Quantitative PET-CT and SPECT-CT


Quantitative PET

Our Center is very active in several research areas pertaining to quantitative positron emission tomography, both in PET-CT and PET-MR. Topics range from absolute quantitation with novel scatter, attenuation, and randoms corrections in model-based iterative reconstructions both in PET-CT and PET-MR to quantitative imaging in non-traditional radioisotopes such as Y-86, and absolute quantitation of neurotransmitter densities in dynamic PET imaging.


Expand all |
Collapse All

 

  • Single-scan rest/stress cardiac imaging

    We are investigating methods to use 18F agents for myocardial blood flow (MBF) in a clinically viable setting with quantitative results. Because of the long half-life, we propose single-scan rest/stress measurements made using [18F]flurpiridaz. Images of pig hearts were taken before and after infarction. Two injections of the tracer were done within the same scan, and the stressor, adenosine, was introduced partway through the scan. Time-activity-curves of regions of interest and parametric images were created and used for quantitation. We found that with a single 15 minute scan we were able to measure changes in rest/stress blood flow between healthy and infarcted hearts comparable to results using the same tracer with injections separated by 1 hour and subtraction of residual activity.

    Parametric maps of K1 in the heart during the rest (top) and stress (bottom) portions of the scan Fits to a time-activity-curves during the two-injection rest/stress scan

    N Guehl, M Normandin, D Wooten, G Rozen, A Sitek, M Mansour, T Shoup, L Ptaszek, G El Fakhri, N Alpert, Single-scan rest/stress imaging with [18F]flurpiridaz: demonstration in a porcine model. J Nucl Med May 1, 2015 vol. 56 no. supplement 3 516. [Link]
     

  • Absolute Quantitation of Myocardial Blood Flow Using Dynamic PET

    Absolute Quantitation of
    Myocardial Blood Flow Using Dynamic PET



    The
    central hypothesis of this proposal is that objective quantitation of
    myocardial perfusion and coronary flow reserve can be achieved using
    82Rb imaging and, furthermore, that these measures are
    important determinants of clinical risk and, thus, useful for optimizing
    management decisions in patients with coronary artery disease (CAD).  We
    use dynamic 82Rb PET along with innovative approaches based
    on generalized factor analysis of dynamic sequences (GFADS), that allow
    automatic estimation of left and right ventricle input functions, as
    well as region-based compartment analysis to characterize and quantify
    the coronary flow reserve (CFR) as well as the severity and extent of
    perfusion abnormalities that occur in CAD.

     

    Related Papers:

    • El Fakhri G., Kardan A., Sitek A., Dorbala S., Abi-Hatem
      N., Lahoud Y., Fischman A.J., Coughlan M., Yasuda T., Di
      Carli M.F. Reproducibility and Accuracy of Quantitative
      Myocardial Blood Flow Assessment Using 82Rb-PET:
      Comparison with 13N-Ammonia. J. Nucl. Med. 2009; 50:
      1062-1071. [PDF]

    • Anagnostopoulos C., Almonacid A., El Fakhri G.,
      Currilova Z., Sitek A., Roughton M., Dorbala S., Popma
      J., Di Carli M. Quantitative Relationship Between
      Coronary Vasodilator Reserve Assessed by Rubidium-82 PET
      Imaging and Coronary Artery Stenosis Severity. Eur. J.
      Nucl. Med. Mol. Imag. 2008; 35: 1593-1601.
      [PDF]

    • Di Carli M.F., Dorbala S., Meserve J., El Fakhri G.,
      Sitek A., Moore S.C. Clinical myocardial perfusion
      PET-CT. J. Nucl. Med; 2007; 48: 783-793.
      [PDF]

    • El Fakhri G., Sitek A., Guérin B., Kijewski M.F., Di
      Carli M.F., and Moore S.C. Quantitative dynamic cardiac
      82Rb-PET imaging using generalized factor and
      compartment analyses. J. Nucl. Med. 2005; 46: 1264-1271
      (2006 Mosby-Year Book of Nuclear Medicine).
      [PDF]

  • Scatter Compensation in PET


    Scatter Compensation in PET



    The overall goal of this project is to develop improved techniques for
    quantitative PET imaging. Recently, we have reformulated the PET
    reconstruction problem in order to include the energy of detected
    photons, which is a measurement not currently used in commercial scanner
    but has the potential of improving significantly the accuracy of scatter
    correction. Specifically, this work involved working a new expression of
    the list-mode PET likelihood function containing the 2D (because there
    are 2 photons per coincidence) energy probability density functions
    (EPDFs) of primary and scatter coincidences. This formulation of the
    likelihood lead to a novel MLEM reconstruction algorithm incorporating
    position and energy dependent corrections. Finally, analytical formulas
    were derived for estimating these EPDFs based on a decomposition of
    energy spectra into primary and scatter components. Our results on
    simulations indicate that 1) the energy information can be used to scale
    the scatter sinogram to the measured data and is more accurate than the
    traditional “tails fitting” approach for large patients; 2) the
    inclusion of energy PDFs in the MLEM reconstruction improves
    significantly the accuracy of the scatter correction in term of bias and
    contrast in cold regions. 

     

    Related Papers:


    Guerin B., Sitek A., and El Fakhri G., "Statistical energy-based scatter correction and reconstruction in list-mode PET". J. Nucl. Med. 2008; 49.
    [PDF]

  • Quantitative Y-86 PET for Personalized Radionuclide Therapy


    Quantitative Y-86 PET for Personalized
    Radionuclide Therapy



    Targeted radionuclide therapy (TRT) and
    radioimmunotherapy (RIT) are at the forefront of molecular cancer
    treatment modalities that involve the use of cancer cell-targeting
    radiopharmaceuticals, such as radiolabeled antibodies.  Y-90 based
    therapy has thus far been hampered by the inability to accurately image
    and quantify the distribution of Y-90 within the body as the latter is a
    pure b emitter.  Quantitative Y-86 PET imaging allows personalized
    patient treatment by enabling tailored Y-90 TRT based on the
    quantitative biodistribution of Y-86 uptake but presents unique
    challenges as it requires compensation for many physical effects
    including gamma cascade that greatly affect image quality and accuracy. 
    We are accurately modeling the physics of PET imaging using isotopes
    with cascade gammas, and developing correction algorithms to achieve
    quantitative Y-86 PET dosimetry.


    Related Papers:


    • Zhu X.and El Fakhri G. Monte Carlo
      modeling of cascade gamma rays in 86Y PET imaging:
      Preliminary results. Phys. Med. Biol.

      2009 Jul 7;54(13):4181-93. [PDF]

  • Quantitative Imaging of Neurotransmitter and Receptor Systems


    Quantitative Imaging of Neurotransmitter and Receptor Systems



    In recent years a number of positron-emitting ligands have
    been developed, allowing study of neurotransmission and reception in
    the brain. The methodology for assessing quantities, such as receptor
    for density and occupancy, in vivo are based on prior in vitro assays.
    However, the in vivo situation is more complex, owing to the necessity
    to account for blood flow, blood-to-tissue transport mechanisms, and
    the dynamics of nonspecific binding of the ligands. Several different
    assays have been developed, ranging from single injection determinations of binding potential, to the use of multiple injections at different
    specific activities to determine absolute receptor density. A novel line of
    investigation for future research has been described along with analytic techniques - that use cognitive activation to modulate receptor
    occupancy.




  • Iterative Brain and Cardiac Image Reconstruction


    Iterative Brain and Cardiac Image Reconstruction



    By using the iterative image reconstruction one may obtain PET images with better quality.
    Our projects are focused on using a priori information in the iterative reconstruction to improve image quality and assist computation of the parametric images.

     



  • Application of Image Processing to Radiology


    Application of Image Processing to
    Radiology



    Research in this area has
    focused on development of algorithms for multi-modality registration of human and
    animal brain images and development of algorithms and techniques
    for parametric imaging of physiological processes. Image registration
    techniques allow image volumes taken under different circumstances
    (e.g. different times, different modalities, even different subjects) to be
    aligned (and possibly deformed) so that they can exactly overlay one
    another. This process is important for research and diagnosis when
    serial quantitative comparisons are necessary. Parametric images are
    formed by analyzing the concentration history of every voxel in PET
    data sets. Kinetic parameters for each voxel are presented as images,
    making it possible to provide a quantitative visualization of physiological process.

  • Direct Reconstruction of PET Kinetic Parameters

    We are interested in an alternative way to estimate kinetic parameters in dynamic PET imaging. In this approach, we directly reconstruct PET parametric images from sinograms as compared to the conventional indirect approach where frame-by-frame images are reconstructed first and then followed by fitting time-activity curves voxel-wise. This direct approach has potential to reduce noise in parametric images because it does not impose assumptions about noise modeling and uncorrelatedness among voxels in the reconstructed images. To this end, we have developed a gradient-based direct reconstruction approach with improved convergence, and have assessed its performance in cardiac PET imaging.


     

    Comparison of the standard deviation images between the direct and indirect approaches in estimating K<sub>1</sub>

    Comparison of the mean images between the direct and indirect approaches in estimating K<sub>1</sub>

    Figure: Comparison of the mean and standard deviation images between the direct and indirect approaches in estimating K1.

  • Non-Local Means Denoising of Dynamic PET Images

    Dutta J, Leahy RM, Li Q. Non-Local Means Denoising of Dynamic PET Images. PLoS One. 2013; 8(12):e81390.
    [Link]

    Dynamic positron emission tomography (PET), which reveals information about both the spatial distribution and temporal kinetics of a radiotracer, enables quantitative interpretation of PET data. Model-based interpretation of dynamic PET images by means of parametric fitting, however, is often a challenging task due to high levels of noise, thus necessitating a denoising step. The objective of this paper is to develop and characterize a denoising framework for dynamic PET based on non-local means (NLM).

    The overall objective of this work is to develop an NLM-based denoising framework for dynamic PET images and to assess the quantitative and qualitative merits of the resultant approach relative to other well-known image denoising approaches. We compare this technique with Gaussian denoising and principal component analysis (PCA) based denoising, both widely used in the context of dynamic PET imaging. In addition, we compare this technique with HighlY constrained backPRojection (HYPR) and conventional NLM denoising based on spatial patches. We perform a realistic simulation study based on a dynamic digital mouse phantom and compare the denoising methods by examining bias-variance characteristics of the denoised dynamic images and the corresponding Patlak parametric images. We then apply the developed method to denoise a preclinical 18F PET dataset from a mouse study and a clinical 18F PET dataset from a patient with hepatocellular carcinoma and perform Patlak analysis on these datasets.

    Figure 4: A coronal slice from the dynamic Digimouse phantom. Figure 7. Patlak parametric imaging for the preclinical study.

Quantitative SPECT

  • Quantitative SPECT Imaging


    Quantitative Methods in SPECT



    We have developed compensation methods for Compton scatter, attenuation, variable collimator response as well as partial volume effect and septal penetration for many radionuclides used in SPECT such as Tc-99m, I-123, In-111, Ga-67, etc. Our methods range from spectral and dynamic factor analyses to fast and Monte Carlo-based iterative reconstruction and were applied to single and dual isotope SPECT in the brain, heart and cancer. Our lab has been one of the first to report successful imaging of simultaneous dual Tc-99m and I-123 in the brain and the heart in humans. We have also developed high-sensitivity centrally peaked collimators for a dedicated brain scanner and shown that a 3 fold gain in sensitivity can be achieved compared to triple head SPECT cameras. This collimator was manufactured and used in human brain studies. Finally, we have been heavily involved in the assessment of image quality in SPECT and the task-based improvement that can be achieved with these compensation methods that we have developed for lesion detection and activity quantitation.
     

     

    Related Papers:

    • Ouyang J., El Fakhri G., Moore S.C. Improved activity estimation with MC-JOSEM versus TEW-JOSEM
      in 111In SPECT. Med. Phys.
      2008; 35: 2029-2040. [PDF]

    • Zaidi H. and El Fakhri G. “Is absolute quantification of
      dopaminergic neurotransmission studies with 123I SPECT
      ready for clinical use?”. Eur. J. Nucl. Med. Mol. Imag.
      2008; 35: 1330-1333.

    • Mamede M., El Fakhri G., Abreu-e-Lima P., Gandler W.,
      Nose V., Gerbaudo V. Pre-operative estimation of
      esophageal tumor metabolic length in FDG PET images with
      surgical pathology confirmation. Ann Nucl Med. 2007; 21:
      553-562. [PDF]

    • El Fakhri G., Ouyang J., Zimmerman R.E., Fischman A.J.,
      Kijewski M.F. Performance of a Novel Collimator for High–Sensitivity Brain SPECT. Med. Phys. 2006; 33:209-215.
      [PDF]

    • El Fakhri G., Sitek A.,
      Zimmerman R.E., Ouyang J. Generalized Five Dimensional
      Dynamic and Spectral Factor Analysis. Med. Phys. 2006;
      33: 1016-1024. [PDF]

    • Moore S.C., Ouyang J., Park M.,
      El Fakhri G. Monte Carlo-based compensation for patient
      scatter, detector scatter, and crosstalk contamination
      in In-111 SPECT imaging. Nucl. Instrum. Meth. A. 2006;
      569: 472-476.
      [PDF]

    • Ouyang J., El Fakhri G., Xia W., Kijewski M.F., Genna S.
      The Design and Manufacture of an Annular
      Variable-focusing Collimator for High-sensitivity Brain
      SPECT. IEEE Trans Nucl Sci 2006; 53: 2613-2618.

      [PDF]

    • Moore S.C., Kijewski M.F., and El Fakhri G. Collimator optimization
      for detection and quantitation tasks: application to gallium-67
      imaging. IEEE Trans. Med. Imag; 2005; 24: 1347-1356.

    • El
      Fakhri G., Kijewski M.F., Albert M.S., Johnson K.A., and Moore
      S.C. Quantitative SPECT leads to improved performance in
      discrimination tasks related to prodromal Alzheimer’s disease.
      J. Nucl. Med. 2004; 45: 2026-2031. [PDF]

    • El
      Fakhri G., Kijewski M.F., Johnson K.A., Syrkin G, Killiany R.J.,
      Becker JA, Zimmerman R.E., Albert M.S. MRI-Guided SPECT
      perfusion measures and volumetric MRI in prodromal Alzheimer’s
      disease. Arch Neurol 2003; 60: 1066-1072. [PDF]

    • El
      Fakhri G., Moore S.C., Maksud P, and Kijewski M.F. The effects
      of compensation for scatter, lead x-rays and high-energy
      contamination on lesion detectability and activity estimation in
      Ga-67 imaging. IEEE Trans Nucl Sci 2003; 50: 439-445.
      [PDF]

    • El
      Fakhri G., Moore S.C., and Kijewski M.F. Optimization of Ga-67
      imaging for detection and estimation tasks: dependence of
      imaging performance on spectral acquisition parameters. Med Phys
      2002; 29: 1859-1866. [PDF]

    • El
      Fakhri G, Moore S.C., Maksud P. A new scatter compensation
      method for Ga-67 Imaging using artificial neural networks. IEEE
      Trans Nucl Sci. [PDF]

    • El
      Fakhri G, Kijewski M.F, Moore S.C. Absolute activity
      quantitation from projections using an analytical approach:
      comparison with iterative methods in brain SPECT. IEEE Trans
      Nucl Sci. [PDF]

    • El
      Fakhri G, Buvat I, Benali H, Todd-Pokropek A, Di Paola R.
      Relative impact of scatter, attenuation, collimator response and
      finite spatial resolution corrections in cardiac SPECT. J Nucl
      Med 2000;41:1400-1408. [PDF]

    • El
      Fakhri G, Maksud P, Kijewski M.F, Habert M.O, Todd-Pokropek A,
      Aurengo A, Moore S.C. Scatter and Cross-Talk Corrections in
      Simultaneous Tc-99m/I-123 Brain SPECT using Constrained Factor
      Analysis and Artificial Neural Networks. IEEE Trans Nucl Sci
      2000;47:1573-1580. [PDF]

    • El
      Fakhri G, Buvat I, Almeida P, Bendriem B, Todd-Pokropek A,
      Benali H. Should scatter be corrected in both transmission and
      emission data for accurate quantitation in cardiac SPECT? Eur J
      Nucl Med 2000;27:1356-1364. [PDF]

    • Pélégrini M, Buvat I, El Fakhri G, Benali H, Grangeat P, Di
      Paola R. A spline-regularized minimal residual algorithm for
      iterative attenuation correction in SPECT. Phys Med Biol 1999;
      10: 2623-2642.

    • El
      Fakhri G, Buvat I, Pélégrini M, Benali H, Almeida P, Bendriem B,
      Todd-Pokropek A, Di Paola R. Respective roles of scatter,
      attenuation, collimator response and partial volume effect in
      cardiac SPECT quantitation: a Monte Carlo study. Eur J Nucl Med
      1999 ; 26 : 437-446. [PDF]

    • El
      Fakhri G, Maksud P, Aurengo A. Evaluation of scatter correction
      methods using Monte Carlo simulation in non uniform media. SPIE
      1998; 3338: 363-369. [PDF]

    • Maksud
      P, Fertil B, Rica C, El Fakhri G, Aurengo A. Artificial neural
      network as a tool to compensate for scatter and attenuation in
      radionuclide imaging. J Nucl Med 1998; 39; 4; 735-745.

    • Pélégrini M, Benali H, Buvat I, El Fakhri G, Di Paola R.
      Two-dimensional statistical model for regularized backprojection
      in SPECT. Phys Med Biol 1998; 43: 421-434.

  • Factor Analysis of Medical Images


    Factor Analysis of Medical Images



    An active area of research in our group pertains to factor analysis of spectral (energy) and/or dynamic (time) sequences. Factor analysis is a powerful technique that allows the decomposition of a spectral or temporal image sequence into a small number of fundamental functions (factors) whose associated spatial distributions are called factor images. This, in turn, yields a synthetic representation of the contents of a relatively large dynamic or spectral image dataset. The major drawback of factor analysis, which precludes quantitation, is the non-uniqueness of the solution and we have developed several approaches to solve the non-uniqueness problem in brain and cardiac imaging.

     

     

    Related Papers:

    • El Fakhri G., Buvat I, Benali H, Todd-Pokropek
      A, Di Paola R. Relative Impact of Scatter, Collimator Response, Attenuation, and Finite Spatial Resolution Corrections in Cardiac SPECT. J Nucl Med  2000 ; 41 : 1400-1408 (published
      in the 2001 Mosby-Year Book of Nuclear Medicine).
      [PDF]

    • El Fakhri G., Maksud P, Kijewski M.F, Habert
      M.O, Todd-Pokropek A, Aurengo A, Moore S.C. Scatter and
      Cross-Talk Corrections in Simultaneous Tc-99m/I-123 Brain SPECT
      using Constrained Factor Analysis and Artificial Neural
      Networks. IEEE Trans Nucl Sci 2000 ; 47 :1573-1580
      [PDF]

    • El Fakhri G., Maksud P., Kijewski M.F., et al.
      Quantitative simultaneous Tc-99m/I-123 SPECT: design study and
      validation with Monte Carlo simulations and physical
      acquisitions. IEEE Trans Nucl Sci 2002 ; 49: 2315-2321.
      [PDF]

    • El Fakhri G., Sitek A., Guérin B., Kijewski
      M.F., Di Carli M.F., and Moore S.C.  Quantitative dynamic
      cardiac 82Rb-PET imaging using generalized factor and
      compartment analyses. J. Nucl. Med. 2005; 46: 1264-1271 (2006
      Mosby-Year Book of Nuclear Medicine). [PDF]

    • El Fakhri G., Sitek A., Zimmerman R.E., Ouyang
      J.  Generalized Five-Dimensional Dynamic and Spectral
      Factor Analysis. Med. Phys. 2006; 33: 1016-1024.
      [PDF]

    • Anagnostopoulos C., Almonacid A., El Fakhri
      G., Currilova Z., Sitek A., Roughton M., Dorbala S., Popma J.,
      Di Carli M. Quantitative Relationship Between Coronary
      Vasodilator Reserve Assessed by Rubidium-82 PET Imaging and
      Coronary Artery Stenosis Severity. Eur. J. Nucl. Med. Mol. Imag.
      2008; 35: 1593-1601. [PDF]

    • El Fakhri G., Kardan A., Sitek A., Dorbala S.,
      Abi-Hatem N., Lahoud Y., Fischman A.J., Coughlan M., Yasuda T.,
      Di Carli M.F.  Reproducibility and Accuracy of Quantitative
      Myocardial Blood Flow Assessment Using  82Rb-PET:
      Comparison with 13N-Ammonia. J. Nucl. Med. 2009; 50: 1062-1071.
      [PDF]

Simultaneous Dual Tracer ECT

  • Simultaneous Dual Tracer ECT

    We are bringing to fruition our successful developments of accurate
    compensation methods for physical factors affecting image quality and
    reconstruction approaches for quantitative dual-isotope SPECT, and are
    expanding our scope to PET. We are presently evaluating the performance
    of two novel scatter correction methods for quantitative single and
    dual-isotope PET that we have developed.  We are also extending our
    previous work in SPECT to novel approaches to dual-isotope dynamic brain
    PET using spatio-temporal information. We are assessing the performance
    of quantitative dual-isotope SPECT and PET in activity estimation tasks
    related to early diagnosis and quantitation of disease extent in
    coronary artery disease (CAD) and early Parkinson disease (PD) with or
    without dementia.

    Dual-tracer PET using generalized factor analysis of dynamic sequences

    We developed a novel approach based on generalized factor analysis of dynamic sequences (GFADS) that exploits spatio-temporal differences between radiotracers and applied it to near-simultaneous imaging of 2-deoxy-2-[18F]fluoro-D-glucose (FDG) (brain metabolism) and 11C-raclopride (D2) with simulated human data and experimental rhesus monkey data. We show theoretically and verify by simulation and measurement that GFADS can separate FDG and raclopride measurements that are made nearly simultaneously.

    Estimation of specific and free + nonspecific 11C-raclopride. (Left) Time activity curves are calculated by subtraction of the (free + nonspecific) factor from the total brain factor after they have been scaled such that the factor images have the same mean values. (Right) (free + nonspecific) factor image is calculated by adding the spatially opposed factor images after they have been scaled such that their addition produces a uniform image. Resultant time activity factors for FDG and specifically and free + nonspecifically bound 11C-raclopride in a rhesus monkey study. Results were obtained by applying GFADS to a mid-striatal brain volume. Ground truth FDG activity is also shown for the initial three time frames.
    Factor images from application of GFADS to a mid-striatal, whole-brain slice of a dual-tracer rhesus monkey study. (Left to right) Ground truth FDG image (data), computed as the average of the initial three time frames; GFADS FDG image, demonstrating the same features as the ground truth image; GFADS specifically bound 11C-raclopride factor; derived free + nonspecifically bound GFADS factor image.

    With kind permission from Springer Science+Business Media: Molecular Imaging and Biology, Dual-Tracer PET Using Generalized Factor Analysis of Dynamic Sequences, Volume 15, Issue 6, 2013, pp 666-674, Georges El Fakhri, Cathryn M. Trott, Arkadiusz Sitek, Ali Bonab, Nathaniel M. Alpert, Figures 1, 5, and 6 [Link]

    Related
    Papers:

    • Trott C. and El Fakhri G. "Sequential and simultaneous
      dual-isotope brain SPECT: comparison with PET, for
      estimation and discrimination tasks in early parkisnon
      disease". Med. Phys., 2008; 35: 3343-3353.
      [PDF]

    • Ouyang J., Zhu X., Trott C., and El
      Fakhri G. Quantitative simultaneous 99m
      Tc/123I
      cardiac SPECT using MC-JOSEM,” Med. Phys. 36, 602-611
      (2009). 
      [PDF]

    • Ouyang J., El Fakhri G., Moore S.C. Fast Monte Carlo
      Simulation Based Joint Iterative Reconstruction for
      Simultaneous 99mTc/123I Brain SPECT Imaging. Med. Phys.
      2007; 34:x 3263-3272. [PDF]

    • El Fakhri G., Habert M.O., Maksud P., Kas A., Malek Z.,
      Kijewski M.F., and Lacomblez L.. Quantitative
      simultaneous 99mTc-ECD/123I-FP-CIT SPECT in Parkinson
      disease and multiple system atrophy. Eur. J. Nucl. Med.
      Mol. Imag. 2006; 33: 87-92. [PDF]

    • El Fakhri G., Maksud P., Kijewski M.F., et al.
      Quantitative simultaneous 99mTc/123I SPECT: design study
      and validation with Monte Carlo simulations and physical
      acquisitions. IEEE Trans Nucl Sci 2002 ; 49: 2315-2321.
      [PDF]

    • El Fakhri G, Moore S.C, Maksud P, Aurengo A, Kijewski
      M.F. Absolute activity quantitation in simultaneous
      I-123/Tc-99m brain SPECT. J Nucl Med 2001; 42: 300-308.
      [PDF]

    • El
      Fakhri G, Maksud P, Kijewski M.F, Habert M.O, Todd-Pokropek A,
      Aurengo A, Moore S.C. Scatter and Cross-Talk Corrections in
      Simultaneous Tc-99m/I-123 Brain SPECT using Constrained Factor
      Analysis and Artificial Neural Networks. IEEE Trans Nucl Sci
      2000;47:1573-1580. [PDF]