Hybrid solid-state SPECT/CT left atrial innervation imaging for identification of left atrial ganglionated plexi: technique and validation in patients with atrial fibrillation

Stirrup J1, Gregg S2, Baavour R3, Roth N3, Breault C3, Agostini D4, Ernst S5,6*, Underwood SR2*

1Department of Cardiology, Royal Berkshire Hospital NHS Foundation Trust, United Kingdom

2Department of Nuclear Medicine, Royal Brompton and Harefield NHS Foundation Trust, United Kingdom

3Spectrum Dynamics Medical, Caesarea, Israel

4Department of Nuclear Medicine, CHU Caen and Normandy University EA 4650, Caen, France

5Department of Cardiology, Royal Brompton and Harefield NHS Foundation Trust, United Kingdom

6Cardiovascular Research Center, Royal Brompton and National Heart and Lung Institute, Imperial College London, UK

*Joint senior authors

Word count: 4083

Address for correspondence:

Dr James Stirrup

Consultant Cardiologist

Royal Berkshire Hospital NHS Foundation Trust

Craven Road

Reading

RG1 5AN

United Kingdom

Email: james.stirrup@royalberkshire@nhs.uk

1

Key words

nuclear medicine; imaging; nervous system, autonomic; atrial fibrillation; ganglionated plexi;

mapping; catheter ablation; mIBG, SPECT/CT, CZT

Acknowledgements

The authors wish to thank U. Voss, Department of Nuclear Medicine, Royal Brompton and Harefield

NHS Foundation Trust, United Kingdom for contributing to data collection; and Samy Bross,

Spectrum Dynamics Medical, Caesarea, Isarel, for contribution to the development of the phantom

model.

2

AbstractBackground

Ablating left atrial (LA) ganglionated plexi (GP), identified invasively by high-frequency stimulation

(HFS) during pulmonary vein isolation (PVI), may reduce atrial fibrillation (AF) recurrence. 123I-

metaiodobenzylguanidine (123I-mIBG) solid-state SPECT LA innervation imaging (LAII) has the spatial

resolution to detect LAGP non-invasively but this has never been demonstrated in clinical practice.

Methods

20 prospective patients with paroxysmal AF scheduled for PVI underwent 123I-mIBG LAII. High-

resolution tomograms, reconstructed where possible using cardiorespiratory gating, were co-

registered with pre-PVI cardiac CT. Location and reader confidence (1 [low] – 3 [high]) in discrete 123I-

mIBG LA uptake areas (DUAs) were recorded and correlated with HFS.

Results

A total of 73 DUAs were identified, of which 59 (81%) were HFS-positive (HFS+). HFS+ likelihood

increased with reader confidence (92% [score 3]). 64% of HFS-negative DUAs occurred over the

lateral and inferior LA. Cardiorespiratory gating reduced the number of DUAs per patient (4 vs. 7,

p=0.001) but improved: HFS+ predictive value (76% vs. 49%); reader confidence (2 vs. 1, p=0.02); and

interobserver, intraobserver and interstudy agreement (κ=0.84 vs. 0.68; 0.82 vs. 0.74; 0.64 vs. 0.53

respectively).

Conclusions123I-mIBG SPECT/CT LAII accurately and reproducibly identifies GPs verified by HFS, particularly when

reconstructed with cardiorespiratory gating.

Abbreviations

AF atrial fibrillation; CT computed tomography; CZT cadmium-zinc-telluride; DMR DUA to

mediastinal ratio; DUA discrete uptake area; ECG electrocardiogram; GP ganglionated plexus; HFS

high frequency stimulation; 123I-mIBG iodine-123 meta-iodobenzylguanidine; SPECT single photon

emission computed tomography;; LA left atrium; LAII left atrial innervation imaging; LV left ventricle;

PVI pulmonary vein isolation; ROS region of search; 99m-Tc technetium-99m

3

Introduction

Pulmonary vein isolation (PVI) is the cornerstone technique for treating patients with symptomatic

paroxysmal atrial fibrillation (AF) resistant to standard pharmacological therapies. Despite advances

in technique since the procedure was described originally (1–3), first-procedure success rate for

abolishing atrial tachyarrhythmia remains only ~60% at one year (4). Ancillary techniques, such as

the addition of linear ablation lines to the roof of the left atrium (LA) or the mitral isthmus, do not

appear consistently to improve outcomes (4). Accordingly, interest has turned to the role of LA

innervation in the modulation of AF substrate (5).

Anatomical studies in human cadavers demonstrate fine networks of interconnecting nerves that

reach the heart via the paravertebral ganglia and thereafter the vagus and sympathetic nerves. In fat

pads on the atrial epicardium, larger “nests” of nerve cell bodies form so-called ganglionated plexi

(GP) containing both sympathetic and parasympathetic nerves. These vary in size but are said

typically to measure in the region of 5-10mm (6,7). Catheter ablation of GPs as an add-on to PVI or

even in isolation in patients with AF has been reported to improve clinical outcomes by several

investigators (8–11) and the presence of residual GP activity after PVI might predict likelihood of

recurrence (12). Localization of GPs has been proposed by invasive multi-site testing with high-

frequency stimulation (HFS) (9,10,13), a process that identifies GPs by their typical vagal

(parasympathetic) response of slowing atrioventricular nodal conduction. The process is time

consuming and takes several hours to map the entire LA. Furthermore, the criteria that define a

positive response to HFS are debated. As sympathetic and parasympathetic fibres co-localise in GPs,

non-invasive imaging of LA sympathetic innervation might identify the same areas in advance of

catheter ablation, potentially shortening procedure time and allowing ablation of paroxysmal AF

with improved outcomes.

Radionuclide imaging using 123I-metaiodobenzylguanindine (mIBG) has been used to assess the

sympathetic nervous system for decades, first for planar imaging of paraganglioma (14) and

subsequently for planar, single photon emission computed tomography (SPECT) and recently

cadmium-zinc-telluride (CZT) imaging of cardiac sympathetic innervation, particularly in heart failure

(15–20). Both indications involve imaging relatively large structures; the small size of LAGPs

precludes reliable identification using conventional Anger gamma camera technology (A-SPECT) due

to limitations in spatial resolution. However, recent years have seen the development of solid-state

gamma camera technology with significantly improved sensitivity, spatial resolution and energy

resolution, allowing high quality SPECT imaging with spatial resolution of <5mm (21). Consequently,

4

solid-state SPECT now has the potential to identify GPs as discrete uptake areas (DUA) on the LA

epicardial surface.

We report our initial experience of solid-state SPECT LA innervation imaging (LAII). Having

demonstrated feasibility in a phantom model (Appendix), we describe methods for image acquisition

and interpretation and thereafter present pilot data in patients with AF, with assessment of inter-

and intra-observer variability, inter-study variability in sequential acquisitions and correlation with

invasive HFS.

Methods

Patient selection

21 patients (13 male, 8 female; median age 61 years [57.2 – 66.3]) with symptomatic paroxysmal AF

were recruited to the pilot study. One or more episode of AF lasting >30 seconds on ambulatory ECG

monitoring was required to constitute the diagnosis, as per European Society of Cardiology

guidelines (22). One patient withdrew prior to imaging, leaving 20 patients for analysis. All patients

were in sinus rhythm at the time of SPECT imaging. Median LA size was 77mL [63.3 – 86.8mL] and

left ventricular (LV) ejection fraction 63% [59.5 – 65%]. The study was approved by Local Research

Ethics Committee, registered with Clinicatrials.gov [Identifier NCT02267889) and all patients signed

written informed consent to participation in the study and to use of their data for research purposes.

SPECT image acquisition

All participants stopped medications known to interfere with mIBG uptake for durations

recommended by published guidelines (19,23) and received thyroid blockade before imaging (24).

Imaging was performed following injection of 370MBq 123I-mIBG on a dedicated cardiac solid-state

SPECT camera with 9-head detector system (D-SPECT, Spectrum Dynamics Medical, Caesarea, Israel).

Images were acquired for 20 minutes with the region of interest set such that the entire heart was

included and the centre of the region of interest was located around the base of the ventricles.

Scans were acquired 4 hours after radionuclide administration in line with previously described

practice (15). For assessment of inter-study reproducibility, 10 patients remained in position on the

camera upon completion of the first acquisition and underwent an immediate second acquisition.

Data were acquired in list mode with (when available) simultaneous cardiorespiratory gating using a

dedicated device (Ivy 7810, Ivy Biomedical Systems Inc, Branford, USA).

Total radiation exposure for the SPECT study was 4.8mSv (23).

5

Cardiac computed tomography

All patients underwent cardiac computed tomography (CT) as part of standard clinical care for

delineation of LA and pulmonary venous anatomy prior to PVI. Images were acquired in diastole and

held-expiration and subsequent tomograms were imported into dedicated software (Shina Systems

Limited, Caesarea, Israel) for cardiac segmentation. Each heart chamber was semi-automatically

partitioned (Figure 1a and 1b) into left and right atrial and ventricular segments, an aortic segment

and a LV myocardial segment. After manual corrections to the segmentation, a representative 3D

surface mesh file was created for each chamber (Figure 1c) that was then used for co-registration

with SPECT tomograms.

Hybrid Image processing

The hybrid image was generated using a dedicated workstation (SUMO D-SPECT, Spectrum Dynamics

Medical, Israel). mIBG data were reconstructed in high resolution. Standard D-SPECT reconstruction

parameters generate a voxel size of 4.92mm3, whereas the high-resolution reconstruction algorithm

generates a voxel size of 2.46mm3, yielding an improved average resolution (full width half maximum

in x and y planes) of 3.9mm compared to 5.6mm with standard D-SPECT reconstruction. Phantom

validation of the high-resolution reconstruction is described in the Appendix.

Whilst recruitment was underway, a software upgrade allowed for the processing of ECG- and

respiratory phase-gated images in order to minimise the effects of cardiorespiratory motion on

spatial localisation of DUAs. SPECT datasets from the final 7 recruited patients were subjected to

further reconstruction using only data acquired in diastole and/or the expiratory phase of the

respiratory cycle. In brief, the ECG-gated LV tomograms and filling curves are reviewed to select the

phases of minimum cardiac motion (Figure 2). Data from these phases are then used to reconstruct

new SPECT tomograms with systolic phases excluded. Similarly, review of the respiratory gating

curve allows selection of the expiratory phase of the respiratory cycle (Figure 2). The first and last

100ms of the selected phase were excluded to ensure complete transition into the expiratory phase.

Where possible, both systolic and inspiratory phases were excluded during SPECT reconstruction in

order to minimise effects of cardiorespiratory motion on DUA localisation. Whilst an ungated study

was reconstructed from 100% of acquired counts, cardiorespiratory gating reduced available counts

for reconstruction to around 30%. However, the 7-fold higher sensitivity of the D-SPECT system

compared to conventional gamma camera technology (25) allows acquisition of higher counts over

the imaging period. LV counts for an ungated study in our cohort was 3.5±1.3 million counts; this fell

to 1.5±0.5 million for a gated study with systolic exclusion only and 1.0±0.2 million counts when full

6

cardiorespiratory gating was applied. For D-SPECT technology, 1 million LV counts appears to be

adequate for reconstructing gated myocardial perfusion scintigrams using 99m-Tc sestamibi (26) and

our data are within a similar range.

The mediastinum served as the reference area for measuring background activity to which other

areas of uptake were compared. This was measured by placing a cuboid in the upper mediastinum of

the SPECT reconstruction, excluding specifically any adjacent lung activity (see Online Resource

Figure 4). The mean activity within this region constituted the mediastinal activity.

SPECT tomograms were automatically co-registered to the CT tomograms using LV uptake as the

SPECT reference and LV myocardial segmentation as the CT reference (3c1d). The SPECT tomograms

may be translated in the (x,y,z) direction in order to achieve ‘best fit’ with the CT-derived myocardial

segmentation but rotation is not possible in the current software version. Co-registration accuracy

was reviewed by scrolling through horizontal long, vertical long and short axis views and, once

satisfactory alignment was confirmed, final hybrid tomograms were generated.

Following registration, the CT-derived LA segment was used as an anatomical constraint to define a

region of search (ROS) around the LA endocardium (Figure 3a and 3b) to facilitate identification of

focal mIBG uptake adjacent to the atria. Predicted errors in registration were as follows: registration

inaccuracies (±5mm), beating heart motion (±3mm), and breathing motion (±3mm). These errors

were combined using the root-sum-square method(27) to generate an overall error of ±7mm. As the

LA endocardium defined the outer bound of the CT-derived LA segment, 3mm was added to the

outer ROS thickness to account for LA myocardium. Hence the respective epi- and endocardial ROS

boundaries were 10mm and 7mm.

Focal increased mIBG activity within the ROS was automatically overlaid on the CT-derived left atrial

surface, generating a hybrid 3D image of left atrial innervation and anatomy. An automated

algorithm identified focal increased mIBG activity within the ROS and these DUAs were reviewed and

then accepted or rejected by the reader, as described below. The reader also reviewed the

tomographic and 3D images and selected any further DUAs not identified by the algorithm. In order

to facilitate DUA identification, the 3D SPECT/CT image was adjusted to exclude all LA activity below

the level of background activity; the upper bound was set individually but was generally in the region

of 15 times background activity. For all identified DUAs, a three-dimensional Gaussian curve was

fitted, with a threshold of 80% maximum of the Gaussian fit used to determine the outer boundary

7

of the DUA (Figure 3c). This threshold was used in order to meet expected anatomical GP size and

was verified by the phantom experiment (see Online Resource). Establishing DUAs in this way

enables definition of DUA size, volume and average activity (corresponding to the average ellipsoid

radius and volume, respectively); and the distance between DUA and LA endocardial surface

(corresponding to the distance between the ellipsoid centre and the closest point of the LA ROS).

DUA average activity was also expressed as a ratio to mediastinal activity (DUA to mediastinal ratio;

DMR).

Selection of DUAs

DUAs were identified and marked according to location by three readers experienced in nuclear

cardiology, reporting separately with cases in randomized order (RB – clinical study; JS, SG –

inter/intra-observer variability). For the purposes of assessing reproducibility, a standardized LA

segment map was created (Figure 4). In brief, the LA was divided into anterior, posterior, lateral,

septal and inferior walls, using the pulmonary veins and the mitral annulus to define the boundaries

between segments. Each segment was further subsegmented to improve granularity of geographic

localization. Four additional segments were created around each of pulmonary vein ostium. In total,

the LA surface was standardized to 32 segments.

Each DUA was assigned a confidence score (1: low; 2: moderate, 3: high) based on discreteness;

extent of overlap with non-LA activity (e.g. LV myocardium, lung) and proximity to known GP

locations (8). Examples are shown in Figure 5. Areas of uptake were categorized as discrete if they

had a well-defined boundary edge in all dimensions that did not extend outside of the Region of

Search. Areas of uptake adjacent to areas of known high activity, such as the basal lateral wall of the

LV, lung or anterior to the oesphagus were scored with lower confidence or excluded from

consideration (Figure 6). DUAs that met all three criteria – discrete, distinct from adjacent

extrcardiac activity and located in an area known to be typical for GPs – were scored with high

confidence (score 3). DUAs meeting only 2 or 1 of the criteria were scored with moderate and low

confidence (score 2 and 1) respectively. DUAs considered too distant from the LA surface – wholly

outside of the region of search – or too close to the mitral annulus were not recorded. DUAs were

marked on a LA segmentation map (Figure 4) along with DMR and confidence score.

Invasive verification of LAGPs

During catheter ablation, DUAs were assessed functionally by HFS. Patients were studied in a fasted

state and under general anaesthesia (intravenous Propofol and Fentanyl infusion). Hybrid SPECT/CT

8

images were imported into the 3D electroanatomical mapping (EAM) system (CARTO 3 or CARTO 3

RMT, all Biosense Webster, Diamond Bar, USA) and merged with the 3D fast anatomical map (FAM).

If AF was not already present, atrial burst pacing was performed to induce sustained arrhythmia. As

a first step, detailed 3D maps were acquired using FAM in combination with user-defined

annotations of complex fractionated atrial electrograms (CFAE). Maps were acquired for all

pulmonary veins (PVs), LA, right atrium (RA) and LA appendage (LAA).

DUA sites were overlaid on the FAM and subjected to HFS according to previously described

techniques (8,28,29). In cases where DUAs were depicted at some distance from the reconstructed

LA surface, care was taken to position the stimulation catheter as close as possible, but without

risking perforation. DUAs considered too close to the ventricular myocardium were not subjected to

HFS in order to avoid induction of ventricular fibrillation by inadvertent ventricular capture. HFS-

positive DUAs were assumed to represent epicardially located LAGPs.

Statistics

Data are presented as median [interquartile range] and, where appropriate, analysed using the

Mann-Whitney test and kappa statistic. Strength of agreement was categorized nominally according

to accepted standard definitions (30).

Results

Inter-observer variability

79 and 83 DUAs were observed by the two readers respectively, with agreement in 62 (75%).

Interobserver agreement for the presence or absence of a DUA in an atrial segment was substantial (

κ=0.73). Median reader confidence was higher for DUAs reported by both observers (3 [2 – 3] vs 2 [1

– 3], p=0.02). Agreement was almost perfect for the studies reconstructed with cardiorespiratory

gating and higher than for those reconstructed without gating (κ=0.84 vs 0.68 [substantial]

respectively).

Intra-observer variability

79 and 72 DUAs were observed in the two reporting session respectively, with agreement in 60

(75%). Intra-observer agreement for the presence or absence of a DUA in an atrial segment was

substantial (κ=0.77). Median reader confidence was higher for DUAs reported in both studies (3 [2 –

3] vs 2 [1 – 2], p<0.001). Agreement was almost perfect for the studies reconstructed with

9

cardiorespiratory gating and higher than for those reconstructed without gating (κ=0.82 vs 0.74

[substantial] respectively).

Reproducibility on serial imaging

10 patients underwent serial scans. 55 and 44 DUAs were identified, with agreement in 33 (60%).

Intra-scan agreement for the presence or absence of a DUA in an atrial segment was substantial (κ

=0.61). Median reader confidence was higher for the DUAs present in both studies (3 [2 – 3] vs 2 [1 –

2], p=0.002). Agreement was substantial for the studies reconstructed with cardiorespiratory gating

and higher than for those reconstructed without gating (κ=0.64 vs 0.53 [moderate] respectively).

Comparison with HFS

After exclusion of DUAs too close to the mitral annulus or considered too distant from the LA to

realistically represent GPs, 73 DUAs were identified and tested by HFS in total (median per patient 4

[3 – 4]), of which 59 (81%) were HFS-positive. Median DUA intensity and DMR were 1540

photons/s/mL [972 – 2014] and 5.2 [3.9 – 7.9]. The likelihood of positive HFS increased with reader

confidence, with 12/18 (67%), 23/29 (79%) and 24/26 (92%) testing positive for confidence scores 1,

2 and 3 respectively.

20/24 (83%) and 39/49 (80%) DUAs were HFS-positive around the pulmonary veins and LA free walls

respectively. 9/14 (64%) HFS-negative DUAs were identified over either the lateral (basal LV,

ligament of Marshall) or inferior wall (oesophageal and pre-aortic activity).

Effects of cardiorespiratory gating

The complete raw data was used to reconstruct images in the first 13 recruited patients (Group 1),

whilst images in the final 7 recruited patients were reconstructed from data in the diastolic and

expiratory phases only (Group 2).

More DUAs were identified per patient in Group 1 than in Group 2 ((7 [6 – 8] vs. 4 [3 – 4.5], p=0.001)

but they had lower predictive value for positive HFS (44/90 (49%) vs. 19/25 (76%) respectively).

Furthermore, reader confidence was lower (1 [0 – 2] vs. 2 [1 – 3] respectively, p=0.02) and more

DUAs were identified as either too distant from LA to represent GPs or too close to LV annulus

(39/90 (43%) vs. 3/25 (12%) respectively).

When these DUAs are removed from analysis, predictive value for positive HFS (40/51 (78%) vs.

19/22 (86%)) and reader confidence (2 [1 – 3] vs. 2 [ 2 – 3], p=0.5) are broadly similar for Groups 1

and 2 respectively.

10

Qualitatively, cardiorespiratory gating improved the discreteness of identified DUAs (Figure 7).

Discussion

We report our pilot experience in hybrid 123I-mIBG solid-state SPECT/CT LA innervation imaging (LAII)

in patients with AF. We demonstrate that DUAs are reproducible on serial acquisitions and show

good inter- and intra-observer reproducibility, particularly when images are reconstructed using

cardiorespiratory gating. Finally, we demonstrate that these regions correspond to GPs verified by

invasive HFS, again with improvement when the SPECT images are reconstructed using

cardiorespiratory gating. mIBG imaging has been performed previously with A-SPECT to assess

patients with cardiac arrhythmia and various underlying heart pathologies (31,32). It predicts

adverse outcomes in patients with heart failure (33,34), after ICD implantation (35–37) and even

predicts outcome of catheter ablation of AF (38). However, these reports have been largely limited

to planar imaging parameters such as heart-to-mediastinum (H/M) ratio and myocardial washout

rate. PET tracers of cardiac innervation also exist and have been studied in similar settings(39,40),

and the use of 18F tracers may allow imaging at a lower cost than current 11C-labelled agents(41). As

in SPECT imaging, PET cardiac innervation studies have focussed on LV myocardial applications. The

better spatial resolution of PET may lend itself to LAII but no studies in this area have yet been

published.

Four aspects of hybrid SPECT/CT LAII in the clinical pilot warrant special focus. First, SPECT and CT

images were co-registered from datasets acquired at separate times, leading to the potential for

misregistration. Although a hybrid D-SPECT/CT device has recently been released, it was not

available commercially at the time of this pilot and hence imaging was necessarily performed on

separate occasions. In order to minimise misregistration, D-SPECT images were acquired supine in a

manner similar to CT and the CT images were acquired in end-diastole in order to match as closely as

possible the cardiac phases used to reconstruct the D-SPECT images. The post-processing software

allowed semi-automated co-registration with manual adjustments, as described above. Other

landmarks beyond the left ventricular myocardium, such as the papillary muscles, were used to

confirm proper SPECT/CT registration. Other modalities for left atrial anatomical assessment were

considered – cardiac MRI(42), intracardiac echocardiography(43) and even radionuclide blood pool

imaging – but cardiac CT is an accepted method for delineating pulmonary venous and left atrial

anatomy prior to ablation and these were performed for clinical reasons in the patients recruited to

this study. We believe that the opportunity for misregistration is no greater than for any other

imaging modality and that registration in our cases was appropriate, particularly when using

cardiorespiratory gating.

11

Second, this reinforces our impression that cardiorespiratory gating appears to be mandatory. Prior

to our capability to reconstruct gated images, more DUAs were distant from the LA surface or too

close to the mitral annulus to safely stimulate. Predictive value for positive HFS rose from 49% to

76% through the use of gating and qualitatively DUA discreteness also improves. We surmise that

removing the blurring effect of cardiorespiratory motion improves spatial localisation of the

relatively small DUAs and hence improves correspondence with invasively defined GP locations.

Accordingly, we recommend that images be reconstructed in the diastolic and expiratory phases to

improve accuracy. Although there may be concern that dividing the data in this way reduces the

number of counts available for image reconstruction, increasing image noise, the 7-fold higher

sensitivity of the D-SPECT camera allows reconstruction of images at a much lower count with

significantly reduced Poisson noise levels(25).

Third, reader confidence plays an important role in the likelihood of a DUA proving HFS-positive. It is

desirable to suppose that GPs are uniformly represented by easily distinguishable discrete areas of

activity on the SPECT images, but this is often not the case. Very discrete areas of mIBG uptake in

locations known to be typical for GPs were easy to call with high confidence. However, although the

highest density of neurons within the GP is at the centre, there is no abrupt boundary defining the

‘edge’ of the plexus. In a number of our cases, LA activity was manifested as a ‘smear’ across the LA

surface that often corresponded to a similar sized area defined by HFS. We surmise that GPs are not

necessarily entirely discrete, as is commonly described. This creates challenges in image

interpretation, particularly as smears of activity may arise from adjacent extra-cardiac activity that

crosses the epicardial boundary of the LA ROS. Unsurprisingly, reader confidence is somewhat lower

when this occurs and there is a correspondingly lower predictive value for positive HFS during

invasive testing. However, even at the lowest level of reader confidence, 67% of DUAs proved to

HFS-positive, suggesting that, at least for now, low-confidence DUAs should still be reported.

The fourth linked issue is that DUA location is an important predictor of likelihood of a GP at that

site. We show that DUAs identified over the lateral and inferior LA walls are less likely to be HFS-

positive. This is due to difficulty in distinguishing true LA epicardial activity from uptake related to

the basal lateral LV wall and ligament of Marshall (lateral LA wall) and oesophageal and pre-aortic

activity (inferior LA wall). DUAs in these regions tend to be less discrete and, as above, reader

confidence in these zones is lower. Although wholly imaging-driven GP ablation is clinically desirable,

our preliminary experience suggests that, even if LAII is validated subsequently in larger trials, DUAs

identified either laterally/inferiorly or with low reader confidence might still require invasive

verification before proceeding to GP ablation.

12

Our results underline that GP size and distribution are highly individual and a simple anatomical

approach to GP ablation may not achieve complete results. We believe the ‘anatomical’ approach

has the potential to overlook GPs not located at typically reported GP sites. Similarly, atrial

myocardium might be ablated when there is no GP at that location, causing unnecessary scarring

and potential negative consequences (pro-arrhythmia or perforation). Individual GP localisation as

demonstrated in our patients might allow safer and more time-efficient GP ablation.

Limitations

Although our results indicate feasibility and reproducibility of SPECT LAII, our pilot study comprises

small numbers, particularly in the inter-study comparison, and a larger study is required to confirm

our findings. Investigations in well-defined, larger patient cohorts with paroxysmal and other forms

of AF are necessary to judge clinical effectiveness. Similarly, the effect of mIBG-guided GP ablation

for management of AF needs investigation in prospective trials, which are under way. HFS served as

the reference standard for invasive GP localisation but the criteria that define a ‘positive’ response

are debated. It is possible that some part of any disagreement between non-invasive and invasive GP

localisation reflects an imperfect gold standard.

Conclusion

Hybrid 123I-mIBG solid-state SPECT/CT LAII is feasible and, in our pilot study, reproducibly identifies

GPs verified by HFS, particularly when performed using cardiorespiratory gating. Reader confidence

is an important predictor of reproducibility and affects the likelihood of a positive HFS result, with

>85% reproducibility and >90% predictive value for positive HFS at the highest level of reader

confidence. Location is also an important determinant, with activity from the ligament of Marshall,

oesophageal plexi and pre-aortic areas probably explaining the lower reader confidence and positive

predictive values in the lateral and inferior LA walls.

New knowledge gained

123I-mIBG solid-state SPECT/CT LAII imaging of left atrial ganglionated plexi is feasible and

reproducible, particularly when cardiorespiratory gating is used to reconstruct the SPECT images.

Compliance with Ethical Standards

Funding: This study was supported by Spectrum Dynamics Medical.

13

Conflict of Interest: JS, SE, DA and SRU receive consultancy fees from Spectrum Dynamics Medical.

JS receives speaker honoraria from Canon Medical Systems Europe (previously Toshiba Medical

Systems Europe). UV was employed by a restricted research grant while NR, RB and CB are

employees of Spectrum Dynamics Medical.

Ethical approval: All procedures performed were in accordance with the ethical standards of the

local Research Ethics Committee (London – Camberwell St Giles REC, reference 14/LO/2207) and

with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

(Clinicatrials.gov Identifier NCT02267889)

Informed consent: Informed consent was obtained from all individual participants included in the

study.

14

References

1. Haïssaguerre M, Jaïs P, Shah DC, Takahashi A, Hocini M, Quiniou G, et al. Spontaneous initiation of atrial fibrillation by ectopic beats originating in the pulmonary veins. N Engl J Med. 1998 Sep 3;339(10):659–66.

2. Haïssaguerre M, Shah DC, Jaïs P, Hocini M, Yamane T, Deisenhofer I, et al. Electrophysiological Breakthroughs From the Left Atrium to the Pulmonary Veins. Circulation. 2000 Nov 14;102(20):2463–5.

3. Wazni OM, Marrouche NF, Martin DO, Verma A, Bhargava M, Saliba W, et al. Radiofrequency Ablation vs Antiarrhythmic Drugs as First-line Treatment of Symptomatic Atrial Fibrillation: A Randomized Trial. JAMA. 2005 Jun 1;293(21):2634–40.

4. Fink T, Schlüter M, Heeger C-H, Lemes C, Maurer T, Reissmann B, et al. Stand-Alone Pulmonary Vein Isolation Versus Pulmonary Vein Isolation With Additional Substrate Modification as Index Ablation Procedures in Patients With Persistent and Long-Standing Persistent Atrial Fibrillation: The Randomized Alster-Lost-AF Trial (Ablation at St. Georg Hospital for Long-Standing Persistent Atrial Fibrillation). Circ Arrhythm Electrophysiol. 2017 Jul 1;10(7):e005114.

5. Lu Z, Scherlag BJ, Lin J, Yu L, Guo J-H, Niu G, et al. Autonomic mechanism for initiation of rapid firing from atria and pulmonary veins: evidence by ablation of ganglionated plexi. Cardiovasc Res. 2009 Nov 1;84(2):245–52.

6. Pauza DH, Skripka V, Pauziene N, Stropus R. Morphology, distribution, and variability of the epicardiac neural ganglionated subplexuses in the human heart. Anat Rec. 2000 Aug 1;259(4):353–82.

7. Armour JA, Murphy DA, Yuan BX, Macdonald S, Hopkins DA. Gross and microscopic anatomy of the human intrinsic cardiac nervous system. Anat Rec. 1997 Feb;247(2):289–98.

8. Po SS, Nakagawa H, Jackman WM. Localization of left atrial ganglionated plexi in patients with atrial fibrillation. J Cardiovasc Electrophysiol. 2009 Oct;20(10):1186–9.

9. Katritsis DG, Giazitzoglou E, Zografos T, Pokushalov E, Po SS, Camm AJ. Rapid pulmonary vein isolation combined with autonomic ganglia modification: a randomized study. Heart Rhythm. 2011 May;8(5):672–8.

10. Katritsis DG, Pokushalov E, Romanov A, Giazitzoglou E, Siontis GCM, Po SS, et al. Autonomic Denervation Added to Pulmonary Vein Isolation for Paroxysmal Atrial Fibrillation: A Randomized Clinical Trial. J Am Coll Cardiol. 2013 Dec 17;62(24):2318–25.

11. Pokushalov E, Romanov A, Artyomenko S, Turov A, Shirokova N, Katritsis DG. Left atrial ablation at the anatomic areas of ganglionated plexi for paroxysmal atrial fibrillation. Pacing Clin Electrophysiol PACE. 2010 Oct;33(10):1231–8.

12. Kurotobi T, Shimada Y, Kino N, Ito K, Tonomura D, Yano K, et al. Features of intrinsic ganglionated plexi in both atria after extensive pulmonary isolation and their clinical significance after catheter ablation in patients with atrial fibrillation. Heart Rhythm. 2015 Mar 1;12(3):470–6.

15

13. Malcolme-Lawes LC, Lim PB, Wright I, Kojodjojo P, Koa-Wing M, Jamil-Copley S, et al. Characterization of the Left Atrial Neural Network and its Impact on Autonomic Modification ProceduresClinical Perspective. Circ Arrhythm Electrophysiol. 2013 Jun 1;6(3):632–40.

14. Jacobson AF, Deng H, Lombard J, Lessig HJ, Black RR. 123I-meta-iodobenzylguanidine scintigraphy for the detection of neuroblastoma and pheochromocytoma: results of a meta-analysis. J Clin Endocrinol Metab. 2010 Jun;95(6):2596–606.

15. Jacobson AF, Senior R, Cerqueira MD, Wong ND, Thomas GS, Lopez VA, et al. Myocardial Iodine-123 Meta-Iodobenzylguanidine Imaging and Cardiac Events in Heart Failure: Results of the Prospective ADMIRE-HF (AdreView Myocardial Imaging for Risk Evaluation in Heart Failure) Study. J Am Coll Cardiol. 2010 May 18;55(20):2212–21.

16. Chirumamilla A, Travin MI. Cardiac applications of 123I-mIBG imaging. Semin Nucl Med. 2011 Sep;41(5):374–87.

17. Raffel DM, Wieland DM. Development of mIBG as a Cardiac Innervation Imaging Agent. JACC Cardiovasc Imaging. 2010 Jan 1;3(1):111–6.

18. Wollenweber T, Bengel FM. Molecular imaging to predict ventricular arrhythmia in heart failure. J Nucl Cardiol Off Publ Am Soc Nucl Cardiol. 2014 Dec;21(6):1096–109.

19. Flotats A, Carrió I, Agostini D, Le Guludec D, Marcassa C, Schäfers M, et al. Proposal for standardization of 123I-metaiodobenzylguanidine (MIBG) cardiac sympathetic imaging by the EANM Cardiovascular Committee and the European Council of Nuclear Cardiology. Eur J Nucl Med Mol Imaging. 2010 Aug;37(9):1802–12.

20. Bellevre D, Manrique A, Legallois D, Bross S, Baavour R, Roth N, et al. First determination of the heart-to-mediastinum ratio using cardiac dual isotope (123I-MIBG/99mTc-tetrofosmin) CZT imaging in patients with heart failure: the ADRECARD study. Eur J Nucl Med Mol Imaging. 2015 Nov;42(12):1912–9.

21. Gambhir SS, Berman DS, Ziffer J, Nagler M, Sandler M, Patton J, et al. A Novel High-Sensitivity Rapid-Acquisition Single-Photon Cardiac Imaging Camera. J Nucl Med. 2009 Apr 1;50(4):635–43.

22. Kirchhof P, Benussi S, Kotecha D, Ahlsson A, Atar D, Casadei B, et al. 2016 ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS. Eur Heart J. 2016 Oct 7;37(38):2893–962.

23. Bombardieri E, Giammarile F, Aktolun C, Baum RP, Bischof Delaloye A, Maffioli L, et al. 131I/123I-metaiodobenzylguanidine (mIBG) scintigraphy: procedure guidelines for tumour imaging. Eur J Nucl Med Mol Imaging. 2010 Dec;37(12):2436–46.

24. Administration of Radioactive Substances Advisory Committee. Notes for Guidance on the Clinical Administration of Radiopharmaceuticals and Use of Sealed Radioactive Sources. Public Health Engl Adm Radioact Subst Advis Comm [Internet]. 2017 Jan [cited 2017 Dec 28]; Available from: https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/590649/ARSAC_NfG_2017.pdf

25. Imbert L, Poussier S, Franken PR, Songy B, Verger A, Morel O, et al. Compared Performance of High-Sensitivity Cameras Dedicated to Myocardial Perfusion SPECT: A Comprehensive Analysis of Phantom and Human Images. J Nucl Med. 2012 Dec 1;53(12):1897–903.

16

26. Nakazato R, Berman DS, Hayes SW, Fish M, Padgett R, Xu Y, et al. Myocardial Perfusion Imaging with a Solid-State Camera: Simulation of a Very Low Dose Imaging Protocol. J Nucl Med. 2013 Mar 1;54(3):373–9.

27. Joint Committee for Guides in Metrology. Evaluation of measurement data - Guide to the expression of uncertainty in measurement (GUM). JCGM [Internet]. 2008 [cited 2018 Aug 29];100. Available from: https://www.bipm.org/en/publications/guides/gum

28. Nakagawa H, Scherlag BJ, Patterson E, Ikeda A, Lockwood D, Jackman WM. Pathophysiologic basis of autonomic ganglionated plexus ablation in patients with atrial fibrillation. Heart Rhythm. 2009 Dec;6(12 Suppl):S26-34.

29. Lemery R, Cleland M, Bernick J, Wells GA. Contact force mapping and voltage thresholds during high-frequency stimulation of human cardiac ganglionated plexuses†. Eur Eur Pacing Arrhythm Card Electrophysiol J Work Groups Card Pacing Arrhythm Card Cell Electrophysiol Eur Soc Cardiol. 2015 Apr;17(4):552–8.

30. Landis JR, Koch GG. The Measurement of Observer Agreement for Categorical Data. Biometrics. 1977;33(1):159–74.

31. Akutsu Y, Kaneko K, Kodama Y, Li H-L, Suyama J, Shinozuka A, et al. Iodine-123 mIBG Imaging for Predicting the Development of Atrial Fibrillation. JACC Cardiovasc Imaging. 2011 Jan;4(1):78–86.

32. Akutsu Y, Kaneko K, Kodama Y, Li H-L, Asano T, Suyama J, et al. Usefulness of severe cardiac sympathetic dysfunction to predict the occurrence of rapid atrial fibrillation in patients with Wolff-Parkinson-White syndrome. Am J Cardiol. 2013 Sep 1;112(5):688–93.

33. Verschure DO, Veltman CE, Manrique A, Somsen GA, Koutelou M, Katsikis A, et al. For what endpoint does myocardial 123I-MIBG scintigraphy have the greatest prognostic value in patients with chronic heart failure? Results of a pooled individual patient data meta-analysis. Eur Heart J Cardiovasc Imaging. 2014 Sep;15(9):996–1003.

34. Shah AM, Bourgoun M, Narula J, Jacobson AF, Solomon SD. Influence of ejection fraction on the prognostic value of sympathetic innervation imaging with iodine-123 MIBG in heart failure. JACC Cardiovasc Imaging. 2012 Nov;5(11):1139–46.

35. Nishisato K, Hashimoto A, Nakata T, Doi T, Yamamoto H, Nagahara D, et al. Impaired cardiac sympathetic innervation and myocardial perfusion are related to lethal arrhythmia: quantification of cardiac tracers in patients with ICDs. J Nucl Med Off Publ Soc Nucl Med. 2010 Aug;51(8):1241–9.

36. Marshall A, Cheetham A, George RS, Mason M, Kelion AD. Cardiac iodine-123 metaiodobenzylguanidine imaging predicts ventricular arrhythmia in heart failure patients receiving an implantable cardioverter-defibrillator for primary prevention. Heart Br Card Soc. 2012 Sep;98(18):1359–65.

37. Boogers MJ, Borleffs CJW, Henneman MM, van Bommel RJ, van Ramshorst J, Boersma E, et al. Cardiac sympathetic denervation assessed with 123-iodine metaiodobenzylguanidine imaging predicts ventricular arrhythmias in implantable cardioverter-defibrillator patients. J Am Coll Cardiol. 2010 Jun 15;55(24):2769–77.

17

38. Arimoto T, Tada H, Igarashi M, Sekiguchi Y, Sato A, Koyama T, et al. High washout rate of iodine-123-metaiodobenzylguanidine imaging predicts the outcome of catheter ablation of atrial fibrillation. J Cardiovasc Electrophysiol. 2011 Dec;22(12):1297–304.

39. Fallavollita JA, Heavey BM, Luisi AJ, Michalek SM, Baldwa S, Mashtare TL, et al. Regional Myocardial Sympathetic Denervation Predicts the Risk of Sudden Cardiac Arrest in Ischemic Cardiomyopathy. J Am Coll Cardiol. 2014 Jan;63(2):141–9.

40. Fujita W, Matsunari I, Aoki H, Nekolla SG, Kajinami K. Prediction of all-cause death using <Superscript>11</Superscript>C-hydroxyephedrine positron emission tomography in Japanese patients with left ventricular dysfunction. Ann Nucl Med. 2016 Aug 1;30(7):461–7.

41. Kobayashi R, Chen X, Werner RA, Lapa C, Javadi MS, Higuchi T. New horizons in cardiac innervation imaging: introduction of novel <Superscript>18</Superscript>F-labeled PET tracers. Eur J Nucl Med Mol Imaging. 2017 Dec 1;44(13):2302–9.

42. Hauser TH, Peters DC, Wylie JV, Manning WJ. Evaluating the left atrium by magnetic resonance imaging. Europace. 2008 Nov;10(Suppl 3):iii22–7.

43. Biermann J, Bode C, Asbach S. Intracardiac Echocardiography during Catheter-Based Ablation of Atrial Fibrillation. Cardiol Res Pract [Internet]. 2012 [cited 2018 Aug 29];2012. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3368317/

18

Figure legends

Figure 1 Transaxial cardiac CT images before (a) and after (b) segmentation into standard cardiac compartments. A 3-dimensional mesh is generated for each compartment (c), with the LA compartment used to generate a region of search for areas of discrete 123I-mIBG uptake. (d) Co-registration of nuclear and CT images using LV uptake and the CT-derived LV segment. Ao aorta; LA left atrium; LV left ventricle; RV right ventricle

Figure 2 Cardiorespiratory gating. Respiratory gating allows identification of the expiratory phase, whilst ECG gating allows selection of the diastolic phase of the cardiac cycle. SPECT images are reconstructed using only counts detected during expiration and diastole (yellow boxes). EXP expiratory; INSP inspiratory; LV left ventricular

Figure 3 (a, b) Co-registered SPECT mIBG and cardiac CT images in transverse (a) and coronal (b) planes. The LA region of search is defined (solid lines, arrows) by the CT-derived LA compartment. (c) Color-coded 3D maps of 123I-mIBG uptake and distribution on CT-derived LA compartments, as well as 3D ellipsoidal DUA representations (ellipsoids, arrows). LSPV left superior pulmonary vein; LIPV left inferior pulmonary vein; RSPV right superior pulmonary vein; RIPV right inferior pulmonary vein; DUA discrete uptake area; LA left atrium; LV left ventricle

Figure 4 Standardised LA segmentation map. The centre of the map, bounded by the pulmonary veins, represents the posterior LA wall. Solid lines demarcate boundaries between the different LA walls whilst dashed lines subdivide the walls to provide granularity of recorded DUA locations. A separate scoring area is included for the LA appendage (dashed box). Various quality assurance (QA) parameters are also recorded. DUAs are marked according to location on 3D morphological-functional fusion maps, along with reader confidence and DMR. LUPV left upper pulmonary vein; LLPV left lower pulmonary vein; RUVP right upper pulmonary vein; RLPV right lower pulmonary vein; DUA discrete uptake area; DMR DUA to mediastinal activity ratio

Figure 5 (a) DUA on the anterosuperior LA surface (arrow), compatible with the typical location of the anterior descending ganglionated plexus. There is also low-level activity anteroseptal to the right superior pulmonary vein (asterisk), although this is much less discrete. Although this could represent activity in the posterior right atrial GP, the confidence is lower. (b) DUA on the anteroseptal LA surface (arrow), compatible with the typical location of the superior right atrial ganglionated plexus. The DUA can be localised more confidently from the coronal co-registered SPECT/CT image (c, arrow) and is located at the superior cavoatrial junction. LS(U)PV left superior (upper) pulmonary vein; RS(U)PV right superior (upper) pulmonary vein; RLPV right lower pulmonary vein; LAA LA appendage; Ant Anterior wall of left atrium; Sept Atrial septum; MA mitral annulus; RA right atrium; Ao aorta; LV left ventricle; SVC superior vena cava

Figure 6 (a) Large smear of activity on the inferoposterior LA surface (left, dotted bracket), with multiple potential DUAs. Review of the transverse and sagittal co-registered SPECT/CT images (right) shows a chain of activity running anterior to the spine over the oesophagus (arrow [top] and dotted bracket [bottom]), suggesting that this activity is most likely to be

19

extracardiac. (b) Encroachment of activity from the basal lateral wall of the left ventricle on to the lateral wall of the left atrium (arrows). LSPV left superior pulmonary vein; RSPV right superior pulmonary vein; RIPV right inferior pulmonary vein; LA left atrium; LAA left atrial appendage; LV left ventricle; S spine

Figure 7 Effects of cardiorespiratory gating. (a) Without gating (top), low intensity activity is seen on the posterior wall (asterisk) and adjacent to the RIPV (arrow). Reconstruction of images using on the diastolic and expiratory phases (bottom) shows that the posterior wall activity disappears and the RIPV activity intensifies (arrow). (b) Without gating (top), there is intense activity running down from LAA to the mitral annulus (bracket) that makes detection of a DUA on the superior aspect of the LA difficult to detect (asterisk). There is also a DUA on the septum bounded also by low-intensity uptake running down to the mitral annulus (arrow). After gating (bottom), the superior DUA is easier to appreciate (asterisk) and there is little activity between the LAA and the mitral annulus (bracket). The septal DUA has also been rendered more discrete (arrow). LSPV left superior pulmonary vein; LIPV left inferior pulmonary vein; RSPV right superior pulmonary vein; RIPV right inferior pulmonary vein; LAA LA appendage; DUA discrete uptake area

20