¹⁸F-NaF PET/MRI for Detection of Carotid Atheroma in Acute Neurovascular Syndrome

Abstract

Background

MRI and fluorine 18–labeled sodium fluoride (¹⁸F-NaF) PET can be used to identify features of plaque instability, rupture, and disease activity, but large studies have not been performed.

Purpose

To evaluate the association between ¹⁸F-NaF activity and culprit carotid plaque in acute neurovascular syndrome.

Materials and Methods

In this prospective observational cohort study (October 2017 to January 2020), participants underwent ¹⁸F-NaF PET/MRI. An experienced clinician determined the culprit carotid artery based on symptoms and record review. ¹⁸F-NaF uptake was quantified using standardized uptake values and tissue-to-background ratios. Statistical significance was assessed with the Welch, χ², Wilcoxon, or Fisher test. Multivariable models were used to evaluate the relationship between the imaging markers and the culprit versus nonculprit vessel.

Results

A total of 110 participants were evaluated (mean age, 68 years ± 10 [SD]; 70 men and 40 women). Of the 110, 34 (32%) had prior cerebrovascular disease, and 26 (24%) presented with amaurosis fugax, 54 (49%) with transient ischemic attack, and 30 (27%) with stroke. Compared with nonculprit carotids, culprit carotids had greater stenoses (≥50% stenosis: 30% vs 15% [P = .02]; ≥70% stenosis: 25% vs 4.5% [P < .001]) and had increased prevalence of MRI-derived adverse plaque features, including intraplaque hemorrhage (42% vs 23%; P = .004), necrotic core (36% vs 18%; P = .004), thrombus (7.3% vs 0%; P = .01), ulceration (18% vs 3.6%; P = .001), and higher ¹⁸F-NaF uptake (maximum tissue-to-background ratio, 1.38 [IQR, 1.12–1.82] vs 1.26 [IQR, 0.99–1.66], respectively; P = .04). Higher ¹⁸F-NaF uptake was positively associated with necrosis, intraplaque hemorrhage, ulceration, and calcification and inversely associated with fibrosis (P = .04 to P < .001). In multivariable analysis, carotid stenosis at or over 70% (odds ratio, 5.72 [95% CI: 2.2, 18]) and MRI-derived adverse plaque characteristics (odds ratio, 2.16 [95% CI: 1.2, 3.9]) were both associated with the culprit versus nonculprit carotid vessel.

Conclusion

Fluorine 18–labeled sodium fluoride PET/MRI characteristics were associated with the culprit carotid vessel in study participants with acute neurovascular syndrome.

Clinical trial registration no. NCT03215550 and NCT03215563

© RSNA, 2022

Online supplemental material is available for this article.

Summary

Fluorine 18–labeled sodium fluoride PET/MRI characteristics were associated with the culprit atherosclerotic plaques in the carotid circulation of study participants with acute neurovascular syndrome; PET/MRI was also usable in the assessment of carotid stenosis, high-risk plaque features, and plaque biologic activity.

Key Results

• ■ In this prospective observational cohort study of 110 participants with acute neurovascular syndrome who underwent fluorine 18–labeled sodium fluoride (¹⁸F-NaF) PET/MRI, culprit carotid vessels had more severe stenoses than nonculprit vessels (≥70% stenosis: 25% vs 4.5% [P < .001]).

• ■ Culprit vessels had higher ¹⁸F-NaF uptake (maximum tissue-to-background ratio, 1.38 vs 1.26 [P = .04]) and more MRI-derived adverse plaque features, including intraplaque hemorrhage (42% vs 23%; P = .004), necrotic core (36% vs 18%; P = .004), thrombus (7.3% vs 0%; P = .01), and ulceration (18% vs 3.6%; P = .001).

Introduction

The identification of the culprit atherosclerotic plaque represents a key goal in carotid artery imaging. The primary focus of current imaging is stenosis severity, but the detection of high-risk plaque features may aid the identification of patients at risk for stroke and potential targets for novel therapeutic intervention. Both PET and MRI have shown promise in the identification of atherosclerotic plaques with the potential to cause a stroke. In addition to stenosis severity, MRI tissue characterization can be used to identify adverse plaque features, such as lipid-rich necrotic core, intraplaque hemorrhage, and fibrous cap thickness (1). The PET radiotracer fluorine 18–labeled sodium fluoride (¹⁸F-NaF) localizes to areas of microcalcification activity indicative of inflammation and necrosis and has shown promise in the identification of active atherosclerosis across the vascular system (2,3). Combining these two techniques in PET/MRI would enable the simultaneous assessment of carotid stenosis, adverse plaque features, and plaque biologic activity.

PET/MRI has the potential to provide MRI-based stenosis assessment and soft-tissue characterization alongside PET-based functional evaluation of disease activity. To date, only a few small studies (n = 10–18) have assessed the feasibility of carotid PET/MRI (4–6). In our study, we assessed ¹⁸F-NaF PET/MRI in 110 participants with a broad range of acute neurovascular syndromes. We aimed to evaluate the association between ¹⁸F-NaF activity and culprit carotid plaque in acute neurovascular syndrome.

Materials and Methods

This prospective observational cohort study was approved by the South East Scotland Research Ethics Committee (17/SS/0097, 17/SS/0098). Written informed consent was obtained, and the study was performed in accordance with the Declaration of Helsinki. The study was registered on ClinicalTrials.gov (identifiers NCT03215550 and NCT03215563).

Participants

Participants over 40 years of age who had experienced a recent neurovascular event (within 14 days) that was compatible with the carotid (anterior) territory and who had undergone carotid Doppler US were recruited consecutively from acute neurovascular clinics (n = 2) or hospitals (n = 2) between October 2017 and January 2020. Diagnostic criteria for acute neurovascular events were based on symptoms supported by radiologic investigations, in line with published guidelines (7). The diagnosis of acute ischemic stroke or retinal infarct included clinically evident infarction persisting longer than 24 hours (7). A transient ischemic attack or amaurosis fugax (transient retinal ischemia) was diagnosed when clinical symptoms lasted less than 24 hours (7).

Exclusion criteria were a modified Rankin Scale score greater than 3, chronic kidney disease (estimated glomerular filtration rate <30 mL/min/1.73 m²), atrial fibrillation, pregnancy, prior ipsilateral carotid intervention or neck radiation therapy, inability to tolerate the supine position, inability to comply with study requirements, prior allergic reaction to study agents, contraindications to MRI, or if participation would delay carotid surgery.

Clinical Assessments

Within 14 days of the index event, participants underwent baseline assessment, which incorporated routine clinical assessment, review of health care records, and previous investigations. The attending stroke clinician (M.D. or W.W., each with >10 years of experience) defined the culprit carotid artery side based on the clinical presentation and medical records, including radiologic investigations.

¹⁸F-NaF PET/MRI Protocol

Participants received a target dose of 125 MBq of ¹⁸F-NaF intravenously, and after 60 minutes of rest, were imaged with a PET/MRI scanner (Biograph mMR; Siemens Healthcare) using a dedicated head and neck coil. A three-dimensional ¹⁸F-NaF PET scan was obtained using list-mode acquisitions with two 15-minute bed positions covering the carotid arteries (from the aortic arch to skull base). An MRI scan from the aortic arch to the circle of Willis was obtained simultaneously, including a black-blood T2 sampling perfection with application-optimized contrasts using different flip angle evolution, three-dimensional time-of-flight MR angiography, and two-dimensional, axial, black-blood T1-weighted, T2-weighted, and proton density–weighted stacks (Appendix E1 [online]). After PET image acquisition, contrast-enhanced MR angiography (gadobutrol 0.1 mL/kg [Gadovist 1.0; Bayer]) was performed, covering the carotid bifurcations and aortic arch. PET images were reconstructed with corrections applied for attenuation, dead time, scatter, and random coincidences (matrix size, 256 × 256; ordered subset expectation maximization reconstruction with point spread function modeling [high-definition PET; Siemens Healthcare]; three iterations; 21 subsets; 2-mm Gaussian filtration). Head and neck coils were part of the attenuation correction map.

MRI Analysis

Visual analysis of carotid plaques was performed using the Carestream Vue picture archiving and communication system (version 11), with the reader (M.C.W., with more than 10 years of experience) blinded to clinical presentation and imaging. The most severe stenosis in each carotid artery was visually assessed at time-of-flight or MR angiography, and area stenosis was categorized as less than 30%, 30%–49%, 50%–69%, or greater than or equal to 70% stenosis compared with a distal reference segment (8). Consistent with previously published criteria, signal intensities on T1-weighted, T2-weighted, proton density–weighted, and time-of-flight sequence images were used to identify intraplaque hemorrhage, lipid-rich necrotic core, calcification, thrombus, fibrosis, and ulceration (Table E1 [online]) (9–13). The sternocleidomastoid muscle was used as a reference for signal intensities, as per a previous study (14).

PET Image Analysis

Visual analysis of ¹⁸F-NaF uptake was performed by one author (M.C.W.) with use of dedicated software (FusionQuant; Cedars-Sinai Medical Center) and with the reader blinded to clinical presentation and imaging. PET was fused with time-of-flight MRI or MR angiography for analysis. ¹⁸F-NaF uptake was visually assessed as positive or negative in culprit and nonculprit vessels. ¹⁸F-NaF uptake was measured in a spherical volume of interest centered on the most severe stenosis in the culprit carotid, contralateral nonculprit carotid, and aortic arch. Care was taken to exclude adjacent bony structures by ensuring accurate coregistration and reviewing MRI scans and PET images. Mean and maximum standardized uptake values were recorded. Mean and maximum tissue-to-background ratios (TBR_mean and TBR_max) were calculated by dividing by the mean or maximum background activity averaged from three regions of interest within the brachiocephalic or internal jugular veins (2,15).

Statistical Analysis

Statistical analysis was performed using R (version 4.0.3; R Foundation for Statistical Computing) by two authors (J.K. and M.C.W., both with >10 years of experience). Normally distributed variables are presented as means ± SDs. Nonnormally distributed data are presented as medians with IQRs. Categorical variables are presented as numbers with percentages. Statistical significance was derived using the Welch two-sample t test, Wilcoxon rank-sum test with continuity correction, Pearson χ² test, or Fisher exact test. Threshold values for optimum TBR_mean and TBR_max to identify culprit versus nonculprit vessels were established using the Youden J statistic from receiver operator characteristic curve analysis. Multivariable models were constructed to identify culprit compared with nonculprit vessels, which included as covariates the presence of stenosis of 70% or greater at MRI, one or more MRI-derived adverse plaque characteristics, and either TBR_mean or TBR_max. TBR_mean and TBR_max were log-transformed for analysis (log base 2). Odds ratios were calculated. Two-sided P < .05 was considered indicative of statistically significant difference. On the basis of PET/CT (2), we estimated that with 80% power and a two-sided P < .05, a sample size of 33 participants would be required to for the identification of differences between culprit and nonculprit plaques. To account for differences in quantification with PET/MRI, participant drop-out, and multiple comparisons, we recruited 110 participants.

Results

Participant Characteristics

Among 500 eligible participants, 117 consented (Fig 1). Subsequently, seven withdrew due to an inability to complete the imaging protocol. The final study sample consisted of 110 participants (20 surgical, 90 nonsurgical). Imaging of the aortic arch was available in all but two participants.

Figure 1:

Consolidated Standards of Reporting Trials diagram of participant recruitment.

Participants had a mean age of 68 years ± 10 (SD), 70 were men, 40 were women, and 34 of 110 (31%) had a history of cerebrovascular disease (Table 1). Of the 110 participants, the presenting event was amaurosis fugax for 26 (24%), transient ischemic attack for 54 (49%), and ischemic stroke for 30 (27%). Compared with those presenting with amaurosis fugax or transient ischemic attack, participants presenting with ischemic stroke had a higher 10-year cardiovascular risk score (Assessing cardiovascular risk using Scottish Intercollegiate Guideline Network guidelines, or ASSIGN) (22 [IQR, 18–30] and 26 [IQR, 17–43] vs 39 [IQR, 24–55], respectively; P = .01), but most other risk factors were similar between subgroups (Table E2 [online]). Twenty-seven participants had stenoses greater than or equal to 70% in the culprit carotid vessel and were of a similar age, sex, and risk factor profile compared with those with stenoses less than 70% (Tables 1, 2). We found no evidence of a difference in presentation with stroke between participants with stenoses of 70% or greater in the culprit vessel compared with those without (11 of 27 participants [41%] vs 19 of 83 participants [23%]; P = .11). Twenty of the 110 participants (18%) underwent clinically indicated carotid endarterectomy after PET/MRI.

Table 1:

Clinical Characteristics of All Participants and Those with Culprit Vessel Stenosis above or below 70%

Table 2:

Therapy in All Participants and Those with Culprit Vessel Stenosis above or below 70%

Culprit versus Nonculprit Vessels

MRI analysis.—Compared with nonculprit vessels, culprit vessels were more likely to have stenosis greater than or equal to 70% (five of 110 [4.5%] vs 27 of 110 vessels [25%]; P < .001) or greater than or equal to 50% (17 of 110 [15%] vs 33 of 110 [30%]; P = .02) (Table 3). Culprit vessels were more likely to have intraplaque hemorrhage (46 of 110 vessels [42%] vs 25 of 110 [23%]; P = .004), necrotic core (40 of 110 [36%] vs 20 of 110 [18%]; P = .004), thrombus (eight of 110 [7.3%] vs 0 of 110 [0%]; P = .01), and ulceration (20 of 110 [18%] vs four of 110 [3.6%]; P = .001) (Table 3, Fig 2). We found no evidence of a difference in the presence of fibrosis (35 of 110 vessels [32%] vs 48 of 110 [44%]; P = .10) or calcification (47 of 110 [43%] vs 39 of 110 [35%]; P = .33) between culprit and nonculprit vessels. Compared with participants with less than 70% culprit stenosis, participants with greater than or equal to 70% culprit stenosis were more likely to have thrombus (one of 83 participants [1.2%] vs seven of 27 participants [26%]; P < .001), calcification (23 of 83 [28%] vs 24 of 27 [89%]; P < .001), and ulceration (nine of 83 [11%] vs 11 of 27 [41%]; P = .001), but there was no evidence of a difference in fibrosis between the two groups (31 of 83 [37%] vs four of 27 [15%]; P = .052) (Tables 4, 5).

Table 3:

¹⁸F-NaF PET/MRI Findings in Culprit and Nonculprit Vessels

Figure 2:

Fluorine 18–labeled sodium fluoride (¹⁸F-NaF) PET/MRI carotid scans and ex vivo histologic analysis sample in a patient who experienced a left hemispheric transient ischemic attack. MR angiograph in a 60-year-old woman with left hemispheric transient ischemic attack shows a left carotid artery stenosis of greater than or equal to 70% severity (box, MR angiography [MRA] panel; arrow, time-of-flight [TOF] panel). MRI plaque analysis was used to identify intraplaque hemorrhage (arrow, T1 panel), lipid-rich necrotic core (arrow, T2 panel), and calcification (arrow, proton density [PD] panel). ¹⁸F-NaF PET/MRI showed focal uptake in the region of the intraplaque hemorrhage and lipid-rich necrotic core (orange arrow). The carotid plaque was excised at operation. Cross-sectional images through the carotid plaque are shown, with hematoxylin and eosin (H & E) and Movat pentachrome stains. The red arrow indicates the thin fibrous cap, the blue arrow highlights plaque erosion, the star denotes a rupture, the purple arrow shows intraplaque hemorrhage with a thrombus, the yellow arrowheads denote inflammation (shoulder and central regions), and the green arrowheads indicate lipid-rich necrotic cores.

Table 4:

MRI Findings in Participants with Culprit Vessel Stenosis above or below 70%

Table 5:

MRI Findings in Participants with Nonculprit Vessel Stenosis above or below 70%

PET analysis.—Culprit vessels were more likely to have visually assessed ¹⁸F-NaF uptake than nonculprit vessels (78 of 110 vessels [71%] vs 41 of 110 [37%]; P < .001). Culprit vessels had a higher ¹⁸F-NaF uptake than nonculprit vessels (TBR_max, 1.38 [95% CI: 1.12, 1.82] vs 1.26 [95% CI: 0.99, 1.66]; P = .04) (Table 3). In the culprit vessel, stenosis greater than or equal to 70% was associated with higher ¹⁸F-NaF uptake (TBR_mean, 1.81 [95% CI: 1.44, 2.07] vs 1.27 [95% CI: 1.09, 1.60]; P = .001) (Figs 3, E1 [online]). However, we found no evidence of a difference in ¹⁸F-NaF TBR_mean in the nonculprit vessel in patients with or without 70% stenosis or greater (TBR_mean, 1.49 [95% CI: 1.00, 1.95] vs 1.25 [95% CI: 0.99, 1.43]; P = .15) (Figs 3, E1 [online]).

Figure 3:

Tukey box and whisker plots show fluorine 18–labeled sodium fluoride uptake (mean tissue-to-background ratio [TBR_mean] and maximum tissue-to-background ratio [TBR_max]) stratified by the degree of carotid stenosis (expressed as percentages) in culprit (red) and nonculprit (blue) vessels. The box bounds the IQR (upper and lower quartile) divided by the median (solid horizontal midline); whiskers extend to the most extreme data points from the edge of the box; outliers beyond the whiskers are individually plotted by the solid dots.

Participants with greater than or equal to 70% culprit stenosis had higher ¹⁸F-NaF uptake in the culprit vessel than those with less than 70% culprit stenosis (TBR_max, 4.23 [95% CI: 2.64, 7.13] vs 2.58 [95% CI: 2.13, 3.57] [P < .001]; TBR_mean, 1.81 [95% CI: 1.44, 2.07] vs 1.27 [95% CI: 1.09, 1.60] [P = .001]) (Fig 4, Table E3 [online]). They also had higher TBR_max in the nonculprit vessel than those with less than 70% culprit stenosis (3.07 [95% CI: 2.15, 4.26] vs 2.27 [95% CI: 1.88, 3.17]; P = .02) and higher TBR_mean in the aortic arch (1.62 [95% CI: 1.40, 1.97] vs 1.42 [95% CI: 1.24, 1.62]; P = .04) (Fig E1, Table E3 [online]).

Figure 4:

Tukey box and whisker plots show comparisons for fluorine 18–labeled sodium fluoride uptake in the (A) culprit vessel, (B) nonculprit vessel, and (C) aortic arch in patients with stenosis above or below 70% in the culprit vessel. The box bounds the IQR (upper and lower quartile) divided by the median (solid horizontal midline); whiskers extend to the most extreme data points from the edge of the box; outliers beyond the whiskers are individually plotted by the solid dots. TBR_mean = mean tissue-to-background ratio.

Combined PET and MRI.— ¹⁸F-NaF TBR_mean was higher in all vessels with necrotic core (1.45 [95% CI: 1.14, 1.82] vs 1.26 [95% CI: 1.01, 1.68]; P = .04), intraplaque hemorrhage (1.38 [95% CI: 1.13, 1.82] vs 1.26 [95% CI: 1.01, 1.61]; P = .049), ulceration (95% CI: 1.70 [1.37, 1.85] vs 1.27 [95% CI: 1.03, 1.70]; P = .006), and calcification (1.71 [95% CI: 1.31, 2.05] vs 1.16 [95% CI: 0.98, 1.39]; P < .001) and lower in vessels with fibrosis (1.14 [95% CI: 0.99, 1.35] vs 1.46 [95% CI: 1.14, 1.92]; P < .001) (Figs 5, E2 [online], E3 [online]).

Figure 5:

Tukey box and whisker plots show comparisons for fluorine 18–labeled sodium fluoride uptake and MRI-defined plaque features. The box bounds the IQR (upper and lower quartile) divided by the median (solid horizontal midline); whiskers extend to the most extreme data points from the edge of the box; outliers beyond the whiskers are individually plotted by the solid dots. TBR_mean = mean tissue-to-background ratio.

With use of the Youden J statistic, the optimum threshold value for the identification of culprit compared with nonculprit vessels was 1.25 for TBR_mean and 2.46 for TBR_max. Culprit vessels were more likely to have a combination of adverse plaque characteristics and TBR_mean greater than 1.25 (88 of 110 vessels [80%] vs 61 of 110 [55%]; P < .001) or TBR_max greater than 2.46 (88 of 110 [80%] vs 67 of 110 [61%]; P = .003).

In a multivariable analysis, MRI-derived stenosis of 70% or greater (odds ratio, 5.72 [95% CI: 2.2, 18]; P < .001) and adverse plaque characteristics (odds ratio, 2.16 [95% CI: 1.2, 3.9]; P = .009) were both associated with the culprit vessel compared with the nonculprit vessel, but TBR_mean (odds ratio, 1.16 [95% CI: 0.7, 1.9]; P = .56) was not. Similarly, in a second model, TBR_max was not associated with the culprit vessel compared with the nonculprit vessel (odds ratio, 1.02 [95% CI: 0.7, 1.6]; P = .91), whereas MRI-derived stenosis of 70% or greater (odds ratio, 5.84 [95% CI: 2.2, 18]; P < .001) and adverse plaque characteristics were (odds ratio, 2.18 [95% CI: 1.2, 3.9]; P = .008) (Fig 6).

Figure 6:

Multivariable models were constructed for the identification of culprit compared with nonculprit vessels, which included as covariates the presence of MRI-derived stenosis greater than or equal to 70%, one or more MRI-derived adverse plaque characteristics, and either mean tissue-to-background ratio (TBR_mean) or maximum tissue-to-background ratio (TBR_max). Dot plots show odds ratios, with blue lines representing 95% CIs. TBR_mean is per doubling [log₂(TBR_mean)]. *** = P < .001, ** = P < .01, * = P < .05, no asterisk = P ≥ .05.

Surgical versus Nonsurgical Participants

The 20 participants (18%) who underwent carotid surgery were more likely to have experienced stroke (nine of 20 [45%] vs 21 of 90 participants [23%]; P = .02) but had similar demographic characteristics to participants who did not undergo surgery (Table E4 [online]).

MRI analysis.—In the culprit vessel, participants undergoing surgery were more likely to have stenoses greater than or equal to 70% or 50% than those who did not undergo surgery (20 of 20 participants [100%] vs 13 of 90 participants [14%] [P < .001] and 19 of 20 [95%] vs eight of 90 [8.9%] [P < .001], respectively) (Table E5 [online]). In the nonculprit vessel, participants undergoing surgery were more likely to have stenoses greater than or equal to 50% than participants who did not undergo surgery (10 of 20 participants [50%] vs seven of 90 participants [7.8%]; P < .001), but we found no evidence of a difference in the frequency of stenoses of 70% or greater (three of 20 [15%] vs two of 90 [2.2%]; P = .06).

In the culprit vessel, participants undergoing surgery were more likely to have a necrotic core, intraplaque hemorrhage, thrombus, ulceration, and calcification and less likely to have fibrosis compared with participants who did not undergo surgery (P = .04 to P < .001) (Table E5 [online]). In the nonculprit vessel, participants undergoing surgery were also more likely to have calcification (19 of 20 participants [95%] vs 20 of 90 [22%]; P < .001) and less likely to have fibrosis (one of 20 [5.0%] vs 47 of 90 [52%]; P < .001), but we found no evidence of a difference in the other plaque characteristics (Table E5 [online]). Participants undergoing surgery were more likely to have any (15 of 20 participants [75%] vs 42 of 90 participants [47%]; P = .04) or a higher number of adverse plaque features (three [95% CI: 0.75, 5] vs 0 [95% CI: 0, 2]; P < .001) in the culprit vessel, but we found no evidence of a difference in the presence of adverse plaque features in the nonculprit vessel (0 [95% CI: 0, 1] vs 0 [95% CI: 0, 1.25]) (Table E5 [online]).

PET analysis.—Participants undergoing surgery had higher ¹⁸F-NaF uptake in the culprit vessel (TBR_mean, 1.79 [95% CI: 1.45, 1.95] vs 1.28 [95% CI: 1.10, 1.72] [P = .007]; TBR_max, 4.05 [95% CI: 2.68, 6.33] vs 2.55 [95% CI: 2.12, 3.59] [P = .001]). However, we found no evidence of a difference in radiotracer uptake in the nonculprit vessel (1.46 [95% CI: 1.06, 1.87] vs TBR_mean, 1.25 [95% CI: 0.98, 1.51]; P = .15) or the arch of the aorta (TBR_mean, 1.58 [95% CI: 1.29, 1.88] vs 1.44 [95% CI: 1.25, 1.65]; P = .26) between participants who did and did not undergo surgery (Table E5 [online]).

PET and MRI.—Participants undergoing surgery were more likely to have a combination of adverse plaque characteristics and TBR_mean greater than 1.25 (20 of 20 participants [100%] vs 64 of 90 [71%]; P = .01) or TBR_max greater than 2.46 (20 of 20 [100%] vs 64 of 90 [71%]; P = .01) than those who did not.

Discussion

Carotid fluorine 18–labeled sodium fluoride (¹⁸F-NaF) PET/MRI has the potential to combine information on stenosis severity, MRI-derived adverse plaque characteristics, and atherosclerotic disease activity. In this large prospective study including participants with a broad range of acute neurovascular syndromes, culprit carotid vessels had a higher degree of stenosis (≥70% stenosis: 25% vs 4.5%; P < .001), higher ¹⁸F-NaF uptake (maximum tissue-to-background ratio, 1.38 vs 1.26; P = .04), and more frequent MRI-derived adverse plaque features (52% vs 30%; P = .002) compared with nonculprit vessels. Moreover, ¹⁸F-NaF uptake was higher in vessels with lipid-rich necrotic core (1.45 [95% CI: 1.14, 1.82] vs 1.26 [95% CI: 1.01, 1.68]; P = .04), intraplaque hemorrhage (1.38 [95% CI: 1.13, 1.82] vs 1.26 [95% CI: 1.01, 1.61]; P = .04), and ulceration (1.70 [95% CI: 1.37, 1.85] vs 1.27 [95% CI: 1.03, 1.70]; P = .006) and lower in areas of fibrosis (1.14 [95% CI: 0.99, 1.35] vs 1.46 [95% CI: 1.14, 1.92]; P < .001), suggesting complementarity between plaque features and plaque biologic activity. These findings may lay the groundwork for future clinical evaluation of strokes with uncertain causes and assessment of carotid disease activity to guide the selection of patients for surgical intervention.

Previous studies established that adverse carotid plaque characteristics can be identified with MRI and that these features are more frequent in culprit compared with nonculprit vessels, including in patients without severe stenoses (16). The MRI-derived adverse carotid plaque characteristics described in this study were intraplaque hemorrhage, lipid-rich necrotic core, thrombus, and ulceration. Identification of these features may therefore help to determine the culprit vessel in cases where this is uncertain. These adverse carotid plaque characteristics have also been linked to an increased risk of subsequent acute events in those with established cerebrovascular disease and in asymptomatic patients (17–19). Recent meta-analyses found that the carotid MRI features associated with the highest risk of subsequent acute neurovascular syndromes were intraplaque hemorrhage, necrotic core, and thinning or rupture of the fibrous cap (18,20). Identification of these high-risk plaque features can therefore provide additional information beyond stenosis severity, which can help to guide management and prognostication.

¹⁸F-NaF is a marker of biologically active microcalcification (3,19), and we have previously shown that ¹⁸F-NaF PET/CT can be used to identify the culprit carotid plaque in patients with stroke or transient ischemic attack (2). In our current study, we expanded these findings demonstrating that ¹⁸F-NaF can be used to identify the culprit carotid plaque across a wider range of disease severity. We showed that radiotracer uptake is higher in vessels with high-risk characteristics at MRI and lower in areas of fibrosis. These findings are consistent with histologic assessment, showing that ¹⁸F-NaF localizes in areas with macrophage infiltration, apoptosis, necrosis, active calcification, and ulceration (21). MRI and PET, therefore, provide overlapping and complementary markers of high-risk atherosclerotic disease.

Several small studies have demonstrated the feasibility of PET/MRI of the carotid arteries using fluorine 18 fluorodeoxyglucose (4,22–28), copper 64 DOTA-0-Tyr3-octreotate (6), and gallium 68 pentixafor (29) in patients with previous stroke and a variety of other conditions. To our knowledge, only one previous study assessed ¹⁸F-NaF PET/MRI, and the study included 12 patients with carotid stenosis of 50% or greater (six symptomatic and six asymptomatic) (5). It was found that ¹⁸F-NaF uptake was higher in culprit compared with nonculprit plaques. However, the authors found no association with MRI features of plaque vulnerability, likely due to the small number of patients. In contrast, our study showed that ¹⁸F-NaF uptake was associated with key features of plaque vulnerability, a finding that, to our knowledge, has not previously been well established in the literature.

Our study had several limitations. First, it assumes that one culprit vessel was the source of the index acute event, whereas other cardioembolic sources or bilateral carotid disease may have been present. Second, the impact of the circle of Willis architecture or other brain MRI findings was not assessed. Third, ¹⁸F-NaF uptake in adjacent bony structures can cause overestimation of carotid uptake. Fourth, the threshold used to define abnormal ¹⁸F-NaF may be higher than the true threshold because a control group was unavailable. Fifth, false-negative ¹⁸F-NaF uptake may occur with a large plaque rupture event where there is little residual metabolically active plaque in the culprit vessel. Sixth, MRI signal characteristics may overlap, and very small lesions may not be apparent due to the MRI resolution. Finally, the clinical use of PET/MRI is limited by its lack of widespread availability and the exclusion of patients with metallic implants or claustrophobia.

Identification of culprit carotid plaques in patients with acute neurovascular syndromes could improve targeting of medical and surgical management, be used to monitor disease progression, and inform future prognostication. This study shows that carotid fluorine 18–labeled sodium fluoride (¹⁸F-NaF) PET/MRI can identify culprit atherosclerotic plaques and features of plaque vulnerability and activity. ¹⁸F-NaF uptake was higher in culprit lesions, areas of severe stenosis, and locations of adverse plaque characteristics. However, only the presence of severe stenosis and adverse plaque characteristics were independent markers of the culprit compared with nonculprit vessel. Future research will assess the optimal ¹⁸F-NaF uptake threshold for the identification of culprit lesions with PET/MRI and incorporate follow-up imaging in larger cohorts. Further technical improvements of PET/MRI will require reduction of image acquisition time and improvement of MRI attenuation correction. Further research is warranted to assess how these features of PET/MRI correspond with subsequent clinical outcomes.