DE102009004576A1 - Method for evaluating multiple two-dimensional spatially resolved projection images of local three-dimensional object, involves determining resulting probability distribution of elongated structure in local three-dimensional object - Google Patents

Abstract

The method involves determining a resulting probability distribution of an elongated structure in the local three-dimensional object for each projection image by a computer (1). The resulting probability distribution is based an initial probability distribution given for the projection image of the elongated structure in a local three-dimensional object and the respective projection image. Both the initial probability distributions and the resulting probability distribution are three dimensionally spatially resolved. An independent claim is also included for a computer program with a machine code.

Description

The The present invention relates to an evaluation method for a number of two-dimensionally spatially resolved projection images a spatially three-dimensional object.

The The present invention further relates to a computer program which Machine code, which can be processed directly by a computer is and its execution by the computer causes the calculator performs such an evaluation process.

After all The present invention relates to a computer equipped with a such computer program is programmed so that it is in operation performs such a programming method.

The The above objects are well known.

For example It is known, based on a variety of two-dimensional projection images, the be detected under different angulations, a three-dimensional To determine the volume data set of the object. This approach is not important in the context of the present invention.

Another example is the re-mapping of the projection images into a given volume data set with resolution of the resulting ambiguities. This procedure is taken in particular in connection with the allocation of blood flow. Purely by way of example is on the US 6,823,204 B2 and the US 6,442,235 B2 directed.

In connection with the allocation of the blood flow, probability-based approaches are also known. Purely by way of example is on the US 2008/243435 A1 directed.

The Medicine is increasingly making an effort to minimize surgical procedures perform invasively. Typical application areas minimal Invasive surgical procedures are vascular diseases such as aneurysms (vessel bulging), Stenoses (vasoconstriction) or aterio-venous Short circuits (AVM). Within the diagnosis of such diseases and partly also in the context of their therapy become three-dimensional imaging techniques used, for example, CT and magnetic resonance imaging.

The Minimally invasive procedures are often performed using a Catheter or a guidewire (hereinafter generally referred to as elongated structure). The elongated structure is introduced in the groin area and from there to navigated to their place of work. The navigation takes place in the state of Technique with the help of two-dimensionally spatially resolved X-ray images (= projection images within the meaning of the present Invention). The x-rays are taken using a C-arm X-ray system added. Since such an intervention often requires considerable time (sometimes over an hour), it is important that Radiation exposure for the patient and for the keep medical staff as low as possible. It is therefore unreasonable to perform a 3D scan over and over again and the object including the elongated structure each reconstruct.

On The reason of only two-dimensional given information will continue the navigation of the elongated structure through coverings and ambiguity difficult. The intellectual dissolution the overlaps and ambiguities required by the attending physician much experience and good anatomical knowledge. Despite all knowledge, however, it can happen that a projection image intellectually can not be resolved. In one In such case, it is necessary to use the X-ray arrangement to position on another angulation and the ambiguities to dissolve through another shot from the new angle. Especially with highly branched vascular systems with many overlaps, as for example in the brain area occur, such realignments extend the operation time considerably. Furthermore, they lead to an enlarged Radiation exposure of the patient and the person involved in the procedure medical staff.

In More and more hospitals are using so-called Biplan C-arm systems used. By means of such systems, it is possible to simultaneously X-rays from two different angulations too create. This will change the number of ambiguities significantly reduced. Nevertheless, even a biplane plant only delivers two-dimensional Pictures used for three-dimensional navigation become. The orientation problems thus continue to exist.

Essential more pleasant and intuitive would be a representation of the elongated Structure in three dimensions, for example in an existing one three-dimensional volume data set of the object. The doctor could ambiguities and overlaps in this case very much quickly and efficiently dissolve by rotation of the 3D representation. However, it is essential that the representation of the elongated Structure in three-dimensional is correct and current. The investigation The location of the elongated structure must therefore be reliable and be fast.

From the US 2007/189457 A1 It is known to estimate the three-dimensional position of the tip of a catheter with the aid of two-dimensional x-ray images and to display it in the three-dimensional data record of the patient. To determine the tip of the catheter, a particle filter is used. For the procedure of US 2007/189457 A1 it is necessary that in the two-dimensional radiographs, the catheter tip is segmented. This is done at the US 2007/189457 A1 manually. This renders real-time use impossible. In addition, the required calculation time of the method is the US 2007/189457 A1 too high to allow real-time deployment during surgery.

An essentially identical disclosure content is the study work "3-D tracking of a catheter in the vascular system of a patient from 2-D X-ray images" by D. Gohlke, Friedrich Schiller University Jena and Siemens Medical Solutions (Forchheim), from the year 2006 refer to.

The Object of the present invention is ways to create, by means of which an elongated structure such as a catheter in a spatially three-dimensional object quickly and reliably by means of a respective, local only two-dimensional projection image can be localized.

The Task is performed by an evaluation method with the characteristics of Claim 1 solved. Advantageous embodiments of the evaluation method are the subject of dependent claims 2 to 11th

The Task is still solved by a computer program having the machine code, which can be processed directly by a computer is and its execution by the computer causes the calculator carries out an evaluation method according to the invention. The computer program may be on a disk in machine-readable form be saved.

After all the task is solved by a computer, which with a such computer program is programmed so that it is in operation performs such an evaluation process.

• – dass ein Rechner für jedes Projektionsbild anhand einer für das jeweilige Projektionsbild gegebenen jeweiligen anfänglichen Wahrscheinlichkeitsverteilung einer langgestreckten Struktur in dem örtlich dreidimensionalen Objekt und des jeweiligen Projektionsbildes eine jeweilige resultierende Wahrscheinlichkeitsverteilung der langgestreckten Struktur in dem örtlich dreidimensionalen Objekt ermittelt,
• – dass sowohl die jeweiligen anfänglichen Wahrscheinlichkeitsverteilungen als auch die jeweiligen resultierenden Wahrscheinlichkeitsverteilungen dreidimensional ortsaufgelöst sind,
• – dass der Rechner für die Projektionsbilder anhand der jeweiligen resultierenden Wahrscheinlichkeitsverteilung dreidimensional ortsaufgelöst einen jeweiligen Anwesenheitsverlauf der langgestreckten Struktur in dem örtlich dreidimensionalen Objekt ermittelt und den jeweiligen ermittelten Anwesenheitsverlauf in einer Darstellung des örtlich dreidimensionalen Objekts anzeigt,
• – dass der Rechner den jeweiligen Anwesenheitsverlauf durch Maximieren einer Kostenfunktion ermittelt, in welche die Länge des jeweiligen ermittelten Anwesenheitsverlaufs und die Wahrscheinlichkeiten gemäß der jeweiligen resultierenden Wahrscheinlichkeitsverteilung über den ermittelten jeweiligen Anwesenheitsverlauf und die Krümmungen des jeweiligen ermittelten Anwesenheitsverlaufs eingehen.

According to the invention is provided in the context of the evaluation process,

• A computer determines a respective resulting probability distribution of the elongated structure in the spatially three-dimensional object for each projection image on the basis of a respective initial probability distribution of an elongated structure in the spatially three-dimensional object and the respective projection image given for the respective projection image,
• That both the respective initial probability distributions and the respective resulting probability distributions are three-dimensionally spatially resolved,
• - That the computer for the projection images based on the respective resulting probability distribution three-dimensional spatially resolved determines a respective presence history of the elongated structure in the spatially three-dimensional object and displays the respective determined presence history in a representation of the spatially three-dimensional object,
• - That the calculator determines the respective course of attendance by maximizing a cost function, in which the length of the respective determined presence history and the probabilities according to the respective resulting probability distribution over the determined respective presence history and the curvatures of the respective determined attendance course received.

The Projection images are usually X-ray images. exceptionally There may be other fluoroscopic images. In the The X-ray images DSA images (DSA = digital subtraction angiography).

The Define probability distributions for each location a respective probability that part of the elongated Structure stays in this place. Minimal are the individual Probabilities according to the probability distribution usually greater than zero. The integral of the probability distribution over the volume is normalized. In general, the individual can Probabilities have a variety of values. Especially the values are not binary.

in the Contrary to the probability distributions is the course of attendance binary: At a certain point of the object is set to "elongated Structure is present "or" elongated Structure is not present "detected. Intermediates are locked out.

In As a rule, the projection images form a temporal sequence. In In this case, it is preferably provided that the computer with respect each pair of immediately consecutive projection images the resulting probability distribution of the earlier one Projection image as initial probability distribution the temporally later projection image used.

Preferably, it is furthermore provided that the computer inserts the respective ascertained presence history into the respective projection image for each projection image. This allows the user immediately verify that the determined attendance is correct.

Of the each determined attendance course can in principle be arbitrary Be nature. It is preferably provided that the respective determined Presence history consists of a sequence of sections, where each section by a number of parameters one for all sections defined uniform parameterizable function is. This approach results in reduced complexity when determining the attendance history. As parameterizable function For example, the individual lines of a traverse or splines of given complexity are used especially cubic B-splines.

• – K die Kostenfunktion ist,
• – a, b und c von Null verschiedene Wichtungsfaktoren sind und x ein weiterer Wichtungsfaktor ist,
• – A ein von der Länge des jeweiligen ermittelten Anwesenheitsverlaufs abhängiger Einflussfaktor ist,
• – B ein von den Wahrscheinlichkeiten gemäß der jeweiligen resultierenden Wahrscheinlichkeitsverteilung über den ermittelten jeweiligen Anwesenheitsverlauf abhängiger Einflussfaktor ist,
• – C ein von den Krümmungen des jeweiligen ermittelten Anwesenheitsverlaufs abhängiger Einflussfaktor ist und
• – X ein weiterer Einflussfaktor ist.

In a preferred embodiment of the present invention, it is provided that the cost function is the form K = aA + bB + cC + xX having,

• - K is the cost function,
• - a, b and c are non-zero weighting factors and x is another weighting factor
• A is an influencing factor dependent on the length of the respectively determined attendance course,
• B is an influencing factor dependent on the probabilities according to the respective resulting probability distribution over the determined respective presence history,
• C is an influencing factor dependent on the curvatures of the respective detected course of the attendance, and
• - X is another influencing factor.

Of the last term xX serves in the context of the present invention as "silent Reserve". Because of the fact that the associated further weighting factor x can have the value zero, it is possible that the last term does not contribute to the cost function. However, the chosen formulation should be the possibility keep that open in addition to the contributions due to the influencing factors A, B and C further influence factors X can be taken into account if this is the case should prove useful.

It it is possible that the weighting factors - in particular the non-zero weighting factors - fixed are. It is alternatively possible that they are from a user can be specified, in particular are interactively changeable.

To the Determining the of the probabilities according to the respective resulting probability distribution over depend on the determined respective attendance course Influence factor is preferably provided that the integral of a probability function on the determined respective presence history is formed, wherein the Probability function of the probability of the respective one Place according to the respective resulting probability distribution depends.

When Probability function can in the simplest case the probability yourself. Even better, as a probability function to use a logarithm of probability. This approach has the particular advantage that at very small probabilities the logarithm tends towards minus infinity and so on simple Way very unlikely gradients are discarded.

To the Determining the determined by the curvatures of the respective Presence-dependent influencing factor is preferred provided that the integral of a curvature function over the determined respective presence history is formed, wherein the curvature function of the amount of curvature the respective determined attendance course at the respective place depends.

The Curvature function, for example, the amount of curvature itself or its square.

In The rule is that the calculator for each projection image for determining the respective resulting probability distribution in addition to the respective initial probability distribution and the respective projection image further information about possible locations of the elongated structure in the spatially three-dimensional object and / or over considers the locally three-dimensional object. This procedure simplifies the determination of the correct ones Probability distribution considerably.

For example is it possible that the further information (at least) a further projection image acquired simultaneously with the respective projection image of the object. The se procedure is particularly suitable then on, when a biplanar X-ray system is available stands.

According to a preferred embodiment of the present invention, it is provided-as an alternative or in addition to the use of further, simultaneously acquired projection images of the object-that the further information comprise a three-dimensionally spatially resolved course of a vascular system, wherein the vascular system is identical to the object. In particular, in this case the plausible assumption is possible that the elongated structure does not leave the vascular system. The initial truth The probability distribution can therefore be set to zero or to a very small value for all locations that are outside the vascular system.

In a further preferred embodiment of the present invention is provided that the calculator for each projection image for determining the respective resulting probability distribution based on the respective initial probability distribution and a three-dimensionally defined transition probability a respective three-dimensional spatially resolved preliminary probability distribution determined and the respective resulting probability distribution by Weighting the probabilities according to the respective provisional probability distribution with the corresponding Chances according to a two-dimensional determined spatially resolved image probability distribution, which calculates the computer on the basis of the respective projection images.

The two-dimensional image probability distribution is - analog to the three-dimensional probability distributions - for that characteristic, whether at the respective locations of the projection image a part of the elongated structure is or not. The minimum value of the probabilities according to the image probability distribution greater than zero. The integral over the projection image is normalized. There are a variety of values possible, so not just binary yes or no.

Also The transition probability can be a variety of Have values. The integral over the volume is up One normalized. As a rule, the minimum value is above zero. In general, the three-dimensional spatially resolved transition probability a monotone - especially strictly monotonous - falling Function of the distance between two places.

If Information about the vascular system as such are not given, the distance is preferably the Euclidean Distance used. In general, however, information is about given the vascular system. In this case, the Distance preferably to a path within the vascular system based. This circumstance will continue to places outside the vascular system a transition probability associated with zero (or near zero). Because there is no path and therefore no meaningful distance from a place of the vascular system to a place outside the vascular system.

• – anhand einer ersten Tabelle für die beiden Orte jeweils mindestens einen nächsten Knotenpunkt und maximal zwei nächste Knotenpunkte ermittelt,
• – die Abstände der beiden Orte von ihren nächsten Knotenpunkten ermittelt, anhand einer zweiten Tabelle die Abstände der Knotenpunkte voneinander ermittelt und
• – anhand der Abstände der beiden Orte von ihren nächsten Knotenpunkten und der Knotenpunkte voneinander den Abstand der beiden Orte voneinander ermittelt,

The distance of two locations of the vascular system from each other can be determined by the computer, for example, that he

• Determined on the basis of a first table for the two locations at least one next node and a maximum of two nearest nodes,
• - Determines the distances between the two locations of their nearest nodes, using a second table, the distances of the nodes determined from each other and
• The distance of the two locations from each other from their nearest nodes and the nodes from each other,

The Nodes include at least all branches of the vasculature.

Preferably is provided in the context of the present invention that the computer for each projection image in the course of determining the respective two-dimensional spatially resolved image probability distribution the respective image probability at a location of the respective Projection image based on the image data of the respective projection image, in an evaluation core around the respective location of the respective Projection image lying around, and a distance-dependent Weighting function determined.

For example can the calculator for each projection image in the context of Determining the respective two-dimensional spatially resolved Image probability distribution the respective projection image fold with a convolution core. The convolution kernel may alternatively fixed or user - also interactive - changeable be.

Prefers However, the calculator is within the scope of determining the image probability a particular place the influences of the in the evaluation core lying places determined on the respective place and as image probability of the respective place the maximum of the influences uses.

The calculator can in principle apply the initial probability distributions in any way. Preferably, however, the calculator uses the initial probability distributions as particle representations. In this case, the computer determines the respective resulting probability distributions using a particulate filter. Particle representations and particle filters are known as such. Purely by way of example is the technical essay "A Tutorial on Particle Filters for Online Nonlinear / Non-Gaussian Bayesian Tracking" by MS Arulampalam et al., IEEE Transactions on Signal Processing, Vol. 50, No. 2, 2002, pages 174-188 directed. Particle representations and particle filtering have the advantage that they require significantly less computational effort than other representations.

Further advantages and details result from the following description of exemplary embodiments in conjunction with the drawings. In a schematic representation:

1 a block diagram of an evaluation arrangement,

2 a flow chart,

3 a vascular system,

4 to 7 Flowcharts,

8th a transition probability,

9 a section of a vascular system and

10 and 11 Flowcharts.

According to 1 is a calculator 1 with a computer program 2 programmed so that he is in operation the computer program 2 executing. The computer program 2 can the calculator 1 for example, via a computer-computer connection (not shown), for example the World Wide Web. Alternatively, it is possible that the computer program 2 on a disk 3 is stored in machine-readable form (especially in electronic form) and the computer 1 over the disk 3 is supplied. The disk 3 in particular a mobile data carrier 3 be. Is shown in 1 Purely by way of example, an embodiment of the data carrier 3 as a USB memory stick. Alternatively, the disk could 3 For example, be designed as a CD-ROM or SD memory card.

The computer program 2 has machine code 4 on. The machine code 4 is from the calculator 1 directly executable. The execution of the machine code 4 through the computer 1 causes the calculator 1 performs an evaluation method, which will be described in detail below in conjunction with the other figures.

According to 2 defines the calculator 1 in a step S1, an initial probability distribution W3a. The initial probability distribution W3a is three-dimensionally spatially resolved. It gives for each location of a spatially three-dimensional object 5 - For example, a vascular system 5 as it is in 3 is shown - a probability w for that at the respective location of the object 5 a part of an elongated structure 6 located. At the elongated structure 6 For example, it may be a catheter or a guidewire.

Since in the execution of the step S1 usually no actual information about the actual location of the elongated structure 6 is known, in the context of step S1 is usually set an equal distribution. Regardless of whether a uniform distribution is applied or another distribution is applied, the integral of the initial probability distribution W3a is over the object 5 normalized to one.

The initial probability distribution W3a can be up different way to be set. Preferred is in the frame of the present invention that the calculator the initial probability distribution W3a attaches as a particle representation.

In a step S2, the computer takes 1 a two-dimensional spatially resolved projection image P of the object 5 opposite. The projection image P can be, for example, an X-ray image that is generated by means of an X-ray system 7 was detected at a time t. In principle, in this case the X-ray image P can be processed as it is from the X-ray system 7 is detected. Preferably, however, the X-ray image P is a DSA image. Using DSA images results in significantly better results.

In a step S3, the computer determines 1 based on the received in step S2 projection image P a two-dimensional spatially resolved image probability distribution W2. In a step S4, the computer determines 1 from the initial probability distribution W3a and the image probability distribution W2 a resulting probability distribution W3r. Especially in the case that the calculator 1 the calculator has determined the initial probability distribution W3a as a particle representation 1 the resulting probability distribution W3r in the course of step S4 using a particle filter.

The resulting probability distribution W3r is also spatially resolved in three dimensions. It gives for every place of the object 5 a likelihood that at the particular place a part of the elongated structure 6 located. The difference from the initial probability distribution W3a is that in the resulting probability distribution W3r, in addition to the initial probability distribution W3a, the information about the actual position of the elongated structure 6 are taken into account, which are given on the basis of the projection image P.

In a step S5, the computer determines 1 from the resulting probability distribution W3r a presence profile V of the elongated structure 6 in the object 5 , Also the Anwe The course of events V is three-dimensionally spatially resolved. He indicates where within the object 5 the elongated structure 6 located. The presence history V is for each location of the object 5 binary (eg, zero = no part of the elongated structure at this location 6 and one = at this location is a part of the elongated structure 6 ). The attendance course V is also continuous in itself. Any two places that are part of the elongated structure 6 According to the course of attendance V, therefore, they are connected to each other by a series of locations immediately adjacent to each other and all also part of the elongate structure 6 are according to the determined attendance course V.

The elongated structure 6 according to the determined presence course V further has a finite length S0. It has exactly one beginning and one end. Furthermore, for each location, the elongate structure exists 6 according to the determined attendance course V, which lies between the beginning and the end, exactly two - no more and no less - immediately adjacent places, which are also part of the elongated structure 6 are according to the determined attendance course V. The course of attendance V does not split up. An example of a possible presence course V is in 3 dashed lines.

In a step S6, the computer determines 1 a representation of the spatially three-dimensional object 5 and gives this representation to a user 8th from, for example, via a viewing device 9 , In this illustration, the calculator shows 1 the determined attendance course V on. For example, the calculator 1 make a corresponding insertion.

The presentation of step S6 is always made. Preferably, a step S7 is additionally present. In step S7, the computer projects 1 the determined presence profile V in the projection image P and displays the computationally determined projection of the presence profile V in the projection image P. Due to the step S7, it is for the user 8th it is readily possible to judge the quality of the determined course of presence V, that is to say its agreement with the projection image P.

It is possible to carry out the procedure described so far only for a single projection image P. In general, the procedure of 2 however, performed for multiple projection images P. If the projection images P (exceptionally) are uncorrelated, the steps S1 to S6 and possibly also S7 are processed separately for each projection image P. As a rule, however, the projection images P form a temporal sequence. In this case, additional steps S8 and S9 are present in addition to the steps S1 to S6 and S7.

In step S8, the computer checks 1 whether the corresponding presence course V has already been determined for all projection images P. If this is the case, the procedure of 2 completed. Otherwise, the calculator goes 1 to step S9 via.

In step S9, the computer sets 1 the initial probability distribution W3a equal to the previously determined resulting probability distribution W3r. This will ensure that the calculator 1 as the initial probability distribution W3a of the next projection image P, the resulting probability distribution W3r of the just processed projection image P is used. From step S9, the computer goes 1 back to step S2, in which the calculator 1 the next projection image P accepts.

In the context of the procedure now described, the resulting probability distribution W3r of the previous run is thus set as the initial probability distribution W3a during each run of steps S2 through S7. This is allowed because the actual location of the elongated structure 6 from projection image P to projection image P usually changes only slightly. This results in a very good convergence of the determined resulting probability distribution W3r to the actual position of the elongated structure even after a few projection images P. 6 ,

• – die Länge S0 des Anwesenheitsverlaufs V,
• – die Wahrscheinlichkeiten w der Orte, die entlang des Anwesenheitsverlaufs V liegen, und
• – die Krümmungen, die der Anwesenheitsverlauf V aufweist.

In the course of determining the presence history V (see the step S5 of FIG 2 ) sets the calculator 1 initially according to 4 in a step S11 a presence history V on. In a step S12, the computer determines 1 for the scheduled attendance course V a cost function K. In the cost function K go

• The length S0 of the course of attendance V,
• The probabilities w of the locations lying along the course of the presence V, and
• The curvatures which the course of occupancy V has.

Preferably, the length S0 of the course of attendance V is positive in the cost function K, so that the attendance course V becomes as long as possible. Preferably, furthermore, the probabilities w of the locations along the course of attendance V also enter positively into the cost function K, so that the presence course V has a high resulting probability. Preferably, however, the curvatures of the course of attendance V enter into the cost function K negatively, so that the course of attendance V is as smooth as possible and is straight. This situation is represented in step S12 by the signs of the weighting factors a, b and c.

In a step S13, the computer varies 1 the attendance course V, so that the cost function K becomes maximum. Corresponding investigations are known to those skilled in the art. By way of example, reference is made to the optimization algorithm of Powell, for example, in the technical paper "An efficient method for finding the minimum of a function of several variables without the generator" of MMD Powell, published in The Computer Journal 7 (2), 1964, pages 155-162 , is described. The thus determined attendance course V is in the further course of the 2 , that is used in the context of steps S6 and S7.

Of the Presence course V can in principle be arbitrary. Preferably the presence history V consists of a sequence of sections, where each section by a number of parameters one for all sections defined uniformly parameterizable function is. For example, the individual sections can be individual lines of a polygon. Alternatively you can the individual sections each Weil as a spline of a predetermined Be formed complexity.

The cost function K can in principle be arbitrary provided that it meets the above criteria. For example, it may have the shape which in step S12 of FIG 4 is specified. a, b, c, and x are weighting factors in the formula of step S12. The weighting factors a and b are greater than zero, the weighting factor c is less than zero. The weighting factor x may have any value. In particular, it may be zero, greater than zero, or less than zero.

A is an influencing factor that is determined by the length S0 of the Occupancy history V is dependent. In the simplest case it is the length S0 itself or a positive one Potency of length S0. The power can be an integer (for example the second or third power) or non-integer (for example the root or the 1.5th power of length S0).

B is an influencing factor which depends on the probabilities w according to the resulting probability distribution W3r over the determined respective presence course V. In the simplest case, it is the integral of the probabilities w itself. Alternatively, it can be - see 5 - to be the integral of a suitable function fB of the probabilities w, hereafter referred to as the probability function. The probability function fB in this case depends on the probability w of the respective location according to the resulting probability distribution W3r. It is particularly preferred that the probability function fB is a logarithm, in particular the natural logarithm.

C is an influencing factor, which depends on the curvatures of the determined presence profile V. For example, the influencing factor C according to 5 the integral of a Krüm mung function fC over the determined presence course V are formed. The curvature function fC in this case depends on the amount of curvature of the determined presence profile V at the respective location. The curvature function fC may in particular be the amount of the curvature itself or the square of the amount of the curvature.

Of the Factor X is a quiet reserve, as it were, of course is possible, in the cost function K also other influences to be taken into account.

The weighting factors a, b and c (and optionally also x) can be fixed. They can have the same value or different values in terms of amount. The weighting factors a, b and c (and optionally also x) are preferably user-specific 8th specifiable, in particular interactive specifiable.

In general, the calculator takes into account 1 for each projection image P for determining the respective resulting probability distribution W3r not only the respective initial probability distribution W3a and the respective projection image P, but additionally further information about possible locations of the elongated structure 6 in the locally three-dimensional object 5 and / or over the locally three-dimensional object 5 yourself. For example, the X-ray machine can 7 be designed as a biplane system, so that simultaneously with the projection images P from another angulation each a further projection image P 'can be detected. In this case, the procedure of 2 corresponding 6 be designed.

6 build up 2 on. 6 therefore also contains the steps S1 to S9. In addition, the procedure of 6 the steps S16 and S17.

Steps S1 to S9 have already been described above in connection with 2 explained. Therefore, there are no weiterge existing versions required. In step S16, the calculator takes 1 the respective further projection image P 'against. The further projection image P 'corresponds in terms of the projection image P, which was mentioned in connection with the step S2. The projection image P 'of the step S16 shows the object 5 however, under a different angulation than the projection image P of step S2. In step S17, the computer modifies 1 taking into account the further projection image P ', the initial probability distribution W3a. The step S17 corresponds essentially to the further projection image P 'of the step S16 of the sequence of steps S3, S4 and S9.

As already mentioned several times, the object is 5 usually a vascular system 5 , It is therefore possible - both as an alternative and in addition to the modifications according to 6 - That the further information is a three-dimensional spatially resolved course of the vascular system 5 include. In this case, the procedure of 2 according to 7 at least by a step S21, preferably also supplemented by a step S22.

In step S21, the calculator takes 1 a volume record 10 opposite. In the volume data set 10 is the object 5 (= the vascular system 5 ) segmented. This is the calculator 1 able to locate outside the vascular system 5 to assign very low probabilities (in extreme cases zero) within the framework of the initial probability distribution W3a.

In step S22, the computer determines 1 a transition probability w '. The transition probability w 'is a probability of being in a particular location of the object 5 a part of the elongated structure 6 is located within the context of the immediately preceding iteration at a different location of the object 5 a part of the elongated structure 6 has found. The transition probability w 'is defined in three dimensions. As a rule, the transition probability w 'is in accordance with 8th a monotone - especially strictly monotonic - decreasing function of the distance d of two places. For example, it may be a Gauss curve or a similar bell curve. The width of the bell curve can be interactively adjustable. Furthermore, the bell curve may be location-dependent, that is, depending on the specific location whose probability is to be determined.

As in 7 is shown, the step S4 of 2 preferably implemented in the form of two steps S26 and S27. In step S26, the computer determines 1 for each location the respective probability w according to a preliminary probability distribution W3v by integration of the product of the probability w according to the initial probability distribution W3a and the transition probability w '. Integrated is the volume. The provisional probability distribution W3v weights the calculator 1 in step S27 with the image probability distribution W2, which he has in step S3 of 2 has determined for each considered projection image P. The result of step S27 corresponds to the resulting probability distribution W3r.

It is possible to determine the distance d Euclidean. Such a determination is useful especially if no information about the vascular system 5 as such. When the vascular system 5 however, the distance d is preferably to a path within the vasculature 5 based. This is schematically in 9 shown.

To determine the distance d, there are various possibilities in this case. It is preferred that the calculator 1 the distance d between two locations of the vascular system 5 determined that it first based on a first table, which is stored inside the computer, for each of the two locations a next node (exception) or two nearest nodes (rule) determined. Then the computer takes out 1 a second table, which is also stored computer-internal, the distances between the possible node combinations of each other. Furthermore, the computer determines the distances of the two locations from their nearest nodes. Based on the information now given, ie the distances of the two locations from their nearest nodes and the next node points from each other, the computer determines 1 the distance of the two places from each other. The nodal points or nodal point combinations contained in the tables comprise at least all branches of the vascular system 5 ,

Details of this procedure are explained in the older, on the filing date of the present invention not yet disclosed German patent application "Abstandbestimmungsverfahren" of the Applicant, which was filed on 07.05.2008 at the German Patent and Trademark Office and the official file number 10 2008 022 532.0 had received.

Different possibilities also exist for determining the two-dimensionally spatially resolved image probability W2. Preferably, the computer determines 1 according to 10 in a step S31 for a particular location of the viewed projection image P, those locations of the viewed projection image P lying in a predetermined evaluation kernel around the given location. For each of the locations located within the scoring core, the calculator determines 1 then, in a step S32, the influence of the image data values of the locations of the viewed projection image P lying in the evaluation core on the specific location. He takes into account a distance-dependent weighting function. In a step S33, the computer 1 finally on the basis of the previously determined influences for the specific location whose local image probability.

For example, it is possible that the calculator 1 to determine the image probability W2 the respective projection image P folds with a convolution kernel. The convolution kernel can be fixed in this case. Alternatively, he can by the user 8th be changeable, even interactively changeable. As a rule, the convolution kernel is a bell curve.

Preferably, however, step S33 is appropriate 11 designed. According to 11 the calculator determines 1 for each specific location of the projection image P under consideration, the influences of the locations lying in the evaluation core for this location on this location. As image probability for this place he then uses the maximum of the influences. This procedure is similar to a convolution. In contrast to a convolution, however, the individual influences are not summed up (or integrated) but only the maximum is utilized. This procedure is particularly advantageous because it is possible that different parts of the elongated structure 6 be imaged on the same location of the two-dimensionally spatially resolved projection image P.

The The present invention has many advantages. In particular, works They are robust and reliable. This is true despite the effects of noise and the artifacts that may be included in the projection images P. Furthermore, the method is real-time capable. In experiments 10 to 15 projection images P per second could be processed. Furthermore, despite using only the projection images P - ie even without biplane - a fast and correct convergence reachable. Any ambiguity arising from the depreciation the two-dimensionally spatially resolved projection images P resulting in the three-dimensional, can be resolved quickly become.

through the procedure of the invention is the Doctor significantly assisted in navigation. Because according to the invention, he can now directly in Navigate three-dimensional volume. He is not on a navigation limited in the two-dimensional projection images P. This makes it much more intuitive. This speeds up surgical procedures. Both the doctor and the other medical Staff and the patient are relieved (physically, mentally and also with regard to the radiation dose). Furthermore is for the inventive method no Special hardware needed. There are neither special catheters nor special sensors needed. The installation and use of the process do not require any additional costs the current approach. Furthermore, unlike others known methods not only the catheter tip shown. Much more The catheter is displayed over its entire length. There is no comparable method in the prior art for this purpose known.

The The above description is for explanation only of the present invention. The scope of the present invention On the other hand, it is intended solely by the attached Claims to be determined.

1

computer

2

computer program

3

disk

4

machine code

5

Object / vascular system

6

elongated structure

7

X-ray system

8th

user

9

vision device

10

Volume data set

a, b, c, x

Weighting factors

A, B, C, X

factors

d

distance

fB, fC

features

K

cost function

P, P '

projection images

S0

length

S1 to S33

steps

t

time

V

presence during

w

probabilities

w '

Transition probability

W2

Picture probability distribution

W3a

initial probability distribution

W3r

resulting probability distribution

W3v

provisional probability distribution

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• - US 6823204 B2 [0006]
• - US 6442235 B2 [0006]
• US 2008/243435 A1 [0007]
• US 2007/189457 A1 [0013, 0013, 0013, 0013]
• - DE 102008022532 [0089]

Cited non-patent literature

• - "3-D tracking of a catheter in the vascular system of a patient from 2-D X-ray images" by D. Gohlke, Friedrich Schiller University Jena and Siemens Medical Solutions (Forchheim), from the year 2006 [0014]
• "A Tutorial on Particle Filters for Online Nonlinear / Non-Gaussian Bayesian Tracking" by MS Arulampalam et al., IEEE Transactions on Signal Processing, Vol. 50, No. 2, 2002, pages 174 to 188 [0045]
• - MMD Powell, "An efficient method for finding the minimum of a function of several variables without generator", published in The Computer Journal 7 (2), 1964, pages 155-162 [0072]

Claims (14)

Evaluation method for a number of two-dimensionally spatially resolved projection images (P) of a spatially three-dimensional object ( 5 ), - whereby a computer ( 1 ) for each projection image (P) on the basis of a respective initial probability distribution (W3a) of an elongated structure given for the respective projection image (P) ( 6 ) in the spatially three-dimensional object ( 5 ) and the respective projection image (P) a respective resulting probability distribution (W3r) of the elongated structure ( 6 ) in the spatially three-dimensional object ( 5 ), wherein both the respective initial probability distributions (W3a) and the respective resulting probability distributions (W3r) are spatially resolved three-dimensionally, - the computer ( 1 ) for the projection images (P) based on the respective resulting probability distribution (W3r) spatially resolved three-dimensionally a respective presence profile (V) of the elongated structure ( 6 ) in the spatially three-dimensional object ( 5 ) and the respective determined presence history (V) in a representation of the spatially three-dimensional object ( 5 ), the computer ( 1 ) determines the respective attendance course (V) by maximizing a cost function (K) into which the length (S0) of the respectively determined presence course (V) and the probabilities (w) according to the respective resulting probability distribution (W3r) over the determined respective presence course ( V) and the curvatures of the respective determined course of attendance (V). Evaluation method according to claim 1, characterized in that the cost function (K) is the form K = aA + bB + cC + xX in which - K is the cost function, - a, b and c are non-zero weighting factors and x is another weighting factor, - A is an influencing factor dependent on the length (S0) of the respectively determined presence profile (V), - B is an influencing factor dependent on the probabilities (w) according to the respective resulting probability distribution (W3r) over the determined respective presence history (V), - C is an influencing factor dependent on the curvatures of the respectively determined presence profile (V), and - X is another influencing factor. Evaluation method according to claim 2, characterized that for determining the of the probabilities (w) according to the respective resulting probability distribution (W3r) depend on the determined respective course of attendance (V) Influence factor (P) the integral of a probability function (fB) over the determined respective attendance course (V), the probability function (fB) of the probability (w) of the respective place according to the depending on the resulting probability distribution (W3r). Evaluation method according to claim 2 or 3, characterized characterized in that for determining the of the curvatures dependent on the respective determined attendance course (V) Influence factor (C) the integral of a curvature function (fC) over the determined respective course of attendance (V), the curvature function (fC) of Amount of the curvature of the respective determined course of attendance (V) depends on the place. Evaluation method according to one of the above claims, characterized in that the computer ( 1 ) for each projection image (P) for determining the respective resulting probability distribution (W3r) in addition to the respective initial probability distribution (W3r) and the respective projection image (P) further information about possible locations of the elongated structure ( 6 ) in the spatially three-dimensional object ( 5 ) and / or via the locally three-dimensional object ( 5 ) considered. Evaluation method according to claim 5, characterized in that the object ( 5 ) a vascular system ( 5 ) and that the further information is a three-dimensionally spatially resolved course of the vascular system ( 5 ). Evaluation method according to one of the above claims, characterized in that the computer ( 1 ) for each projection image (P) for determining the respective resulting probability distribution (W3r) on the basis of the respective initial probability distribution (W3a) and a three-dimensionally defined transition probability (w ') determines a respective three-dimensionally spatially resolved provisional probability distribution (W3v) and the respective resulting probability distribution (W3r ) by weighting the probabilities according to the respective provisional probability distribution (W3v) with the corresponding probabilities according to a two-dimensionally spatially resolved image probability distribution (W2) which the computer ( 1 ) determined on the basis of the respective projection image (P). Evaluation method according to claim 7, characterized in that the three-dimensional spatially resolved transition probability (w ') is a monotone - especially strictly monotonic - falling function of the distance (d) of two places respectively. Evaluation method according to claim 6 and 8, characterized in that the distance (d) to a path within the vascular system ( 5 ) is related. Evaluation method according to one of the above claims, characterized in that the computer ( 1 for each projection image (P) within the framework of determining the respective two-dimensionally spatially resolved image probability distribution (W2), the respective image probability at a location of the respective projection image (P) on the basis of the image data values of the respective projection image (P) which are in an evaluation kernel around the respective location of the respective projection image respective projection image (P), and a distance-dependent weighting function determined. Evaluation method according to claim 10, characterized in that the computer ( 1 ) In the context of determining the image probability of a respective location, the influences of the locations lying in the evaluation core are determined on the respective location and the maximum likelihood of the influences is taken as the image probability of the respective location. Computer program, the machine code ( 4 ) generated by a computer ( 1 ) is immediately executable and its execution by the computer ( 1 ) causes the computer ( 1 ) carries out an evaluation method according to one of the above claims. Computer program according to claim, characterized in that it is stored on a data medium ( 3 ) is stored in machine readable form. Calculator that works with a computer program ( 2 ) is programmed according to claim 12, so that it performs an evaluation method according to one of claims 1 to 11 in operation.