Medical image processing device, x-ray ct device, and medical image processing method

By identifying and correcting partial volume errors in photon counting detectors using micropixel data, the method improves material discrimination and image accuracy in CT systems, overcoming PVE challenges.

JP2024156936AActive Publication Date: 2024-11-06CANON MEDICAL SYST CORP
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
JP2024131220
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2018-11-06
Filing Date
2024-08-07
Publication Date
2024-11-06
Estimated Expiration
2039-04-22

AI Technical Summary

Technical Problem

Photon counting detectors (PCDs) in CT systems face challenges with partial volume errors (PVE) due to nonlinear responses at high X-ray flux rates, leading to incomplete material discrimination in reconstructed images.

Method used

The method involves identifying macropixels with PVE by checking for non-convergence in material discrimination processing and correcting these errors using projection data from smaller micropixels instead of macropixels, employing a look-up table to calibrate and correct the material discrimination process.

Benefits of technology

This approach enhances the accuracy of material discrimination and image reconstruction by effectively addressing PVE, resulting in more precise material component differentiation in CT scans.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024156936000001_ABST
    Figure 2024156936000001_ABST
Patent Text Reader

Abstract

To correct a partial volume error appropriately.SOLUTION: A medical image processing device according to this embodiment includes an acquisition unit, a generation unit, a selection unit, and a correction unit. The acquisition unit acquires first projection data indicating intensity of a radiation ray detected in a plurality of detector elements. The generation unit generates second projection data on the basis of the first projection data corresponding to the plurality of detector elements included in each macro pixel for each macro pixel that groups two or more detector elements of the plurality of detector elements. The selection unit selects a macro pixel in which material discrimination processing based on the second projection data does not converge. The correction unit generates correction material discrimination projection data by executing material discrimination processing on the basis of the first projection data corresponding to part of the detector elements included in the selected macro pixel.SELECTED DRAWING: Figure 6
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a medical image processing apparatus, an X-ray CT apparatus, and a medical image processing method. [Background technology]

[0002] Computed tomography (CT) systems and methods are widely used, particularly for medical imaging and diagnosis. CT systems generally create projection images of one or more cross-sectional slices through a subject's body. A radiation source, such as an X-ray source, irradiates the body from one side. A collimator, generally adjacent to the X-ray source, limits the angular range of the X-ray beam, such that radiation striking the body is substantially confined to a planar region (i.e., an X-ray projection plane) that defines a cross-sectional slice of the body. At least one detector (and generally many more than one detector) on the opposite side of the body receives the radiation propagated through the body in the projection plane. Attenuation of the radiation passing through the body is measured by processing electrical signals received from the detector. In some implementations, a multi-slice detector configuration is used, providing a volumetric projection of the body rather than a planar projection.

[0003] Typically, the X-ray source is mounted on a gantry that rotates around the long axis of the body. A detector is similarly mounted on the gantry on the opposite side of the X-ray source. Cross-sectional images of the body are obtained by making projection attenuation measurements at a series of gantry rotation angles, sending the projection data / sinograms to a processor via slip rings configured between the gantry rotor and the stationary part, and then processing the projection data using a CT reconstruction algorithm (e.g., inverse Radon transform, filtered back projection, Feldkamp-based cone beam reconstruction, iterative reconstruction or other methods). For example, the reconstructed image may be a digital CT image that is a square matrix of elements (pixels), each of which represents a volume element (volumetric pixel or voxel) of the patient's body. In some CT systems, the combination of the body translation and the rotation of the gantry relative to the body is such that the X-ray source traverses a helical or spiral trajectory relative to the body. Multiple views are then used to reconstruct a CT image showing the internal structure of the slice or of multiple such slices.

[0004] Traditionally, energy-integrating detectors have been used to measure CT projection data. Photon-counting detectors (PCDs) have then presented a viable alternative to energy-integrating detectors. PCDs have many advantages, including their ability to perform spectral CT and the ability to divide the scan area into many smaller "pixels" of the detector for greater resolution. Semiconductor-based PCDs offer unique advantages for spectral CT, but also pose unique challenges. For example, due to pulse pile-up, PCDs can exhibit a nonlinear response to x-ray flux. Without correcting for the nonlinearity and spectral shifts in the detector response, images reconstructed from semiconductor-based PCDs may have worse image quality.

[0005] One advantage of PCD is that it can be used for spectral CT because PCD gives information about the change in x-ray attenuation depending on the energy of the x-ray. Spectral CT is desirable because different materials such as bone and water exhibit different spectral absorption signatures, allowing the spectrally resolved CT scan to be discriminated into material components. However, when the pixels of the x-ray detector are large enough that x-rays that fall within the same pixel of the x-ray detector path pass through different materials, this material discriminator process can produce partial volume errors (PVEs). Therefore, better ways to identify and correct PVEs are desired. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] US Patent Application Publication No. 2017 / 0023496 [Patent Document 2] U.S. Patent No. 9,687,207 [Patent Document 3] U.S. Patent No. 8,422,826 Summary of the Invention [Problem to be solved by the invention]

[0007] The problem to be solved by the present invention is to appropriately correct partial volume errors. [Means for solving the problem]

[0008] The medical image processing apparatus according to the present embodiment includes an acquisition unit, a generation unit, a selection unit, and a correction unit. The acquisition unit acquires first projection data representing the intensity of radiation detected in a plurality of detector elements. The generation unit generates second projection data for each macropixel formed by grouping two or more detector elements among the plurality of detector elements, based on the first projection data corresponding to the plurality of detector elements included in each macropixel. The selection unit selects a macropixel for which material decomposition processing based on the second projection data does not converge. The correction unit generates corrected material decomposition projection data by executing material decomposition processing based on the first projection data corresponding to some of the detector elements included in the selected macropixel. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of a partial volume effect according to an exemplary embodiment of the present disclosure. [Figure 2A] FIG. 2A is a schematic diagram of a perspective view of partial volume averaging voxels having two materials arranged in an orthogonal configuration relative to the beam direction according to an exemplary embodiment of the present disclosure. [Figure 2B] FIG. 2B is a schematic diagram of a perspective view of partial volume averaging voxels having a homogenous mixture of two materials according to an exemplary embodiment of the present disclosure. [Figure 2C] FIG. 2C is a schematic diagram of a perspective view of partial volume averaging voxels having two materials arranged in a parallel configuration relative to the beam direction according to an exemplary embodiment of the present disclosure. [Figure 3A] FIG. 3A is a schematic diagram of a side view of partial volume averaging voxels having two materials arranged in an orthogonal configuration relative to the beam direction according to an exemplary embodiment of the present disclosure. [Figure 3B] FIG. 3B is a schematic diagram of a side view of partial volume averaging voxels having a homogenous mixture of two materials according to an exemplary embodiment of the present disclosure. [Figure 3C]FIG. 3C is a schematic diagram of a side view of partial volume averaging voxels having two materials arranged in a parallel configuration relative to the beam direction according to an exemplary embodiment of the present disclosure. [Figure 4] FIG. 4 illustrates a flow chart for a method for identifying partial volumes during a scan according to an exemplary embodiment of the present disclosure. [Diagram 5] FIG. 5 is a diagram illustrating a method for selecting a micro-pixel according to an exemplary embodiment of the present disclosure. [Figure 6] FIG. 6 is a schematic diagram of an implementation of a computed tomography scanner according to an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] The embodiments described herein relate generally to material decomposition of photon-counting spectral computed tomography data, and more particularly to identifying partial volume errors (PVEs) when the material decomposition process fails to converge, and then correcting the PVEs by using projection data from micropixels rather than macropixels for material decomposition, where a macropixel is a virtual pixel that aggregates signals from multiple pixels in a detector array, (e.g., micropixels). The following description, together with the accompanying drawings, describes various aspects of the disclosed subject matter, and does not necessarily represent the only aspects. In some instances, the description includes specific details to provide an understanding of the disclosed subject matter. However, it is clear to those skilled in the art that the aspects may be practiced without these specific details. In some cases, well-known structures and components may be shown in block diagram form to avoid obscuring the concepts of the disclosed subject matter.

[0011] References throughout this specification to "one embodiment" or "an embodiment" mean that a particular feature, structure, property, operation, or function described with respect to an embodiment is included in at least one embodiment of the disclosed subject matter. Thus, any occurrence of the phrase "in one embodiment" or "in one embodiment" herein does not necessarily refer to the same embodiment. Furthermore, particular features, structures, properties, operations, or functions may be combined in any suitable manner in one or more embodiments. Furthermore, the embodiments of the disclosed subject matter can and shall cover modifications and variations of the described embodiments.

[0012] Additionally, terms such as "top," "bottom," "front," "rear," "side," "interior," "exterior," and the like, that may be used herein, merely describe points of reference and do not necessarily limit aspects of the disclosed subject matter to any particular orientation or configuration. Additionally, terms such as "first," "second," "third," and the like, merely identify one of several parts, components, points of reference, operations, and / or functions described herein and similarly do not necessarily limit aspects of the disclosed subject matter to any particular configuration or orientation.

[0013] As explained above, photon counting detectors (PCDs) offer several advantages to CT, including the advantage that material decomposition can be performed using spectrally resolved projection data. However, several obstacles may limit the realization of the full potential of PCDs. As explained above, at high x-ray flux rates, pulse pile-up produces a nonlinear response for PCDs. To avoid pile-up, the cross-sectional area of ​​the PCD can be reduced to suppress the occurrence of pulse pile-up even at higher x-ray fluxes. For example, a single large PCD can be subdivided into four smaller PCDs, each covering 1 / 4 of the area of ​​the large PCD, with the flux rate per smaller PCD being 1 / 4 that of the large PCD. In practice, charge sharing and other effects may fundamentally limit how small PCDs can be performed.

[0014] For example, subdividing the PCD into smaller PCDs increases the total number of detector elements / pixels in the CT scanner and can create a communication bottleneck between the rotor and the stationary part of the CT gantry based on how quickly the projection data can be transmitted from the rotating annular structure in which the x-ray detectors are configured across the slip rings to the stationary gantry structure housing one or more computer processors performing image reconstruction, etc. That is, collecting data at a higher resolution results in more data to be transmitted across the slip rings in a given amount of time, exceeding the limited communication bandwidth available through the slip rings. The slip rings are a non-limiting example of a communication device between the rotor and the stationary part, and in some implementations, fiber optic rotary joints or other communication devices may be used in place of the slip rings, as will be appreciated by those skilled in the art.

[0015] To overcome this communication bottleneck, counts from small PCDs (eg, micropixels) may be aggregated as counts of a virtual large PCD (eg, macropixel) to reduce the amount of data transmitted.

[0016] While this scheme helps address the pile-up problem, summing the micropixel counts into the macropixel counts can leave the projection data susceptible to partial volume error (PVE). The method described herein addresses PVE by identifying the macropixels where PVE occurs and then using the micropixel counts rather than the macropixel counts to correct for the PVE.

[0017] Referring now to the drawings, in which like reference numbers designate the same or corresponding parts throughout the several views, FIG. 1 illustrates a cross-sectional view of an object and two parallel x-ray trajectories passing through the object on their way to respective micropixels of a detector array. The micropixels corresponding to the two x-ray trajectories are summed / averaged to generate a signal for a macropixel. In the two x-ray trajectories illustrated, one x-ray trajectory passes through only soft tissue, while the other x-ray trajectory passes through bone (e.g., ribs) in addition to passing through soft tissue. Thus, when two micropixel signals corresponding to these two x-ray trajectories are combined to generate a macropixel signal, material decomposition of the resulting macropixel signal may be incomplete due to partial volume effects, ultimately resulting in a PVE. A PVE is identified based on the material decomposition process failing to converge. The material decomposition process may then be corrected by performing the material decomposition process on the micropixel signal rather than the macropixel signal, as described in more detail below. The lack of convergence of the material decomposition process is due to the measured count N of the macropixels. ~ (Hereinafter, the superscript tilde indicates that the tilde is placed directly above the character) and the material projection length L → (Hereafter, the superscript arrow indicates a vector) based on the calibrated / theoretical count N → (L → ) can be shown by a non-negligible difference between

[0018] Thus, PVEs can occur when different materials occupy different regions within the cross-sectional area of ​​an X-ray trajectory that falls within a recorded pixel (e.g., a macropixel) of a CT scan. The methods described herein provide the advantageous effect of identifying and correcting these PVEs, thereby producing more accurate material decomposition and reconstructed images.

[0019] Materials with atoms with different atomic numbers Z have different spectral profiles, allowing material decomposition of spectral CT data. In part from the k-edge effect, the spectral shape of X-ray attenuation in biological materials is determined by two physical processes: photoelectric attenuation and Compton scattering. Thus, the attenuation coefficient as a function of energy can be approximated by the decomposition of

[0020]

number

[0021] Here, μ PE (E,x,y) is the photoelectric decay, μ C (E,x,y) is the Compton attenuation. This attenuation coefficient can instead be reconstructed into a discrimination between material 1 (e.g., a high-Z material such as bone) and material 2 (e.g., a low-Z material such as water), yielding

[0022]

number

[0023] where c 1,2 (x,y) is a spatial function describing the concentrations of material 1 and material 2 located at position (x,y). When the material composition structure of a voxel is approximately spatially uniform, the intensity along a given x-ray trajectory can be given by:

[0024]

number

[0025] where L1 = ∫dlc1(x,y) is the projected length of the first material component, L2 = ∫dlc2(x,y) is the projected length of the second material component, and the integrals are line integrals along the line l(x,y) of the X-ray trajectory.

[0026] In spectral CT utilizing photon-counting detectors (PCDs), image reconstruction is preceded by a pre-reconstruction step that involves correcting for detector response and material decomposition. In particular, partial volume effects refer to the averaging / aggregation of adjacent pixel values ​​corresponding to different material components that can result in anomalous spectral signatures.

[0027] In general, the intensity I of a beam, e.g., an X-ray beam, propagated through an absorbing material is expressed as:

[0028]

number

[0029] where I0 is the initial beam intensity, L is the path length, and μ is the attenuation coefficient. The beam can be monochromatic or polychromatic with respect to energy. A polychromatic beam contains multiple X-ray energies and is used to extract the energy-dependent attenuation coefficient. The spectral term S(E) is the incident spectrum of energy E on the detector. This spectral term for two materials is given by:

[0030]

number

[0031] where S0 is the radiation energy spectrum from the radiation source, μ1 and μ2 are the attenuation coefficients of the reference materials for material decomposition, and L1 and L2 are the respective transmission path lengths. When a PCD is used to detect X-rays, the energy spectrum is expressed as the number of counts N in the i-th energy bin. i The energy bins are divided into a set of n energy bins with the range {E i ,E i+1}.

[0032]

number

[0033] where C is a calibration constant and the mth detector element is {(x m ,y m ),(x m+1 ,y m+1 )}.

[0034] 2A, 2B, 3A, and 3B provide non-limiting, simplified examples of scenarios in which a PVE does not occur, while FIGS. 2C and 3C provide non-limiting, simplified examples of scenarios in which a PVE does occur.

[0035] FIG. 2A shows a perspective view of a partial volume averaging voxel 100 having two materials arranged in an orthogonal configuration according to an exemplary embodiment of the present disclosure. Orthogonal configuration refers to a positional relationship in which multiple materials are arranged along the transmission path of X-rays. The orientation of the first material 105 and the second material 110 in the voxel 100 can be simplified as a sandwich structure, where the interface between the first material 105 and the second material 110 is substantially orthogonal to the direction of the beam 115 of photons that strikes the detector 120. The beam 115 is represented as a single ray that traverses the voxel 100, or as multiple small rays that are averaged together to comprise a single ray. The detector 120 can be configured to receive multiple photons from the beam 115 and convert the energy of the impinging photons into an electrical signal. The detector 120 is a photon counting detector (PCD), where the generated electrical signal is converted to photon counts and analyzed by a processing circuit. In the orthogonal direction, a beam 115 having intensity I0 when entering a voxel 100 passes through a first material 105 with attenuation μ1, then passes through a second material 110 with attenuation μ2 having intensity I when exiting the voxel 100. If the beam 115 passes through both materials with the same thickness as shown, the path length is half the length of the voxel 100 in each material. Thus, L / 2=L1=L2, and the intensity I of the monochromatic beam 115 may be expressed as:

[0036]

number

[0037] A similar formula based on the above is given for the number of propagated photons, N, where N is the incident number of photons, yielding:

[0038]

number

[0039] 2B shows a perspective view of a partial volume averaging voxel 100 having a homogenous mixture of two materials, according to an exemplary embodiment of the present disclosure. Here, the mixing of a first material 105 and a second material 110 results in the same propagation as in FIG. 2A, since the beam 115 effectively travels through a path length L / 2 that matches the thickness of each material. Thus, the order of materials through which the beam travels does not matter due to the exponential form of attenuation, as long as it traverses an effective path length of L / 2 through the voxel 100, which has length L for both materials. Thus, the number of propagated photons N can again be expressed as explained above.

[0040] 2C shows a perspective view of a partial volume averaging voxel 100 having two materials arranged in a parallel configuration according to an exemplary embodiment of the present disclosure. A parallel configuration refers to a positional relationship in which the materials are arranged along substantially orthogonal axes of the transmission path of the x-rays. As the beam 115 rotates around the object during the scan, for example, an additional 90° rotation, the beam 115 can move to a position where the interface plane between the first material 105 and the second material 110 is oriented substantially parallel to the beam 115. The beam 115 passing through the voxel 100 can include a narrower ray that passes only through the first material 105 and not the second material 110, or vice versa. 2A and 2B as being composed of an average of many rays, the beam 115 in FIG. 2C can be split into a first ray 115a that passes through the first material 105 and a second ray 115b that passes through the second material 110.

[0041] Here, the first light beam 115a has a different number of propagated photons compared to the second light beam 115b when measured by the detector 120. The detector 120 may include a grid of at least one macropixel 125, where each of the at least one macropixel 125 may be divided into a plurality of micropixels 130. That is, the micropixels 130 are a portion of detector elements (e.g., a detector element group formed from two or more detector elements) included in the macropixel 125. The micropixels 130 may be configured to detect the narrower first and second light beams 115a, 115b and convert the energy imparted from the photons through each light beam passing through each material into an electrical signal at each micropixel 130 to register a separate photon count. Because the first material 105 and the second material 110 have different attenuation coefficients, the propagated number of photons (for a polychromatic beam 115) is given by:

[0042]

number

[0043] 2A and 2B. Therefore, an attempt to discriminate the spectral counts into material components will in this case result in partial volume error (PVE). However, this PVE is mitigated if the measurement volume is subdivided into smaller volumes corresponding to micropixels 130, and the counts for the micropixels 130 are utilized for material discrimination instead of the counts for the macropixels 125.

[0044] 3A shows a cross-sectional view of partial volume averaging of a voxel 100 having two materials arranged in an orthogonal configuration, according to an exemplary embodiment of the present disclosure. When employing layer stacking, as described in FIG. 2A, a first material 105 can form an interface with a second material 110. In a simplified diagram, this interface can be planar and substantially perpendicular to the direction of the beam 115. In this configuration, the partial volume effect does not contribute to an error in the reading of N, since the beam 115 passes through both materials 105, 110.

[0045] 3B illustrates a cross-sectional view of partial volume averaging of a voxel 100 having two homogeneously mixed materials, according to an exemplary embodiment of the present disclosure. As described in FIG. 2B, the first material 105 and the second material 110 can be a homogeneous mixture such that there is no boundary between the materials. However, because the mixture is homogeneous, the beam 115 still passes through the voxel 100 with the same attenuation as it did in FIG. 2A and FIG. 2B, and no partial volume effect occurs.

[0046] FIG. 3C shows a cross-sectional view of partial volume averaging of a voxel 100 having two materials configured in a parallel configuration, according to an exemplary embodiment of the present disclosure. As described in FIG. 2C, the first material 105 can form an interface with the second material 110, where the interface can be planar and substantially parallel to the direction of the beam 115. This orientation of the first material 105 and the second material 110 can cause measurement errors due to partial volume effects when the beam 115 is composed of many smaller rays 115a, 115b that are summed and averaged to generate an effective transmitted beam passing through the voxel 100. Therefore, to reduce measurement errors, individual measurements of multiple micropixels 130 (determined via averaging the values ​​of multiple micropixels 130 for each macropixel 125) can be used for image reconstruction instead of the value of at least one macropixel 125.

[0047] FIG. 4 shows a flow chart of a non-limiting example of a method for identifying PVE and correcting it by switching from macropixel measurements to micropixel measurements.

[0048] In step S300, image acquisition begins. In step S301, the detector 120 generates projection data representing the intensity of radiation detected at the multiple detector elements. The projection data here is combined for each macropixel, which is a group of two or more detector elements among the multiple detector elements, to generate projection data corresponding to the macropixel. Specifically, for the m-th micropixel of the micropixels 130 and for the i-th energy bin, a count N i,m (micro) These counts are then calculated as N i are summed to produce a count of , which is given by:

[0049]

number

[0050] In step S301, the measured count N ~ ={N1 (macro) , N1 (macro) , ,N n (macro)} is the coefficient N for material decomposition → (L → ), where the material decomposition coefficient N → is the projection length L → That is, for each energy bin, the projection data corresponding to each macropixel is combined by summing the counts of the projection data corresponding to each macropixel to generate a first count. For example, this comparison may be performed using an objective function φ (also called a cost function or error measure) to compare the measured counts N ~ and the material decomposition count N → The measured count N i ~ and the modeled count N m Several different definitions of the cost function φ(L1,L2) may be used to represent the difference between L1,L2 and N'. In one implementation, the cost function is m and the material decomposition count N → the least squares difference between

[0051]

number

[0052] In one implementation, the cost function is calculated by multiplying the measured count N' by m and the modeled count N m the weighted least squares fit of the difference between

[0053]

number

[0054] Here, σi is the measured count N i ~ is the standard deviation of

[0055] Alternatively, the comparison can be any known distance measure, for example the Euclidean distance, or an error measure between the measured and discriminated counts.

[0056] Discriminated count N → (L → ) is obtained, for example, by calculation using the following formula:

[0057]

number

[0058] where C is a calibration constant. In some implementations, the above equation can be modified to include the detector response. As discussed in U.S. patent application Ser. No. 13 / 866,965, the entirety of which is incorporated herein by reference, the response function of the radiation detector can be calibrated to give improved results. In one implementation, the detector model for the counts of each given radiation detector is:

[0059]

number

[0060] Here, the integral time T, the linear response function R0, the nonlinear response function R1, and the dead time τ are each known for each radiation detector and energy component as a result of a calibration performed prior to projection measurement on the object OBJ.

[0061] Furthermore, the discriminated count N → (L →) is obtained by performing various calibrations to measure and store the counts for various known lengths (path lengths) of the material phantom, and interpolating between these counts to determine length counts between the lengths of the material phantom. The look-up table contains counts grouped according to respective values ​​of projected length of the material components.

[0062] Furthermore, the discriminated count N → (L → ) may be provided and stored in a look-up table (LUT). Furthermore, the LUT may be → It is indexed by N → (L → )=LUT(L → ) In other words, the LUT contains count values ​​N → (L → ) and the projection length L → are stored in correspondence with each other.

[0063] The objective function φ is the projection length L → is used to adjust ,i.e., to construct the optimization problem by

[0064]

number

[0065] The minimum of the objective function φ is then returned to represent an estimate that quantifies the PVE. ~ Count N discriminated from → (L → ) is minimized. → (L → ) is selected. In other words, for each macropixel, material decomposition is performed on the projection data corresponding to the macropixel to generate material-decomposed projection data (also called material-decomposed projection data). When PVE is small, N(L → ) is the value of the measured count N~ and the value of the objective function φ approaches a minimum, e.g., 0. In other words, the material decomposition process converges. However, N(L → ) and the measured count N ~ When a poor match is obtained between x, y, and y, the poor match may be due to PVE, and the value of the objective function φ does not approach 0. In other words, the material decomposition process does not converge. Therefore, when the material decomposition process does not converge, one method is to reduce the pixel size from macropixels to micropixels, and repeat the material decomposition process again using micropixels instead.

[0066] In some implementations, a predetermined predicted value N in the LUT → (L → ) may be a value of N calculated based on the expected reference material path length of the material, alone or in combination with one or more additional materials (e.g., the material may be bone and soft tissue composed primarily of water).

[0067] N → (L → The values ​​of ) may be generated via a simulation program or empirically (e.g., using calibration measurements of a known material phantom). The values ​​of macropixel 125 may be used to estimate the path length of a constituent reference material in the scan, e.g., a combination of first material 105 and second material 110.

[0068] In step S303, the macropixel count N ~ and the calculated / calibrated count N(L →) (e.g., the minimum value of the objective function φ) is compared to a predefined threshold. When the minimum value of the objective function φ is less than the predefined threshold, it is determined that the material decomposition process has converged, it is determined that PVE correction should not be performed, and the method proceeds from step S303 to step S305. When the minimum value of the objective function φ is greater than the predefined threshold, it is determined that the material decomposition process has not converged, it is determined that PVE correction should be performed, and the method proceeds from step S303 to step S307.

[0069] In some implementations, the LUTs are used to calculate the N → Set of reference material path lengths for L = {L1 → ,L1 → ,…, L k → ), where N → =LUT(L → ) Argument L that minimizes discrepancies → A search is performed on the LUT to find:

[0070]

number

[0071] Subsequent macropixel measurements N ~ A predefined threshold ε is applied to the minimized difference between and the value in the LUT. Thus, PVE correction is performed when the following inequality is not satisfied:

[0072]

number

[0073] To reach step S307, N for macropixel 125 ~ and N →It has been determined that the measured difference between is outside a predefined deviation range (e.g., greater than a predefined threshold ε) from the value in the LUT. Thus, a partial volume error (PVE) has been identified and micropixel measurements are used instead of macropixel measurements to provide better spatial resolution of different material components. That is, a material decomposition process is performed on some detector elements (i.e., micropixels) contained in macropixels for which the minimum value of the objective function φ is above a predefined threshold. This corrects the partial volume error and produces corrected material decomposition projection data (also referred to as corrected material decomposition projection data). The corrected material decomposition data includes material decomposition projection data for micropixels included in the macropixels selected for PVE correction and includes material decomposition projection data for macropixels not selected for PVE correction.

[0074] The measurements of the micropixels 130 are calculated using the same method as used for the macropixels 125 to obtain the material components (e.g., the projected length L → For example, the measured counts can be differentiated into N ~ =4×{N1 (macro) , N2 (macro) , ,N n (macro)} where the factor 4 accounts for the fact that micropixels 130 are four times smaller than macropixels 125. Therefore, to use a LUT based on / calibrated for macropixels 125, the factor must be scaled up by the number of micropixels 130 per macropixel 125. A material decomposition process is then performed for each of the macropixels 125, for example, by solving the following optimization problem:

[0075]

number

[0076] Other methods of performing material decomposition processing on the micropixel measurements may be used without departing from the spirit and scope of the present invention.

[0077] The threshold ε may be a predetermined value greater or less than the value in the LUT, for example, ±1%, ±10%, ±20%, ±30%, or ±50%.

[0078] The LUT interpolation may be based on the arguments of the LUT, e.g., L1 and L2. Alternatively, the arguments of the LUT may be p and L, where L=L1+L2, L1=p*L, and L2=(1-p)*L. The interpolation may be performed between the L values. For example, to determine values ​​of L1=2.5 and L2=1.5, no values ​​in the LUT may be given because the lookup table is discretized at integer values ​​of the projection length (i.e., L1={0,1,2,...,N} and L2={0,1,2,...,N}). Thus, linear interpolation may be performed by determining the average LUT value between discretized values ​​adjacent to the L1 and L2 values, e.g., by the following equation:

[0079]

number

[0080] In step S305, a material component image is reconstructed from the material decomposition of the projection images.

[0081] In some implementations, the reconstruction can use different spatial grid sizes depending on the spatial distribution in which macropixels are selected for material decomposition to be performed using the micropixel measurements. That is, in areas where micropixel measurements are used for material decomposition, the material decomposition projection data is resolved at micropixel resolution. In areas where macropixel measurements are used for material decomposition, the material decomposition projection data is resolved at macropixel resolution. For example, a material component image is reconstructed using material decomposition projection data for micropixels included in the macropixels selected for PVE correction and using material decomposition projection data for macropixels not selected for PVE correction.

[0082] In some implementations, the reconstruction may be performed using a uniform spatial grid size that corresponds to the macropixel resolution. For example, in regions where micropixel measurements were used for material decomposition, the projection lengths of the micropixels may be averaged to generate average projection lengths for the macropixels, and these average projection lengths for the macropixels may be used for image reconstruction.

[0083] In some implementations, a communication bottleneck may result in too little data bandwidth to send all of the micropixel data across the slip ring to the fixed part of the CT gantry. In this case, a comparison / difference between the measured count of a given macropixel and the closest count in the LUT may be used as a quick check whether the macropixel count represents a PVE. For those macropixels where the difference between the measured count and the closest count in the LUT exceeds a threshold, the macropixel may be flagged and the micropixel count for that macropixel may be included in the data communicated across the slip ring to be processed later during the material decomposition process. When only a small subset of the macropixels exceed the PVE threshold ε, the increase in the amount of data communicated across the slip ring may be negligible, especially compared to the increase if all of the micropixel counts were communicated across the slip ring.

[0084] In some implementations, values ​​from macropixels less than the threshold ε and values ​​from micropixels corresponding to macropixel measurements greater than the threshold ε are then utilized to reconstruct the scan image in step S305, i.e., macropixels not affected by the PVE are processed during reconstruction at a coarser macropixel resolution, and macropixels affected by the PVE are processed at a finer micropixel resolution.

[0085] In some implementations, the selection of the value of the predefined threshold ε may be empirically determined. For example, adjusting the threshold ε may change the percentage of macropixels 125 identified as potentially having PVE that should be corrected using micropixel counting. As the predefined threshold ε is made smaller, there is a point of diminishing returns below which only small gains in image quality can be achieved. Furthermore, the selection of the value of the predefined threshold ε may be guided by the maximum communication bandwidth between the rotating and fixed parts of the CT scanner, which is used for calculations during image reconstruction, thus adjusting the transmission bandwidth requirements and processing time.

[0086] In some implementations, the material decomposition process includes, in step S303, determining whether a predefined threshold ε is equal to or smaller than the projection length-based count N → and the measured count N ~ This is then performed until a first convergence criterion is met to determine whether the difference / discrepancy between the projection length L from this initial discrimination exceeds the → may be stored and used later in step S305 or step S307 for a second material decomposition process that may be performed up to a second convergence criterion. For example, in the second material decomposition process, the results from the LUT are → may be interpolated to more accurately determine F. Additionally, any known method of material decomposition may be used for the first and second material decomposition processes.

[0087] In some implementations, the material decomposition process is performed only once per macropixel in step S301, and when the material decomposition process is performed on micropixels, the material decomposition process is performed only once on the macropixels in step S307.

[0088] In step S305, a material component image is reconstructed from the material component projection data generated by discriminating the spectral projection data for the macropixels and for the selected micropixels into material components. Any known method of image reconstruction may be used. For example, the image reconstruction process may be performed using any of the following: a filtered backprojection method, an iterative image reconstruction method (e.g., using a total variation minimization regularization term), a Fourier-based reconstruction method, or a probabilistic image reconstruction method.

[0089] Next, an example of a method for selecting micro-pixels will be described with reference to FIG. When using micropixel measurements in step S307, ie, selecting micropixels from macropixels, micropixels may be selected in stages from the macropixels. As shown in FIG. 5, a macropixel 400 is assumed to be composed of a total of nine pixels, three pixels vertically and three pixels horizontally. The macropixel measurement value N ~ and the value in the LUT exceeds a threshold ε, a micropixel 420 formed of four pixels, two pixels vertically and two pixels horizontally, is selected as a micropixel smaller in size than the macropixel 400. In the macropixel 400 shown in Fig. 5, a different micropixel 420 can be selected by shifting the macropixel 400 vertically and horizontally by one pixel. That is, in the example shown in Fig. 5, there are four micropixels 420, so at the stage of selecting the micropixel 420, an appropriate micropixel can be selected from the four micropixels 420.

[0090] Subsequently, if micropixel 420 is selected such that the minimized difference between the micropixel measurement value corresponding to micropixel 420 and the value in the LUT exceeds a threshold, then micropixel 440, which is formed by one pixel, may be selected as the smaller micropixel. In this way, by selecting micropixels 420 or micropixels 440 so that the pixel size is gradually reduced, material decomposition processing and the like can be performed using micropixels 420, which have a pixel size smaller than macropixels 400 but larger than the smallest pixel unit, and the image quality of the subsequent reconstructed image can be improved compared to processing using the smallest pixel unit.

[0091] The above description illustrates the methods described herein using a non-limiting example of a PCD. However, without departing from the spirit of the methods described herein, these methods may also be implemented using other types of X-ray detectors, including, for example, energy-integrating detectors, in which signals from each smaller detector (e.g., micropixels) are summed to generate a signal for a virtual larger detector (e.g., macropixel).

[0092] FIG. 6 shows a schematic diagram of an implementation of a CT scanner according to an exemplary embodiment of the present disclosure. Referring to FIG. 6, an X-ray imaging gantry 500 is shown in a side view, which further includes an X-ray tube 501, an annular frame 502, and a multi-row or two-dimensional array type X-ray detector 503. The X-ray tube 501, the X-ray detector 503, and the DAS 504 are mounted diametrically across the object OBJ on the annular frame 502, which is rotatably supported about a rotation axis RA (or axis of rotation). A rotation unit 507 rotates the annular frame 502 at a high speed, for example, 0.4 seconds / revolution, while the object OBJ is moved along the axis RA toward or away from the illustrated page.

[0093] X-ray CT devices include various types of devices, such as rotating / rotating devices in which the X-ray tube and the X-ray detector rotate together around the object to be inspected, and fixed / rotating devices in which many detector elements are arranged in the form of a ring or a plane and only the X-ray tube rotates around the object to be inspected. This disclosure applies to either type. The rotating / rotating type will be used as an example for clarity.

[0094] The multi-slice X-ray CT apparatus further includes a high voltage generator 509 that generates a tube voltage that is applied to the X-ray tube 501 through a slip ring 508 so that the X-ray tube 501 generates X-rays. The X-rays are emitted into an object OBJ whose cross-sectional area is represented by a circle. For example, the X-ray tube 501 has an average X-ray energy during a first scan that is smaller than an average X-ray energy during a second scan. Thus, two or more scans corresponding to different X-ray energies may be acquired. An X-ray detector 503 is located on the opposite side of the X-ray tube 501 across the object OBJ to detect the emitted X-rays propagated through the object OBJ. The X-ray detector 503 further includes individual detector elements or units.

[0095] The CT apparatus further includes other devices for processing the detection signal from the X-ray detector 503. The data acquisition circuit or data acquisition system (DAS) 504 converts the signal output from the X-ray detector 503 of each channel into a voltage signal, amplifies the signal, and further converts the signal into a digital signal. The X-ray detector 503 and the DAS 504 are configured to handle a predetermined total number of projections per rotation (TPPR). The DAS 504 may send values ​​of the projection data for each macropixel above a threshold. The DAS 504 may also send values ​​of the projection data for the macropixels where the material decomposition process converges, such as to the pre-processing device 506 or the reconstruction device 514.

[0096] The DAS 504 may also transmit values ​​of the projection data for all of the macro-pixels, such as to a pre-processing device 506 or a reconstruction device 514 . In this case, the reconstruction device 514 selects the selected macro-pixels for which the material decomposition process has not converged. The reconstruction device 514 sends a transmission request to the DAS 504 requesting transmission of values ​​of projection data corresponding to the selected macro-pixels. The reconstruction device 514 receives the transmission of values ​​of projection data corresponding to the selected macro-pixels from the DAS 504. The reconstruction device 514 performs the PVE correction described above on the selected macro-pixels to generate corrected material decomposition projection data. The reconstruction device 514 reconstructs a material component image from the corrected material decomposition projection data.

[0097] The data described above is sent through a non-contact data transmitter 505 to a pre-processing device 506, which is stored in a console outside the X-ray gantry 500. The pre-processing device 506 performs some corrections, such as sensitivity correction, on the raw data. A memory 512 stores the acquired data, also called projection data, at a stage just before the reconstruction process. The memory 512 is connected to a system controller 510 through a data / control bus 511, along with a reconstruction device 514, an input device 515, and a display 516. The system controller 510 controls a current regulator 513, which limits the current to a level sufficient to drive the CT system.

[0098] The detectors are rotated and / or fixed relative to the patient among various generations of CT scanner systems. In one implementation, the CT system described above may be an example of a combination of third and fourth generation geometry systems. In a third generation system, the x-ray tube 501 and x-ray detector 503 are diametrically mounted on an annular frame 502 and rotate around the object OBJ as the annular frame 502 rotates around the rotation axis RA. In a fourth generation geometry system, the detectors are fixedly positioned around the patient and the x-ray tube rotates around the patient. In an alternative embodiment, the radiography gantry 500 has multiple detectors configured on an annular frame 502 supported by a C-arm and a stand.

[0099] The memory 512 can store measurements representative of the intensity of X-rays at the X-ray detector unit 503. Furthermore, the memory 512 can store, for example, a dedicated program for performing various steps of a method for training a neural network and reducing imaging artifacts, for example, for identifying PVEs and performing PVE correction as described in FIG.

[0100] The reconstruction device 514 may have the acquisition, generation, selection, correction and reconstruction functions according to the present embodiment, and may perform pre-reconstruction image processing such as volume rendering and image differencing, if necessary.

[0101] Pre-reconstruction processing of the projection data performed by pre-processing device 506 may include, for example, detector calibration, correcting for detector non-linearity and polar effects.

[0102] Post-reconstruction processing performed by the reconstruction device 514 may include image filtering and smoothing, volume rendering, and image subtraction, as necessary. The image reconstruction process may implement various CT image reconstruction methods. The reconstruction device 514 may use memory to store, for example, projection data, reconstructed images, calibration data and parameters, and computer programs.

[0103] The reconfiguration device 514 may include a CPU (processing circuitry) that may be implemented as discrete logic gates, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other complex programmable logic device (CPLD). The FPGA or CPLD implementation may be coded in VHDL, Verilog, or any other hardware description language, and the code may be stored in electronic memory, either directly within the FPGA or CPLD or as a separate electronic memory. Furthermore, the memory 512 may be non-volatile, such as a ROM, EPROM, EEPROM, or flash memory. The memory 512 may also be volatile, such as a static or dynamic RAM, and a processor, such as a microcontroller or microprocessor, may be provided to manage the electronic memory and the interaction between the FPGA or CPLD and the memory.

[0104] Alternatively, the CPU in the reconstruction device 514 may execute a computer program including a set of computer readable instructions to perform the functions described herein, the program being stored in any of the non-transitory electronic memories described above and / or hard disk drives, CDs, DVDs, flash drives or any other known storage media. Furthermore, the computer readable instructions may be provided as a utility application, background daemon, or operating system component, or combinations thereof, that executes in conjunction with a processor, such as a Xenon processor from Intel®, USA, or an Opteron processor from AMD, USA, and an operating system, such as Microsoft VISTA®, UNIX®, Solaris, LINUX®, Apple®, MAC-OS, and other operating systems known to those skilled in the art. Furthermore, the CPU may be implemented as multiple processors that work cooperatively in parallel to execute instructions.

[0105] In one implementation, the reconstructed image may be displayed on a display 516. The display 516 may be an LCD display, a CRT display, a plasma display, an OLED, an LED, or any other display known in the art.

[0106] The memory 512 may be a hard disk drive, a CD-ROM drive, a DVD drive, a flash drive, RAM, ROM, or any other electronic storage known in the art.

[0107] PCDs can use direct conversion x-ray detectors based on semiconductors such as cadmium telluride (CdTe), cadmium zinc telluride (CZT), silicon (Si), mercury iodide (HgI2), and gallium arsenide (GaAs). Semiconductor-based direct conversion x-ray detectors generally have a much faster time response than indirect conversion x-ray detectors such as scintillator detectors. The fast time response of direct conversion x-ray detectors allows them to resolve individual x-ray detection events. However, at high x-ray fluxes typical of clinical x-ray applications, some pile-up of detection events will occur. The energy of the detected x-rays is proportional to the signal generated by the direct conversion x-ray detector, and the detection events are organized into energy bins that result in spectrally resolved x-ray data for spectral CT.

[0108] According to at least one embodiment described above, partial volume errors (PVEs) are identified as occurring in macropixels where the material decomposition process does not converge. The PVEs are then corrected by using projection data from micropixels rather than macropixels for material decomposition. This allows the partial volume errors to be appropriately corrected, resulting in more accurate material decomposition and image reconstruction.

[0109] Although several implementations have been described, these implementations are presented by way of example only and are not intended to limit the teachings of the present disclosure. Indeed, the novel methods, apparatus and systems described herein may be embodied in a variety of other forms, and further, various omissions, substitutions and changes in the form of the methods, apparatus and systems described herein may be made without departing from the spirit of the present disclosure. [Explanation of symbols]

[0110] 100 Voxels 105 First Substance 110 Second Substance 115 Beam (monochromatic beam) 115a Ray of light 115b Ray of light 120 Detector 125,400 macro pixels 130,420,440 micropixels 500 X-ray imaging gantry 501 X-ray tube 502 Annular Frame 503 X-ray detector (X-ray detector unit) 504 Data Collection System 505 Contactless Data Transmitter 506 Pretreatment Device 507 Rotating Unit 508 Slip Ring 509 High Voltage Generator 510 System Controller 511 Data / Control Bus 512 Memory 513 Current regulator 514 Reconstruction Device 515 Input Devices 516 Display

Claims

[Claim 1] an acquisition unit that acquires first projection data representative of an intensity of radiation detected at a plurality of detector elements; a generating unit that generates second projection data for each macro pixel formed by grouping two or more detector elements among the plurality of detector elements, based on the first projection data corresponding to the plurality of detector elements included in each macro pixel; a selection unit for selecting a macropixel for which material decomposition processing based on the second projection data does not converge; a correction unit that generates corrected material decomposition projection data by performing a material decomposition process based on first projection data corresponding to a part of detector elements included in the selected macro pixel; A medical image processing device comprising:

Citation Information

Patent Citations

  • Spectral x-ray imaging

    US20170023496A1

  • Method and apparatus to facilitate using fused images to identify materials

    US8422826B2

  • Pre-reconstruction calibration, data correction, and material decomposition method and apparatus for photon-counting spectrally-resolving X-ray detectors and X-ray imaging

    US9687207B2