WO2023119904A1 - 画像処理装置、画像処理方法、及び画像処理プログラム - Google Patents

画像処理装置、画像処理方法、及び画像処理プログラム Download PDF

Info

Publication number
WO2023119904A1
WO2023119904A1 PCT/JP2022/041084 JP2022041084W WO2023119904A1 WO 2023119904 A1 WO2023119904 A1 WO 2023119904A1 JP 2022041084 W JP2022041084 W JP 2022041084W WO 2023119904 A1 WO2023119904 A1 WO 2023119904A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
calcification
distribution
image processing
low
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/JP2022/041084
Other languages
English (en)
French (fr)
Japanese (ja)
Inventor
湧介 町井
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fujifilm Corp
Original Assignee
Fujifilm Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fujifilm Corp filed Critical Fujifilm Corp
Priority to EP22910624.0A priority Critical patent/EP4454566A4/en
Priority to JP2023569137A priority patent/JPWO2023119904A1/ja
Publication of WO2023119904A1 publication Critical patent/WO2023119904A1/ja
Priority to US18/747,369 priority patent/US20240338825A1/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
    • A61B6/502—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of breast, i.e. mammography
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
    • A61B6/5217—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data extracting a diagnostic or physiological parameter from medical diagnostic data
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/54—Control of apparatus or devices for radiation diagnosis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10116—X-ray image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30068—Mammography; Breast
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30096—Tumor; Lesion

Definitions

  • the present disclosure relates to an image processing device, an image processing method, and an image processing program.
  • Japanese Unexamined Patent Application Publication No. 2018-161405 describes a technique for intuitively determining a portion where a mammary gland region and a minute circular calcification region overlap.
  • Japanese Patent Application Laid-Open No. 2016-22143 describes a technique for intuitively grasping the dense state of minute circular calcified tissues.
  • the present disclosure has been made in view of the above circumstances, and provides an image processing device, an image processing method, and an image processing program capable of accurately determining the type of distribution state of calcification in the breast.
  • the image processing apparatus of the first aspect of the present disclosure includes at least one processor, the processor detects calcification from a radiographic image obtained by irradiating radiation and imaging the breast, and detects calcification.
  • a low-resolution image having a resolution lower than that of the radiographic image is generated from the calcification distribution image representing the result, and the type of calcification distribution state is determined from the low-resolution image.
  • An image processing apparatus is the image processing apparatus according to the first aspect, wherein the calcification distribution image is a grayscale image or a binary image.
  • An image processing apparatus is the image processing apparatus according to the first aspect, wherein the calcification distribution image is a mask image in which areas other than the calcification region are masked.
  • An image processing apparatus is the image processing apparatus according to any one of the first to third aspects, wherein the processor enlarges the calcification image included in the calcification distribution image.
  • the calcification image is configured to generate a low resolution image from the magnified calcification distribution image.
  • An image processing apparatus is the image processing apparatus according to any one of the first to fourth aspects, wherein the processor converts the calcification distribution image into a calcification signal value based filtering and generating a low resolution image from the filtered calcification distribution image.
  • An image processing apparatus is the image processing apparatus according to the fifth aspect, wherein the processor derives the mammary gland volume of the breast from the radiation image, and sets a threshold for filtering based on the mammary gland volume. Configured.
  • An image processing apparatus is the image processing apparatus according to any one of the first to sixth aspects, wherein the processor extracts at least one of breast skin lines and nipples from the radiographic image. Configured to detect, the calcification distribution image includes at least one of the detected skin lines and nipples.
  • An image processing apparatus is the image processing apparatus according to any one aspect of the first aspect to the seventh aspect, wherein if the calcification distribution image includes a plurality of distributions, the processor , to determine the type of distribution state for each distribution.
  • An image processing apparatus is the image processing apparatus according to any one aspect of the first to eighth aspects, wherein the radiographic image is obtained by imaging the left breast of the subject.
  • a left breast radiographic image and a right breast radiographic image obtained by imaging the right breast of the subject are included, and the processor detects calcification from each of the left breast radiographic image and the right breast radiographic image,
  • a left breast low-resolution image is generated from the left breast calcification distribution image representing the detection result of breast calcification
  • a right breast low-resolution image is generated from the right breast calcification distribution image representing the detection result of right breast calcification. and determine the type of calcification distribution from the left breast low resolution image and the right breast low resolution image.
  • An image processing device is the image processing device according to any one aspect of the first aspect to the ninth aspect, wherein the processor determines the type of calcification shape from the low-resolution image. configured as follows.
  • An image processing apparatus is the image processing apparatus according to the tenth aspect, wherein the processor, based on the type of distribution state of calcification and the type of shape of calcification, determines the degree of malignancy of calcification is configured to estimate
  • An image processing apparatus is the image processing apparatus according to any one aspect of the first to eleventh aspects, wherein the processor determines whether the calcification is benign or malignant from the low-resolution image. is configured to estimate whether there is
  • An image processing device is the image processing device according to any one aspect of the first aspect to the twelfth aspect, wherein the processor superimposes the determination result of the type of distribution state on the radiographic image. configured to display
  • An image processing apparatus is the image processing apparatus according to any one of the first to thirteenth aspects, wherein the processor uses a rule-based calcification detection model from the radiographic image to , configured to detect calcification.
  • An image processing apparatus is the image processing apparatus according to any one of the first to thirteenth aspects, wherein the processor uses a learning-based calcification detection model from the radiographic image to , configured to detect calcification.
  • An image processing apparatus is the image processing apparatus according to any one aspect of the first to fifteenth aspects, wherein the radiation image is obtained by normal radiography of the subject's breast. a dimensional image, a plurality of tomographic images obtained from a series of projection images obtained by tomosynthesis of the breast, or a composite obtained by combining at least part of a series of projection images or a plurality of tomographic images It is a two-dimensional image.
  • calcification is detected from a radiographic image obtained by irradiating radiation and imaging a breast, and from a calcification distribution image representing the detection result of calcification, is an image processing method in which a computer executes a process of generating a low-resolution image whose resolution is lower than that of a radiographic image and determining the type of calcification distribution state from the low-resolution image.
  • the image processing program detects calcification from a radiographic image obtained by irradiating radiation and imaging a breast, and detects calcification from a calcification distribution image representing the detection result of calcification. , to generate a low-resolution image lower in resolution than the radiographic image, and to cause the computer to execute a process of determining the type of calcification distribution state from the low-resolution image.
  • FIG. 1 is a configuration diagram schematically showing an example of the overall configuration of a radiographic imaging system according to an embodiment
  • FIG. FIG. 4 is a diagram for explaining an example of tomosynthesis imaging
  • 1 is a block diagram showing an example of the configuration of an image processing apparatus according to an embodiment
  • FIG. It is a figure for demonstrating the classification and kind of calcification.
  • FIG. 4 is a diagram for explaining an example of a calcification distribution determination model; It is a figure for demonstrating a convolution process.
  • FIG. 4 is a diagram for explaining an example of machine learning of a calcification distribution determination model;
  • FIG. 4 is a schematic diagram for explaining an overview of the flow of determining the type of calcification distribution state in the image processing apparatus of the embodiment;
  • 1 is a functional block diagram of an example configuration of an image processing apparatus according to an embodiment;
  • FIG. 4 is a flowchart showing an example of the flow of image processing by the image processing apparatus of the embodiment;
  • FIG. 10 is a diagram showing an example of a display of a determination result of a type of calcification distribution state;
  • FIG. 10 is a diagram showing an example of a display of a determination result of a type of calcification distribution state;
  • FIG. 11 is a block diagram showing an example of a storage unit of an image processing apparatus according to modification 1;
  • FIG. 11 is a diagram for explaining an example of learning by machine learning of the calcification detection model of modification 1;
  • FIG. 11 is a schematic diagram for explaining an overview of the flow of determination of the type of calcification distribution state in the image processing apparatus of Modification 2;
  • 10 is a flow chart showing an example of the flow of image processing by the image processing apparatus of modification 2;
  • FIG. 12 is a schematic diagram for explaining an overview of the flow of determination of the type of calcification distribution state in the image processing apparatus of Modification 3;
  • 10 is a flow chart showing an example of the flow of image processing by the image processing apparatus of modification 3;
  • FIG. 11 is a schematic diagram for explaining another example of the flow of determination of the type of calcification distribution state in the image processing apparatus of Modification 3;
  • FIG. 12 is a schematic diagram for explaining an overview of the flow of determination of the type of calcification distribution state in the image processing apparatus of Modification 4;
  • FIG. 12 is a schematic diagram for explaining an overview of the flow of determination of the type of calcification distribution state in the image processing apparatus of Modification 5;
  • FIG. 20 is a schematic diagram for explaining an overview of the flow of determination of the type of calcification distribution state in the image processing apparatus of modification 6;
  • FIG. 21 is a schematic diagram for explaining an outline of another example of the flow of determination of the type of calcification shape in the image processing apparatus of modification 7;
  • FIG. 4 is a diagram for explaining an example of machine learning of a calcification shape determination model;
  • 19 is a flow chart showing an example of the flow of image processing by the image processing apparatus of modification 7;
  • FIG. 21 is a schematic diagram for explaining another example of the outline of the flow of estimating whether calcification is benign or malignant in the image processing apparatus of modification 8;
  • FIG. 4 is a diagram for explaining an example of machine learning of a benign/malignant presumed model;
  • 16 is a flow chart showing an example of the flow of image processing by the image processing apparatus of modification 8;
  • FIG. 1 shows a configuration diagram showing an example of the overall configuration of a radiographic imaging system 1 of this embodiment.
  • the radiation imaging system 1 of this embodiment includes a mammography apparatus 10 , a console 12 and an image processing apparatus 16 .
  • the console 12 and the image processing device 16 are connected via a network 17 by wired communication or wireless communication.
  • FIG. 1 shows a side view showing an example of the appearance of a mammography apparatus 10 of this embodiment. Note that FIG. 1 shows an example of the appearance of the mammography apparatus 10 when viewed from the left side of the subject.
  • the mammography apparatus 10 of the present embodiment operates under the control of the console 12, irradiates the breast of the subject with radiation R (for example, X-rays) from the radiation source 29, and produces a radiation image of the breast. It is a device that shoots
  • the mammography apparatus 10 of the present embodiment performs normal imaging in which the radiation source 29 is positioned along the normal direction of the detection surface 20A of the radiation detector 20, and the radiation source 29 is positioned at each of a plurality of irradiation positions. It has a function of performing so-called tomosynthesis imaging (details will be described later), in which imaging is performed by moving the camera to the
  • the mammography apparatus 10 includes an imaging table 24, a base 26, an arm portion 28, and a compression unit 32.
  • a radiation detector 20 is arranged inside the imaging table 24 .
  • the user positions the breast U of the subject on the imaging surface 24A of the imaging table 24 when performing imaging.
  • the radiation detector 20 detects radiation R that has passed through the breast U, which is a subject. Specifically, the radiation detector 20 detects the radiation R that enters the subject's breast U and the imaging table 24 and reaches the detection surface 20A of the radiation detector 20, and based on the detected radiation R An image is generated and image data representing the generated radiographic image is output.
  • imaging a series of operations in which radiation R is emitted from the radiation source 29 and a radiographic image is generated by the radiation detector 20 may be referred to as "imaging".
  • imaging a series of operations in which radiation R is emitted from the radiation source 29 and a radiographic image is generated by the radiation detector 20 may be referred to as "imaging".
  • the type of the radiation detector 20 of the present embodiment is not particularly limited. A direct conversion type radiation detector that directly converts R into charge may be used.
  • a compression plate 30 used for compressing the breast during imaging is attached to a compression unit 32 provided on the imaging table 24, and a compression plate driving section (not shown) provided on the compression unit 32 is used to perform imaging. It is moved toward or away from the platform 24 (hereinafter referred to as "vertical direction").
  • the compression plate 30 compresses the breast of the subject with respect to the imaging table 24 by moving in the vertical direction.
  • the arm part 28 can be rotated with respect to the base 26 by the shaft part 27 .
  • the shaft portion 27 is fixed to the base 26, and the shaft portion 27 and the arm portion 28 rotate together.
  • the shaft portion 27 and the compression unit 32 of the imaging table 24 are respectively provided with gears, and by switching between the engaged state and the non-engaged state of the gears, the compression unit 32 of the imaging table 24 and the shaft portion 27 are connected. It is possible to switch between a state in which the shaft portion 27 rotates integrally and a state in which the shaft portion 27 is separated from the imaging table 24 and idles. Switching between transmission and non-transmission of the power of the shaft portion 27 is not limited to the gear described above, and various mechanical elements can be used.
  • the arm portion 28 and the imaging table 24 are separately rotatable relative to the base 26 with the shaft portion 27 as a rotation axis.
  • the radiation source 29 When tomosynthesis imaging is performed in the mammography apparatus 10, the radiation source 29 is sequentially moved to each of a plurality of irradiation positions with different irradiation angles by rotating the arm portion 28.
  • FIG. The radiation source 29 has a radiation tube (not shown) that generates radiation R, and the radiation tube is moved to each of a plurality of irradiation positions according to the movement of the radiation source 29 .
  • FIG. 2 shows a diagram for explaining an example of tomosynthesis imaging. 2, illustration of the compression plate 30 is omitted.
  • the maximum value is 7
  • the irradiation angle of the radiation R with respect to the detection surface 20A of the radiation detector 20 is moved to a different position.
  • the radiation source 29 irradiates the breast U with the radiation R according to the instruction from the console 12, and the radiation detector 20 captures a radiographic image.
  • the radiographic imaging system 1 when tomosynthesis imaging is performed by moving the radiation source 29 to each of the irradiation positions 19t and radiographic images are captured at each irradiation position 19t , in the example of FIG. An image is obtained.
  • radiographic image captured at each irradiation position 19 in tomosynthesis imaging will be referred to as a "projection image” when distinguished from other radiographic images.
  • projection image radiation images
  • radiation images such as tomographic images and normal two-dimensional images, which will be described later, are collectively referred to simply as “radiation images” regardless of their types.
  • the irradiation angle of the radiation R refers to the angle ⁇ between the normal CL of the detection surface 20A of the radiation detector 20 and the radiation axis RC.
  • the radiation axis RC refers to an axis that connects the focal point of the radiation source 29 at each irradiation position 19 and a preset position such as the center of the detection surface 20A.
  • the detection surface 20A of the radiation detector 20 is assumed to be substantially parallel to the imaging surface 24A.
  • the radiation source 29 is placed at an irradiation position 19 t where the irradiation angle ⁇ is 0 degrees (irradiation position 19 t along the normal direction, irradiation position 19 4 in FIG. 2). is left as is.
  • Radiation R is emitted from the radiation source 29 according to an instruction from the console 12 , and a radiographic image is captured by the radiation detector 20 .
  • a radiographic image obtained by normal imaging is referred to as a "normal two-dimensional image" when distinguishing it from other radiographic images such as projection images.
  • the mammography apparatus 10 and console 12 are connected by wired communication or wireless communication.
  • a radiation image captured by the radiation detector 20 in the mammography apparatus 10 is output to the console 12 via a communication I/F (Interface) section (not shown) by wired or wireless communication.
  • I/F Interface
  • the console 12 of this embodiment includes a control unit 40, a storage unit 42, a user I/F unit 44, and a communication I/F unit 46.
  • the control unit 40 of the console 12 has a function of controlling radiographic imaging of the breast by the mammography apparatus 10, as described above.
  • Examples of the control unit 40 include a computer system including a CPU (Central Processing Unit), ROM (Read Only Memory), and RAM (Random Access Memory).
  • the storage unit 42 has a function of storing information related to radiographic imaging, radiographic images obtained from the mammography apparatus 10, and the like.
  • the storage unit 42 is a nonvolatile storage unit, and includes, for example, an HDD (Hard Disk Drive), an SSD (Solid State Drive), and the like.
  • the user I/F unit 44 includes input devices such as various buttons and switches that are operated by a user such as an engineer, and display devices such as lamps and displays for displaying information on radiography and radiographic images. .
  • the communication I/F unit 46 communicates with the mammography apparatus 10 various data such as information related to radiographic imaging and radiographic images obtained by radiographic imaging by wired or wireless communication. Further, the communication I/F unit 46 communicates various data such as radiation images with the image processing apparatus 16 via the network 17 by wired communication or wireless communication.
  • the image processing device 16 is used when a doctor or the like (hereinafter simply referred to as "doctor") interprets radiographic images.
  • the image processing apparatus 16 of this embodiment has a function of detecting pale calcification from a radiographic image and displaying a calcification distribution image representing pale calcification in the breast of the radiographic image. is not required.
  • FIG. 3 shows a block diagram showing an example of the configuration of the image processing device 16 of this embodiment.
  • the image processing apparatus 16 of this embodiment includes a control section 60 , a storage section 62 , a display section 70 , an operation section 72 and a communication I/F section 74 .
  • the control unit 60, the storage unit 62, the display unit 70, the operation unit 72, and the communication I/F unit 74 are connected via a bus 79 such as a system bus or a control bus so that various information can be exchanged with each other.
  • a bus 79 such as a system bus or a control bus
  • the control unit 60 controls the overall operation of the image processing device 16 .
  • the control unit 60 includes a CPU 60A, a ROM 60B, and a RAM 60C.
  • the ROM 60B stores in advance various programs and the like for performing control by the CPU 60A.
  • the RAM 60C temporarily stores various data.
  • the storage unit 62 is a non-volatile storage unit, and specific examples include an HDD, an SSD, and the like.
  • the storage unit 62 stores various information such as an image processing program 63 and a calcification distribution determination model 64 .
  • the calcification distribution determination model 64 is a model that, when a low-resolution image including calcification is input, outputs a determination result of determining the type of calcification distribution state.
  • calcifications observed by doctors For example, according to mammography guidelines, calcifications observed in radiographic images are classified into clear benign calcifications and calcifications that require benign/malignant discrimination. Obvious benign calcifications include vascular calcification, central lucid calcification, calcareous calcification, and suture calcification. On the other hand, calcifications that require identification of benign or malignant are calcifications that do not belong to clear benign calcifications, and are classified mainly according to the morphology and distribution of calcifications, as shown in FIG.
  • the morphology of calcification is divided into 'microcircular', 'pale and obscure', 'polymorphic or heterogeneous', and 'microlinear or microbranched'.
  • “Microcircular” calcifications are calcifications that are circular or elliptical in size of 1 mm or less and have clear margins.
  • “Pale and indistinct” calcifications are mainly round or flaky calcifications, often small and pale. Because it is faint and unclear, it is difficult to classify the morphology and tends to be difficult to see in radiographic images. It should be noted that pale, indistinct calcifications tend to be smaller than microcircular calcifications.
  • “Polymorphic or heterogeneous” calcifications are irregularly shaped calcifications of various sizes and concentrations, typically with the shape of a cracked stone. "Fine linear or finely branched” calcifications are long and irregularly shaped calcifications, which are often linearly recognized. On the other hand, the distribution of calcification is classified into “diffuse or disseminated”, “regional”, “concentrated”, “linear” and “segmental”. "Diffuse or disseminated” is calcification that is scattered without a uniform distribution pattern throughout the breast. "Regional” is calcification that is widespread, but not throughout the mammary gland. “Conglomerate” is multiple calcifications that are confined to a small area. A “line” is a calcification that is arranged in a line. “Segmental” refers to calcifications consistent with the ductal lobar system, suggesting the possibility of lobular or segmental spread of breast cancer.
  • categories are associated with the distribution and morphology of calcification, as shown in FIG. "Category 2" is benign, “Category 3” is benign but malignancy cannot be ruled out, “Category 4" is suspected malignancy, and “Category 5" is malignant.
  • diagnosis of benign and malignant calcifications is made.
  • the calcification distribution determination model 64 of the present embodiment has "diffuse or sporadic", “regional”, “aggregated”, “linear”, and “segmental” as the state of the distribution of calcification. It is determined whether it is any of
  • calcification distribution determination model 64 a convolutional neural network (CNN: Convolutional Neural Network) that has been machine-learned by deep learning using learning data is used.
  • CNN Convolutional Neural Network
  • FIG. 5 shows an example of the calcification distribution determination model 64 of this embodiment.
  • the calcification distribution determination model 64 shown in FIG. 5 includes an input layer 200, an intermediate layer 201, a flat layer 210, and an output layer 212.
  • An image to be processed (in this embodiment, a low-resolution image whose details will be described later) is input to the input layer 200 .
  • the intermediate layer 201 includes a convolution layer 202 and a convolution layer 206 that perform convolution processing (conv), and a pooling layer 204 and a pooling layer 208 that perform pooling processing (pool).
  • conv convolution processing
  • pool pooling processing
  • the convolution processing performed in the convolution layers 202 and 206 will be described with reference to FIG.
  • the pixel value Ip (x, y) of the target pixel Ip of the input data DI is set to "e”
  • the pixel values of the surrounding adjacent pixels are set to "a” to "d”
  • "f ” ⁇ “i” and the coefficients of the 3 ⁇ 3 filter F are “r” ⁇ “z”
  • the pixel value Icp (x, y ) is obtained, for example, according to the following formula (1).
  • the coefficient of this filter F corresponds to the weight indicating the strength of the connection between the nodes in the preceding and succeeding layers.
  • Icp(x, y) a ⁇ z+b ⁇ y+c ⁇ x+d ⁇ w+e ⁇ v+f ⁇ u+g ⁇ t+h ⁇ s+i ⁇ r (1)
  • the convolution process the above-described convolution operation is performed on each pixel, and the pixel value Icp(x, y) corresponding to each pixel of interest Ip is output.
  • output data DIc having pixel values Icp(x, y) arranged two-dimensionally is output.
  • One output data DIc is output for one filter F.
  • FIG. When a plurality of filters F of different types are used, output data DIc is output for each filter F.
  • a filter F means a neuron (node) of a convolutional layer, and features that can be extracted for each filter F are determined.
  • pooling processing is performed to reduce the original image while retaining the features.
  • pooling processing is performed to reduce the image size by reducing the resolution of the input image by selecting a local representative value. For example, if pooling is performed by selecting a representative value from a block of 2 ⁇ 2 pixels with a stride of “1”, i.e., shifting by one pixel, a reduced image that is half the size of the input image is output. be done.
  • the convolutional layers 202 and 206 and the pooling layers 204 and 208 described above are arranged in order from the input layer 200 to the convolutional layer 202, the pooling layer 204, the convolutional layer 204, and the convolutional layer 204, respectively.
  • Layer 206 and pooling layer 208 are arranged in this order.
  • the convolution layer 202 applies a 3 ⁇ 3 filter F1 to the input (propagated) image, and performs the above-described convolution operation to obtain the characteristics of the input image. is extracted, and an image feature map cmp1 in which pixel values are arranged two-dimensionally is output. As described above, the number of image feature maps cmp1 is the number corresponding to the type of filter F1.
  • the pooling layer 204 reduces the size of the image feature map cmp1 to 1/4 (the vertical and horizontal sizes are 1/4) by performing a pooling process of selecting a representative value from a block of 2 ⁇ 2 pixels for the image feature map cmp1. 2) output a plurality of reduced image feature maps cmp2.
  • the convolution layer 206 applies a 3 ⁇ 3 filter F2 and performs the above-described convolution operation to extract the features of the input image feature map cmp2, and extract two-dimensional pixels. Output a plurality of image feature maps cmp3 in which values are arranged.
  • the pooling layer 208 reduces the size of the image feature map cmp3 to 1/4 by performing a pooling process of selecting a representative value from a block of 2 ⁇ 2 pixels on the image feature map cmp3. A plurality of image feature maps cmp4 reduced to 1/2 in length and width are output.
  • the numerical values of the data themselves are rearranged while the image feature map cmp4 remains unchanged.
  • three-dimensional data represented by a plurality of image feature maps cmp4 are rearranged as one-dimensional data.
  • the value of each node 211 included in the flat layer 210 corresponds to the pixel value of each pixel of multiple image feature maps cmp4.
  • the output layer 212 is a fully connected layer in which all the nodes 211 are connected.
  • Node 213B corresponds to the decision to be "aggregate”
  • node 213C corresponds to the decision to be “linear”
  • node 213D corresponds to the decision to be "regional”. It includes a node 213E corresponding to the decision.
  • the output layer 212 uses the softmax function, which is an example of an activation function, to generate the probability corresponding to the determination that the type of calcification distribution state corresponding to the node 213A is diffuse or sporadic, and the probability corresponding to the determination that the type of distribution state of corresponding calcifications is regional, the probability corresponding to the determination that the type of distribution state of calcifications corresponding to node 213C is convergence, Probability corresponding to determination that the type of calcification distribution state corresponding to node 213D is linear, and probability corresponding to determination that the type of calcification distribution state corresponding to node 213E is regional and Note that the calcification distribution determination model 64 may output information representing the probability of each of the nodes 213A to 213E instead of outputting labels such as "area” and "aggregation" as determination results. .
  • the calcification distribution determination model 64 is generated by machine-learning a machine-learning model using the learning data 120 .
  • the learning data 120 is composed of a set of a low-resolution image 95 and correct data 122 .
  • the low-resolution image 95 is an image obtained by reducing the resolution of the calcification distribution image 93 representing the detection result of detecting the calcification 50 from the radiation image 91 captured by the mammography apparatus 10 .
  • Schematically, the low-resolution image 95 is illustrated by making the image itself smaller than the calcification distribution image 93 relatively.
  • the types of the distribution state of the calcifications 50 included in the radiographic image 91 are "diffuse or sporadic", “regional”, “concentrated”, “linear”, and “segmental”. It is information indicating which of In this embodiment, machine learning of the calcification distribution determination model 64 is performed using the error backpropagation method.
  • the low-resolution image 95 of the learning data 120 is input to the calcification distribution determination model 64 .
  • the entire low-resolution image 95 is input to the calcification distribution determination model 64 for learning.
  • the calcification distribution determination model 64 uses the determination result of the type of distribution state of the calcifications 50 included in the low-resolution image 95, specifically, the values of the nodes 213A to 213E included in the output layer 212 of the calcification distribution determination model 64. to output
  • the node 213A has a value of "1” and the nodes 213B to 213E have a value of "0". should be.
  • the correct data 122 for the low-resolution image 95 input to the calcification distribution determination model 64 is "area"
  • the value of the node 213B is "1” and the values of the nodes 213A, 213C to 213E are "0". should be.
  • the correct data 122 for the low-resolution image 95 input to the calcification distribution determination model 64 is "segmental"
  • the node 213E has a value of "1”
  • the nodes 213A to 213A The value of 213D should be "0".
  • the difference (error) between the values of the nodes 213A to 213E output from the calcification distribution determination model 64 and the values to be taken by the nodes 213A to 213E corresponding to the correct data 122 is calculated. Then, using the error propagation method according to the error, the weight of each neuron is updated from the output layer 212 to the input layer 200 so as to reduce the error, and the calcification distribution determination model 64 is updated according to the update setting. Updated.
  • the learning of the calcification distribution determination model 64 may be performed by the image processing device 16, or may be performed by an external learning device, and the image processing device 16 acquires the learned calcification distribution determination model 64. and stored in the storage unit 62.
  • the display unit 70 of the image processing device 16 displays radiation images and various information.
  • the display unit 70 is not particularly limited, and may be various displays.
  • the operation unit 72 is used by the user to input instructions, various information, and the like for the doctor's diagnosis of breast lesions using radiographic images.
  • the operation unit 72 is not particularly limited, and examples thereof include various switches, a touch panel, a touch pen, and a mouse. Note that the display unit 70 and the operation unit 72 may be integrated to form a touch panel display.
  • the communication I/F unit 74 communicates various information with the console 12 via the network 17 by wireless communication or wired communication.
  • the function of the image processing device 16 to output the determination result of determining the type of distribution state of calcification in the breast in the radiographic image captured by the mammography device 10 will be described.
  • a radiographic image for which the classification of the distribution state of calcification is to be determined is a normal two-dimensional image.
  • FIG. 8 shows a schematic diagram for explaining the outline of the flow of determination of the type of calcification distribution state in the image processing apparatus 16 of the present embodiment.
  • FIG. 9 shows a functional block diagram of an example of the configuration of the image processing device 16 of this embodiment.
  • the image processing device 16 includes an acquisition section 80 , a calcification detection section 82 , a resolution reduction section 84 , a low resolution image processing section 86 and a display control section 88 .
  • the CPU 60A of the control unit 60 executes the image processing program 63 stored in the storage unit 62, so that the CPU 60A performs the acquisition unit 80, the calcification detection unit 82, the low-resolution function as a conversion unit 84 , a low-resolution image processing unit 86 , and a display control unit 88 .
  • the acquisition unit 80 has a function of acquiring a radiographic image 90 of a breast obtained by the mammography apparatus 10 . Specifically, the acquisition unit 80 acquires image data representing the radiation image 90 captured by the radiation detector 20 of the mammography apparatus 10 via the communication I/F unit 46 and the communication I/F unit 74 . The acquisition unit 80 outputs the acquired radiographic image 90 to the calcification detection unit 82 .
  • the calcification detection unit 82 has a function of detecting calcification 50 from the radiographic image 90 and generating a calcification distribution image 92 representing the type of distribution state of the calcification 50 .
  • the calcification distribution image 92 of the present embodiment detects the calcifications 50 from the radiation image 90 using a rule-based calcification detection model.
  • the calcification detection unit 82 sets an ROI (Region Of Interest) for each pixel of the radiation image 90, and based on the variance value ⁇ calc 2 derived for the ROI, the following equation (2) Detected by Threshold ⁇ ⁇ calc 2 (2)
  • Threshold in the above equation (2) is a threshold for distinguishing between noise and calcification.
  • ACR American College Radiology
  • the threshold value is calculated and used for each apparatus as a dispersion value that can detect 1.00 mm of calcified calcification.
  • the threshold may be adjusted according to the amount of mammary glands.
  • the amount of mammary glands in a breast is large, the structure of the mammary glands overlaps and becomes difficult to see, so the value of Threshold may be increased.
  • the method by which the calcification detection unit 82 derives the mammary gland volume from the radiation image 90 is not particularly limited.
  • the calcification detection unit 82 derives, for each pixel of the radiographic image 90, the mammary gland content rate representing the mammary gland content rate in the thickness direction of the breast, which is the irradiation direction of the radiation R, as the mammary gland amount.
  • the mammary gland content When there is no mammary gland and only fat, the mammary gland content is "0", and the higher the mammary gland density value, the larger the mammary gland content.
  • the calcification detection unit 82 detects the pixel values of a region where no breast is shown in the radiographic image 90, a so-called bare region, the pixel value of a pixel corresponding to fat, and the pixel value of a pixel from which the mammary gland content is to be derived.
  • the average attenuation coefficient ratio between mammary gland and fat mean attenuation coefficient of mammary gland/average attenuation coefficient of fat
  • the calcification detection unit 82 generates a calcification distribution image 92 containing information on the type of distribution of the calcification detected from the radiographic image 90 in this way.
  • the breast of the radiographic image 90 shown in FIG. 8 includes calcifications 50 and masses 54 that are other lesions.
  • a calcification detector 82 detects calcification 50 and generates a calcification distribution image 92 .
  • the calcification distribution image 92 is a black-and-white, binary image in which calcification is represented by "1" and other areas are represented by "0". As shown in FIG. 8, only the calcifications 50 appear as white images in the calcification distribution image 92 .
  • the calcification distribution image 92 is not limited to a binary image, and may be, for example, a grayscale image that represents stepwise from white to black. Alternatively, for example, it may be a mask image in which areas other than the calcification 50 area are masked.
  • the calcification detection unit 82 outputs the generated calcification distribution image 92 to the resolution reduction unit 84 .
  • the resolution reduction unit 84 has a function of generating a low resolution image 94 having a resolution lower than that of the radiation image 90 from the calcification distribution image 92 .
  • the method by which the resolution reduction unit 84 generates the low resolution image 94 from the calcification distribution image 92 is not particularly limited.
  • the low-resolution image 94 may be generated by setting a predetermined number of adjacent pixels in the calcification distribution image 92 as one pixel. In this case, one of the average value, the maximum value, and the minimum value of the pixel values of a predetermined number of pixels may be used as the pixel value of one pixel in the low-resolution image 94 .
  • the degree to which the resolution of the calcification distribution image 92 is reduced is not particularly limited.
  • the size of the low-resolution image 94 is determined according to the size of the calcification 50 in the low-resolution image 94, the size of the image input to the calcification distribution determination model 64, and the like. Depending on the situation, the resolution of the calcification distribution image 92 may be reduced.
  • the resolution reduction section 84 outputs the generated low resolution image 94 to the low resolution image processing section 86 .
  • the low-resolution image processing unit 86 has a function of determining the type of calcification distribution state from the low-resolution image 94 . Specifically, the low-resolution image processing unit 86 inputs the low-resolution image 94 to the calcification distribution determination model 64 described above, and acquires the output determination result 110 . The low resolution image processing section 86 outputs the determination result 110 to the display control section 88 .
  • the display control unit 88 has a function of controlling the display of information representing the determination result 110 obtained by the low-resolution image processing unit 86 on the display unit 70 .
  • FIG. 10 The image processing shown in FIG. 10 is executed by the CPU 60A executing the image processing program 63 stored in the storage unit 62.
  • FIG. 10 The image processing shown in FIG. 10 is executed by the CPU 60A executing the image processing program 63 stored in the storage unit 62.
  • the acquisition unit 80 acquires the radiation image 90 from the console 12 as described above.
  • the calcification detection unit 82 detects calcification from the radiographic image 90 acquired in step S100 and generates a calcification distribution image 92. As described above, the calcification detection unit 82 sets the ROI for each pixel of the radiographic image 90 and detects calcification based on the variance. In addition, as described above, the calcification detection unit 82 generates a binary image representing calcification and other regions as the calcification distribution image 92 .
  • the resolution reduction unit 84 In the next step S104, the resolution reduction unit 84 generates a low resolution image 94 from the calcification distribution image 92 generated in step S102, as described above.
  • the low-resolution image processing unit 86 determines the type of calcification distribution state from the low-resolution image 94 generated in step S104, as described above. As described above, the low-resolution image processing unit 86 inputs the low-resolution image 94 to the calcification distribution determination model 64 and acquires the output determination result 110 .
  • the display control unit 88 controls the display unit 70 to display the determination result of the type of calcification distribution state in step S106.
  • the display mode in which the display control unit 88 causes the display unit 70 to display the type of calcification distribution state is not particularly limited.
  • the display unit 70 displays characters indicating which of "diffuse or disseminated", “regional”, “aggregate”, “linear”, and “segmental". You may let Further, for example, as shown in FIG. 11A, the determination result 110 may be displayed together with the radiation image 90 acquired in step S100. Further, for example, as shown in FIG. 11B, a determination result 110 may be displayed together with the calcification distribution image 92 generated in step S102.
  • the radiographic image 90 and the calcification distribution image 92 are each superimposed with the determination result 110 .
  • a form in which the determination result 110 is displayed outside the distribution image 92 may be employed.
  • the calcifications 50 in each of the radiographic image 90 and the calcification distribution image 92 may be displayed in an emphasized manner, for example, by changing the color.
  • step S108 ends, the image processing shown in FIG. 10 ends.
  • Modification 1 a modification of the calcification detection method by the calcification detection unit 82 will be described.
  • the calcification detection unit 82 of the above embodiment detects calcification from the radiographic image 90 using a rule-based calcification detection model.
  • the calcification detection unit 82 of this modification detects calcification from the radiographic image 90 using a learning-based calcification detection model.
  • a trained model learned by machine learning is used as a calcification detection model so that calcification is detected from the radiographic image 90 and a calcification distribution image 92 is output as a detection result. Therefore, as shown in FIG. 12, a calcification detection model 66 is further stored in the storage unit 62 of the image processing apparatus 16 of this modified example.
  • a CNN Convolutional Neural Network
  • CNN Convolutional Neural Network
  • MLP MultiLayer Perceptron
  • the calcification detection model 66 is machine-learned using the learning data 65 .
  • a set of a radiation image 90 and an annotation image 99 is used as the learning data 65 .
  • Radiographic images 90 used as learning data 65 include those that do not include calcification 50 .
  • the annotation image 99 is an image in which the calcification 50 has been previously annotated, for example, manually.
  • the annotation image 99 is an image for matching an answer with the calcification distribution image 92 output from the calcification detection model 66 according to the radiographic image 90 , and is compared with the calcification distribution image 92 .
  • the radiation image 90 is input to the calcification detection model 66.
  • a calcification distribution image 92 is output from the calcification detection model 66 .
  • the calcification distribution image 92 output from the calcification detection model 66 and the annotation image 99 are compared, and the detection accuracy of the calcification 50 in the calcification detection model 66 is evaluated.
  • the calcification detection model 66 is updated according to this evaluation result.
  • the radiographic image 90 is input to the calcification detection model 66, the calcification distribution image 92 is output from the calcification detection model 66, and the calcification distribution image 92 is compared with the annotation image 99. Evaluation of the detection accuracy of the detection model 66 and updating of the calcification detection model 66 are performed while changing the set of the radiation image 90 and the annotation image 99, and are repeated until the detection accuracy of the calcification detection model 66 reaches a desired level.
  • the learning of the calcification detection model 66 may be performed by, for example, the image processing device 16.
  • an external learning device may perform the learning, and the learned calcification detection model 66 may be transferred to the image.
  • the processing device 16 may acquire from an external learning device.
  • step S102 of the image processing of the above embodiment the calcification detection unit 82 stores the radiation image 90 in the calcification detection model 66 stored in the storage unit 62. is input, the calcification detection model 66 is caused to detect calcification, and the calcification distribution image 92 output from the calcification detection model 66 is acquired to generate the calcification distribution image 92 .
  • a model that detects only faint and unclear calcifications may be constructed using a learned model learned by machine learning. Therefore, by using an appropriate model, Detection accuracy can be improved.
  • FIG. 14 shows a schematic diagram for explaining the outline of the flow of determination of the type of calcification distribution state in the image processing apparatus 16 of this modified example.
  • the calcification detection unit 82 of this modification enlarges the image of the calcification 50 included in the calcification distribution image 92 .
  • the resolution reduction unit 84 generates a low-resolution image 94 from the calcification distribution image 92 obtained by enlarging the image of the calcification 50 from the calcification distribution image 92 .
  • the calcification detection unit 82 performs calcification enlargement processing to enlarge the image of the calcification 50 included in the calcification distribution image 92, so that even if the resolution is reduced by the low-resolution processing, Information on the calcification 50 required for distribution determination is not lost. It should be noted that how much the calcification distribution image 92 enlarges the size of the image of the calcification 50 is not particularly limited.
  • the calcification detection unit 82 detects the calcification distribution image so that the size of the image of the calcification 50 included in the radiographic image 90 and the size of the image of the calcification 50 included in the low-resolution image 94 are the same.
  • the image of the calcifications 50 contained in 92 is enlarged.
  • FIG. 15 shows a flowchart showing an example of the flow of image processing by the image processing device 16 of this modified example.
  • the image processing shown in FIG. 15 differs from the image processing of the above embodiment (see FIG. 10) in that the processing of step S103 is provided between steps S102 and S104.
  • step S103 of FIG. 15 the calcification detection unit 82 performs the above-described calcification enlargement process to enlarge the image of the calcification 50 included in the calcification distribution image 92.
  • FIG. 16 shows a schematic diagram for explaining an overview of the flow of determination of the type of calcification distribution state in the image processing apparatus 16 of this modified example.
  • the calcification detection unit 82 of this modification performs filtering processing on the calcification distribution image 92 based on the signal value of the calcification 50 .
  • the resolution reduction unit 84 generates a low-resolution image 94 from the calcification distribution image 92 that has undergone calcification 50 filtering processing on the calcification distribution image 92 .
  • calcification can take many forms (see Figure 4).
  • microcircular calcifications have a higher signal value than pale, indistinct calcifications. Therefore, only the desired calcifications 50 can be extracted from the calcification distribution image 92 by performing a filtering process using a threshold corresponding to the calcifications to be detected.
  • the radiographic image 90 includes a pale and unclear calcification 50-1 and a micro-circular calcification 50-2 .
  • a filter for detecting the faint and unclear calcification 50-1 is set to a threshold value such as the intermediate value between the signal value of the faint and unclear calcification 50-1 and the signal value of the minute circular calcification 50-2 .
  • a light and unclear calcification 501 is extracted from the calcification distribution image 92 .
  • the resolution reduction unit 84 generates a low resolution image 94 from the calcification distribution image 92 in which only the faint and unclear calcifications 501 are represented.
  • FIG. 17 shows a flowchart showing an example of the flow of image processing by the image processing device 16 of this modified example.
  • the image processing shown in FIG. 17 differs from the image processing of the above-described embodiment (see FIG. 10) in that the processing of step S103A is provided between steps S102 and S104.
  • step S103A of FIG. 15 the calcification detection unit 82 performs the filtering process on the calcification distribution image 92 as described above, and detects the desired calcification 50 (light and unclear calcification 50 1 in FIG. 16). generates a calcification distribution image 92 in which only .
  • a calcification distribution image 92 containing only the desired form of calcification 50 is obtained. Therefore, it is possible to determine the type of distribution state according to the form of calcification 50 .
  • the threshold for filtering may be set based on the mammary gland volume of the breast.
  • FIG. 18 shows a schematic diagram for explaining an overview of the flow of determination of the type of calcification distribution state in the image processing apparatus 16 in this case.
  • the larger the mammary gland volume the smaller the threshold for filtering.
  • the calcification detection unit 82 derives the mammary gland volume of the breast from the radiation image 90 .
  • the method for deriving the mammary gland volume is not particularly limited.
  • the mammary gland content rate may be derived based on the pixel value of the pixel for which the rate is to be derived and the average attenuation coefficient ratio between mammary gland and fat (average attenuation coefficient of mammary gland/average attenuation coefficient of fat).
  • the calcification detection unit 82 derives the mammary gland volume from the radiation image 90 before performing the filtering process. For example, when the derived mammary gland volume exceeds a predetermined mammary gland threshold, the calcification detection unit 82 changes the threshold by filtering to a threshold determined to be used when the mammary gland volume is large. A filtering process is performed on the calcification distribution image 92 using a threshold value. Further, for example, a correspondence relationship between the mammary gland mass and a threshold value used for filtering is determined in advance, and the calcification detection unit 82 specifies a threshold value corresponding to the derived mammary gland mass based on the correspondence relationship, and uses the specified threshold value. A filtering process is performed on the calcification distribution image 92 .
  • a calcification distribution image 92 containing only the desired form of calcification 50 can be obtained with high accuracy.
  • FIG. 19 shows a schematic diagram for explaining the outline of the flow of determination of the type of calcification distribution state in the image processing apparatus 16 of this modified example.
  • a calcification distribution image 92 of this modified example includes not only the calcification 50 but also the breast skin line 56 and the nipple 57 .
  • the image processing device 16 of this modification further includes a skin line detection section 83A and a nipple detection section 83B.
  • the skin line detection unit 83A detects breast skin lines 56 from the radiation image 90 .
  • the method for detecting the breast skin line 56 by the skin line detection unit 83A is not particularly limited, and a known technique can be applied.
  • Japanese Unexamined Patent Application Publication No. 2008-086389 discloses a method of examining the density of a radiographic image 90, detecting positions where the density difference is greater than or equal to a predetermined value, and defining a set of pixels having the density difference greater than or equal to the predetermined value as a skin line. is exemplified. Further, for example, in Japanese Patent Application Laid-Open No.
  • the radiographic image is divided into a breast region and a bare region, and a boundary between the breast region and the bare region is disclosed.
  • a method of generating by connecting pixels that become points is exemplified.
  • the skin line detection unit 83 A superimposes the detected skin line 56 , or more precisely, the image of the skin line 56 on the calcification distribution image 92 .
  • the nipple detection unit 83B detects the breast nipple 57 from the radiographic image 90 .
  • the method by which the nipple detector 83B detects the nipple 57 is not particularly limited, and a known technique can be applied.
  • a part of the skin line of the breast that protrudes into the exposed area may be detected as a nipple.
  • the nipple detection unit 83B superimposes the detected nipple 57, strictly speaking, an image representing the nipple 57 on the calcification distribution image 92.
  • the resolution reduction unit 84 generates a low resolution image 94 from a calcification distribution image 92 including calcifications 50 , skin lines 56 and nipples 57 .
  • the skin line 56 and the nipple 57 are included in the calcification distribution image 92, so that the classification of the distribution state of the calcification 50 can be determined by referring to the skin line 56 and the nipple 57. can be done. Therefore, it is possible to easily determine the type of distribution state of the calcifications 50, and to improve the determination accuracy.
  • FIG. 20 shows a schematic diagram for explaining an overview of the flow of determination of the type of calcification distribution state in the image processing apparatus 16 of this modification.
  • a radiation image 90 shown in FIG. 20 includes calcifications 50A and 50B included in different distributions, and a calcification distribution image 92 includes calcifications 50A and 50B.
  • the low-resolution image processing unit 86 of this modified example determines the type of distribution state for each distribution.
  • the low-resolution image processing unit 86 determines whether or not a plurality of distributions of the calcifications 50 are included in the calcification distribution image 92, and if a plurality of distributions are included, the low-resolution image 94 , to generate a low-resolution image 94 for each distribution.
  • the method by which the low-resolution image processing unit 86 determines whether or not the calcification distribution image 92 includes a plurality of distributions is not particularly limited. In the example shown in FIG. 20, a distribution clustering model 67 that detects the distribution of the calcifications 50 by clustering the calcifications 50 (50A, 50B) included in the low-resolution image 94 is applied. By determining whether or not the low-resolution image 94 includes a plurality of distributions, it may be determined whether or not the calcification distribution image 92 includes a plurality of distributions.
  • the low-resolution image processor 86 applies the distribution clustering model 67 to the low-resolution image 94 to generate a low-resolution image 94A containing calcifications 50A and a low-resolution image 94A containing calcifications 50B. Generate a resolution image 94B.
  • the low-resolution image processing unit 86 applies the calcification distribution determination model 64 to the low-resolution image 94A to derive a determination result 110A, and applies the calcification distribution determination model 64 to the low-resolution image 94B to derive a determination result 110B.
  • the distribution clustering model 67 for example, it is conceivable to apply a hierarchical clustering method such as the centroid method.
  • a dendrogram is created by recursively performing a process of grouping calcification groups that are close to each other into the same cluster. If the distance between clusters is greater than or equal to a threshold value, the clusters are separated as separate calcification aggregates.
  • a method of determining clusters by applying the x-means method which is an unsupervised learning method, is also conceivable.
  • FIG. 21 shows a schematic diagram for explaining an outline of the flow of determination of the type of calcification distribution state in the image processing apparatus 16 of this modified example.
  • a low-resolution image 94L generated based on the radiographic image 90L of the left breast of the same subject and a low-resolution image 94R generated based on the radiographic image 90R of the right breast of the same subject.
  • a uniform distribution decision model 64 is applied to derive a decision result 110L for the left breast and a decision result 110R for the right breast.
  • the calcification detection unit 82 detects calcification 50L from the radiation image 90L of the left breast and generates a calcification distribution image 92L.
  • the resolution reduction unit 84 generates a low resolution image 94L from the calcification distribution image 92L.
  • the calcification detection unit 82 detects calcification 50R from the radiation image 90R of the right breast and generates a calcification distribution image 92R.
  • the resolution reduction unit 84 generates a low resolution image 94R from the calcification distribution image 92R.
  • the low-resolution image processing unit 86 inputs the low-resolution image 94L and the low-resolution image 94R to the calcification distribution determination model 64, and acquires the output determination result 110R and determination result 110L.
  • the distribution of calcifications 50L in the left breast and the distribution of calcifications 50R in the right breast can be compared to determine the type of distribution state in each. . Therefore, according to this modified example, it is possible to more accurately determine the type of the distribution state of calcification in the breast.
  • Modification 7 the image processing device 16 estimates the category to which the calcifications 50 belong based on the type of the distribution state of the calcifications 50 and the type of the shape of the calcifications 50.
  • the low-resolution image processing unit 86 of this modified example further has a function of determining the type of shape of the calcification 50 from the low-resolution image 94 .
  • FIG. 22 shows a schematic diagram for explaining the outline of the flow of determining the type of calcification shape in the low-resolution image processing unit 86 of the image processing device 16 of this modified example.
  • the processing until the low-resolution image 94 is input to the low-resolution image processing unit 86 is as described above, so the description is omitted.
  • the low-resolution image processing unit 86 inputs the low-resolution image 94 to the calcification shape determination model 68 and acquires the output determination result 111 .
  • the calcification shape determination model 68 is a model that outputs a determination result 111 of determining the type of the shape of the calcification 50 when the low-resolution image 94 including the calcification 50 is input.
  • the calcification shape determination model 68 of this modified example includes the types of the calcification 50, which are the above-described calcification forms, such as “microcircular”, “pale and unclear”, “polymorphic or heterogeneous”, and whether it is "finely linear or finely branched”.
  • the calcification shape determination model 68 is, for example, similar to the calcification distribution determination model 64, a convolutional neural network (CNN) machine-learned by deep learning using learning data (see FIG. 5). can be used.
  • CNN convolutional neural network
  • the calcification shape determination model 68 is generated by subjecting the machine learning model to machine learning using the learning data 121 .
  • the learning data 121 is composed of a set of a low-resolution image 95 and correct data 123 .
  • the low-resolution image 95 is an image obtained by reducing the resolution of the calcification distribution image 93 representing the detection result of detecting the calcification 50 from the radiation image 91 captured by the mammography apparatus 10 .
  • the correct data 123 indicates that the types of the shape of the calcifications 50 included in the radiographic image 91 are "microcircular”, “pale and unclear", “polymorphic or heterogeneous", and “fine linear or fine branching". This is information indicating which one.
  • the correct data 123 is information representing the type of shape of the calcification 50 included in the low-resolution image 95 corresponding to the radiographic image 91 .
  • the correct data 123 stores information representing the type of shape that is estimated to have the highest degree of malignancy among the types of shapes of each of the plurality of calcifications 50.
  • the low-resolution image 95 is the correct answer for the image 95.
  • the data 123 is information representing "fine linear or fine branch”.
  • the plurality of calcifications 50 included in the low-resolution image 95 include calcifications 50 whose shape type is “polymorphic or heterogeneous”, and calcifications whose shape is “microlinear or microbranched”. 50 is not included, the correct data 123 for the low-resolution image 95 is information representing "polymorphism or heterogeneity".
  • the plurality of calcifications 50 included in the low-resolution image 95 include calcifications 50 whose shape type is “pale and unclear”, and calcifications 50 that are “fine linear or fine branching” and If the calcifications 50 that are "polymorphic or heterogeneous" are not included, the correct data 123 for the low-resolution image 95 will be information representing "faint and unclear”.
  • the plurality of calcifications 50 included in the low-resolution image 95 include calcifications 50 whose shape type is "microcircular", and “microlinear or microbranching” calcifications 50, " If the "polymorphic or heterogeneous” calcifications 50 and the “pale and unclear” calcifications 50 are not included, the correct data 123 for the low-resolution image 95 is information representing "microcircles.”
  • machine learning of the calcification shape determination model 68 is performed using the error backpropagation method.
  • the low-resolution image 95 of the learning data 121 is input to the calcification shape determination model 68 .
  • the entire low-resolution image 95 is input to the calcification shape determination model 68 for learning.
  • the calcification shape determination model 68 determines the type of shape of the calcifications 50 included in the low-resolution image 95, specifically, the values of the nodes 213F to 213I included in the output layer 212 of the calcification shape determination model 68. Output.
  • the correct data 123 for the low-resolution image 95 input to the calcification shape determination model 68 is "fine circle"
  • the value of the node 213F should be "1"
  • the values of the nodes 213G to 213I should be "0". is.
  • the node 213G has a value of "1” and the nodes 213F, 213H, and 213I have a value of "0.” should be.
  • the correct data 123 for the low-resolution image 95 input to the calcification shape determination model 68 is "polymorphic or heterogeneous"
  • the value of the node 213H is "1”
  • the values of the nodes 213F, 213G, and 213I are " 0”.
  • the difference (error) between the values of the nodes 213F to 213I output from the calcification shape determination model 68 and the values to be taken by the nodes 213F to 213I corresponding to the correct data 123 is calculated. Then, using the error propagation method according to the error, the weight of each neuron is updated from the output layer 212 to the input layer 200 so as to reduce the error, and the calcification shape determination model 68 is updated according to the update setting. Updated.
  • the learning of the calcification shape determination model 68 may be performed by the image processing device 16, or may be performed by an external learning device, and the image processing device 16 acquires the learned calcification shape determination model 68. and stored in the storage unit 62.
  • the low-resolution image processing unit 86 outputs the calcification shape type determination result 111 output from the calcification shape determination model 68 and the calcification distribution state output from the calcification distribution determination model 64. Based on the type determination result 110, the category to which the calcification 50 belongs is estimated. As an example, the low-resolution image processing unit 86 of this modified example refers to the correspondence relationship shown in FIG. 110, the category to which calcification 50 belongs is estimated. The low-resolution image processing section 86 outputs the determination results 110 and 111 and information representing the estimated category to the display control section 88 .
  • FIG. 24 shows a flowchart showing an example of the flow of image processing by the image processing device 16 of this modified example.
  • the image processing shown in FIG. 24 differs from the image processing of the above-described embodiment (see FIG. 10) in that steps S110 to S114 are provided instead of step S108.
  • the low-resolution image processing unit 86 determines the type of calcification shape from the low-resolution image 94 generated at step S104. As described above, the low-resolution image processing unit 86 inputs the low-resolution image 94 to the calcification shape determination model 68 and acquires the output determination result 111 .
  • the low-resolution image processing unit 86 estimates the category to which the calcification belongs from the type of calcification distribution determined in step S106 and the type of calcification shape determined in step S108. As described above, the low-resolution image processing unit 86 refers to the correspondence shown in FIG. 4 and estimates the category.
  • step S114 the display control unit 88 outputs the determination result 110 of the type of calcification distribution state obtained in step S106, the determination result 111 of the type of calcification shape obtained in step S110, and the estimation of the category obtained in step S112. Control is performed to display the result on the display unit 70 .
  • the display control unit 88 determines the type of calcification distribution state and the type of calcification shape.
  • a display form in which the estimated category is displayed on the display unit 70 is not particularly limited.
  • the type of shape of the calcifications 50 is determined from the low-resolution image 94, the type of the calcifications 50 is determined based on the type of shape of the calcifications 50 and the type of distribution state. Categories can be inferred.
  • the low-resolution image processing unit 86 of the image processing device 16 estimates the category to which the calcifications 50 belong. Estimation of the category of calcification 50 based on the type of calcification need not be performed. For example, the category of calcification 50 may be determined by a doctor or the like. In this case, step S112 of the series of steps of the image processing shown in FIG. 24 is omitted. Further, in step S114, the display control unit 88 controls the display unit 70 to display only the determination result 110 of the type of calcification distribution state obtained in step S106 and the determination result 111 of the type of calcification shape obtained in step S110. should be done. In the case of this embodiment, the doctor or the like refers to the type of distribution state and the type of shape of the calcifications 50 displayed on the display unit 70, and from the correspondence shown in FIG. It is possible to judge whether it is any of
  • the calcification shape determination model 68 and its learning method are not limited to the methods described above.
  • the calcification shape determination model 68 may be a semantic segmentation model, and the type of shape may be determined for each of the plurality of calcifications 50 included in the low-resolution image 94 .
  • the learning data 121 is composed of a set of a low-resolution image 95 and a correct label image corresponding to the low-resolution image 95 .
  • the correct label image an image is used in which the shapes of the calcifications 50 are treated as independent classes, and IDs (Identifiers) corresponding to each shape are pixel values.
  • the correct label image is an image in which the pixel values of the plurality of calcifications 50 included in the low-resolution image 95 are set to ID values (1 to 4) according to the shape type. becomes.
  • the learning data 121 is composed of a set of the low-resolution image 95, the set ROI image, and the correct label assigned to the ROI image.
  • Modification 8 In this modified example, a mode in which the low-resolution image processing unit 86 further has a function of estimating whether the calcification 50 is benign or malignant from the low-resolution image 94 will be described.
  • FIG. 25 shows a schematic diagram for explaining the outline of the flow of estimating whether calcification is benign or malignant in the low-resolution image processing unit 86 of the image processing device 16 of this modification. .
  • the processing until the low-resolution image 94 is input to the low-resolution image processing unit 86 is as described above, so the description is omitted.
  • the low-resolution image processing unit 86 inputs the low-resolution image 94 to the benign/malignant estimation model 69 and acquires the output estimation result 113 .
  • the benign/malignant estimation model 69 is a model that outputs an estimation result 113 estimating whether the calcification 50 is benign or malignant when the low-resolution image 94 including the calcification 50 is input.
  • the benign/malignant estimation model 69 is, for example, similar to the calcification distribution determination model 64, a convolutional neural network (CNN: Convolutional Neural Network) (see FIG. 5) that has been machine-learned by deep learning using learning data. can be used.
  • CNN Convolutional Neural Network
  • the benign/malignant presumed model 69 is generated by subjecting the machine learning model to machine learning using the learning data 124 .
  • the learning data 124 is composed of a set of a low-resolution image 95 and correct data 126 .
  • the low-resolution image 95 is an image obtained by reducing the resolution of the calcification distribution image 93 representing the detection result of detecting the calcification 50 from the radiation image 91 captured by the mammography apparatus 10 .
  • the correct data 126 is information indicating whether the calcification 50 included in the radiation image 91 is benign or malignant.
  • machine learning of the benign/malignant estimation model 69 is performed using the error backpropagation method.
  • the low-resolution image 95 of the learning data 124 is input to the benign/malignant estimation model 69 .
  • the entire low-resolution image 95 is input to the benign/malignant estimation model 69 for learning.
  • the benign/malignant estimation model 69 determines whether the calcifications 50 included in the low-resolution image 95 are benign or malignant. , outputs 213K values.
  • the correct data 126 for the low resolution image 95 input to the benign/malignant estimation model 69 is "benign"
  • the value of the node 213J should be "1”
  • the value of the node 213K should be "0”.
  • the correct data 126 for the low resolution image 95 input to the benign/malignant estimation model 69 is "malignant”
  • the value of the node 213K should be "1" and the value of the node 213J should be "0". be.
  • the difference (error) between the values of the nodes 213J and 213K output from the benign/malignant estimation model 69 and the values to be taken by the nodes 213J and 213K corresponding to the correct data 126 is calculated. Then, using the error propagation method according to the error, the weight of each neuron is updated from the output layer 212 to the input layer 200 so as to reduce the error, and the benign/malignant estimation model 69 is generated according to the update setting. Updated.
  • the low-resolution image 95 of the learning data 124 is input to the benign/malignant estimation model 69, and the nodes 213J and 213K included in the output layer 212 from the benign/malignant estimation model 69 are used.
  • a series of processes including value output, error calculation based on the values of nodes 213J and 213K and correct answer data 126, weight update setting, and benign/malignant estimation model 69 update are repeated.
  • the learning of the benign/malignant estimation model 69 may be performed by the image processing device 16, or may be performed by an external learning device, and the image processing device 16 acquires the trained benign/malignant estimation model 69. and stored in the storage unit 62.
  • the low-resolution image processing unit 86 of this modified example controls the display of the estimation result 113 as to whether the calcification 50 is benign or malignant and the determination result 110 of the calcification distribution state. Output to unit 88 .
  • FIG. 27 shows a flowchart showing an example of the flow of image processing by the image processing device 16 of this modified example.
  • the image processing shown in FIG. 27 differs from the image processing of the above-described embodiment (see FIG. 10) in that steps S120 and S122 are provided instead of step S108.
  • the low-resolution image processing unit 86 estimates whether the calcification is benign or malignant from the low-resolution image 94 generated at step S104. As described above, the low-resolution image processing unit 86 inputs the low-resolution image 94 to the benign/malignant estimation model 69 and acquires the output estimation result 113 .
  • step S122 the display control unit 88 displays, on the display unit 70, the determination result 110 of the type of calcification distribution state obtained in step S106 and the estimation result 113 as to whether it is benign or malignant in step S120.
  • control to allow The display mode in which the display control unit 88 causes the display unit 70 to display the classification of the calcification distribution state and the estimation result as to whether the calcification is benign or malignant is not particularly limited.
  • the shape of the calcification 50 may be further determined. That is, this modified example and modified example 7 may be combined.
  • the calcification detection unit 82 detects the calcification 50 from the radiation image 90 obtained by irradiating radiation and imaging the breast.
  • a low-resolution unit 84 generates a low-resolution image 94 whose resolution is lower than that of the radiographic image 90 from the calcification distribution image 92 representing the detection result of the calcification 50 .
  • a low-resolution image processing unit 86 determines the type of distribution of calcifications 50 from the low-resolution image 94 .
  • the image processing device 16 of each of the above forms detects the calcification 50 from the radiographic image 90 and lowers the resolution of the calcification distribution image 92 . Therefore, even calcifications with weak signals can be reduced in resolution without being buried in surrounding structures. Also, the determination of the type of distribution state of the calcifications 50 is performed from the low-resolution image 94 .
  • the low resolution image 94 is a low resolution image that is smaller than the radiation image 90 . Therefore, the distribution can be determined from the entire low-resolution image 94, and a global determination can be performed by looking at the entire image. Therefore, according to the image processing apparatus 16 of the present embodiment, it is possible to accurately determine the type of distribution state of calcification in the breast. Further, according to each of the above embodiments, since the type of distribution state is determined from the low-resolution image 94 that does not include the surrounding structure, the number of learning data 120 used for learning the calcification distribution determination model 64 can be reduced. .
  • the radiographic image 90 is a normal two-dimensional image
  • the radiographic image 90 is not limited to a normal two-dimensional image.
  • the radiographic image 90 is a plurality of tomographic images obtained from a series of multiple projection images, or a composite two-dimensional image obtained by synthesizing at least part of a series of multiple projection images or multiple tomographic images. good too.
  • a processing unit that executes various processes such as the acquisition unit 80, the calcification detection unit 82, the resolution reduction unit 84, the low resolution image processing unit 86, and the display control unit 88
  • various processors shown below can be used.
  • the various processors include, in addition to the CPU, which is a general-purpose processor that executes software (programs) and functions as various processing units, circuits such as FPGAs (Field Programmable Gate Arrays), etc.
  • Programmable Logic Device which is a processor whose configuration can be changed, ASIC (Application Specific Integrated Circuit) etc. Circuits, etc. are included.
  • One processing unit may be composed of one of these various processors, or a combination of two or more processors of the same or different type (for example, a combination of a plurality of FPGAs, or a combination of a CPU and an FPGA). combination). Also, a plurality of processing units may be configured by one processor.
  • a single processor is configured by combining one or more CPUs and software.
  • a processor functions as multiple processing units.
  • SoC System On Chip
  • a processor that realizes the functions of the entire system including multiple processing units with a single IC (Integrated Circuit) chip. be.
  • various processing units are configured using one or more of the above various processors as a hardware structure.
  • an electric circuit combining circuit elements such as semiconductor elements can be used.
  • the image processing program 63 is pre-stored (installed) in the storage unit 62, but is not limited to this.
  • the image processing program 63 is provided in a form recorded in a recording medium such as a CD-ROM (Compact Disc Read Only Memory), a DVD-ROM (Digital Versatile Disc Read Only Memory), and a USB (Universal Serial Bus) memory. good too.
  • the image processing program 63 may be downloaded from an external device via a network.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • Surgery (AREA)
  • Public Health (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Animal Behavior & Ethology (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Biomedical Technology (AREA)
  • Veterinary Medicine (AREA)
  • Optics & Photonics (AREA)
  • Pathology (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Dentistry (AREA)
  • Physiology (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
PCT/JP2022/041084 2021-12-21 2022-11-02 画像処理装置、画像処理方法、及び画像処理プログラム Ceased WO2023119904A1 (ja)

Priority Applications (3)

Application Number Priority Date Filing Date Title
EP22910624.0A EP4454566A4 (en) 2021-12-21 2022-11-02 IMAGE PROCESSING DEVICE, IMAGE PROCESSING METHOD AND IMAGE PROCESSING PROGRAM
JP2023569137A JPWO2023119904A1 (https=) 2021-12-21 2022-11-02
US18/747,369 US20240338825A1 (en) 2021-12-21 2024-06-18 Image processing apparatus, image processing method, and image processing program

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2021-207601 2021-12-21
JP2021207601 2021-12-21

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/747,369 Continuation US20240338825A1 (en) 2021-12-21 2024-06-18 Image processing apparatus, image processing method, and image processing program

Publications (1)

Publication Number Publication Date
WO2023119904A1 true WO2023119904A1 (ja) 2023-06-29

Family

ID=86902068

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2022/041084 Ceased WO2023119904A1 (ja) 2021-12-21 2022-11-02 画像処理装置、画像処理方法、及び画像処理プログラム

Country Status (4)

Country Link
US (1) US20240338825A1 (https=)
EP (1) EP4454566A4 (https=)
JP (1) JPWO2023119904A1 (https=)
WO (1) WO2023119904A1 (https=)

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH08294479A (ja) * 1995-01-23 1996-11-12 Fuji Photo Film Co Ltd 計算機支援画像診断装置
JPH09251535A (ja) * 1996-03-14 1997-09-22 Konica Corp ２次元多階調画像におけるパタン検出方法及び装置
US5857030A (en) * 1995-08-18 1999-01-05 Eastman Kodak Company Automated method and system for digital image processing of radiologic images utilizing artificial neural networks
US5946407A (en) * 1993-08-13 1999-08-31 Bamberger; Philippe System and method for scanning medical images using adjustable exposure time and brightness
JP2002109512A (ja) * 2000-09-28 2002-04-12 Fuji Photo Film Co Ltd 異常陰影候補検出装置および記録媒体
JP2005177037A (ja) * 2003-12-18 2005-07-07 Fuji Photo Film Co Ltd 石灰化陰影判定方法、石灰化陰影判定装置およびプログラム
JP2008086389A (ja) 2006-09-29 2008-04-17 Fujifilm Corp 放射線撮影装置および放射線撮影方法
JP2010051456A (ja) 2008-08-27 2010-03-11 Fujifilm Corp 医用画像の評価装置、方法及びプログラム
JP2016022143A (ja) 2014-07-18 2016-02-08 ジーイー・メディカル・システムズ・グローバル・テクノロジー・カンパニー・エルエルシー 石灰化表示装置及び撮影装置並びにプログラム
JP2018161405A (ja) 2017-03-27 2018-10-18 キヤノン株式会社 画像処理装置及び放射線画像データの表示方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107798679B (zh) * 2017-12-11 2021-04-27 福建师范大学 乳腺钼靶图像乳房区域分割与钙化点检测方法

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5946407A (en) * 1993-08-13 1999-08-31 Bamberger; Philippe System and method for scanning medical images using adjustable exposure time and brightness
JPH08294479A (ja) * 1995-01-23 1996-11-12 Fuji Photo Film Co Ltd 計算機支援画像診断装置
US5857030A (en) * 1995-08-18 1999-01-05 Eastman Kodak Company Automated method and system for digital image processing of radiologic images utilizing artificial neural networks
JPH09251535A (ja) * 1996-03-14 1997-09-22 Konica Corp ２次元多階調画像におけるパタン検出方法及び装置
JP2002109512A (ja) * 2000-09-28 2002-04-12 Fuji Photo Film Co Ltd 異常陰影候補検出装置および記録媒体
JP2005177037A (ja) * 2003-12-18 2005-07-07 Fuji Photo Film Co Ltd 石灰化陰影判定方法、石灰化陰影判定装置およびプログラム
JP2008086389A (ja) 2006-09-29 2008-04-17 Fujifilm Corp 放射線撮影装置および放射線撮影方法
JP2010051456A (ja) 2008-08-27 2010-03-11 Fujifilm Corp 医用画像の評価装置、方法及びプログラム
JP2016022143A (ja) 2014-07-18 2016-02-08 ジーイー・メディカル・システムズ・グローバル・テクノロジー・カンパニー・エルエルシー 石灰化表示装置及び撮影装置並びにプログラム
JP2018161405A (ja) 2017-03-27 2018-10-18 キヤノン株式会社 画像処理装置及び放射線画像データの表示方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4454566A4

Also Published As

Publication number Publication date
EP4454566A1 (en) 2024-10-30
EP4454566A4 (en) 2025-04-16
US20240338825A1 (en) 2024-10-10
JPWO2023119904A1 (https=) 2023-06-29

Similar Documents

Publication Publication Date Title
JP5260892B2 (ja) 放射線医学的徴候の検出のためのトモシンセシスにおける放射線画像の処理方法
JP5138910B2 (ja) 投影画像を用いた３ｄｃａｄのシステム及び方法
JP7237809B2 (ja) 深層学習ベースの画像再構成のためのシステムおよび方法
JP5041821B2 (ja) 放射線医学的徴候の検出のためのトモシンセシス投影画像の処理装置
JP5162182B2 (ja) 異常陰影の検出のために放射線画像を処理するための方法
WO2021021329A1 (en) System and method for interpretation of multiple medical images using deep learning
JP2002109510A (ja) 異常陰影候補検出処理システム
US12524874B2 (en) Image processing device, learning device, image processing method, learning method, image processing program, and learning program that can accurately detect a lesion from a medical image using a lesion detection model
Liu et al. 3d tomographic pattern synthesis for enhancing the quantification of covid-19
Mabrouk et al. Computer aided detection of large lung nodules using chest computer tomography images
US7983463B2 (en) Methods and apparatus for suppressing tagging material in prepless CT colonography
WO2023119904A1 (ja) 画像処理装置、画像処理方法、及び画像処理プログラム
CN1692377A (zh) 形成人体结构的独立显影的方法和设备
JP5048233B2 (ja) Ｃａｄシステムにおける解剖学的形状の検出のための方法及びシステム
JP7710950B2 (ja) 画像処理装置、画像処理方法、及び画像処理プログラム
JP2006346094A (ja) 検出情報の出力方法及び医用画像処理システム
Thangaraju et al. Detection of microcalcification clusters using hessian matrix and foveal segmentation method on multiscale analysis in digital mammograms
US12475556B2 (en) Image processing device, image processing method, and image processing program
JP7840696B2 (ja) 画像処理装置、画像処理方法、及びプログラム
Taljeh et al. Segnet Deep Architecture for Breast Tumor Segmentation
Chan et al. Computer-aided diagnosis of breast cancer with tomosynthesis imaging
Boudouh et al. Breast Micro-Calcification Classification: Xception and DenseNet201 Feature Fusion in Mammography
Kim et al. Automatic multi-class classification of 3D relationships between the mandibular third molar and canal on cone-beam computed tomography using a geometry-aware network
Samei et al. Biplane correlation imaging: a feasibility study based on phantom and human data
JP2002330952A (ja) 異常陰影候補検出方法および装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22910624

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 2023569137

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2022910624

Country of ref document: EP

Effective date: 20240722