Performance evaluation of ¹⁸F radioluminescence microscopy using computational simulation

Abstract

Purpose Radioluminescence microscopy can visualize the distribution of beta-emitting radiotracers in live single cells with high resolution. Here, we perform a computational simulation of 18F positron imaging using this modality to better understand how radioluminescence signals are formed and to assist in optimizing the experimental setup and image processing. Methods First, the transport of charged particles through the cell and scintillator and the resulting scintillation is modeled using the GEANT4 Monte-Carlo simulation. Then, the propagation of the scintillation light through the microscope is modeled by a convolution with a depth-dependent point-spread function, which models the microscope response. Finally, the physical measurement of the scintillation light using an electron-multiplying charge-coupled device (EMCCD) camera is modeled using a stochastic numerical photosensor model, which accounts for various sources of noise. The simulated output of the EMCCD camera is further processed using our ORBIT image reconstruction methodology to evaluate the endpoint images. Results The EMCCD camera model was validated against experimentally acquired images and the simulated noise, as measured by the standard deviation of a blank image, was found to be accurate within 2% of the actual detection. Furthermore, point-source simulations found that a reconstructed spatial resolution of 18.5 μm can be achieved near the scintillator. As the source is moved away from the scintillator, spatial resolution degrades at a rate of 3.5 μm per μm distance. These results agree well with the experimentally measured spatial resolution of 30–40 μm (live cells). The simulation also shows that the system sensitivity is 26.5%, which is also consistent with our previous experiments. Finally, an image of a simulated sparse set of single cells is visually similar to the measured cell image. Conclusions Our simulation methodology agrees with experimental measurements taken with radioluminescence microscopy. This in silico approach can be used to guide further instrumentation developments and to provide a framework for improving image reconstruction.

In silico optimization of radioluminescence microscopy

“…The 2 sets of parameters and corresponding performance metrics (using an 18 F point source, 4 μm from the scintillator) are given in Table . Further, we compared the simulated radioluminescence cell images (illustrated in our previous paper ) obtained using experimental and optimized parameters, as shown in Figure A and B, respectively, given the same planar cell source which is 4 μm from the scintillator and same total radioactive decays: 10 6 over all cells. Positrons are assumed to be emitted evenly from the cells extracted from the bright field cell image.…”

“…Numerical models describing radiation transport and scintillation, light propagation through the microscope and photon sensing by the EMCCD camera were established and validated in our previous study . One advantage of computational modeling is that the space of the system's parameters can be efficiently explored to derive optimal device parameters for imaging.…”

“…This component helps suppress readout noise but it introduces excess noise . A method to model this noise has been presented in our previous paper . The standard deviation of the total output noise N can be estimated as where M is the total gain of cascaded electron multiplication elements and F is the excess noise factor, which approaches when the number of incident electrons and multiplying stages are large.…”

“…At the end of the pipeline, the simulated camera frames are processed by a reconstruction method to produce an image of the radiotracer distribution in the microscope field of view. We use a simplified version of the optical reconstruction of beta‐ionization track (ORBIT) method, which is described in previous papers . Briefly, the method segments individual positron ionization tracks from raw camera frames, then applies a center‐of‐mass calculation to a selected subset of ionization tracks to estimate the emission location of the positron.…”

“…However, while this empirical performance optimization can yield useful information, the optimal combination of system parameters is still unknown. This study aims at solving this problem using a computational simulation model of the RLM we previously established and validated . Here we aim to find the optimal parameters for (1) the material and thickness of the scintillator, (2) the magnification of the microscope, and (3) the settings of the electron multiplying charge coupled device (EMCCD) camera (EM gain and binning).…”