients were included from two multicenter trials and were divided into two groups based on their modified arterial occlusive lesion grade. CTP parameter maps were generated with three methods—a commercial method (ISP), block-circulant singular value decomposition (bSVD), and non-linear regression (NLR). Follow-up non-contrast CT defined the follow-up infarct region. Conventional thresholds for individual parameter maps were established with a receiver operating characteristic curve analysis. Probabilistic classification was carried out with a logistic regression model combining the available CTP parameters into a single probability. Results A total of 225 CTP data sets were included, divided into a group of 166 patients with successful recanalization and 59 with persistent occlusion. The precision and recall of the CTP parameters were lower individually than when combined into a probability. The median difference [interquartile range] in mL between the estimated and follow-up infarct volume was 29/23/23 [52/50/52] (ISP/bSVD/NLR) for conventional thresholding and was 4/6/11 [31/25/30] (ISP/bSVD/NLR) for the probabilistic classification. Conclusions Multivariable probability maps outperform thresholded CTP parameter maps in estimating the infarct lesion as observed on follow-up non-contrast CT. A multivariable probabilistic approach may harmonize the classification of ischemic stroke regions. Key Points • Combining CTP parameters with a logistic regression model increases the precision and recall in estimating ischemic stroke regions. • Volumes following from a probabilistic analysis predict follow-up infarct volumes better than volumes following from a threshold-based analysis. • A multivariable probabilistic approach may harmonize the classification of ischemic stroke regions.

Tissue Outcome Prediction in Patients with Proximal Vessel Occlusion and Mechanical Thrombectomy Using Logistic Models

“…In addition, spatial accuracy (Dice) was significantly improved for the logistic model in comparison to the thresholding-based model in our internal cross-validation. This superiority of the multiparametric logistic approach is well in line with a recent study which demonstrated higher predictive performance based on precision-recall plots and mean (but not mean absolute) volume difference [ 7 ]. In our analyses, the improved spatial accuracy did not generalize to the external validation cohort.…”

“…After deconvolution and calculation of perfusion parameter maps [ 4 ], individual ischemic core and penumbra are visualized and quantified by thresholding these parameter maps. Established criteria [ 2 , 3 ] are CBF rel (i.e., cerebral blood flow relative to the unaffected hemisphere) < 30% for the ischemic core [ 5 ] and T max (i.e., time-to-maximum of the flow-scaled residue function) > 6 s for the entire hypoperfused area including the ischemic penumbra [ 6 ], although different criteria are also applied in various commercially available software packages [ 7 ]. However, simple thresholding-based methods are not suited to exploit the full potential hidden in the high-dimensional perfusion data.…”

“…However, simple thresholding-based methods are not suited to exploit the full potential hidden in the high-dimensional perfusion data. Indeed, multiparametric machine learning methods (e.g., logistic regression and random forest classification) were shown to have the potential to outperform the simpler thresholding-based methods in predicting tissue outcome in case of successful (i.e., ischemic core) or unsuccessful (i.e., core and penumbra) recanalization [ 7 – 9 ]. These kinds of predictive models could therefore improve the selection of eligible patients with proximal anterior circulation vessel occlusion for mechanical thrombectomy by more accurate visualization and quantification of the tissue potentially salvageable (i.e., ischemic penumbra) as well as tissue that is no longer salvageable (i.e., ischemic core).…”

Spatial CT perfusion data helpful in automatically locating vessel occlusions for acute ischemic stroke patients

“…The arterial input function was determined automatically as described elsewhere (14). An in-house developed model-based non-linear regression method generated the perfusion maps of the cerebral blood flow, the cerebral blood volume, the mean transit time, and the time to peak (15). From these perfusion maps, a logistic model (that is described elsewhere) determined the hypoperfused region (i.e., the infarct core and the penumbra taken together) (16).…”

Neuroimaging Modalities Used for Ischemic Stroke Diagnosis and Monitoring

“…To this day, there is no standardized classification between the different processors to define the core and penumbra, since vendors establish their variations to define infarct cores and penumbras in their software [49]. However, it is the match or mismatch between the CBV and CBF with the MTT and TTP that will determine if the area is a penumbra or infarct core; this means that, if all the variants match on having altered levels, as indicated in Table 2, it is classified as an infarct.…”