METHOD FOR PRODUCEING A GAMMA IMAGE TAKING INTO ACCOUNT SPATIAL UNEVENNESS OF SENSITIVITY

Description

DOMAINE TECHNIQUE

[0001] The technical field of the invention is X or gamma imaging, and more particularly the reconstruction of the position of irradiating sources using an image acquired by a gamma camera. ART ANTERIEUR

[0002] Gamma cameras are devices that create images to map radiation sources in a given environment. One application is visualizing radiation sources within an organism for medical diagnosis. Another application is locating radiation sources within a facility, particularly a nuclear facility.

[0003] The use of gamma cameras in the medical field is relatively old. In the nuclear sector, this type of device was developed in the 1990s and is becoming increasingly common in nuclear facilities for radiological characterization. The goal is to identify the main radiation sources present in a facility. Indeed, radiation sources are not distributed homogeneously. They are often concentrated locally, in the form of "hot spots," a common term in the field of radiation protection. A gamma camera offers the advantage of remotely locating these hot spots.

[0004] Some gamma cameras consist of a two-dimensional pixel array connected to a detector material. The detector material is usually a semiconductor material, such as CdTe or CdZnTe. When ionizing radiation interacts with the detector material, one or more pixels generate an electrical pulse whose amplitude is correlated with the energy released by the radiation during the interaction. Each pixel is connected to an electronic pulse processing circuit.

[0005] Each pixel consists of an electrode, which usually acts as the anode. When incident radiation interacts with the detector material, electrons are released. These electrons are collected by the anode. The anode generates a pulse whose amplitude depends on the number of electrons collected, a number generally proportional to the energy lost by the ionizing radiation in the detector material.

[0006] Each pixel spans a few millimeters on each side. For compactness, the pixel array typically contains one hundred, or a few hundred, pixels per row and per column. To achieve sufficient spatial resolution, each pixel can be "sub-pixelized," that is, subdivided into virtual pixels. Methods have already been developed that allow each interaction to be associated not with a pixel, but with a virtual pixel. Such methods exploit the fact that when an interaction occurs in the detector material, charge carriers migrating toward the anode array generate a signal detectable by several adjacent pixels. Thus, these methods are based on a combination of the signals detected by several adjacent pixels. One such method is described, for example, in the publications of Warburton W.K, “An approach to sub-pixel spatial resolution in room temperature X-ray detector arrays with good energy resolution” as well as in Montemont et al. “Studying spatial resolution of CZT detectors using sub-pixel positioning for SPECT”, IEEE Transactions on Nuclear Science, Vol. 61, No. 5, October 2014, and also in patent US9322937B2. Using these methods, the size of the virtual pixels can reach, for example, 0.5 mm * 0.5 mm, or 0.1 mm by 0.1 mm.

[0007] The publication Yuefeng Zhu et al "Sub-pixel position sensing for Pixelated 3-D position sensitive, wide band-gap, semiconductor,r gamma ray detector" describes a sub-pixelation method for a Compton-type gamma camera.

[0008] However, the inventor observed that applying such methods leads to problems of non-uniformity in the gamma camera's response. When the pixel array is irradiated by a uniform photon flux, the sub-pixelation into virtual pixels results in a spatially non-uniform response from the gamma camera: the number of interactions associated with "central" virtual pixels, located at the center of a pixel, is overestimated, to the detriment of the number of interactions associated with "peripheral" virtual pixels, located at the periphery of a pixel. Such non-uniformity can be detrimental when the image formed by the gamma camera is processed to reconstruct an image of the irradiating sources in the field of view. This is the case, for example, with gamma cameras using a coded mask collimator. Non-uniformities can generate artifacts during the reconstruction of the map showing the position of the irradiating sources.

[0009] There are other causes of non-uniformity in the response of a gamma camera. In nuclear facilities, radiation sources, known as "out-of-field" sources, are located outside or at the edge of the field of view. Such radiation sources generate stray radiation that can affect the image formed by the gamma camera. Gamma cameras have a collimator, which defines the field of view. When the collimator is a coded mask type, the image formed by the gamma camera undergoes reconstruction to account for the presence of the coded mask. Radiation sources located at the edge of the field of view can generate artifacts during reconstruction. When the collimator is a pinhole collimator, highly irradiating sources located outside the field of view can produce a non-uniform veil affecting the image acquired by the gamma camera.

[0010] Another cause affecting gamma camera non-uniformity is the presence of defects in the detector material, locally altering detection sensitivity.

[0011] The invention described below addresses this problem, and makes it possible to reduce the spatial non-uniformities of a gamma camera. EXPOSE DE L'INVENTION

[0012] A first object of the invention is a method for determining a spatial sensitivity function of a gamma camera according to claim 1,

[0013] According to a first embodiment: The pixels are distributed according to rows and columns on the detection surface; each group of pixels contains pixels belonging to the same row or the same column; step e) comprises the following sub-steps: ei) formation of a first vector, comprising first terms, each first term being associated with a column of pixels, each first term comprising a sum of the quantity of interactions positioned in each pixel of the column; e-2) formation of a second vector, comprising second terms, each second term being associated with a row of pixels, each second term comprising a sum of the quantity of interactions positioned in each pixel of the row; step f) comprises the following sub-steps: f-1) for each virtual pixel, calculation of a product of a first term, associated with the column of the pixel, by a second term, associated with the row of the pixel;f-2) determination of the weight assigned to the pixel as a function of the product calculated during sub-step f-1). ;

[0014] Substep f-2) may include: calculation of an average value of the products respectively calculated, during substep f-1), for a set of pixels; for each pixel, normalization of the product resulting from substep f-1) by the average value. - .

[0015] Regardless of the method of implementation, the process may include: g) development of a sensitivity matrix, each point of the sensitivity matrix corresponding to a virtual pixel of the gamma camera, the value of the sensitivity matrix at each point corresponding to the weight assigned to said pixel.

[0016] The gamma camera may include a processing unit configured to process interactions stored in memory. The method includes a step (h) of reconstructing a spatial distribution of the irradiating sources in the field of view, based on the interactions stored in step (c) and the weights assigned to each pixel in step (f). The processing unit can normalize the number of interactions detected by each virtual pixel by the weight assigned to that pixel.

[0017] Preferably, Each pixel is linked to a spectrometry unit, the spectrometry unit being configured to classify each detection signal into an energy band among several energy bands; in step b), each interaction positioned in a pixel is associated with a detected energy band; step h) includes: taking into account at least one isotope, the isotope emitting photons in at least one emission energy; reconstructing a spatial distribution of irradiating sources including the isotope from the interactions positioned in step b).

[0018] Step h) may include: taking into account a spectral response function of each pixel, the spectral response function representing a probability of detection, in different energy bands, of each photon emitted by the isotope and detected in the pixel; taking into account a spatial response function, the spatial response function representing a probability of detection, in different pixels, of a photon emitted from a point in the field of observation.

[0019] According to one possibility, the gamma camera includes a collimator defining the field of view, in particular a coded mask collimator.

[0020] A second object of the invention is a gamma camera, intended to detect the presence of radiating sources in an observation field, the gamma camera comprising: a detector material; pixels, distributed over a detection surface of the detector material, each pixel being configured to form a detection signal under the effect of a detection of an interaction of an ionizing photon in the detector material; a memory, configured to store a quantity of interactions detected during an acquisition period and respectively assigned to each pixel; a processing unit, configured to process the interactions stored in the memory, the processing unit being configured to implement steps d) to f) of a process according to the first object of the invention from the detected interactions.

[0021] The processing unit is configured to implement steps d) to f) of the variant of the first embodiment previously described in relation to the first object of the invention.

[0022] A third object of the invention is a method for reconstructing a spatial distribution of radiating sources in the field of observation of a gamma camera according to the second object of the invention, the method comprising the following steps: i) acquisition of detection signals by the pixels of the gamma camera during an acquisition period, each detection signal being associated with a detected interaction; ii) as a function of the detection signals, assignment of a position of each detected interaction, during the acquisition period, the position corresponding to a pixel I; iii) storage, by memory, of a number of detected and assigned interactions in each pixel; the process being characterized in that it comprises: (iv) taking into account weights, respectively assigned to each pixel I, each weight being determined by implementing steps (a) to (f) of the method according to the first object of the invention, step (a) being carried out by exposing the gamma camera to an observation field identical or different from the observation field to which the gamma camera is exposed during step (i); (v) reconstruction of the spatial distribution of the irradiating sources, in the observation field, from the interactions memorized during step (iii) and the weights assigned to each pixel during step (iv).

[0023] According to one embodiment: in step i), the field of observation is identical to the field of observation of the gamma camera in step a); step ii) is coincident with step b); step iii) is coincident with step c).

[0024] According to one embodiment: in step i), the field of view is different from the field of view of the gamma camera in step a); step iv) is implemented taking into account weights assigned to each pixel, the weights having been established during an implementation of steps a) to f), prior to or subsequent to step i).

[0025] The invention will be better understood by reading the explanation of the examples of embodiment presented, in the continuation of the description, in connection with the figures listed below. FIGURES

[0026] There figure 1A The diagram illustrates the main components of a gamma camera enabling the implementation of the invention. figure 1B shows a segmentation of pixels into virtual pixels (or sub-pixels). figure 2A shows a probability distribution of an estimated position, per unit of sub-pixelation of the gamma camera, as a function of the actual position of an interaction, the position being defined parallel to the detection surface. figure 2B illustrates a non-uniformity in the spatial sensitivity function of the gamma camera. figures 2C et 2D represent a number of interactions assigned to different virtual pixels, the latter being defined within pixels delimited by solid lines or columns. figures 2C et 2D respectively illustrate the actual positions of the interactions and the positions of the interactions resulting from the sub-pixelation unit. figure 2E shows an example of a gamma image acquired by a gamma camera as schematically depicted on the figure 1A The spatial sensitivity function is non-uniform. The gray level of each virtual pixel corresponds to a number of interactions detected during an acquisition period. figure 3A This diagram outlines the main steps in a process for determining the spatial sensitivity function of different virtual pixels of a gamma camera (steps 100 to 120), as well as the step of reconstructing an object image, showing a spatial distribution of radiating sources in the field of view (step 130). figure 3B illustrates components of the gamma camera described in connection with the figure 1A , as well as spatial or spectral response functions associated with these components. The figures 4A à 4D illustrate a first embodiment of the invention. figure 4A shows an example of a gamma image acquired by the gamma camera described in connection with the figure 1A . THE figures 4B et 4C These diagrams respectively represent the terms of a first vector and a second vector that allow the determination of the spatial sensitivity function of the gamma camera. figure 4D is a matrix representation of a spatial sensitivity function of the gamma camera obtained according to the first embodiment. figures 5A à 5C illustrate a second embodiment of the invention. The figure 5A This shows an example of a gamma image acquired under field conditions, with a strong contribution from scattered radiation. figure 5B represents an average image, in which the value of each point corresponds to an average of the number of interactions detected by virtual pixels of the same rank. figure 5C is a matrix representation of a spatial sensitivity function of the gamma camera obtained according to the second embodiment. figure 6 This shows a spectral response function of a pixel from the gamma camera, for two different isotopes. The x-axis corresponds to an energy channel and the y-axis represents a detection probability within the energy channel. figures 7A et 7B represent respectively a gamma image acquired by a gamma camera and an object image reconstructed from the gamma image, without implementation of the invention. figure 7C shows the gamma image of the figure 7A after application of a spatial sensitivity function, the latter being determined according to a first embodiment of the invention. figure 7D represents respectively an object image reconstructed from the gamma image shown on the figure 7C . There figure 7E shows the gamma image of the figure 7A after application of a spatial sensitivity function, the latter being determined according to a second embodiment of the invention. figure 7F represents respectively an object image reconstructed from the gamma image shown on the figure 7E . EXPOSE DE MODES DE REALISATION PARTICULIERS

[0027] There figure 1A represents a gamma camera 1, enabling an implementation of the invention. The gamma camera 1 is configured to detect ionizing photons, of the X-ray or gamma type, whose energy is generally between 10 keV and 10 MeV, in an observation field Ω. The observation field Ω extends around a central axis Δ. The gamma camera 1 can be coupled to a visible camera, enabling the formation of a visible image of the observation field. The term gamma camera refers to an imager presenting an observation field and configured to form an object image. O allowing localization of irradiation sources in the observation field Ω.

[0028] The gamma camera 1 comprises a detector material 11, usually a semiconductor material that generates charge carriers (electron / hole pairs) upon interaction with X-rays or gamma rays. Examples include CdTe or CdZnTe. Generally, the detector material is well-suited to interact with ionizing photons to generate charge carriers. The detector material is preferably a semiconductor. Alternatively, it can be a scintillator material coupled to a photodetector.

[0029] The gamma camera 1 has 121 ... 12i ... 12l pixels, distributed across a detection area. The pixels are represented on the figure 1B The index i represents a rank of each pixel 12 i. Iis the number of pixels. Generally, the 12i pixels are coplanar and arranged in a two-dimensional, preferably regular, matrix. The matrix may, for example, contain a few dozen or even a few hundred pixels. Nxi corresponds to the number of pixels along the X-axis and Nyi to the number of pixels along the Y-axis. Each 12i pixel is an elementary radiation detector. In the example described, each pixel is an anode, whose polarization allows it to collect electrons produced during an interaction occurring in the detector material.

[0030] The surface area of ​​each pixel 12i can be relatively large, on the order of a few mm². When an ionizing photon interacts with the detector material 11, charge carriers, for example electrons, migrate towards one or more pixels, these being called hit pixels: each hit pixel is a charge-carrier pixel. An interaction can result in one or more hit pixels. As described in the prior art, during their migration towards a hit pixel, the charge carriers generate a signal, usually referred to as the induced signal, on the pixels adjacent to the hit pixel(s).

[0031] In general, each interaction results in the formation of a detection signal by at least one pixel, and most often several pixels. The detection signal can be a signal resulting from the collection of charge carriers by pixel 12i or a signal induced by the migration of charge carriers in the detector material 11. In order to improve the spatial resolution of the gamma image, the gamma camera includes, in an unclaimed embodiment, a subpixelization unit 14, programmed to assign a position (x,y) of each detected interaction, parallel to the detection surface 12, based on detection signals formed by several pixels 12i following each interaction. In the following description, the position of an interaction corresponds to a position of the interaction parallel to the detection surface defined by pixels 12i.

[0032] The gamma camera may include a collimator 10 to define the field of view Ω, which may contain irradiating sources 5. The pixels are exposed to irradiation from the irradiating sources in the field of view during an acquisition period. During this acquisition period, the pixels acquire detection signals resulting from interactions of ionizing photons emitted by the irradiating sources located in the field of view. The collimator 10 may be of the pinhole or coded mask type. In the following example, the collimator 10 is of the coded mask type.

[0033] The gamma camera can be of the Compton gamma camera type, in which case a collimator is not necessary. A Compton gamma camera has a specific electronic circuit that allows it to estimate the respective positions, within the detector material, of two temporally coincident interactions, and to estimate a direction of propagation of the incident radiation.

[0034] In an unclaimed embodiment, the subpixelization unit 14 divides each pixel 12i into virtual pixels (or subpixels) 13i,j. The term "virtual pixel" refers to the fact that a virtual pixel 13i,j has no physical reality: it results from a virtual segmentation of each physical pixel 12i. The index j is an integer corresponding to the rank of each virtual pixel 13i,j within a pixel 12i, with 1 ≤ j ≤ J. J corresponds to the number of virtual pixels in each pixel. The rank jThe position of a virtual pixel 13j,j defines the position of the virtual pixel within a pixel 12i. Virtual pixels 13i,j of the same rank are arranged in the same position relative to pixel 12i. In the example shown on the figure 1B the rank j = 1 corresponds to the top left corner and the row j = J corresponds to the lower right corner. In this figure, J = 9: each pixel is segmented into 9 virtual pixels. Subpixelization allows for the formation of a gamma image G comprising a number of virtual pixels 13 i,j equal to I × J. The virtual pixels 13 i,j are aligned along rows (parallel to an X-axis) and columns (parallel to a Y-axis). The position of each virtual pixel 13 ij on the detection surface 12 is located by coordinates ( x , y ) . The link between each virtual pixel 13 ij and each position ( x , y) is bijective, such that at each coordinate position ( x , y ) corresponding to a single virtual pixel 13 j,j , and vice versa.

[0035] The gamma camera 1 may include a unit programmed to determine the depth of interaction in the detector material, from detection signals formed by several pixels.

[0036] Under the effect of charge carrier collection, each pixel 12i generates a pulse whose amplitude depends on the energy released in the detector material by an ionizing photon during an interaction; this energy is usually referred to as the "interaction energy." Optionally and advantageously, the gamma camera 1 includes a spectrometry unit 15. The spectrometry unit 15 allows for the most accurate possible estimation of the amplitude of the pulses resulting from charge carrier collection following an interaction. The spectrometry unit 15 can include both electronic means (pulse shaping circuit, multichannel analyzer, analog-to-digital converter) and software means. Estimating the amplitude of a pulse allows for the estimation of the interaction energy. This energy must be estimated as precisely as possible.The energy range addressed is generally between 10 keV and a few hundred keV, or even a few MeV. It is desirable that the energy accuracy be on the order of a percent, or even lower.

[0037] Thus, the spectrometry unit 15 allows us to obtain a spectrum of the radiation detected by each pixel. The spectrometry unit 15 allows us to select energy bands of interest, corresponding to unscattered photons, that is, photons that have not been deflected since their emission by the radiation source. Their selection, within predetermined energy bands, makes it possible to eliminate noise corresponding to scattered photons. Since these scattered photons have been deflected since their emission, they do not provide useful information for locating the radiation sources. Scattering is therefore a source of noise, which spectrometry allows us to significantly limit. Each energy band E extends between E ± δE. 2 δE then corresponds to the spectral width of each energy band. For example, 2 δE = 0.2 keV.

[0038] Another advantage of gamma spectrometric cameras is that knowing the energy of the detected photons allows for the identification of the isotopes responsible for the irradiation. This is important information in the fields of radiation protection, radioactive waste management, nuclear facility decommissioning, and post-accident radiological characterization.

[0039] Gamma camera 1 has a memory 16, configured to store a quantity of interactions G ( x, y ) respectively assigned to each virtual pixel 13 ij of position ( x, y ) . The gamma camera may include a processing unit 17 configured to form a gamma image Gfrom the interactions positioned, by the sub-pixelation unit 14, in each virtual pixel 13 i,j. The gamma image G is defined according to coordinates ( x, y ), parallel to the detection surface 12, each coordinate ( x, y ) corresponding to a virtual pixel 13 i,j . Each point G ( x, y ) of the gamma image G corresponds to a quantity of interactions assigned, by the sub-pixelation unit 14, to a virtual pixel 13 i,j of coordinates ( x, y ) on the detection surface 12. Typically, the processing unit may include a microprocessor programmed to execute instructions to implement certain steps described below, in connection with the figure 3A .

[0040] When the spectrometry circuit 15 is implemented, the memory 16 can store a quantity of interactions G E ( x, y ) in several energy bands Erespectively assigned to each virtual pixel with coordinates ( x, y ) . The processing unit 17 can generate gamma images from the same field of view G E respectively representative of an energy band E. Knowing the emission spectrum of an isotope, it is also possible to combine different energy bands to form a gamma image G k ( x, y ) corresponding to the isotope. The index kThis refers to an isotope. The combination of different energy bands is based on the isotope's known energy emission probabilities. The decay patterns of isotopes likely to constitute the irradiating sources of the observation field are then taken into account. An isotope's decay pattern refers to its emission energy(s) as well as its branching rates (the emission probabilities of a photon at different emission energies).

[0041] The gamma camera 1 includes a processing unit 18, configured to form an object image O from the gamma image G resulting from the subpixelization unit 14, or, more generally, from a quantity of interactions detected by each virtual pixel, possibly by energy bands. The object image formation can follow a reconstruction algorithm, as described below.

[0042] The inventors observed that the segmentation into virtual pixels 13 ij is a source of non-uniformity in the gamma camera's sensitivity. figure 2A This represents, for a pixel 12l, a spatial response function of the subpixelization unit. This figure was obtained by simulation, using a subpixelization unit as described in US9322937B2, and a pixel with a side length of 2.5 mm. The y-axis corresponds to the actual position of an interaction, along a longitudinal axis X, within a pixel. The x-axis corresponds to a "virtual" position assigned by the subpixelization unit 14. In this example, the pixel was virtually subdivided into 100 increments along the X-axis. Each increment corresponds to a spatial step of 25 µm. The gray level corresponds to the probability value of assigning the position located on the x-axis, while the actual position corresponds to the position located on the y-axis.Each row of the matrix corresponds to a probability distribution for assigning a position to an interaction, resulting from the sub-pixelation unit, given the actual position of the interaction. Each column of the matrix corresponds to a probability distribution for the actual position of an interaction, given an interaction position resulting from the sub-pixelation unit.

[0043] We observe that the probability of assignment is more widely distributed at the center of the pixel (coordinate 50 on the figure 2A ) than at the edges of the pixel (coordinates 0 and 100 on the figure 2A ). There figure 2A was simulated taking into account a photon energy of 122 keV. figure 2A depends on the energy of the photons. The higher the energy, the narrower the probability distribution for assigning a position.

[0044] On the figure 2B The solid line curve corresponds to a projection of the figure 2A on the x-axis. This corresponds to a probability distribution for assigning each "virtual" position along the X-axis when a pixel is irradiated with spatially homogeneous radiation. On the figure 2B The x-axis represents a position within the pixel, with x-coordinate 0 corresponding to the pixel center and x-coordinates 0.5 and -0.5 corresponding to the left and right edges of the pixel. The y-axis represents the probability that a virtual pixel, corresponding to the x-coordinate, is considered by the sub-pixelation unit 14 as corresponding to the position of an interaction, while the spatial distribution of the irradiation is homogeneous. It is observed that the sub-pixelation unit 14 tends to overestimate the number of interactions considered to be located at the center of the pixel, at the expense of the number of interactions considered to be located at the edge of each pixel. In other words, central virtual pixels (i.e., at the center of a pixel) have a higher sensitivity than peripheral virtual pixels (i.e., at the periphery of a pixel).The dotted line shows the probability distribution of the position assigned to each interaction in the case of a subpixelization unit exhibiting uniform spatial sensitivity.

[0045] THE figures 2C et 2D represent a portion of a sub-pixelated gamma image, i.e., resulting from the sub-pixelation unit 14, a grid of black lines representing the outlines of different pixels. In these figures, each virtual pixel corresponds to a point. The figures 2C et 2D results from simulations considering homogeneous irradiation. figure 2C corresponds to the actual position of the interactions. The figure 2D shows the positions of the interactions by implementing the subpixelization unit 14. In these figures, the darker the gray level, the greater the quantity of interactions detected. We observe that the subpixelization unit induces a dithering of the gamma image: the virtual pixels located at the center of each pixel are overweighted, to the detriment of the virtual pixels located at the edges of the pixels. This is a consequence of the spatial distribution represented on the figure 2B .

[0046] There figure 2E is a gamma image G of uniform radiation, essentially emitted by 60< Co, (1173 keV and 1332 keV). This is an image acquired on-site, with a contribution from scattered radiation. In this example, the detection area 12 has rows of 16 pixels and columns of 16 pixels, for a total of 256 pixels. Each pixel covers an area of ​​2.5 mm by 2.5 mm. Each pixel is subpixelized inJ = 64 virtual pixels (8 x 8 virtual pixels). Thus, the resulting gamma image extends along 128 pixels on each side. On the figure 2E The lighter the shade of gray, the greater the number of interactions detected. The dithering mentioned in connection with the figure 2D is evident: the edges of each pixel can be visualized, since they have fewer detected interactions than the central parts of each pixel. Thus, spatially homogeneous irradiation leads to the formation of a spatially non-homogeneous gamma image. When the gamma image undergoes a reconstruction process, as described later, such non-uniformity can lead to the formation of artifacts on the reconstructed image: cf. figures 7A et 7B described later.

[0047] THE figures 2A à 2E reflect the fact that the sub-pixelation unit 14 is affected by a non-uniform spatial sensitivity function, the sensitivity of the virtual pixels at the center of each pixel being greater than the sensitivity of the peripheral virtual pixels, the latter being adjacent to the edges of each pixel.

[0048] To correct the non-homogeneity of the non-uniform spatial sensitivity of the subpixel unit 14, one solution is to expose the gamma camera to homogeneous radiation, thus forming a calibration image that takes the response function into account. This is a technically simple solution, but difficult to implement. Indeed, the spatial sensitivity function depends on the energy of the incident radiation. However, the energy of the incident radiation can vary depending on the isotopes composing the irradiating sources. Moreover, it is difficult to predict, particularly when the irradiating sources are composed of several isotopes in varying and unknown proportions, and / or when the irradiation to which the gamma camera is exposed contains a high proportion of scattered radiation.

[0049] The inventor proposes a method for determining a spatial sensitivity function. This spatial sensitivity function corrects the non-homogeneous spatial response of the subpixelization unit. The correction can be implemented in the field without the need for homogeneous calibration radiation. A particularly important aspect is that the spatial sensitivity function can be derived from the gamma image acquired by the gamma camera and reconstructed to determine the position of the irradiating sources within the field of view. The method does not require a calibration image acquired beforehand under laboratory conditions. The calibration image can be created directly from the gamma image acquired by the camera in the field.

[0050] There figure 3A The diagram outlines the main steps of the process, which are described below. figure 3B represents the different quantities used and schematically represents the response functions, spatial or spectral, respectively associated with the collimator 10, the sub-pixelation unit 14, and the spectrometry circuit 15.

[0051] Etape 100 : detection of interactions. The gamma camera 1 is arranged in an observation field Ω potentially containing irradiating sources 5. During an acquisition period, from detection signals emanating from pixels 12i, the sub-pixelation unit 14 assigns each detected interaction to a virtual pixel 13j,j, with coordinate ( x, y ) . This yields a quantity G ( x, y ) of interactions assigned to each virtual pixel.

[0052] Etape 110 : formation of a gamma image G.

[0053] The quantities G ( x, y The values ​​assigned to each pixel can be gathered and ordered into a gamma image G. Each gamma image G is defined by rows and columns. figure 4A is an example of a gamma image G 128 x 128 pixels.

[0054] The value G ( x, y ) of the gamma image, for each virtual pixel of position ( x, y ) corresponds to the number of interactions detected and positioned in the virtual pixel.

[0055] Etape 120 Determining a weight H ( x, y ) for each virtual pixel.

[0056] Based on the quantity of interactions G ( x, y ) detected by each virtual pixel 13 ij of coordinates ( x, y ), step 120 determines a weight H ( x, y ) for each virtual pixel. The weight H ( x, y ) represents the gamma sensitivity of the camera for each virtual pixel. In the example described, the higher the weight, the higher the sensitivity of the virtual pixel: this corresponds to the fact that the probability of an interaction affecting the virtual pixel is high. The weight H ( x, y ) is an estimate of the probability that an interaction, whose coordinates, on the detection surface 12, is assigned to said virtual pixel 13 ij of coordinates ( x , y ). When the spatial sensitivity function of the subpixelization unit 14 is homogeneous, the probability is equal for all virtual pixels 13 ij .

[0057] From the gamma image G As a result of the image formation unit, a calibration image can be formed. H , representative of the subpixelization unit's response. The calibration image His defined for each virtual pixel 13 i,j. Each point H ( x, y ) of the calibration image H is a weight H ( x, y ) assigned to the virtual pixel with coordinates ( x, y ) .

[0058] Generally speaking, the weight H ( x, y The interaction quantity of each virtual pixel is obtained by dividing the virtual pixels into different groups and adding the interaction quantity values. G ( x, y ) assigned to the virtual pixels of the same group. Different embodiments are envisaged, in which the groups of virtual pixels are: virtual pixels arranged in the same row or column: see substeps 121 to 123. virtual pixels of the same rank j : cf. sub-steps 125 to 126.

[0059] Two embodiments can be envisaged. A first embodiment is described in relation to substeps 121 to 123. A second embodiment is described in relation to substeps 125 to 126.

[0060] Sous-étape 121 : formation of a first vector H X , representative of the total quantity of interactions G ( x, y ) detected in the virtual pixels of each column (same coordinate) y ).

[0061] Based on the quantities of interaction G ( x, y ) detected by each virtual pixel, we establish a first vector H X . The first vector H X is determined for different columns of virtual pixels, which correspond to different columns of the gamma image G. Each term H X ( y The first term of the first vector is designated as the "first term". Each first term H X ( y) is associated with a column of virtual pixels. Each first term corresponds to a number of interactions assigned to virtual pixels in the same column. Thus, each first term is, for example, such that: H X y = ∑ x = 1 x = N x G x y Or N x is the number of virtual pixel lines considered, which corresponds to the number of lines in the gamma G image.

[0062] The dimension of H X East ( N y , 1), where N y is the number of virtual pixel columns. The vector H X can be considered as a projection of the gamma image G, along the axis Y, on the axis X.

[0063] There figure 4B is a graphical representation of a first vector H X established according to the gamma image shown on the figure 4A The x-axis corresponds to a position of each term in the first vector. H X , between 1 and 128. The y-axis corresponds to the value of each first term of the first vector.

[0064] Sous-étape 122 : formation of a second vector H Y , representative of the total quantity of interactions G ( x, y ) detected in the virtual pixels of each line (same coordinate) x ) .

[0065] Similarly, based on the quantities of interaction G ( x, y ) detected by each virtual pixel, a second vector is established H Y . The second vector H Y is determined for different virtual pixel lines, which correspond to different lines of the gamma image G. Each term H Y ( x The term of the second vector is designated as the "second term". Each second term H Y ( x) is associated with a line of virtual pixels. Each second term corresponds to a number of interactions assigned to virtual pixels in the same line. Thus, each second term is, for example, such that: H Y x = ∑ y = 1 y = N y G x y

[0066] The dimension of H Y is (1, N x The vector H Y can be considered as a projection of the gamma image G, along the axis X, on the axis Y.

[0067] There figure 4C is a graphical representation of a second vector H Y , established according to the gamma image shown on the figure 4A The x-axis corresponds to a rank of each term in the vector H Y , between 1 and 128. The y-axis corresponds to the value of each term. Substep 123: Training of the sensitivity matrix.

[0068] We perform the dyadic product of HX And HY , in order to obtain a matrix H XY , in size, N x N y : H XY = H X ⊗ H Y ⊗ is the "dyadic product" operator.

[0069] Thus, the value H XY ( x , y ) of each term of the matrix H XY East HX ( y ) HY ( x ) : H HY x y = H X y × H Y x

[0070] The matrix H XY is preferably normalized by the mean H XY ( x, y ) of the set of terms H XY ( x, y ) . Thus, the sensitivity matrix H is such that H x y = H XY x y H XY ¯ x y

[0071] The sensitivity matrix H is the same size as the gamma image G. Each term H ( x, y ) of the sensitivity matrix H corresponds to a weight, assigned to the virtual pixel with coordinates ( x, y ) .

[0072] There Figure 4D shows a sensitivity matrix H established by combining the vectors HX And HY .We can observe that the sensitivity matrix H highlights the gamma image dithering discussed above. The sensitivity matrix H features dark horizontal or vertical lines, which correspond to areas where the weight H ( x, y ) assigned to each virtual pixel is low.

[0073] According to another embodiment, described in connection with substeps 125 and 126, each group of virtual pixels gathers virtual pixels of the same rank.

[0074] Substage 125 : Based on the quantities of interaction G ( x, y ) detected by each virtual pixel, resulting from step 100, we calculate an average of the values ​​of virtual pixels of the same rank. The Figure 5A represents another example of a gamma image G, acquired in the field, with a significant amount of scattered radiation. The Figure 5B shows an average image H'each point of which corresponds to an average of the virtual pixels of the same rank in the gamma image G. The average image H' includes J points, J corresponding to the number of virtual pixels per pixel.

[0075] Each term H' ( xj , yj ) of the average image H' is such that H ′ x j y j = mean G x y j Or G ( x, y ) j corresponds to the quantity of interactions detected by each virtual pixel of rank j . mean denotes the average operator.

[0076] According to one possibility, the average image H' is standardized, such that: H ′ x j y j = mean G x y j mean G x y

[0077] There Figure 5B represents an average image H' obtained from the gamma image shown on the Figure 5A .

[0078] Substep 126 : from the average image H', calibration image formation Hby concatenation of the average image: the average image H' is duplicated N Xi times along the X-axis and N Yi times along the Y-axis, N Xi and N Yi denoting the number of pixels 12 i per row and column of the detection surface 12 respectively. This yields a calibration image H of the same size as the gamma image G. There Figure 5C represents an example of a calibration image obtained by concatenating an average image H'.

[0079] The embodiment described in connection with substeps 125 and 126 assumes a certain spatial homogeneity of the radiation detected by the gamma camera. This embodiment is particularly suitable for correcting non-uniformities resulting from the sub-pixelation unit.

[0080] Step 120 allows us to define, for each virtual pixel, a weight H ( x, y ) which is then used to process the quantities of interactions G ( x, y ) Or GE ( x, y ) assigned to each virtual pixel during exposure, and stored in memory 16. The processing aims to form an object image O of the observation field Ω. By object image, we mean a spatial distribution of radiating sources in the observation field Ω. Depending on one possibility, an object image is formed O k representative of a spatial distribution of radiating sources containing an isotope k.

[0081] When using a pinhole collimator 10, the processing performed by the processing unit 18 is relatively simple. Indeed, the gamma image allows for an immediate representation of the object image. The processing can be a simple normalization of the gamma image. G by the calibration image H.

[0082] In this example, gamma camera 1 has a coded mask collimator 10. This type of collimator is known to those skilled in the art. With this type of collimator, the image acquired by the imager does not allow a direct representation of the radiating sources in the field of view. The gamma image G undergoes processing, taking into account a camera response function, in order to obtain an image of the observation field, representative of the position of the sources in the observation field Ω. The transition from the gamma image, representative of a quantity of interactions detected by each virtual pixel, after application of the spatial sensitivity function, to the image O The field of observation Ω is described below. The image O is discretized into different coordinates ( u , v ). Step 130 reconstruction

[0083] The overall image formation model is such that: G E x y = H x y ∑ k S E k M u v ∗ O k u v x y Or : GE is a gamma image formed in an energy band E, the image being defined at different points ( x , y ), each point corresponding to a virtual pixel. This is a measured quantity, obtained from the interactions detected during step 100, the energy band selection being carried out by the spectrometry unit 15. H ( x, y ) is the weight assigned to each virtual pixel with coordinates ( x, y ), resulting from step 120. S ( k, E ) is a probability of detecting a photon in the energy band E for a unit activity (for example 1 Bq) of an isotope k. S is a spectral response matrix of size ( N k , NE ), Or I k is the number of isotopes considered and NE is the number of energy bands addressed. This is an input data. Stakes into account a spectral response function of the gamma camera. Figure 6 represents two rows of a spectral response matrix S, corresponding respectively to 241< Am and 57< Co. On the Figure 6 The x-axis corresponds to a channel number, between 1 and 1024. We observe the emission peaks of 241< Am and 57< Co, corresponding respectively to energies of 59 keV (241< Am) and 122 keV and 136 keV (57< Co). M is a spatial response function of the gamma camera for a photon emitted from a position ( u, v ) of the observation field Ω and reaching the virtual pixel ( x, y ) . This is an input data point, which depends on the imaging modality used. O k ( u, v ) is an activity of the isotope k to the position ( u, v ) of the field of observation. This is what we wish to estimate. O k is an object image of size ( N u , N v ), Or N u And N v are the coordinate numbers u And v according to which the field of observation is discretized.

[0084] The general form of the spatial response function of the gamma camera is M ( x, y, u, v ) . When gamma camera 1 has a collimator 10, for example a coded mask collimator, the coordinates u , v are spatial coordinates. When the gamma camera is a Compton camera type, without a collimator, the coordinates u , v are angular coordinates, corresponding to the angles of incidence of the detected radiation.

[0085] Expression (10) assumes that the field of observation is considered as a surface S Ω, called the object surface, discretized according to coordinates ( u , v ) .

[0086] There Figure 3BDiagram the main components of the gamma camera and their respective response functions: Collimator 10 determines the spatial response function M, which conditions a probability that an interaction detected in a virtual pixel of position ( x , y ) comes from a coordinate ( u , v ) of the object surface S Ω. The spectrometry circuit 15, along with the detector material 11 and the pixels of the detection surface 12, determine the spectral response function S. The sub-pixelation unit 14 determines the sensitivity matrix H, which includes all the weights H ( x, y ) assigned to the different virtual pixels.

[0087] The transition from the detection surface 12 to the object surface S Ω is a backprojection R of the gamma image G,on the object surface. The passage of the object surface S Ω at the detection surface 12 is a projection P of the object image O on the detection surface 12.

[0088] Expression (10) corresponds to a projection P. Expression (10) can be written as follows: G ^ E x y = H x y ∑ k S k E ∑ u , v M x − u , y − v O k u v

[0089] The objective of the reconstruction stage is to estimate Ô k ( u, v ) from one or more gamma images GE , respectively determined in one or more energy bands E.

[0090] The estimate of Ô k ( u, v ) assumes knowing the spatial response M, the spectral response S and the sensitivity matrix H, resulting from step 120, and representing the spatial sensitivity function.

[0091] According to a probabilistic approach, O ^ k u v = ∑ E , x , y p k , u , v E x y G E x y p ( k, u, v | E, x, y) is a probability of the presence of an isotope k to the position ( u , v ) knowing a measure of an interaction in an energy band E by the virtual pixel ( x, y This probability can be estimated by applying Bayes' theorem. p k u v E , x , y = p E x y k , u , v p k u v p E x y

[0092] Expression (13) then becomes O ^ k u v = ∑ E , x , y p E x y k , u , v p k u v p E x y G E x y

[0093] The probabilities p ( E, x, y | k, u, v ) And p ( E, x, y ) are established from the direct model (projection), corresponding to (12).

[0094] Expression (15) then becomes: O ^ k u v = ∑ E , x , y S k E M u − x , v − y O k u v ∑ k , u , v H x y ∑ k S k E ∑ u , v M u − x , v − y O k u v G E x y O ^ k u v = O u v ∑ E , x , y S k E M u − x , v − y O k u v ∑ k , u , v H x y ∑ k S k E ∑ u , v M u − x , v − y O k u v G E x y

[0095] The estimate Ô k ( u, v ) can be performed iteratively. Each iteration, of rank n, aims to estimate O ^ k n + 1 u v based on a previous estimate O ^ k n u v . When n = 0, the estimation is performed from an initial estimate O ^ k n = 0 u v The initial estimate could, for example, be a uniform distribution of the isotope. k on the object surface.

[0096] Based on an initial estimate O ^ k n u v each estimate O ^ k n + 1 u v is such that: O ^ k n + 1 u v = ∑ E , x , y S k E M u − x , v − y O ^ k n u v ∑ k , u , v H x y ∑ k S k E ∑ u , v M u − x , v − y O ^ k n u v G E x y O ^ k n + 1 u v = O ^ k n u v ∑ x , y M u − x , v − y ∑ E S k E H x y ∑ k S k E ∑ u , v M u − x , v − y O ^ k n u v G E x y

[0097] The iterations are performed until a stopping criterion is reached. The stopping criterion can be a predetermined number of iterations or an error criterion considered sufficiently small. The error criterion can be an error between the acquired image GE and the projection Ĝ E of O ^ k n u v on the detection surface, the projection Ĝ E being obtained from expression (12). This may be a quadratic error or a Küllback-Leibler divergence between the acquired image GE and the projection Ĝ E ,

[0098] In order to minimize the size of the mathematical quantities used, the gamma image GE ( x, y ) can be formed according to a list mode, whereby the image is formed by a sum of detected interactions. Each interaction is assigned a rank. l , which can be established chronologically. Each interaction of rank l is assimilated to a Dirac function δ at the coordinate ( xl , y ). For this reason, G E x y = ∑ l δ x l , y l , E l = E x l And y l designating the coordinates of the interaction of rank l ; E l is the energy detected for the interaction of rank l.

[0099] Expression (19) becomes: O ^ k n + 1 u v = O ^ k n u v ∑ x , y M u − x , v − y ∑ l S k E l H x l y l ∑ k S k E l ∑ u , v M u − x l , v − y l O ^ k n u v δ x l , y l x y E l O ^ k n + 1 u v = O ^ k n u v ∑ x , y M u − x , v − y ∑ l p kl n δ x l , y l x y E l Or p kl n is a weighting factor assigned to the interaction l in the formation of spatial distribution O ^ k n + 1 of the k isotope in the field of observation. p kl n = S k E l H x l y l ∑ k S k E l ∑ u , v M u − x l , v − y l O ^ k n u v Soit w kl n = S k E l ∑ u , v M u − x l , v − y l O ^ k n u v w kl n is a probability of detecting an interaction of rank l, energy E l ,corresponding to an isotope k, taking into account the spatial distribution O ^ k n .

[0100] The sum ∑ u , v M u − x l , v − y l O ^ k n u v is a convolution product of type M * O ^ k n This corresponds to the direct model of gamma image formation based on knowledge of the spatial distribution of the isotope. k in the field of view. This is an estimate of the gamma image G, at the point with coordinates ( xl , y ), taking into account the spatial distribution of O ^ k n the isotope k in the field of observation.

[0101] S ( k, E l ) corresponds to an energy component of the probability w kl n , while ∑ u , v M u − x l , v − y l O ^ k n u v corresponds to a spatial component of the probability w kl n The sum ∑ k w kl n = ∑ k S k E l ∑ u , v M u − x l , v − y l O ^ k n u v corresponds to the sum of probabilities w kl n across all isotopes k.

[0102] Given (23) and (24), p kl n = S k E l H x l y l ∑ k w kl

[0103] In expression (22), ∑ l p kl n δ x l y l E l can be considered a relative error ε k n x y , for the isotope k, whose rear projection gives an updated image U k n for the isotope k.

[0104] Expression (22) can be reduced to: O ^ k n + 1 u v = O ^ k n u v U k n u v with U k n u v = ∑ x , y M u − x , v − y ε k n x y

[0105] Equation (27) corresponds to a backprojection of the relative error term ε k n in the observation field Ω, by convolution with the response function M of the collimator.

[0106] Equation (26) is a spatial distribution update equation O ^ k n u v of the isotope k in the field of observation, at each iteration n.

[0107] The reconstruction process described above shows that the sensitivity matrix His taken into account to compensate for the non-uniformity of the camera's gamma sensitivity, resulting from the sub-pixelation process. The weight H ( x, y ) thus allows for weighting the consideration of interactions positioned at coordinates ( x, y Thus, depending on the value H ( x, y ), plus the value of H ( x, y The lower the weight, the lower the weight. p kl n assigned to an interaction positioned at ( x, y ) is important. This allows for an overweighting of interactions detected at coordinates where the value of H ( x, y ) is weak.

[0108] In the steps just described, steps 120 and 130 are performed based on the interactions detected during step 100. Thus, the object image is reconstructed O is carried out using interactions G ( x, y) detected having been used to form the sensitivity matrix H.

[0109] According to one variant, the sensitivity matrix H is updated periodically. Between updates, the gamma camera is exposed to different fields of view. During each acquisition period, the object image is reconstructed using a sensitivity matrix H previously determined. In other words, reconstruction step 130 can be implemented using weights H ( x, y ) previously established, from a gamma image different from the one on which the reconstruction step is based. However, it is preferable to use a calibration function H determined by exposing the gamma camera to radiation comparable, from an energy point of view, to the radiation to which the gamma camera is exposed during the acquisition of the gamma image being reconstructed.

[0110] A notable advantage of the invention is that the weights H ( x, y These parameters can be updated during each gamma image acquisition, allowing for detection uniformity to be taken into account by considering the acquisition conditions. This notably enables the consideration of effects induced by radiation sources located outside the field of view, but which can influence the acquired image. Thus, the sensitivity matrix resulting from step 120 can be used to compensate for non-uniform sensitivity resulting from the subpixel unit, but also to compensate for the influence of off-field radiation sources or other causes of sensitivity non-uniformity, such as a defect in the detector material.

[0111] The inventors implemented the process described in connection with steps 100 to 130. A gamma camera was used, comprising a 6 mm thick CdZnTe semiconductor detector with a 16x16 pixel array, each pixel measuring 2.5 mm on a side. Each pixel was sub-pixelated into 8x8 virtual pixels, resulting in a detection area segmented into 128x128 virtual pixels. The field of view consisted primarily of 60 < Co sources, with a significant contribution from scattered radiation. The irradiation level at the camera was 70 µSv / h.

[0112] THE Figures 7A , 7C And 7E are gamma images obtained respectively: without treatment; after normalization by weights H ( x, y ) determined from the gamma image shown on the Figure 7A , by implementing the first embodiment (substeps 121 to 123); after normalization by weights H ( x, y) determined from the gamma image shown on the Figure 7A , by implementing the second embodiment (substeps 125 to 126).

[0113] THE Figures 7B , 7D And 7F are object images obtained by performing a reconstruction, as described in step 130, respectively: without taking weight into account H ( x, y ): Step 130 is implemented considering H ( x, y ) = 1; taking into account weights as determined according to the first embodiment (substeps 121 to 123); taking into account weights as determined according to the second embodiment (substeps 125 to 126).

[0114] Without implementing the invention, the reconstructed image contains artifacts, indicated by arrows on the Figure 7B The actual radiating source is surrounded by dotted lines on the Figure 7B . THE Figures 7D And7F show that the implementation of the invention makes it possible to limit the presence of artifacts. The irradiating source is localized with a significantly improved signal-to-noise ratio. The invention allows for weighting of the amount of interactions detected by each virtual pixel upstream of the reconstruction. It is particularly advantageous when the collimator is of the coded mask type, for which the non-uniformity of the detection sensitivity can generate reconstruction artifacts, as shown in the Figure 7B .

[0115] A significant advantage of the invention is that the weights assigned to each virtual pixel can be determined by acquiring detection signals while the gamma camera is deployed in the field, without requiring laboratory conditions in which the irradiation of each virtual pixel is uniform. Since the spatial sensitivity function is likely to vary depending on the energy of the radiation to which the gamma camera is exposed, the ability to establish a spatial sensitivity function based on the actual field conditions is particularly advantageous. Indeed, these conditions are difficult to predict and reproduce in the laboratory, especially when there is a significant contribution from scattered radiation.

[0116] The weights assigned to each virtual pixel can be updated regularly using gamma images acquired in the field. Thus, the same spatial sensitivity function H can be used to reconstruct an object image from a gamma image acquired before or after the spatial sensitivity function was determined. In this case, it is preferable that the radiation to which the gamma camera is exposed during the acquisition of the gamma image be comparable, from an energy point of view, to the radiation detected during the establishment of the spatial sensitivity function used in the reconstruction.

[0117] Although described in relation to a sub-pixelation unit, the embodiment described in relation to steps 121 to 123 can be implemented on a gamma camera with small pixels, for example, less than mm². Subdivision into virtual pixels is not necessary. Thus, according to a variant covered by the attached claims, the sensitivity matrix HThis can be constructed by considering the interactions assigned to each pixel and grouping these pixels into rows and columns. Equations (1) and (2) are performed by considering the pixels in the same column and row, respectively. This yields a sensitivity matrix defined for each pixel, with a weight assigned to each pixel. Using this sensitivity matrix allows for the consideration of non-uniform sensitivity of the gamma camera due to the presence of off-camera sources or local defects in the detector material.

Claims

1. Method for determining a spatial-sensitivity function of a gamma camera, the gamma camera (1) being configured to locate radiation sources (5) in a field of observation (Ω), the field of observation being liable to contain radiation sources (5), the gamma camera comprising: - a detector material (11); - pixels (12i), distributed over a detecting area (12) of the detector material, each pixel being configured to form a detection signal under the effect of detection of an interaction of an ionising photon in the detector material; - a memory (16), configured to store a quantity of interactions (G(x, y), GE(x, y)) detected in the course of an acquisition period and respectively assigned to each pixel; the method comprising the following steps: - a) acquiring detection signals with the pixels (12i) during one acquisition period, each detection signal being associated with one detected interaction; - b) depending on the detection signals, attributing a position (x, y) of each interaction detected, during the acquisition period, to one pixel (12i); - c) storing, in the memory, a number of detected interactions (G(x, y), GE(x, y)); assigned to each pixel; the method being characterized in that it comprises, following step c), - d) defining groups of pixels, each group containing a plurality of pixels - e) computing a value for each group (HX(y), HY(x), H'(xj, yj)), the value of each group being computed on the basis of a sum of the number of interactions positioned in each pixel belonging to said group, - f) on the basis of the value computed for each group, assigning a weight (H(x, y)) to each pixel, the weight assigned to each pixel being representative of a detection sensitivity of said pixel, all of the weights respectively assigned to each pixel forming the spatial-sensitivity function (H) of the gamma camera.

2. Method according to Claim 1, wherein: - the pixels are distributed in rows and columns over the detecting area; - each group of pixels contains pixels belonging to the same row or to the same column; - step e) comprises the following sub-steps: • e-i) forming a first vector (HX), containing first terms (HX(y)), each first term being associated with one column of pixels, each first term comprising a sum of the quantity of interactions positioned in each pixel of the column; • e-2) forming a second vector (HY), containing second terms (HY(x)), each second term being associated with one row of pixels, each second term comprising a sum of the quantity of interactions positioned in each pixel of the row; - step f) comprises the following sub-steps: • f-1) for each pixel, computing a product of multiplication of a first term (HX(y)), associated with the column of the pixel, by a second term (HY(x)), associated with the row of the pixel; • f-2) determining the weight (H(x, y)) assigned to the pixel depending on the product computed in sub-step f-1).

3. Method according to Claim 2, wherein sub-step f-2) comprises: - computing a mean value of the products respectively computed, in sub-step f-1), for a set of pixels; - for each pixel, normalising the product resulting from sub-step f-1) by the mean value.

4. Method according to any one of the preceding claims, comprising: - g) generating a sensitivity matrix (H), each point of the sensitivity matrix corresponding to one pixel of the gamma camera, the value (H(x, y)) of the sensitivity matrix at each point corresponding to the weight assigned to said pixel.

5. Method according to any one of the preceding claims, wherein the gamma camera comprises a processing unit (18) configured to process the interactions stored in the memory, the method comprising a step h) of reconstructing a spatial distribution (O, Ok) of the radiation sources (5), in the field of observation, on the basis of the interactions stored in step c) and of the weights (H(x, y)) assigned to each pixel in step f).

6. Method according to Claim 5, wherein the processing unit normalizes a number of interactions detected by each pixel by the weight assigned to said pixel.

7. Method according to Claim 6, wherein: - each pixel is connected to a spectrometry unit (15), the spectrometry unit being configured to classify each detection signal into one energy band (E) among a plurality of energy bands; - in step b), each interaction positioned in a pixel is associated with one detected energy band; - step h) comprises: • taking into account at least one isotope (k), the isotope emitting photons in at least one emission energy; • reconstructing a spatial distribution of radiation sources (Ok) comprising the isotope on the basis of the interactions positioned in step b).

8. Method according to Claim 7, wherein step h) comprises: - taking into account a spectral response function (S) of each pixel, the spectral response function representing a detection probability (S(k, E)), in various energy bands, of each photon emitted by the isotope (k) and detected in the pixel; - taking into account a spatial response function (M), the spatial response function representing a probability of detection, in various pixels, of a photon emitted from one point in the field of observation.

9. Method according to any one of the preceding claims, wherein the gamma camera comprises a collimator (10) defining the field of observation (Ω).

10. Gamma camera (1) intended to detect a presence of radiation sources (5) in a field of observation (Ω), the gamma camera comprising: - a detector material (11); - pixels (12i), distributed over a detecting area of the detector material, each pixel being configured to form a detection signal under the effect of detection of an interaction of an ionising photon in the detector material; - a memory (16), configured to store a quantity of interactions detected in the course of an acquisition period and respectively assigned to each pixel; - a processing unit (17), configured to process the interactions stored in the memory, the gamma camera being characterized in that the processing unit is configured to implement steps d) to f) of a method according to any one of claims 1 to 10 on the basis of the detected interactions.

11. Method for reconstructing a spatial distribution (O, Ok) of radiation sources in the field of observation of a gamma camera (1) according to Claim 10, the method comprising the following steps: - i) acquiring detection signals with the pixels (12i) of the gamma camera during an acquisition period, each detection signal being associated with one detected interaction; - ii) storing, in the memory, a number of detected interactions assigned to each pixel; the method being characterized in that it comprises: - iii) taking into account weights (H(x, y)) respectively assigned to each pixel, each weight being determined by implementing the method according to any one of claims 1 to 9, step a) being performed by exposing the gamma camera to a field of observation identical to or different from the field of observation to which the gamma camera is exposed in step i); - iv) reconstructing the spatial distribution of the radiation sources, in the field of observation, on the basis of the interactions stored in step iii) and of the weights assigned to each pixel in step iii).

12. Method according to Claim 11, wherein: - in step i), the field of observation is identical to the field of observation of the gamma camera in step a); - step ii) and step b) are one and the same; - step iii) and step c) are one and the same.

13. Method according to Claim 12, wherein: - in step i), the field of observation is different from the field of observation of the gamma camera in step a); - step iv) is implemented taking into account weights assigned to each pixel, the weights having been established in the course of an implementation of steps a) to f), prior or subsequent to step i).