Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
  • H04N1/40—Picture signal circuits
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T3/00—Geometric image transformation in the plane of the image
  • G06T3/40—Scaling the whole image or part thereof
  • G06T3/4084—Transform-based scaling, e.g. FFT domain scaling
• • G06T5/73—
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/30—Subject of image; Context of image processing
  • G06T2207/30004—Biomedical image processing
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N2201/00—Indexing scheme relating to scanning, transmission or reproduction of documents or the like, and to details thereof
  • H04N2201/04—Scanning arrangements
  • H04N2201/0402—Arrangements not specific to a particular one of the scanning methods covered by groups H04N1/04 - H04N1/207
  • H04N2201/0458—Additional arrangements for improving or optimising scanning resolution or quality

Definitions

• the present invention relates in general to digital image processing. Specifically, the present invention relates to methods for improving image resolution and improving image data transmission in medical imaging, movies and video games, teleimage, and any other areas involving digital image acquisition and transmission. More specifically, the present invention provides methods for increasing image resolution - spatially and/or temporally - and methods for transmitting image data using back pixelation, a technique that involves data processing and reconstruction of overlaid images from multiple acquisitions or multiple sampling.
• MRI Magnetic resonance imaging
• MR magnetic resonance
• the pulse sequence determines the image contrast, the speed of image acquisition, and the spatial resolution of the resulting MR image.
• the maximal spatial and temporal resolutions are limited by the strength of the magnetic field and the scanner's gradient hardware.
• High resolution scanning is desired for better morphologic depiction and lesion characterization.
• Low spatial resolution on one or more dimensions may cause a partial volume effect that results in poor differentiation of structures. That is, clinically relevant information may be masked if a lesion is mixed with other tissue(s) in the space represented by a single pixel or voxel.
• Schreiner, S.; et al Journal of Computer Assisted Tomography 20(l):56-67.
• the pixel or voxel size needs to be no larger than half the size of the lesion. See, Id. High spatial resolution for acquiring images is thus vital in these situations.
• Clinical imaging such as MRI also requires a balance between temporal resolution and spatial resolution, the later represented by the pixel or voxel size and the former the scan time.
• Fast acquisition or high temporal resolution is important for reducing the length of the exam time and the overall exam cost; and more importantly, it is critical in assessing dynamic changes and monitoring structures in real-time imaging, especially during the use of contrast agents for tissue and organ perfusion imaging.
• fast imaging is often performed at the expense of spatial resolution. That is, for example, imaging a fixed field of view (FON) with a l28 x l28 matrix may be performed more quickly than with a 256 x 256 matrix, but the resulting lower spatial resolution information of the 128 x 128 acquisition may be insufficient for confident diagnosis of structures.
• FON field of view
• the 256 x 256 acquisition - or the spatial resolution achieved thereby - may be required for structural identification and depiction. Yet, a 256 x 256 image requires a longer acquisition period, yielding a lower temporal resolution, and hence its diagnostic utility for dynamic or real time imaging may be limited, e.g., in the cases where contrast media enhancement is evaluated for medical diagnosis.
• S/N signal-to-noise ratio
• the spatial resolution achieved may be sufficient for depiction of a target structure but the S/N of the relatively small voxel or pixel size may be insufficient for actual visualization of the structures.
• a 128 x 128 matrix (with a larger pixel or voxel size) may be performed quicker and yields a higher S/N per pixel or per voxel.
• high spatial resolution, high temporal resolution, and high S/N are all preferable for MRI; however, they often represent competing factors in image acquisition that call for appropriate balancing.
• S Ns may be improved by a variety of methods such as increasing the number of excitation averages, decreasing the receiver bandwidth, or increasing the acquisition repetition time. These methods improve S/N but slow image acquisition, thus result in decreased temporal resolution. On the other hand, S/N may be compromised by decreasing pixel or voxel size, which marks improved spatial resolution.
• Certain post-processing techniques such as zero-filled interpolation and voxel shift interpolation, have been proposed to address spatial resolution related partial volume effect problem in some applications. See, e.g., Du Y. et al., JMRI September/October 1994 p. 733-741.
• these filtering methods do not inherently improve spatial resolution; the method of Du et al., for example, only interpolates intermediate voxels.
• These methods can not, nor can other existing image processing or restoration methods, accurately determine a signal that would be returned from a portion of tissues with a size smaller than one pixel or one voxel.
• the size of a digital image in application areas such as medical imaging is typically fairly large such that the transmission of such images poses a significant challenge on transmission capacity.
• To achieve satisfactory resolution after transmission is a lasting battle for the engineers and researchers alike.
• Certain data segmentation, compression, or reduction techniques have been used (e.g., preserving every fourth point of the image). But, they may result in voids in the transmitted image and therefore may not be desirable in some situations where high image integrity is required.
• a method for improving resolution of a two-dimensional image capable of being acquired from an object at a first pixel size comprising: defining a sampling region for producing the image at a second pixel size, the second pixel size being no larger than the first pixel size; sampling, with an overlapping increment, the sampling region thereby producing a multiplicity of sample layers, the overlapping increment being no larger than the first pixel size thereby determining the second pixel size; obtaining values of the pixels of the second pixel size in each of the multiplicity of sample layers; computing the value of each pixel of the second pixel size from the values of the corresponding pixels in the multiplicity of the sample layers, thereby producing an intermediate image; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing the image at the second pixel size, the filter function being capable of representing artifacts from the multiple sampling.
• a method for improving resolution of a three-dimensional image capable of being acquired from an object at a first voxel size comprising: defining a sampling space for producing the image at a second voxel size, the second voxel size being no larger than the first voxel size; sampling, with an overlapping increment, the sampling space thereby producing a multiplicity of sample areas, the overlapping increment being no larger than the first voxel size thereby determining the second voxel size; obtaining values of the voxels of the second voxel size in each of the multiplicity of sample areas; computing the value of each voxel of the second voxel size from the values of the corresponding voxels in the multiplicity of the sample areas, thereby producing an intermediate image; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing the image at the second voxel size, the filter function being
• the second pixel size or the second voxel size is no larger than, respectively, the first pixel size or the first voxel size thereby determining, respectively, the second pixel size or the second voxel size.
• the second pixel size or the second voxel size is equal to, respectively, the overlapping increment.
• the second pixel size or the second voxel size is equal to, respectively, the first pixel size or the first voxel size divided by the multiplicity.
• the filter function is a point response function.
• the point response function is defined by a multiple sampling factor, which is equal to the multiplicity.
• the sampling is performed with an overlapping increment having a fixed or varied size.
• the overlapping increment is taken along one or more dimensions.
• the overlapping increment is taken equally on one or more dimensions; in other embodiments, the overlapping increment is taken unequally on one or more dimensions.
• the overlapping increment is taken angularly, whether equally or unequally on one or more dimensions.
• the overlapping increment is defined in terms of any kind of reference coordinates, such as Cartesian coordinates or polar coordinates.
• a method for improving resolution of a two-dimensional image of an object from a plurality of images taken from the object at a first pixel size, wherein each image in the plurality is capable of overlaying one other images in the plurality at an increment, wherein the increment is no larger than the first pixel size comprising: defining a region for producing the two-dimensional image at a second pixel size, the second pixel size being determined by the increment; obtaining values of the pixels of the second pixel size in each image in the plurality; computing the value of each pixel of the second pixel size from the values of the corresponding pixels in the plurality, thereby producing an intermediate image; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing the two-dimensional image at the second pixel size, the filter function being capable of representing artifacts from the multiple overlaying.
• a method for improving resolution of a three- dimensional image of an object from a plurality of images taken from the object at a first voxel size, wherein each image in the plurality is capable of overlaying one other images in the plurality at an increment, wherein the increment is no larger than the first voxel size comprising: defining an area for producing the three-dimensional image at a second voxel size, the second voxel size being determined by the increment; obtaining values of the voxels of the second voxel size in each image in the plurality; computing the value of each voxel of the second voxel size from the values of the corresponding voxels in the plurality, thereby producing an intermediate image; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing the three-dimensional image at the second voxel size, the filter function being capable of representing artifacts from
• each image in the plurality overlays one other image in the plurality at an increment having a varied size and the second pixel size or the second voxel size is equal to, respectively, the smallest of the increment.
• the increment has a fixed size and the second pixel size or the second voxel size is equal to the respective increment.
• the increment is defined in terms of any kind of reference coordinates, such as Cartesian coordinates or polar coordinates
• the filter function is a point response function.
• the point response function is defined by a multiple overlaying factor, which is equal to the plurality.
• the overlaying is performed along one or more dimensions. According to still another embodiment, the overlaying is performed equally on one or more dimensions. According to a further embodiment, the overlaying is performed unequally on one or more dimensions. In another embodiment, the overlaying is performed angularly, whether equally or unequally on one or more dimensions.
• a method for producing a two-dimensional image of an object from multiple acquisitions using a two-dimensional acquisition matrix, wherein the acquisition matrix defines the resulting pixel size of the two- dimensional image comprising: acquiring, one at a time, a multiplicity of images from the object by shifting one or more units in the acquisition matrix; obtaining values of the pixels in each image in the multiplicity; computing the value of each pixel from the values of the corresponding pixels in the multiplicity, thereby producing an intermediate image; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing the two-dimensional image, the filter function being capable of representing artifacts from multiple acquisitions.
• the method comprises: acquiring, one at a time, a multiplicity of images from the object by shifting one or more units in the acquisition matrix; transmitting, one at a time, the multiplicity of images; obtaining values of the pixels in each transmitted image in the multiplicity; computing the value of each pixel from the values of the corresponding pixels in the multiplicity, thereby producing an intermediate image; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing the two-dimensional image, the filter function being capable of representing artifacts from multiple acquisitions.
• a method for producing a three-dimensional image of an object from multiple acquisitions using a three-dimensional acquisition matrix, wherein the acquisition matrix defines the resulting voxel size of the three- dimensional image comprising: acquiring, one at a time, a multiplicity of images from the object by shifting one or more units in the acquisition matrix; obtaining values of the voxels in each image in the multiplicity; computing the value of each voxel from the values of the corresponding voxels in the multiplicity, thereby producing an intermediate image; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing the three-dimensional image, the filter function being capable of representing artifacts from multiple acquisitions.
• a method for producing and transmitting a three- dimensional image of an object from multiple acquisitions using a three- dimensional acquisition matrix, wherein the acquisition matrix defines the resulting voxel size of the three-dimensional image is provided.
• the method comprises: acquiring, one at a time, a multiplicity of images from the object by shifting one or more units in the acquisition matrix; transmitting, one at a time, the multiplicity of images; obtaining values of the voxels in each transmitted image in the multiplicity; computing the value of each voxel from the values of the corresponding voxels in the multiplicity, thereby producing an intermediate image; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing the two-dimensional image, the filter function being capable of representing artifacts from multiple acquisitions.
• the resulting pixel size of the two-dimensional image or the resulting voxel size of the three- dimensional image is equal to the size of one or more units of the respective acquisition matrix.
• the acquisition matrix comprises equally-spaced units. According to yet another embodiment, the acquisition matrix comprises unequally-spaced units. In various embodiments, the acquisition matrix is defined in terms of any kind of reference coordinates, such as Cartesian coordinates or polar coordinates.
• the multiplicity of images are acquired at defined time points over a period of time. According to a still further embodiment, one or more preliminary test acquisitions for determining the scheduling of the time points are performed. According to another embodiment, the multiplicity of images are acquired, one at a time, by shifting a decimal number of units in the acquisition matrix.
• the filter function is a point response function.
• the point response function is defined by a multiple acquisition factor, which is equal to the multiplicity.
• the acquisition matrix is adaptively determined by adjusting the unit size or scheme for the subsequent acquisitions based on assessment of variance for one or more comparators (factors or parameters to compare) for at least two prior acquisitions.
• the one or more comparators are selected from the group consisting of pixel or voxel data, k-space data, phase data, and signal- to-noise ratio data.
• the multiple acquisitions further comprises one or more preliminary test acquisitions for determining the unit size and scheme of the acquisition matrix.
• the method comprises: decomposing the image into a plurality of images taken from the object, each image in the plurality being capable of overlaying one other images in the plurality at an increment; transmitting, one at a time, at least two images in the plurality; computing the value of each pixel of an intermediate image from the values of the corresponding pixels in the transmitted images in the plurality; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing a resultant two-dimensional image having a second pixel size, the filter function being capable of representing artifacts from the multiple overlaying.
• a method for transmitting a three-dimensional image of an object comprising: decomposing the image into a plurality of images taken from the object, each image in the plurality being capable of overlaying one other images in the plurality at an increment; transmitting, one at a time, at least two images in the plurality; computing the value of each voxel of an intermediate image from the values of the corresponding voxels in the transmitted images in the plurality; and dividing, in the Fourier domain, the intermediate image by a filtering function thereby producing a resultant two-dimensional image having a second voxel size, the filter function being capable of representing artifacts from the multiple overlaying.
• the second pixel size is equal to the first pixel size and the second voxel size is equal to said first voxel size.
• all images in the plurality is transmitted.
• the computing is performed by taking arithmetic means in some embodiments. In other embodiments, the computing is performed by a heuristic function or an arithmetic function according to suitable mathematics and artificial intelligence procedures.
• the two-dimensional image or the three- dimensional image of this invention is selected from the group consisting of a magnetic resonance image, an image produced by a digital scanner, an image produced by a digital camera or digital video, a pathological image, a histological image, and a radiological image.
• Fig. 1 depicts the initial resolution level of an image from multiple sampling in a computer simulation of image reconstruction using the back pixelation technique.
• the multiple sampling factor is equal to eight.
• Fig. 2 depicts, in the same computer simulation as that of Fig. 1, the resulting intermediate image after image data reconstruction.
• Fig. 3 depicts, in the same computer simulation as that of Fig. 1 and 2, the final reconstructed image after removing multi-sampling artifacts.
• Fig. 4 shows the original image with a size of 256x256, which was subjected to the computer simulation depicted in Fig. 1, 2, and 3.
• a digital image may be a two-dimensional image or a three-dimensional image; it may be captured or acquired by a digital device such as a digital camera or video, a digital scanner, a MRI scanner, or transformed from a conventional image through digitization process.
• a digital device such as a digital camera or video, a digital scanner, a MRI scanner, or transformed from a conventional image through digitization process.
• pixel and "voxel” are interchangeable, both referring to the discrete unit or element of a digital image that bears image data; the former is applicable when two-dimensional images are concerned whereas the later is applicable when three-dimensional images are concerned.
• pixelation refers to the summation of all data points from an object within a predetermined region or space; such summation forms a basis for a two- or three-dimensional representation of such data.
• back pixelation is defined herein to represent a process or technique whereby image acquisition is performed by multiple layered sampling which results in a multiplicity of images that are capable of overlaying one another by an increment and whereby the final image is derived by consolidating the multiplicity of images using mathematical computations and artificial intelligence procedures (e.g., taking arithmetic means) which derive the final pixel/voxel values as a function of the corresponding pixels or voxels in the multiplicity of images.
• the incremented multiple sampling and overlaying of the multiplicity of images are therefore deterministic of the quality of the resulting final image. That is, the scheduling and organization of the multiple sampling and multiple overlaying, the size and variation of the overlaying increment dictate the resultant pixel or voxel size of the final image and ultimately, the amount and reliability of the spatial and temporal information captured therein.
• the multiplicity is also referred to, interchangeably, as a multiple sampling factor, a multiple overlaying factor, a multiple acquisition factor, or an over-sampling factor; this number is equal to the number of times sampling (or overlaying, image acquisition) is performed and the number of images produced during sampling.
• the increment refers to a shift in position in the two-dimensional or three-dimensional space, depending upon the dimensionality of the image being processed.
• the increment can be taken on one or more dimensions or axis (x, y, or z axis); the size of the increment may be fixed or may vary during the multiple sampling and overlaying.
• the increments may be taken angularly, whether equally or unequally on one or more dimensions.
• the overlapping increment may be defined in terms of Cartesian coordinates or polar coordinates. The use of angular increments and/or polar coordinates may be advantageous for certain imaging applications, e.g., spiral MRI or edge detection of certain target structures.
• the back pixelation process may be applied prospectively or retrospectively. For example, given a set of pre-acquired images of an object, whose relevant positions are such that they can overlay one another by an increment, a new image of the object may be constructed by overlaying the set of pre-acquired images and taking arithmetic means thereof (or performing any other suitable mathematical computations or artificial intelligence procedures according to this invention), an application of retrospective back pixelation.
• an acquisition matrix may be designed for imaging an object of interest; multiple acquisition may be performed according to the acquisition matrix, in the same manner as multiple sampling, thereby producing a multiplicity of images.
• the final image can then be constructed by overlaying the multiplicity of images and taking arithmetic means thereof (or performing any other suitable mathematical computations or artificial intelligence procedures according to this invention), an application of prospective back pixelation.
• arithmetic means thereof or performing any other suitable mathematical computations or artificial intelligence procedures according to this invention.
• the method of this invention improves, in one embodiment, the spatial resolution of an image. That is, it reduces the pixel or voxel size; it does so not by adjusting or enhancing hardware capacities of image acquisition instruments, but by applying the back pixelation technique.
• the initial step of back pixelation is the multiple sampling, as discussed supra.
• the multiple sampling is performed with an increment, preferably an increment smaller than a pixel or voxel, thereby producing a multiplicity of images, which, together, capture the feature information contained in a unit with a size of the increment which is smaller than the pixel or voxel.
• the resulting image manifests the feature information of the small unit and hence achieves a smaller pixel or voxel size.
• this image is ordinarily obscured by artifacts - similar to Gaussian blurs - from the overlaying reconstruction; it is an intermediate image that needs to be subject to further processing according to the present invention.
• a final reconstruction or filtering step is performed to remove the artifacts.
• the artifacts may be defined as a point response function which is used as a filtering mask.
• the final image is derived by dividing, in the Fourier domain, the intermediate image by the filtering mask.
• the region R is sampled in such a way that the image is segmented into pixels of real size.
• the value of each pixel is the average value of all the points that are contained within the region defined by the pixel.
• a square pixel of size 1 is used.
• the upper left corner of the first pixel is at the origin and all of the other pixels follow in columns and rows to sample the entire region R without overlapping.
• S(0,0) is represented as a two dimensional array of pixel values and would be the standard output of a typical MR image.
• the sample region is defined by the location of its top left comer relative to the origin; in this case the top left corner is located precisely at the origin.
• the resolution of the sampling region In order to accurately depict the position of a point of interest within R that is smaller than an individual pixel, the resolution of the sampling region must be increased. However, the pixel size 1 has a lower limit in MRI and therefore the resolution has an upper limit.
• I this point of interest be called I.
• the size of I is some square with a size smaller than 1 so that the length of a side of I is 2*l/n, where n is an over-sampling factor that is to be used to generate a system of pixel arrays from which to generate the region R with sufficient resolution to accurately display I.
• the factor n is determined by the minimum resolution needed to display I with a reasonable accuracy level; n determines, in turn, the number of pixel arrays that must be gathered - over- sampled - along each dimension.
• the process of pixelation is repeated in order to build a system of pixel arrays that can be used to calculate the values of the pixels in the resulting image.
• This process is also referred to as multiple sampling or over-sampling.
• this resulting image be represented as a region F that is the same size as R.
• the resolution of F must be high enough to accurately depict the point of interest I.
• an individual pixel in F is less than half the size of I. See, e.g., Kamen EW and Heck BS; 1997 Prentice Hall; New Jersey. Therefore, the size of an individual pixel in F is (1/2 * 2(l/n)) or (1/n).
• the pixelation process is repeated on region R with the origin shifted by an increment that is equal to the size of a pixel in F - (1/n) - so that every pixel in F that would be contained in the region defined by the first pixel in S(0,0) becomes the origin for one of the sampling arrays in the system of pixel arrays.
• the pixelation process is repeated to obtain a sampling array S(j, k).
• the system of pixel arrays is termed the sampling region or space S and an individual sampling array is termed a layer.
• sampling or acquisitions may be performed (i) from a given origin, at a predetermined resolution, and by a random increment along orthogonal dimensions; (ii) from the same origin but at a varied resolution; (iii) at the same resolution but with rotated frames or matrixes; or any combinations of (i), (ii), and (iii).
• suitable for the methods of the present invention is any process by which a multiplicity of sampling layers may be produced from imaging an object where each layer consists of more than redundant image information vis-a-vis the rest of the layers; to wit, no single image layer may be deduced from the rest of the layers.
• the value of any pixel in F is determined as the average value (the arithmetic mean) of the corresponding pixel in all layers of S. In practice, this is computed as the summation of all layers of S projected into the region F and divided by the number of layers.
• Taking arithmetic means is an approach of consolidating multi-sampled layers of S in certain embodiments of this invention. In other embodiments, various other approaches may be used, including, e.g., taking harmonic means, geometric means, or simple summations, or applying any other mathematical computations or artificial intelligent procedures by which a single value is deduced as a function (e.g., an arithmetical function or a heuristic function) of the corresponding pixel values in each layer of S. This process yields a final region F that contains all of the information in S.
• the reconstruction artifact can be defined as the response of the forgoing sampling, overlaying, and consolidating process to a single point in the original region R.
• DSP digital signal processing
• it is an impulse response function that can also operate as a filtering mask.
• the point response function is a square region with the highest intensity in the center and the lowest intensity on the corners.
• the size of the region is one pixel in F smaller than two pixels in S, or (2n-l) pixels in F.
• the relative intensities of the point response may vary, as the artifacts are similarly defined by the response of this sampling, overlaying, and consolidating process on a single point within one pixel in F. If the impulse point used to generate this function is a unit impulse with intensity of one, the impulse response function for an over-sampling factor of three, m(3), would be as follows.
• I is the reconstructed image space obscured by the artifact and (w,z) designates that the function is represented in Fourier space.
• This division in Fourier space by the impulse response function concludes the back pixelation processing and produces a final image that has desired higher resolution in F.
• the aforementioned steps may be similarly applied to three- dimensional images, an example of which is set forth infra in Example 4.
• the back pixelation process may be applied to reduce pixel or voxel size.
• Reconstructing image data from over-sampled layers permits recapture of the image information contained in a region that is smaller than the original pixel or voxel in size.
• the process may be employed using the same field of view (FOV) but a different matrix size; a different FOV but the same matrix; or a different FOV and a different matrix size. It is important that the pixels or voxels are not identically super-imposed such that redundant image data is gathered by multiple sampling. That is, the multiple sampling and overlaying is with a shift and/or by an increment, in order to allow different spatial data to be captured by each sampling or acquisition.
• the increment remain constant; in other embodiment, the increment varies throughout the sampling process. In a preferred embodiment, the increment has a size that is smaller than the size of the original pixel or voxel.
• the increment may be taken, in various embodiments, on one or more dimensions; i.e., along the x axis, y axis, and/or z axis.
• the relative positions of the over-sampled image layers i.e., the organization and/or scheduling of the increments provides a linkage between the resulting image and the matrix position.
• the pixelation process may be applied both retrospectively and prospectively.
• the back pixelation reconstruction may be performed on these images as discussed supra, by taking arithmetic means therefrom (or performing any other suitable mathematical computations and artificial intelligence procedures according to this invention), and removing artifacts using a filtering mask.
• the final image produced would therefore have a higher resolution or lower pixel/voxel size, owing to the size of the increment and the position shift of the matrix.
• the final image would be constructed from multiple overlaid images acquired in each acquisition - whose acquisition resolution may be lower than the desired resolution of the final image - by taking arithmetic means therefrom (or performing any other mathematical computation and artificial intelligence procedures according to this invention) and removing artifacts using a filtering mask. Because the acquisition matrix is pre-designed, the pattern of the multiple overlaying or the super-imposition is predetermined.
• the matrix consists of even-sized or equally-spaced units; in other embodiment, the matrix consists of one or more uneven-sized or unequally-spaced units.
• the unit size of the acquisition matrix is smaller than the size of the pixel or voxel used during the multiple acquisitions.
• the pixel or voxel size of the final image may be computed based on the unit size and the pattern of the acquisition matrix. In some embodiments, the pixel or voxel size of the final image is equal to the unit size of the acquisition matrix, an integral number of times of the unit size, or a decimal number of times of the acquisition matrix.
• the prospective determination of the acquisition matrix - the size of the overlapping increment (unit size of the matrix) and the overlapping scheme (scheme of the matrix) - may be adaptive in alternative embodiments of this invention. That is, the matrix may be determined in real time using artificial intelligence selection based on degree changes between image data sets. For example, suppose the initial overlapping increment of the second acquisition is 10% of the pixel size on a dimension, the resultant digital image data set from the second acquisition may be assessed for variance from the initial digital image data set based on any number of comparators (factors or parameters to be compared), such as raw pixel or voxel data, phase shift information (phase data), k-space data, S/N per pixel, or voxel location.
• comparators factors or parameters to be compared
• the subsequent (e.g., the third) acquisition(s) may be augmented by applying a greater overlapping increment or a greater shift (i.e. >10%) or, by adjusting to an alternative matrix or overlapping scheme that permits effective procurement of new (different to the earlier acquisitions) image data, as assessed by a significant variance.
• This adaptive acquisition process thus allows one to achieve an optimal spatial resolution by advantageously taking efficient use of time during acquisitions.
• the prospective and retrospective applications of the back pixelation technique may be carried out in any reference coordinates systems, such as a Cartesian coordinates system or a polar coordinate system in various embodiments of this invention. That is, the increment or overlapping increment and the acquisition matrix may be defined in terms of any reference coordinates, such as Cartesian coordinates or polar coordinates.
• these prospective and retrospective back pixelation methods may be employed on any digital images, which includes, among other things, images produced by a digital scanner, images produced by a digital camera or digital video, and medical images such as MR images, pathological images, histological images, and radiological images. These methods are thus particular useful in medical imaging, movies and video games, teleimage, and any other areas involving the use of digital images.
• the back pixelation technique also enables improvements on temporal resolution according to the present invention.
• the prospective application of the back pixelation technique using a predetermined acquisition matrix permits scheduling a spatial image acquisition pattern for multiple acquisitions over time.
• the temporal resolution of the resulting imaging would be that of the actual acquisition speed of the images.
• the multiple acquisitions may be adaptive, as described in the preceding subsection. Such methods are particularly useful where temporal considerations are important.
• a further enhancement to the adaptive determination of the acquisition matrix or scheme and the temporal scheduling of the acquisitions is the use of a small-scaled initial acquisition(s) (i.e., a preliminary test acquisition) to determine the desired matrix parameters.
• a small-scaled initial acquisition(s) i.e., a preliminary test acquisition
• a smaller contrast media dose a "test bolus" can be administered prior to the diagnostic larger contrast bolus exam. See, e.g., Earls JP et al. Radiology 1996, 201:705-710.
• Scanning during the test bolus using a trial matrix may be used as such to determine optimal spatial and temporal resolution for back pixelation acquisition in order to monitor for the desired dynamic change, that is, in this example, the change of contrast media enhancement during the subsequent full diagnostic contrast media dose administration.
• a preliminary test acquisition(s) may be particularly useful when performed in conjunction with an operator-controlled or subject-dependent intervention or time- referenced change in a clinical diagnostics and/or treatment context, such as the intravascular administration of a test bolus of contrast media, breathing motion, cardiac contraction, and the movement of a structure, etc.
• the segmentation of image data into small time “packets” also facilitates the improved ability to synchronize or gate data acquisition for imaging regions where physiologic motion is a dominant concern. For example, cardiac imaging requires both cardiac and respiratory gating.
• the back pixelation methods would enable the shortening of the acquisition window per cardiac or respiratory cycle for the minimization of image blurring, thereby improving structural depiction and enabling improved temporal synchronization of image data acquisition for specific periods of the cycle.
• back pixelation methods of this invention improve image resolution both spatially and temporality, and hence allow a better balance between the two. These methods become particular useful where, other than the requirement of a reasonable spatial resolution, a requirement of high temporal resolution is critical; the examples of such situations include imaging contrast enhancement and determining bolus kinetics such as wash- in, time-to-peak, equilibrium, and wash-out times. Contrast media is commonly used in the assessment of vascularity and tissue or tumor perfusion and in the assessment of vessels.
• the back pixelation technique may also be used for transmission of image data according to the present invention.
• Advantageous reduction of data volume may be achieved by reducing or decomposing an image data set into smaller lower resolution data packets for transmission. That is, for example, a two-dimensional image of an object may be decomposed through back pixelation process into a plurality of images taken from the object, each image in the plurality is capable of overlaying one other images in the plurality at an increment.
• the images in the plurality may be transmitted separately, one at a time.
• one may choose to transmit only part of the plurality (hence achieve lower image resolution) or the entire plurality (hence achieve better resolution, same as or similar to the resolution of the original image).
• An intermediate image may then be derived as its pixel values are computed from the values of the corresponding pixels in the transmitted images in the plurality. Subsequently, a resultant two- dimensional image having a second pixel size may be produced by dividing, in the Fourier domain, the intermediate image by a filtering function; the filter function is capable of representing artifacts from the multiple overlaying.
• the filter function is a point response function according to one embodiment of the invention.
• the computing may be performed by taking arithmetic means in some embodiments. In other embodiments, the computing is performed by a heuristic function or an arithmetic function according to suitable mathematics and artificial intelligence procedures.
• Image data transmission and data acquisition may be coupled in real time according to this invention.
• a two-dimensional image of an object may be produced and transmitted from multiple acquisitions using a two-dimensional acquisition matrix.
• the acquisition matrix and the transmission process defines the resulting pixel size of the two-dimensional image.
• a multiplicity of images from the object are acquired, one at a time, by shifting one or more units in the acquisition matrix.
• the multiplicity of images are transmitted, one at a time.
• a intermediate image is derived as its pixel values are computed from the pixel values in each transmitted image in the multiplicity.
• the resulting image after transmission is obtained by dividing, in the Fourier domain, the intermediate image by a filtering function; the filter function is capable of representing artifacts from multiple acquisitions.
• the resulting image resolution after transmission may be adjusted on an as needed basis.
• Image packets containing resolution data may be sent via remote communication to a client where they are reconstructed in real time. Transmission may be halted or stopped once the client determines that the resolution of the reconstructed image from transmitted packets becomes satisfactory.
• this procedure enables time efficient data transmission. It in essence can yield a gradually improving image at the receiving end and hence allows early initiation of a subsequent process before completion of the data transmission, which may be critical in medical monitoring and treatment context, for example.
• Decomposition of image data based on spatial resolution according to this invention is advantageous over other forms of data segmentation or reduction (e.g., preserve every fourth point or a quadrant of the image data), because the preserved data packet represents an overlay(s) of the entire image data set. If one data packet is lost on transmission, the outcome is an image with a slightly lowered spatial resolution. Whereas, with other data segmentation or reduction schemes, the resultant image may have voids at random areas of the image. Moreover, in such situations, the quality of transmitted image may be readily restored according to the present invention by enabling retransmission of only the lost image data packet.
• Example 1 A Computer Simulation Of Image Reconstruction Using The Back Pixelation Technique
• Image data processing is implemented with the C++ programming language.
• Examples of the C++ source code segments are included in Examples 2 and 3 infra.
• an intermediate image was derived from the first stage of image data reconstruction. This procedure was coded using C++ in Example 3 infra. Essentially, the arithmetic mean of the corresponding pixel values for all images from multiple sampling is taken and assigned as the value of each pixel. The resulting intermediate image shows an improvement in the quality of information contained therein. However, the reconstruction artifact from multiple sampling and overlaying obscures most of the detailed features of the structure. As a final step, the intermediate image was further processed to remove the reconstruction artifact. A filter mask was generated based on the size of the pixel used in the simulation space, which was equal to the pixel size of the finally reconstructed image. Image restoration was then performed by dividing the intermediate image by the filter mask in the Fourier domain.
• Example 2 A C++ Source Code Segment Implementing Simulated Acquisition Or Sampling
• BOOL IsOpen FALSE; CFile savefile;
• IsOpen savefile.Open(filename, CFile: :modeCreate
• Example 3 A C++ Source Code Segment Implementing Simulated Image Reconstruction And Filtering
• IsSharp FALSE
• IsHarmonic FALSE
• m_xPxls nPxls+2*x ⁇ ad
• m_yPxls nPxls+2*ypad
• m_xOrigin xpad
• m_yOrigin ypad
• IsOpen savefile.0pen(file ⁇ athname, CFile: :modeCreate
• ⁇ buffer[i] (byte)m_array[i];
• M (fft * w_complex*)malloc(m_yPxls*(m_xPxls/2+l)*sizeof(fftw_complex)); (fftw_complex*)malloc(m_yPxls*(m_xPxls/2+l)*sizeof(fftw_complex));
• R NULL)
• invplan rffrw2d_create_plan(m_yPxls,m_xPxls,FFTW_COMPLEX_TO_REAL,
• the region V is sampled with a cuboidal sample region S of size 1 in such a way that S just overlaps V at point (xl,yl,zl) by some value dl. This is sample position (1,1,1).
• a new region of space C is created which is to be used for computing the reconstruction of V that is of the size V + 21 in each dimension. In this region of space at sample location (0,0,0), the sample region would be in the extreme near bottom left corner of C. At the final sample location (x2 + 1, y2 + 1, z2 + 1), the sample region would be in the extreme far upper right corner of C. Every point in C is assigned a value of 0.
• the sampling and reconstruction computation process here is performed by stepping the sample region S through each sample location using step sizes of dl. By way of pseudo-code, the following routine illustrates this process:
• zn is the number of steps to cross the region in the z axis, yn the y axis, xn the x axis,
• this process in the three-dimensional space may increase image resolution without increasing the magnetic field strength.
• the resulting artifact around a point in space would be a three-dimensional Gaussian blur, which can be removed through the aforementioned filtering process that is applied to two- dimensional images extrapolated to three-dimensional data sets.

Applications Claiming Priority (5)

Publications (2)

Family

ID=26996595

Family Applications (1)

Country Status (5)

Cited By (1)

Family Cites Families (4)

• 2002
  • 2002-04-09 CA CA2443969A patent/CA2443969C/en not_active Expired - Fee Related
  • 2002-04-09 WO PCT/US2002/010994 patent/WO2002082357A1/en active Application Filing
  • 2002-04-09 JP JP2002580246A patent/JP2004534295A/ja active Pending
  • 2002-04-09 AU AU2002256121A patent/AU2002256121B2/en not_active Ceased
  • 2002-04-09 EP EP02725565A patent/EP1382005A4/de not_active Withdrawn

Non-Patent Citations (5)

Also Published As

Similar Documents

Legal Events