JP2010082031A - X-ray ct image reconstruction method using energy information - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that when merely obtaining an image by using only energy of a small range close to monochrome, influence of beam hardening can be removed but count is reduced in a narrow window and a statistics error is increased and the removal cannot be attained by a present X-ray tube performance and data of each window has to be added (averaged) in some form, and that even when performing weighed averaging at the time of counting, remarkable control of the beam hardening is followed by side reaction of raising a weighing function of only a narrow energy portion, that is, increasing the influence of statistics noise. <P>SOLUTION: An image reconstruction method for an X-ray CT device subjects a count measured values obtained in a plurality of energy windows to transformation to the line integral of an X-ray linear attenuation coefficient and thereafter to weighed averaging. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an X-ray CT imaging apparatus, and more particularly to a spectrum measurement type X-ray CT imaging apparatus that acquires photon energy information.

  In Patent Document 1, in order to provide an image based on transmitted radiation equivalent to that obtained in the conventional integration mode in a state in which artifacts due to the beam hardening phenomenon and a decrease in contrast resolution of soft tissue are prevented. In addition, the count data of each of the plurality of energy regions is weighted by a weighting factor given for each energy region, and the count data of each of the plurality of energy regions for each weighted collection pixel is added to each other to obtain a radiation image for each collection pixel Output as data for generation.

JP 2006-101926 A

  The process is a general process of spectrum measurement until a count is obtained for each energy, and thereafter, an arbitrary weighted average (addition) process is performed on the count value. The effect of beam hardening can be eliminated even if an image is obtained using only a narrow range of energy close to a single color, but this is not possible with the current X-ray tube performance because the count is reduced in a narrow window and the statistical error is increased. However, it is necessary to add (average) the data for each window in some form.

  However, even if weighted averaging is performed at the time of counting, there is a side effect of increasing the weighting function only for a narrow energy portion, that is, increasing the influence of statistical noise, in order to significantly suppress beam hardening.

  An image reconstruction method for an X-ray CT apparatus, characterized in that weighted averaging is performed after the count measurement values obtained in a plurality of energy windows are converted into a line integral of an X-ray attenuation coefficient.

  It is possible to obtain a good reconstructed image that is free from beam hardening artifacts or is significantly suppressed and statistical noise is suppressed to a conventional level.

  Embodiments will be described below with reference to the drawings.

  FIG. 1 shows an outline of an X-ray CT apparatus. The X-ray CT apparatus 1 includes a gantry 2, an X-ray tube 3 which is a radiation source for emitting X-rays, a detector panel 4 for detecting radiation, a bed on which a subject sleeps, and an operation panel 7 for operating the X-ray CT apparatus in the field. An image processing calculation device 8 for generating an image based on a detection signal from the detector, an input / operation device 9 for inputting a command to the image processing calculation device, and a display device 10 for displaying an image generated by the image processing calculation device. Have. The detector panel 4 has a plurality of detection elements for converting X-ray energy information into electrical signals, and the detection elements have a pulse mode measurement circuit in the subsequent stage. The X-ray tube 3 and the detector panel 4 are arranged opposite to each other with the subject 20 interposed therebetween, and are rotated around the subject 20 in a 360-degree direction by a rotation drive unit of the gantry. The subject 20 stands still on the bed 6 and performs imaging while the set of the X-ray tube 3 and the detector panel 4 moves in the direction of the body axis in the direction perpendicular to the paper surface (longitudinal direction of the bed).

FIG. 2 shows a spectrum measurement type CT image reconstruction procedure. The image reconstruction procedure in FIG. 2 is executed by the image processing calculation device 8 in FIG. The image processing calculation device 8 includes a memory that records the contents of each calculation process and a CPU that is a calculation process execution unit. When the X-ray tube 3 and the detector panel 4 are irradiating and detecting X-rays while rotating the gantry 2, the number of photons (hereinafter referred to as count C) that is a measured value is 1 pixel within a certain time (eg, 1 ms). Obtained in the form of the energy spectrum of the incident photons. When the representative energy of the _N energy windows is E _N , the projection count data is C (ρ, θ, E). Here, θ is an angle in the gantry rotation direction at which the center of the detector panel 4 can be considered to be localized within a short period of time. For simplification, ρ is a value after parallel beam conversion. Focusing on an arbitrary slice, the body axis direction z is not considered, and the fluctuation of μ is treated as a sufficiently small constant value within the energy window. Does not affect the nature of

  Here, the reconstruction procedure is roughly divided into a weighted average operation 51, count → μ-line integral conversion operation 52, μ → HU value conversion operation 53, and image reconstruction operation 54 (line integral → map). The weighted average operation 51 is irreversible (= losing part of the information). The count → μ line integral conversion operation 52 and the μ → HU value conversion operation 53 are reversible but non-linear. The image reconstruction operation 54 is reversible and linear if the discretization error is ignored. These processes are executed by the CPU of the image processing calculation device 8, and constitute a conversion operation unit, a weighted averaging unit, a normalization unit, and an image reconstruction unit of the image processing calculation device 8. The conversion operation unit, weighted averaging unit, normalization unit, and image reconstruction unit can be configured by a circuit instead of software processing.

  In the figure, the following is adopted as a notation with a reduced number of characters to avoid complications.

Projection C (ρ, θ, E) × N window: C × N
Reconstructed image μ (x, y, E) × N window: μ-map × N
Weighted average: Horizontal line on variable X-ray line attenuation coefficient (μ) [unit length ^-1 ], which is the image reconstruction target value of the CT device, is energy-dependent, and has an absorption edge in the energy region below MeV With the exception of (1), μ increases as the energy increases, that is, the transmission probability per transmission distance increases. Accordingly, when considering an X-ray photon group distributed in continuous (non-monochromatic) energy, the thicker the material to be transmitted, the more the existence ratio in the remaining beam is biased toward the higher energy side. In X-rays, the high energy side is described as hard and the low energy side is described as soft, and this phenomenon is generally called beam hardening.

The radiation detection method used in CT apparatuses that are commercially available in the past is the current mode. The signal I obtained in the current mode is an output proportional to the sum of the product of the incident photon energy E and the number C, and the energy information of the individual photons is lost. If C is a function of E,
I∝∫ (E · C (E)) dE (1)
It is.

The line integral of the linear attenuation coefficient when the monochromatic X-ray passes through the substance having the linear attenuation coefficient μ (ξ) in the ξ direction is defined as ∫μ (ξ) dξ (hereinafter referred to as ∫μ). If the initial number of photons is C ₀ , the number of photons after 透過 μ transmission is C = C ₀ exp (−∫μ), and from the measured values C and C ₀ ∫μ is ∫μ = log (C ₀ / C) Formula (2)
Can be obtained as In CT, in a certain xy plane, this ∫μ is obtained as a width ρ (ξ is orthogonal to ρ), ∫μ (ρ, θ) is obtained over the entire circumference θ, and Radon inverse transformation ∫μ (ρ, θ) → μ ( An image reconstruction μ value distribution on the xy plane, that is, a tomographic image is obtained by x, y).

In a conventional CT apparatus, the radiation source is a non-monochromatic X-ray tube,
∫μ _current = log (I ₀ / I) = log ((∫E · C ₀ (E) dE) / (∫E · C (E) dE))
... Formula (3)
Because it is trying to find ∫μ in the same form as a single color, it is affected by beam hardening. Specifically, the thicker the transmission path, the more the beam is biased toward the hard side, the denominator of the remaining amount of the beam becomes larger, and ∫μ per unit length obtained is the unit length when passing through the extremely thin same material The value is smaller than ∫μ per unit.

If a flat μ-distributed cylinder is imaged, the path closer to the outer periphery has a larger ∫μ _{current and the} path closer to the center has a smaller ∫μ _current . As a result, μ appears to be small. This is an artifact (false structure) that does not reflect the actual structure, and is a typical beam hardening artifact called a cupping effect. In addition, considering the uneven μ distribution like the bone in the human body, beam hardening is especially lacking ∫μ in the path through the bone, and the reconstructed μ value appears to be low in the gap between the bone and bone. , Causing various artifacts.

  Further, the non-reproducibility of the reconstruction value is a beam hardening effect that is not a local artifact. In Expression (3), for example, the μ value obtained by reconstructing water depends on the size of the imaging target and is not uniquely determined. Conversion to Hounsfield unit values based on μ values of water also uses μ values of water measured by phantoms, and the HU values were not universal. Furthermore, even in individual imaging, substances with the same μ will show different values in the reconstructed value due to differences in patient size and the like.

  On the other hand, instead of current mode, pulse mode measurement, in particular, spectrum measurement type CT that obtains energy information is being developed. Until the count is obtained for each energy bin, it is a general process of spectrum measurement, and thereafter, an arbitrary weighted average (addition) process of the count value is performed. The effect of beam hardening can be eliminated even if an image is obtained using only a narrow range of energy close to a single color, but this is not possible with the current X-ray tube performance because the count is reduced in a narrow window and the statistical error is increased. However, it is necessary to add (average) the data for each window in some form.

  However, even if weighted averaging is performed at the time of counting, there is a side effect of increasing the weighting function only for a narrow energy portion, that is, increasing the influence of statistical noise, in order to significantly suppress beam hardening. In addition, the weighting function for weighted averaging is highly arbitrary, and no optimization guideline has been obtained for common use in general.

  Therefore, in the conventional CT apparatus, the Hounsfield unit value obtained by both the current mode type and the spectrum measurement type cannot uniquely determine the correspondence with the actual substance, and there is a problem in quantitative diagnosis such as density change.

  Further, if weighted addition is performed after image reconstruction, image reconstruction with a high calculation cost is N times when the number of windows is N, and this is a big problem in scenes where high resolution is required in the in-plane direction, tomographic direction, and time direction. sell.

  When the energy of continuous X-rays exceeds the region where Compton scattering is dominant and the pair generation is high, μ becomes larger on the higher energy side, so the term of beam hardening is not appropriate. There is no change in energy dependence. In other words, the above discussion is similarly problematic in industrial X-ray CT using high energy of MeV or higher.

  As a comparative example, the current measurement value obtained in the current mode type CT (integration mode) is obtained by using a weighted average operation 51 having a weighting function W (E) in spectrum measurement as W (E) = E as a first operation. Indicates that

  We paid attention to the order of conversion from count (or current), which is a nonlinear operation, to ∫μ, and weighted average, which is an irreversible operation (= a part of information is lost). Solves the fact that the statistics are insufficient for the information of each energy window alone by weighted averaging.

  In each energy window and each projection data, the weighted average is not performed at the time of counting but after conversion from counting to ∫μ, thereby removing or strongly suppressing beam hardening artifacts after image reconstruction. At this time, it is not necessary to bias the weighting function to a narrow range, and there is no particular side effect of increasing statistical noise.

  Furthermore, it has been found that there are two major HU values depending on whether Hounsfield unit value conversion is performed before weighted averaging, and one is superior in terms of universality. This is defined as a method for realizing a new HU value. In each projection data, (1) count to ∫μ, (2) ∫μ to ∫HU, (3) weighted averaging, and (4) image reconstruction operations The order is the image reconstruction technique unique to the spectrum measurement CT.

  By considering a simple addition average (flat weighting function) as a special case, a uniquely determined HU value suitable as a standard for comparison with a conventional current measurement HU value can be provided.

  By preparing this comparative standard condition, it is possible to obtain both a high-quality image and a universal (quantitative) HU value with an arbitrary weighting function (for example, enhancement of low energy components, minimization of statistical noise, etc.). . Both can be associated by segmentation processing on the high-quality image side.

  A good reconstructed image with no statistically reduced statistical noise, with no or no beam hardening artifacts, is obtained with a single number of image reconstructions for multiple energy windows. Furthermore, with respect to the reconstructed HU value, an image that is universal and suitable as a comparison reference can be obtained, and a high-contrast image such as low-energy component enhancement can be managed quantitatively.

  However, it is assumed that the image reconstruction operation 54 is performed after the count → μ line integral conversion operation 52 and the μ → HU value conversion operation 53 is performed after the count → μ line integral conversion operation 52. It is found that there is a large difference in the obtained image reconstruction results depending on the combination order of the operation groups, and a suitable image reconstruction technique that takes the operation cost into consideration is provided. In an X-ray computed tomography system that has a means of obtaining photon energy information and has a spectrum measurement type radiation detector as a pixel, the data acquisition system has a plurality of energy windows. The X-ray of water is converted to the X-ray attenuation coefficient line integral using the logarithm of the ratio of the count measurement value to the non-attenuated (air transmission) count measurement value acquired in advance for each energy window. An image reconstruction method that performs normalization using the attenuation coefficient, image reconstruction, and weighted averaging will be described.

The count → μ-line integral conversion operation 52 uses a natural phenomenon in which the decay of the number of photons occurs exponentially with monochromatic photons (considered as photons in each energy window). The cause of the beam hardening artifact is that the relationship between the transmission distance and the weighted count collapses from the exponential relationship depending on the operation order of the weighted average operation 51 at the time of counting. Here, what is called beam hardening effect is roughly divided. (BH 1) μ depends on energy (BH 2) (Under (BH 1)) ∫ μ depends on the transmission path length If it understands that it is mixing, it turns out that the influence of (BH2) can be removed by confirming a 1st operation to count-> micro-integral conversion operation 52. FIG. In other words, the conventional current measurement and weighted average at the time of counting can completely eliminate a problematic beam hardening artifact that diffuses from a certain reconstruction value to surrounding pixels (sharing the transmission path).

  As described above, the weighting average operation 51 may be performed after the count → μ-line integral conversion operation 52, and the μ → HU conversion operation 53 and the image reconstruction operation 54 may be performed in any order in the entire image reconstruction process. .

  Furthermore, the beam measurement artifacts can be reduced by the image reconstruction method of the X-ray CT apparatus that performs weighted averaging after converting the count measurement values obtained in a plurality of energy windows into the line integral of the X-ray attenuation coefficient. A good reconstructed image in which statistical noise is suppressed to a conventional level without or significantly suppressed can be obtained.

  In addition, a bed on which the subject is placed, a radiation source that emits radiation around the bed, a detector that detects the radiation, and an image processing calculation device that generates an image based on a detection signal from the detector, The image processing calculation device includes a conversion operation unit that converts the count measurement values obtained in a plurality of energy windows into a line integral of an X-ray attenuation coefficient based on a detection signal, and a line integral of an X-ray source weak coefficient. By using an X-ray CT apparatus having a weighted averaging unit that performs weighted averaging, a reconstructed image having no statistically reduced statistical noise and no beam hardening artifacts can be obtained. Can do.

  The description will be made with reference to FIG. 2 as in the first embodiment.

The X-ray attenuation coefficient line integral data existing as projection data is normalized using the water X-ray attenuation coefficient known for each energy window, and the water-integrated X-ray attenuation coefficient line integral is obtained. An image reconstruction technique for performing weighted averaging after conversion into the above will be described. Whether or not the μ → HU value conversion operation 53 is performed before the weighted average operation 51 will be discussed. In the spectrum measurement type CT, a universal HU value using water μ known for each energy is obtained for each window. Furthermore, considering the weighted average between windows, two systems can be realized. Respectively, E is the energy value in the window, W (E) is the weighting function, Σ is the sum of the number of windows N, and HU value (A) = (ΣW (E) · μ (E)) / (ΣW (E) · Water μ (E)) ... Formula (4)
HU value (B) = Σ (W (E) · μ (E) / (W (E) · water μ (E))) / ΣW (E) (5)
It is. The HU value (A) corresponds to the value obtained by performing the μ → HU value conversion operation 53 after the weighted average operation 51, and the HU value (B) corresponds to the value obtained by performing the weighted average operation 51 after the μ → HU value conversion operation 53. Since μ (E) / water μ (E) generally has energy dependence, the HU value (A) and the HU value (B) are not the same value.

  In this embodiment, the HU value (B) is a weighted average based on HU value = μ (E) / water (E) for each energy that can be defined without using the concept of window width, and is intuitive. In the HU value (A), there is a wasteful need to introduce a correction term for canceling the influence of the water μ to the weighting function (for example, statistical noise correction) for μ, but the weighting function for the HU value in the HU value (B). Is not necessary for the HU value (A), it has been found that it is superior in terms of universality.

  As described above, the X-ray attenuation coefficient line integral data is normalized using the known water X-ray attenuation coefficient for each energy window, and the water normalized X-ray attenuation coefficient line integral is obtained. By performing weighted averaging after conversion to, waste of introducing a correction term for canceling the influence of water μ on the weighting function for μ can be eliminated.

  Further, the image processing calculation apparatus standardizes the line integral of the X-ray attenuation coefficient, which is the output of the conversion operation unit, using a known water X-ray attenuation coefficient for each energy window, and normalizes the water. An X-ray CT apparatus having a conversion unit for converting to an X-ray attenuation coefficient line integral and a weighted averaging unit for performing weighted averaging on the line integral of a water-standardized X-ray source weak coefficient is used as a weighting function for μ. The waste of introducing a correction term for canceling the influence of water μ can be eliminated.

  This will be described with reference to FIG.

  An image reconstruction method for performing image reconstruction after performing weighted averaging will be described.

The HU (B) described in the second embodiment is defined as a method for realizing a new HU value. (1) Count → μ-line integral conversion operation 52 for each projection data
(2) μ → HU value conversion operation 53
(3) Weighted average operation 51
(4) Image reconstruction operation 54
Each of the operation orders is used as an image reconstruction technique specific to the spectrum measurement CT. This is the order indicated by the solid line in FIG. 2, but according to this, the image reconstruction operation 54 needs to be performed only once, and the effects of the first and second embodiments are also achieved.

  Further, not only for HU (B), but also for HU (A), after count → μ-line integral conversion operation 52, weighted average operation 51 is performed, and after this weighted average operation 51 is performed, image reconstruction operation 54 is performed. In fact, the image reconstruction process may be performed once, and the effects of the first embodiment are also achieved. The image reconstruction operation 54 may be performed after the weighted average operation 51, and the μ → HU value conversion operation 53 may be performed before or after the image reconstruction operation 54.

  This will be described with reference to FIG.

In particular, the water normalized X-ray attenuation coefficient value tomographic image obtained by image reconstruction using a flat weighting function is stored as a constant reference water normalized X-ray attenuation coefficient,
An image reconstruction method using a representative value of a water standardized X-ray attenuation coefficient group depending on a change in the weighting function will be described.

  Now consider the beam hardening effect (BH 1). Conventionally, the influence of (BH 2) was mixed, so μ (E) (or μ (E) / water μ (E)) depends on complicated input conditions such as irradiation spectrum distribution, imaging target size and shape. Therefore, the HU value determined on a one-to-one basis could not be defined for the substance, but until the previous section, the HU value for the substance was greatly reduced to two systems of HU value (A) and HU value (B). ) Was selected. The remaining degree of freedom is only the weighting function used to obtain one representative numerical value from a certain function.

  If the simple addition average (flat weighting function) is selected as the simplest case, the HU value uniquely determined for each substance (atomic number, density, mixing ratio) suitable as a standard for comparison with the conventional current measurement HU value Can be given.

  If this comparative standard condition is prepared, the weighting function was adjusted even if the HU value changed when adjusting the weighting function (for example, emphasizing low energy components, minimizing statistical noise, etc.) to obtain a high-quality image. A universal (quantitative) HU value can be managed by a method such as performing segmentation with an image on the high image quality side and obtaining an average HU value within a region determined by segmentation with a comparative standard image.

  This is not limited to a certain apparatus, and can be used as a universal HU value standard that can be shared by CT apparatus groups using the same method.

  As mentioned above, by using a radiation source that emits continuous X-rays and an energy-dependent X-ray attenuation coefficient, beam hardening artifacts that occur in CT reconstructed images are eliminated, and materials that could not be achieved in the past One-to-one reproducibility of (atomic number, density, mixing ratio) and reconstruction value can be obtained.

  Further, the reconstruction operation group performs the conversion from the count to the μ-line integral value before the weighted averaging, thereby localizing the beam hardening effect to one pixel. As a method of HU, it is determined that HU is also performed before weighted averaging and image reconstruction, and the remaining energy dependency is managed by using a flattened weighting function as a reference, and the substance and reconstruction value. 1 to 1 reproducibility can be obtained.

Outline of X-ray CT apparatus. Spectral measurement type CT image reconstruction procedure.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 X-ray CT apparatus 2 Gantry 3 X-ray tube 4 Detector panel 6 Bed 7 Operation panel 8 Image processing calculation apparatus 9 Input / operation apparatus 10 Display apparatus 20 Subject 51 Weighted average operation 52 Count → μ-ray integral conversion operation 53 μ → HU value conversion operation 54 Image reconstruction operation

Claims (10)

After converting the count measurement value obtained in multiple energy windows to the line integral of the X-ray attenuation coefficient,
An image reconstruction method for an X-ray CT apparatus, characterized by performing weighted averaging. The image reconstruction method according to claim 1,
After the line integral data of the X-ray attenuation coefficient is normalized using a known water X-ray attenuation coefficient for each energy window, and converted to a water-integrated X-ray attenuation coefficient line integral. ,
An image reconstruction method characterized by performing the weighted averaging. The image reconstruction method according to claim 1,
After performing the weighted averaging,
An image reconstruction method characterized by performing image reconstruction. The image reconstruction method according to claim 2,
After performing the weighted averaging,
An image reconstruction method characterized by performing image reconstruction. In claim 2, the water standardized X-ray attenuation coefficient value tomographic image obtained by image reconstruction using a flat weighting function is recorded as a constant reference water standardized X-ray attenuation coefficient,
An image reconstruction method characterized in that a representative value of a water normalized X-ray attenuation coefficient group depending on a change in a weighting function is used. A bed on which the subject is placed;
A radiation source emitting radiation around the bed;
A detector for detecting the radiation;
An image processing calculation device for generating an image based on a detection signal from the detector;
The image processing calculation device comprises:
Based on the detection signal, a conversion operation unit that converts a count measurement value obtained in a plurality of energy windows into a line integral of an X-ray attenuation coefficient;
An X-ray CT apparatus comprising: a weighted averaging unit that performs weighted averaging on a line integral of the X-ray source weak coefficient. The X-ray CT apparatus according to claim 6,
The image processing calculation device comprises:
With respect to the line integral of the X-ray attenuation coefficient that is the output of the conversion operation unit, normalization is performed using a known water X-ray attenuation coefficient for each energy window, and the water normalized X-ray attenuation coefficient is calculated. It has a normalization part that converts to line integrals,
The X-ray CT apparatus, wherein the weighted averaging unit performs weighted averaging on a line integral of the water normalized X-ray source weak coefficient. The X-ray CT apparatus according to claim 6,
An X-ray CT apparatus comprising an image reconstruction unit that performs image reconstruction using an output of the weighted averaging unit. The X-ray CT apparatus according to claim 7,
An X-ray CT apparatus comprising an image reconstruction unit that performs image reconstruction using an output of the weighted averaging unit. The X-ray CT apparatus according to claim 7,
An X-ray comprising a recording unit for recording a water-standardized X-ray attenuation coefficient value tomographic image obtained by image reconstruction using a flat weighting function as a constant reference water-normalized X-ray attenuation coefficient CT device.

Images