WO2004006745A2

WO2004006745A2 - System and method for performing tomosynthesis of an object

Info

Publication number: WO2004006745A2
Application number: PCT/EP2003/006137
Authority: WO
Inventors: Eckhard Sperschneider
Original assignee: Macrotron Process Technologies Gmbh
Priority date: 2002-06-11
Filing date: 2003-06-11
Publication date: 2004-01-22
Also published as: WO2004006745A3

Abstract

The invention relates to a method and a system for performing tomosynthesis of an object (O) made of a material. Said system comprises a device (XRAY) for forming projections, particularly an x-ray device which forms projections (P1(0)... P1(m)) of the object, and a computer (CC) to which a control program is assigned. In order to determine the interior space of the object (O), the projections (P1(0)...P1(m)) thereof are compared with projections (E) of a reference object, and the projections (E) of the reference object are approximated to the projections (P1(0)...P1(m)) of the object (O).

Description

System and method for tomosynthesis of an object

The invention relates to a system and method for tomosynthesis of an object according to the preamble of claim 1 or 6.

With computer tomography, the internal structure of objects can be made non-destructive and non-contact, whereby individual layers (2D) as well as volumes (3D) can be reconstructed. According to information published by the Fraunhofer-Gesellschaft on the Internet at http://www.iis.fhg.de/xrt/tech/ct/index.html, spatial resolutions of up to 10 μm are achieved.

According to the above information, a 3-D industrial computer tomography system enables the detection of density changes and errors as well as characterization of their type, geometry and location in a component. In addition, internal, hidden structures can be measured. Software modules allow the display of arbitrary levels from the completely reconstructed volume data as well as a spatial representation of individual volume areas.

For objects that are not suitable for axial computed tomography due to their geometry, planar computer tomography can be used to visualize density changes and errors that are characterized by their geometry and location in the object can. Digital tomosynthesis allows the reconstruction of multiple layers from one set of projections, while classic laminography produces only one slice per measurement.

To determine the material composition of an object (in terms of geometry, volume and material distribution and structure), object data are metrologically recorded. For example, the object to be determined is recorded by X-ray technology under defined recording geometries. From several (n-1) dimensional measurement data (object projections), a (n) dimensional object is synthesized and thus reconstructed. This is done by means of the projection data, a mathematical description of the recording geometry and an algorithm that describes the imaging system.

This method requires a large number of measurement data from different acquisition geometries. Their extraction and their computational processing require much time because of the relatively large amounts of data.

Based on this prior art, the present invention seeks to provide a system and a method of the type mentioned, wherein the number of projections can be further reduced.

The invention is associated with a plurality of advantages. By reducing the number of projections, the processing time can be reduced, enabling the system to process a larger number of objects per unit of time. At the same time the energy input for the processing of an object is reduced.

The invention will now be described with reference to the drawing.

It shows

Figures 1 to 18 representations in connection with that of the system according to the invention feasible method, inter alia, in comparison with conventional methods;

Figure 19 is a block diagram of the system according to the invention; and

Figure 20 steps of the method according to the invention.

The system shown in FIG. 19 comprises a device (XRAY) for forming projections, in particular from an X-ray recording device, the projections (P1 (0)... P1 (m)) of the object layers (01 (5)... -5)), from a detector unit DET and from a computing unit CC, which is assigned a control program. This defines the method according to the invention. Furthermore, the system is assigned an input / output unit I / O. In the unit are first data D1, which designates the geometry of the object (O) (= setpoint geometry of a reference object), and / or second data D2, which denotes the material of the object (O) (= material of the reference object), and / or Additional data that designate the object can be entered. On a monitor of the unit I / O, it is possible to represent layers of the object as well as layers of a reconstructed space, such as e.g. shown in Figures 10 to 18.

The method practicable by the system is as follows:

An initial object Oorigin _a i is located in the room, cf. eg Figure 19. After

The process can be shown in several layers, see FIG. 1 and the sections O1 (5)... O1 (-5) and O2 (5)... O2 (-5) in FIGS. 10 and 11.

By means of X-ray images projections P are produced at different angles, figures 2 (projections Pi: -45 °, P ₂ : 0 °, P ₃ : + 45%), 10 and 11.

These projections generated by the original object O (in real, for example, created by X-ray technology) form the starting point for a reconstruction of the original object O. To reconstruct the original object 0, use the following algorithm. The starting point is a homogeneous material distribution in the virtual object space, FIG. 3 (homogeneous object, virtual object space in the first reconstruction _step V _rec ).

Based on the imaging geometry of the original projections, new virtual projections (E0... N) are now calculated from this initial object, FIG. 4 (calculated projections Ei: -45%, E ₀ : 0 °, E ₂ : + 45%).

Based on the difference between the original projection P and the virtual calculated projection E, a correction of the object space is made, with the aim of aligning the calculated projections E in several iteration steps to the original projections O.

If it is possible to modify the virtual object space such that the calculated projections E are identical to the original projections P, the object space has been reconstructed.

Thus, in the course of several iteration steps based on the original projections and the calculated virtual projections, a correction of the object space to be reconstructed is undertaken

V _r ec (n + 1) = V _rec (n) ⁺ Function (Eo (n) -Pθ (n), E i _(n) -Pi (n), E ₂ _(n) -P2 (n) - Em (n) -PM (n))

(NeuRaum = AltRaum + Korrektur) for additive reconstruction methods (NeuRaum = AltRaum * Korrektur) for multiplicative reconstruction methods

In the type of correction calculation, the different algebraic reconstruction methods differ. However, the basic procedure (iterative change of the object space and approximation of the virtually calculated projections to the original projections) is common to all these methods.

As part of the method, projections P1 (0)... P1 (m) of the object O are formed; for the determination of the interior of the object O, the projections P1 (0)... P1 (m) of the object O are compared with projections E of a reference Object and the projections E of the reference object are approximated to the projections P1 (0)... P1 (m) of the object O. Furthermore, to determine the interior of the object O, first data D1, which designate the geometry of the reference object, and / or second data D2, which designate the material of the object O, and / or further data, which designate the object, are used.

Specifically, deviations between data indicating the reference object and data indicating the object are formed.

For determining the interior of the object O, when using at least two data, e.g. D1 and D2 the data used are weighted differently (e.g., 0.4 and 0.6). Thus, data that designates a highly absorbent object material is weighted less than data that designate the target geometry of the object (= geometry of the reference object).

After determining the interior of the object O, data which designate the specific object interior can be compared with predefinable reference data. Depending on the comparison, the object O is released or not released for a predefinable use.

A major problem with all reconstruction methods is that with small numbers of projections different material distributions in the object space can produce the same projections. That the smaller the number of projections, the greater the uncertainty in the reconstructed object space.

For example, both material distributions shown in FIG. 5 could produce the same projections in space.

(dark = material with strong absorption, lighter = material with lower absorption)

Only through additional projections can the object space be correctly reconstructed, FIG. 6. In the case of the homogenous object material, this problem can be reduced.

But even in this case and with a small number of projections, the reconstruction of the object space can only be done with a certain degree of uncertainty.

In the example, even with a homogeneous material, the same projections could occur in both material distributions in space, Figure 7. In the example shown, the material is homogeneous, but the size and distribution in space is different. Nevertheless, the three beams produce nearly identical projections. But since only the projections are known, the space can only be reconstructed with uncertainty.

In this example too, the reconstruction quality could be improved by increasing the number of projections, FIG. 8.

Assuming, however, that one knows a large part of the material distribution, the uncertainty described above can be reduced, cf. Figure 9 (left: actual material distribution, middle: expected material distribution, right: unknown material distribution).

In the process, the actual material distribution is constructed from a good approximation (expected material distribution), which proves to be simpler, and thus high, than to calculate the actual material distribution from an unknown material distribution. Accordingly, in the context of the method for determining the interior of an object O, said first and / or second data are used. It is also possible to use data representing an error to be found, e.g. to denote a crack.

The method is suitable, inter alia, for cast-iron objects whose (desired) outer geometry is known and which consist of exactly one - known - material. Even with 3 projections, very good spatial reconstruction results can be achieved. The method which can be carried out by the system according to the invention can in principle also be applied to objects which consist of more than one material.

For the images shown in FIG. 10 (1st column (left): n layers of the

Original object O; 2nd column: projections P of O at different angles, 3.

Column: n layers of the reconstructed space, all projections were used; State of the art in iteration> 10, 4th column: n layers of the reconstructed space using all projections and the Inform.

"Homogeneous material" at iteration> 10) became a calculated phantom

(Reference object) used, which consists of a large ball, which contains inside 2 smaller spherical holes. These pictures are the result of the first reconstruction attempts.

It can be seen clearly the qualitative difference that is achieved by the use of the information "homogeneous material", see Figures 10, 3 and 4.

Column.

Artifacts in the picture (a double cross in the middle of the picture) are the result of problems in the calculation of the geometric relationship between projections and images

Object space.

Also for the images shown in Figure 11 (1st column (left): n layers of the original object O; 2nd column: n layers of the reconstructed space after the 1st iteration, 3 projections were used, system according to the invention; 3rd column: n layers of the reconstructed space after the 3rd iteration, 3 projections were used, system according to the invention), a calculated phantom (reference body) was used, which consists of a large ball containing 2 smaller spherical holes inside. The upper hole is smaller than the lower one. Both holes merge into each other. Artifacts are largely missing. The geometry calculation could be decidedly improved. For the reconstruction, the following 3 projections of the original were used. Projections of the original (0, -45 °, + 45 °), Figure 12

These 3 projections alone were used for the reconstruction. According to the above-described algorithm, an attempt is made to approximate the calculated projections to the measured original projections by varying the material distribution in the space. Projections of the reconstruction after the 1st iteration, Figure 13 Projections of the reconstruction after the 3rd iteration; FIG. 14

After the third iteration there is no improvement of the reconstruction result by further iteration steps.

The projections after the third iteration are very similar to the projections of the original.

As computation time requirement for the above small objects the following times resulted: I .Iteration ~ 50ms 2nd iteration ~ 10ms 3rd iteration <1ms

The following assumptions were made:

1. Homogeneous material, i. every place in object space is made of material or nothing (air)

2. Basic geometry is widely known, i. For the reconstruction, it is assumed that the large sphere is located in the object space.

The holes are reconstructed only by the reconstruction in position and size in space. For the qualitative control of the reconstruction, the sections of the original space can be compared with the sections of the reconstructed space in FIGS. 12 to 14.

In order to clarify the decided advantage of the method, in which a good approximation of the initial geometry is taken into account, the results of the reconstruction method without consideration of the geometry are clarified below.

It is assumed that the third projections of the original, Figure 15

By varying the material distribution in the room, the following calculated projections were determined after the first iteration step:

Projections of the reconstruction after the 1st iteration, Figure 16;

Projections of the reconstruction after the 10th iteration, Figure 17.

It can be seen that a correspondingly good approximation of the calculated projections to the original projections is also achieved here by a larger number of iteration steps.

The calculation times required for the above small objects were - with an unknown object distribution - the following times: 90 ms 90 ms 71 ms 60 ms 60 ms 51 ms 50 ms 40 ms 50 ms

40ms The sections of the original space and the reconstructed space shown in FIG. 18 show considerable differences. It becomes clear that despite a very good approximation of the projections, a qualitatively sufficient reconstruction of the room based on only 3 projections is not possible. Only with additional projections can this result be improved.

In summary, it can be stated that in the method which can be carried out by the system according to the invention, whose steps are shown schematically in FIG. 20, starting from a largely known geometry of an object (eg CAD data are present) and assuming that it is a homogeneous material (FIG. eg aluminum casting) good reconstruction results can be achieved. It is not necessary to reconstruct the entire 3D object, only deviations from an existing object geometry are reconstructed. Thus one obtains a difference model between (target) initial model (desired geometry) and actually existing actual model.

Defects in objects (voids) can thus be well described and represented in terms of volume and arrangement.

The (3D) method described above, which is performed by the system according to the invention, can be further simplified. Using the acquisition technique (LDA), the calculation is limited to the reconstruction of 2D slices based on n-1 D projections. This considerably reduces the volume of data to be processed and the computing time.

LIST OF REFERENCE NUMBERS

O object

01 (5) ... 01 (-5) Object layers

P1 (0) ... P1 (m)) Projection of the object (01 (5) ... O1 (-5))

E Projections of the reference object

XRAY device for the formation of projections P

DET detector

CC arithmetic unit

I / O input / output unit

Claims

claims

A system for tomosynthesis of an object (O) made of a material, comprising means (XRAY) for forming projections, in particular an X-ray recording device, which forms projections (P1 (0) ... P1 (m)) of the object , and with a computing unit (CC), which is assigned a control program, characterized in that the control program is further configured in such a way that for the determination of the interior of the object (O) the projections (P1 (0) ... P1 ( m)) of the object (O) are compared with projections (E) of a reference object, and that the projections (E) of the reference object are approximated to the projections (P1 (0) ... P1 (m)) of the object (O) ,

2. System according to claim 1, characterized in that the control program is further configured in such a way that for determining the interior of the object (O) first data (D1), which designate the geometry of the reference object, and / or second data (D2 ) indicating the material of the object (O) and / or other data denoting the object.

3. A system according to any one of the preceding claims, characterized in that the control program is further configured in such a way that deviations between data indicating the reference object, and data denoting the object are formed.

4. System according to claim 2 or 3, characterized in that the control program is further configured in such a way that for determining the interior of the object (O) in the use of at least two data (D1 D2), the data used, the geometry of the reference object, the material of the object (O) and / or the object denote be used in different weighting.

5. System according to any one of the preceding claims, characterized in that the control program is further configured in such a way that after determining the interior of the object (O) data, which denote the particular object interior, are compared with predetermined reference data, and that in dependence of the comparison, the object (O) is released or not released for a predefinable use.

6. A method for tomosynthesis of an object (O), which consists of a material, wopbei projections (P1 (0) ... P1 (m)) of the object are formed, characterized in that for determining the interior of the object (O) the projections (P1 (0) ... P1 (m)) of the object (O) are compared with projections (E) of a reference object, and that the projections (E) of the reference object correspond to the projections (P1 (0) ... P1 (m)) of the object (O).

7. The method according to claim 6, characterized in that for determining the interior of the object (O) first data (D1), which designate the geometry of the reference object, and / or second data (D2), the material of the object (O) and / or other data designating the object.

Method according to claim 6 or 7, characterized in that deviations between data designating the reference object and data denoting the object are formed.

9. The method according to claim 7 or 8, characterized in that for determining the interior of the object (O) when using at least two data (D1 D2) the data used, the geometry of the reference object, the material of the object (O) and / or designate the object to be used in different weighting.

10. The method according to any one of claims 6 to 9, characterized in that after determining the interior of the object (O) data that denote the specific object interior, are compared with predetermined reference data, and that depending on the comparison, the object (O) for a predefinable use is released or not released.