PhD
Research (2004)
Simultaneous Imaging of Optical and Isotopic Photon Distributions in Mice Using a Combined PH SPECT-OT Scanner - A Monte Carlo Simulation Study
Introduction: For imaging regional distribution and time variation of gene expression in mice, radionuclide imaging (SPECT, PET) and, most recently, bioluminescence / fluorescence imaging (OT) technologies have demonstrated separately to be highly sensitive on the molecular level. In order to investigate how markers are activated by specific enzymatic interactions researchers sometimes do employ both modalities sequentially, i.e. image the same object initially using a PET or SPECT scanner and thereafter using a CCD camera. Besides a longer total acquisition time, comparing marker kinetics can be impossible. Therefore, unified simultaneous dual-modality acquisition along with integrated reconstruction strategies and tracer / protein-kinetic models are desirable.
Methods: We have designed and propose to build a dual-modality single pinhole aperture small animal SPECT device with an integrated CCD camera to image simultaneously and from the same projection angle gene expression with radionuclide and optical markers. The PH SPECT sub-system is designed with energy-specific asymmetrical channelized PH inserts, a flat-sided collimator, a 10cm2 area with pixellated 0.1cm2 x 0.5cm NaI(Tl) scintillator crystals, and four position-sensitive flat panel photo-multiplier tubes (PMTs) H8500 (Hamamatsu Photonics). For the OT sub-system a cooled SensiCam QE670KS (PCO Imaging) CCD camera is used. Both detector systems are mounted on a common gantry and a system of mirrors is used to strip off optical photons from the multi-energetic photon flux and to compose an optical projection that is aligned with the angular field of view of the PH SPECT camera (Fig. 1). A Monte Carlo simulation program has been developed und was used to investigate and to optimize the pre-construction imaging system. Simulations have been performed for optical photons and for photons up to 600keV using either analytically defined point and line sources or a 110µm resolution tomographic mouse phantom, both with wavelength/energy- and tissue-specific absorption, scattering and anisotropy coefficients.
Results: In order to assess the system's physical performance a series of application-specific Monte Carlo simulations have been performed. For the SPECT sub-system, the geometry of PH inserts to be manufactured has been optimized for sensitivity/resolution with respect to the energy range of specific isotopes and the resolution properties of the position sensitive PMTs. The optical reflective surface which is located in the field of view of the SPECT camera has been carefully investigated for possible introduction of imaging-degrading artifacts. None have been found for energies above 10keV if the reflective aluminum layer is less than 5µm and the substrate layer less than 0.8mm thick. Primary subject of investigation for the OT sub-system has been the evaluation of the much higher interaction probability of optical photons in opaque tissue.
Discussion: Because of the very low penetration depth of optical photons, an enormous number of optical photon emissions is essential for performing fully 3D optical tomography. Therefore, the operational mode of the imaging system has not yet been entirely assessed and dedicated reconstruction algorithms for diffuse optical photon distributions need to be developed. The PH SPECT is well understood. Compared to state-of-the-art dedicated micro-SPECT systems performance is limited only by an object-pinhole gap introduced by the optical reflective surface extending the pinhole which depends on collimator insert thickness and pinhole geometry. The proposed design can also be used to detect optical or isotopic photons separately at highest sub-system sensitivity.
Evaluation of Pharmacokinetic Models by Means of Monte Carlo Simulated Tracerkinetic Time Frame Data
Objectives: We have extended our previously introduced SPECT/PET Monte Carlo (MC) simulation tool to enable direct investigation and comparison of pharmacokinetic (PK) models based on single-parameter and acquisition schedule variations using simulated time frame projection data. The primary purpose of this study is to provide confidence intervals for the application of specific PK models with regard to spatial resolution and noise characteristics in the volume/input data, to optimize sampling schedules, and to support the development of population-based PK approaches.
Material: In order to perform the studies, we have developed and implemented device-specific MC simulation algorithms that solve physically exact photon path integrals through non-uniform tissue phantoms (compartmental analytical as well as high-resolution tomographic) for photons up to 750keV.
Methods: A library of measured and derived prototype regional time activity distributions along with associated input curves has been composed for a variety of quantitative physiological imaging applications such as blood flow with [15O]water, glucose metabolism with [18F]FDG, or receptor interactions with [11C]NNC112. Series of application-specific MC simulation studies have been performed for a range of sampling schedules/signal-to-noise ratios and for varying distribution/input curve kinetics. Projections were reconstructed and the PK model under investigation was applied to yield functional parameters.
Results: Confidence intervals with respect to noise and single-parameter variation were derived. We have found that graphical methods and low-compartmental PK approaches are quite robust concerning data statistics as long as the kinetics is resolved. Significant fluctuations have been seen, however, in low-count data within small regions which are caused basically by resolution limitations of the imaging system. A cause of significant discrepancy of calculated parameters has been observed in case there is a divergence in timing between the input curve and the volume activity kinetics.
Conclusion: We evaluated quantitatively the effect of alterations in volume data/input functions on calculated physiological parameters. Aspects of limitations of PK models with regard to the relationship between activity measurements and the underlying physiological parameters cannot be addressed by a simulation approach that merely approximates the physics of imaging. Therefore, for a particular biological question methods for approximating bio-chemistry, physiology, and further a priori knowledge that also link the input curve to the volume data measurement need to be investigated and included into this approach.
Integrating Kinetic Models for Simulating Tumor Growth in Monte Carlo Simulation of ECT Systems
Summary: We have developed an integrated framework for linking tumor growth models directly into a Monte Carlo simulation algorithm for PET and SPECT systems. Tumors are approximated either by analytically defined five-dimensional (x, y, z, tgeometry, tactivity) compartments or by compound cellular lattice inserts. Both representation models can be placed into arbitrarily complex tomographic or mathematical phantoms. Various models for tumor growth approximation have been developed or are implemented such as sigmoidal growth according to the Gompertz equation, compartment models for heterogeneous metastatic tumors including model extensions that account for various therapy strategies, and self-organizing multi-particle systems in the form of cellular automata. With this novel approach, Monte Carlo simulation studies can be performed repetitively in static or dynamic acquisition mode at any given time of projected tumor growth. The proposed simulation technique provides a basis for deriving allometric relationships between growth rate and tumor representation in a sequence of simulated tomographic images. We propose that the introduced framework provides an ideal experimental model environment for evaluating growth theories and acquisition/schedule strategies, especially under controlled conditions of the simulated imaging system under investigation.