Journal of Nuclear Cardiology

, Volume 24, Issue 1, pp 142–144 | Cite as

Advances in image reconstruction software in nuclear cardiology: Is all that glitters gold?

  • Claudio Marcassa
  • Orazio Zoccarato


The cornerstone results of nuclear cardiology in the last 25 years were obtained with the Filtered Back Projection as the preferred reconstruction method for tomographic studies. Recently, evolution of the OSEM iterative reconstruction algorithms was implemented by different vendors. The value and limitations of the new methods are briefly addressed.


Gated-SPECT iterative algorithms image reconstruction 

Myocardial perfusion imaging with radionuclide tracers is the most worldwide used non-invasive imaging technique for the diagnostic evaluation and risk stratification of patients with known or suspected coronary heart disease.1, 2, 3, 4 It accurately stratifies the risk in a variety of sub-populations including women, patients with diabetes, previous myocardial infarction, or coronary revascularization procedures.

The major, well consolidated, strengths of nuclear cardiology are the unique capability of assessing perfusion and function in the same session, the very low event rate in subjects with a normal MPI (<1%/year, comparable to that observed in the general population), and a close correlation between the amount of extent and severity of perfusion defects or the amount of left ventricular dysfunction with the adverse events rate.

All these pivotal results were obtained with the instrumentation invented by Hal Anger back in the ‘50 and the Filtered Back Projection (FBP) as the preferred reconstruction method for tomographic studies. This methodology was the basis of nuclear cardiology up to the ‘90, when iterative methods entered the nuclear medicine field.

Unlike analytical methods (e.g., FBP), the iterative reconstruction methods, besides taking into account the probabilistic nature of the radiation detection phenomenon, allow the definition of a more realistic model of the whole tomographic system. The Maximum Likelihood Expectation Maximization algorithm (MLEM) is the most popular statistical reconstruction method; rigorous mathematical demonstrations indicate that MLEM convergences to the real image but, at the same time, is greatly limited by its slowness of convergence and by the computational resources request.5 The Ordered Subsets Expectation Maximization (OSEM) iterative algorithm was introduced in 1994 by Hudson and Larkin with the aim of reduce the reconstruction time typical of the conventional MLEM.6 OSEM algorithm divides the acquired projections in N subsets and, starting from the first subset, applies the MLEM algorithm to each subgroup of projection using the result of the previous subset as input for the next subset. In this way, it is possible to obtain an acceleration proportional to N. However, it must be considered that, despite being based on the MLEM, it is not guaranteed that OSEM converges to the same ML solution; moreover, at the increase in the number of subsets, there is an increment in the image noise suggesting that this parameter represents a critical point of the algorithm that has to be modified carefully. Despite these limitations, due to its speed, the OSEM algorithm quickly became the reconstruction method of choice in nuclear medicine.

More recently, some advanced iterative reconstruction algorithms based on OSEM method were implemented with resolution recovery and noise suppression techniques (IRR). The major differences between IRR and FBP are clearly summarized in this issue by Olga et al7; in summary, IRR algorithms employ an iterative reconstruction process that compensates for the non-stationarity of the collimator’s response (“beam spread function” effect), take into consideration the statistical behavior of the emission and noise, which may vary from one application to the other, and correct for the variation in the detector-to-patient distance according to a fixed or variable type of orbit. All these points play a role in the better resolution of IRR over conventional FBP reconstruction, and in the superior image contrast, and allow reconstruction of images with low counts statistics, obtained either with shorter acquisition time or lower radionuclide activity.

The pathway of implementation of the IRR in the clinical practice includes the following:
  1. (1)

    Experimental phantom studies: they documented that the new IRR algorithms resulted in better resolution when compared to conventional FBP and OSEM.8, 9, 10 In addition, in an anthropomorphic cardiac phantom, WBR showed a comparable performance with respect to conventional FBP, either in half-time SPECT with a standard dose or with SPECT acquired at a standard time/frame but with half isotope activity.10

  2. (2)

    Single-center clinical studies: they documented that the image quality and the interpretation of half-time or half-dose gated-SPECT reconstructed with the new algorithms were comparable respectively to that of a standard-time or standard-dose gated-SPECT acquisition reconstructed with conventional FBP.11, 12, 13, 14, 15, 16

  3. (3)

    Clinical validation diagnosis and prognosis. Most of the published studies comparing new IRR with conventional FBP focused on demonstrating the “non inferiority” of IRR images in terms of image quality and the “equivalence” in terms of image interpretation and perfusion defect quantification.17, 18, 19, 20, 21 A recent report by Patil et al, documented in 93 patients a sensitivity of 92% and specificity of 87% in the detection of significant coronary artery disease.19 Gutstein et al, in a larger group of 290 patients reported sensitivity and a specificity values of 89% and 75%, respectively.20 Moreover, the utilization of IRR increased the diagnostic interpretation of equivocal studies, reducing the need of further testing.21 These accuracy values favorably compare with that reported in the literature using gated-SPECT acquisitions reconstructed with the conventional FBP.2,4 More recently, the prognostic value of MPI with the new softwares was also demonstrated.22


Finally, the new IRR softwares are able to provide the physicians high-quality SPECT images, with shorter acquisition time or lower study tracer activities, hampering patients’ and staff’s exposition.

Is this a sort of Holy Grail of nuclear cardiology?

Several pitfalls should be taken into account.

Most manufacturers have implemented their own version of IRR algorithms into conventional softwares for OSEM iterative reconstruction, optimized for their own specific camera. As result, the recommended reconstruction parameters vary widely between systems, making the comparison a difficult task.23

The performances of the different coupled hardware-software systems available in the market have been investigated in a experimental phantom study.24 It was demonstrated that there is an effective dependence of the image quality indexes from the scanner/software combination, which is not univocal but also depends on the image quality index considered. This could be due to the fact that the different manufacturers seem to emphasize different aspects of the myocardial image quality.24 As a matter of fact, the experimental results obtained in phantom studies show that IRR algorithms produce better results than conventional FBP and OSEM in terms of lesion contrast, wall thickness and noise indices, independently from the scanner/software combination considered.24 Moreover, the best results are obtained applying attenuation and scatter corrections, a practice still performed by the minority of laboratories. As a final point, also the polar maps obtained by these different models are significantly different; consequently, normal reference databases must be specifically implemented for each algorithm and system used.25

All the above-mentioned observations are related to the effect of the IRR on the perfusion pattern. Several concerns also affect the left ventricular volumes and ejection fraction measurements. As correctly emphasized by Dendy and Tilkemeier in this issue, in addition to the perfusion values, the functional parameters obtained by gated SPECT studies as well are strongly dependent on the patient’s characteristics and the reconstruction model used.23 The values obtained with the new IRR algorithms are critically related to the scanner/software combination used. This issue, however, is well known: volumetric quantification values in gated-SPECT acquisitions are strongly dependent on the quantification package used.26

As a consequence, normal reference limits specific for each scanner/software combination should be considered, either for polar maps quantification or left ventricular volumes and ejection fraction values.

Conclusions The new iterative reconstruction softwares are able to provide images of good quality from low count density acquisitions, allowing a significant reduction in acquisition time and/or dose/radiation, providing clinical information comparable to conventional FBP images reconstruction. Images interpretation, however, should take into account the differences between the different software/hardware packages.

Although we are still far from the numbers provided by the last 25 years of nuclear cardiology literature with conventional reconstruction methods, giving experimental and clinical studies available so far, the use of the new IRR to reduce the radiation burden to patients and operators,27 without losing diagnostic and prognostic accuracy, should be encouraged.


  1. 1.
    Shaw LJ, Hendel R, Borges-Neto S, et al. Prognostic value of normal exercise and adenosine 99 m Tc-tetrofosmin SPECT imaging: Results from the multicenter registry of 4,728 patients. J Nucl Med 2003;44:134–9.PubMedGoogle Scholar
  2. 2.
    Underwood SR, Anagnostopoulos C, Cerqueira M, et al. Myocardial perfusion scintigraphy: The evidence. Eur J Nucl Med Mol Imaging 2004;31:261–91.CrossRefPubMedGoogle Scholar
  3. 3.
    Parker MW, Iskandar A, Limone B, et al. Diagnostic accuracy of cardiac positron emission tomography versus single photon emission computed tomography for coronary artery disease. Clinical perspective: A bivariate meta-analysis. CircCardiovasc. Imaging 2012;5:700–7.Google Scholar
  4. 4.
    Cremer P, Hachamovicth R, Tamarappoo B. Clinical decision making with myocardial perfusion imaging in patients with known or suspected coronary artery disease. Semin Nucl Med 2014;44:320–9.CrossRefPubMedPubMedCentralGoogle Scholar
  5. 5.
    Shepp LA, Vardi Y. Maximum likelihood reconstruction for emission tomography. IEEE Trans Med Imaging 1982;1:113–22.CrossRefPubMedGoogle Scholar
  6. 6.
    Hudson HM, Larkin RS. Accelerated image reconstruction using ordered subsets of projection data. IEEE Trans Med Imaging. 1994;13:601–9.CrossRefPubMedGoogle Scholar
  7. 7.
    Olga J, Pagnanelli R, Borges Neto S. Resolution recovery and noise regularization in nuclear cardiology. J Nucl Cardiol 2016. doi: 10.1007/s12350-016-0532-1.Google Scholar
  8. 8.
    DePuey EG, Gadiraju R, Clark J, et al. Ordered subset expectation maximization and wide beam reconstruction “half-time” gated myocardial perfusion SPECT functional imaging: A comparison to “full-time” filtered backprojection. J Nucl Cardiol 2008;15:547–63.CrossRefPubMedGoogle Scholar
  9. 9.
    DePuey EG, Bommireddipalli S, Clark J, et al. A comparison of the image quality of full-time myocardial perfusion SPECT versus Wide Beam Reconstruction half-time and half-dose SPECT. J Nucl Cardiol 2011;18:273–80.CrossRefPubMedGoogle Scholar
  10. 10.
    Zoccarato O, Campini R, Marcassa C, Calza P. Performance of a new iterative reconstruction algorithm for cardiac short-time SPECT: Preliminary results in an anthropomorphic cardiac phantom study. Comput Cardiol 2008;35:329–32.Google Scholar
  11. 11.
    Borges-Neto S, Pagnanelli RA, Shaw LK, et al. Clinical results of a novel wide beam reconstruction method for shortening scan time of Tc-99 m cardiac SPECT perfusion studies. J Nucl Cardiol 2007;14:555–65.CrossRefPubMedGoogle Scholar
  12. 12.
    Marcassa C, Campini R, Zoccarato O, et al. Wide beam reconstruction for half-dose or half-time cardiac gated-SPECT acquisitions: Optimization of resources and reduction in radiation exposure. Eur J Nucl Med 2010;38:499–508.CrossRefGoogle Scholar
  13. 13.
    Cullom SJ, Saha K, Heller GV, et al. An optimized iterative reconstruction and processing protocol for “half-time” (32 projections) REST/STRESS Tc99 m-sestamibi myocardial perfusion SPECT. J Nucl Cardiol 2008;15:S6.CrossRefGoogle Scholar
  14. 14.
    Sun XX, Tian YQ, Wang DY, et al. Shortened acquisition time or reduced activity dose for gated myocardial perfusion SPECT with new reconstruction algorithm. Int J Cardiovasc Imaging 2013;29:1287–93.CrossRefPubMedGoogle Scholar
  15. 15.
    DePuey EG, Bommireddipalli S, Clark J, et al. Wide beam reconstruction “quarter-time” gated myocardial perfusion SPECT functional imaging: A comparison to “full-time” ordered subset expectation maximum. J Nucl Cardiol 2009;16:736–52.CrossRefPubMedGoogle Scholar
  16. 16.
    Druz R, Phillips L, Chugkowski M, et al. Wide-beam reconstruction half-time SPECT improves diagnostic certainty and preserves normalcy and accuracy: A quantitative perfusion analysis. J Nucl Cardiol 2011;18:52–61.CrossRefPubMedGoogle Scholar
  17. 17.
    Venero CV, Heller GV, Bateman TM, et al. A multicenter evaluation of a new post-processing method with depth-dependent collimator resolution applied to full-time and half-time acquisitions without and with simultaneously acquired attenuation correction. J Nucl Cardiol 2009;16:714–25.CrossRefPubMedGoogle Scholar
  18. 18.
    De Lorenzo A, Fonseca LM, Landesmann MC, Lima RS. Comparison between short-acquisition myocardial perfusion SPECT reconstructed with a new algorithm and conventional acquisition with filtered backprojection processing. Nucl Med Commun 2010;31:552–7.PubMedGoogle Scholar
  19. 19.
    Patil HR, Bateman TM, McGhie AI, et al. Diagnostic accuracy of high-resolution attenuation-corrected Anger-camera SPECT in the detection of coronary artery disease. J Nucl Cardiol 2014;21:127–34.CrossRefPubMedGoogle Scholar
  20. 20.
    Gutstein A, Navzorov R, Solodky A, et al. Angiographic correlation of myocardial perfusion imaging with half the radiation dose using ordered-subset expectation maximization with resolution recovery software. J Nucl Cardiol 2013;20:539–44.CrossRefPubMedGoogle Scholar
  21. 21.
    Qutub MA, Dowsley T, Ali I, Wells RG, et al. Incremental diagnostic benefit of resolution recovery software in patients with equivocal myocardial perfusion single-photon emission computed tomography (SPECT). J Nucl Cardiol 2013;20:545–52.CrossRefPubMedGoogle Scholar
  22. 22.
    Lima R, De Lorenzo A, Camargo G, et al. Prognostic value of myocardium perfusion imaging with a new reconstruction algorithm. J Nucl Cardiol 2014;21:149–57.CrossRefPubMedGoogle Scholar
  23. 23.
    Dendy JM, Tilkemeier P. Successful innovation: A time for change? J Nucl Cardiol 2016. doi: 10.1007/s12350-016-0533-0.PubMedGoogle Scholar
  24. 24.
    Zoccarato O, Scabbio C, De Ponti E, et al. Comparative analysis of iterative reconstruction algorithms with resolution recovery for cardiac SPECT studies. A multi-center phantom study. J Nucl Cardiol 2014;21:135–48.CrossRefPubMedGoogle Scholar
  25. 25.
    Zoccarato O, Marcassa C, Lizio D, et al. Differences in polar-map patterns using the novel technologies for myocardial perfusion imaging. J Nucl Cardiol 2016. doi: 10.1007/s12350-016-0500-9.Google Scholar
  26. 26.
    Hedeer F, Palmer J, Arheden H, Ugander M. Gated myocardial perfusion SPECT underestimates left ventricular volumes and shows high variability compared to cardiac magnetic resonance imaging: a comparison of four different commercial automated software packages. BMC Med Imaging 2010;25(10):10.CrossRefGoogle Scholar
  27. 27.
    Marcassa C, Zoccarato O, Campini R, Calza P. Temporal evolution of administered activity in cardiac gated-SPECT and patients’ effective dose. Analysis of an historical series. Eur J Nucl Med Mol Imaging 2013;40:325–30.CrossRefPubMedGoogle Scholar

Copyright information

© American Society of Nuclear Cardiology 2016

Authors and Affiliations

  1. 1.Cardiology DepartmentSalvatore Maugeri Foundation, IRCCS, Scientific Institute of Veruno (NO)VerunoItaly
  2. 2.Nuclear Medicine DepartmentSalvatore Maugeri Foundation, IRCCS, Scientific Institute of Veruno (NO)VerunoItaly

Personalised recommendations