Multiple Sensor Position Locating System

This invention relates generally to systems for locating a position of a source of a detectable emittance, and more particularly, to a system which is particularly adapted for locating the position of a scintillation event in single photon emission computer tomography (SPECT).

The estimation of a position of an event or other interaction which affects a plurality of sensors has received considerable attention in a variety of disciplines. Of particular interest is the estimation of the position of a scintillation event in a radiation detector, particularly of the type used in single photon emission computer tomography. Most commonly, SPECT systems utilize detectors which are comprised of arrays of small detector elements, such as photomultiplier tubes. In order to maximize the useful field of view, the photomultiplier tubes are arranged in a predetermined configuration, which may be an array having a rectilinear or hexagonal configuration. The essential purpose of such a system, particularly a SPECT configuration as used in nuclear medicine, is to facilitate the imaging of radionuclides distributions in a predetermined region.

The essential elements of a known scintillation camera of the modular type employs a scintillation crystal which is arranged to communicate with a light guide. The light guide is in turn connected to a plurality of photomultiplier tubes which produce respective electrical signals responsive to the number of optical photons which reach each photomultiplier tube as a result of a scintillation event. As is known, the number of optical photons which impinge on a photomultiplier tube result in the production of a multiplicity of photoelectrons in each photomultiplier tube. It is significant to note, however, that almost all of the events incident to scintillation events and their detection must be analyzed statistically. For example, both, the number of photons generated in each scintillation event, as well as the number of photoelectrons produced in each photomultiplier tube, are statistical. These numbers, therefore, fluctuate about a mean value which, with respect to the number of photoelectrons produced in each photomultiplier tube, is a function of the position of the scintillation event. The uncertainties attendant any given event result in difficulties in ascertaining the precise position of the event.

The prior art has thrust at the problem of position estimation from a variety of approaches. One known method utilizes a completely digital system which employs look-up tables stored in memories. It is a problem with the known system that it requires a very large memory for more than two or three sample inputs. Also, the method depends upon the spatial resolution of the position estimation. In situations where four phototubes are employed, this known system could easily require sixteen megabytes of memory to estimate only sixty-four discrete positions.

A further known system utilizes both analog and digital aspects, and employs resistor networks and summing amplifiers. It is a problem with this method that it is inflexible and quite expensive. Again using four phototubes for example, this known system would require sixty-four amplifiers and 256 precision resistors to achieve the same estimation. The resistors may need to be nonstandard values, thereby increasing the cost. Moreover, such components occupy considerable space and must be mounted individually on circuit boards.

It is, therefore, an object of this invention to provide a simple and inexpensive system for estimating the position of a source of an emittance, or an event, such as a scintillation event.

It is another object of this invention to provide a precision estimating system which can be rapidly and easily configured to suit given detector module characteristics, without requiring changing resistors or other circuit board components.

It is a further object of this invention to provide a position estimator wherein no penalty is incurred as a result of dividing more finely the interpolated position values.

It is also an object of this invention to provide a position estimating system which is easily extended to include a greater number of phototubes.

It is additionally an object of this invention to provide a position estimating system which can be extended to perform multidimensional maximum liklihood estimates.

It is still another object of this invention to provide a precision estimating system which can be speeded up simply and with very little increase in cost.

It is still a further object of this invention to provide a position estimating system which can use relatively slow and inexpensive components.

The foregoing and other objects are achieved by this invention which provides a system for locating a position of a source of a detectable emittance, such as a scintillation event, utilizing a plurality of sensors arranged in a predetermined configuration with respect to one another. Each such sensor detects the detectable emittance and produces at an associated output a respective sensor signal which has a value which corresponds substantially to the magnitude of the detectable emittance which impinges on the associated sensor. A plurality of weighting values, which are determined during an initialization, or calibration cycle, are stored in a memory and are combined with the signals from the sensors to produce weighted sensor signals. The weighted sensor signals are combined in a summing arrangement which produces at its output a composite signal. The value of the composite signal is monitored while the weighting factors are recalled from the memory. The addresses of the weighting factors in the memory are correlated to position information, illustratively by way of a position look-up table, the correct position value being selected in response to the composite signal assuming a predetermined value. In a preferred embodiment, the predetermined value corresponds to zero. In an alternative, highly advantageous embodiment of the invention, the predetermined value may be set equal to a bias term which is determined from the total energy on an event-by-event basis.

In one embodiment of the invention, the weighted sensor signals are produced by combining each sensor signal with a respectively associated weighting factor in a digitial-to-analog converter (DAC). The output of the various DACs are combined in a summing arrangement which produces at its output the composite weighted signal. Increases in speed of operation of the system can be achieved by using additional DACs in parallel.

In one embodiment of the invention for estimating position along two or more coordinates, it is advantageous to cross-couple information between the estimation circuit for each coordinate, or dimensional address element. In the case of estimating the x and y coordinates of a scintillation event, improved results are obtained by supplying the estimated x coordinate to the y estimation circuit where this information is used to select a subset of y weights which are specific to events at or near the determined x location. In similar fashion, the estimate of y location may be returned to the x estimation circuit where it is used to select a subset of x weights which are specific to events which have occurred at or near the determined y location. It is possible in principle to continue this iterative process to produce a maximum likelihood estimate of both, the x and y locations. This process is particularly advantageous when the shape and sampling of the light distribution is spatially variant. In a simple embodiment of this concept, information pertaining to the photomultiplier which is issuing the maximum signal in one coordinate is coupled to the position computers for the other coordinate(s) where the information is used to select the most appropriate subset of weights for the estimation process.

In a specific illustrative embodiment of the invention, the initial x and y position estimates are assumed to be independent. Such, however, is not necessarily the case, as lack of independence is to an extent the result of asymmetrical light distribution within the light bars. However, iterative cross-coupling, illustratively such that the y estimation is dependent upon the initial x value estimate, yields substantial improvement.

In one advantageous embodiment of the invention, processing efficiency is improved by identifying the one of the sensors, which may be photomultiplier tubes, which is issuing the highest magnitude signal. Such identification establishes a neighborhood in which a search for the position of the event or interaction is conducted. The neighborhood is identified by a range of addresses of the weight factor memory. Thus, as the weight factors are clocked out of the memory, illustratively in response to a clock, the composited weighted signal is monitored, illustratively by means of a zero-crossing detector. The maximum likelihood position of the event or interaction corresponds to the location at which the composited weighted signal assumes a zero value. Of course, persons skilled in the art can configure the circuitry which, will be disclosed in detail hereinbelow so that the maximum likelihood position corresponds to some other value of the composite weighted signal. In the present embodiment of the invention, the use of a zero-crossing detector is enabled by using DACs having complementary current outputs. In such an embodiment, the complementary current outputs are coupled to inverting and noninverting inputs of an amplifier, the output of which is a true representation of the product of the weight factors in the sensor signal values.

Since it is expected that, in tomographic embodiments, a scintillation effect may occur during processing of a prior event, a plurality of sample and hold arrangements are provided for each photomultiplier tube. Thus, while one sample and hold arrangement is holding the sensor signal value for processing, the other sample and hold arrangement is open to receive a signal value corresponding to a subsequent scintillation event. A multiplexer is provided to enable selection between the sample and hold arrangements.

Comprehension of the invention is facilitated by reading the following detailed description in conjunction with the annexed drawing, in which:

FIG. 1 is an isometric representation of a detector arrangement for use in SPECT wherein scintillation bars are optically coupled to photomultiplier tubes via a segmented light guide;

FIG. 2 is a schematic representation of a one-dimensional geometry which is useful for explaining the position estimation problem;

FIG. 3 is a graphical representation of photomultiplier tube response functions plotted with respect to distance x in mm;

FIG. 4 is a graphical representation of weight factors plotted with respect to distance x in mm; and

FIG. 5 is a function block and line representation of a specific illustrative embodiment of the invention.

A detector intended for SPECT application is required to meet several requirments. These include high photo peak sensitivity for .sup.123 I and .sup.111 In. In addition, the detector must have greater than 90% packing fraction in ring geometry, an intrinsic resolution of 3 mm or better over the full module width, and continuous sampling in the axial (y) direction.

FIG. 1 is an isometric representation of a detector arrangement 10 which meets the foregoing conflicting requirements. As shown, detector arrangement 10 is provided with a plurality of sodium iodide bars 11 which may be 3 mm wide by 12.5 mm thick, and 15 cm long. The sodium iodide (NaI) bars are optically coupled to a Pyrex face plate 12 which, in this specific illustrative embodiment, has a thickness of 5 mm. Pyrex face plate 12 is optically coupled to a segmented light guide 13 which is designed to match the x and y light-spread functions to the diameters of a plurality of photomultiplier tubes 15. Segmented light guide 13 is dimensioned to space the photomultiplier tubes radially so that the scintillator edge does not extend much beyond the centers of those photomultiplier tubes arranged on the edge, such as photomultiplier tube 16.

NaI bars 11, which are thin scintillation bars, provide continuous sampling in the y direction and limit the light-spread function to 12.5 mm full width half maximum (FWHM) in the x direction. The light distribution is only minimally distorted at the module edge so that resolution and linearity can be maintained over essentially the full module width. The shape of the light-spread function from thin bars, such as those of the present detector, is much less dependent on interaction depth than is the case for continuous crystals. Thus, the detector response will be less sensitive to gamma ray energy.

FIG. 2 is a schematic representation which illustrates the one-dimensional geometry for the position estimation problem. Elements of structure having analogous correspondence to those of FIG. 1 are similarly designated. In FIG. 2, if the expected value of scintillation light intensity at the surface of the light guide for an interaction occurring at x is given by S(x-u), the mean output of the ith phototube of radius r as a function of source position may be written as: ##EQU1## where u.sub.i locates the center of photomultiplier tube PMT.sub.i. In accordance with a method described in an article entitled "One Dimensional Scintillation Cameras for Positron ECT Ring Detectors," by Burnham, et al., IEEE Transactions on Nuclear Science, Vol. NS-28, No. 1, February, 1981, which article is incorporated herein by reference, let M.sub.i represent a noisy measurement of S.sub.i. For a single event, assuming that the number of photons reaching the tube is governed by Poisson statistics, the conditional probatility of observing M.sub.i given x is: ##EQU2## If the M.sub.i are independent, the joint probability of observing a set of M.sub.i, M, for the N phototubes is: ##EQU3## If x is the value of x that maximizes PROB[M.vertline.x], it will also maximize the 1n(PROB[M.vertline.x]) so that we can write: ##EQU4##

Equation (7) describes a continuous sequence of inner products of the phototube outputs with the spatially-variant weighting functions W.sub.i (x) which have been calculated from the means tube responses. That value of x for which the product equals zero corresponds to x, the most likely value of x. In the actual case, the spatial coordinate is sampled discretely, and the sets of weights are represented as W.sub.ij.

The second term of the explicit weighting function shown in Equation (6) is a bias related to the spatial derivative of the total light collected. It compensates for changes in light collection at edges and between tubes. The weighting functions are also seen to be energy normalized which means if the light distribution is constant, the position estimator is energy independent. Instead of using an approximation in Equation (6) which assumes that the sum over M.sub.i (x) sum over S.sub.i, it is possible to use the first term of Equation (5) as the weight, and the condition for most likely position is satisfied when the weighted sum of photomultiplier outputs is equal to the second term (bias term) in Equation (5). This formation is of greater accuracy since the statistical fluctuations in M.sub.i do not multiply the bias term as they do in Equation (6). This form, however, makes the assumptions that each interaction deposits equal energy in the detector. If this is not the case, then the first form gives superior results.

If the two-dimensional sampled light-spread function, S.sub.ij (x,y), were separable and spatially invariant so that it could be written as S.sub.i (x)S.sub.j (y), then the two-dimensional problem could legitimately be reduced to two independent one-dimensional problems involving row and column sums of the phototube outputs and separate x and y weighting functions. This assumption is clearly violated in the case of a hexagonally packed array of round photomultiplier tubes. The correct solution is to perform a full two-dimensional likelihood calculation as described in an article entitle "Digital Position Estimation for the Modular Scintillation Camera, " by Milster, et al., IEEE Transactions on Nuclear Science, Vol. NS-32, pp. 748-752, 1985. This can be accomplished, but at the expense of greatly increased memory and/or computation time. The two-dimensional geometry was therefore simulated to determine how much distortion would result from approximating the two-dimensional solution by two one-dimensional solutions. Based on simulations and measurements on thin NaI bars, a light-spread function of the form: ##EQU5## was used with the FWHM in x and y equal to 38 and 55 mm, respectively.

An alternative solution is to cross-couple the x and y estimation process in an iterative fashion. The initial x and y position estimates are transmitted to the y and x estimation circuits, respectively. This information is used to select a new set of weights from weight memory which are most appropriate to the specific x,y region of the detector. This process can is then continued by iteration to obtain the correct two-dimensional solution.

FIG. 3 is a graphical representation of a plurality of measured photomultiplier tube response functions plotted with respect to distance x. As shown, the x-coordinate is calibrated in millimeters. FIG. 4 is a graphical representation of a plurality of weight factors plotted with respect to distance x. The weight factors in the embodiment of the invention assume both negative and positive values. As will be described hereinbelow, the responses of the photomultiplier tubes will be combined with the respectively associated weight factors to produce a weighted photomultiplier tube response signal for each photomultiplier tube. The weighted photomultiplier tube response signals are then summed to produce a composite weighted photomultiplier tube response signal. When this composite signal assumes a predetermined value, which is zero in the specific illustrative embodiment of the invention described herein, the corresponding value of x is the position of maximum liklihood for the event.

FIG. 5 is a function block and line representation of a specific illustrative embodiment of the invention. The circuit of FIG. 5 is used to compute the x or the y coordinates of each event. Thus, two such circuits are utilized to produce the coordinate pair in a two-dimensional embodiment.

In this embodiment, a plurality of input terminals 50 receive pulses from each row or column summing amplifier (not shown) which are coupled to the photomultiplier tubes (not shown in this figure). The pulses from each row or column are fed to two respectively associated sample and hold amplifiers 51A and 51B. In practice, the sample and hold amplifiers may be of a type known as Harris 5320. Pulses which satisfy a minimum energy criterion are stretched, by the holding operation of the sample and hold amplifiers, and presented at a reference input of a multiplying digital-to-analog converter (DAC) 53. The selection of whether the output of sample and hold amplifier 51A or sample and hold amplifier 51B is conducted to DAC 53 is performed by an analog multiplexer 54. In this specific embodiment, DAC 53 is a Signetics 4-quadrant 8-bit high-speed multiplying type. Thus, a single stage of analog buffering is used to derandomize the input pulses.

The stretched pulses from each input channel are also conducted to a respectively associated pair of diodes 61 and 62 which are coupled at their cathodes to respective inputs of a respective comparator 64. The maximum of the eight input voltages, minus the drop across the respective one of diodes 61, appears across resistor R0, and therefore at one input of all eight comparators 64 in this embodiment. Only for a channel with the maximum input voltage will both diodes 61 and 62 be conducting, and since the value of R0, is smaller than the values of resistors R1-R8, the voltage drop across the associated diode 61 of the pair will exceed the voltage drop across the associated diode 62. Thus, the corresponding comparator 64 will change states.

Comparators 64 are each connected at respective inputs of a priority encoder 65 which issues at its output a signal which identifies the particular one of comparators 64 which changed its state. In this manner, the search for the maximum liklihood position can be confined to the neighborhood of the maximum row or column by identifying the particular one of comparators 64 which is associated with the channel having the maximum input signal.

The output signal of priority encoder 65 is combined in register 83 with a search range offset selected in ROM 82 according to the position signal calculated for the other coordinate. The output of register 83 is used to select search start address from ROM 67 and a search stop address to be selected from ROM 68. These addresses are conducted to a weight memory 70 which contains within it memory locations for each of the eight channels in this embodiment. Once the start and stop addresses have been determined from ROMS 67 and 68, the weights arc synchronously clocked out in parallel to their respectively associated ones of DACs 53 in response to a clock 71. In this specific embodiment, DACs 53 provide complementary current outputs which permit negative weights, as shown in FIG. 4 hereinabove, to be used. The output currents of DACs 53 are each proportional to the product of the weight being clocked into them and the summed output of a row or column of phototubes (not shown). The sum of these products, which is equal to the local value of the superposition integral (Equation (7)), is conducted to a threshold-crossing detector 75. The threshold may be zero when the weights are calculated from Equation (6) or the threshold may be set equal to the bias term when the weights are taken as the first term of Equation (5). In this case, the bias voltage must be derived from values stored in a bias memory for each value of position by loading the digital bias values in to the bias digital-to-analog converter 84. In this specific illustrative embodiment of the invention, the bias memory is part of weight memory 70.

In accordance with this specific embodiment, the sensor signals at the inputs of DACs 53 is transformed to respective weighted sensor signals at the outputs thereof. These weighted signals are then summed algebraically in summing network 76 which delivers to threshold-crossing detector 75 a summed, or composite, weighted sensor signal.

In order to save processing time, the initial search is conducted in a coarse step mode. When zero crossing is detected at zero-crossing detector 75, a signal is conducted to control logic system 80 which controls the step size of the increments in weight memory 70. Consequently, the step size is reduced, and the scan direction is reversed to obtain a more precise location of the zero crossing. Scintillation events which do not satisfy a minimum energy criteria can be rejected at any time by control logic 80. In addition, control logic 80 controls analog multiplexers 54, controlling the system output at output register 81, and other system functions, including the iterative estimation of position.

Although the invention has been described in terms of specific embodiments and applications, persons skilled in the art can, in light of this teaching, generate additional embodiments without exceeding the scope or departing from the spirit of the claimed invention. Accordingly, it is to be understood that the drawing and description in this disclosure are proffered to facilitate comprehension of the invention and should not be construed to limit the scope thereof.