Prognosis of ischemia recurrence in patients with moderate intracranial atherosclerotic disease using quantitative MRA measurements

Abstract

Purpose:

To investigate the relation between delayed ischemic stroke and the intracranial atherosclerotic disease (ICAD) hemodynamics as determined by Non-invasive Optimal Vessel Analysis (NOVA) MRI measurements.

Materials and Methods:

Thirty-three patients with ICAD were enrolled in this study. All patients underwent clinically indicated angioplasty followed by 2-dimensional phase contrast MR (2D PCMR) performed on a 3.0 Tesla MRI scanner using either a 16-channel neurovascular coil or 32-channel head coil. The volumetric flow rate measurements were calculated from 2D PCMR with Non-invasive Optimal Vessel Analysis (NOVA) software (VasSol, Chicago, IL, USA). Flow rate measurements were obtained in 20 major arteries distal, proximal and within the Circle of Willis. Patients were followed up for six month, and ischemia reoccurrence and location were recorded. Receiver operating characteristic (ROC) analysis was performed using flow rates measurements in the ipsilateral side of the ischemic event occurrence.

Results

Complete set of measurements was achieved in n=34. Left and right hemisphere ischemia recurrence was observed in seven and three cases respectively. Best predictor of ischemic event reoccurrence was flow rate in the middle cerebral artery with area under the ROC of 0.821±0.109.

Conclusions:

This is an effectiveness study to determine whether blood flow measurements in the intracranial vasculature may be predictive of future ischemic events. Our results demonstrated significant correlation between the blood flow measurements using 2D PCMR processed with the NOVA software and the reoccurrence of ischemia. These results support further investigation for using this method for risk stratification of ICAD patients.

1. INTRODUCTION

Atherosclerotic stenosis in the major intracranial arteries is one of the most common causes of stroke worldwide [1]. About 20 to 40 out of every 100,000 people worldwide are estimated to suffer an ischemic event related to atherosclerosis in intracranial arteries [2]. In the United States, 8% to 10% of ischemic strokes are a result of intracranial atherosclerosis. Furthermore, the condition also poses a 15% risk of recurrent stroke in affected patients [3]. In particular, intracranial atherosclerotic disease (ICAD) is historically seen to be more prevalent in the African American, Asian, Hispanic, and Indian populations when compared to the Caucasian population. In fact, while ICAD is the cause of only about 5–10% of strokes in white people, 15–29% of the transient ischemic attacks or strokes in African American people and up to 30–50% of the strokes in Asian people are a result of ICAD. While the direct cause of this disproportionate distribution of intracranial atherosclerosis is yet to be conclusively determined, some studies attribute the high prevalence in African American populations to higher hypertension, diabetes mellitus, and hyperlipidemia [4]. Meanwhile, genetic and environmental factors appear to play a bigger role in the development of the condition in Asians [4]. Moreover, intracranial artery examination during autopsy revealed severe intracranial atherosclerosis in 43% aged 60–69, 65% among 70–79 and 80% among those aged >=80 [4]. Early diagnosis would be crucial in the case of intracranial steno-occlusive diseases like intracranial atherosclerosis due to the inherent risk of stroke and the financial burden it would inflict. Once an occlusion becomes severe, external carotid-internal carotid (EC-IC) bypass surgery is often employed for revascularization [5, 6]. Therefore, the ability to establish a relation between a delayed ischemic stroke and intracranial atherosclerosis disease is essential.

Unfortunately, ICAD as a cause of stroke remains under-studied as compared to carotid stenosis and atrial fibrillation. Clinical evidence shows high-risk of recurrence or secondary stroke in patients with symptomatic ICAD; up to 12% of ICAD patients will experience a recurrent stroke within 1^st year of follow-up [7, 8]. In order to avoid such devastating consequences, neurointerventionalists and neurologists routinely assess the risk of recurrent stroke and its potential mechanism. Determining the lesion’s risk is essential for choosing the right therapy, which can include aggressive management of risk factors and medical treatment as a first treatment option, or endovascular repair, which has a higher risk and thus is an alternative treatment option reserved only for high-risk cases who do not respond to medical therapy [1, 4, 9].

Current diagnosis of ICAD severity for medical decision is guided by the percentage of vessel narrowing on 2D angiograms [10]. Studies performed for other disease states, such as coronary diseases, have shown that this metric does not accurately reflect the underlying hemodynamic impairment that can precede ischemia [11]. Functional information such as fractional flow reserve (FFR), given by the ratio of distal to proximal pressure, has been found to be a key complementary physiological measurement to identify patients at real risk of myocardial ischemia. Moreover, FFR has been shown to be more sensitive compared to stenosis measurement [12]. Since clinical use of minimally invasive pressure probes [11, 13] is unfeasible in the neuro space, non-invasive flow estimation using MRI could become an alternative solution.

In this paper, we aim to directly evaluate MRI-based cerebral hemodynamics measurements of intracranial stenosis and study its association with probability of recurrent stroke. We used intracranial vascular hemodynamics, measured by MRI, to estimate the odds of ischemic event reoccurrence in patients with intracranial atherosclerotic disease (ICAD).

2. MATERIALS AND METHODS

2.1. Study design

This study was approved by the IRB at University at Buffalo. Data acquired for this study involved patients with symptoms of intracranial atherosclerotic disease (ICAD). The study schema is displayed in Figure 1. ICAD patients underwent diagnosis with CT and MRI imaging followed by angioplasty. Next, they underwent MRI that included 2D PCMR to measure the flow rate in target vessels. From the total number of patients examined, 33 provided data that was usable for the study, with one patient having two distinct imaging sessions acquired two months apart. 10/34 studies showed recurrence of ischemia while 24/34 showed no recurrence. 7/10 had recurrence on the left side of the brain and 3/10 showed recurrence on the right side, respectively.

Figure 1:

Stepwise depiction of the project design.

In this study flow was evaluated in most patent arteries of the Circle of Willis. For reference the name of the arteries are presented in Table 1. The side of the brain is indicated with letters R or L for the right and left hemisphere respectively. The number at the end of the acronym indicates the branching level.

Table 1.

Artery names and their acronyms.

Acronym	Artery
ACA	Anterior Cerebral Artery
BA	Basilar Artery
EC-IC	External Carotid-Internal Carotid
ICA	Internal Carotid Artery
MCA	Middle Cerebral Artery
PCA	Posterior Cerebral Artery
PCOM	Posterior Communicating Artery
SCA	Superior Cerebellar Artery

2.2. Image data acquisition and NOVA velocity calculation

All subjects underwent qMRA on a 3.0 Tesla MRI Scanner (Philips, Best, Netherlands) using either a 16-channel neurovascular coil or 32 channel head coil. Intracranial arteries were first visualized using 3D Time-of-Flight MRA with the following parameters: TR/TE=25/3.5 millisecond; flip angle=25°; FOV=200 mm; slice thickness=1 mm; matrix=324×282. The volume flow rate measurements were calculated from 2D PCMR with the Non-invasive Optimal Vessel Analysis (NOVA. VasSol, Chicago, IL, USA) software (See Figure 2). The MRA images were sent to the NOVA software running on a separate workstation to reconstruct a three-dimensional surface-rendering of the vasculature used for automated determination of the double oblique scan plane perpendicular to flow of the vessels of interest. The 2D PCMR scans were cardiac-gated and had the following parameters: TR=10–15 ms; TE=4–7 ms; flip angle=15°, NSA=4; slice thickness=3 mm for intracranial arteries and 5 mm for neck arteries; FOV=140 mm for intracranial arteries and 180 mm for neck arteries; matrix=256 × 192 for intracranial arteries and 256 × 128 for neck arteries. This technique of flow measurement has been previously published [14].

Figure 2.

Schematic map shows volumetric flow rates obtained in individual vessels with NOVA from a patient with ICAD.

Blood flow measurement methodology has been previously described [15] and is briefly summarized here. Phase-contrast MRI is used for calculation of blood flow velocity. A bipolar gradient is applied to distinguish the moving protons in blood from stationary protons in other tissues. Due to the bipolar nature of the gradient, stationary protons do not acquire a phase shift as the opposing gradients will offset the phase shift each causes. Moving protons, which change location in the gradient unlike stationary protons, acquire differing amounts of phase shift based on their velocity. This bipolar gradient is applied during signal acquisition in the direction of blood flow [16].

The velocity of moving protons can be directly calculated from the amount of phase shift acquired from the bipolar gradient using the following equation:

$$\Delta \varphi =\gamma \overrightarrow{v}\Delta M_1$$ (1)

Where Δϕ is the phase shift, γ is the gyromagnetic ratio, $\overrightarrow{v}$ is the proton velocity, and ΔM₁ is the change in the magnetic moment. Solving for the proton velocity results in the following equation:

$$\overrightarrow{v}=\frac{\Delta \varphi }{\gamma \Delta M_1}$$ (2)

Due to the nature of angles where an angle above 360° or below −360° returns to 0°, aliasing occurs for angles greater than 360° or less than −360°. Also, a positive angle between 180° and 360° can give the same phase shift as an angle between 0° and −180°, and vice versa. For instance, a 190° phase shift gives the same result as a −170° phase shift. Therefore, only velocities that cause a phase shift between −180° and 180° can be measured correctly to allow for direction of flow to be known. Aliasing can be avoided by reducing the field strength of the gradient because this reduces the phase shift caused by a particular velocity, allowing for greater velocities to be measured [16].

NOVA uses the principles of phase-contrast MRI to calculate the velocity of blood in the vessels of the brain. The NOVA algorithm is able to do automatic contouring of the blood vessels using Time-of-Flight MRA. Each artery and vein of interest can be marked by the operator using this 3D model. Then, the NOVA algorithm finds a perpendicular scan plane to each vessel, and 2D phase-contrast imaging is used to calculate the velocities and estimate the cross-section of the vessel. The volumetric flow rate of the blood can be determined by multiplying the velocity and area as can be seen in the following Equation 3:

$$Q=vA$$ (3)

Where Q is the volumetric flow rate, v is the proton velocity, and A is the cross-sectional area of the vessel.

2.3. Data processing and interpretation

We attempted analysis in maximum 22 locations within the brain vasculature which included main arteries and in certain cases daughter branches. Complete data set of measurements was achieved on both the left (L) and right (R) sides only in five main arteries: Internal Carotid Artery (ICA), Middle Cerebral Artery (MCA), A1 segment of Anterior Cerebral Artery (ACA), A2 segment of Anterior Cerebral Artery (ACA2) and Posterior Cerebral Artery (PCA).

For each measurement we performed an adjudication study where we verified the conservation of mass at every bifurcation/trifurcation. We performed an error analysis for deviation from the conservation of the mass for the following arteries bifurcations/trifurcation: the RICA, LICA, ACA, and BA. The RICA bifurcation consists of RICA, RMCA, RACA, and Right Posterior Communicating Artery (RPCOM). The LICA bifurcation consists of LICA, LMCA, LACA, and Left Posterior Communicating Artery (LPCOM). The ACA bifurcation consists of RACA, LACA, RACA2, and LACA2. The BA bifurcation consists of BA, RSCA, LSCA, RPCA, LPCA, RPCOM, and LPCOM. First, we investigated the MRA images to determine whether one of the daughter branches was missed during the phase contrast. If the artery was missed due to small signal or stenotic vessel, we manually added the artery to the study, and the corresponding flow rate was derived using the conservation of mass principle. Each case was adjudicated by an experienced neuro-radiologist with five years of experience. Next, we performed an error analysis as to whether the data is within the 10% accuracy of the NOVA software for each of these. If the error was above 10% we investigated whether the automated cross section plane was properly placed. A comparison of the measurement plane being placed perpendicular versus non-perpendicular can be seen in Figure 3.

Figure 3.

An example of the volumetric flow rate being measured in the basilar artery, perpendicular to the vessel lumen (left) versus being measured non-perpendicular to the vessel lumen (right).

Flow estimation is accurate only if the cross-section plane is perpendicular to the vessel lumen. Overestimation of flow can occur due to the measurement plane not being perpendicular to the vessel lumen. In the patients with errors above 10% in mass conservation we investigated placement of the plane in the parent and daughter arteries. If the plane was incorrectly placed we applied a correction factor given by Equation 4.

$$\mathit{\text{Corrected}} \mathit{\text{flow}} = (\mathit{\text{Measured}} \mathit{\text{flow}}) \times cos(\mathit{\text{angle}} \mathit{\text{from}} \mathit{\text{horizontal}} \mathit{\text{to}} \mathit{\text{measurement}} \mathit{\text{plane}})$$ (4)

The data collected from the NOVA imaging analysis of the patients showing both recurrence and non-recurrence of ischemia were analyzed. The potential of using flow measurements in the Left and Right ICAs, MCAs, ACAs, ACA2s, and PCAs to predict ischemia recurrence was assessed using receiver operating characteristic (ROC) curves within SPSS.

3. RESULTS

3.1. Demographics

From the 44 cases originally chosen, 8 were excluded for insufficient MRA data, and 2 were excluded based on the judgment of a neuroradiologist. Additionally, one patient provided two unique cases for examination, so there were 33 patients overall and 34 unique studies. Of the 33 patients who were considered for the study, 14 self-reported as male and 19 were self-reported as female. The median age of the patients was 71 with an average age of 69.7 years and a standard deviation of 10.9 years. History of hyperlipidemia, hypertension, diabetes mellitus, atrial fibrillation and smoking were registered, along with the original ischemic event. Fifteen patients had diabetes mellitus, 24 had hypertension, 25 had hyperlipidemia, and 4 reported suffering from atrial fibrillation. From the study population, 17 patients reported a history of smoking, with 11 reported to have quit smoking while 6 were still active smokers. Two patients chose to not disclose their smoking habits. The demographics data can be seen in Table 2.

Table 2.

Summary of the demographics data.

Parameter	Condition	Number of Cases	Percentage
Diabetes Mellitus	Diabetic	15	45.5%
	Non-Diabetic	18	54.5%
Smoking Habits	Active Smokers	6	18.2%
	Former Smokers	11	33.3%
	Non-Smokers	14	42.4%
	Undisclosed	2	6.1%
Sex	Male	14	42.4%
	Female	19	57.6%
Hypertension	Yes	24	72.7%
	No	9	27.3%
Hyperlipidemia	Yes	25	75.8%
	No	8	24.2%
Atrial Fibrillation	Yes	4	12.1%
	No	29	87.9%
Presentation	Transient Ischemic Attack	12	36.4%
	Stroke	21	63.6%
Laterality	Left	14	42.4%
	Midline	4	12.1%
	Right	9	27.3%
	Bilateral	4	12.1%
	Undisclosed	2	6.1%

3.2. Flow measurement results

Each MRA data was assessed individually with the neuro radiologist. Four bifurcations were established for analysis: the RICA bifurcation, the LICA bifurcation, the ACA-ACA2 bifurcation, and the BA bifurcation. Mass balance was checked at these main bifurcations. To assess the effect of the inherent 10% error, we estimated the range of mass balance inaccuracies. The maximum value was determined by taking the NOVA measured flow for each artery and adding/subtracting 10% to/from the inflow/outflow measurements, respectively. The minimum value was determined by inversing the sign of the added error, i.e., 10% was subtracted from the inflow and 10% was added to the outflow. A study for 20% was also performed. In some situations, flow measurements were not present due to small artery signal in the MRA. A special rule was created for cases when the mass conservation was satisfied but MRA images demonstrated the presence of an additional communicating arteries such as PCOM. For each case we assigned arbitrarily a flow of 10 ml/min for SCAs and 5 ml/min for PCOMs which was within the 10% error. For the direction of the flow a neuro-radiologist analyzed the data in the posterior and anterior circulation in a case-by-case basis and assigned the direction of flow.

In Figure 4 we display the error range in the mass conservation at the four bifurcations. Most of the mass balance points lie within the expected 10% error in mass balance calculations. As an academic exercise we also performed the same evaluation for a 20% error in the measurements. Based on the error analysis 8 cases were ideal as assessment of the flow analysis at the bifurcations/trifurcations was within 10% error at all bifurcations. 26/34 cases had at least one bifurcation which was not within the 10% threshold. Table 3 provides a summary of the measurements which indicate that most of the mass balances lie within the 10% expected error.

Figure 4.

Volumetric flow rate balance at the RICA, LICA, ACA-ACA2, and BA bifurcations with error bars representing 10% error (left) and 20% error (right) in the volumetric flow rate measurements.

Table 3.

Summary of the volumetric flow rate error analysis.

Bifurcation	0% Error Due to Giving Unmeasured Vessel(s) Exact Missing Flow	0% < Error < 10%	10% < Error < 20%	> 20% Error
RICA Bifurcation	2	29	3	0
LICA Bifurcation	2	27	5	0
ACA Bifurcation	3	15	11	5
BA Bifurcation	9	18	3	4

Flow summary for each side of the brain in the main arteries dichotomized for no-reoccurrence vs. reoccurrence of the ischemia is shown in Figure 5. Significant differences between the positive and negative ischemia patients can be observed for the MCA measurements and ICA to a lesser degree. ACA and PCA measurements show complete overlap of the data.

Figure 5.

Flow measurements in the left and right sides for patients with ischemia recurrence. 0 and 1 denotes no recurrence and recurrence, respectively.

The ROC analysis using six-month ischemia reoccurrence status is shown in Figure 6. Table 4 gives a summary of the results of the ROC analysis. From the 34 studies that presented usable data for the study, 10 experienced reoccurring ischemic events. Seven showed new ischemia in the left anterior circulation and three showed new ischemia in the right anterior.

Figure 6.

Receiver Operator Characteristics, for reoccurrence of ischemia in the left side and right side respectively using ipsilateral blood flow measurements in the Internal Carotid Artery (ICA), Middle Cerebral Artery (MCA), Anterior Cerebral Artery (ACA), Anterior Cerebral Artery 2nd branching (ACA2), and Posterior Cerebral Artery (PCA).

Table 4.

Summary of the ROC analysis.

Left Side			Right Side			Average
Location	AUROC	Std. Error	Location	AUROC	Std. Error	AUROC	Std. Error
LICA	.583	.155	RICA	.670	.153	0.627	0.154
LMCA	.797	.137	RMCA	.844	.081	0.821	0.109
LACA	.431	.174	RACA	.598	.113	0.514	0.144
LACA2	.414	.141	RACA2	.607	.133	0.511	0.137
LPCA	.393	.132	RPCA	.808	.076	0.400	0.104

4. DISCUSSIONS

This is an effectiveness study for using 2D phase-contrast MRI-based intracranial flow measurements as a diagnosis for potential ischemia recurrence in patients with ICAD. Flow related parameters are used in other vascular disease for disease severity assessment and triaging of the patients for various medical management [17–21]. Cardiac Fractional Flow Reserve (FFR) in this case is a classic example for patient selection for angioplasty [22, 23]. In the neurovascular space, there is no quantitative biomarker-based standard for patient triaging for angioplasty or medical therapy. These initial results indicate that 2D phase-contrast MRI-based cerebral flow measurement could be used for early identification of the patients with high risk of recurrent ischemia.

Our initial assessment of a small group of ICAD patients (n=34) revealed that NOVA measurements could be used within 10% accuracy to measure the flow in the intracranial arteries. However, we determined that in a significant number of cases the measurements were inaccurate, especially with small vessel size or when the noise could affect placement of the perpendicular plane used for flow measurement. We investigated methods to correct for the missing data using post acquisition analysis. Using simple mass balance at each bifurcation we can identify those measurements which could be erroneous. Next, we could use adjuvant imaging data and location measurement setting to determine whether the measurement was due to the missing artery or missing flow measurement over an intact artery.

There are some limitations of this study related to patient variability, image resolution, and flow measurement methodology. These challenges can be addressed in a case-by-case situation, which currently is not optimal for clinical applications. In order to derive the flow within unmeasured vessels, the MRA scans of each case were first reviewed to determine whether vessels were not present versus present but not measured. Any vessels that were determined to be not present were given a flow value of zero. Then, the flow values for vessels deemed present but not measured were derived based on the flow measurements at each bifurcation using the concept of mass conservation, except for certain cases which are described in the following paragraph.

Another challenge is related to the inherent error in measurement. Although mass conservation should apply, there is an inherent error in the measured volumetric flow rates of the vessels. Therefore, there are cases where the inflow and outflow values were not equal, whether that was greater inflow than outflow, or vice versa. To account for this, certain estimates were made based on the neuroradiologist’s judgment. For example, if the RPCOM were present but not measured, and mass conservation at the RICA bifurcation was calculated to be nonzero but within 10% error of each vessel’s measurement, then RPCOM was assigned a volumetric flow rate of 5 ml/min flowing retrograde to the BA bifurcation. This estimation of 5 ml/min retrograde to the BA bifurcation was used for both RPCOM and LPCOM. A similar estimation was used in the case of unmeasured but present SCAs within the BA bifurcation where they were assigned a volumetric flow rate of 10 ml/min.

5. CONCLUSIONS

Utilization of 2D PCMR-based flow quantification of intracranial arteries, such as NOVA, in patients with moderate intracranial atherosclerotic disease shows potential of predicting reoccurrence of ischemia. Such predictions could provide valuable information to plan preventive medical treatments for ICAD patients to reduce the risk of recurrent ischemic events.