Magnetocardiography detects left atrial dysfunction in paroxysmal atrial fibrillation

Abstract

BACKGROUND:

Magnetocardiography (MCG) is a non-invasive technique and to characterize the magnetic field, a pseudo-current conversion was used. The role of MCG in detecting left atrial (LA) dysfunction in patients with paroxysmal atrial fibrillation (PAF) is unknown.

OBJECTIVE:

The aim of this study was to evaluate LA function using MCG in patients with PAF and healthy subjects, to identify possible indices to diagnose PAF.

METHODS:

We enrolled a total of 70 subjects including 26 healthy volunteers (group 1) and 22 marathon runners (group 2) who did not exhibit any cardiac abnormalities, and 22 patients with PAF (group 3) which was documented by electrocardiography (ECG). Spatiotemporal activation graph (STAG) in base-apex and left-right direction was reconstructed. The maximum value of LA pseudo-current under rest and peak exercise were measured between the end of the P wave and beginning of the Q wave.

RESULTS:

LA pseudo-current increase at peak exercise in PAF patients was significantly lower than in healthy volunteers and marathon runners (0.4±0.3 pT in group 3 vs. 0.8±0.3 pT in group 1 vs. 1.1±0.5 pT in group 2, p < 0.001). PAF patients had less pseudo-current increase in STAG at peak exercise than healthy volunteers and marathon runners (46% of 26 PAF patients, 81% of 22 healthy subjects vs. 81% of 22 marathon runners, p = 0.002). Sensitivity, specificity, and the area under the receiver-operator characteristics curve of LA pseudo-current increase at peak exercise for differentiating PAF patients from healthy subjects were 77%, 92%, and 0.896.

CONCLUSIONS:

MCG can provide important non-invasive information for detecting LA dysfunction in PAF patients. Therefore, MCG may help in differentiating PAF patients from healthy subjects.

Keywords

Left atrium dysfunction magnetocardiography paroxysmal atrial fibrillation

1. Introduction

Paroxysmal atrial fibrillation (PAF), a self-terminating recurrent form of cardiac arrhythmia that comprises between 25 % and 60 % of atrial fibrillation (AF), is thought to precede sustained AF, and leads to progressive atrial electrical and structural remodeling. In addition, it has been suggested as a possible cause for cryptogenic strokes [1, 2]. Previous studies reported that PAF is more prevalent than persistent AF in patients with strokes and transient ischemic attacks (TIA) [3]. Despite these serious adverse outcomes, PAF has been underdiagnosed and treated with relatively less effective strategies.

The diagnosis of PAF poses a challenge due to its brief duration, episodic frequency, asymptomatic presentation and the low yield afforded by conventional electrocardiography (ECG) monitoring. Although PAF has been defined by the American Heart Association as “recurrent (two or more) episodes of AF that terminate spontaneously lasting between 30 seconds and less than seven days” [4], there is controversy over the duration and morphology of the ECG data used in defining an event qualifying for AF [5].

On the other hand, left atrial (LA) remodeling such as alterations in LA function has been established to be a hallmark for PAF and plays an important role in thrombogenesis being a potential cardiac source of embolism [6, 7]. Although many echocardiographic studies have been performed to evaluate LA function in PAF, echocardiographic indices have important pitfalls such as obligation of highly trained specialists and suboptimal image quality [8].

Magnetocardiography (MCG) has been known as a unique non-invasive, contactless technique for diagnosing heart disease [9, 10]. A pseudo-current pattern reflecting the bioelectric currents in the heart can be obtained from multichannel MCG data without the need to apply specific source and volume-conductor models [11]. However, the role of MCG in detecting LA dysfunction in patients with PAF is relatively unknown. Recently, we detected the exercise-induced LA activation pattern by MCG using the derivatives of LA pseudo-current and density mapping. Therefore, the aim of this study was to evaluate LA function using MCG in patients with PAF and healthy subjects, to identify possible indices to diagnose PAF.

2. Materials and methods

2.1. Study population

This study was conducted as a prospective registry at Asklepios Hospital Harburg, Hamburg, Germany and at Coburg Hospital, Coburg, Germany. We enrolled a total of 70 subjects including 26 healthy volunteers (group 1) and 22 marathon runners (group 2) who did not exhibit any cardiac abnormalities, who were all in sinus rhythm and act as controls, and 22 patients with PAF (group 3) which was documented by conventional or ambulatory ECG between October 2011 and July 2014. Exclusion criteria included the history of AF, significant valvular heart disease, dyspnea ≥New York Heart Association functional class II, history of ischemic heart disease, left ventricular (LV) ejection fraction <50%, hyperthyroidism, primary pulmonary hypertension, and respiratory disease. At the time of enrollment, clinical information was obtained from hospital records. This study was carried out according to the Declaration of Helsinki guidelines and was approved by the Institutional Review Board ethics committee. The informed consent was obtained from all subjects.

2.2. Magnetocardiographic imaging

All examinations were carried out during sinus rhythm. The MCG recordings were performed using a 64-channel axial gradiometer system in a magnetically shielded room (MSR) (CS-MAG II, BMP GmbH, Hamburg, Germany) [12]. The MCG system utilizes double relaxation oscillation superconducting quantum interference device (DROSQUID) sensors [13]. The average noise spectral density of the entire system in the MSR room is 10 fT/√Hz at 1 Hz and 5 fT/√Hz over 100 Hz. Tangential components of the cardiomagnetic fields were measured which is effective in obtaining the overall heart information with a relatively small area of the sensor array [14]. However, in order to apply the well-known magnetic field map parameters, the tri-polar field map patterns were changed into ordinary dipolar field maps using minimum norm estimation [15]. The signal processing software provided automatic digital filtering, averaging, synthetic gradiometer formation and baseline correction of the acquired recordings. We used the same MCG and protocol at both Asklepios hospital and Coburg hospital.

2.3. MCG data acquisition

The MCG signals were digitally recorded at rest, during every exercise step, and post-exercise for 60 s at a sampling rate of 500 Hz, with the patient in the supine position and the SQUID’s 2-D arrayed sensors positioned close to, but not in contact with, the left chest wall. Stress recordings were acquired by bicycle exercise test in all subjects. The stress test is performed according to the medical guidelines [16]. The aim of the stress test was to record the cardiac magnetic field at maximum physical stress (at individual maximum heart rate) and to compare with those at rest. An independent investigator performed quality evaluation and analysis of ECG and MCG.

2.4. PQ segment fluctuation score

For the calculation of the PQ segment fluctuation score, after averaging and broadband filtering with a binomial bandpass filter (37 Hz – 90 Hz), the fluctuation of the PQ interval (between the beginning of the P wave and beginning of the QRS complex) is quantified by calculating the sum of the absolute values of the differences in neighboring extrema (spans). In addition, the absolute values of the first and the last remaining extrema are added to this sum. Thus the PQ segment fluctuation score is calculated as the multiplication of the determined sum by the number of extrema [17].

2.5. Spatiotemporal activation graph of PQ interval

We used the spatiotemporal activation graph (STAG) to quantify LA activity during the PQ interval. The STAG shows pseudo-current amplitude over time along the two spatial dimensions “base - apex” and “left - right”. If activity increases during stress, this indicates a positive score. A large increase is denominated + 3, a moderate increase + 2 and a slight increase + 1. No change is defined as 0 and a reduced activation during stress is denominated -1. Figure 1 shows the representative case that present the activity in the PQ interval increasing during stress.

2.6. LA pseudo-current under rest and peak exercise between end of p-wave and beginning of QRS complex

Fig. 1.

The representative case of spatiotemporal activation graph in the PQ interval. This figure presents that the activity in the PQ interval is increased during stress.

This is the root mean square (RMS) magnetic field strength under rest and peak exercise between end of P-wave and beginning of QRS complex. What is measured is the largest value of the RMS signal of the magnetic field measured in all sensors in this interval. The unit is picoTesla (pT). We measured the difference between rest and peak exercise in all subjects. Figure 2 shows the representative case of LA pseudo-current increase from rest to peak exercise.

Fig. 2.

The representative case of LA pseudo-current increase under rest and peak exercise. This figure shows LA pseudo-current increase under rest and peak exercise between end of P-wave and beginning of QRS complex. The cursor of top plane is in the PQ segment and the RMS field strength at this position is 0.218 pT, which is the parameter value at rest. The cursor of bottom plane is in the PQ segment and the RMS field strength at this position is 0.577 pT. Therefore, the increase is 0.359 pT.

2.7. Statistical analysis

All statistical analyses were done using SPSS version 19.0 (SPSS, Inc., Chicago, Illinois, USA). Descriptive statistical methods were used to describe the data. Results are presented as mean±standard deviation (SD) for continuous variables and categorical variables as numbers and percentages. We performed that the Gaussian distribution of all samples was tested using Kolmogorov-Smirnov test. Differences between two groups were examined with Student’s t-test or Mann-Whitney U-test (in the cases of non-Gaussian distribution) for continuous variables and χ²-test or Fisher exact test for categorical variables. Differences between multiple groups were examined with Analysis of variance (ANOVA) followed by a multiple comparison test for continuous variables and χ²-test for categorical variables. Receiver operating characteristic (ROC) curve analysis was performed to determine the diagnostic accuracy and optimal cutoff values of MCG parameters which demonstrated detection of PAF. Areas under the ROC curves (AUCs) were compared using the DeLong method. Applicable tests were 2-tailed, and p < 0.05 was considered statistically significant.

3. Results

3.1. Patient characteristics

Table 1 shows baseline clinical characteristics of all patients. There are more males in healthy subjects (group 1) and PAF patients (group 3). The body mass index (BMI) (26.3±2.5 kg/m² in group 3 vs. 23.2±2.0 kg/m² in group 1 vs. 22.7±2.2 kg/m² in group 2, p < 0.001) and systolic blood pressure (SBP) (138±20 mmHg in group 3 vs. 120±9 mmHg in group 1 vs. 118±9 mmHg in group 2, p < 0.001) were significantly higher in the PAF patients (group 3) compared to other groups.

Table 1.
Baseline characteristics of groups

Variables Group 1 (Healthy, n = 26) Group 2 (Runners, n = 22) Group 3 (PAF, n = 22) p-value

Age, years 51.2±8.3 51.9±9.4 56.1±7.3 0.107

Male, n (%) 23 (88) 12 (55) 19 (86) 0.009

BMI, kg/m² 23.2±2.0 22.7±2.2 26.3±2.5 <0.001

SBP, mmHg 120±9 118±9 138±20 <0.001

DBP, mmHg 80±4 78±6 83±19 0.521

HR, beats/min 69±8 66±13 65±11 0.408

Variables	Group 1 (Healthy, n = 26)	Group 2 (Runners, n = 22)	Group 3 (PAF, n = 22)	p-value
Age, years	51.2±8.3	51.9±9.4	56.1±7.3	0.107
Male, n (%)	23 (88)	12 (55)	19 (86)	0.009
BMI, kg/m²	23.2±2.0	22.7±2.2	26.3±2.5	<0.001
SBP, mmHg	120±9	118±9	138±20	<0.001
DBP, mmHg	80±4	78±6	83±19	0.521
HR, beats/min	69±8	66±13	65±11	0.408

Data are mean±SD or number (percentage). PAF = paroxysmal atrial fibrillation; BMI = body mass index; SBP = systolic blood pressure; DBP = diastolic blood pressure; HR = heart rate.

3.2. PQ segment fluctuation score

Table 2 shows the MCG data. The peak value of PQ segment fluctuation score at rest and peak exercise was higher in PAF patients (group 3) compared to the other groups (1.6±1.1% in group 3 vs. 0.7±0.6% in group 1 vs. 0.9±0.9% in group 2, p = 0.003). The relative percent change of PQ segment fluctuation score at rest and peak exercise was not lower in PAF patients (group 3) compared to the other groups (– 18.6±64.9% in group 3 vs. – 62.5±0.6% in group 1 vs. – 55.4±39.3% in group 2, p = 0.003).

Table 2.
Magnetocardiographic data of groups at rest and stress

Variables Group 1 (Healthy, n = 26) Group 2 (Runners, n = 22) Group 3 (PAF, n = 22) p-value

The peak value of PQ fluctuation score, % 0.7±0.6 0.9±0.9 1.6±1.1 0.003

The percent change of PQ segment fluctuation score at rest and stress, % –62.5±0.6 –55.4±39.3 –18.6±64.9 0.003

The increase of STAG of PQ interval at stress, n (%) 21 (81) 21 (81) 12 (46) 0.002

The peak value of LA pseudo-current, pT 0.8±0.3 1.1±0.5 0.4±0.3 <0.001

The difference of LA pseudo-current between rest and stress, pT 0.5±0.3 0.7±0.4 0.2±0.3 <0.001

The percent change of LA pseudo-current at rest and stress, % 151.6±101.8 222.5±132.7 80.2±126.3 0.001

Variables	Group 1 (Healthy, n = 26)	Group 2 (Runners, n = 22)	Group 3 (PAF, n = 22)	p-value
The peak value of PQ fluctuation score, %	0.7±0.6	0.9±0.9	1.6±1.1	0.003
The percent change of PQ segment fluctuation score at rest and stress, %	–62.5±0.6	–55.4±39.3	–18.6±64.9	0.003
The increase of STAG of PQ interval at stress, n (%)	21 (81)	21 (81)	12 (46)	0.002
The peak value of LA pseudo-current, pT	0.8±0.3	1.1±0.5	0.4±0.3	<0.001
The difference of LA pseudo-current between rest and stress, pT	0.5±0.3	0.7±0.4	0.2±0.3	<0.001
The percent change of LA pseudo-current at rest and stress, %	151.6±101.8	222.5±132.7	80.2±126.3	0.001

Data are mean±SD or number (percentage). PAF = paroxysmal atrial fibrillation; STAG = spatiotemporal activation graph; LA = left atrial.

3.3. Spatiotemporal activation graph of PQ interval

In the control groups (group 1 and 2) 81% of subjects showed an increase of STAG at peak exercise compared to only 46% of PAF patients (group 3) (21 subjects (81%) in group 1 vs. 21 subjects (81%) in group 2 vs. 12 patients (46%) in group 3, p = 0.002).

3.4. LA pseudo-current under rest and peak exercise between end of p-wave and beginning of QRS complex

The peak value of LA pseudo-current at peak exercise was lower in PAF patients (group 3) compared to the other groups (0.4±0.3 pT in group 3 vs. 0.8±0.3 pT in group 1 vs. 1.1±0.5 pT in group 2, p < 0.001) (Fig. 3). The difference in LA pseudo-current between rest and peak exercise was also lower in PAF patients (group 3) compared to other groups (0.2±0.3 pT in group 3 vs. 0.5±0.3 pT in group 1 vs. 0.7±0.4 pT in group 2, p < 0.001). Figure 4 shows the changes in LA pseudo-current in correlation to exercise work load stage amongst the three groups. The percent change in LA pseudo-current between rest and peak exercise was lower in PAF patients (group 3) compared to other the groups (80.2±126.3% in group 3 vs. 151.6±101.8% in group 1 vs. 222.5±132.7% in group 2, p = 0.001).

Fig. 3.

The peak value of LA pseudo-current of groups. The peak value of LA pseudo-current at peak exercise was lower in PAF patients (group 3) compared to other groups (p < 0.001).

Fig. 4.

The changes of LA pseudo-current according to exercise stage of groups. These figures show that healthy volunteers and marathon runners had LA pseudo-current increase according to exercise stage, but PAF patients had no definite LA pseudo-current increase according to exercise stage.

3.5. Receiver-operator characteristics curves for detection of PAF

The best cut-off value for the peak value of LA pseudo-current for detection of PAF was 0.5 pT with a sensitivity of 77% and specificity of 92%. The positive predictive value, negative predictive value, and the area under the receiver-operator characteristics (ROC) curve of the peak value of LA pseudo-current for detection of PAF were 51%, 97%, and 0.896 (Fig. 5).

Fig. 5.

The ROC curve of LA pseudo-current for detection of PAF. The positive predictive value, negative predictive value, and the area under the receiver-operator characteristics (ROC) curve of the peak value of LA pseudo-current for detection of PAF were 51%, 97%, and 0.896.

4. Discussion

The results of this study demonstrate the following: 1) MCG detected that patients with PAF had an increased exercise-induced PQ-segment fluctuation score compared to healthy subjects; 2) MCG detected that patients with PAF had a lower exercise-induced atrial pseudo-current mapping (PQ mapping) STAG increase compared to healthy subjects; 3) MCG detected that patients with PAF had a decreased exercise-induced LA pseudo-current increase compared to healthy subjects.

PAF may present as a brief single episode of arrhythmia or as clusters of abnormal rhythm of variable duration, sometimes evolving into a more persistent or permanent form [2]. Therefore, the true prevalence of PAF is unknown and PAF is generally underdiagnosed. However, 6 and 28 % of cryptogenic strokes have been found to be secondary to PAF [18]. Systematic reviews assessing the detection of AF with external ECG monitoring in patients after cryptogenic stroke have shown a detection rate of newly diagnosed AF of 5 to 20% [19]. A recent study showed that AF after cryptogenic stroke was most often asymptomatic and paroxysmal and thus unlikely to be detected by strategies based on symptom-driven monitoring or intermittent short-term recordings [5]. Although this previous study suggested that AF was more frequently detected with insertable cardiac monitoring devices rather than conventional follow-up, these devices are more expensive and invasive.

It is clear that studies of LA function provide new insights into the contribution of LA performance in the aetiology of cardiovascular diseases, such as PAF, and are promising tools for predicting cardiovascular events in a wide range of patients [20]. For the diagnosis of PAF and prediction of AF progression, many recent echocardiographic studies have suggested that echocardiography enables detection of LA dysfunction in patients with PAF [21, 22]. However, echocardiography has important pitfalls such as the need for highly trained specialists and suboptimal image quality [8]. In addition, Tao et al. showed that single-plane tissue-tracking cardiac magnetic resonance (CMR) imaging can measure LA structure and function with high accuracy and reproducibility [23]. However, the quality of LA tissue-tracking of CMR can be affected by image quality, anatomy and slice orientation.

On the other hand, MCG is a non-contact, non-invasive, relatively easy technique for the assessment of electromagnetic activity of the heart. Many studies have demonstrated the potential benefit of MCG for some clinical applications. MCG has been used to non-invasively determine the location of conduction pathways in the heart, making MCG potentially beneficial for the localization of arrhythmia sources for catheter ablation [24, 25]. Therefore, MCG may have a role in diagnosing PAF using a surrogate such as LA dysfunction, however data supporting this are currently limited. Yoshida et al. suggested that MCG can more precisely detect the dominant frequency of AF in the LA and the coronary sinus compared with an ECG, and may be helpful in informing patients about the possible therapeutic risks, benefits, and procedural success rate [26]. However, this study included only patients with persistent AF.

In this study, we used three parameters for cardiac magnetic field analysis in order to detect LA dysfunction for the diagnosis of PAF. 1) We measured two quantitative parameters: the peak value of PQ fluctuation score and LA pseudo-current increase at rest and peak exercise. 2) We also used a qualitative parameter: PQ mapping STAG. In our previous study, the change of ST-segment fluctuation score accurately predicted the presence of hemodynamically significant coronary artery disease [17]. Similarly, the PQ segment is the most sensitive phase to detect the atrial electromagnetic deviation. Our results suggest that the irregular fluctuation of the filtered PQ segment provides a means to identify atrial dysfunction including conduction impairment related to the occurrence of PAF. Specifically, the exercise-induced LA pseudo-current increase between the end of the P wave and beginning of the QRS complex was an independent predictor of PAF. The sensitivity and negative predictive value of this parameter were 92% and 97%, respectively. Although the positive predictive value is rather low with 51%, we can exclude PAF using this parameter as the negative predictive value is very high with 97%. In previous pacing-induced AF models, transition from acute to persistent AF resulted in loss of spatiotemporal organization accompanied by an increase in LA volume [27]. Jacquemet et al. suggested that increased density and length of a collagenous septa reduce the conduction velocity and increase the amount of fractionation and asymmetry in the electrograms, indicating that complex fractionated atrial electrogram formation is consistently associated with a region of fibrosis [28]. Ashihara et al. reported that heterogeneous fibroblast proliferation in the myocardial sheet resulted in the frequent spiral wave breakups, and bipolar electrograms recorded at the fibroblast proliferation area exhibited complex fractionated atrial electrogram formation [29]. Consistent with previous studies, we showed that patients with PAF had decreased exercise-induced atrial pseudo-current mapping (PQ mapping), and STAG increase suggesting electrogram disorganization is related to the advanced fibrosis and impairment of wave propagation.

4.1. Limitations

First, although we excluded patients with coronary artery disease in our study, these MCG parameters suggesting LA dysfunction are not specific to detect PAF. Second, because of its small sample size, our study may be underpowered to demonstrate the clear application of MCG to predict PAF. However, the credibility of this study is strengthened by the higher value of sensitivity and negative predictive value of MCG parameter.

5. Conclusions

MCG can provide important information in detecting LA dysfunction in patients with PAF. Therefore, MCG may help in differentiating PAF patients from healthy subjects.

Conflict of interest

All authors report no relationships that could be construed as a conflict of interest.

References

Seet

, Friedman

, Rabinstein

. Prolonged rhythm monitoring for the detection of occult paroxysmal atrial fibrillation in ischemic stroke of unknown cause. Circulation. 2011;124(4):477–86.

Lévy

, Novella

, Ricard

, Paganelli

. Paroxysmal atrial fibrillation: A need for classification. J Cardiovasc Electro-physiol. 1995;6(1):69–74.

Rizos

, Wagner

, Jenetzky

, Ringleb

, Becker

, Hacke

, et al. Paroxysmal atrial fibrillation is more prevalent than persistent atrial fibrillation in acute stroke and transient ischemic attack patients. Cerebrovasc Dis. 2011;32(3):276–82.

Fuster

, Rydén

, Cannom

, Crijns

, Curtis

, Ellenbogen

, et al.; American College of Cardiology/American Heart Association Task Force on Practice Guidelines; European Society of Cardiology Committee for Practice Guidelines; European Heart Rhythm Association; Heart Rhythm Society. ACC/AHA/ESC 2006 Guidelines for the Management of Patients with Atrial Fibrillation: A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the European Society of Cardiology Committee for Practice Guidelines (Writing Committee to Revise the 2001 Guidelines for the Management of Patients With Atrial Fibrillation): Developed in collaboration with the European Heart Rhythm Association and the Heart Rhythm Society. Circulation. 2006;114(7):e257–354.

Sanna

, Diener

, Passman

, Di Lazzaro

, Bernstein

, Morillo

, et al.; CRYSTAL AF Investigators. Cryptogenic stroke and underlying atrial fibrillation. N Engl J Med. 2014;370(26):2478–86.

Thijssen

, Ausma

, Borgers

. Structural remodelling during chronic atrial fibrillation: Act of programmed cell survival. Cardiovasc Res. 2001;52(1):14–24.

Taguchi

, Takashima

, Hirai

, Fukuda

, Ohara

, Nakagawa

, et al. Significant impairment of left atrial function in patients with cardioembolic stroke caused by paroxysmal atrial fibrillation. Intern Med. 2010;49(16):1727–32.

Howlett

, Hatch

, Alexeenko

, Jabr

, Leatham

, Fry

. Diagnosing Paroxysmal Atrial Fibrillation: Are Biomarkers the Solution to This Elusive Arrhythmia? Biomed Res Int. 2015;2015:910267.

Koskinen

, Lehto

, Väänänen

, Rantonen

, Voipio-Pulkki

, Mäkijärvi

, et al. Measurement and reproducibility of magnetocardiographic filtered atrial signal in patients with paroxysmal lone atrial fibrillation and in healthy subjects. J Electrocardiol. 2005;38(4):330–6.

10.

Jurkko

, Mäntynen

, Tapanainen

, Montonen

, Väänanen

, Parikka

, et al. Non-invasive detection of conduction pathways to left atrium using magnetocardiography: Validation by intra-cardiac electroanatomic mapping. Europace. 2009;11(2):169–77.

11.

Jurkko

, Mäntynen

, Lehto

, Tapanainen

, Montonen

, Parikka

, et al. Interatrial conduction in patients with paroxysmal atrial fibrillation and in healthy subjects. Int J Cardiol. 2010;145(3):455–60.

12.

Lee

, Yu

, Kim

, Kwon

, Kim

. A high-sensitivity magnetocardiography system with a divided gradiometer array inside a low boil-off Dewar. Supercond Sci Technol. 2009;22:114003.

13.

Lee

, Kwon

, Kim

, Park

. Double relaxation oscillation SQUID with high flux-to-voltage transfer and its application to a biomagnetic multichannel system. J Korean Phys Soc. 1998;32:600–5.

14.

Kim

, Lee

, Kwon

, Kim

, Park

. Optimal sensor distribution for measuring the tangential field components in MCG. Neurol Clin Neurophysiol. 2004;2004:60.

15.

Numminen

, Ahlfors

, Ilmoniemi

, Montonen

, Nenonen

. Transformation of multichannel magnetocardiographic signals to standard grid form. IEEE Trans Biomed Eng. 1995;42(1):72–8.

16.

Gibbons

, Balady

, Bricker

, Chaitman

, Fletcher

, Froelicher

, et al. ACC/AHA 2002 Guideline Update for Exercise Testing: Summary Article: A Report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee to Update the 1997 Exercise Testing Guidelines). J Am Coll Cardiol. 2002;40(8):1531–40.

17.

Park

, Shin

, Ann

, Gödde

, Park

, Brachmann

, et al. Validation of magnetocardiography versus fractional flow reserve for detection of coronary artery disease. Clin Hemorheol Microcirc. 2015;59(3):267–81.

18.

Alhadramy

, Jeerakathil

, Majumdar

, Najjar

, Choy

, Saggur M. Prevalence and predictors of paroxysmal atrial fibrillation on Holter monitor in patients with stroke or transient ischemic attack. Stroke. 2010;41(11):2596–600.

19.

Seet

, Friedman

, Rabinstein

. Prolonged rhythm monitoring for the detection of occult paroxysmal atrial fibrillation in ischemic stroke of unknown cause. Circulation. 2011;124(4):477–86.

20.

Hoit

. Left atrial size and function. J Am Coll Cardiol. 2014;63(6):493–505.

21.

Mori

, Kanzaki

, Amaki

, Ohara

, Hasegawa

, Takahama

, et al. Impact of reduced left atrial functions on diagnosis of paroxysmal atrial fibrillation: Results from analysis of time-left atrial volume curve determined by two-dimensional speckle tracking. J Cardiol. 2011;57(1):89–94.

22.

Yoon

, Oh

, Kim

, Park

, Kim

, Park

, et al. Echocardiographic predictors of progression to persistent or permanent atrial fibrillation in patients with paroxysmal atrial fibrillation (E6P Study). J Am Soc Echocardiogr. 2015;28(6):709–17.

23.

Tao

, Ciuffo

, Lima

JAC

, Wu

, Ashikaga

. Quantifying left atrial structure and function using single-plane tissue-tracking cardiac magnetic resonance. Magn Reson Imaging. 2017;42:130–8.

24.

Mori

, Nakaya

. Present status of clinical magnetocardiography. CV World Rep. 1988;1:78–86.

25.

Nomura

, Nakaya

, Saito

, Kishi

, Watatsuki

, Miyoshi

, et al. Noninvasive localisation of accessory pathway by magnetocardiographic imaging. Clin Cardiol. 1994;17(5):239–44.

26.

Yoshida

, Ogata

, Inaba

, Nakazawa

, Ito

, Yamaguchi

, et al. Ability of magnetocardiography to detect regional dominant frequencies of atrial fibrillation. J Arrhythm. 2015;31(6):345–51.

27.

Akar

, Everett

THIV

, Kok

, Moorman

, Haines

. Effect of electrical and structural remodeling on spatiotemporal organization in acute and persistent atrial fibrillation. J Cardiovasc Electrophysiol. 2002;13(10):1027–34.

28.

Jacquemet

, Henriquez

. Genesis of complex fractionated atrial electrograms in zones of slow conduction: A computer model of microfibrosis. Heart Rhyth. 2009;6(6):803–10.

29.

Ashihara

, Haraguchi

, Nakazawa

, Namba

, Ikeda

, Nakazawa

, et al. The role of fibroblasts in complex fractionated electrograms during persistent/permanent atrial fibrillation: Implications for electrogram-based catheter ablation. Circ Res. 2012;110(2):275–84.