Application of Neuroimaging for the Prediction of Hemorrhagic Transformation after Intravenous Thrombolysis in Acute Ischemic Stroke Free

Stroke is the primary cause of long-term disability and the second leading cause of mortality worldwide, with more than 13.7 million people experiencing a stroke annually in the USA and approximately 87% of stroke cases being ischemic stroke [1‒4]. Ischemic stroke, a common cerebrovascular disease, is a sudden occlusion of the cerebral artery due to a thrombus or embolus that results in hypoxia, ischemia, and loss of glucose in the local cerebral tissue [5]. A series of pathological cascade events, such as excitotoxicity, neuroinflammation, and oxidative stress, are activated following brain ischemia, which in turn contributes to blood-brain barrier (BBB) dysfunction, vasogenic edema, intracranial hemorrhage, and neuronal death [5‒7].

Currently, recombinant tissue plasminogen activator is the main intravenous thrombolysis drug licensed by the Food and Drug Administration to treat patients with acute ischemic stroke within 4.5 h after onset. However, only a few patients have access to receive intravenous thrombolysis because of the time window limitation, and delayed treatment can significantly increase the risk of hemorrhagic transformation (HT) [8]. HT is the most serious complication after intravenous thrombolysis that can aggravate clinical poor prognosis [9]. Currently, there are no effective therapeutic drugs to prevent the incidence of HT. Therefore, it is important to early predict the risk of post-thrombolysis HT in patients with acute ischemic stroke. Previous research studies showed that the predictive value for HT is mainly based on clinical data and serum biomarkers. Recently, neuroimaging techniques, such as multimodal computed tomography (CT) [10, 11] and magnetic resonance imaging (MRI) [12, 13], have been valuable in predicting HT after intravenous thrombolysis. Imaging parameters such as cerebral blood flow (CBF) or cerebral blood volume (CBV), apparent diffusion coefficient (ADC) value, and permeability surface (PS) might be potential predictors of HT after intravenous thrombolysis [10, 11, 13].

Hence, recognizing and understanding the predictive performance of neuroimaging parameters may help assess the risk of HT after intravenous thrombolysis in patients with acute ischemic stroke. In this review, we mainly focus on the advances in the current application of neuroimaging in predicting HT after intravenous thrombolysis.

Radiographic classification of post-thrombolysis HT had distinguished between hemorrhagic infarction and parenchymal hematoma (PH) according to the European Cooperative Acute Stroke Study (Table 1) [14]. HT is also divided into asymptomatic intracranial hemorrhage and symptomatic intracranial hemorrhage (SICH) based on the clinical deterioration or not. SICH is defined as hemorrhage in any part of the brain within the first 24 h after thrombolysis, accompanied by worsening of clinical symptoms or an increase of >4 points in the National Institutes of Health Stroke Scale score [15].

Non-contrast CT is the preferred method to assess intracranial hemorrhage in stroke patients. Previous studies indicated that the hyperdense middle cerebral artery sign was the independent predictor of HT after intravenous thrombolysis [16‒18]. A lower Alberta stroke program early CT score (ASPECTS), especially a score of ≤7 (sensitivity 66% and specificity 92%), was significantly associated with an increased risk of HT after intravenous thrombolysis [19, 20]. Therefore, the combination of the hyperdense middle cerebral artery sign and ASPECTS can be used to assess the risk of HT after intravenous thrombolysis.

In addition, single-photon emission CT (SPECT) was previously reported as an imaging method for predicting HT after intravenous thrombolysis. The lower baseline CBF on SPECT reflected severe hypoperfusion in the ischemia region, which was associated with a high risk of HT after intravenous thrombolysis [21‒23]. Even though SPECT may be an ideal method for evaluating CBF by obtaining semiquantitative hemodynamic data after acute ischemic stroke, this imaging model is not routinely used on patients with stroke because of its high radiation.

CT perfusion (CTP) was first proposed by Axel [24] in 1980s, which was not widely used due to the slow CT acquisition and post-processing systems at that time. With the development of technology, it has been gradually applied in current clinical practice. CTP can obtain cerebral tissue/capillary-level hemodynamics parameters and form cerebral perfusion maps by injecting a contrast agent. The main perfusion parameters include CBF, CBV, mean transit time (MTT), time to maximum (Tmax), and PS [25‒27].

CTP plays a critical role in determining “infarct core” and ischemic “penumbra” via the quantitative evaluation of perfusion parameters [25, 28]. The measurement of the perfusion parameters CBF and CBV is considered suitable for assessing the optimal infarct core threshold [28‒31]. While the optimal penumbra threshold was reported to exist on MTT or Tmax maps, a Tmax of >6 s is the optimal penumbra threshold for CTP [28, 31, 32].

Multiple research studies also showed that CTP can predict the risk of post-thrombolysis HT [11, 26, 27, 33‒35]. Adebayo et al. [35] revealed that the sensitivity and specificity of CTP in predicting HT in patients with acute ischemic stroke were 85.9% and 73.9%, respectively. Relative CBF is an effective CTP parameter for predicting HT, in which reduction is significantly related to a higher risk of HT after intravenous thrombolysis [26, 34]. The ischemia-hypoperfusion area with relative CBF ≤0.48 and relative MTT ≥1.3 was significantly related to post-thrombolysis HT [36]. Others believed that lower relative CBV has a better predictive value for HT [26, 27]. However, a significant difference was noted in PS between patients with hemorrhage and those without hemorrhage but not in CBF or CBV [37, 38]. Several researchers were interested in studying the relationship between the PS or contrast volume transfer coefficient (K^trans) and HT in acute ischemic stroke [10, 11, 37‒41]. The PS refers to the rate of contrast agent extravasation into the extravascular space through the damaged BBB [42]. An increased PS or K^trans value reflects the elevation of BBB permeability, which tends to occur in patients with HT after intravenous thrombolysis [11, 40]. Ozkul-Wermester et al. [43] reported that the infarct PS was an independent predictor of post-thrombolysis HT (odds ratio 28, 95% confidence interval: 1.75–452.98; p = 0.02). Li et al. [38] showed that a PS threshold of 1.179 mL/100 g/min is the optimal value for predicting HT in the deep middle cerebral artery. A recent study indicated that a PS threshold of 0.94 mL/100 g/min had a sensitivity of 95.5% and specificity of 78.6% in predicting post-thrombolysis HT [10]. Hence, a higher PS appears to be the promising predictor of HT in acute ischemic stroke, and so did higher K^trans values. A cutoff value for K^trans of 0.35/min promotes the occurrence of HT after thrombolytic therapy in acute ischemic stroke [39]. Chen et al. [41] also revealed that a similar K^trans cutoff value (0.334/min) had a sensitivity of 95% and specificity of 75% in predicting HT, respectively. Additionally, the relationship between Tmax and HT has been investigated by Yassi et al. [34] The result showed that a Tmax value of >14 s was the independent predictor of parenchymal hemorrhage, with a sensitivity of 79% and specificity of 68%. CTP parameter thresholds are not consistent in different research studies, which may be related to various factors, such as type of study, patient selection, sample size, and CTP imaging time. Therefore, further and larger studies are needed to validate the diagnostic performance of CTP and calculate the optimal parameter threshold for predicting HT after intravenous thrombolysis.

In recent years, CT angiography (CTA) has been widely used in many stroke centers. Because of the limitation of the treatment time window, neurologists need to make a quick decision on the treatment option for patients with acute ischemic stroke. CTA can provide information on cerebral vascular anatomy and the site of arterial occlusion. Two score systems, clot burden score (CBS) and collateral score (CS), can be collected from CTA imaging and play a vital application value in predicting HT after acute ischemic stroke [10, 44, 45].

The CBS is mainly administrated to reflect the extent and location of the thrombus in the anterior circulation vessels and assess scores according to the presence or absence of a contrast agent on CTA images. The CBS ranges from 0 to 10 (Table 2). A score of 0 indicates the anterior circulation vessels are complete occlusion, while a score of 10 indicates the anterior circulation vessels are normal. Higher CBSs indicate lower clot burden, higher likelihood of vascular recanalization after reperfusion therapy, and better clinical and functional outcomes [44‒46]. Several studies reported that the CBS may be a predictor of HT after acute ischemic stroke [40, 45, 47, 48]. For example, a CBS of ≤3 significantly increases the risk of HT after endovascular therapy [45], while a higher CBS is associated with a lower likelihood of HT development [48]. Recently, a meta-analysis suggested that CBS is the predictor of vascular recanalization and clinical functional outcomes after reperfusion therapy in patients with acute ischemic stroke [47].

The CS is another score system that can provide collateral supply information in occluded anterior circulation territory. The criterion of CS is shown in Table 3. A lower CS indicates a worse collateral supply in the infarct area, while a higher CS reflects a good collateral status [44]. Moreover, collateral circulation status may be related to the development of HT after acute ischemic stroke. Studies have reported that patients with poor collateral supply have an increased risk of the development of post-thrombolysis HT [10, 49]. Feng et al. [50] suggested that a lower collateral status is closely related to the increased risk of HT after mechanical thrombectomy. Furthermore, patients with poor baseline collaterals were associated with not only a higher risk of post-thrombolysis SICH but also in-hospital mortality and early functional outcome deterioration [51].

Although CT has long been considered the fastest method for detecting acute intracranial hemorrhage, MRI was also reported as the method used to assess HT after intravenous thrombolysis. MRI contains multiple sequences, including T1-weighted image, T2- or T2*-weighted image, fluid-attenuated inversion recovery (FLAIR), diffusion-weighted imaging, and perfusion-weighted imaging. Partial MRI parameters have a potential predictive value for HT after intravenous thrombolysis [52‒55].

T2*-weighted sequences of MRI, especially gradient recall echo (GRE), are extremely sensitive to iron-contained complexes, which have a similar effect in detecting early cerebral hemorrhage as a CT scan [56, 57]. Meanwhile, GRE was used to assess the post-thrombolysis HT by detecting cerebral microbleed (CMB) [54, 56]. The relationship between CMB and HT remains controversial. Earlier studies suggested the CMB was considered as the independent predictor of HT [58, 59]. Bokura et al. [60] conducted a longitudinal study based on healthy elderly individuals and found that CMB was a risk factor for intracerebral hemorrhage after intravenous thrombolysis. Additionally, patients with high CMB burden (>10 CMBs) were tightly associated with post-thrombolysis SICH [61‒63]. Contrary to the aforementioned results, several studies considered that CMB may be independent of HT. Two multicenter studies reported that CMB on the T2*GRE sequence did not appear to increase the risk of post-thrombolysis HT [64, 65]. Takahashi et al. [66] suggested HT is not related to CMB but the infarct size or neurological deficit. Wang et al. [67] also demonstrated that there was no relationship between CMB and HT. Therefore, although the use of GRE sequence can help detect HT, further research studies are needed to validate the predictive performance of CMB for HT after intravenous thrombolysis.

FLAIR is one of the MRI sequences that can contribute to determine the onset time of patients with acute ischemic stroke [68, 69]. Similarly, FLAIR plays a positive role in predicting post-thrombolysis HT. It is understood that FLAIR hyperintensity is the net increase in water content in infarcted tissues caused by BBB disruption after acute ischemic stroke [70, 71]. Cho et al. [72] found that the location of FLAIR hyperintensity in the cerebral ischemia region may co-localize with post-thrombolysis SICH. The result indicated that FLAIR hyperintensity may be the predictor of post-thrombolysis SICH after acute ischemic stroke. A similar result was reported in the following studies. Kufner et al. [73] revealed that the early development of FLAIR hyperintensity in the cerebral infarcted area is related to post-thrombolysis HT. A retrospective study on 134 patients with acute ischemic stroke showed that FLAIR hyperintensity significantly increased the risk of HT after intravenous thrombolysis [71]. Jha et al. [70] found that FLAIR hyperintensity is not only associated with HT but also related to higher MMP-9 levels after intravenous thrombolysis. However, the relationship between FLAIR hyperintensity and HT is not confirmed in all studies. Campbell et al. [74] reported that FLAIR hyperintensity is prevalent in patients within 3–6 h of stroke onset that could not predict HT. The reason for this result may be the small sample size of the study, while the possibility of FLAIR hyperintensity for predicting post-thrombolysis HT could not be ruled out. Therefore, further research studies are needed to identify the predictive performance of FLAIR hyperintensity for HT after intravenous thrombolysis.

The hyperintense acute reperfusion marker (HARM) on FLAIR sequence has been reported as an indicator of early BBB disruption that was the critical mechanism of HT after intravenous thrombolysis [75, 76]. The relationship between HARM and HT remains controversial. In several studies, patients with HARM had a higher risk of post-thrombolysis HT than those without HARM, which indicated HARM was significantly associated with HT [77, 78]. However, others did not find the relationship between HARM and post-thrombolysis HT [79, 80]. Additionally, the white matter hypoperfusion on FLAIR imaging reflected severe leukoaraiosis, which was significantly associated with post-thrombolysis HT. Recent studies showed that the white matter hypoperfusion may be related to the increased risk of HT after intravenous thrombolysis [81‒84]. Hence, leukoaraiosis may be an independent predictor of HT risk after intravenous thrombolysis.

Diffusion-weighted imaging (DWI) is particularly sensitive to brain ischemia and conducive to the evaluation and diagnosis of acute ischemic stroke [85, 86]. The ADC value extracted from DWI is an indicator for calculating the brain infarction volume and assessing the extent of brain ischemic lesions [87]. The combination of DWI and ADC imaging can distinguish the infarcted core from the ischemic penumbra [87, 88], which has similar accuracy with CTP in predicting post-ischemia HT [36]. Experimental stroke models and clinical studies observed that the ADC value <550^×10⁻⁶ mm²/s is closely correlated with the occurrence of HT after brain ischemia or thrombolytic therapy [13, 89‒91]. Relative ADC ratio is defined as the ratio of ADC pixel values in the infarcted area to the corresponding ADC pixel values in the contralateral normal brain area. Shinoda et al. [92, 93] conducted single-center retrospective studies that showed a relative ADC ratio of <0.6 may be predictive of post-thrombolysis HT. Similarly, Liu et al. [94] considered that a relative ADC ratio may be a reliable predictor of HT after acute ischemic stroke. Moreover, a DWI-FLAIR mismatch was reported to play a positive role in making a thrombolysis decision for stroke patients with unknown time of onset or wake-up stroke, and intravenous thrombolysis may be safe and effective in these patients [95, 96]. The DWI-FLAIR mismatch refers to an acute ischemic lesion on DWI without marked hyperintensity in the same area on FLAIR [95]. Thomalla et al. [97] showed a DWI-FLAIR mismatch had a sensitivity of 62% and specificity of 78% in identifying patients with stroke within 4.5 h of symptom onset. Recently, a single-center retrospective study reported that a DWI-FLAIR mismatch may be correlated with a lower risk of SICH and better functional outcomes in patients with mechanical thrombectomy [98]. Therefore, DWI may have a certain predictive value for post-thrombolysis HT.

Apart from the above sequences, perfusion-weighted imaging (PWI) is also applied in some clinical stroke centers to predict post-thrombolysis HT risk and provide imaging evidence for intravenous thrombolysis in patients beyond the time window [12, 99, 100]. PWI is the same as CTP that can be used to reflect the information on cerebral tissue perfusion. Hemodynamic parameters, including CBV, CBF, MTT, and BBB permeability maps, are obtained from PWI images [101, 102]. A recent meta-analysis showed that PWI-based high BBB permeability and low CBV were related to HT [55]. Alsop et al. [103] also reported that a low CBV may be a better predictor of HT after intravenous thrombolysis in patients with acute ischemic stroke. A very low CBV value indicates severe brain ischemia or hypoperfusion and may be used to stratify risk in patients who will undergo endovascular therapy [104, 105]. Studies suggested that a very low CBV is closely correlated with PH [91, 104‒106]. Campbell et al. [91] described that a region of a very low CBV has a better predictive performance of PH after intravenous thrombolysis than DWI or ADC. A relative CBV ratio of <0.42 and a region of a very low CBV of ≥3.55 mL may independently predict PH, which was demonstrated by Mishra et al. [105]. Additionally, a Tmax of >14 s derived from PWI images was also considered as the optimal threshold for predicting post-thrombolysis PH in patients with acute ischemic stroke [106]. However, PWI is rarely used to evaluate cerebral perfusion in the emergency department because it takes a long time to complete one scan.

This review mainly introduces the application of neuroimaging in predicting HT after intravenous thrombolysis. It is understood that neuroimaging techniques, such as CTP, CTA, FLAIR, DWI, and PWI, may be used to assess the risk of HT after intravenous thrombolysis. Imaging parameters, including CBF or CBV, ADC value, or DWI-FLAIR mismatch, may be potential predictors of HT after intravenous thrombolysis. The clinical application of neuroimaging techniques and the predictive performance of each imaging parameter for predicting HT are shown in Table 4. In summary, a specific neuroimaging technique and the optimal cutoff value of its parameters can be used to guide the decision-making of intravenous thrombolysis in patients with acute ischemic stroke.

The authors have no conflicts of interest to declare.

The authors did not receive any funding.

Miaomiao Yang participated in literature retrieval and drafted the manuscript. Lisha Tang participated in literature retrieval and prepared the tables. Zhiping Hu revised the manuscript. Xiangqi Tang designed the study, helped with summarizing the manuscript, and contributed to the manuscript revision.