Advertisement
Journal Home
Search for

Volume 33, Issue 1, Pages 1-6 (January 2009)


View previous. 2 of 11 View next.

Comparison of current density viability imaging at rest with FDG-PET in patients after myocardial infarction

M. GoernigaCorresponding Author Informationemail address, J. Haueisenbc, J. Schreiberb, U. Ledera, H. Hänninende, T. Mäkeläg, P. Takaladg, J. Nenonendg, K. Lauermaf, J. Knuutih, M. Mäkijärvie, L. Toivonene, T. Katilag

Received 15 May 2008; accepted 22 September 2008.

Abstract 

The assessment of myocardial viability is a major diagnostic challenge in patients with coronary artery disease (CAD) after myocardial infarction. Novel threedimensional current density (CD) imaging algorithms use high-resolution magnetic field mapping to determine the electrical activity of myocardial segments at rest. We, for the first time, compared CD activity obtained with several algorithms to 18-F-fluoro-deoxyglucose positron emission tomography (FDG-PET) in evaluation of myocardial viability. Magnetic field maps were obtained in nine adult patients (pt) with CAD and a history of infarction. The criterion for non-viable myocardium was an FDG-PET uptake with less than 45% of the maximum in the respective segments. CD imaging was applied to the left ventricle by using six different methods to solve the inverse problem. Mean CD activity was calculated for a close meshed grid of 90 locations of the left ventricle. A cardiologist compared bull's eye plots of CD and FDG-PET activity by eye. Spearman's correlation coefficients and specificity at a given level of sensitivity (70%) were calculated. Bull's eye plots revealed a significant correlation of CD/PET in 5 pt and no correlation in 3 pt. One pt had a negative correlation. The six different CD reconstruction methods performed similar. While CD reconstruction has the principal potential to image viable myocardium, we found that the reconstructed CD magnitude was low in scar segments but also reduced in some segments with preserved metabolic activity under resting conditions. New vector measurement techniques, the use of additional stress testing and advances in mathematical methodology are expected to improve CD imaging in future.

Article Outline

Abstract

1. Introduction

2. Methods

2.1. Patients

2.2. MRI

2.3. PET

2.4. MCG

2.5. CD reconstruction

2.6. Comparison of CD versus PET

3. Results

3.1. CD methods

3.2. Comparison of CD versus PET

3.3. CD scar detection

4. Discussion

Acknowledgment

References

Biography

Copyright

1. Introduction 

return to Article Outline

In patients with coronary artery disease after myocardial infarction and left ventricular dysfunction the evaluation of the extent of viable myocardium still is a major clinical challenge. Many studies show that the amount of viable myocardium prior to revascularization procedures correlates with both short- and long-term survival after coronary bypass surgery [1].

Therefore, noninvasive assessment of myocardial viability is of paramount clinical importance prior to planning revascularization strategies. Diagnostic methods utilize catecholaminergic recruitment of contractile myocardium (dobutamine stress echocardiography), tracer uptake at rest and exercise by Single Photos Emission Tomography (SPECT), contrast agent uptake by late enhancement magnetic resonance imaging (MRI) and metabolic activity by 18-F-fluoro-deoxyglucose positron emission tomography (FDG-PET). Whereas exercise electrocardiography is the key procedure in detecting ischemia, lack of 3D information and inability to localize ischemia hinders the use of electrocardiographic measurements in localizing viable myocardium in infarcted myocardium prior to revascularization [2].

Loss of electrical activity may be assumed to predict functional damage in a similar fashion as metabolic alterations detects abnormal glucose utilization and the lacking of contractile reserve during low-dose catecholamine infusion [3]. The beat-to-beat signal of the MCG has the same principal shape as the electrocardiogram (ECG), but its information content might be different, which can be explained by physical and methodical difference between MCG and ECG [4], [5] Therefore, solving the inverse problem by advanced current density (CD) reconstruction methods for the entire left ventricle could, thus, identify viable myocardial segments based on electrical activity.

The aim of our study was to test the clinical utility of different CD imaging methods developed recently. FDG-PET was our gold standard in viability assessment [6].

2. Methods 

return to Article Outline

2.1. Patients 

We studied 9 patients (pt) (mean age of 69years, 7 males, 2 females) suffering from CAD and having a history of myocardial infarction. The presence of CAD was proven by coronary angiography and wall motion abnormalities in left ventricular angiography [7]. All pt underwent MRI, magnetocardiography (MCG) and FDG-PET imaging within 10days. 7/9 pt had an FDG-PET uptake of less than 45% of the maximum uptake in at least one segment.

2.2. MRI 

MRI data were recorded with a 1.5 T Siemens Magnetom Vision (Siemens, Erlangen, Germany) at the Department of Radiology at Helsinki University Central Hospital (HUCH). 39 ECG-gated contiguous transversal images were recorded during free respiration using a TurboFLASH sequence. The pixel size and the slice thickness were 1.95mm×1.95mm and 10mm, respectively. Patient-specific boundary element models (including triangulated surfaces of the torso, the left and right lungs, the left and right ventricles) were constructed out of the MR datasets [8].

2.3. PET 

PET data were recorded with an ECAT 931/08-12 (Siemens/CTI, Knoxville, USA) PET scanner at the Turku PET Centre (Finland). A series of 16 contiguous transmission and emission images were recorded. Transmission images were used for attenuation correction of emission images and also provided topographical information that was utilized for registration purposes. For both transmission and emission images, the pixel size and the slice thickness were 2.41mm×2.41mm and 6.75mm, respectively.

The PET images were assessed by an experienced cardiologist, who was blinded to the classification of the results obtained by CD reconstruction. Viability was assumed for myocardial segments with an uptake of more than 45% of the maximum of any of the segments [9].

2.4. MCG 

Magnetic field maps were recorded in the Biomag Laboratory at Helsinki University Central Hospital using a 67-channel biomagnetometer with 7 axial and 60 planar first-order gradiometers (Neuromag, Helsinki, Finland). Recordings were taken at rest in supine position. The sampling rate was 1000Hz, the signals were band-pass filtered at 0.03–300Hz.

2.5. CD reconstruction 

CD imaging was performed using a commercial software package (Curry® V4.51, Neuroscan, USA) commonly applied in the neurosciences. We compared the L2, L1, L1.7, L1.5, L1.3 and the L2-LORETA algorithm [10], where the numbers denote the norm parameter. The sequential QRS signals were used for the CD calculations (Fig. 1). The CD was calculated for a source space consisting of 90 locations of the left ventricle for each instant in time during the QRS interval.


View full-size image.

Fig. 1. Individually constructed model of the human body (a) including the torso, the lungs and the surface of the left ventricle. The latter was used as source space. Measured magnetic field data of selected channels (b) with the time interval analyzed (QRS complex, gray). CD reconstruction result (c) with discretization points (black dots) and the normalized CD scaling (d).


The maximum CD at any instant in time of the QRS complex was calculated for each point and plotted using the common bull's eye method (Fig. 2).


View full-size image.

Fig. 2. CD bull's eye plot. Time course of the CD activity in an anteroseptal (a) and an apical location (b) during the QRS. The maximum CD activity value at any instant in time was assigned to the respective point in the CD bull's eye image (c). CD activity was scaled to the maximum activity (d).


2.6. Comparison of CD versus PET 

CD and PET images were normalized to the respective maximum intensities. An experienced nuclear cardiologist compared the images by eye. We plotted FDG-PET uptake versus CD activity (scatter plots), and for quantitative evaluation Spearman's correlation coefficients were computed.

Furthermore the capability of CD reconstruction for detecting scarred myocardium was tested. Therefore, we defined a sensitivity of 70% and calculated the corresponding specificity for the identification of myocardial segments with less than 45% FDG-PET uptake.

3. Results 

return to Article Outline

3.1. CD methods 

We found the six different CD methods performing similar with respect to the general distribution of the spatial zones of high and low current in the bull's eye plots for each patient (Fig. 3). The contrast of the images increased with decreasing L-norm (L2–L1), thus showing a merely blurred image with L2-norm and edgy patterns with L1-norm (Fig. 3b). The L2-LORETA-norm CD reconstructions did hardly differ from the L2-norm CD reconstructions when analyzing by eye.


View full-size image.

Fig. 3. Example of the comparison of the PET image (a) with the different CD images (b) (pt 4, see Table 1). Both PET and CD activity were scaled to the respective maximum (c).


3.2. Comparison of CD versus PET 

Expert analysis by eye revealed similarity of CD and PET bull's eye images in 5/9 pt. One of the remaining 4 pt had opposite patterns, while 3 pt showed clear match (Fig. 4, Table 1).


View full-size image.

Fig. 4. Bull's eye plots for PET and CD reconstruction (L1.5-norm) and scatter plots for all patients. Spearman's correlation coefficients (r) and their levels of significance (p).


Table 1.

Ejection fraction, coronary vessels involved, result of the visual comparison and the correlation coefficient between PET and CD activity.

Pt no.EFCoronory vessels involvedBull's eye visual CD versus PETCorrelation coefficient CD versus PET
1453No correlation−0.003
2362 (LCX, RCA)Positive correlation0.517*
3402 (LAD, RCA)Positive correlation0.587*
4402 (LAD, LCX)Positive correlation0.489*
5323Inverse correlation−0.475*
6501 (RCA)No correlation−0.127
7803No correlation0.184
8573Positive correlation0.467*
9303Positive correlation0.254*
*

A significance level of 5%.

Visual analysis was confirmed by correlation analysis (Table 1). Correlation coefficients ranged from −0.5 to 0.6 (4 pt: moderate correlation of about 0.5; 1 pt: weak correlation of about 0.25; 3 pt: no correlation; 1 pt: moderate negative correlation of about −0.5).

3.3. CD scar detection 

The specificities for the detection of scarred myocardium (i.e. PET activity <45%) are given in Table 2. Among the different CD reconstruction methods the L1.5-norm method performed best with an average specificity of 53.3%. Looking at the CD method performing best in the individual subject (Table 2, cells shaded) it became obvious that there was no systematic specificity ranking of the different methods. When selecting the method with maximum average specificity (L1.5-norm) a specificity of above 70% was seen in 4/9 pt.

Table 2.

Specificity of the different CD reconstruction methods for detecting non-viable myocardium.

The sensitivity was set to 70%. The best value for each patient has gray shading.

4. Discussion 

return to Article Outline

Our study demonstrates that the diagnostic performance of the used CD imaging methods is not sufficient for myocardial viability testing under resting conditions. According to the underlying normalization principle, the L1-norm presented sharply demarcated areas of activation (bright areas, Fig. 3), whereas the L2-norm produced a smooth transition from low to high CD activity regions. However, the basic CD image was similar for all CD methods.

We found a match of FDG-PET uptake with CD activity in 5 of 9 pt. FDG-PET glucose uptake indicates both function of myocardial muscle mass of the different segments as well as viability in contrast to no uptake in scar regions. CD activity describes electric currents in the myocardium. These matching cases indicate a link between the two imaging principles which may constitute the basis of future investigations.

The technology underlying the study is subject to some limitations. Although our modeling is based on realistic geometries, it does not take fiber anisotropy into account. The inclusion of fiber anisotropy can potentially improve the CD reconstruction results [11], [12]. New vector measurement techniques [13] and advanced inverse algorithms, which are currently under development, may additionally improve the CD reconstruction results.

One major limitation of this study is the heterogeneity and small size of the study group. Our patients, however, reflect the average patient in a clinical setting. A prospective study is necessary which investigates the CD status in connection with recovery of the left ventricular function after revascularization and inclusion of stress/rest perfusion data to evaluate the effects of ischemia, stunning and hibernating on CD reconstruction. A larger patient study group with well-defined “punch hole” myocardial infarction and single vessel disease would increase statistical power and support a unique interpretation of the results.

Our study compared CD under resting conditions and PET images. Benchmarking of a variety of CD methods with PET as the gold standard for myocardial viability imaging uncovered an insufficient diagnostic performance of the inverse methods applied. Taking into account the high dynamics in the research on inverse methods, this status might improve soon, especially with the additional use of stress testing.

Acknowledgements 

return to Article Outline

This work was supported by the German Academic Exchange Program (DAAD) and the Academy of Finland.

References 

return to Article Outline

[1]. [1]Schinkel AF, Bax JJ, Poldermans D, Elhendy A, Ferrari R, Rahimtoola SH. Hibernating myocardium: diagnosis and patient outcomes. Curr Probl Cardiol. 2007;32:375–410. Abstract | Full Text | Full-Text PDF (896 KB) | CrossRef

[2]. [2]Camici PG, Prasad SK, Rimoldi OE. Stunning, hibernation, and assessment of myocardial viability. Circulation. 2008;117:103–114. CrossRef

[3]. [3]Janse MJ, Wit AL. Electrophysiological mechanisms of ventricular arrhythmias resulting from myocardial ischemia and infarction. Physiol Rev. 1989;69:1049–1069. MEDLINE

[4]. [4]Lant J, Stroink G, ten Voorde B, Horacek BM, Montague TJ. Complementary nature of electrocardiographic and magnetocardiographic data in patients with ischemic heart disease. J Electrocardiol. 1990;23:315–322. MEDLINE | CrossRef

[5]. [5]Liehr M, Haueisen J, Goernig M, Seidel P, Nenonen J, Katila T. Vortex shaped current sources in a physical torso phantom. Ann Biomed Eng. 2005;33:240–247. MEDLINE | CrossRef

[6]. [6]Hartiala J, Knuuti J. Imaging of heart by MRI and PET. Ann Med. 1995;27:35–45. MEDLINE | CrossRef

[7]. [7]Lauerma K, Niemi P, Hänninen H, Janatuinen T, Voipio-Pulkki L, Knuuti J, et al. Multimodality MR imaging assessment of myocardial viability: combination of first-pass and late contrast enhancement to wall motion dynamics and comparison with FDG-PET. Radiology. 2000;217:729–736. MEDLINE

[8]. [8]Haueisn J, Schreiber J, Brauer H, Knösche TR. The dependence of the inverse solution accuracy in magnetocardiography on the boundary element discretization. IEEE Trans Mag. 2002;38:1045–1048.

[9]. [9]Gerber BL, Ordoubadi FF, Wijns W, Vanoverschelde JLJ, Knuuti MJ, Janier M, et al. Positron emission tomography using 18-F-fluoro-deoxyglucose and euglycaemic hyperinsulinaemic glucose clamp: optimal criteria for the prediction of recovery of post-ischaemic left ventricular dysfunction. Results from the European Community Concerted Action Multicenter study on use of 18-F-fluoro-deoxyglucose positron emission tomography for the detection of myocardial viability. Eur Heart J. 2001;22:1691–1701. CrossRef

[10]. [10]Fuchs M, Wagner M, Köhler T, Wischmann HA. Linear and nonlinear current density reconstruction. J Clin Neurophysiol. 1999;16:267–295. MEDLINE | CrossRef

[11]. [11]Haueisen J, Tuch DS, Ramon C, Schimpf PH, Wedeen VJ, George JS, et al. The influence of brain tissue anisotropy on human EEG and MEG. Neuroimage. 2002;15:159–166. MEDLINE | CrossRef

[12]. [12]Ramon C, Schmpf PH, Wang Y, Haueisen J, Ishimaru A. The effect of volume currents due to myocardial anisotropy on body surface potentials. Phys Med Biol. 2002;47:1167–1184. MEDLINE | CrossRef

[13]. [13]Kobayashi K, Uchikawa Y. Estimation of multiple sources using spatio-temporal data on a three-dimensional measurement of MEG. IEEE Trans Mag. 2001;37:2915–2917.

Dr. Matthias Goernig, received a M.D. at the University of Hamburg, Germany, in 1993. From 1993 to 1994 he worked as an Internship at the Max von Pettenkofer Institute for Microbiology in Munich and at the National Cancer Institute in Bethesda, USA. From 1995 to 1998 he was working at the Clinic for Dermatology, Friedrich-Schiller-University and in 1999 at the Dermatology Hospital of Leutenberg/Thuringia, Germany. Since 2000 he is at the Clinic of Internal Medicine (Department of Cardiology), Friedrich-Schiller-University. His research interests are in biological signal analysis and the clinical applications of bioelectric and biomagnetic fields.

Prof. Jens Haueisen received a M.S. and a Ph.D. in electrical engineering from the Technical University Ilmenau, Germany, in 1992 and 1996, respectively. From 1996 to 1998 he worked as a Post-Doc and from 1998 to 2005 as the head of the Biomagnetic Center, Friedrich-Schiller-University. Since 2005 he is professor of biomedical engineering and directs the Institute of Biomedical Engineering and Informatics at the Technical University Ilmenau, Germany. His research interests are in the numerical computation of bioelectric and biomagnetic fields and biological signal analysis.

Dr. Joerg, Schreiber, was born in Suhl, Germany, in 1976. He received his Physics-Diploma from the Friedrich-Schiller-University, Jena, Germany Jena, Germany, in 2002 and a Ph.D., from the University Munich, Germany, in 2006. Since 2002 he his working with the Max Planck Institute of Quantum Optics in Garching, Germany. His research interests focus on high-intensity laser physics.

Dr. Uwe Leder was born in Zella-Mehlis, Thuringia, Germany, in 1964. He received his M.D. from the Friedrich Schiller University, in 1992. In 2000 he became a cardiologist at the University Hospital of Jena. In 2002 he received his habilitation, facultas docendi und venia legendi for Internal Medicine. Since 2003, he worked for the University administration of the University Hospital of Jena, Germany. His research interests focus on biological signal analyses and magnetic field imaging.

Dr. Helena Hänninen, specialist in internal medicine and cardiology (born 1 April 1968) is currently working as consulting cardiologist in Helsinki University Hospital. Her research interests focus on coronary artery disease, body surface potential mapping, magnetocardiography and cardiac imaging.

Dr. Timo Mäkelä, was born in Evijarvi, Finland, in 1970. He received a degree from the University of Oulu, Finland, in 1986 and a Ph.D., from the Helsinki University of Technology, in 2003. Since 2003, he has been working as a hospital physicist in Internal Medicine Cardiology Department of Oulu University Hospital. His research interests focus on cardiac image and signal processing and biomedical engineering.

Dr. Pekka Takala was born in Helsinki, Finland, in 1970. He received a degree from the Helsinki University of Technology, Finland, in 1997 and a Ph.D., from the same University, in 2001. Since 2001, he has been with GE Healthcare, where he is senior engineer in patient monitoring R&D. His research interests focus on algorithm development and signal processing.

Dr. Jukka Nenonen was born in Joensuu, Finland, in 1961. He received the Doctor of Technology degree in technical physics at Helsinki University of Technology in 1992. He was working as a researcher at the Laboratory of Biomedical Engineering of Helsinki University of Technology until 2003, and thereafter as the method development manager with Elekta Neuromag Oy, Helsinki, Finland. His research interests include modeling, inverse problem, and signal processing bioelectricity and biomagnetism.

Prof. Kirsi Lauerma, was born in Kauhava, Finland, in 1961. She received a medical degree from the University of Helsinki, Finland, in 1988 and a Ph.D., from the same University, in 1998. Since 1994, he has been with the Central University Hospital of Helsinki, Finland, where she is associate professor in radiology and a chief in pediatric radiology. Her research interests focus on MRI of the heart.

Prof. Juhani Knuuti, (8 April 1960) is director and professor of Turku PET Centre, University of Turku, Finland. His main research field is noninvasive cardiovascular imaging especially using PET, SPECT, MRI and CT. He has published over 200 articles in peer-reviewed journals. He has actively worked in several associations of cardiology and nuclear medicine as well as editorial boards of scientific journals. He is currently a chairman of the Working Group of Nuclear Cardiology and Cardiac CT at European Society of Cardiology and member of the Cardiovascular Committee of EANM.

Prof. Markku Mäkijärvi, was born in Savonlinna, Finland, in 1958. He received a M.D. degree from the University of Helsinki, Finland, in 1984 and a Ph.D., from the same University, in 1993. Since 1987, he has been working in the Helsinki University Central Hospital, where he is chief physician and consultant of cardiology. His research interests are cardiac arrhythmias and ischemic heart disease.

Prof. Lauri Toivonen (7 January 1950) is associate professor in Helsinki University and director of the Cardiac electrophysiology and pacing clinic in Helsinki University Hospital. His research is focused on cardiac electrophysiology using invasive and noninvasive techniques, ischemic heart disease and inherited cardiac arrhythmias. He has published 200 articles in peer-reviewed journals.

Prof. Toivo Katila was born in Eura, Finland, in 1941. He received the M.Sc. (technol.) degree in 1966 and the Dr. Technol. in 1970, both from the Helsinki University of Technology (HUT), Finland. Since 1973, he has been an associate professor/professor in biomedical engineering at HUT, from where he was retired in 2005. His main interests in research of biomedical engineering are bioelectromagnetism, medical image processing and biomedical optics.

a Clinic of Internal Medicine I, University Hospital of Jena, Erlanger Allee 101, 07747 Jena, Germany

b Biomagnetic Center, Clinic of Neurology, University Hospital of Jena, Jena, Germany

c University of Technology Ilmenau, Ilmenau, Germany

d Engineering Centre, BioMag Laboratory, Helsinki University Central Hospital, Helsinki, Finland

e Division of Cardiology, Department of Medicine, Helsinki University Central Hospital, Helsinki, Finland

f Department of Radiology, Helsinki University Central Hospital, Helsinki, Finland

g Laboratory of Biomedical Engineering, Helsinki University of Technology, Espoo, Finland

h Turku PET Centre, Turku, Finland

Corresponding Author InformationCorresponding author. Tel.: +49 3641 9324529; fax: +49 3641 9324177.

PII: S0895-6111(08)00096-7

doi:10.1016/j.compmedimag.2008.09.002


View previous. 2 of 11 View next.