0000000000222455
AUTHOR
Jorge D. Martinez
Depth of /spl gamma/-ray interaction within continuous crystals from the width of its scintillation light-distribution
We have studied a new and inexpensive method of measuring the depth of interaction (DOI) in /spl gamma/-ray detectors with large-sized scintillation crystals. This method takes advantage of the strong correlation between the width of the undisturbed light-distribution in continuous crystals and the /spl gamma/-ray's DOI. In order to quantify the dependence of the distribution's width with respect to the DOI, we first studied an analytical model of the light-distribution and tested it by means of Monte Carlo (MC) simulations of the light transport inside the crystal. Further we present an inexpensive modification of the commonly used charge division circuit that allows analog and instantaneo…
Position sensitive scintillator based detector improvements by means of an integrated front-end
PESIC is an integrated front-end for multianode photomultiplier based nuclear imaging devices. Its architecture has been designed to improve position sensitive detectors behavior by equalizing its response over its whole area. Its preamplying stage introduces two main benefits: digitally programmable gain adjustment for every photomultiplier output, and isolation from other front-end electronics by means of current buffers. This last feature allows to use different types of photomultipliers and optimizes front-end deadtime, reducing impact position dependent output delay. PESIC also includes an indirect measurement of the depth of interaction of the gamma ray inside the scintillator crystal…
Multiprocessor SoC Implementation of Neural Network Training on FPGA
Software implementations of artificial neural networks (ANNs) and their training on a sequential processor are inefficient because they do not take advantage of parallelism. ASIC and FPGA implementations employ specific hardware structures to exploit parallelism in order to improve processing speed; however, optimizing resource usage requires the use of fixed-point arithmetic, thereby losing precision, and the final system is restricted to a particular network topology. This paper presents a mixed approach based on a multiprocessor system-on-chip (SoC) on a FPGA. The use of software-driven embedded microprocessors with custom floating-point extensions for ANN related functions allows for gr…
SoC-Based Implementation of the Backpropagation Algorithm for MLP
The backpropagation algorithm used for the training of multilayer perceptrons (MLPs) has a high degree of parallelism and is therefore well-suited for hardware implementation on an ASIC or FPGA. However, most implementations are lacking in generality of application, either by limiting the range of trainable network topologies or by resorting to fixed-point arithmetic to increase processing speed. We propose a parallel backpropagation implementation on a multiprocessor system-on-chip (SoC) with a large number of independent floating-point processing units, controlled by software running on embedded processors in order to allow flexibility in the selection of the network topology to be traine…
Maximum likelihood positioning for gamma-ray imaging detectors with depth of interaction measurement
Abstract The center of gravity algorithm leads to strong artifacts for gamma-ray imaging detectors that are based on monolithic scintillation crystals and position sensitive photo-detectors. This is a consequence of using the centroids as position estimates. The fact that charge division circuits can also be used to compute the standard deviation of the scintillation light distribution opens a way out of this drawback. We studied the feasibility of maximum likelihood estimation for computing the true gamma-ray photo-conversion position from the centroids and the standard deviation of the light distribution. The method was evaluated on a test detector that consists of the position sensitive …
A Tool for Implementing and Exploring SBM Models: Universal 1D Invertible Cellular Automata
The easiest form of designing Cellular Automata rules with features such as invertibility or particle conserving is to rely on a partitioning scheme, the most important of which is the 2D Margolus neighborhood. In this paper we introduce a 1D Margolus-like neighborhood that gives support to a complete set of Cellular Automata models. We present a set of models called Sliding Ball Models based on this neighborhood and capable of universal computation. We show the way of designing logic gates with these models, propose a digital structure to implement them and finally we present SBMTool, a software development system capable of working with the new models.