Detection and Characterization of Thrombosis in Humans using Fibrin-Targeted Positron Emission Tomography and Magnetic Resonance

Abstract

Background:

The detection of thrombus in the left atrial appendage (LAA) is vital in the prevention of stroke and is currently performed using transesophageal echocardiography (TEE).

Objectives:

We present a novel technique to detect and characterize LAA thrombus in humans using combined positron emission tomography-magnetic resonance imaging (PET/MRI) of a fibrin-binding radiotracer, [⁶⁴Cu]FBP8.

Methods:

The metabolism and pharmacokinetics of [⁶⁴Cu]FBP8 were studied in 8 healthy volunteers. Patients with atrial fibrillation (AF) and recent TEEs of the LAA (positive n=12, negative n=12) were injected with [⁶⁴Cu]FBP8 and imaged with PET/MRI, including mapping the longitudinal magnetic relaxation time (T1) in the LAA .

Results:

[⁶⁴Cu]FBP8 was stable to metabolism and was rapidly eliminated. The maximum standardized uptake value (SUV_Max) in the LAA was significantly higher in the TEE positive than negative subjects, median [interquartile range] of 4.0 [3.0-6.0] vs. 2.3 [2.1-2.5]; p < 0.001, with an area-under-the-receiver operating characteristic curve of 0.97. A SUV_Max threshold of 2.6 provided a sensitivity of 100% and specificity of 84%. The minimum T1 (T1_Min) in the LAA was 970ms [780-1080] vs. 1380ms [1120-1620] (TEE positive vs. negative), p < 0.05, with some overlap between the groups. Logistic regression using SUV_Max and T1_Min allowed all TEE positive and negative subjects to be classified with 100% accuracy.

Conclusion:

PET/MR of [⁶⁴Cu]FBP8 is able to detect acute as well as older platelet-poor thrombi with excellent accuracy. Furthermore, the integrated PET/MRI approach provides useful information on the biological properties of thrombus such as fibrin and methemoglobin content.

Clinical Trial:

Imaging of LAA Thrombosis URL: https://clinicaltrials.gov/ct2/show/NCT03830320 Registration number: NCT03830320

Introduction

Thrombosis plays a central role in numerous cardiovascular diseases (1). While major advances have been made in the molecular understanding and treatment of thrombosis, its diagnosis remains indirect and imprecise (2,3). The presence of a filling defect on computed tomography (CT) and magnetic resonance imaging (MRI), or the detection of echogenic material on ultrasound, currently form the basis of non-invasive thrombus detection. These approaches are of major clinical utility but provide no information on the molecular and cellular properties of the thrombus. In addition, these techniques frequently employ nephrotoxic contrast agents, have limited anatomical coverage and, in the case of transesophageal echo (TEE), involve heavy sedation and the risk of esophageal trauma.

A number of molecular imaging approaches targeting platelets (4,5), fibrin (6,7), d-dimers (8,9), or fibrinogen (10), have been proposed, with some progressing to clinical trials (3,11). However these agents have been limited by suboptimal pharmacokinetics, high blood background, low specificity, or sensitivity only to hyperacute/acute thrombi (3). We and others have previously described the use of a small fibrin-avid peptide to image thrombosis by MRI (6,12,13). The translational potential of this agent (EP-2104R), however, was limited at the time due to concerns regarding the potential toxicity of gadolinium (14,15). While these concerns have not materialized, and no toxicity from EP-2104R has emerged, it did motivate us to design a new agent with a less complicated pathway towards clinical translation.

We have recently reported the preclinical development of a small peptide based fibrin-specific probe labeled with the copper-64 isotope to detect thrombus by positron emission tomography (PET) (16). This probe, [⁶⁴Cu]FBP8, has submicromolar affinity for fibrin, and facilitated the rapid detection of venous, arterial, and embolic thrombi in a number of animal models (16–18). Here, we report the first human experience with [⁶⁴Cu]FBP8, which we use to detect left atrial appendage (LAA) thrombus in subjects with atrial fibrillation (Afib). The detection of LAA thrombus was selected due to the clinical prevalence of Afib and stroke (19), because TEE provides an accurate gold standard for comparison (20), and because the use of LAA occlusion devices in selected individuals allows the development of the thrombus to be precisely timed (21,22). We further aimed in the study to assess whether an integrated PET/MRI approach, enabling the simultaneous acquisition of PET and longitudinal magnetic relaxation (T1) maps of the heart would enhance diagnostic accuracy and help characterize the detected thrombi.

Methods

Study Subjects

This study was approved by the Mass General Brigham (MGB) Institutional Review Board (protocol number 2015P002385) and registered on the ClinicalTrials.gov website (NCT03830320). All subjects were aware of the benefits/risks of the study and provided written consent. Healthy controls (HC) were recruited via the MGB Research Study Volunteer Program, and had to be 18 years of age or older with no known medical issues, have no history of Afib or thrombosis, and a negative drug screen. Patients with Afib, presenting to the electrophysiology laboratory at the Massachusetts General Hospital (MGH) for TEE guided cardioversion, were recruited for the LAA arm of the study. These subjects fell into three groups: those with no sign of LAA thrombus in the TEE (n=12), those with spontaneous thrombi in the LAA on TEE (n=4), and those in whom thrombus in the LAA was purposefully induced through the placement of a LAA closure device (Watchman, Boston Scientific, n=8). Most of the subjects (7/8) with Watchman devices were imaged within 12 days of the device being placed. One Watchman subject was recruited from the echocardiography laboratory at the MGH when presenting for his scheduled TEE 6 weeks after the placement of the device. All subjects underwent PET/MRI with [⁶⁴Cu]FBP8 within 14 days of their reference TEE. Subjects were excluded if they had electrical implants such as a cardiac pacemaker/defibrillator, exceeded the weight limit of the PET/MRI scanner or had a contraindication to MRI. Further details on the clinical characteristics of the TEE negative and TEE positive study subjects are provided in Table 1 and in the supplement. The synthesis of [⁶⁴Cu]FBP8 and the chemical analysis of blood samples are also described in the supplement.

Table 1:

Characteristics of TEE Positive and TEE Negative Subjects

Characteristic	TEE Positive Patients (N = 12)	TEE Negative Patients (N = 12)	p-value
Age – yr
Median	80	61	0.0035
Interquartile range	76-84	55-73
Female sex – no. (%)	4 (33.3%)	4 (33.3%)	1.0000
CHA2DS2-VASc score
Median	4.50	2.00	0.0101
Interquartile range	4.00-5.75	0.25-3.75
GFR (ml/min) – mean ± sd	54.08 ± 14.23	65.55 ± 21.55	0.1542
Duration of atrial fibrillation - yr
Median	1.33	0.62	0.1748
Interquartile range	0.63 - 2.74	0.07 - 1.32
Type of atrial fibrillation			0.0996
Paroxysmal– no. (%)	3 (25.0%)	8 (66.7%)
Persistent/Permanent– no. (%)	8 (66.7%)	4 (33.3%)
Unspecified– no. (%)	1 (8.3%)	0 (0%)
Current smoking status – no. (%)	0 (0%)	1 (8.3%)	1.0000
Concurrent diagnosis of hypertension – no. (%)	11 (91.7%)	8 (66.7%)	0.3168
Concurrent diagnosis of diabetes – no. (%)	5 (41.7%)	0 (0%)	0.0373
Prior stroke, TIA, or thromboembolism – no. (%)	5 (41.7%)	1 (8.3%)	0.1550
Hypercoagulability – no. (%)	0 (0%)	1 (8.3%)	1.0000
Anticoagulation status			0.2174
DOAC – no. (%)	9 (75.0%)	12 (91.7%)
Warfarin – no. (%)	2 (16.7%)	0 (0%)
Warfarin INR
Median	3.30
Interquartile range	3.25 - 3.35
None – no. (%)	1 (8.3%)	0 (0%)
ASA only – no. (%)	1 (8.3%)	0 (0%)	1.0000
Antiplatelet only – no. (%)	0 (0%)	0 (0%)	1.0000
Dual Rx (anticoagulation + ASA) – no. (%)	8 (66.7%)	3 (25%)	0.0995
Triple Rx (anticoagulation + ASA + antiplatelet) – no. (%)	2 (16.7%)	1 (8.3%)	1.0000
No. of days from qualifying TEE to imaging
Median	9.50	11.00	0.4305
Interquartile range	5.00-12.00	8.00-12.75
Previous Watchman device placement – no. (%)	8 (66.7%)	0 (0%)
LAA Emptying Velocity (cm/s)
Median	14.00	35.50	0.1134
Interquartile range	11.00-64.50	19.50-52.75
LA Volume Index (ml/m2) – mean ± sd	44.95 ± 12.65	36.79 ± 13.34	0.1680

PET/MRI Data Acquisition

All images were acquired on a simultaneous PET/MRI scanner (Biograph mMR, Siemens Heathineers, USA) at a single station (~45 min) for the Afib subjects and at multiple stations (~ 2h) for the HCs. For all subjects, PET images were retrospectively reconstructed around 70 min post injection of approximately 380 and 250 MBq of [⁶⁴Cu]FBP8 for the HC and Afib subjects respectively. Simultaneously to the PET imaging, the following MRI sequences were acquired: a dual-echo Dixon-VIBE and HASTE sequences (both HC and Afib), and ECG-gated balanced steady state free precession (bSSFP) cine images and Modified Look-Locker imaging (MOLLI) sequences for Afib subjects only.

PET Image Analysis

Signal was measured in two regions of interest (ROIs) in the PET images: The LAA and the left ventricular blood pool. To avoid bias, the ROIs were drawn using OsiriX MD (v. 10.3, Pixmeo, Switzerland) on the T2 Haste images (HC) or the Dixon opposite phase images (Afib subjects) and then copied without modification to the PET images. A large spherical ROI was drawn to enclose the entire LAA and was standardized as much as possible across the subjects.

PET images were converted into standardized uptake values (SUV) using the subject’s weight as a normalization factor (23). Mean and maximum SUVs (SUV and SUV_MAX respectively) were obtained from the ROIs after exporting the OsiriX ROIs into Matlab for data analysis. Receiver operating characteristic (ROC) analysis of the SUV_MAX values was performed in GraphPad Prism and a SUV_MAX threshold of 2.6 was chosen from the ROC curve to provide 100% sensitivity.

T1 Map Analysis

T1 maps were produced using the reconstruction package provided by the manufacturer (Myomaps, Siemens Healthineers). ROIs were drawn using OsiriX in each coronal and axial T1 map in the three following regions: 1) LAA, a manually drawn ROI that closely followed the anatomical contour of the LAA; 2) BloodPool_T1, a manually drawn ROI in either the left atrium or the right pulmonary artery; 3) Fat, a subcutaneous fat ROI near the LAA. ROIs were exported into an in-house Matlab program and mean and minimum T1 values (T1 and T1_min respectively) were obtained from the LAA. To prevent voxels within the adipose tissue surrounding the LAA from biasing the analysis, voxels within the LAA with values lower than two standard deviations above the mean of the Fat region were excluded.

Combined PET/MRI Analysis: Z-Score Analysis with Logistic Regression

Z-scores are defined as the number of standard deviations above (for PET images) or below (for T1 maps) the mean of the background region, as follows:

$$Z_{PET}=\frac{SU{V}_{max}-\overline{BloodPool}}{S{D}_{BloodPool}},Z_{T1}=\frac{\overline{BloodPoo{l}_{T1}}-T{1}_{min}}{S{D}_{BloodPoo{l}_{T1}}}$$

Z-scores for PET and T1 were used in a logistic regression analysis to estimate the predictive probability of combined PET and MR measurements to define a subject i as positive/negative, with respect to the gold standard of TEE, as follows:

$$p_i=\frac{e^{{\beta }_0+{\beta }_1\cdot Z_{PET}^i+{\beta }_2\cdot Z_{MR}^i}}{1+e^{{\beta }_0+{\beta }_1\cdot Z_{PET}^i+{\beta }_2\cdot Z_{MR}^i}}$$

Qualitative Analysis

Qualitative analysis of the PET images, identifying a patient as positive or negative for increased FBP8 uptake over the LAA, was performed by a radiologist with more than 15y of experience (O.C.), who was blinded to the TEE data. Due to the heterogeneous elimination of the probe from the blood pool, images were individually windowed to produce a consistent appearance of the left ventricular blood pool.

Statistical Analysis

Descriptive statistics were calculated to summarize clinical characteristics. Differences between the TEE positive and negative groups were assessed using a two-tailed Student’s t-test for normally distributed characteristics, and a Mann-Whitney test for non-normally distributed characteristics. A Fisher’s exact test (with Freeman-Halton extension for multiple comparisons) was used for dichotomous outcomes. All datasets were tested for normality (GraphPad Prism) using the D’Agostino-Pearson omnibus K2 test. The maximum SUV value (SUV_MAX) in the LAA of the TEE positive and negative subjects was compared using a Mann-Whitney test (GraphPad Prism). The minimum T1 value (T1_MIN) in the LAA of the TEE positive and negative subjects was also compared using a Mann-Whitney test (GraphPad Prism). Median values with interquartile ranges are reported for both SUV_MAX and T1_MIN. Correlations were calculated using the Spearman coefficient. Logistic regression was used, as explained above, to estimate the predictive probability of the combined PET/MRI measurements, using the predict and GLM functions in R with a binomial family. All statistical tests were run on Graphpad Prism v.8.4.0 or Rstudio v.1.2 using R package v.3.6.2.

Results

Safety and Pharmacokinetics of [⁶⁴Cu]FBP8 in Healthy Human Volunteers

The pharmacokinetics of [⁶⁴Cu]FBP8, the structure of which is provided in Figure 1A, were highly consistent. The relatively low molecular weight of 2300 Da facilitated rapid and efficient excretion into the urine (Figure 1B). A smaller fraction of the radioactivity was excreted via the hepatobiliary system. Time activity curves (TACs) from ROIs drawn in the left ventricle (blood pool), lung, and background (measured in a large ROI capturing the gluteus maximus and/or the quadriceps muscles) showed that [⁶⁴Cu]FBP8 was rapidly cleared from these tissues (Figure 1B), with little activity remaining by 4 hours. High performance liquid chromatography (HPLC), metabolite analysis, binding activity results are shown in Figure 1C–1F. A visual depiction of [⁶⁴Cu]FBP8 biodistribution is provided in Figure 1G, and a magnified view of the heart in Figure 1H. The background signal in the heart and thorax was low within 1 hour of injection and was virtually eliminated within 4 hours of injection (see supplement).

Figure 1. Structure and properties of ⁶⁴Cu-FBP8.

(A) The probe consists of a 6 amino-acid cyclic peptide with a high affinity for fibrin, conjugated to a ⁶⁴Cu-containing chelator (NODAGA). (B) Average time activity curves (TACs) from 5 healthy subjects, expressed as SUV (g/ml), show a relatively rapid reduction of the PET signal in the blood and lung, accompanied by renal and hepatic elimination. Image analysis of the blood pool showed that ⁶⁴Cu-FBP8 was eliminated from the blood in a bi-exponential manner typical of radiotracers with extracellular distribution and renal clearance, with an initial distribution phase (alpha t_1/2 = 6.9 ± 3.0 min) followed by an elimination phase (beta t_1/2 = 68 ± 2 min). (C-F) Chemical analysis of blood samples taken from healthy volunteers injected with [⁶⁴Cu]FBP8. (C) High Performance Liquid Chromatography (HPLC) of ⁶⁴Cu-containing moieties in the blood at various times postinjection shows that the majority of ⁶⁴Cu activity is intact [⁶⁴Cu]FBP8 which has a retention time of 12 minutes. (D) [⁶⁴Cu]FBP8 is stable in blood, with >60% of the probe still intact by HPLC one hour after injection. (E) Over 50% of the probe remains functional and able to bind fibrin 1 hour after injection. (F) The clearance of ⁶⁴Cu activity from the blood, measured by venous blood sampling and expressed as %ID/g, indicates a one-hour blood elimination half-life (beta t_1/2 = 57 ± 7 min). (G-H) Biodistribution of [⁶⁴Cu]FBP8 in a healthy control subject shows a substantial reduction in the background thoracic signal within 1 hour of injection, and a near-complete loss of background by 4 hours. Elimination of the probe in the bladder and gallbladder is clearly seen. (H) Magnified view of the thorax at 1 hour showing no evidence of focal [⁶⁴Cu]FBP8 uptake in either the heart or lungs.

Comparison of [⁶⁴Cu]FBP8 versus TEE in Subjects with Atrial Fibrillation

The characteristics of the subjects in the TEE positive and negative cohorts are described in the Table 1. The majority (21/24) of subjects were taking direct oral anticoagulants (DOACs) at the time of imaging, 2 were taking warfarin and 1 was not taking any anticoagulant. Five of the subjects in the TEE positive group, and one in the TEE negative group, had a history of TIA, stroke or thromboembolism. One of the subjects in the study (in the TEE negative group) had a history of a hypercoagulable state (Factor V Leiden).

The [⁶⁴Cu]FBP8 images in the TEE positive patients typically showed a region of high focal uptake in the LAA compared to the PET signal in the rest of the blood pool and the surrounding tissue. This is shown in Figure 2A–F, where several spontaneous thrombi are present in the LAA by TEE. PET imaging with [⁶⁴Cu]FBP8 detected all of these thrombi. Foci of intense [⁶⁴Cu]FBP8 uptake were detected in the LAA of all the TEE positive patients. In contrast, the majority of TEE negative subjects showed no evidence of focal [⁶⁴Cu]FBP8 uptake in the LAA. This is well demonstrated in Figure 2G–I in a subject with spontaneous echo contrast (SEC) but no thrombus within the LAA. The maximum standardized uptake values (SUV_MAX) in the LAA were significantly higher in the TEE positive than TEE negative group (median and interquartile range of 4.0 [3.0-6.0] vs. 2.3 [2.1-2.5]; p < 0.001). A SUV_MAX threshold of 2.6 yielded a sensitivity of 100% and specificity of 84% (Figure 2J–K). ROC analysis of SUV_MAX produced an area-under-the-ROC-curve (AUROC) of 0.97 (Figure 2L).

Figure 2. Detection of spontaneous thrombi in the LAA.

TEE, PET and fused PET/MRI images of a subject with multiple LAA thrombi are shown in two orthogonal planes, (A-C) and (D-F). The thrombi produce echogenic/bright foci (yellow arrows) on the TEE images (A, D). The PET (B, E) and PET/MRI (C, F) images reveal multiple foci of [⁶⁴Cu]FBP8 uptake in the LAA (white arrows) in both planes. In addition, thrombus is also identified in the RAA (green arrow). (G-I) [⁶⁴Cu]FBP8 does not accumulate in the LAA in a subject with spontaneous echo contrast (SEC) but no thrombus. (G) Echogenic signal is seen within the LAA (yellow arrow), but color Doppler (inset) reveals extensive flow consistent with SEC. No evidence of LAA thrombus is seen on the [⁶⁴Cu]FBP8 PET images in the axial (H) or coronal (I) views. (J-L) The maximum SUV value in the LAA (SUV_MAX) accurately distinguishes TEE positive and negative subjects. (J) SUV_MAX in the TEE positive subjects was significantly higher than the TEE negative subjects. (J-K) A SUV_MAX threshold of 2.6 correctly classified all (12/12) TEE positive subjects and 10/12 TEE negative subjects. The area in the rectangle in panel J is magnified in panel K and confirms that a SUV_MAX threshold of 2.6 produces no false negative and only 2 false positive cases. (L) Receiver operating characteristic (ROC) curve demonstrates the robustness of [⁶⁴Cu]FBP8 SUV_MAX in distinguishing TEE positive and negative subjects, with an area under the curve (AUC) of 0.97.

The LAA thrombi induced by the placement of the occlusion device (Watchman, Boston Scientific) were extremely conspicuous (Central Illustration), likely reflecting the size and high fibrin content in these thrombi. The majority of the Watchman subjects (7/8) were imaged 5-12 days after the procedure. The other Watchman subject was imaged at day 54 and the thrombus in the LAA remained extremely conspicuous (SUV_MAX of 7.9). All thrombi in the Watchman subjects were confined to the LAA, behind the device, and no evidence of communication between the LAA and left atrium was seen on the bSSFP cine images.

Central Illustration. Thrombus in the LAA induced by recent placement of a Watchman LAA closure device.

A large thrombus is seen in the LAA behind the device on (A, B) the coronal PET and PET/MRI images as well as (C, D) the axial PET and PET/MRI images. (E, F) 2D bSSFP cines stacks in the coronal and axial planes have been combined into a single 3D dataset and fused with the 3D PET data. In these multi-modal images the PET and MR data are both displayed in 3D, creating a volumetric depiction of the heart and the thrombus/[⁶⁴Cu]FBP8 containing LAA. Volume rendered images in the oblique coronal and axial planes confirm the presence of thrombus in the LAA. (A-F) The thrombus produced by the closure device is contained within the LAA, and no evidence of thrombus is seen elsewhere in the heart or thorax.

Integrated Approach with [⁶⁴Cu]FBP8 PET and MR T1 Mapping

Inspection of the individual T1 weighted images in the TEE positive subjects frequently revealed the presence of hyperintense signal in the LAA (Figure 3A, 3E), consistent with LAA thrombus. T1 mapping confirmed the presence of large foci of reduced T1 in the LAA, which were extremely conspicuous in the Watchman subjects (Figure 3B–C). Quantitative analysis of the T1 maps showed that T1_MIN in the LAA was significantly shorter in the TEE positive subjects compared to the TEE negative subjects 970 [780-1080] versus 1380 [1120-1620], respectively, p < 0.05; although there was some overlap in the T1_MIN values between the groups. Joint analysis of the SUV and T1 data was then performed using logistic regression. Logistic regression using the combined z-scores of SUV_MAX and T1_MIN as inputs was extremely accurate (Figure 3J). All TEE positive cases were detected with a probability of 100% and no TEE negative case had a probability > 30%. The diagnostic accuracy of this approach was 100%, with an AUROC of 1.

Figure 3. Integrated analysis of magnetic relaxation (T1) and SUV_MAX in the LAA.

(A-D) Subject with a LAA closure device. (A) T1-weighted image in the coronal plane showing a hyperintense region in the LAA (white arrow). (B) T1 map in the same view demonstrating low values (green-purple; black arrow) in the LAA, similar to the T1 of myocardium (purple) but far lower than the blood pool (orange). (C) Axial T1 map confirms the presence of short-T1 (paramagnetic) species in the LAA, consistent with methemoglobin (met-heme) accumulation in thrombus. (D) Magnified view of the axial T1map fused with the [⁶⁴Cu]FBP8 PET image demonstrates a high degree of the overlap between low T1 and high [⁶⁴Cu]FBP8 uptake. (E-H) Spontaneous LAA thrombus. (E) Axial T1W image showing a hypertintense focus (arrow) in the LAA. (F) Axial T1 map shows a focus of reduced T1 (arrow) in the LAA. (G) Magnified view of T1 map and short T1 focus (arrow). (H) Magnified view of R1 map (units s⁻¹). The thrombus in the LAA (arrow) has a high R1 value. (I) The minimum T1 value (T1_MIN) in the LAA is significantly shorter in the TEE positive than negative group. (J) Probability of having a LAA thrombus based on logistic regression of the z-scores for both SUV_MAX and T1_MIN. WA = Watchman, SP = spontaneous. (K) Biological properties of the LAA (fibrin and met-heme content) in those with precisely aged recent thrombi (WA group) and no thrombi (Neg group). A strong correlation is seen between fibrin content (SUV_MAX) and met-heme content (R1_MAX = 1/T1_MIN). (L) Within the TEE negative group an association may be present between higher SUV and R1 values and thrombotic risk. Factor VL = Factor V Leiden.

The relationship between fibrin and methemoglobin (met-heme) content in the LAA, measured by SUV_MAX and R1_MAX (1/T1_MIN), respectively, was complex. The correlation between these measures in spontaneous LAA thrombi was poor, perhaps reflecting the heterogeneous nature of this group. However, in the remaining 14 subjects in whom both SUV and R1 data were available (8 Watchman and 6 TEE negative) a strong correlation (r = 0.76) was seen between fibrin and met-heme content in the LAA (Figure 3K). Further, in the 6 TEE negative subjects, the presence of both high SUV_MAX and R1_MAX values in the LAA was associated with high risk features, such as a Factor V Leiden mutation and a history of cerebral thromboembolism (Figure 3L).

Composite Detection of Cardiac and Extracardiac Thrombi

While quantitative analysis with the SUV_MAX threshold described above produced a sensitivity of 100%, qualitative inspection of the images correctly classified 11/12 TEE positive subjects and 12/12 TEE negative subjects. Co-incidental thrombi were relatively uncommon and included 1 pulmonary embolism (Figure 4A–C) and right atrial appendage thrombus in 4/24 subjects.

Figure 4. [⁶⁴Cu]FBP8 facilitates whole-body thrombus detection.

(A-C) Right pulmonary artery embolism in a subject with a spontaneous LAA thrombus. (A) Coronal PET image with a focus of high [⁶⁴Cu]FBP8 activity (arrow). (B) Fusion of the PET and Dixon water images reveals that the thrombus is located within a branch of the right pulmonary artery. (C) Volume rendered image showing the pulmonary embolism (arrow,) in the right lung. (D-F) TEE positive subject with a history of a large intracranial bleed. (D) Sagittal PET image, (E) fused PET and T1-weigted image and, (F) fused PET and susceptibility weighted image (SWI). The PET image shows that a low level of [⁶⁴Cu]FBP8 uptake persists at the site of the bleed (arrow).

Discussion

Anticoagulation is recommended in many clinical scenarios but is frequently empiric and complicated by the risk of major, and potentially fatal, hemorrhage (1,2). We show in this first-in-human study that PET/MRI of the fibrin-binding PET probe [⁶⁴Cu]FBP8 can detect acute to subacute LAA thrombi with a sensitivity of 100% and specificity of 84%, matching the performance of current more invasive approaches. The probe allows the composite burden of arterial thrombosis and venous thrombi, particularly those in larger veins with the potential to embolize, to be imaged and could potentially play a valuable role in elucidating the source of embolism in patients with TIAs and stroke, although this needs to be validated with further studies. The integrated PET/MRI approach also provides specific signatures that directly reflect the biological properties of the thrombus.

Human thrombus molecular imaging has been attempted in the past with limited success (3,9,10). The development and application of [⁶⁴Cu]FBP8 was based on the rational design of its preclinical precursors (16–18), and prior human studies with a fibrin-binding MR-detectable agent, EP-2104R (6). While no cases of nephrogenic systemic fibrosis have been reported with EP-2104R, the use of gadolinium-based probes in subjects with renal impairment remains a concern. In addition, the imaging of EP-2104R frequently involved two separate MRI sessions for pre- and post-contrast imaging, although post-contrast imaging alone may be sufficient (13), particularly with newer approaches such as T1 mapping. A strong case existed, therefore, to develop a PET-based probe based on the fibrin-binding peptide in EP-2104R. [⁶⁴Cu]FBP8 emerged from this process with nanomolar affinity to fibrin, rapid blood clearance, absence of metabolism in rodents, and high sustained and specific uptake in different animal models of arterial, venous, and embolic thrombi (17,18).

Many targets that have been explored for thrombus detection are highly expressed in acute but not chronic thrombi. The glycoprotein_IIb/IIIa receptor, for instance, is only suited to the detection of acute platelet-rich thrombi (4). While the level of fibrin in thrombi does decrease with time (18), our data suggest that it remains sufficient to support detection with [⁶⁴Cu]FBP8 for at least 8-10 weeks. We estimate based on serial TEEs that one of the spontaneous thrombi detected in the study may have been up to 9 weeks old. In addition, one of the subjects with a Watchman device was imaged 54 days after implantation. The SUVs in the thrombi over this range (< 10 weeks) remained stable. Further study will be needed to determine the sensitivity of [⁶⁴Cu]FBP8 for chronic thrombi (> 10 weeks), the utility of adding a gadolinium-based contrast agent to facilitate their detection, and the embolic risk of these fibrin-poor organized thrombi.

The degradation of red blood cells in subacute thrombi results in the production of the strongly paramagnetic met-heme species, which can be detected with T1 mapping (24). When integrated with PET imaging of [⁶⁴Cu]FBP8, the combined approach detected the presence or absence or thrombus with 100% accuracy. In addition, relevant biological information on the properties of the thrombus could be obtained (24). In large acute to subacute thrombi (Watchman subjects), and interestingly in the TEE negative subjects as well, a strong correlation was seen between fibrin and met-heme content in the LAA. High SUV_MAX and R1_MAX values in the TEE negative subjects were also associated with high-risk features such as the presence of a hypercoagulable state (Factor V Leiden) or history of cerebral thromboembolism. While very preliminary, these observations raise the possibility that small microthrombi, undetectable by TEE, may be present in the LAA during Afib (25). Studies using transorbital Doppler in TEE negative subjects have documented multiple embolic signals to the brain, supporting this concept. In subjects with spontaneous LAA thrombi the correlation between fibrin and met-heme content in the thrombus was poor. This likely represents the heterogeneous nature of this cohort and the older age of their thrombi. Collectively, while very preliminary, the observations above demonstrate the potential value of integrated PET/MRI imaging with [⁶⁴Cu]FBP8, T1 mapping and other forms of endogenous MR contrast (magnetization transfer, diffusion) with sensitivity to thrombi (26).

Study Limitations

The PET images of [⁶⁴Cu]FBP8 were not gated or motion corrected, which could result in volume averaging and the underestimation of probe uptake. Although we minimized potential artifacts introduced by the attenuation correction method supplied by the scanner manufacturer, remaining inaccuracies could have biased our measurements. The pharmacokinetic studies in healthy volunteers indicated that imaging of [⁶⁴Cu]FBP8 should ideally be performed 4 hours after injection. However, for logistical reasons, imaging in this study was performed 1 hour after probe injection. Since the rate of renal excretion of the probe varies significantly, this resulted in heterogeneous levels of background/blood pool signal both within and between the healthy controls and the TEE groups. However, despite the higher background signal, the LAA thrombi were highly conspicuous with high thrombus to background ratios. The number of spontaneous LAA thrombi in the study was low but sufficient, we believe, to demonstrate proof-of-principle in this first-in-human study. Further studies will be required to validate our results in larger prospective clinical trials.

Conclusions

We present the first human experience with a novel fibrin-binding radiotracer, [⁶⁴Cu]FBP8, and show that it facilitates the detection of acute-subacute LAA thrombi with a sensitivity of 100% and a diagnostic accuracy greater than 90%. We also show that the integrated use of [⁶⁴Cu]FBP8 and T1 mapping of the LAA can further increase diagnostic accuracy and characterize aspects of thrombus biology. The probe is well tolerated, demonstrates favorable pharmacokinetics and can support whole body detection of thrombosis. [⁶⁴Cu]FBP8 has the potential to advance the understanding of thrombus biology and to transform the diagnosis and treatment of thrombosis in a wide range of disease settings.

Clinical Perspectives.

Clinical Competency

Medical Knowledge

• The presence of fibrin in thrombus allow both acute and older thrombi to be detected by positron emission tomography (PET)

• Mapping the longitudinal relaxation time (T1) in the left atrial appendage (LAA) may allow thrombi to be detected via a reduction in T1

Translational Outlook

• Serial imaging with [⁶⁴Cu]FBP8 could be used to better understand the natural history of thrombus in the cardiovascular system

• The detection of LAA thrombus with [⁶⁴Cu]FBP8 could be used to refine the risk for thrombo-embolism in those at high risk of anticoagulation.

• [⁶⁴Cu]FBP8 could play a valuable role in detecting the source of embolism/thrombus in those with transient ischemic attacks and stroke

• The presence of high fibrin levels ([⁶⁴Cu]FBP8 uptake) in a thrombus could be used to predict the likelihood of successful thrombolysis

Abbreviation list:

PET/MRI

Positron emission tomography / Magnetic resonance imaging

TEE

Transesophageal echocardiography

LAA

Left atrial appendage

HC

Healthy Controls

Afib

Atrial fibrillation

ECG

Electrocardiogram

ROI

Region of interest

SUV

Standardized uptake values

ROC

Receiver operating characteristic curve

TACs

Time activity curves

SEC

spontaneous echo contrast

bSSFP

balanced steady state free precession