Infarct-core CT perfusion parameters in predicting post-thrombolysis hemorrhagic transformation of acute ischemic stroke

Ischemic stroke is defined as an episode of neurological dysfunction caused by focal cerebral, spinal or retinal infarction accompanied by overt symptoms.¹ About 90% of all strokes are ischemic, while roughly 10% are hemorrhagic (including intracerebral and subarachnoid hemorrhage).²

Patients presenting signs and symptoms of acute ischemic stroke (AIS) undergo a series of CT imaging procedures, including non-contrast CT to exclude pre-treatment intracranial hemorrhage, CT angiography (CTA) to determine the precise location of vessel occlusion, CT perfusion (CTP) to differentiate between potentially salvageable and irreversibly damaged brain tissue, and post-treatment non-contrast CT to exclude thrombolysis- related hemorrhage.^{3, 4, 5, 6}

The standard AIS treatment method within 4.5 hours of symptom onset is intravenous thrombolysis (IVT) with tissue plasminogen activator (tPA) injection.^{7, 8} Patients with IVT-therapy contraindications or ineffectiveness may be eligible for endovascular mechanical thrombectomy (MT).^{8, 9}

Hemorrhagic transformation (HT) is a conversion of ischemic brain tissue into a hemorrhagic lesion due to blood-brain barrier disruption. It may occur spontaneously in ischemic brain tissue but may also be triggered by reperfusion.^{10, 11} HT occurs in 4.5 - 68.0% of AIS clinical cases and has a higher incidence in patients treated with IVT than in patients without such treatment.^{12, 13, 14} While mild to moderate HT may not seriously impact the clinical outcome, severe HT is a significant predictor of neurological deterioration and higher mortality.^{14, 15}

CTP is an imaging technique that measures brain tissue blood perfusion by analyzing time-attenuation curves of contrast agent in input artery and parenchyma, generating maps of CT perfusion parameters (CTPPs). CTPPs are cerebral blood volume (CBV), mean transit time (MTT) and cerebral blood flow (CBF). CBV is defined as the total volume of flowing blood in a given volume of brain. MTT is defined as the average transit time of blood through a given brain region. CBF is defined as the volume of flowing blood moving through a given volume of brain in a specific amount of time. The three CTPPs are associated by the equation: CBF = CBV / MTT.¹⁶

Absolute CTPPs are values of a certain brain region while relative CTPPs are values of a certain brain region divided by values of the contralateral brain region. In the context of AIS, relative CTPPs are values measured in the pathological hemisphere expressed as a percentage of the values measured in the contralateral normal hemisphere.¹⁷

Previous studies have shown that CTPPs of the whole infarct area (penumbra and infarct core as a single region) could be used to predict HT, finding relative CBV (rCBV) ≤ 1.09 and T_max > 14 s , respectively, to be the most predictive of HT, with relative CBF (rCBF) < 30% also being of considerable utility in predicting HT.^{18, 19} One study found neither CBF nor CBV to be significantly different between cases and controls, while not examining relative CTPPs.²⁰ Another study examined infarct-core CTPPs, finding CBV ≤ 0.5 mL/100 g to be predictive of symptomatic intra-cerebral hemorrhage, while also not investigating relative CTPPs.²¹ One study initially proposed separate analysis of infarct-core CTPPs but eventually dismissed the idea due to insufficient sample size.¹⁸

The rationale for separate analysis of the infarct core subregion is that it may provide a different insight into HT prediction than a whole-infarct approach, due to the elimination of the “average-out” effect.²¹ The reasoning behind the use of relative rather than absolute CTPPs is to account for potential interpatient variability, while also avoiding the inter-vendor variability of postprocessing software.^{18, 22} Our study thus aimed to investigate CTPPs of the infarct core in predicting HT, with an emphasis on relative CTPPs

Seventy-five patients with AIS who had undergone CTP imaging and had been treated with IVT were included in the study. Significant differences in mean values between cases and controls were observed (p = < 0.000–0.009) for CBF, CBV, rCBF and rCBV.

The area under the curve (AUC) of the receiver operating characteristic (ROC) of rCBF (0.736) showed a satisfactory ability to differentiate between the occurrence and non-occurrence of HT, while AUCs of CBF, CBV and rCBV showed comparatively inferior differentiation abilities (0.704– 0.676).

HT is a potentially grave complication of IVT, occurring in 4.5–68.0% of clinical AIS cases. Severe

Figure 1

Figure 2

Figure 3

HT is a significant predictor of neurological deterioration and higher mortality.¹⁵ Various HT prediction methods have been investigated, including CT, CTP, SPECT and diffusion- and perfusion-weighted MR imaging.^{21, 22, 23, 24, 25, 26} According to previously published data, CTPPs of the whole infarct area effectively predict HT. Jain et al. and Yassi et al. found relative CTPPs – rCBV and rCBF, respectively, – to be the most predictive of HT.^{18, 19}

To the best of our knowledge, this study is the first to analyze relative CTPPs of infarct core in predicting HT. On the one hand, the infarct core itself seems not to have been the main focal point of investigations so far, due to various circumstances, including limited sample size, as was the case with Jain et al.¹⁸ On the other hand, the sole study that did analyze infarct core, carried out by Lin et al., investigated absolute CTPPs only. The findings of Zussman et al. cautioned against using absolute CTPPs due to inter-vendor variability of post-processing software, while Jain et al. also warned of possible interpatient variability of absolute CTPPs, encouraging the use of relative CTPPs instead.^{21, 22}

The above arguments prompted us to focus on analyzing relative CTPPs of the infarct core, with the full knowledge that there might currently be no point of reference to compare our results directly. We found additional reasoning for an infarct-core approach in the fact that our attempts at free-hand whole-infarct designation with image segmentation to eliminate structures that were irrelevant for CTP (e.g., large vessels and sulci) proved futile in many cases; eliminating a large vessel completely often required such a large Hounsfield units (HU) exclusion interval that it inadvertently also deselected the majority of tissue viable for CTP analysis. Additionally, we found the non-segmentation single-ROI infarct-core approach to be fast and straightforward – a potential advantage when using CTPPs in an emergency clinical setting.

Our study of infarct-core CTPPs demonstrates that rCBF < 4.5% of the contralateral mean best predicts the occurrence of HT (sensitivity 71.0%, specificity 52.5%). Considering studies that opted for whole-infarct measurement of relative CTPPs, it should be noted that our approach offered considerably inferior sensitivity but better specificity than whole-infarct rCBF < 30% (sensitivity 100%, specificity 39.0%) studied by Yassi et al., while whole-infarct rCBV < 1.09 (sensitivity 100%, specificity 58.3%) researched by Jain et al. proved to be superior to both infarct-core rCBF and whole-infarct rCBF. Infarct-core rCBV < 8.5% (sensitivity 71.4%, specificity 42.5%) examined by our study proved to be inferior in HT prediction to the aforementioned CTPPs but might be considered as an additional parameter to rCBF when evaluating the infarct core due to similar sensitivity.

A possible reason for the extremely low infarct-core rCBF examined in our study being less sensitive in prediction of HT than the moderately low whole-infarct rCBF examined by Yassi et al. might be that the severe hypo-perfused stroke region contains very low levels of contrast, which in certain cases could be undetectable by CTP.¹⁹ Additionally, our ROI placement protocol limited the maximum diameter of circular ROI to 20 mm, and while smaller ROIs help eliminate the “average-out” effect that is associated with large, whole-infarct freehand ROIs, they may also be more susceptible to the effect of random pixel noise.²¹ Another possible source of the comparatively higher data heterogeneity in our study may be the potential temporal truncation of the contrast bolus in the infarct region.^{27, 28}

Our study has several limitations. It was a retrospective study. The effects of stroke severity and anatomical location were not controlled by matching, nor was the anatomical location of HT. HT was not stratified into subtypes. The time to treatment was controlled by matching; this, however, decreased the final cohort size. ROI placement was performed by a single experienced operator but no estimation of the intra-observer reproducibility of this procedure was made. Our results represent only a single-institution experience.

In conclusion, our analysis indicates that infarct-core CTPPs - low rCBF in particular - can predict HT in patients with AIS. Should this be further verified by larger multi-centre studies, CTP imaging could become the method of choice for identification of patients at low risk of HT, thus helping decide on IVT treatment.