Systems and methods for visualizing and characterizing objects in images

Abstract

Convergence-based systems and methods for visualizing and characterizing objects in images. [Solution] A method for visualizing, characterizing, and detecting objects in an image by applying a local micro-contrast convergence algorithm to a first image to generate a second image that is different from the first image, wherein all similar objects converge to a similar pattern or color in the second image.

Description

[Technical Field]

[0001]

[0002] The present invention relates generally to image processing, and more particularly to a convergence-based system and method for visualizing and characterizing objects in images. [Background technology]

[0002] One embodiment of the present invention relates to a system and method for visualizing and characterizing objects, for example, in an image. Summary of the Invention [Problem to be solved by the invention]

[0003] One embodiment of the present invention relates to a system and method for visualizing and characterizing objects, for example, in an image. [Means for solving the problem]

[0004]

[0003] In one embodiment, there is a convergence-based method that visualizes and characterizes all features in a first grayscale image so that the first image is replicated into at least two channels with equivalent luminance values, then applies a local micro-contrast convergence algorithm that transforms at least some of the input values ​​of each copy's channel so that the output pixel values ​​of each copy's channel differ from both its input pixel values ​​and the output pixel values ​​of any other copy's channel, and then uses a lookup table to map values ​​for each vector in each channel, which in the process collectively generate a second image that differs from the first image.

[0005]

[0004] The channels may be created as grayscale channels, alpha channels, color information channels, or a combination of the three.

[0005] In a further embodiment, a second local micro-contrast convergence algorithm, separate and distinct from the first local micro-contrast convergence algorithm, is applied to the second image to generate a third image, separate and distinct from the first image and separate and distinct from the second image.

[0006] In a further embodiment, modifying the third image by successively applying one or more additional local micro-contrast convergence algorithms to generate a fourth image.

[0007]

[0007] In a further embodiment, one or more of the first image, second image, third image, or fourth image are combined to generate a fifth image that is separate and distinct from the first image, second image, third image, or fourth image.

[0008] In a further embodiment, the local micro-contrast convergence algorithm sequence includes one or more of the preceding types of multi-dimensional (multi-channel) image transformations.

[0009] In a further embodiment, the multidimensional image transformation may be represented as a profile look-up table (PLUT) in a digital file format as hexadecimal code or text.

[0010] In a further embodiment, the multidimensional image transformation may be stored as a PLUT in a digital file format as one or more matrices.

[0011] In a further embodiment, a local micro-contrast convergence algorithm can define and process a sequence of transformations utilizing metrics specified in the PLUT that transform image input pixel values ​​representing specific material types into image output pixel values ​​in order to converge the relationships between neighboring pixel groups into predictable color and luminosity patterns that follow the structure of the material and its relationship to the imaging modality, allowing each material to be uniquely characterized and visually distinguishable.

[0011]

[0012] In a further embodiment, the local micro-contrast convergence, multi-dimensional image transformation uses spatial variables (x,y) and luminance variables (i) to compute two-dimensional input functions F1(x,y,i), F2(x,y,i), ..., F N The set of (x,y,i) is the two-dimensional output functions G1(x,y,i), G2(x,y,i), . . ., G N It can be stored as a PLUT in a digital file format, which maps to a set of (x,y,i).

[0012]

[0013] In a further embodiment, the multidimensional image transformation uses spatial variables (x,y), luminance variables (i), and alpha or color channels (j,k,l) ​​to generate two-dimensional input functions F1(x,y,i), F2(x,y,i), ..., F N The set of (x,y,i) is the submatrix G1(x,y,i,j,k,l), G2(x,y,i,j,k,l), , G N It can be stored as a PLUT in a digital file format, which is mapped to a set of more than two-dimensional output functions of the form (x,y,i,j,k,l).

[0013]

[0014] In a further embodiment, the first grayscale image may be replicated in a first multi-dimensional space where each layer dimension of the multi-dimensional space is a replica of the first image.

[0015] In a further embodiment, the number of dimensions in the multidimensional space is equal to two or more.

[0014]

[0016] In a further embodiment, the number of dimensions in the multidimensional space is equal to four, including luminance and the color components red, green, and blue.

[0017] In a further embodiment, the number of dimensions in the multidimensional space is equal to the N dimensions of a color space such as HSL, XYZ, and CMYK.

[0015]

[0018] In a further embodiment, converting the multi-dimensional color space image produced by the local micro-contrast convergence algorithm sequence into a single channel grayscale image.

[0016]

[0019] In a further embodiment, converting a multi-dimensional color space image to a single channel grayscale image by differentially modifying the luminance values ​​of colors in the first image so that they are represented in a grayscale (de-saturated) image.

[0017]

[0020] In further embodiments, the function utilized within the local micro-contrast convergence algorithm sequence may include a superposition additive or differential operator utilizing two or more resulting images from two different local micro-contrast algorithm sequences.

[0018]

[0021] In further embodiments, one or more local micro-contrast convergence algorithm sequences may employ finite area convolution filters with an M×M impulse response array to either sharpen or reduce noise in the image.

[0019]

[0022] In further embodiments, the resulting visualized and characterized features may be expressed in the context of a given first grayscale image, with each object or material type converging to a similar pattern or color characteristic of its type, thereby expressing a unique characteristic in response to the algorithmic sequence.

[0020]

[0023] In further embodiments, different local micro-contrast convergence algorithm sequences may be utilized for the same given first grayscale image to represent different converged visualizations and characterizations of materials within that image by converging all similar materials into similar patterns or colors.

[0021]

[0024] In further embodiments, different algorithm sequences may be created and applied to optimize the characterization of distinct material properties in the image, such as object boundaries, textures, fine structures, and variations within the object.

[0022]

[0025] In further embodiments, the first image is an image generated by x-ray, ultrasound, infrared, ultraviolet, MRI, CT scan, PET scan, grayscale, color, visible light, millimeter wave, or laser scan.

[0023]

[0026] In further embodiments, cancers, cysts, or any abnormalities in breast tissue, breast, prostate, kidney, liver, bone, lung, brain, or skin of either a human or animal can be visualized and characterized within the context and pattern of all other structures in the image.

[0024]

[0027] In further embodiments, biomarkers for cardiovascular disease, Alzheimer's disease, eye disease, or multiple sclerosis pathology can be visualized and characterized within the context and pattern of all other structures in the image.

[0025]

[0028] In a further embodiment, chemical markers for solid or liquid organic compounds can be visualized and characterized within the context and pattern of all other structures in the image.

[0026]

[0029] In a further embodiment, structural defects or anomalies can be visualized and characterized within the context and pattern of all other structures in the image.

[0030] In one embodiment, there is a system for reducing false positive error rates for visually or digitally representing the presence of a feature in an image according to any of the methods described herein.

[0027]

[0031] In one embodiment, there is a method for reducing false negative error rates for visually or digitally representing the presence of features in an image, comprising applying a local micro-contrast tissue convergence algorithm to a first image to generate a second image that differs from the first image.

[0028]

[0032] In further embodiments, the first image is an image generated by x-ray, ultrasound, infrared, ultraviolet, MRI, CT scan, PET scan, grayscale, color, visible light, millimeter wave, or laser scan.

[0029]

[0033] In further embodiments, cancers, cysts, or any abnormalities in breast tissue, breast, prostate, kidney, liver, bone, lung, brain, or skin of either a human or animal can be visualized and characterized within the context and pattern of all other structures in the image.

[0030]

[0034] In further embodiments, biomarkers for cardiovascular disease, Alzheimer's disease, eye disease, or multiple sclerosis pathology can be visualized and characterized within the context and pattern of all other structures in the image.

[0031]

[0035] In a further embodiment, chemical markers for solid or liquid organic compounds can be visualized and characterized within the context and pattern of all other structures in the image.

[0032]

[0036] In a further embodiment, structural defects or anomalies can be visualized and characterized within the context and pattern of all other structures in the image.

[0037] In further embodiments, the false negative rate for breast cancer detected or visualized by a radiologist in a second (i.e., subsequent) image is less than 16% for normal breasts and less than 60% for breasts with areas of dense tissue.

[0033]

[0038] In one embodiment, there is a system for reducing the false negative error rate for detecting or revealing features in an image according to any of the methods described herein.

[0039] In one embodiment, a system includes one or more memory units each operable to store at least one program; and at least one processor communicatively coupled to the one or more memory units, wherein the at least one program, when executed by the at least one processor, causes the at least one processor to perform the following steps: receiving an image; mapping pixel values ​​of the image to an initial multidimensional color space; applying one or more local micro-contrast convergence transfer functions to the initial multidimensional color space of the image to cause local micro-contrast convergence and create a processed image having a multidimensional color space; and displaying a visualization of the image based on the processed multidimensional color space.

[0034]

[0040] In a further embodiment, converting the processed multi-dimensional color space image to a single channel grayscale image.

[0041] In a further embodiment, the multi-dimensional color space image includes a luminance dimension having luminance values.

[0035]

[0042] In a further embodiment, converting the processed multi-dimensional color space to a single channel grayscale image by differentially modifying the luminance values ​​of the colors in the first image so that they are represented as a grayscale (de-saturated) image for image display or analysis purposes.

[0036]

[0043] In a further embodiment, the multi-dimensional color space is an RGB color space.

[0044] In some embodiments, the multi-dimensional color space may be one of HSV, HLS, HSB, XYZ, CMYK, CIEXYZ, or CIELAB.

[0037]

[0045] In a further embodiment, the system further includes processing a breast image (mammogram, CT, MRI, or ultrasound) and applying a median filter to the initial multidimensional color space, wherein applying one or more PLUT functions to the initial multidimensional color space includes applying a first set of PLUT functions to attenuate low-density fatty breast tissue; applying a second set of PLUT functions to make the fatty breast tissue appear as a first color and distinguish denser breast tissue using other colors; applying a third set of PLUT functions to amplify low pixel values ​​and attenuate high pixel values ​​in the color space layer associated with the first color; and applying a fourth set of PLUT functions to change the background of the image to black or other desired luminance or color value when displayed. include.

[0038]

[0046] In a further embodiment, the system further includes receiving a second image, the second image being substantially similar to the first image; mapping pixel values ​​of the second image to a second initial multidimensional color space; applying a median filter and a convolution filter to the initial multidimensional color space to create a second processed multidimensional color space; and displaying an image visualization based on the processed multidimensional color space associated with the first image and the second processed multidimensional color space associated with the second image, wherein applying one or more PLUT functions to the initial multidimensional color space associated with the first image includes applying a first set of PLUT functions to boost darker values ​​and attenuate mid-tones of the image, applying a second set of PLUT functions to the multidimensional color space to add subtle hues, and applying a third set of PLUT functions to enhance tonal values ​​associated with cancer.

[0039]

[0047] In a further embodiment, the system further includes adjusting a gamma level of the multidimensional color space to adjust contrast and highlight structural details of the first image, and applying one or more PLUT functions to the initial multidimensional color space associated with the first image includes applying a first set of PLUT functions to slightly reduce luminance levels and applying a second set of PLUT functions to invert values ​​in the initial multidimensional color space associated with luminance.

[0040]

[0048] In a further embodiment, the first image is a mammogram including dense and fatty tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is separate and distinct from the first image includes mapping pixel values ​​of the first image to a first multidimensional color space; applying a median filter to the first multidimensional color space to generate the second multidimensional color space; inverting the second multidimensional color space to generate a third multidimensional color space; applying a first set of one or more nonlinear transfer functions to the third multidimensional color space to generate a fourth multidimensional color space and make fatty breast tissue appear as one color and distinguish denser breast tissue using another color; applying a second set of one or more transfer functions to the fourth multidimensional color space to generate a fifth multidimensional color space and amplify high pixel values, attenuate low pixel values, and highlight breast area structures; and displaying an image visualization based on the fifth multidimensional color space.

[0041]

[0049] In a further embodiment, the first image is a mammogram including dense and fatty tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is separate and distinct from the first image includes mapping pixel values ​​of the first image to a first multidimensional color space; applying a first set of one or more transfer functions to the first multidimensional color space to generate the second multidimensional color space, making fatty breast tissue appear as one color and distinguishing denser breast tissue using another color; converting the second multidimensional color space to a third multidimensional color space in the HLS color space; and displaying an image visualization based on the third multidimensional color space.

[0042]

[0050] In a further embodiment, the first image is a mammogram including dense and fatty tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is separate and distinct from the first image includes mapping pixel values ​​of the first image to a first multidimensional color space, applying a first set of one or more transfer functions to the first multidimensional color space to generate a second multidimensional color space, making fatty breast tissue appear as one color and using other colors to distinguish and reveal detailed structures in the denser breast tissue, and viewing the image based on the second multidimensional color space. and displaying a visualization.

[0043]

[0051] In a further embodiment, the first image is a mammogram including dense and fatty tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is separate and distinct from the first image includes mapping pixel values ​​of the first image to a first multidimensional color space, applying a first set of one or more transfer functions to the first multidimensional color space to generate a second multidimensional color space, making fatty breast tissue appear translucent and using other colors to distinguish denser breast tissue and distinguish small dot-like structures, and displaying an image visualization based on the second multidimensional color space.

[0044]

[0052] In a further embodiment, the first image is a mammogram including dense tissue and fatty tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is separate and distinct from the first image includes mapping pixel values ​​of the first image to a first multidimensional color space, applying a median filter to the first multidimensional color space to generate a second multidimensional color space, applying a convolution filter to the second multidimensional color space to generate a third multidimensional color space, importing a duplicate first image, mapping the image pixel values ​​to a fourth multidimensional color space, and applying a first set of one or more transfer functions to the fourth multidimensional color space to generate a fifth multidimensional color space, build contrast, and darken fatty tissue, and applying a second set of one or more transfer functions to the fifth multidimensional color space to build contrast and darken fatty tissue to generate a sixth multidimensional color space; applying a third set of one or more transfer functions to the sixth multidimensional color space to generate a seventh multidimensional color space to invert fatty breast tissue luminance to appear as one color and use another color to distinguish and reveal detailed structure in the denser breast tissue; applying a fourth set of one or more transfer functions to the seventh multidimensional color space to generate an eighth multidimensional color space to define a breast boundary; merging the third multidimensional color space with the eighth multidimensional color space to generate a ninth multidimensional color space; and converting the ninth multidimensional color space to grayscale values ​​and displaying an image representing the ninth multidimensional color space.

[0045]

[0053] In a further embodiment, the first image is a mammogram including dense and fatty tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is separate and distinct from the first image includes mapping pixel values ​​of the first image into a first multidimensional color space, applying a first set of one or more transfer functions to the first multidimensional color space to generate a second multidimensional color space and non-linearly inverting the image pixel values, and applying a second set of one or more transfer functions to the second multidimensional color space to generate a third multidimensional color space and making fatty breast tissue appear as one color and using other colors to distinguish and reveal detailed structure in the denser breast tissue. applying a third set of one or more transfer functions to the third multidimensional color space to generate a fourth multidimensional color space, making fatty breast tissue appear as one color and using other colors to distinguish and reveal detailed structures in the denser breast tissue; converting the fourth multidimensional color space to a fifth multidimensional color space in the HLS color space; merging the fifth multidimensional color space with the first multidimensional color space by employing darkening blending to generate a sixth multidimensional color space; adjusting the opacity of the sixth multidimensional color space to generate a seventh multidimensional color space; converting the seventh multidimensional color space to grayscale values; and displaying the image representing the seventh multidimensional color space.

[0046]

[0054] In a further embodiment, the first image is a mammogram containing dense tissue and fatty tissue. and applying a first local micro-contrast convergence algorithm to the first image to generate a second image, the second image being separate and distinct from the first image; mapping pixel values ​​of the first image to a first multidimensional color space; applying a median filter to the first multidimensional color space to generate a second multidimensional color space; applying a first set of one or more transfer functions to the second multidimensional color space to generate a third multidimensional color space and to modify contrast and reduce luminosity of fatty tissue; and applying a second set of one or more transfer functions to the third multidimensional color space to generate a fourth multidimensional color space and to colorize all breast tissue except that of higher density. applying a third set of one or more transfer functions to the fourth multidimensional color space to generate a fifth multidimensional color space and reduce adipose tissue to a nearly solid color; inverting the colors of the fifth multidimensional color space to generate a sixth multidimensional color space; applying a fourth set of one or more transfer functions to the sixth multidimensional color space to generate a seventh multidimensional color space and distinguish the breast from outside its boundary; converting the seventh multidimensional color space to an eighth multidimensional color space in the HLS color space and adjusting HLS properties of the eighth multidimensional color space to generate a ninth multidimensional color space; and displaying the image visualization based on the ninth multidimensional color space.

[0047]

[0055] In one embodiment, there is a method performed by the system described herein.

[0056] The foregoing summary, as well as the following detailed description of embodiments of the invention, will be better understood when read in conjunction with the accompanying drawings of illustrative embodiments, it being understood, however, that the invention is not limited to the precise arrangements and instrumentalities shown.

[0048]

[0057] In the drawing: [Brief explanation of the drawings]

[0049] [Figure 1a]

[0058] FIG. 2 illustrates elements of an iterative function module, in accordance with at least some embodiments of the present invention. [Figure 1b]

[0059] 1A-1C show two resulting image representations after processing an original mammogram with two different local micro-contrast convergence algorithm sequences to reveal cancer patterns, in accordance with at least some embodiments of the present invention. [Figure 1c]

[0060] FIG. 1 illustrates a characteristic nonlinear luminance transformation “tone adjustment curve” with two nodal (anchor) points where the shape of the curve is adjusted to change the input value to the output value, in accordance with at least some embodiments of the present invention. [Figure 1d]

[0061] FIG. 1 illustrates a hierarchy of levels of image processing and analysis implemented in accordance with at least some embodiments of the present invention. [Figure 1e]

[0062] FIG. 1 illustrates a local micro-contrast convergence algorithm sequence, in accordance with at least some embodiments of the present invention. [Figure 1f]

[0063] FIG. 1 illustrates a plot in a coordinate system representing a nonlinear transfer function utilized by at least some embodiments of the present invention. [Figure 1g]

[0064] FIG. 1 illustrates a plot in a coordinate system representing breast tissue color in a grayscale image utilized in accordance with at least some embodiments of the present invention. [Figure 2a]

[0065] FIG. 1 illustrates mathematical functions that may be used to generate new profile lookup table (PLUT) input and output values, according to at least some embodiments of the present invention. [Figure 2b]

[0066] FIG. 1 illustrates a lookup table (LUT) for an 8-bit grayscale image, in accordance with at least some embodiments of the present invention. [Figure 2c]

[0067] FIG. 1 is a block diagram illustrating an electronic device for implementing one or more methods in accordance with at least some embodiments of the present invention. [Figure 3]

[0068] FIG. 3a shows an exemplary high density original x-ray mammogram containing cancer in the brightest area of ​​the image.

[0069] FIG. 3b shows an exemplary mammogram image after application of a local micro-contrast convergence algorithm sequence to create a resultant image using one or more methods according to at least some embodiments of the present invention. [Figure 4a]

[0070] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4b] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4c] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4d] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4e] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4f]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4g] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4h] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4i] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4j] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 4k] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 5a]

[0071] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 5b]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 5c] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 5d] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 5e] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 5f] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 5g] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 5h] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 5i]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal low-attenuation breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6a]

[0072] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6b] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6c] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6d] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6e] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6f] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6g]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6h] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 6i] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal details in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7a]

[0073] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7b] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7c] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7d]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7e] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7f] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7h] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7i] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 7j] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal the presence of microcalcifications in dense breast tissue in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8a]

[0074] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8b] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8c] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8d] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8e] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8f] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8g] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8h] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8i] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8j] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8k] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8l] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8m] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8n]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8o] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8p] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8q] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8r] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8s] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8t] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 8u] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal very fine breast tissue structure details in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9a]

[0075] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9b] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9c] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9d] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9e] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9f] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9g]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9h] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9i] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9j] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9k] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9l] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9m] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9n]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9o] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9p] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 9q] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to reveal breast abnormalities in the resulting grayscale image, in accordance with at least some embodiments of the present invention. [Figure 10a]

[0076] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10b] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10c] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10d]FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10e] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10f] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10g] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10h] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10i] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10j] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10k] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10l] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10m] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10n] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10o] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10p] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10q] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10r] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10s] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10t] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10u] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10v] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 10w] FIG. 1 illustrates an exemplary local micro-contrast convergence algorithm sequence for processing mammography images to isolate breast abnormalities in the resulting color image, in accordance with at least some embodiments of the present invention. [Figure 11a]

[0077] FIG. 10 illustrates an exemplary local micro-contrast convergence algorithm sequence applied to four different mammograms generated from four different image acquisition modalities showing the same pattern from one local micro-contrast convergence algorithm, in accordance with at least some embodiments of the present invention. [Figure 11b] FIG. 10 illustrates an exemplary local micro-contrast convergence algorithm sequence applied to four different mammograms generated from four different image acquisition modalities showing the same pattern from one local micro-contrast convergence algorithm, in accordance with at least some embodiments of the present invention. [Figure 11c]FIG. 10 illustrates an exemplary local micro-contrast convergence algorithm sequence applied to four different mammograms generated from four different image acquisition modalities showing the same pattern from one local micro-contrast convergence algorithm, in accordance with at least some embodiments of the present invention. [Figure 11d] FIG. 10 illustrates an exemplary local micro-contrast convergence algorithm sequence applied to four different mammograms generated from four different image acquisition modalities showing the same pattern from one local micro-contrast convergence algorithm, in accordance with at least some embodiments of the present invention. [Figure 11e]

[0078] FIG. 1 shows an X-ray image of surgically removed breast cancer tissue. [Figure 11f]

[0079] FIG. 11c shows the results after applying an exemplary local micro-contrast convergence algorithm sequence to the x-ray in FIG. 11e, in accordance with at least some embodiments of the present invention. [Figure 11g]

[0080] FIG. 1 shows a magnification of a mammography X-ray image revealing the presence of cancer. [Figure 11h]

[0081] FIG. 11g shows the results after applying an exemplary local micro-contrast convergence algorithm sequence to the x-ray in FIG. DETAILED DESCRIPTION OF THE INVENTION

[0050]

[0082] Referring in detail to the drawings, wherein like reference numerals indicate like elements throughout, there are shown in Figures 1a-11h systems, devices and methods, generally designated, in accordance with exemplary embodiments of the present invention.

[0051]

[0083] introduction

[0084] Most image processing and analysis methodologies in medicine are designed to diverge, branch, or isolate areas within an image as areas of interest (AOIs). In these processes, the AOIs are isolated by applying one or more sequences of segmentation algorithms. Many image processing and analysis methodologies, known as computer-aided detection (CAD) processes, are designed to identify the presence of breast cancer in mammograms, other diseases in other modalities, and for non-medical applications. Research has shown that CAD processes used in breast image analysis have false positive rates of up to 5,000 to 1. The false positive rate is the ratio between the number of negative events incorrectly categorized as positive (false positives) and the total number of actual negative events.

[0052]

[0085] It is the process of visual or data segmentation of an object of interest, bifurcation of the object in the image, and / or subsequent isolation (divergence) from other tissues in the image that significantly limits the usefulness of such techniques to the clinician. Because the bifurcation / segmentation process removes the context of the surrounding objects / tissues from the larger context in which the AOI arises, the diagnostic value of such a process to the physician is significantly limited, as localization of disease or abnormalities within the breast and its surrounding tissues limits its use in making improved clinical decisions regarding potential outcomes and treatments.

[0053]

[0086] Exploiting local properties in the image such as luminance values, running an average filter, rubber-band straightening transform, measuring circularity in suspect regions of interest, texture Many mathematical techniques have been devised to examine the original grayscale image through gradient histograms and gray-level incremental analysis. Many of these techniques fail to produce acceptable results in areas of the image where the object to be detected is quite similar in value to its surrounding neighbors. While cancer may be detected, the cancer's borders (boundaries) may not be clearly established. Still others utilize machine learning, where an atlas of known pathologies is compared to the image being processed to determine likelihood probabilities based on the similarity between the atlas and an unknown set of image metrics in the image being analyzed.

[0054]

[0087] Furthermore, many CAD methodologies do not improve the visualization and characterization of objects in the processed images to aid radiologists in visually determining the extent of abnormalities or distinguishing normal tissue from abnormalities. Instead, CAD techniques simply place location markers within the original mammogram image. This further presents a dilemma for radiologists in that no additional discriminatory visual information is available to assess the effectiveness of the markers. Using CAD methodologies, radiologists must not only assess the original image for the presence of cancer or other abnormalities, but also assess the effectiveness of a given marker, while being aware of the extremely high false positive rate associated with the CAD process. Similar deficiencies exist across a wide spectrum of fields that use CAD methodologies or image segmentation algorithm approaches.

[0055]

[0088] Therefore, there is a need in the art for improved image processing techniques other than those of CAD, bifurcation, or divergence-based processes.

[0089] Breast cancer imaging application example

[0090] Mammography is the use of X-ray radiographs to generate images of a person's breast to detect the possible presence of breast cancer or other abnormalities. While the use of mammograms is currently the best methodology available for screening to detect breast cancer, between 10% and 30% of women with cancer report negative (i.e., cancer-free). This may be due in part to the highly complex and often very subtle nature of detecting cancer in mammographic images, and is particularly problematic for women with dense breast tissue, who are more likely to suffer from breast cancer. While cancers appear white on mammograms, breasts contain non-cancerous elements that also appear white (e.g., dense breast tissue) and dark (e.g., fatty breast tissue). While radiologists more easily observe cancers in fatty tissue, cancers arising in dense breast tissue are very difficult to distinguish from the surrounding tissue. Approximately 40% of women have breasts that contain at least a portion of dense tissue; therefore, there is a great need to be able to distinguish cancerous lesions regardless of the level or relative amount of density in a woman's breast tissue.

[0056]

[0091] Furthermore, when a radiologist determines that breast cancer may be present in a mammogram, several possible follow-up procedures may be employed. These may include the use of ultrasound, MRI with contrast, breast CT scan, and biopsy. These follow-up procedures are expensive and often emotionally traumatic for the patient and their family, and in some cases, may cause physical trauma. When indicating the need for a biopsy, ultrasound has a positive predictive value of only 9%. Clinically, 91% of patients who undergo ultrasound followed by biopsy are confirmed by pathology as cancer-free. Similarly, 60% of patients who undergo MRI and proceed to biopsy do not have cancer. As used herein, positive predictive value refers to the probability that a subject with a positive screening test has disease. As used herein, negative predictive value refers to the probability that a subject with a negative screening test does not have disease.

[0057]

[0092] Patients who have a possible disease flagged in their mammogram may be sent for an ultrasound or an MRI with contrast. When an ultrasound is performed and the radiologist determines from the ultrasound image that cancer may be present, a biopsy is often performed. Of those patients who underwent ultrasound-based follow-up biopsies, 91% were cancer-free.

[0058]

[0093] A highly sensitive and specific technique that can reveal cancer using only standard screening and inexpensive imaging (e.g., mammograms) would be revolutionary in today's cancer detection environment. Approximately 90% of breast cancers arise in the cells lining the ducts of breast tissue. Early detection of breast cancer relies on the clinical ability to distinguish changes that may be present in images. Again, the presence of localized or globally dense breast tissue makes early detection of breast cancer a challenging task. Depending on breast density, a dense breast can be understood to contain 5% to 95% dense breast tissue. Generally, density varies throughout the breast volume, with some localized areas having greater or lesser density than other (e.g., different or nearby) areas. There may be certain areas in a woman's breast that are very dense overall, and other areas with very low density containing fatty tissue. In some women, the entire breast is very dense, while in others, there are only spot-like areas of high density. Despite the amount of density as a percentage of a woman's breast, cancers occurring within dense areas are easily misdiagnosed because breast cancer appears white in a mammogram, as does dense breast tissue, often leading to radiologists being unable to distinguish between the high density and the cancer itself.

[0059]

[0094] Breast cancer can develop from normal tissue in one or more different stages of change. Abnormal tissue development can progress from normal to hyperplasia, atypical hyperplasia, ductal carcinoma in situ (DCIS), and invasive DCIS. Tissue can develop from normal to invasive carcinoma without any intervening steps. When the tumor grows beyond the duct, it is called an invasive carcinoma.

[0060]

[0095] Currently, only 1% of breast cancers can be detected when the lesions are 1 mm or less in size.

[0096] The challenge of using computer-aided detection and machine learning techniques to detect cancer in images that exhibit varying local or global tissue densities is exacerbated by the variability associated with dynamic structural changes that can occur in biological tissue. Disease segmentation, with its large number of possible combinations, makes it extremely difficult to train a computer to consistently detect cancer while maintaining a low number of false positives.

[0061]

[0097] Techniques such as standard machine learning protocols, the use of segmentation algorithms, and processes for isolating only disease-related pixels in an image (i.e., segmenting or bifurcation) suffer from having too many possible combinations to correctly identify disease. These processes work best when there is a single object with a unique boundary associated with the object of interest. For example, identifying bacteria in an image generated through a microscope is aided because bacteria have a distinct shape and size, and cell boundaries limit other possible combinations. As the name suggests, bifurcation of an image produces abrupt changes that lead to a binary (yes / no) result and does not allow for subtle differences in boundaries within a given region of image content.

[0062]

[0098] In contrast, breast cancer, as well as other diseases and abnormalities, has a diffuse boundary. Cancer is often amorphous and multi-patterned. The tissue can also be in various transition states. Lesions can be in atypical hyperplasia, as well as ductal carcinoma in situ and have cells becoming invasive. Furthermore, both normal and abnormal breast conditions can include or be affected by the following:

[0063]

[0099] Presence of spiculations and calcifications

[0100] Presence of necrotic tissue

[0101] Abundance of dense fibroglandular tissue associated with buried cancer

[0102] Previous surgery, biopsy, or weight gain

[0103] Changes in women during their menstrual cycle or changes from menopause

[0104] Traditional CAD methods

[0105] Generally, radiographic findings related to breast cancer are generally of two different types: The diagnostic procedure involves identifying the presence of ductal structures, masses, and microcalcifications. Microcalcifications associated with pathology commonly occur in the ducts and in association with neoplasms. Masses most often correlate with abnormalities and can be either benign or cancerous. Fibroglandular tissue within the breast can obscure masses, making them difficult to detect in raw images.

[0064]

[0106] In mammography, ensuring that all breast parenchyma is included in the view To achieve this, two mammographic views are typically created for each breast (cranial / caudal CC and medial / lateral oblique MLO), which further complicates the task of cancer detection and quantification in that it is difficult to correlate the presence and dimensionality of structures between the two different views.

[0065]

[0107] Existing computerized diagnostic methodologies involve the following sequence of operations: The general procedure is to suspect a lesion > extract lesions > extract features > classify > predict the probability of malignancy > report the probability.

[0066]

[0108] These methodologies allow for the analysis of areas for possible malignancies. As such, it is important to segment or extract (e.g., divide) areas of concern. For example, applying an equalization or divergence process to an image distinguishes fatty tissue from denser tissue. The equalization process is limited in that it is a linear process and does not have a specific thresholding process that is optimal for all mammograms. While a divergence-type segmentation algorithm can be used to separate fatty tissue from dense tissue, it does not effectively support distinguishing white cancerous areas within white dense breast tissue.

[0067]

[0109] Binary processes are generally designed to look for specific diseases, but They do not address other diagnostically important features in mammograms or other medical images, such as architectural distortion, the degree of asymmetry between breasts, nipple cupping, dilated ducts, and skin lesions. While not cancerous, these features remain important to clinicians and their patients. While segmentation and branch-diverge algorithm approaches focus on cancer, they are not necessarily designed to address the overall structure of all tissues in an image.

[0068]

[0110] These segmentation techniques use a set of pixels to define the boundaries of possible lesions. Analysis of the gray level increment in the cell is often used. Other techniques use probabilistic interpolation of pixel data, but interpolation methods are again limited by the extreme similarity between lesions and dense tissue.

[0069]

[0111] Local micro-contrast based convergence

[0112] In some embodiments of the present invention, all structures within the context of a given image are Systems and methods relating to image processing methodologies designed to improve visualization and maintain the context of all tissues by differentially and predictably visualizing and characterizing structures and features are disclosed. These embodiments employ a process of iterative sequencing of image processing functions that causes the local micro-contrast patterns associated with each material type to coalesce (or converge) and consistently manifest as distinctive characteristic patterns within the resulting processed image. In other words, these embodiments provide a method for characterization of all tissue types within the context of the rest of the tissue, rather than attempting to extract or remove identified tissue types outside of the context of the rest of the tissue. can.

[0070]

[0113] Biological growth, neuronal patterns, river branching, pipe corrosion, and snowflakes Many objects in the real world, such as the formation of a sphere, are statistically self-similar, with patterns of evolution exhibiting the same statistical properties at many scales of magnification. In these patterns, small portions of the object or pattern are similar to patterns at larger scales. These self-similar natural patterns are represented as neighborhoods of individual pixels captured in an image. As utilized in at least some embodiments of the invention described herein, an iterative process that can be used in the local micro-contrast convergence methodology can be designed and operates in a manner that explicitly visualizes and characterizes these self-similar patterns at any scale in an image.

[0071]

[0114] Figure 1a shows the local micro-contrast convergence algorithm sequence path. 1 illustrates one embodiment of the method. An original image 101, e.g., a grayscale image 101, is input into an iterative function module processing sequence. Image 101 is then processed by image processing function 102, which either results in a resulting image 103, or is further processed by applying a second, but different, image processing function in function 102. The iterative process 104 can be applied 0 to n times.

[0072]

[0115] Diseases such as cancer exhibit such self-similarity in their growth, and their growth Using local micro-contrast processes, extremely small cancerous lesions can be characterized and visualized at any scale, exhibiting the same pattern of expression as larger lesions.

[0073]

[0116] Fractal geometry is the generation of natural patterns through the repetition of mathematical functions. The techniques illustrated in this set of embodiments mathematically decompose fractal patterns generated in biological systems into identifiable and measurable representations of pixel data within an image. Thus, the local micro-contrast convergence algorithm described herein can mathematically resemble an iterative process, allowing visualization of tissue patterns such as breast boundaries, cancerous and benign lesion margins and cores, and characteristics of breast asymmetries that may be present in mammography images.

[0074]

[0117] As used herein, local micro-contrast convergence refers to the profile It can refer to a repetitive sequencing of image transformations that utilize lookup table (PLUT) functions.

[0075]

[0118] As used herein, the PLUT function refers to the localized self-correction in the source image. Similar image contrast refers to a mathematical formula in a matrix / array that specifies image input and output values ​​for an image such that pixel variables (such as statistically-based co-occurrences of pixel neighborhood relationships, e.g., texture) have a discrete set of values ​​(called reconstruction levels), and pixels in each local neighborhood in the source image (e.g., pixels with similar characteristics) are assigned a single color or luminance value in the resulting output image.

[0076]

[0119] PLUT, and other factors involved in the local microcontrast convergence process The single or repeated application of the algorithm can cause the relationships between neighboring pixel groups to converge or aggregate into a repeatable and predictable color and / or luminosity pattern that conforms to the structure of the material and its relationship to the imaging modality. While tissue / material types can vary significantly, each tissue / material type has a common underlying pixel neighborhood relationship. The resulting local micro-contrast convergence pattern, represented in each area of ​​the image, can be used to determine the luminance values ​​for each object or material, regardless of the presence of different types of surrounding and overlying materials. It is possible to visually represent their characteristic color patterns based on, for example, statistically-based distributions of . For example, using a local micro-contrast convergence algorithm, breast cancer lesions in mammograms, whether they are dark fatty tissue or high luminance dense breast tissue, can be characterized with specific visually observable and uniquely quantifiable patterns.

[0077]

[0120] FIG. 1b shows an original mammogram image 105 and at least some embodiments of the present invention. Two resulting images 107, 108 produced using an embodiment are shown. A box outlining the area of ​​cancer is shown at 106. The two resulting images are created by two different local micro-contrast convergence algorithm sequences that reveal the characteristic patterns of cancer, as shown at 107 and 108. The iterative processing sequences transformed the subtle grayscale patterns of the original breast x-ray into distinctive pattern responses, such as edges, boundaries, internal structures, textures, spicules, and luminance values ​​and colors associated with the cancer response.

[0078]

[0121] Figure 1c shows the image transformation plot using two nodes at 109. , which shows the standard photographic coordinate system used in . As used herein, a node refers to a singular point on a curve where the direction of the curve is changed. Moving a node on a curve changes the circumferential aspect of the curve. The original image input values ​​are shown along the bottom of the plot (x-axis), and the output image values ​​are shown on the vertical axis. This approach has limitations. The nodes change the shape of the "curve" and modify the relationship between the image input and output values. However, the nodes must be linked so that all parts of the curve are continuous. Therefore, the nodes are limited to those that can be mapped using continuous and linked values. Nonlinear transformations that utilize nodes do not perform well when separation of objects of nearly equal density is desired.

[0079]

[0122] Currently, feature extraction is performed when an object is successfully segmented or extracted from the pixel data of an image. is entirely dependent on the degree of extraction. Existing algorithms are optimally designed to locate the brightest areas of a possible lesion, but they often cannot distinguish the outer boundary of the lesion, an area that is diagnostically important for determining where vascularization occurs.

[0080]

[0123] In this application, one or more local micro-contrast convergence functions are No knots are used so that the image can be processed to properly define the possible outer boundaries of the object (or other feature of interest).

[0081]

[0124] Figure 1d shows the hierarchy for the implementation of the local micro-contrast convergence process. Illustrating the layer approach, the sequence progresses from the bottom to the top of the triangle as it relates to higher levels of processing integration.

[0082]

[0125] Multi-image modality fusion leads to local micro-contrast convergence process Modality fusion, in the context of an embodiment of the present application, is the process of adapting image inputs from different types of imaging modalities so that the same or slightly modified local micro-contrast convergence algorithm sequence can visualize and characterize the same type of tissue across different imaging modalities. In that case, the local micro-contrast convergence pattern will be similar for a patient's cancer when seen in X-ray, ultrasound, breast CT, and MRI scans. This allows for the information from different input modalities to be combined in a coherent manner. Imaging-based fusion techniques facilitate early fusion, where signals are incorporated at the image feature level, and late fusion, where information is incorporated at the semantic level using post-processing image feature analysis tools.

[0083]

[0126] These data characterize materials such as normal, benign, and cancerous breast tissue types. It can be used to generate one or more probability distribution functions correlated to local response patterns in one or more vector coordinates in order to identify tissues / materials and to correlate that data from multiple x-ray, MRI, or ultrasound images, even when other tissue / material types are overlaid on top of the tissue / material.

[0084]

[0127] In some embodiments, multi-processing sequencing, multi-image synthesis, and In modality fusion, the acquired images can be analyzed and data can be correlated between them in an expert system. Because all tissues are visualized in a local microcontrast convergence process, disease can be detected and disease pathology can be correlated to the disease's manifestation in the organ of origin. This provides opportunities for advanced research in disease prevention and drug / treatment therapy.

[0085]

[0128] At least some embodiments of the invention described herein are useful in medicine. The complex stochastic non-Euclidean fractal-like shapes with growth / development patterns represented in images of natural processes such as cellular processes allow for consistent characterization of tissue / material types in images where other mathematical models built purely deterministically or deterministically with simple random components fail.

[0086]

[0129] In some embodiments, the method particularly comprises identifying structures within structures. The imaging system is designed to be capable of detecting abnormal tissue structures. For example, in medical imaging applications, the final image provides visual evidence regarding the presence and structure of abnormal tissue in the context of the remaining structures in the image. The final image may also provide a mechanism for correlating abnormal objects to other normal and abnormal tissue types. For example, a cancerous lesion in a breast duct will have a different level of concern than a lesion that has become invasive or appears to be associated with a lymph node. Similarly, a carcinoma in close proximity to microcalcifications will require a different clinical interpretation compared to a carcinoma next to the chest wall or in situations where there is significant asymmetry in the breast.

[0087]

[0130] An example of an iterative image process is shown in Figure 1e. In detail, Figure 1e shows 1 illustrates an exemplary basic sequencing of a local micro-contrast convergence process, whereby an iterative function module 110 technique takes a first image 111 and processes the first image 111 with a first set of one or more non-linear transfer functions 112 (e.g., a local micro-contrast convergence algorithm). The second image created becomes the final resulting image 120, or, if a next processing step is designed as part of the algorithm, the first iterative image 113 is further processed with a second function 114 (e.g., a second set of one or more non-linear transfer functions) to produce image 115. The process may be iterated one or more times with a different set of non-linear transfer functions applied within a given algorithm sequence 116-119 (e.g., a third set of one or more non-linear transfer functions or a fourth set of one or more non-linear transfer functions) to output the resulting image 120.

[0088]

[0131] In some embodiments, the same source image 111 is used to generate a second iterative function The modules may be applied to the same image 111, but with different functions and iterations to reveal different characterizations and relationships between tissues. Thus, this multi-process sequencing approach can give two distinct characterizations of the same object in the same original image.

[0089]

[0132] In some embodiments, two or more of the images obtained are a new image, which is a composite of the acquired images or a composite of one acquired image and the original image They may be combined or merged in a multi-image synthesis process to create a derived image, which may be further processed or combined with other derived images.

[0090]

[0133] FIG. 1f illustrates a discontinuous nonlinear transfer function in accordance with at least one embodiment of the present invention. 121. FIG. 1f shows an example of mapping input values ​​of an input image along the x-axis and output values ​​of an output image along the y-axis. Graphic plots generated from PLUTs demonstrate the possibility of designing discontinuous transforms to be applied to images. By using PLUTs with discontinuities in the design of local micro-contrast convergence algorithms, at least some embodiments of the iterative transform module process can better distinguish the margins of a cancer from surrounding tissue, even when the cancer is embedded in dense breast tissue. This may be only limited or not possible at all with the use of nodal plots when converting input values ​​in an image to output values.

[0091]

[0134] Figure 1g shows a plot of the luminance values ​​of breast tissue in a coordinate system representing the luminance values ​​of breast tissue in a mammogram image. Figure 1g shows an example of mapping input values ​​along the x-axis and output values ​​along the y-axis. Adipose tissue representation 122 is shown in luminance areas of the breast image containing adipose tissue, and dense tissue representation 123 shows luminance areas of the breast image containing dense tissue. Breast cancer generally has a much higher luminance than adipose tissue. Therefore, it is important to separate adipose tissue from dense tissue. Remapping luminance below the red diagonal line makes that part of the image darker, decreasing density 124, while remapping luminance above the line makes values ​​brighter, increasing density 125. This correlation of image property distributions with the discontinuous nonlinear transformation built into the PLUT design reduces the time required to develop new algorithms for new diseases and imaging modalities.

[0092]

[0135] Figure 2a shows the results for new disease applications, modalities, and non-medical applications. 1 illustrates one embodiment in which multiple mathematical functions may be utilized to generate possible PLUT values ​​for multiple image channels to generate different iterations of a local micro-contrast convergence algorithm for use in a diagnostic imaging system. Utilizing computer-based generation of PLUT sequences can significantly speed up the process of developing new algorithmic sequences for visualizing new diseases or abnormalities.

[0093]

[0136] In Figure 2a, the x and y axes reflect the input and output values ​​of the image, with the midpoint 126 specifies one possible location of the midpoint relative to the coordinate system. Figure 2a represents the luminance and color values ​​of an 8-bit image with 256 possible data points for luminance and multiple color channel mapping. Three mathematical functions 127 were automatically plotted, and their values ​​are shown in the plot. The blue curve (blue channel) 128 was created using f(x) = sin(x). The red channel 129 was created using g(x) = tan(x), and the luminance channel 130 was created using h(x) = sec(x). The midpoint 126 (or zero point) can be placed at a location within the coordinate system that best supports the mapping of mathematical functions that can be mapped to the PLUT for optimal tissue / material visualization and characterization in an automated process rather than a tedious manual process.

[0094]

[0137] Figure 2b shows the grayscale 2D lookup table for an 8-bit grayscale image. The matrix representing the grid is shown in 131. Level 0, representing black, is at the top left corner of the grid at 131. The grayscale luminance levels increase in steps from left to right and top to bottom until they reach the pure white level of 255 in the bottom right corner at 132.

[0095]

[0138] Exemplary Computer System

[0139] FIG. 2c illustrates one or more methods according to one or more embodiments of the present invention. 2 shows a block diagram illustrating an electronic device 250 for implementing the present invention.

[0096]

[0140] The electronic device 250 receives data from a user or a remote device; The electronic device 250 may be any computing device for processing, generating, and / or displaying data. The electronic device 250 may include a communication infrastructure 251, a processor 252, a memory 253, a user interface 254, and a communication interface 255.

[0097]

[0141] Processor 252 may be, but is not limited to, a special purpose or general purpose digital signal processor. The processor 252 may be any type of processor, including a processor. In this embodiment, the processor 252 is connected to a communications infrastructure 251 (e.g., a bus or network). Various software implementations are described in terms of this exemplary computer system.

[0098]

[0142] The memory 253 may be a floppy disk drive, a magnetic tape drive, or The electronic device 250 may include at least one of random access memory (RAM), a hard disk drive, and a removable storage drive, such as an optical disk drive. The removable storage drive reads from and / or writes to the removable storage unit. The removable storage unit may be a floppy disk, magnetic tape, optical disk, etc., that is read from and written to by the removable storage drive. The memory 253 may include a computer-usable storage medium that stores computer software programs and / or data for performing any of the computational functions of the electronic device 250. The computer software programs (also referred to as computer control logic), when executed, enable the electronic device 250 to implement embodiments of the present invention as described herein. Thus, such computer software programs represent the controller of the electronic device 250. The memory 253 may include one or more data stores that store imaging data, software files, or any other type of data files.

[0099]

[0143] The user interface 254 is displayed on the display of the electronic device 250 (as shown). The user interface 254 may be a program that controls the electronic device 250 (without the need for a keyboard or mouse). The user interface 254 may include one or more peripheral user interface components, such as a keyboard or mouse. A user may use the peripheral user interface components to interact with the electronic device 250. The user interface 254 may receive user input, such as mouse input or keyboard input, from a mouse or keyboard user interface component. The user interface 254 may display imaging data on a display of the electronic device 250.

[0100]

[0144] The communication interface 255 is used to exchange image data between the electronic device 250 and a remote device. The communication interface 251 allows imaging data to be transferred to and from a remote device. Examples of the communication interface 255 may include a modem, a network interface (such as an Ethernet card), a communication port, a Personal Computer Memory Card International Association (PCMCIA) slot and card, etc. The imaging data transferred via the communication interface 251 is in the form of a signal, which may be an electronic signal, an electromagnetic signal, an optical signal, or any other signal capable of being transmitted or received by a communication interface. These signals are provided to or received from the communication interface 251.

[0101]

[0145] Exemplary Local Micro-Contrast Algorithm

[0146] Figure 3a contains a very dense breast, where the density is outlined at 300. 3 shows a mammogram of a breast cancer. The contour at 301 defines the boundary of the extreme density at 302 that contains the cancer.

[0102]

[0147] FIG. 3b shows a process for processing an image using one or more methods described herein. 3 shows an exemplary mammogram image after 303. In this embodiment, only the densest areas of the breast are revealed in color. Fat and other low-density areas of the breast image are shown in black at 303. Density increase is shown in steps, progressing from the outer boundary at green 300, inward to blue 302, and finally to the black area at the center 303 where the most developed cancer is located. Each color represents a quantitatively different level of breast density. This quantification provides accurate reporting per the American College of Radiology BI-RADS specification for indicating the presence of dense breasts in a woman's mammogram. However, this process can also extend the BI-RADS reporting system beyond a simple overall percentage of breast density. This process can quantify multiple levels of breast density, specify their distribution, and estimate the woman's potential risk. These methods are adaptive and compensate for the extreme variability in mammographic image presentation affected by differences in breast size, breast density, changes during pregnancy, changes associated with aging and menopause, alterations due to the development of cysts, fibroadenomas, calcifications, the presence of lesions, and scarring from trauma, surgery, and biopsy.

[0103]

[0148] CI Algorithm

[0149] An embodiment of the CI algorithm processes the original grayscale mammogram The algorithm sequence is designed to optimize the representation of dense abnormalities in breast tissue by highlighting their boundaries and revealing their internal structure. The algorithm sequence provides significant color and brightness distinction between abnormalities and other normal tissue so that areas of concern are easier for clinicians and patients to readily observe.

[0104]

[0150] FIG. 4k illustrates a method for extracting a pixel value from a grayscale image in accordance with at least one embodiment of the present invention. 4 is a flowchart illustrating a method 400 for creating a visualization from a

[0151] In step 401, processor 252 imports a grayscale image. FIG. 4a shows an exemplary grayscale image of a mammogram in accordance with at least one embodiment of the present invention. FIG. 4d shows a horizontal gradient representation of FIG. 4a. The gradient grayscale image provides a full range of luminance levels compared to the range that different mammograms have, so that the full range of colors represented in the local micro-contrast convergence algorithm sequence can be shown. Each step of the algorithm sequence described in FIG. 4k can again be applied to both the mammogram and the gradient for purposes of explanation and comparison.

[0105]

[0152] In some embodiments, the processor 252 may process an image (e.g., a grayscale image) In some embodiments, the images are imported from memory 253. In other implementations, the images are imported from a remote device via communication interface 251.

[0106]

[0153] In some embodiments, the grayscale image has dimensions such as x and y pixel dimensions. and z bits of grayscale or color depth are imported for processing as an input array or matrix. In some embodiments, the matrix may contain 8, 10, 12, 14, or 16 bits of luminance values ​​(Lp) per pixel. (Lp) is the original is the luminance value of each pixel (p) at location (x,y) in the image. As the number of bits increases, the greater the numerical variation in pixel values ​​also increases. For example, if 8 bits are used, 2 8 possible pixel values ​​can be assigned to each pixel, whereas if 16 bits are used, 2 16 possible pixel values ​​can be assigned to each pixel. By increasing the number of possible pixel values, the image processing methods described herein can increase the variability in the final image.

[0107]

[0154] In step 402, the processor 252 converts the grayscale image into a multidimensional image. Mapping to color space.

[0155] In some embodiments, to map a grayscale image, a grayscale The n-scale image is replicated into additional matrices of equivalent x / y coordinates for each color component and luminance value to form an n-dimensional overlap matrix space of color space layers, where n>1 forms a new set of matrices containing the voxels.

[0108]

[0156] In some embodiments, the grayscale image is replicated using the following formula: do. f(Lp)=Cp, Here, the pixel value at each x / y coordinate in the original is mapped to the corresponding x / y coordinate in each color space layer of C's multidimensional color space.

[0109]

[0157] In one embodiment where n=4, the RGB multidimensional color space has four different components, That is, it may be defined in terms of luminance, red, green, and blue. In these embodiments, the RGB multidimensional color space includes a luminance color space layer and first, second, and third color space layers corresponding to blue, red, and green, respectively. The new matrix C contains pixel values, where R = G = B = luminance for each pixel value, and these pixel values ​​are equal to the grayscale image luminance value (Lp). In some embodiments, there may be a separate luminance-only channel, or in other embodiments, luminance may be generated as a composite of three other channels. In another embodiment, the values ​​may also be expressed in terms of other values ​​of n, for example, where n has three values: luminance, saturation, and hue.

[0110]

[0158] These embodiments may be used in a wide range of color image formats other than those described herein. Those skilled in the art will appreciate that it is possible to operate on n-dimensional matrices that can be visualized in a color image format. The processing of each mammogram (or other image) begins with a multi-channel matrix or image. Additional color spaces can also occur in color spaces such as HSV, CMYK, CIEXYZ, or CIELAB, using either xyz or cylindrical color spaces.

[0111]

[0159] In step 403, the processor 252 calculates a median filter in the multidimensional color space. In some embodiments, a median filter may refer to a nonlinear digital image processing technique that removes noise while preserving the edges of objects in a multidimensional color space. Noise reduction can improve the results of subsequent processing.

[0112]

[0160] In some embodiments, the median filter calculates each pixel value from the sum of the values ​​of neighboring pixels. The median filter is applied to each pixel in the multidimensional color space by replacing it with the median of the values. The neighborhood pattern is sometimes called a "window" that slides pixel by pixel across the image. In some embodiments, the median filter is a 3x3 or radius = 1 median filter. In other embodiments, radii greater than 1 and matrix combinations such as 5x5, 7x7, etc. may be used.

[0113]

[0161] In step 404, the processor 252 inverts the image, thereby Black (0) becomes white (255) and white becomes black. All other values ​​are inverted proportionally except for the midpoint of the image value.

[0114]

[0162] In step 405, the processor 252 calculates a multi-dimensional color space (e.g., R A first set of one or more (e.g., PLUT) nonlinear transfer functions is applied to the GB. Representations of the resulting images are shown in Figures 4b and 4e.

[0115]

[0163] Figure 4g shows the local microcontroller after being applied to the image in Figure 4a. The color values ​​of CI PLUT1 (2D lookup table) optimized to reveal breast structures in the Trust Convergence Algorithm sequence are shown.

[0116]

[0164] FIG. 4i illustrates a method for attenuating low-density breast tissue in accordance with at least one embodiment of the present invention. 2 shows an orthogonal plot illustrating a representation of an exemplary (e.g., PLUT) transfer function applied by processor 252 to a multi-dimensional color space to achieve a desired color space. In this orthogonal plot, the color space layer input is shown on the x-axis, with values ​​ranging from −128 to +128. The corresponding output after the (e.g., PLUT) transfer function is shown on the y-axis, with values ​​ranging from −128 to +128, with 0 being the midpoint of the image luminance level. It can be observed that the location of 0 in the coordinate plot can be located anywhere in the x / y coordinate space.

[0117]

[0165] In FIG. 4i, the red channel is shown at 408 and the green channel at 409. and the luminance channel is 410. In some embodiments, a first (e.g., PLUT) transfer function (as shown in FIG. 4i) is applied to the luminance color space layer to attenuate the low-density fatty breast tissue. In some embodiments, the low-density fatty breast tissue has luminance values ​​in the lower 50% range, lower 40% range, lower 30% range, lower 20% range, or lower 10% range.

[0118]

[0166] At this stage in the process, areas that do not have the potential to harbor cancer are considered The lesions in the image are separated from areas where cancer or other abnormalities may occur. Furthermore, as micro-contrast neighborhoods converge, lesions in the image then form boundaries and begin to represent internal morphological structures. Compared to diffuse grayscale mammography images, visually distinct boundaries are formed based on tissue structure. Problems associated with the phenomenon known as the center-surround effect and limited human visual perception are minimized or eliminated. Gray values ​​are differentially interpreted by the human visual system based on what surrounds an object. The same object may appear lighter against a dark background and darker against a light background. At least some embodiments of the present invention enable PLUT values ​​to be determined that eliminate the center-surround problem that affects cancer perception and detection in mammograms. Based on optimal settings for human visual discrimination based on color perception theory, the image viewed by clinicians after transformation provides significantly improved diagnostic potential for the tissue under examination.

[0119]

[0167] Returning to FIG. 4k, in step 406, processor 252 generates a multidimensional color space. A second set of one or more transfer functions is applied in between.

[0168] Figure 4h shows the local microcontroller after being applied to the image in Figure 4b. The color values ​​of CI PLUT2 (2D lookup table) optimized to reveal breast structures in the Trust Convergence Algorithm sequence are shown.

[0120]

[0169] FIG. 4j is an example of a transfer function applied by processor 252 to a multi-dimensional color space. 4j shows an orthogonal plot illustrating a representation of a representative (e.g., PLUT) set of the red channel at 411 and the luminance channel at 412. In FIG. 4h, a graphical representation of the CI PLUT2 lookup table is shown.

[0121]

[0170] In this orthogonal plot Figure 4j, the color space layer inputs range in value from -128 to +12 The values ​​are shown on the x-axis, ranging from -128 to +128. The corresponding output after the transfer function (visually shown in FIG. 4j) is shown on the y-axis, ranging from -128 to +128, with the midpoint of the image luminance level at 0. In these embodiments, values ​​are applied to the resulting image in FIG. 4b to make fatty breast tissue appear as one color (e.g., blue and magenta) in FIG. 4c, and use other colors to distinguish between denser breast tissue (gold and red) and the breast border (green).

[0122]

[0171] 4c and 4f illustrate a low-density breast in accordance with at least one embodiment of the present invention. 4c shows an exemplary image representation of a mammogram and gradient image based on a multidimensional color space after applying an exemplary second set of one or more nonlinear transfer functions to make tissue appear as a single color and distinguish dense breast tissue. In FIG. 4c, the cancer is revealed in gold 413 and surrounded by black.

[0123]

[0172] 413, the high density of the breast image measured in RGB values ​​in Fig. 4c. Area values ​​are Red > 250 / Green > 165 / Blue < 50.

[0173] In some embodiments, the design philosophy of these transfer functions is to It is employed to attenuate pixel values ​​in areas of the mammogram, thereby reducing clinician eye strain in the final image by assigning values ​​to areas of the mammogram outside the breast that do not interfere with patterns within the breast area.

[0124]

[0174] In step 407, the processor 252 performs a color space conversion based on the processed multi-dimensional color space. A visualization image (e.g., Figure 4c) is displayed based on the results.

[0175] Each step in this process is further processed in grayscale, regardless of size. It converts mammograms (which also work well for MRI and ultrasound images of the breast) into color patterns that clearly demarcate the boundaries of abnormal tissue, as well as reveal the structure of normal breast tissue. In this image visualization, cancerous lesions have a distinctive pattern that separates them from all other abnormal and normal tissue structures.

[0125]

[0176] CI visualization shows differences in the characterization of both cancer and benign lesions in visualization. can be distinguished using histogram analysis. The borders of cancers are clearly defined in CI visualization. Furthermore, differences in structures within the borders of cancers are indicated using characteristic colors and shapes. This makes it easier for radiologists to identify the borders of cancerous and benign structures. For example, in CI visualization, the greater the number of color changes within the borders of cancers, the more advanced the cancerous tissue development. Changes in the tissues surrounding cancerous and benign lesions are also revealed in CI visualization. CI visualization can also reveal angiogenesis occurring at the borders of cancerous lesions.

[0126]

[0177] In addition to the distinctions explained above, CI visualization also distinguishes between radial scars versus cancerous lesions and CI visualization can differentiate between ductal and fibroadenomatous lesions. CI visualization can also reveal the presence of developing cancer within the ducts before it becomes invasive and surrounds the breast tissue. Cancerous tissue can be correlated with the presence of microcalcifications.

[0127]

[0178] The cancerous lesion, as well as all other structures, are viewed from the cranial-caudal (CC) or superior view. Correlations between different views of mammograms for women, such as mediolateral oblique (MLO or oblique view), can be used to correlate data between studies at different times. Characterizing the internal structure of cancer by these methods can be useful for guiding surgeons performing biopsies, lumpectomies, and for women undergoing treatment for cancer. It is so accurate that it can be used to determine progression for a patient.

[0128]

[0179] LD algorithm

[0180] An embodiment of the present invention relating to the LD algorithm is to measure the image (e.g., the original mammogram) The system provides visualization designed to highlight extremely fine structures and details in images occurring in areas of very low density (e.g., 1000 x g). Diagnostically important structures, such as spicules and low-attenuation lesions, are clearly defined.

[0129]

[0181] FIG. 5i illustrates a method for extracting a grayscale image in accordance with at least one embodiment of the present invention. 5 is a flowchart illustrating a method 500 for creating an LD visualization from a

[0182] In step 501, processor 252 imports a grayscale image. Figure 5a shows an exemplary grayscale image of a mammogram in accordance with at least one embodiment of the present invention, and Figure 5d shows a horizontal gradient representation of 256 grayscale values ​​from black to white.

[0130]

[0183] In step 502, the processor 252 converts the grayscale image into a multidimensional image. Mapping to Color Space The grayscale mapping in step 502 is substantially similar to the grayscale mapping in step 402 above.

[0131]

[0184] In step 503, the processor 252 calculates one or more colors in a multi-dimensional color space. 5g and 5h.

[0132]

[0185] FIG. 5h illustrates a method for implementing the method by processor 252 in accordance with at least one embodiment of the present invention. 5 shows an orthogonal plot illustrating a representation of exemplary (e.g., PLUT) transfer functions applied by the luminance color space layer 508. In some embodiments, a first transfer function is applied to the luminance color space layer 508 to brighten low-density areas of the breast image while attenuating high-density breast areas. A second transfer function, representing the red channel 509, colorizes the breast parenchyma while leaving high-density tissue dark. In some embodiments, low-density fatty breast tissue has luminance values ​​in the lower 50% range, the lower 40% range, the lower 30% range, the lower 20% range, or the lower 10% range. This local micro-contrast convergence algorithm design and its associated PLUT values ​​function to reveal details in any portion of the image, regardless of the percentage of low density in the breast.

[0133]

[0186] A representation of the resulting image generated after step 503 is shown in Figures 5b and 5e. is shown in.

[0187] In step 504, multiple images (represented as color images shown in FIG. 5b) are The dimensional color space is then converted to the HSL color space. In this embodiment, the RGB values ​​are converted to luminance, hue, and saturation values ​​as shown below in the example below.

[0134] (Hue, Saturation, Brightness, Zone) (0.0,0.0,0.2,red) (0.0,0.0,0.1,cyan) (0.0,-1.0,0,master)

[0188] The image is first converted into RGB color in step 505 or can be displayed in HSL color space after

[0135]

[0189] The image in Figure 5c (and the corresponding image 5f) are all in the HSL color space. 5b and 5e by setting the master saturation for the hues of red, red, and cyan to -100% saturation. As a result, hue is no longer a factor in the representation of the image. However, the luminance values ​​are still adjustable, and varying the luminance values ​​of various hues in the color space can change the grayscale representation of those values. In some embodiments, the red and cyan luminance values ​​are adjusted to 0.2 and 0.1, respectively. This brightens the gray values ​​of the overall breast background, highlights the inner parts of dense tissue, such as cancerous lesions, and creates separation between the fine structure of the breast and the fatty tissue. The image can be converted to a single-channel image containing only luminance in step 507 (and shown in FIG. 5c).

[0136]

[0190] At this stage in the processing, areas associated with low density luminance values ​​are degraded to very fine structure. The structures are separated from the low density, low frequency areas of the breast parenchyma, border, and chest wall 510. Visually distinct boundaries were formed based on the tissue structures in comparison to the diffuse grayscale mammography image.

[0137]

[0191] HD Algorithm

[0192] The HD algorithm embodiment of the present invention focuses on the very densest areas of the image. It provides visualizations designed to reveal details in the image generated during scanning (e.g., the original mammogram). Structures such as breast abnormalities and cancerous lesions are revealed and clearly defined from the surrounding dense bright / white areas.

[0138]

[0193] FIG. 6i illustrates a method for extracting a pixel value from a grayscale image in accordance with at least one embodiment of the present invention. 6 is a flowchart illustrating a method 600 for creating HD visualizations from

[0194] In step 601, processor 252 imports a grayscale image. Figure 6a shows an exemplary grayscale image of a mammogram, in accordance with at least one embodiment of the present invention.

[0139]

[0195] In step 602, the processor 252 converts the grayscale image into a multidimensional image. Mapping to color space.

[0196] In step 603, the processor 252 calculates one or more colors in a multi-dimensional color space. The first set of one or more nonlinear transfer functions (e.g., the HD PLUT1 local micro-contrast algorithm) is applied. Representations of the first set of one or more nonlinear transfer functions are shown in FIGS. 6g and 6h, respectively. FIG. 6g shows color values ​​of an LD PLUT (lookup table) optimized to reveal breast structure in mammography images. FIG. 6h shows a graphical representation in a coordinate system (e.g., which may be created from the PLUT in FIG. 6h). In these embodiments, the first transfer function is applied to the luminance color space layer to invert the luminance values ​​606 of the breast image. The red channel 607 amplifies low-density areas of the image while attenuating high-density breast areas. The green channel 608, shown diagrammatically in FIG. 6h as a discrete mapping of green channel values, colors the breast boundary and contributes to the red channel, making the breast background yellow. In some embodiments, dense breast tissue is greater than the lower 50% range, the lower 40% range, the lower 30% range, the lower 20% range, or the lower 10% range. The blue channel 609 adds color to define the outer boundary of the breast. This local micro-contrast convergence algorithm design and its associated PLUT values ​​can function to reveal detail in any portion of the image, regardless of the proportion of high density in the breast.

[0140]

[0197] At this stage in the processing, areas of the image containing very dense structures 610 , separated from the low density areas 611 of the breast parenchyma, border, and chest wall, the cancer is further distinguished from other high density areas of the breast. In comparison, a visually distinct boundary was formed based on tissue structure.

[0141]

[0198] The image is then processed in step 604 (e.g., as shown in FIG. 6b). 6c) or may be converted to a grayscale image in step 605 before being displayed (e.g., FIG. 6c) using a weighted conversion of the R, G, and B values ​​to achieve a luminance value according to the following formula: 0.30*R+0.59*G+0.11*B=luminance value.

[0142]

[0199] MC algorithm

[0200] An embodiment of the present invention relating to the MC algorithm is to measure the image (e.g., the original mammogram) This provides visualization designed to reveal details in the image occurring in the very densest areas of the image (grams), with primarily small structures such as calcifications revealed and clearly defined from the surrounding dense light / white areas.

[0143]

[0201] FIG. 7j illustrates a method for extracting a pixel value from a grayscale image in accordance with at least one embodiment of the present invention. 7 is a flowchart illustrating a method 700 for creating an MC visualization from a

[0202] In step 701, processor 252 imports a grayscale image. 7a shows an exemplary grayscale image of a mammogram, in accordance with at least one embodiment of the present invention.

[0144]

[0203] In step 702, the processor 252 converts the grayscale image into a multidimensional image. Mapping to color space.

[0204] In step 703, the processor 252 calculates one or more colors in a multi-dimensional color space. The first set of numerical transfer functions (e.g., the MC PLUT1 local micro-contrast convergence algorithm) is applied. Representations of the local micro-contrast convergence algorithm are shown in Figures 7h and 7i. Figure 7h shows the color values ​​of the MC PLUT (look-up table) optimized to reveal breast structures in mammography images. Figure 7i shows a graphical representation in a coordinate system. In these embodiments, the transfer functions are applied in luminance space 706 to discontinuously invert the luminance values ​​of the breast image. The red channel 707 attenuates most of the image, employing a discontinuous mapping of red channel values. The green channel 708 values ​​are responsible for creating brown tones in dense areas of the breast. The blue channel 709 slightly tints fatty tissue areas of the breast.

[0145]

[0205] The design of this local micro-contrast convergence algorithm and its implications The PLUT value functions to reveal the presence of microcalcifications in any part of the image, regardless of the proportion of high density in the breast.

[0146]

[0206] At this stage in the processing, microcalcification structures are detected in very dense areas of the image. Visually distinct calcifications were more clearly revealed compared to diffuse grayscale mammography images, even when isolated from other dense areas of the breast.

[0147]

[0207] The image is then displayed in a multi-dimensional color space in step 704 (or For example, the image may be converted to a grayscale image in step 705 using a weighted conversion of the R, G, and B values ​​to achieve a luminance value according to the following formula (e.g., FIG. 7b), or 0.30*R+0.59*G+0.11*B=luminance value (e.g., FIG. 7c), which is an enlarged cross-section of the image in FIG. 7b after being converted to grayscale. Small black microcalcifications 710 can be seen more easily from the light background than in the original image.

[0148]

[0208] RF Algorithm

[0209] An embodiment of the present invention relating to the RF algorithm is to The reduced visualization provides a visualization designed to highlight the finest structures and details in the original image (e.g., radiologist). Structures such as spicules and ducts are clearly defined, as are structures within the remaining dense areas, including cancerous structures. In some embodiments, the reduced visualization is shown as an overlay on the original image to improve visibility by the user (e.g., radiologist).

[0149]

[0210] FIG. 8u illustrates a method for extracting a pixel value from a grayscale image in accordance with at least one embodiment of the present invention. 8 is a flowchart illustrating a method 800 for creating an RF visualization from a signal.

[0211] Figures 8b-8c-8l-8m-8s show the original mammogram in Figure 8a. Figures 8e-8f-8n-8o-8t show the results obtained by iteratively applying multiple local micro-contrast convergent transforms, starting from the original gradient grayscale image in Figure 8d.

[0150]

[0212] Figures 8g, 8h, 8p, and 8q show breast structures in mammography images. 8a, 8b, 8c, and 8d show color values ​​of an RF PLUT (look-up table) optimized to reveal . Figures 8i, 8j, 8k, and 8r show graphical representations in a coordinate system (which may be created, for example, from the PLUTs in Figures 8g, 8h, 8p, and 8q).

[0151]

[0213] In step 801, processor 252 imports a grayscale image. 8a shows an exemplary grayscale image of a mammogram, in accordance with at least one embodiment of the present invention.

[0152]

[0214] In step 802, the processor 252 multiplies the original grayscale image. dimensional color space.

[0215] In step 803, processor 252 processes multiple images of the original grayscale image. Apply a median filter of radius 1 to the 3-dimensional color space.

[0153]

[0216] In step 804, the processor 252 performs a folding operation on the multi-dimensional color space of the original image. Applying a Convolution Filter In some embodiments, convolution filtering may be used to modify the spatial frequency characteristics of an image.

[0154]

[0217] In operation, the convolution filter 804 converts each pixel value into a pixel value and its A convolution filter is applied to each pixel in the multidimensional color space by replacing it with a weighted average of neighboring pixel values. The pattern of neighboring pixel values ​​is called a "window" that is applied pixel-by-pixel across the entire image. In some embodiments, the convolution filter is a 3x3 or radius=1 convolution filter. In other embodiments, a 5x5, 8x8, etc. matrix combination may be used.

[0155]

[0218] In one embodiment, the values ​​of the 3x3 convolution filter matrix are as shown in Table 1 below: It has been done.

[0156] [Table 1]

[0157]

[0219] In step 805, processor 252 determines whether the processed data is Copy the multidimensional color space of the image.

[0220] In step 806, processor 252 Import a copy of the original image in the same grayscale as it was created.

[0158]

[0221] In step 807, processor 252 converts the image of the copy into a multidimensional color space. Map it.

[0222] In step 808, the processor 252 calculates the multi-dimensional color space of the image of the copy. Apply a first set of one or more transfer functions (e.g., local micro-contrast convergence transfer function RF PLUT1). In these embodiments, the first transfer function (e.g., of the local micro-contrast convergence function RF PLUT1) is applied to the luminance color space 814 to boost the darker values ​​and attenuate the mid-tones of the image.

[0159]

[0223] In these embodiments, (e.g., local micro-contrast convergence function A second transfer function (RF PLUT2) is applied to the luminance color space 815 to further attenuate the mid-tones. In these embodiments, the mid-tones are attenuated to a minimum at a luminance value of 1 in the image in the 8-bit grayscale luminance range (0-255). In some embodiments, the adipose tissue is slightly boosted at a maximum peak level of 47, which is converted to 71. As a result, the adipose tissue 816 is separated from the dense areas of the breast 817.

[0160]

[0224] 8i, 8j, 8k and 8r show at least one embodiment of the present invention. 2a and 2b show orthogonal plots illustrating representations of exemplary PLUT transfer functions (e.g., generated from PLUT applied by processor 252) according to the color space, coordinates, and values ​​in these orthogonal plots were previously described and illustrated in FIG.

[0161]

[0225] FIG. 8b illustrates an example of increasing the darker values ​​of an image in accordance with at least one embodiment of the present invention. 1 shows an example image of a mammogram based on a multidimensional color space after applying a first set of one or more transfer functions to boost the color and attenuate the mid-tones.

[0162]

[0226] FIG. 8c shows a further reduction in mid-tones in accordance with at least one embodiment of the present invention. 10 shows an exemplary image of a mammogram based on a multidimensional color space after applying a second set of one or more transfer functions to attenuate the mammogram.

[0163]

[0227] In FIG. 8l, in step 810, processor 252 A third set of one or more transfer functions (e.g., local micro-contrast convergence function RF PLUT3) is applied to the multi-dimensional color space of the image to produce the image shown in Figure 8l. In these embodiments, the third transfer function is applied to the luminance color space 818 to create a discontinuous inversion in the luminance values.

[0164]

[0228] In these embodiments, another "color" function 819 in the third set of transfer functions is It can be applied to a color space layer to add subtle hues.

[0229] In step 811, processor 252 performs a multi-dimensional A fourth set of one or more transfer functions (e.g., local micro-contrast convergence function RF PLUT4) is applied to the color space to produce the image shown in Figure 8m. In some embodiments, RF PLUT4, also shown diagrammatically in Figure 8q, is applied to the luminance channel 820 to result in an increase in the lower density luminance values ​​of the image, enhance tonal values ​​associated with cancer, and further define the breast boundary.

[0165]

[0230] In step 812, processor 252 determines whether multiplicative blending is employed. The processed multi-dimensional color space from the image in step 811 (e.g., FIG. 8m) is merged with the copied image from step 805 (e.g., FIG. 8a) by

[0046] . In some embodiments, the two images are blended at 100% opacity. The resulting merged image has emphasis on high-frequency structure and attenuation of low-frequency information, leaving the highest density in color.

[0166]

[0231] In these embodiments, and other embodiments employing a merge function, the merge function The numbers can be utilized to allow a mathematical function to be applied to one or more derived images that utilizes optimal qualities from each of the composite images for a particular purpose. For example, an image representing the boundaries of cancerous tissue in an image can be combined with an image representing high-frequency information. Such a combination can simultaneously show the extent of the cancer, as cancer relates to possible high-frequency structures, such as spicules and calcifications, within the tumor.

[0167]

[0232] FIG. 8t illustrates a method for performing the steps in step 812 in accordance with at least one embodiment of the present invention. 8 shows an example image of a mammogram after merging the color spaces of the two images from 805 and 811, applying a 50% merging function, and converting to grayscale in step 813.

[0168]

[0233] In some embodiments, the image may be converted to black and white with additional It may be overlaid with additional matrices (layers) that contain image or processing functions or incorporate layers generated from previous processing, such as from high-pass filtering. Features include, but are not limited to, create new, paste, flatten, copy, make adjustment layer, and merge functions.

[0169]

[0234] GI algorithm

[0235] An embodiment of the present invention relating to a GI algorithm is to measure the image (e.g., the original mammogram) The GI visualization provides a visualization designed to isolate, visualize, and characterize dense structures and details in the image (gram) and display them in a grayscale resulting image. Variations within dense breast tissue are reflected in the darker areas of the image. Structures such as cancerous and benign lesions are clearly defined, as are structures within dense areas. In some embodiments, the GI visualization is designed to improve the visibility of abnormalities by a user (e.g., a radiologist).

[0170]

[0236] FIG. 9q illustrates a method for extracting a pixel value from a grayscale image in accordance with at least one embodiment of the present invention. 9 is a flowchart illustrating a method 900 for creating a GI visualization from a

[0237] Figures 9b-9c-9m-9n show the mammograms starting from the original mammogram in Figure 9a. 9e-9f-9o-9p show the results obtained by iteratively applying multiple local micro-contrast convergence transforms. Figures 9e-9f-9o-9p show the results of the same RF transform sequence steps applied to the original gradient grayscale image in 9d.

[0171]

[0238] Figures 9g-9k reveal breast structures in mammography images. 9i, 9j, and 9l show graphical representations in a coordinate system (e.g., created from the PLUTs in FIGS. 9g, 9h, and 9k, respectively).

[0172]

[0239] Referring now to FIG. 9q, in step 901, processor 252 9a shows an exemplary grayscale image of a mammogram, in accordance with at least one embodiment of the present invention.

[0173]

[0240] In step 902, the processor 252 multiplies the original grayscale image. dimensional color space.

[0241] In step 903, the processor 252 calculates one or more pixels in the multi-dimensional color space of the image. or a first set of transfer functions (e.g., a local micro-contrast convergence transfer function GI PLUT1). In these embodiments, the one or more transfer functions are applied to the luminance color space 912 to non-linearly invert the luminance values ​​of the image (e.g., as seen in the GI PLUT1 lookup table of FIG. 9g and the graphical representation of the PLUT in FIG. 9i).

[0174]

[0242] In step 904, processor 252 generates a multi-dimensional color space image shown in FIG. To process the inter-image, a second set of one or more transfer functions (e.g., local micro-contrast convergence functions Figure 9h GI PLUT2) is applied.

[0175]

[0243] FIG. 9c illustrates a method for increasing the density of a mammogram in accordance with at least one embodiment of the present invention. An exemplary image of a mammogram based on a multi-dimensional color space after performing step 904 to further isolate areas is shown.

[0176]

[0244] The process performed in step 904 involves removing discontinuities in the red channel 914. The luminance channel 913 is discontinuously modified while adding color to the image using a low-value nonlinear set of mappings and values ​​in the green channel 915. In these embodiments, the resulting image in Figure 9c shows that low-density tones are colored orange. In some embodiments, low-density red values ​​have values ​​between 174 and 175, depending on their distribution in the original image. High-density areas are bright, and the boundaries of high-density areas are dark.

[0177]

[0245] In step 905, processor 252 performs a multi-dimensional image reconstructing process on the image in FIG. A third set of one or more transfer functions (e.g., local micro-contrast convergence function GI PLUT3) is applied to the color space to produce the image shown in Figure 9m. In these embodiments, the third transfer function is applied to the luminance channel 916 to amplify low, mid, and high values ​​while attenuating values ​​between the amplified values, as seen in Figures 9k and 9l. This significantly separates the tonal values ​​in the resulting image, separating the breast from the background, highlighting possible cancerous areas of the breast, and further defining the core of a possible lesion in blue 918. Values ​​in some lesions have a blue value = 200 + / - 5.

[0178]

[0246] The red channel 917 of the third set of transfer functions is the hue characteristic of the breast 919. is applied to the color space layer to add

[0247] The color image shown in FIG. 9m is then converted to the HSL color space in step 904. In between, the RGB values ​​are converted to luminance, hue, and saturation values. The image can be displayed in step 906 either initially in RGB color or in HSL color space after conversion.

[0179]

[0248] The resulting image (e.g., Figure 9n) is then scanned based on the processed multidimensional color space. This may be displayed in step 907.

[0249] The image in Figure 9m shows the saturation of all hues in the HSL color space set to -1. This is changed in step 908 by setting it to 00% saturation, so that hue is no longer a factor in the presentation of the image.

[0180]

[0250] In step 909, the desaturated HSL color image in FIG. 9m is darkened Using blending, the processed image is merged (blended) with the original image in Figure 9a. If the pixels in the processed image are darker than the pixels on the original image, they are kept in the image. If the pixels in the processed image are lighter, they are replaced with the tone on the original. do.

[0181]

[0251] In step 910, processor 252 determines whether the blend is 6 of its total effect. Adjust the opacity so that it changes to 0%.

[0252] The blended and then merged image is then The image is then converted to a single luminance channel to form a grayscale image such as the one shown in Figure 1. Details in the final image reveal the large cancerous tumor in the upper portion of the breast. The GI local micro-contrast convergence algorithm process revealed the extent of the lesion 920, defined its boundaries, and revealed details within the core of the lesion. Use of other local micro-contrast convergence algorithm sequences implemented herein can then be correlated to the identified areas for further analysis and to differentiate between normal dense tissue, benign, and cancerous lesions.

[0182]

[0253] The image has a luminance value according to the following formula: 0.30*R+0.59* It may be converted to a single channel image containing only luminance in step 911 using a weighted conversion of the R, G, and B values ​​to achieve G+0.11*B=luminance value.

[0183]

[0254] RB algorithm

[0255] The RB algorithm embodiment of the present invention renders the rest of the breast as dark grey. While revealing cancerous tissue, the imaging provides visualization designed to isolate well-defined boundaries and internal structures within dense areas of the breast, including cancer.

[0184]

[0256] FIG. 10w illustrates a grayscale image in accordance with at least one embodiment of the present invention. 10 is a flowchart illustrating a method 1000 for creating an RB visualization from a

[0257] Figures 10b-10c-10m-10n-10s-10t are 10e-10f-10o-10p-10u-10v show the results obtained by iteratively applying multiple local micro-contrast convergence transforms, starting from the original mammogram in Fig. 10e-10f-10o-10p-10u-10v, respectively, showing the results of the same R-B transform sequence steps applied to the original gradient grayscale image as shown in Fig. 10d.

[0185]

[0258] Figures 10g, 10h, 10k, and 10q show the results of the mammography images. Figures 10i, 10j, 10l, and 10r show color values ​​of an RB PLUT (lookup table) optimized to reveal breast structure. Figures 10i, 10j, 10l, and 10r show graphical representations in a coordinate system (e.g., created from the RB PLUTs in Figures 10g-10h, 10k, and 10q, respectively).

[0186]

[0259] In step 1001, processor 252 imports a grayscale image. 10a shows an exemplary grayscale image of a mammogram in accordance with at least one embodiment of the present invention.

[0187]

[0260] In step 1002, the processor 252 converts the original grayscale image into Mapping to a multidimensional color space.

[0261] In step 1003, processor 252 converts the original grayscale image A median filter of radius 3 is applied to the multidimensional color space.

[0188]

[0262] In step 1004, the processor 252 calculates the multidimensional color space of the image of the copy. In these embodiments, a first set of one or more transfer functions (e.g., local micro-contrast convergence transfer function RB PLUT1) is applied to the luminance transfer function 1012 in FIG. 10g, and FIG. 10i). The first set of functions is designed to discontinuously darken the luminance channel 1012 to darken low and medium density area values ​​of the image, as shown in Figures 10b and 10e.

[0189]

[0263] In these embodiments, in step 1005, processor 252 10c shows an example image of a mammogram based on a multi-dimensional color space after applying the second set of one or more transfer functions (e.g., local micro-contrast convergence function RB PLUT2) 10h to the multi-dimensional color space. For example, in FIG. 10j, transfer functions are applied to the luminance 1013, red 1014, and blue 1015 color space layers. FIG. 10c shows an example image of a mammogram based on a multi-dimensional color space after applying the second set of one or more transfer functions, in accordance with at least one embodiment of the present invention.

[0190]

[0264] The luminance channel is modified to increase the contrast of the image. The red channel discontinuously boosts dark areas of the image, reduces highlights, and "flattens" midtones. The blue channel is reduced in value to control tonal values ​​in a color image.

[0191]

[0265] In step 1006, processor 252 performs a multi-step process on the image in FIG. A third set of one or more transfer functions (e.g., third local micro-contrast convergence function RB PLUT3 FIG. 10k and Plot 10l) is applied to the three-dimensional color space to produce the image shown in FIG. 10m. In some embodiments, a transfer function is applied to the luminance channel 1016 to create discontinuous "flat lines" in low-density areas of the image, attenuating mid-tones and slightly reducing high-density luminance values. The red, green, and blue channels 1017 have transfer functions applied that colorize low-density areas of the breast area. In these embodiments, other "color" functions from the third set of transfer functions are applied to the color space layers to add a uniform hue to the breast image.

[0192]

[0266] In step 1007, the colors of the image shown in FIG. 10m are inverted, The resulting images in Figure 10n in mammogram and Figure 10p in gradient are produced.

[0193]

[0267] In step 1008, processor 252 performs the multi-dimensional color A fourth set of one or more transfer functions (e.g., a fourth local micro-contrast convergence function RB PLUT4) 10q is applied to the spatial image, resulting in the image shown in FIG. 10s. FIG. 10r shows that low-density luminance values ​​1018 are forced to a maximum level of 255 for all luminance values ​​<74, with separate peaks for midtones and the brightest areas of the image. The red channel 1019 attenuates low densities and maximizes high densities, with a value set to 255 for all luminance values ​​>160. The green channel 1020 is responsible for the hue of the background and breast tissue. In these embodiments, RB PLUT4 FIG. 10q, also shown diagrammatically in FIG. 10r, is applied to the luminance color space to distinguish the breast from outside its boundaries.

[0194]

[0268] In step 1009, the color image shown in FIG. 10s is converted into the HSL color space and the RGB values ​​are converted to luminance, hue, and saturation values. The image can be displayed initially in RGB color or in HSL color space after conversion in step 1010. An exemplary HSL color space conversion is as follows:

[0195] (Hue, Saturation, Brightness, Zone) (0.0,-1.0,-0.3,magenta) (0.0,-1.0,0.3,red) (0.0,-1.0,-0.4,yellow) (0.0,-1.0,-0.4,cyan) (0.0,-1.0,0.2,blue) (0.0,-1.0,-0.1,green)

[0269] The final image in Figure 10t is a matrix of all hues in the HSL color space. This is created from the image in Figure 10s by setting the data saturation to -100% saturation. As a result, hue is no longer a factor in the representation of the image. However, the luminance values ​​are still adjustable, and changing the luminance values ​​of various hues in the color space can change the grayscale representation of those values.

[0196]

[0270] In step 1011, the image is extracted as a single channel image containing only luminance. In this embodiment, all areas of non-pathology are revealed in a uniform gray 1021 of the breast image area, which may have an average luminance value of 130. This separation of possible areas of abnormality 1022 reduces the radiologist's "dwell time," i.e., the time they must spend examining all areas of the image to locate areas with the highest probability of cancer occurring.

[0197]

[0271] Consistency of local micro-contrast convergence algorithms

[0272] 11a-11d illustrate how an embodiment of the present application can be used across different imaging modalities. The results demonstrate the consistency achieved by the local micro-contrast convergence technique. The pattern responses for breast images reveal consistent color and tissue characterization for modality 3D tomosynthesis in Fig. 11a, 3D-to-2D synthesis in Fig. 11b, full-field digital mammography (FFDM) in Fig. 11c, and digitized film in Fig. 11d. This gives radiologists and their patients the ability to compare changes over time using only one set of algorithms, even when patient images were historically generated using different imaging modalities. These results validate one of the inherent capabilities of the local micro-contrast convergence technique, as shown in the local micro-contrast convergence hierarchy of features identified as modality fusion in Fig. 1d.

[0198]

[0273] FIG. 11e is an example generated from excised breast tissue removed during surgery. 11A shows an X-ray view of a cancer in a typical mammogram image. FIG. 11F shows an exemplary mammogram image after processing the image using one or more methods described herein. The original image was processed using the CI algorithm previously described herein. The black and magenta boundaries of the cancer 1101 are clearly defined, as are the color variations inside boundary 1102, which indicate the spread of cancer development.

[0199]

[0274] The embodiments of the invention described herein provide a method for detecting, in context, These methods utilize a multi-algorithm, multi-dimensional, computer-based process for feature visualization and characterization. These local micro-contrast focusing methods are applicable in applications where features are less than 1 mm in size, less than 900 microns in size, less than 850 microns in size, less than 800 microns in size, or less than 750 microns in size.

[0200]

[0275] FIG. 11g shows a mammography x-ray known to contain cancer 1103. A close-up view is shown. Figure 11h shows an exemplary mammogram image after processing the image using one or more of the methods described herein. In Figure 11h, the black border of the cancer 1104 using the CI algorithm process previously described in Figures 4a-4k is clearly defined, as are the details inside the cancer's core. The progression from yellow to red to blue within the cancer is evident in the small size of the blue core 1105, which is only 980 microns in size. The multiple algorithmic representations of the present invention provide different characterizations and visualizations of the same tissue.

[0201]

[0276] These methods are suitable for applications where the feature of interest is located within another feature, or where the feature of interest is of a certain size. The present invention is applicable to applications where the first feature is less than 900 microns in size, less than 850 microns in size, less than 800 microns in size, or less than 750 microns in size, and even in applications where the first feature is 1 mm in size or larger. In some embodiments, the feature of interest is between 700 and 900 microns in size.

[0202]

[0277] In some embodiments, structures as small as 750 μm (microns) can be Using the methods described above, cancer cores can be identified. Based on X-ray images in which pixels represent dimensions of breast tissue that are 75 μm in size, cancer cores can be described and characterized in sizes ranging from 75 μm to 1 mm. Through clinical testing, it has been determined that structures as small as 50 μm (microns) can be revealed or distinguished in images with pixel dimensions of 50 μm or less. Thus, various forms of cancer, as well as ductal carcinoma in situ and precancerous atypical hyperplasia, have been revealed using these methods.

[0203]

[0278] Alternative Embodiments - Different Processing Combinations

[0279] The previous paragraph describes different methods for image visualization of local micro-contrast convergence. Although embodiments are described, those skilled in the art will appreciate that one or more of the processing steps performed in one embodiment may be applied in any order and / or to other embodiments, including, but not limited to, gamma level adjustment or equalization, convolution filtering, sharpening filters, smoothing filters, median filters, high-pass filters, low-pass filters, merge functions, image multiplication functions, image subtraction functions, image addition functions, image blending functions, wavelet functions, and image layering functions, among others described herein.

[0204]

[0280] Alternative Embodiments - Different Modalities

[0281] Embodiments of the present invention are particularly, but not exclusively, used in medical imaging (e.g., mammography). The present invention has applicability to several different fields, including imaging, imaging, and forensic science (e.g., imaging, imaging, imaging technology ...

[0205]

[0282] Alternative Embodiments - Cancer / Diseases

[0283] Embodiments of the present invention include, but are not limited to, the prostate, kidney, liver, and liver tissue in both humans and animals. The present invention has applicability to visualizing, characterizing, and detecting several different cancers, including those of the liver, bone, lung, brain, and skin. Those skilled in the art will also appreciate that embodiments of the present invention are not limited to the cancers described herein, but instead are applicable to other similar cancers.

[0206]

[0284] Embodiments of the present invention may be used in a variety of applications, including, but not limited to, cardiovascular disease, arsenic in retinal scans, and the like. It has applicability to detecting several different diseases, including detection of Zheimer's disease, eye diseases, multiple sclerosis lesion mapping, and photomicroscopy. Those skilled in the art will also appreciate that the present invention is not limited to the lesions described herein, but is instead applicable to other similar lesions.

[0207]

[0285] Embodiments for Improving False Positive / False Negative Rates

[0286] For example, the micro-contrast focusing described herein in medical applications Applying one or more of the algorithms produces image visualizations that facilitate a user (e.g., a radiologist) in detecting structures of interest (e.g., cancer), resulting in significantly reduced false positive and false negative rates.

[0208]

[0287] In some embodiments, the breast is a dense breast over a series of 100 trials. The false positive rate for detecting breast cancer in mammogram images, where the breast contains at least a portion of dense breast tissue, is less than 10% as determined by a physician. In some embodiments, the false positive rate for detecting breast cancer in mammogram images, where the breast contains at least a portion of dense breast tissue, over a series of 100 trials is less than 5% as determined by a physician. In some embodiments, the false positive rate for detecting breast cancer in mammogram images, where the breast contains at least a portion of dense breast tissue, over a series of 100 trials is less than 1% as determined by a physician.

[0209]

[0288] In some embodiments, the breast is a dense breast over a series of 100 trials. The false negative rate for breast cancer detection in mammogram images, where the breast includes at least a portion of dense breast tissue, is less than 60% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast includes at least a portion of dense breast tissue, over a series of 100 trials is less than 50% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast includes at least a portion of dense breast tissue, over a series of 100 trials is less than 45% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast includes at least a portion of dense breast tissue, over a series of 100 trials is less than 40% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast includes at least a portion of dense breast tissue, over a series of 100 trials is less than 35% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 30% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 25% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 20% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 15% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 10% as determined by a physician.In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 5% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 4% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 3% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images, where the breast comprises at least a portion of dense breast tissue, over a series of 100 trials is less than 3% as determined by a physician. The false negative rate for detecting breast cancer in mammogram images where the breast includes at least a portion of dense breast tissue is less than 2% as determined by a physician. In some embodiments, the false negative rate for detecting breast cancer in mammogram images where the breast includes at least a portion of dense breast tissue over a series of 100 trials is less than 1% as determined by a physician.

[0210]

[0289] In some embodiments, the breasts are compared to normal breast sets over a series of 100 trials. The false negative rate for breast cancer detection in mammogram images where the breast contains normal breast tissue is less than 16% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images where the breast is normal breast tissue over a series of 100 trials is less than 15% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images where the breast contains normal breast tissue over a series of 100 trials is less than 10% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images where the breast contains normal breast tissue over a series of 100 trials is less than 5% as determined by a physician. In some embodiments, the false negative rate for breast cancer detection in mammogram images where the breast contains normal breast tissue over a series of 100 trials is less than 1% as determined by a physician.

[0211]

[0290] Feature Extraction

[0291] By implementing embodiments of the present invention, (e.g., by a radiologist) Images are generated that visualize and characterize tissue structures in an enhanced manner that improves feature discrimination.

[0212]

[0292] In some embodiments, the processor 252 performs the following steps to identify the cancerous structures: One or more computer-aided detection (CAD) techniques may be implemented on one or more generated image visualizations. Large-scale pattern recognition systems applicable to millions of information features may include primary, secondary, and tertiary image analysis of such features, which may employ image comparisons (e.g., between known cancerous structures and portions of the image visualization).

[0213]

[0293] The local micro-contrast convergence algorithm method used in this application The process allows such tissue types in images, such as mammograms, to adopt characteristic color and grayscale properties that uniquely characterize the tissues and their boundaries, making feature identification and extraction highly effective for accurate identification. These properties include, but are not limited to, morphology, geometry, color, texture, relationships between different tissue structures (such as correlating the presence of a lesion with microcalcifications in breast tissue), shape of the lesion boundary, presence of spicules, edge gradient, cumulative edge gradient distribution, architectural distortion, color distribution within the lesion, contrast, temporal stability (variation between mammographic exams), and correlation of features between different views (multiple view correlation between CC and MLO mammographic image views).

[0214]

[0294] Machine learning processes in the breast cancer detection domain include benign cysts, fibroadenomas, carcinomas, and The method begins by extracting features that correlate with diseases such as cancer and invasive cancer. A training set of images is used to develop a benchmark for comparison between cancerous and non-cancerous areas of mammograms.

[0215]

[0295] The relevant features are then extracted from the local micro- images from a given coordinate area in each processed image. The micro-contrast convergence process extracts tissue characterization patterns as clusters of pixel luminance and color values. Multiple local micro-contrast convergence processed images can be analyzed, and features can be extracted from each of the separate images created through one or more visualization algorithm sequences described herein. All processed images examined can be superimposed, so there is perfect alignment of the area of ​​interest between different processed images.

[0216]

[0296] In some embodiments, the processor 252 performs the following steps to identify the feature of interest: One or more nonlinear transfer functions may be generated for application to the image. In these embodiments, the processor 252 may run different trials with different sets of local micro-contrast convergence transfer functions used for each trial. In some embodiments, the local micro-contrast convergence transfer functions may be randomly generated. In some embodiments, the local micro-contrast convergence transfer functions are generated based on a default function (e.g., a trigonometric function). An example for generating local micro-contrast convergence transfer functions based on a default function is shown in FIG. 2a.

[0217]

[0297] Luminance values ​​available for mapping into this coordinate plot The range of values ​​for the nonlinear transfer functions is infinite. As a result of the trials, the processor 252 may select a preferred set of nonlinear transfer functions to apply to the image based on the lowest probability of false positives and / or false negatives.

[0218]

[0298] Feature analysis includes co-occurrence, Gabor, SIFT, LBP, Histogram of Oriented Gradient, and R. This may include high separability feature extraction (HSFE) from data based on standard and progressive properties of images and time series, including Geometric Ferns, and Hough Forests.

[0219]

[0299] Machine learning, data mining, and statistical modeling techniques are used in Adaboo It can be applied for real-time object recognition and localization in images processed using processes such as st, genetic programming, support vector machines, neural networks, global optimization, and learning vector quantization.

[0220]

[0300] There is a theoretical limit to the number of features that can be extracted or the number of correlations that can be made between them. The sky's the limit: algorithm development can be employed for big data applications using R, Pig, Storm, MySQL, MongoDB, and Hadoop.

[0221]

[0301] *****

[0302] In at least one embodiment, one or more processors and memory ( For example, one or more computers having one or more non-volatile storage devices. In some embodiments, the memory or computer-readable storage medium of the memory stores programs, modules, and data structures, or a subset thereof, for a processor to control and operate the various systems and methods disclosed herein. In one embodiment, a non-transitory computer-readable storage medium having stored thereon computer-executable instructions that, when executed by a processor, perform one or more of the methods disclosed herein.

[0222]

[0303] From the broad inventive concept thereof, the exemplary embodiments shown and described above Those skilled in the art will appreciate that modifications may be made without departing from the spirit and scope of the present invention. Accordingly, it will be understood that the present invention is not limited to the exemplary embodiments shown and described, but encompasses modifications within the spirit and scope of the present invention as defined by the claims. For example, certain features of the exemplary embodiments may not be a part of the claimed invention, and features of the disclosed embodiments may be combined. Unless otherwise stated herein, the terms "a," "an," and "the" are not limited to one element, but instead should be read to mean "at least one."

[0223]

[0304] At least some of the drawings and descriptions of the invention provide a clear understanding of the invention. It should be understood that the description has been simplified to focus on elements relevant for the purpose, and for clarity has excluded other elements that one skilled in the art would appreciate and that may also comprise part of the present invention. However, because such elements are well known in the art, and because they do not necessarily facilitate a better understanding of the present invention, descriptions of such elements are not provided herein.

[0224]

[0305] Furthermore, to the extent that the method does not rely on the particular order of steps described herein, Thus, the particular order of steps should not be construed as a limitation on the scope of the claims. Those skilled in the art will readily appreciate that the claims directed to the present methods of the present invention are not limited to the performance of those steps in the order written, and that the steps can be varied and still remain within the spirit and scope of the present invention.

Claims

1. 1. A method for visualizing and characterizing features in an image, comprising:

1. A method comprising: applying a first local micro-contrast convergence algorithm to a first image to generate a second image that is separate from the first image, the second image including the features; applying the first local micro-contrast convergence algorithm includes iteratively and sequentially applying each of a plurality of nonlinear discontinuous transfer functions (121) to a corresponding image in a sequential set of images, the sequential set including the first image and one or more intermediate images; local micro-contrast convergence represents a sequence of transfer functions including the two or more nonlinear discontinuous transfer functions (121), the sequence of transfer functions being employed to aggregate relationships between neighboring pixel groups into a predefined color and luminosity pattern; each subsequent image in the sequential set of images is generated by applying one of the plurality of nonlinear discontinuous transfer functions (121) to a previous image in the sequential set of images, thereby resulting in the generation of the second image that distinguishes and reveals visualized and characterized patterns of two or more physical or biological structures that were present but not distinguished in the first image; applying one or more non-linear discontinuous transfer functions (121) to the first image includes utilizing one or more grayscale or color profile lookup tables that represent the non-linear discontinuous transfer functions (121); the profile lookup table is a mathematical representation in a matrix that specifies image input and output values ​​such that a locally self-similar image contrast pixel variable of the first image has a discrete set of values; the one or more pixels in the first image that have similar characteristics are assigned a single color or luminance value in the resulting output image; the two or more nonlinear discontinuous transfer functions (121) including the at least one discontinuity cause a tissue feature distinction to be included in the visualized and characterized pattern of the two or more physical or biological structures; the two or more nonlinear discontinuous transfer functions (121) including the at least one discontinuity cause the visualized and characterized patterns of the two or more physical or biological structures to exhibit a boundary line distinguishing between cancerous and benign tissue structures in the second image; the two or more nonlinear discontinuous transfer functions (121) including the at least one discontinuity cause the visualized and characterized patterns of the two or more physical or biological structures to include internal morphology of cancerous and / or benign tissue structures in the second image; method.

2. applying a second local micro-contrast convergence algorithm, distinct from the first local micro-contrast convergence algorithm, to the first image to generate a third image, distinct from the first image and the second image. The method of claim 1 further comprising:

3. combining one or more of the first image, the second image, the third image, or the fourth image to generate a fifth image that is distinct from the original image, the first image, the second image, the third image, or the fourth image. The method of claim 1 further comprising:

4. 2. The method of claim 1, wherein applying the first local micro-contrast convergence algorithm comprises applying one or more nonlinear discontinuous transfer functions (121) to the first image.

5. the first image is a grayscale image having pixel values, and the method comprises: replicating the pixel values ​​of the grayscale image into a first multi-dimensional color space, each dimension of the first multi-dimensional color space being a replica of a pixel value of the grayscale image; The method of claim 1 further comprising:

6. 6. The method of claim 5, wherein the first multidimensional color space includes four dimensions including four different components, namely, luminance, red, green, and blue, and the second image is in an RGB multidimensional color space including luminance and three different color dimensions, namely, luminance, red, green, and blue.

7. The method of claim 1 , wherein the first image and the second image are multi-dimensional color space images.

8. 8. The method of claim 7, wherein converting the second multi-dimensional color space image to a single-dimensional grayscale image comprises modifying the luminance value of each color space of each pixel in the second multi-dimensional color space image to convert to the single-dimensional grayscale image.

9. 10. The method of claim 1, wherein the first image is an image generated by x-ray, ultrasound, infrared, ultraviolet, MRI, CT scan, PET scan, grayscale, color, visible light, mm wave, or laser scan.

10. 10. The method of claim 1, wherein the characteristic is cancer of the breast, prostate, kidney, liver, bone, lung, brain, or skin.

11. 10. The method of claim 1, wherein the feature is a biomarker for cardiovascular disease, Alzheimer's disease, eye disease, or multiple sclerosis pathology.

12. The method of claim 1 , wherein the feature is a chemical marker for a solid or liquid organic compound.

13. The method of claim 1 , wherein the feature is a structural defect.

14. The method of claim 10 , wherein the features are 1 mm or larger in size.

15. the first image contains a first feature of interest, the first feature of interest contains a second feature; the second feature within the first feature is 700-900 micrometers in size, and the first feature is 1 mm or larger in size; further comprising the step of displaying a second feature within the first feature. The method of claim 1.

16. The method of claim 15, wherein the first image has a resolution of 35 to 125 micrometers per pixel.

17. applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is distinct from the first image; receiving the first image; mapping pixel values ​​of the first image into an initial multi-dimensional color space; applying one or more nonlinear transfer functions to the initial multi-dimensional color space to induce local micro-contrast convergence and create a processed multi-dimensional color space; displaying an image visualization of the second image based on the processed multi-dimensional color space; 17. The method of claim 16, comprising:

18. 18. The method of claim 17, wherein the multi-dimensional color space is one of an RGB, HSV, HLS, HSB, XYZ, CMYK, CIEXYZ, or CIELAB color space.

19. applying a median filter to the initial multi-dimensional color space; applying the one or more non-linear transfer functions (121) to the initial multi-dimensional color space, applying a first set of nonlinear transfer functions to attenuate low density fatty breast tissue; applying a second set of non-linear transfer functions to make fatty breast tissue appear as a first color and to distinguish said denser breast tissue using other colors; applying a third set of non-linear transfer functions to amplify low pixel values ​​and attenuate high pixel values ​​in the color space layer associated with the first color; applying a fourth set of non-linear transfer functions to change the background of said image to black when displayed; 18. The method of claim 17, comprising:

20. adjusting a gamma level of the multi-dimensional color space to adjust the contrast of the first image and highlight structural details; applying the nonlinear transfer function to the initial multi-dimensional color space associated with the first image, applying a first set of non-linear transfer functions to reduce luminance levels in the multi-dimensional color space; applying a second set of non-linear transfer functions to invert values ​​in the initial multi-dimensional color space related to luminance; Including, 18. The method of claim 17.

21. 18. The method of claim 17, wherein the received image is an image generated by x-ray, ultrasound, infrared, ultraviolet, MRI, CT scan, PET scan, grayscale, color, visible light, mm wave, or laser scan.

22. 18. The method of claim 17, wherein the received image includes a feature, and the feature is cancer of the breast, prostate, kidney, liver, bone, lung, brain, or skin.

23. 18. The method of claim 17, wherein the received image includes features, the features being biomarkers for cardiovascular disease, Alzheimer's disease, eye disease, or multiple sclerosis.

24. 20. The method of claim 17, wherein the received image includes features, the features being chemical markers for solid or liquid organic compounds.

25. The method of claim 17 , wherein the received image includes a feature, the feature being a structural defect.

26. the first image is a mammogram including a representation of dense breast tissue and fatty breast tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is distinct from the first image, the second image comprising: mapping pixel values ​​of the first image into a first multi-dimensional color space; applying a median filter to the first multi-dimensional color space to generate a second multi-dimensional color space; inverting the values ​​of the second multi-dimensional color space to generate a third multi-dimensional color space; applying a first set of one or more non-linear transfer functions to the third multidimensional color space to generate a fourth multidimensional color space, making the fatty breast tissue appear as one color and distinguishing the dense breast tissue using another color; applying a second set of one or more transfer functions to the fourth multidimensional color space to generate a fifth multidimensional color space, amplifying high pixel values, attenuating low pixel values, and highlighting breast area structures; displaying an image visualization based on the fifth multidimensional color space; Including, The method of claim 1.

27. the first image is a mammogram including a representation of dense breast tissue and fatty breast tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is distinct from the first image, the second image comprising: mapping pixel values ​​of the first image into a first multi-dimensional color space; applying a first set of one or more transfer functions to the first multidimensional color space to generate a second multidimensional color space, making the fatty breast tissue appear as one color and distinguishing the dense breast tissue using another color; converting the second multidimensional color space into a third multidimensional color space in the HLS color space; displaying an image visualization based on the third multidimensional color space; The method of claim 1 , comprising:

28. the first image is a mammogram including a representation of dense breast tissue and fatty breast tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is distinct from the first image, the second image comprising: mapping pixel values ​​of the first image into a first multi-dimensional color space; applying a first set of one or more transfer functions to the first multidimensional color space to generate a second multidimensional color space, making the fatty breast tissue appear as one color and using another color to distinguish and reveal detailed structures in the dense breast tissue; displaying an image visualization based on the second multi-dimensional color space; The method of claim 1 , comprising:

29. the first image is a mammogram including a representation of dense breast tissue and fatty breast tissue, and applying a first local micro-contrast convergence algorithm to the first image to generate a second image that is distinct from the first image, the second image comprising: mapping pixel values ​​of the first image into a first multi-dimensional color space; applying a first set of one or more transfer functions to the first multidimensional color space to generate a second multidimensional color space, making the fatty breast tissue appear translucent and distinguishing dense breast tissue using other colors, and distinguishing small dot-like structures in the fatty and dense breast tissue; displaying an image visualization based on the second multi-dimensional color space; The method of claim 1 , comprising:

30. one or more memory units each operable to store at least one program; and at least one processor communicatively coupled to the one or more memory units, wherein the at least one program, when executed by the at least one processor, causes the at least one processor to perform a method as recited in any of claims 1 to 29.

31. A non-transitory computer-readable storage medium having stored thereon computer-executable instructions which, when executed by a processor, perform the method of any of claims 1 to 29.

Images