The present invention relates to a computed tomography system using radiation.

Tomography systems using radiation, which are well known, calculates an original image, or a picture, i.e., a distribution of radiation absorption coefficients defining a slice-like sectional portion or a sectional area of a subject to be reconstructed, in such a manner that radiation beams are projected into the sectional portion in parallel with the plane and in various directions and the picture is calculated on the basis of the measured projection data thus obtained including a lot of information relating to the image of the selected sectional portion. In this type apparatus, many calculation methods for reproducing an original image of the sectional portion from the measured projection data thereof have been known and one of them is disclosed in the Japanese patent application publication No. 28385/75. Excellent methods are known to be the convolution method and the filtered back-projection method. Both methods are basically common in (1) a modifying operation to obtain modified projection data or filtered projection data by the filtering of the measured projection data in order to improve poor image definition and (2) a back-projection operation for calculating the original image of the sectional portion by back-projecting based on the modified projection data. The modifying operation and the back-projection operation (both operations will be collectively referred to as the original image reconstruction operation for the sake of simplicity) are usually executed by using a digital computer or a special computer operable at a high speed.

As a result, all the data used in the above operations are sampled data. Necessary sampled data to execute the original image reconstruction operation are obtained by detecting numerous radiation beams, which beams are projected parallel at given sampling intervals at each specified sampling angular interval in the range of angles generally from 0° to π. The intensity of each radiation beam penetrating the sectional portion is detected and converted to a logarithmic value which value is referred to as a sampled projection data or measured projection data value. A set of the sampled projection data values at each angular sampling position is referred to as a profile of sampled projection data.

A distribution of the radiation absorption coefficients of the sectional portion, i.e. the original image of the sectional portion, is obtained by calculating the distribution of the radiation absorption coefficients of a large number of picture elements which are defined by segmenting the sectional portion in matrix fashion. In order to reconstruct the original image, or the picture from the projection data, it is necessary to obtain the sampled projection data from the radiation beams passing through the respective elements in various directions. However, the sampled projection data values are discrete values and the picture elements are discretely arranged. Therefore, it is very difficult to obtain the radiation beam exactly passing through a desired picture element, and thus to obtain the sampled projection data passing through the picture element. To overcome this difficulty, in a conventional data processing system, the sampled projection data is first subjected to the filtering operation for enhancing definition of the picture to calculate the sampled modified projection data or filtered projection data, then the sampled modified projection data representing the radiation passing through the picture elements is calculated from the sampled modified projection data resulting from the filtering operation, by using an interpolation.

In the conventional original image reconstruction method, the weight coefficients necessary for the interpolation operation are calculated for each sampling angular position and each picture element. Accordingly, the operation is cumbersome and extremely time consuming.

Accordingly, an object of the invention is to provide a computed tomography system capable of high speed reconstruction of an original image of the desired sectional portion of a subject to be reconstructed.

To achieve the object mentioned above, the tomographic system according to this invention comprises means for transmitting a large number of radiation beams through a selected slice-like sectional portion of a subject to be reconstructed including a given plane of the subject, substantially in parallel with said plane; detecting means for receiving each radiation beam transmitted through said sectional portion and generating output signals; means for obtaining from said output signals a profile of projection data based on a set of measured radiation beams transmitted parallel through said sectional portion at first sampling intervals ΔX₁ substantially at each specified sampling angular interval; means for obtaining a first profile of modified projection data by subjecting said profile of measured projection data to a specified filtering function; means for computing from said first profile of modified projection data a second profile of modified projection data sampled at second sampling intervals ΔX₂ determined corresponding to said specified sampling angular interval; means for obtaining interpolated projection data from said second profile of modified projection data corresponding to the respective picture elements formed by segmenting said sectional portion in matrix fashion; means for computing by a back-projection operation an original image of said sectional portion from the interpolated projection data for the respective picture element.

When modified projection data are calculated by means of interpolation from the profile of recast projection data the required calculations may be considerably reduced by properly selecting the second sampling interval which is determined with respect to the pitch of the picture element arrangement and penetration direction of the radiation beam. Accordingly, the time taken for the image reconstruction is remarkably reduced.

According to a preferred embodiment, when the second sampling interval is designated by ΔX₂, the pitch in the picture element arrangement by Δ, a projection angle θ_k calculated by substracting π/2 from the angle defined by the direction in which a radiation beam is transmitted or penetrates and the direction in which the rows of matrix arrangement extend by θ_k and such a positive integer as n makes ΔX₂ and ΔX₁ equal and as is specific to a given value of θ_k, the second sampling intervals ΔX₂ is calculated as follows: ##EQU1## With the selection of the second sampling interval ΔX₂, when the sampled modified projection data resulting from the radiation beams passing through a particular picture element is calculated by means of interpolation from the second modified projection data resulting from the radiation beams traveling parallel both sides of the radiation beams passing through the picture elements, the weight coefficients required for the interpolation operation are all equal for the following picture elements respectively; (a) picture elements arranged on x-axis (θ_k =0) of Cartesian coodinates fixed on an original image, when 0≦θ_k <π/4 or 3/4π≦θ_k <π; (b) picture elements arranged on lines parallel to the x-axis when 0≦θ_k <π/4 or 3/4π≦θ_k <π; (c) picture elements arranged on the y-axis when π/4≦θ_k <3/4π; (d) picture elements arranged on lines parallel to the y-axis when π/4≦θ_k <3/4π.

As a result, the amount of calculations of the sampled modified projection data required for all the picture elements is considerably reduced.

As mentioned above, the original image reconstruction method makes it possible to speed up the back-projection operation by sampling the first modified projection data at the second sampling interval. However, the following data processing methods may also shorten the operation time of the entire system. The first is to use a special apparatus operable at a high speed for the filtering operation, e.g. a convolver. The convolver is capable of vector operation at a high speed and thus is suitable for an operation in which the operation to obtain a scalar product of vectors is repeated, such as a convolution operation. The interpolation operation to calculate the interpolated projection data relating to the respective picture elements by using the second sampled modified projection data, the operation for calculation of the weight coefficients, and the operation for adding into the memory the interpolated projection data thus calculated as the corresponding data for picture elements are substantially the same as the filtering operation conducted by the special apparatus. Accordingly, when the special apparatus is applied to the filtering operation, and the interpolation operation and the addition operation, the time required for these operations is considerably shortened and therefore the efficiency of the entire apparatus is remarkably improved.

Other objects and features of the invention will be apparent from the following description taken in connection with the accompanying drawings, in which:

FIG. 1 shows a principle for forming sampled projection data and a profile of projection data by using radiation beams;

FIG. 2 shows a principle for reconstructing an original image of a sectional portion of an object from the profiles of modified projection data;

FIGS. 3 and 4 illustrate relations among the sampling positions at first and second sampling intervals, picture elements and weight coefficients;

FIG. 5 shows in block form a basic construction of an embodiment of the invention;

FIG. 6 is a block diagram of the part for performing an original image reconstruction operation in the embodiment shown in FIG. 5;

FIG. 7 illustrates a flow chart for generally illustrating the original image reconstruction operation; and

FIGS. 8a, 8b and 8c show operations of a special operation unit shown in FIG. 6.

Like a conventional computed tomography system using radiation, the system of this invention operates in such a manner that, a number of radiation beams (X-ray beams in this embodiment) are projected into a sectional portion of a subject in various directions. The intensities of the radiation beams projected in a given direction are measured and the projection data based on the measurement data i.e. measured projection data are collected in many directions. Each profile of measured projection data is subjected to a filtering operation to produce a first profile of modified projection data. The measurement and the filtering operations are well known and will be briefly referred to.

In FIG. 1, reference numeral 10 designates a thin slice-like sectional portion of the subject, the portion including a section or a plane across the subject. In the figure, two orthogonal coordinate systems each having the origin O are used; one is an x-y coordinate system and the other is an X-Y coordinate system corresponding to the x-y coordinate system rotated counterclockwise around the origin O by an angle θ. Denoted by reference numeral 12 is a radiation beam penetrating in a direction parallel with the section at a given first sampling interval in the direction as indicated by an arrow, i.e. in the Y-axis direction turned by the angle θ from the y-axis. The radiation beam 12 passes through the point (X₁, O) on the X-axis. Such a set of parallel radiation beams is formed by using parallel beams; however, it may be formed in other suitable manners. In one alternative form, radiation beams projected in a fan-shaped fashion to the sectional portion of the subject are turned around the sectional portion. The turning of the radiation beams is stopped at given angular intervals to be detected or measured. In this case, the angle by which the direction of the radiation beams is measured with respect to O-X axis is shown in FIG. 1 and is called "scan angle". The measured projection data thus discretely collected are inputted into a computer where a set of radiation beams passing substantially parallel with one another through the sectional portion in a given direction at a given interval is selected at each given angular direction. The angle defined by the O-X axis shown in FIG. 1 and the direction in which the radiation beams are transmitted parallel to one another is called the "projection angle" or "sampling angle". This selection is executed by using interpolation if desired. See FIG. 5. The same data processing steps succeeding to this step are applicable for both the methods.

The intensity of the radiation beam, e.g. X-ray beam, denoted by reference numeral 12 is detected after penetration through the sectional portion 10 by a detector 22 (FIG. 5) and the detected intensity is converted into the corresponding logarithmic value by a suitable electronic circuit. The logarithmic value is a sampled projection data value g(X₁, θ) and is depicted as g(X₁, θ) on the profile curve 14 shown on an X-g coordinate system in FIG. 1. The sampled projection data values relating to the X-ray beams parallel to the X-ray beam 12 would be shown by a number of bars disposed at first sampling intervals and in parallel with the value g(X₁, θ). However, such sampled projection data are shown by a continuous curve 14 representing an envelope or a profile connecting the top ends of the bars, for simplicity. A set of the sampled projection data designated by 14 is a profile of projection data relating to the angle θ and is generally expressed by g(X, θ).

In order to reconstruct the original image, i.e. the picture, of the sectional portion, it is necessary to obtain a profile g(X, θ) for each sampling position or each projection direction disposed at a different angular interval, wherein each angle θ is selected within the range from O to π.

FIG. 2 is used to explain the principle for reconstructing the original image of the sectional portion 10. In the figure, among numerous profiles of sampled projection data, two profiles 14a and 14b are illustrated by way of example; the former is collected at the projection angular position in which the radiation beams penetrate the sectional portion in the direction of the Y-axis which is displaced by θ₁₀ from the X-axis and is expressed by g[X, θ₁₀ -(π/2)], and similarly the latter is expressed by g[X, θ₁₂ -(π/2)]. The profiles 14a and 14b are subjected to the well-known filtering operation to calculate profiles of first modified projection data 16a and 16b which are expressed by q[X, θ₁₀ -(π/2)] and q[X, θ₁₂ -(π/2)], respectively. Each of the profile first modified projection data consists of a first sampled modified projection data or filtered projection data sampled at a first sampling interval. Then, by means of a back projection operation, the X-ray absorption coefficients of the picture elements forming the original image 18 shown in FIG. 2 are calculated from the profiles of the first modified projection data q(X, θ) with respect to many angles θ. As previously referred to, a distribution of the X-ray absorption coefficients represents the original image, or the picture. The coordinates of the picture elements are expressed generally by (x,y). In FIG. 2, one picture element P at (x,y) is illustrated for simplicity. The filtering operation is performed in accordance with the following equation. ##EQU2## and |ω|·H(ω) is the filter function in a Fourier region.

The filtering operation can also be expressed by the following equation ##EQU3## Equation (1) corresponds to the filtering process in the filtered back-projection method and equation (2) to that in the convolution method. In the above equations (1) and (2), ω and X' are integration variables introduced into the integrating operations.

The back-projection operation for getting the original image or picture data is performed by the following equation. ##EQU4##

In general, the reconstruction operation of the original image is discretely conducted by using a digital computer or a special calculator operable at a high speed.

The sampling operation for the original image reconstruction will be described referring to FIG. 3. As shown, a number of sampling points are disposed on the X-axis of an X-Y coordinate system rotated by θ_k from the x-y coordinate system. The X-ray beams penetrating through the sectional portions progress in the Y-axis direction through these sampling points, for example, (X_l-1,θ_k), (X_l,θ_k), (X_l+1,θ_k) . . . (X_l+3,θ_k). X_l and ΔX₁ are related by the equation (4), wherein ΔX₁ is the sampling interval, i.e. first sampling interval.

Xl =l·ΔX1, θk =k·Δθ(4)

Thus, g(X_l,θ_k) can be written in the form of g(l·ΔX₁,θ_k). In the above relations, l is 0 or a negative or positive integer, k is 0 or a positive integer, 0, 1, 2, . . . N.sub.θ and Δθ is a sampling angular interval of the angle θ defining the direction of the X-axis.

The original image, or the picture, i.e. f(x,y) of the sectional portion of the subject, is also expressed by a discrete value f(iΔ,jΔ). Here, Δ designates the pitch in the picture element arrangement formed by segmenting the original image in matrix fashion in the x- and y-directions; i and j indicate an integer within the range from -N to +N. Therefore, the equation (3) can be rewritten in the form ##EQU5## In the above equaton, if X is given

X=iΔ·cos θk +jΔ·sin θk (6)

the X in (6) (for example, one length of X is indicated by X_Q1 in FIG. 3) is not coincident with the X_l. In FIG. 3, P₁ and P₂ designate two picture elements adjacently disposed on the line parallel with the x-axis and are expressed by coordinates (iΔ,jΔ) and ((i+l)Δ,jΔ). Points Q₁ and Q₂ are feet of the perpendiculars from the points P₁ and P₂ to the X-axis, respectively. The points Q₁ and Q₂ generally do not coincide with the point (X_l, θ_k), because the point (X_l,θ_k) is generally selected independently from the points P₁ and P₂. To calculate the radiation absorption coefficients of the picture element P₁ or P₂, it is necessary to obtain the sampled modified projection data i.e. interpolated projection data, by the radiation beam which penetrates through the point Q₁ or Q₂, accordingly the point P₁ or P₂, in the Y-axis direction. However, such data does not exist in fact. For this, the interpolated projection data at the point Q₁, for example, is calculated from the first sampled modified projection data at the points (X_l,θ_k) and (X_l+1,θ_k) by an interpolation. The interpolation projection data for the point Q₂ is calculated from the first sampled modified projection data at the points (X_l+2,θ_k) and (X_l+3,θ_k). Many interpolation methods may be used in this case. In this example, the linear interpolation widely used will be employed for calculating the interpolated projection data at the points Q₁ and Q₂. First, the first sampled modified projection data relating to the points (X_l,θ_k) and (X_l+1,θ_k) are expressed by q(X_l,θ_k) and q(X_l+1,θ_k), respectively. Then, weight coefficients ω_2l and ω_1l to be multiplied by the values q(X_l,θ_k) and q(X_l+1,θ_k), respectively are calculated by the following equations: ##EQU6## The final equation indicates that the l is smaller than X_Q1 /ΔX₁ and the nearest integer. The parentheses [ ] hereinafter described indicate the same. With use of the ω_1l and ω_2l, the interpolated projection data q(X_Q1,θ_k) is given by the following equation:

q(XQ1,θk)=ω1l ·q(Xl,θk)+ω2l ·q(X1+1,θk)                      (9)

The above calculation is executed for each angle θ_k over all the picture elements to obtain the interpolated projection data relating to the respective sectional portions of the subject.

The above-mentioned interpolation operation is performed at each of the angles θ_k and for all of the picture elements, thus resulting in many calculations. However, the invention employs a new operation method described below to reduce remarkably the necessary calculations for the interpolation and thus the time taken for the operation is greatly decreased.

In brief, the fundamental principle of the new operation method is such that the second sampled modified projection data or recast projection data q'(X_m,θ_k) sampled at a second sampling interval ΔX₂ is calculated from the first sampled modified projection data q(X_l,θ_k) sampled at the first sampling interval ΔX₁ for the angle θ_k. In this case, the envelope of a set of the second sampled modified projection data q'(X_m,θ_k) for various values of m will collectively be referred to as the profile of the second modified projection data values q'(X,θ_k).

The second sampling interval, or sampling interval for recasting, ΔX₂ is given by the following equation, relating to the pitch Δ of the picture elements and the inclination θ_k of the X axis. ##EQU7## Where n is a positive integer and, preferably, is so selected that the first and second sampling intervals ΔX₁ and ΔX₂ are substantially equal to each other.

Referring now to FIG. 4, some of the new sampling points are disposed along the X-axis at the second sampling intervals ΔX₂ and are denoted as (X_m,θ_k), (X_m+1,θ_k), (X_m+n,θ_k) and (X_m+n+1,θ_k). Here, the X_m is selected to have the following relation with the ΔX₂

Xm =m·ΔX2                         (11)

where m is 0 or a positive or negative integer. The operation for calculating the second sampled modified projection data q'(X_m,θ_k), i.e. q'(m·ΔX₂,θ_k) from the first sampled projection data q(ΔX_l,θ_k) is executed in the same method as the above discribed linear interpolation. Accordingly, the explanation of it will be omitted.

Let us calculate the interpolated projection data values q'(X_Q1,θ_k) and q'(X_Q2,θ_k) corresponding to points Q₁ and Q₂ shown in FIG. 4 respectively, from the second sampled modified or recast projection data q'(X_m,θ_k). The points Q₁ and Q₂, and points P₁ and P₂ are the same as those shown in FIG. 3. In order to calculate the interpolated projection data q'(X_Q1,θ_k), weight coefficients ω_1m and ω_2m to be multiplied by the second sampled modified projection data q'(X_m,θ_k) and q'(X_m+1,θ_k) will be calculated by the following equations: ##EQU8## As seen from equation (12), ω_1m and ω_2m depend on θ_k and j and not i. Accordingly, the interpolated projection data q'(X_Q1,θ_k) for the point P₁ (iΔ,jΔ) (simply expressed α(i)) is given

α(i)=ω1m ·q'(Xm,θk)+ω2m ·q'(Xm+1,θk)                     (14)

The interpolated projection data q'(X_Q2,θ_k) for the point P₂ ((i+1)Δ,jΔ) (simply expressed α(i+1) is given

α(i+1)=ω1m ·q'(Xm+n,θk)+ω2m ·q'(Xm+n+1,θk)

Thus, for the picture elements disposed along the j line expressed by y=jΔ, the interpolation may be executed by using the same weight coefficients ω_1m and ω_2m. As a result, it is sufficient to calculation only the weight coefficients corresponding to the number of the line j. Consequently, when the second modified projection data is calculated through the change of i from -N to +N at each value of j being changed from -N to +N, the (2 N+1) operations may be executed by using the same weight coefficients. For this, it is unnecessary to calculate the weight coefficients for the (2 N+1)² picture elements defining the entire area, unlike the conventional operation. Therefore, the calculation of the interpolated projection data for all the picture elements, i.e. the reconstruction of the original image of the sectional portion, or the picture thereof, is performed in reduced time. The operation shown in the equation (14) is the convolution operation and can be rapidly conducted by using the special operation apparatus.

FIG. 5 shows a block diagram of an embodiment of the computed tomography apparatus according to the invention. An X-ray source 20 projects a number of X-ray beams 12a, for example, an R number of beams, into the sectional portion 10 of a subject. The X-ray beams penetrating through the sectional portion 10 are intensity-detected, after penetration, by a detecting means such as a detector 22. The detector 22 comprises detector elements 26a and an A-D converter 26b. The number of the detector elements 26a corresponds to that of the X-ray beams 12a. Upon receipt of the X-ray beams, the detector elements 26a produce analogue electrical signals which in turn are converted into digital signals through an A/D converter. The digital signals from the AD converter 26b are stored in a memory unit 28.

The detector elements 26a and the X-ray source 20 are turned around the sectional portion 10 in the direction of arrow 24. The above-mentioned measuring is performed at each given angular position or scan angle.

The digital signals stored in the memory 28 are converted into the corresponding logarithmic values by a computer 30, and then again loaded into the memory 28. Each logarithmic value represents the sampled projection data including the information relating to the X-ray absorption coefficient distribution along a path through which the X-ray beam penetrates the sectional portion 10. Various sampled projection data thus obtained are classified into data groups each including data with respect to X-ray beams of R number penetrating in a given angular direction, which direction is changed for each given angular interval. As a result, a profile of the sampled projection data relating to each projection angle θ (FIG. 1) is formed. The classification is performed by computer 30 and the result is again stored in the memory 28. In FIG. 1, reference numeral 14 designates the profile of the sampled projection data g(X,θ) and g(X₁,θ) expresses the sampled projection data.

The memory 28 comprises a program part of the computer 30 adapted to store the profile of the sampled projection data g(X,θ), a filter function h(X) memory part (the filter function h(X) being composed of discrete values corresponding to g(X,θ)), and a picture reconstruction memory part, etc. The computer 30 performs the filtering operation and the image reconstruction operation by given programs and stores the data representing the reconstructed image into the memory 28. The image-reconstruction data stored in the memory is read out therefrom when necessary and converted into analogue signals. The signals are then applied to a display unit 34 and visualized. The image reconstruction data and other data are read out from the memory 28 by the computer 30 may also be printed out by a printer 36. In the figure, an input unit 38 is shown connected to the computer.

FIG. 6 illustrates in greater detail the memory 28 and the computer 30 shown in FIG. 5. The circuit is so designed as to permit high speed processing of the filtering operation and the interpolation operation. In the figure, reference numeral 40 designates a central processing unit (CPU) of a minicomputer used for a computer 30 in FIG. 5, numeral 42 designates a special apparatus or special arithmetic unit, and numeral 44 is a bus permitting data transmission between the CPU 40 and the arithmetic unit 42. The arithmetic unit 42 includes a data memory 46, a function memory 48, a processing unit 50, a picture memory 52 and an address computation unit 54. The processing unit 50 is used for executing the filtering operation, and usually is constructed by multipliers and adders. For this arithmetic unit, a so-called convolver, exclusively used for a convolution processing, may be used.

The operation of the arithmetic unit 42 shown in FIG. 6 and the data processing flow from the filtering operation to the back projection operation will be described referring to the drawings of FIG. 8a, FIG. 8b, FIG. 8c and FIG. 7. FIG. 8a to FIG. 8c illustrate different operation modes of the arithmetic unit 42. FIG. 8a illustrates the case of the filtering operation performed by the arithmetic unit 42. In this case, the data memory 46, function memory 48, processing unit 50 and picture memory 52 are used as a projection data memory, filter function memory, filtering unit, and first modified projection data memory, respectively. In the filtering operation, the sampled projection data values g(l·ΔX₁,θ_k) collected at first sampling intervals ΔX₁ and the filter function are stored in the projection data memory 46 and the filter function memory 48, respectively. Through the operation of the address computation unit 54, the sampled projection data and the filter function at the respective sampling points are supplied to the filtering unit 50 where the filtering operation is carried out. The result of the filtering operation which is the first sampled modified projection data q(l·ΔX₁,θ_k), accordingly the profile of the first sampled modified projection data q(X,θ_k) is supplied to the modified projection data memory 52. As will be described in the order of processes indicated in FIG. 7, the second sampling interval ΔX₂ is calculated in accordance with the equation (10) by the CPU 40. In this case, the value of n is inputted by an operator. Then, the CPU 40 calculates the second sampled modified projection data q'(m·ΔX₂,θ_k) sampled at interval ΔX₂ from the first sampled modified projection data q(l·ΔX₁,θ_k), calculates the weight coefficients ω_1m and ω_2m shown in the equation (12) by using the n and the angular data θ_k, and further computes the X-coordinate of the foot m₀ (not shown) of the perpendicular from a picture element (not shown) located at the left end of j line given by y=jΔ to the X-axis.

Then, the arithmetic unit 42 executes, executes from the second modified projection data q'(mΔX₂,θ_k), as shown in FIG. 8b, the interpolation operation for calculating the interpolated projection data q'(X_Q,θ_k), i.e. the interpolation value α(i), which should be obtained from each of the X-ray beams passing the respective picture elements of the sectional portion. In this case, the data memory 46, function memory 48, processing unit 50 and picture memory 52 operate as the second sampled modified projection data memory, weight coefficient memory, interpolation unit, and interpolated modified projection data memory, the memory 46 and the weight coefficient memory 48 store the second sampled modified projection data q'(m·ΔX₂,θ_k) and the weight coefficients ω_1m and ω_2m, respectively. The address computation unit 54 receives signals m and m₀ transferred from the CPU 40 and successively reads out second sampled modified projection data and the weight coefficients ω_1m and ω_2m from the memories 46 and 48 and supplies the readout values to the interpolation unit 50. The interpolation unit 50 calculates the interpolation values i.e. data for back projection α(i) defined by the equation (14), then the interpolation value is stored in the interpolated data memory 52 in response to an instruction signal issued from the address computation unit 54. The interpolation operation may be performed by the CPU 40. This operation is substantially the same as the filtering operation so that it can be performed at a high speed by using this arithmetic unit 42.

Thus the calculated interpolation value α(i) is the interpolated modified projection data relating to the picture element located at the point (iΔ,jΔ) of the sectional portion and is added into the data of f(iΔ,jΔ) as shown in FIG. 8c. The interpolation values α(i) of all the picture elements (iΔ,jΔ) are calculated and the calculated interpolation values are added into the data of f(iΔ,jΔ) to reconstruct the picture of the sectional portion. This operation may be performed by the CPU 40 and also the arithmetic unit 42.

In the operation mode shown in FIG. 8c, the arithmetic unit 42 operates to add the interpolation value into the original image data, i.e. the picture data of f(iΔ,jΔ). In this mode, the data memory 46, processing unit 50 and picture memory 52 are used as f-α memory for storing picture data f(iΔ,jΔ) and interpolation value α(i), adding unit and picture data memory, respectively. In this case, the function memory 48 is not used and therefore the block 48 is blanked in FIG. 8c. In FIG. 8c, the picture data f(iΔ,jΔ) and the interpolation value α(i) are stored in the f-α memory 46. The adding unit 50 successively receives the picture data f(iΔ,jΔ) and the interpolation value α(i) from the memory 46 by the signals from the address computation unit and adds the interpolation value α(i) into the corresponding picture data f(iΔ,jΔ), then the result of the addition is stored in the memory 52. In this manner, when the arithmetic unit 42 is used for adding of the interpolation value, the adding operation may be performed at a high speed and thus the image may also be reconstructed at a high speed.

The data processing steps from the filtering operation to the back-projection operation will be described with reference to FIG. 7.

In the figure, reference numeral 100 designates a step to set the arithmetic circuit to the state k=0 so that the operation initiates from the time point where the X- and x-axis are coincident in FIG. 3. A step 101 selects one of two equations according to the value θ_k whereupon the second sampling interval ΔX₂ is calculated. A step 103 calculates the first sampled modified projection data q(l·ΔX₁,θ_k) from the sampled projection data g(l·ΔX₁,θ_k). A step 104 calculates the second sampling interval ΔX₂ in accordance with equation (10) by using CPU 40, with a given n. A step 105 calculates the second sampled modified projection data q'(m·ΔX₂,θ_k) from the first sampled modified projection data q(l·ΔX₁,θ_k). A step 106 sets the arithmetic unit 42 to j=-N in order to calculate the weight coefficients of the picture elements disposed along the j line (i.e. y=jΔ line) from y=-NΔ line to y=+NΔ line. A step 107 calculates the X-coordinate at the foot m₀ of the perpendicular from the point x=-NΔ and y=+jΔ to the X-axis and the weight coefficients ω_1m and ω_2m in accordance with the equation (12) by using the values of designated n and the angle θ_k, by means of the CPU 40. A step 108 causes the step 109 and succeeding ones to initiate from i=-N, m=m₀. A step 109 calculates the interpolation value α(i) in accordance with the equation (14). A step 110 changes m into m+n in order that the α(i) operation having been conducted for the point Q₁ is shifted to that for the point Q₂. A step 111 causes steps 109 and 110 to be repeated after the i is changed to i+1 by the step 112 when i is other than i N, and shifts to the step 113 when i N. Step 113 adds the interpolation value α(i) into the picture data f(iΔ,jΔ). A step 114 shifts to the next step 116 when j N, and repeats step 107 and succeeding steps after i is changed to j+1 by step 115 when j is other than j N. A step 116 completes the calculation of the interpolation value for the respective picture elements of the sectional portion of the subject and the addition when k N.sub.θ, and repeats step 101 and succeeding steps after k is changed to k+1 when k is other than k N₇₄.

The calculation of the second sampled modified projection data q'(m·ΔX₂,θ_k) from the first sampled modified projection data q(l·ΔX₁,θ_k) through the changing of the sampling interval in the step 105 is performed in the same manner as the interpolated projection data relating to the points Q₁ and Q₂ and thus P₁ and P₂ in FIG. 4 are calculated from the second sampled modified projection data through linear interpolation. Therefore, the elaboration of the calculation of the second sampled modified projection data q'(m·ΔX₂,θ_k) will be omitted.

Having described a specific embodiment of the invention, it is believed obvious that modification and variation of the invention is possible in light of the above teachings. The radiation beams penetrating the sectional portion may be arranged in parallel with one another or arranged in fan-shape fashion. A turning movement, a parallel movement and a combination of them are applicable for the movement of the radiation source and the detector. The radiation source may be placed within the subject itself, unlike the above-mentioned embodiment placing the radiation source on the outside of the subject. In place of the change of the sampling method of the modified projection data by the second sampling interval, another method is allowable in which the radiation beams are measured at sampling intervals depending on the penetration direction of the radiation beams, and the modified projection data sampled at the sampling interval are directly used. Further, when the original image is back projected on the basis of the second sampled modified projection data, a higher order interpolation method may be employed. The interpolated projection data obtained in this method are more precise.