A retrospective analysis of a newly proposed imaging-etiologic classification for acute ischemic stroke with large vascular occlusion based on MRI and pathogenesis

Introduction

Acute ischemic stroke (AIS) is the most prevalent type of stroke, comprising approximately 69.6–70.8% of all stroke cases in China (Wang et al., 2017a, 2017b). Currently, endovascular treatment (EVT) bolstered by the outcomes of five pivotal randomized controlled trials-MR CLEAN, ESCAPE, EXTEND-IA, SWIFT PRIME and REVASCAT (Berkhemer et al., 2015; Campbell et al., 2015; Goyal et al., 2015; Jovin et al., 2015; Saver et al., 2015), has emerged as the frontline therapeutic option for AIS patients with large vascular occlusion (LVO). The classification of AIS significantly influences patient management, prognostic assessment, secondary stroke prevention strategies, and stroke research endeavors. Currently, the most popular stroke classification system is the TOAST criteria, which provides a theoretical basis for different treatments. However the system does not specifically cater to EVT. Chinese researchers have attempted to devise typing systems based on digital substraction angiography, yet these classifications are often intricate, hindering their widespread adoption and application. With the extension of therapeutic time window, the rates of AIS patients who can be candidates for emergency EVT will increase significantly (Albers et al., 2018; Casetta et al., 2021; Nogueira et al., 2018). Hence, the ability to swiftly, succinctly, and effectively leverage the preoperative data (MRI, medical history, etc.) to identify different types of AIS, guide the EVT and expedite the vascular recanalization time is of great importance. This imperative underpins the inception of our study.

Methods

Imaging-etiologic classification of AIS with LVO

The imaging-etiologic classification of AIS with LVO was based on MRI and pathogenesis, and estimated by comparison with the gold standard intra-operative finding of EVT. All image assessments were performed by two independent neurointerventionists before EVT, and discrepancies were resolved by consensus.

Our classification consists of two main components: imaging subclassification and etiologic subclassification, as detailed in Table 1. The imaging subclassification aims at locating the vascular occlusion or severe stenosis, and is based on preoperative MRI which requires at least two sequences, including DWI and MRA. According to the position of large vascular occlusion, anterior circulation infarction (ACI) and Posterior circulation infarction (PCI) were stratified into four categories respectively (A-D & a-d) (Figs. 1, 2).

Table 1:

Imaging-etiologic classification.

		Type	Definition	Segment	DWI	MRA (Disappear)	Major etiology	Other etiology
Image	ACI	A	Distal M1	About distal third of M1 (Including bifurcation and M2)	Infarction in cortex, corona radiata and borderzone area	MCA or terminal ICA + MCA
		B	Mid-proximal M1	About proximal two-thirds of M1	Basal ganglia infarction	MCA or terminal ICA + MCA
		C	Intracranial ICA	C6-C7	PLIC infarction or infarction in ACA＆MCA simultaneously	ICA or intracranial ICA or ICA + MCA + ACA
		D	Extracranial ICA	C1-C5	PLIC evasion	ICA or ICA + MCA + ACA
	PCI	a	Distal BA	Distal BA (Including SCA)	Superior cerebellar peduncle level of pons, midbrain, thalamus, occipital lobe infarction or infarction in SCA	Mid-distal BA
		b	Mid-proximal BA	From initial BA to SCA (exclusive)	Pons infarction or infarction in AICA	VBA or junction of VBA
		c	Intracranial VA	V4	Medullary infarction or infarction in PICA	VBA or junction of VBA
		d	Extracranial VA	V1-V3	The same as Type-a or large amounts of infarctions in one side	VBA or mid-distal BA+ partial signal of ipsilateral VA
Etiology		1	LAA				Risk factors for atherosclerosis	Age ≥ 40
		2	Thrombosis		Multiple infarctions in bilateral anterior, anterior and posterior circulations		Cardiogenic embolism	Risk factors for HCS, age < 40
		3	Dissection				Atherosclerosis	Trauma, Iatrogenic, etc.

DOI: 10.7717/peerj.18342/table-1

Anterior circulation infarction (ACI) (Fig. 3):

Type-A is defined as occlusion of the distal middle cerebral artery (MCA, M1), specifically the distal one- third of M1, including the M1 bifurcation and M2; DWI shows that the infarctions locate in the cortex, corona radiata and borderzone area. The MCA or terminal internal carotid artery (ICA) disappears on MRA;

Type-B is defined as occlusion of the mid-proximal M1, covering approximately the proximal two-thirds of M1. DWI shows isolated or combined infarctions in the basal ganglia region (such as the caput nuclei caudati, lenticular nucleus, anterior limb or genu of internal capsule infarctions). The MCA or terminal internal carotid artery (ICA) disappears on MRA;

Type-C is defined as occlusion of the intracranial ICA (bounded by the ophthalmic artery), from the ophthalmic segment to bifurcation of the ICA (i.e. C6-C7, based on the ICA classification proposed by Bouthillier). The DWI is characterized by posterior limb of internal capsule (PLIC) infarctions. Another characteristic of DWI is the simultaneous occurrence of infarctions in the territories supplied by the anterior cerebral artery (ACA) and MCA. MRA shows that the ICA or intracranial segment of the ICA, or the ICA combined with the MCA and ACA, disappears.

Type-D is defined as occlusion of the extracranial ICA, from the initial ICA to clinoid segment of the ICA (i.e. C1-C5). DWI shows there is no infarction related to the PLIC, which is described as PLIC evasion. MRA shows that the ICA or ICA combined with the MCA/ACA disappears.

Posterior circulation infarction (PCI) (Fig. 4):

Type-a is defined as occlusion of the distal basilar artery (BA), including the superior cerebellar artery (SCA). DWI shows that acute infarctions occur solely in the superior cerebellar peduncle level of the pons, midbrain, thalamus, occipital lobe and/or the territory supplied by the SCA. MRA reveals that the mid-distal basilar artery disappears;

Type-b is defined as occlusion of the mid-proximal BA, from the origin of the BA up to but excluding the SCA. DWI shows that the infarctions involve the pons (excluding the superior cerebellar peduncle level)or regions supplied by the anterior inferior cerebellar artery (AICA). MRA indicates that the entire segment or the junction of the vertebral-basilar artery (VBA) disappears;

Type-c is defined as occlusion of the intracranial segment of vertebral artery (VA), specifically the fourth segment of VA (i.e. V4). DWI shows that the infarctions involve the medulla or the area supplied by the posterior inferior cerebellar artery (PICA). MRA shows the entire segment or the junction of the vertebral-basilar artery (VBA) disappears;

Type-d is defined as occlusion of the extracranial VA, from V1 to V3. The DWI findings for type-d are either similar to those of type a or show large amounts of infarctions localized to one side of the posterior circulation. This is because large thrombi dislodged from the extracranial VA are influenced solely by the flow of the contralateral VA. MRA reveals that the VBA completely disappears, or the mid-distal BA and ipsilateral VA lose partial signal.

In the PCI classification, the vessel of type-a is so short that atherosclerosis or dissection is not considered to occur in this section solely.

Etiologic classification:

Etiologic classification, which is divided into three types (1–3), is based on the pathogenesis of AIS. The patients’ medical history (including hypertension, diabetes, cardiopathy, tumor, smoking and so on) and electrocardiogram are analyzed to determine the pathogenesis of AIS. Patients with dissecting stroke often present with the typical characteristic of tearing pain in the head and neck, and patients with this characteristic should be given priority consideration for type 3. Among stroke patients exhibiting atrial fibrillation, atrial fibrillation can account for approximately 79% of the causes of stroke. Thus, when multiple determinants are present at the same time, the priority of etiologic classification should be type 3, then 2, and then 1.

Type-1 is defined as large artery atherosclerosis (LAA), with or without thrombosis. Hypertension, diabetes, hyperlipidemia and smoking are the major risk factors for LAA; Patients who are over 40 years old but deny having any illness can also be classified into this type.

Type-2 is defined as thrombotic stroke. Cardiac disease, such as atrial fibrillation, atrial flutter and mitral valvular disease, is the most common risk factor causing thrombotic stroke. The other risk factors that have potential evidence for embolism and result in hypercoagulable state (HCS) include malignancy, bacteremia, septicemia, heavy use of hormones and renal failure. The patients who are younger than 40 years and deny any illness can fall into this type too, as well as acute multiple infarctions involving bilateral anterior and/or anterior and posterior circulations;

Type-3 is defined as vascular dissection, with or without thrombosis. Patients usually have a history of head-and-neck tearing pain.

Results

In our analysis of preoperative MRI, medical history, and intra-operative findings for 184 consecutive subjects with AIS, we found ACI represented 146 cases, accounting for 79.3% (146/184) and PCI 38 cases, accounting for 20.7% (38/184). Using the intra-operative variables of EVT as gold standard, the overall positive rate of the imaging classification meeting the gold standard was 93.5% (172/184), the etiologic classification was 94.6% (174/184), and the imaging-etiologic classification was 90.2% (166/184).

In this study, all vascular dissections occurred in the extracranial segments of the ICA and VA; It was observed that type-b only had atherosclerotic occlusion, while type-c and type-d were also primarily atherosclerosis-based, accounting for 76.5% (13/17) and 85.7% (6/7) respectively; Type-d was totally located in V1 or V1+V2 segments; In type-A, thromboembolism was more common, accounting for 62.5% (20/32). However, the proportions of atherosclerotic and thrombotic occlusion in type-B were nearly the same; Thrombotic occlusion was dominant in type-C, accounting for 75.9% (22/29); Type-D was predominantly caused by atherosclerosis, accounting for 65.8% (25/38), and mainly involved the initial and cervical ICA (that is C1), accounting for 78.9% (30/38).

Each type of the imaging-etiologic classification was evaluated separately using a screening test with the gold standard, and the results demonstrated its excellent validity (Sensitivity ≥ 75%, Specificity ≥ 90% and Youden’s index ≥ 0.75) and reliability (Kappa ≥ 0.80) (Table 3).

Discussion

AIS stands as one of the numerous public health issues that plague the world, frequently contributing to death and disability. The TOAST classification system, a categorization framework used to evaluate stroke types and determine their underlying causes, categorizes strokes into five main categories: large artery atherosclerosis, cardioembolic stroke, small artery occlusion, strokes of other determined etiology, and strokes of undetermined etiology. Presently, classification systems like TOAST, rooted in etiology and pathogenesis, offer a theoretical basis for selecting distinct treatment protocols and medications in clinical practice. However, their directive significance for endovascular treatments remains somewhat limited. The newly conceived classification system, with a primary emphasis on evaluating large vessel occlusive strokes, endeavors to pinpoint the location and etiology of vessel occlusion, thereby facilitating targeted thrombectomy procedures. Nowadays, EVT for AIS patients with LVO is the recommended first-line treatment (Powers et al., 2018). In our center, MRI is the routine examination item for assessing AIS patients who have no contraindication to MRI and can cooperate during the examination. MRI requires at least six sequences, including T2-weighted, FLAIR, DWI, ADC, TOF-MRA and SWI. Changes in DWI and ADC can be identified even within minutes of stroke onset, making them more sensitive than CT in the first 6 h from onset (Nael et al., 2014). In the development of parenchymal edema, however, FLAIR changes may not be apparent until several hours from onset. This DWI/ADC-FLAIR mismatch has been utilized to estimate the time of stroke onset, and to assess thrombolysis or thrombectomy in patients within the unwitnessed or extended time window (Thomalla et al., 2011). The slow flow in arteries often results in complete loss of signal in TOF-MRA, which demonstrates intracranial vessels without using contrast agent. Hence, TOF-MRA may underestimate flow in distal branches, overestimate arterial stenosis or occlusion severity, and fail to fully estimate vascular status (Boujan et al., 2018). Although CTA and contrast-enhanced MRA can show occlusive vessels more clearly, they are not routine examination of AIS in our center due to objective reasons. Therefore, it is often necessary to make treatment decisions after MRI/CT examination in order to shorten the time before EVT. Generally, we use MRI to screen the candidates for thrombolysis, thrombectomy, craniotomy or conservative treatment, rather than to reveal the mechanism and precise position of vascular occlusion, which can only be obtained from procedures of EVT. This potentially increases the risk of EVT, extends the recanalization time and even causes serious consequence by choosing an improper EVT program. For AIS patients with LVO, the shorter the vascular recanalization time, the greater the benefit. We proposed a new imaging-etiologic classification based on MRI and pathogenesis to locate the vascular occlusion and reveal its pathogenesis as quickly as possible. This classification can guide neurointerventionists in choosing the appropriate EVT program and shortening the vascular recanalization time.

In the new system, our imaging-etiologic classification achieved an overall positive rate of 90.2% (166/184) when compared to the gold standard. The screening test of each type demonstrated excellent validity and reliability: Sensitivity ≥ 75%, Specificity ≥ 90%, Youden’s index ≥ 0.75 and Kappa ≥ 0.80. When the Youden index is ≥0.75, it indicates that the diagnostic test achieves a good balance between sensitivity and specificity, enabling accurate identification while effectively reducing misdiagnoses and missed diagnoses, thereby improving the accuracy and reliability of the diagnosis. When Kappa is ≥0.80, it signifies a very high level of agreement among raters, and their judgment results can be considered highly reliable. In this study, we found ACI accounted for 146, 79.3% (146/184) and PCI accounted for 38, 20.7% (38/184), that was consistent with the results of other studies in China. All of the vascular dissection occurred in the extracranial sections of the ICA and VA, which were the most predisposed sites. Additionally, thromboembolism frequently manifested in vascular bifurcations and Coarctation, as evident from our statistical data. Remarkably, throughout the study period, we did not encounter any adverse events associated with the imaging procedures or the endovascular treatments administered. These reassuring data underscore the credibility and suitability of the newly proposed imaging-etiologic classification system for clinical applications.

However, the real situation of AIS is occasionally complex, with occurrences such as thrombus migration and tandem occlusion. In order to fully uncover all the major occlusions, we must not confine ourselves solely to subclassifying the primary occlusion. For example, fragments of plaques or thrombi originating from the initial part of the ICA may dislodge and travel to distal branch, resulting in M1 embolism. In such cases, relying solely on type D-1 does not adequately capture the entire process, so we give D-1→A as a supplement; Similarly, Cardiac-origin thrombi may dislodge and travel to the distal ICA and proximal MCA, leading to a combined ICA and M1 embolism. Here, we give C/B-2 as a supplement to C-2. In the statistical analysis of data, these two scenarios are counted as D-1 and C-2 respectively, with other similar cases being treated analogously.

Vasculopathy in AIS is a intricate and dynamic process, and no relatively simple grading can accurately classify all the variations of such complex vascular lesions. For example, Cardiac-originated thrombi lead to occlusion of all intracranial and extracranial segments of the ICA, which will be wrongly classified as C-2 instead of D-2 (D/C-2) according to the imaging-etiologic classification system. The D-2 (D/C-2) typing indicates a higher thrombus load, which affects the vascular recanalization time and patients’ prognosis. The anatomical variants, collateral circulation, and thrombus migration can make the interpretation of DWI lesions very challenging. In this classification system, the accuracy and consistency are the highest for type A and type c, whereas they are relatively low for type C and type a. This is primarily due to the fact that a larger thrombus load and thrombus displacement can easily lead to misdiagnosis. However, currently, there is a lack of rapid examination methods to accurately determine the thrombus load in emergency situations. In the future, we can explore quick thrombus load examination methods to improve the accuracy of this classification. In spite of occasional misclassifications of the assigned grade, we still believe that the rapidity and simplicity of the classification system have great advantages. To take into consideration every possible anatomic and etiologic permutation of AIS, the system would have to be overly complex, making it impractical. This classification system which has demonstrated its validity and reliability, is simple, easy to learn, and easy to apply.

Some of our own suggestions for applying this classification system are as follows: 1, For large artery atherosclerosis, we propose to predict the thrombus load using the microcatheter “First-Pass Effect” first (Yi et al., 2019), then carry out mechanical thrombectomy with a stent to reduce the thrombus load. If there is higher residual vascular stenosis, perform angioplasty finally; 2, If it is defined as type-2, the thrombotic stroke, we prefer direct aspiration first-pass thrombectomy (ADAPT) and solitaire with the intention for thrombectomy (SWIM), which have been supported by many studies (Lapergue et al., 2021, 2017; Turk et al., 2019); 3, Unless the vascular dissection affects hemodynamics, we tend to select conservative treatment over EVT; Reducing the thrombus load and finding the true lumen as soon as possible are the top priorities of EVT for type-3; 4, For strokes identified as type-A, B, a, and b, ensure an adequate length and preferable diameter of the intermediate catheter (>115 cm, ≤5F); 5, As type-C is mainly thrombotic infarction, we recommend a large-diameter intermediate catheter (6F) to facilitate thrombus aspiration; 6, type-D is predominantly caused by atherosclerosis and often involves thrombus migration. Therefore, the balloon assisted track (BAT) technique or using a balloon guiding tube is a good choice; For cases with tandem occlusion due to thrombus migration, it is controversial whether to preferentially treat extracranial or intracranial lesions (Wilson et al., 2018); In our center, we primarily rely on the individual experience of the doctors.

As discussed above, the two parts of this classification are not independent of each other but require comprehensive consideration to assist the neurointerventionist in determining the main locations and etiologies of vascular occlusion and selecting the most appropriate solution.

Conclusions

Our pilot study results show that our proposed imaging-etiologic classification can simply, rapidly and effectively reveal the location of vascular occlusion, the pathogenesis of AIS with LVO, and be used by interventionists to guide the choice of the appropriate EVT program.

Supplemental Information