Method for generating a combined projection image and imaging device

The present invention relates to a method for generating a combined projection image (SM) from a medical inspection object and an imaging device for generating a combined projection image (SM) from a medical inspection object.

Computed tomography (CT), usually X-ray computed tomography, is a widely-used medical diagnostic tool, which allows the generation of tomographic images or slices of an area of interest in a patient. Radiologists often prefer using thick slab or thick-slice volume for diagnostic and screening reading. For example, 5.0 mm slabs are often used for clinical reading and data storage in thoracic CT, although thin, high resolution 0.5 mm slabs are available and used for computer applications such as three-dimensional volume rendering and computer aided diagnosis algorithms.

It is relatively easy to compute 3.0-5.0 mm slabs in thoracic CT images without losing diagnostically relevant information, for example by applying methods such as Average Intensity Projection (AIP) or Maximum Intensity Projection (MIP). The slab thickness is related to the combined thickness of the original slices in the volume. Generally, a slab is rendered as a two-dimensional image, so that the image can be printed or viewed on a computer screen.

In Digital Breast Tomosynthesis (DBT) data volumes are often reconstructed at a higher resolution, for example at a resolution of 85 microns for a slice thickness of 1.0 mm. Such a high-resolution volume allows visualization of the fine clinical details required for accurate medical diagnosis of breast cancer, for example microcalcifications of only 100 microns in size, very fine spiculations of masses, etc. A correct depiction of microcalcification morphology and spiculations is critical for the radiologist in order to be able to differentiate between benign and malignant lesions in breast tissue. This means that the amount of data in DBT volumes is very large. For example, the total volume of data for a single patient (e.g. including two DBT views and projection images) can exceed one gigabyte. The large data volume complicates data transfer and increases radiologist workload.

Another reason for generating slabs from the initial data is that a radiologist often needs to evaluate the distribution of an entire cluster of calcifications embedded in the surrounding tissue, including any masses and architectural distortions in that region.

In breast tissue, a lesion can often extend over 10.0 mm or more, and can extend in any random direction. Therefore, in order to be able to perform such an evaluation, a radiologist should be provided with very thick slabs, i.e. slabs with a thickness in excess of 10.0 mm. In some cases, it might be desirable to perform such an evaluation over the total volume (essentially the whole volume collapsed into a single slab). In the context of DBT such a single slab has also been called synthetic mammogram (SM) due to its similarity to conventional 2D mammograms.

It is an object of the present invention to produce a projection image from a medical object with high sharpness and contrast.

In a first aspect this object is solved by a method for generating a combined projection image from a medical inspection object, comprising the steps of capturing a set of initial projection images; reconstructing a first and a second three-dimensional volume from the set of initial projection images; generating a first re-projection image from the first three-dimensional volume and a second re-projection image from the second three-dimensional volume; weighting the first re-projection image and the second re-projection image; and combining the weighted first re-projection image and second re-projection image for generating the combined projection image. By the method the quality of the combined projection image is improved and noise is reduced. Diagnostically relevant information can be highlighted while image information with less relevance is down-weighed.

In a preferred embodiment of the method the first and/or second re-projection image is generated under an orthographic projection geometry, cone beam projection geometry or fan beam projection geometry from the first and/or second three-dimensional volume. In this embodiment geometries are employed that increase sharpness of the combined projection image.

In a further preferred embodiment of the method the first re-projection image and second re-projection image are linearly combined for generating the combined projection image. In this embodiment the combined projection image can be generated with few computational steps.

In a further preferred embodiment of the method reconstructing the first and second three-dimensional volume is based on a filtered back-projection algorithm. In this embodiment the three-dimensional volumes can be generated efficiently.

In a further preferred embodiment of the method the first or/and second three-dimensional volume are reconstructed from a reduced subset of initial projection images. In this embodiment an accumulation of artifacts can be avoided.

In a further preferred embodiment of the method the reduced subset of initial projection images comprises a number of initial projection images from view positions adjacent to the re-projection view position. In this embodiment a blurring is avoided to a great extent.

In a further preferred embodiment of the method the first and/or second re-projection image is generated by a maximum intensity projection, an average intensity projection or a highest-interest-projection from the first and/or second three-dimensional volume. In this embodiment projections are used that produce projection images with high contrast.

In a further preferred embodiment of the method capturing a set of initial projection images is based on x-ray imaging from different angles on a circular trajectory. In this embodiment initial images are captured from which the three-dimensional volumes can be constructed efficiently.

In a further preferred embodiment of the method capturing a set of initial projection images is based on a tomosynthesis acquisition. In this embodiment three-dimensional volumes can be reconstructed easily.

In a further preferred embodiment of the method the tomosynthesis acquisition is performed by a computer tomography apparatus. In this embodiment is well suited for investigating medical inspection objects.

In a further preferred embodiment of the method each pixel in the first re-projection image and the second re-projection is weighted independently. In this embodiment contrast and sharpness can be increased to a large extend.

In a further preferred embodiment of the method weighting is performed by multiplying the first and/or the second re-projection image with a weighting factor. In this embodiment weighting can be performed quick and with low complexity.

In a further preferred embodiment of the method the initial projection images, the reconstructed volumes, the re-projection images and/or the combined projection image are noise filtered. In this embodiment the quality of the resulting combined projection image is improved.

In a second aspect this object is solved by a method for generating a combined projection image from a medical inspection object, comprising the steps of capturing a set of initial projection images; reconstructing a three-dimensional volume from the set of initial projection images; generating a first re-projection image and a second re-projection image from the three-dimensional volume; weighting the first re-projection image and the second re-projection image; and combining the weighted first re-projection image and second re-projection image for generating the combined projection image. In this embodiment same advantages as in the method according to the first aspect are achieved.

In a third aspect this object is solved by an imaging device for generating a combined projection image from a medical inspection object, comprising a capturing means for capturing a set of initial projection images; a reconstruction means for reconstructing a first and a second three-dimensional volume from the set of initial projection images; a generation means for generating a first re-projection image from the first three-dimensional volume and a second re-projection image from the second three-dimensional volume; a weighting means for weighting the first re-projection image and the second re-projection image; and a combination means for combining the weighted first re-projection image and second re-projection image for generating the combined projection image. In this embodiment same advantages as in the method according to the first aspect are achieved.

Other features which are considered as characteristic for the invention are set forth in the appended claims.

Although the invention is illustrated and described herein as embodied in method for generating a combined projection image, it is nevertheless not intended to be limited to the details shown, since various modifications and structural changes may be made therein without departing from the spirit of the invention and within the scope and range of equivalents of the claims.

The construction and method of operation of the invention, however, together with additional objects and advantages thereof will be best understood from the following description of specific embodiments when read in connection with the accompanying drawings.

FIG. 1 shows a schematic view of a tomosynthesis acquisition;

FIG. 2 shows a multiple projection synthetic mammogram example;

FIG. 3 shows a diagram of method steps;

FIG. 4 shows an imaging device;

FIG. 5 shows a principle of multiple projection synthetic mammogram;

FIG. 6 shows AIP projections of a reconstructed volume;

FIG. 7 shows MIP projections of a reconstructed volume; and

FIG. 8 shows a synthetic mammogram.

FIG. 1 shows a schematic view of a tomosynthesis acquisition. The medical inspection object 101 (compressed breast) is imaged with x-ray from different angles on a circular trajectory. The breast is compressed by a compression paddle 107 . X-Ray projection images of the medical inspection object 101 are captured by the x-ray detector 105 .

The X-ray source 103 is moved on a circular path around the medical inspection object 101 in order to capture initial projection images PI from different projection directions by the x-ray detector 105 . The x-ray detector 105 is positioned on the opposite side of the circle from the X-ray source 103 .

X-ray computed tomography (x-ray CT) is a technology that uses computer-processed x-rays to produce tomographic images (virtual ‘slices’) of specific areas of the scanned object, allowing a user to see what is inside it without cutting it open.

Digital geometry processing is used to generate a three-dimensional volume of the inside of the medical inspection object 105 from a large series of two-dimensional radiographic initial projection images PI captured around a single axis of rotation. Medical imaging is a common application of x-ray CT. Its cross-sectional images are used for diagnostic and therapeutic purposes in various medical disciplines.

FIG. 2 shows a multiple projection synthetic mammogram example as a combined projection image SM. The three-dimensional volumes RV 1 and RV 2 are reconstructed from the initial projections images PI in this example, e.g. by means of a microprocessor. While both reconstructed three-dimensional volumes RV 1 and RV 2 are done with the same kind of filtered back-projection (FBP) algorithm, the reconstructed volume RV 1 is based on 25 projections and the reconstructed volume RV 2 is based on a subset of only 3 central projections. The central projection directions as a subset from all projection directions are drawn with solid lines in FIG. 1 . While the reconstructed volume RV 2 from the central projections is expected to exhibit more noise and more severe out-of-plane artifacts, the angles at which these artifacts occur are narrower. This causes less blurring from the out-of-plane artifacts, when re-projecting these volumes.

The processed initial tomosynthesis projection images are transformed in a single re-projected image RP on the basis of single reconstructed volume RV.

FIG. 3 shows a diagram of method steps for generating a combined projection image SM from a medical inspection object 101 . In step S 101 a set of initial projection images PI is captured by the tomosynthesis acquisition. Capturing is can be done for example with a solid state X-ray detector.

In step S 102 at least a first and a second three-dimensional volume RV 1 , RV 2 are reconstructed from the set of initial projection images PI, e.g. by a means of a computer algorithm. Reconstructing essentially involves solving the inverse Radon transformation on the basis of the captured initial projection images.

Then in step S 103 at least a first re-projection image RP 1 is generated from the first three-dimensional volume RV 1 and at least a second re-projection image RP 2 is generated from the second three-dimensional volume RV 2 . Generating re-projection images can be performed by summing pixels on a virtual projection beam considering a particular geometry, like a cone beam.

In step S 104 the first re-projection image RP 1 and the second re-projection image RP 2 are weighted. Weighting can be done either by independently weighting each pixel in the re-projection images RP 1 and RP 2 or by multiplying each of the entire re-projection images RP 1 and RP 2 with a weighting factor.

Finally in step S 105 the weighted first re-projection image RP 1 and second re-projection image RP 2 is combined, e.g. added or subtracted, for generating the combined projection image (SM).

FIG. 4 shows an imaging device 200 for generating the combined projection image SM from the medical inspection object 101 . The imaging device 200 is for example a computer tomography apparatus. The imaging device comprises a capturing means 201 for capturing a set of initial projection images PI, e.g. an x-ray detector 105 . A reconstruction means 203 serves for reconstructing at least a first and at least a second three-dimensional volume RV 1 , RV 2 from the set of initial projection images PI by a mathematical calculation. A generation means 205 serves for generating at least a first re-projection image RP 1 from the first three-dimensional volume RV 1 and at least a second re-projection image RP 2 from the second three-dimensional volume RV 2 by a calculated re-projection. A weighting means 207 for weighting the first re-projection image RP 1 and the second re-projection image RP 2 . Finally the imaging device 200 comprises a combination means 209 for combining the weighted first re-projection image RP 1 and second re-projection image RP 2 for generating the combined projection image SM.

The reconstruction means 203 , the generation means 205 , the weighting means 207 and the combination means 209 can be implemented by a computer with a microprocessor and a data storage that is capable of processing the captured initial projection images PI.

FIG. 5 shows a principle of a multiple projection synthetic mammogram as a combined projection image. PI denotes initial projection images, RV denotes reconstructed volumes, RP denotes re-projection images, and SM denotes a combined projection image, like the synthetic mammogram.

From the initial projection images PI captured by the detector 105 n three-dimensional volumes RV 1 . . . RV n are reconstructed. Thus n reconstructed volumes RV 1 . . . RV n with n>0 are given. From the n th reconstructed volume RV n m two-dimensional projection images RP mn on the same virtual detector are computed. The projections of the projection images RP mn may vary in type, e.g., Maximum-Intensity-Projection (MIP), Average-Intensity-Projection (AIP) and Highest-Interest-Projection (HIP) as well as projection geometry, e.g. orthographic, cone beam or fan beam.

The synthetic mammogram SM as combined projection image is computed as linear combination of the re-projection images RP mn , for example by:

SM( u,v )=Σ n=1 N Σ m=1 M n w ( m,n,u,v )·RP mn ( u,v ),  (1)

where N>0, Σ n=1 N M n >1 and w(m,n,u,v)≧0.

This produces a weighted linear combination of multiple re-projection images RP, in which the weights can be chosen independently for each pixel in each projection image. The re-projection images RP can be obtained from one or more reconstructed three-dimensional volumes RV.

Instead of the linear combination described in Eq. (1), other ways of combining/mixing of re-projection images are conceivable. In addition filtering of the initial projection images PI, the reconstructed volumes RV, the re-projection images RP and/or the combined projection image SM can improve the final result.

Multiple projections and/or multiple reconstructions can be used to synthesize a 2D mammogram as a combined projection image. In particular, advantages from different projections and reconstructions can be exploited to generate a synthetic mammogram that highlights diagnostically relevant features. Reconstructing the three dimensional volume RV from a subset of initial projections PI avoids accumulating artifacts when re-projecting.

FIG. 6 shows examples for Average Intensity Projections (AIPs) from a reconstructed volume RV which has been reconstructed using 1, 3, 5 and all 25 projections respectively (from left to right). The medical inspection object 101 is a human breast.

While the contrast-to-noise ratio improves with the number of projection images PI, structures tend to get more and more blurred. While the image noise decreases from left to right, the images also get increasingly blurred.

A similar tendency can be seen for the Maximum-Intensity-Projections (MIPs) in FIG. 5 . Due to the different noise behavior of AIPs and MIPs, however, for a MIP preferably a volume reconstructed with all projections is used, while for the AIP, a volume reconstructed with a subset of less projections, for example 3, is preferred.

The different kind of projections, such as MIP and AIP but also different Highest-Interest-Projections (HIP) are well-suited for highlighting particular diagnostically relevant structures. In particular, these projections exhibit differences regarding contrast-to-noise ratio and spatial noise homogeneity.

However, different three-dimensional reconstructions RV with varying contrast, noise and artifact-behavior can also be constructed from the initial projection images PI. Differences may come from using different reconstruction algorithms (e.g. FBP—filtered back projection or IR—iterative reconstruction) or from different parameterizations (e.g. usage of different backprojection filters in FBP or different regularizers in IR).

FIG. 7 shows MIPs of a volume with 1, 3, 5, and 25 projections (from left to right). While the image noise decreases from left to right, the images also get increasingly blurred.

FIG. 8 shows a synthetic mammogram on the basis of a weighted MIP (weighting factor: 1) from a reconstructed volume RV with 25 projections and a weighted AIP (weighting facto. 0.75) from a reconstructed volume RV with 3 projections. Mixing the re-projection images RP leads to sharper combined projection images SM with more details.

The scope of the invention is defined by the claims and is not restricted by special features discussed in the description or shown in the figures. All features discussed with respect to different embodiments can be combined variously and independently in order to simultaneously realize their technical effect.

The following is a summary list of reference numerals and the corresponding structure used in the above description of the invention:

101 medical inspection object 103 x-ray source 105 detector 107 compression paddle 200 imaging device 201 capturing means 203 reconstruction means 205 generation means 207 weighting means 209 combination means PI initial projection images RV reconstructed three-dimensional volume RP re-projection image SM combined projection image/synthetic mammogram