JP2006181025A - Abnormal shadow detecting method, device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enable a processing with a superior detecting property with consideration given to anatomic features of a subject in detecting an abnormal shadow of a digital medical image. <P>SOLUTION: When detecting the abnormal shadow, a region to be a detection object is divided into small regions R1-R3, and the abnormal shadow is detected for every divided region using a medical image of a type with improved detectability of the abnormal shadow, for example either one of a simple radiation image Px, a time-lapse subtraction image Pt, and an energy subtraction image Pe with consideration given to the anatomic features of the subject (a lung field) in the divided region. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an abnormal shadow detection method and apparatus for detecting an abnormal shadow in a medical image represented by the data based on medical image data, and a program therefor.

  2. Description of the Related Art Conventionally, in the medical field, a computer that analyzes and diagnoses a digital medical image (hereinafter simply referred to as a medical image) using a computer for the purpose of reducing the burden on an interpreter such as a doctor in image diagnosis. Diagnosis support, so-called CAD (Computer Aided Diagnosis), is performed. As such CAD, for example, an abnormal shadow detection process for automatically detecting an abnormal shadow in a medical image using a computer is known.

  By the way, the QL value (pixel signal value) representing the density of each pixel in the medical image has a characteristic for each anatomical position of the subject. For example, in a chest X-ray image in which the subject is the chest of a human body, due to the influence of tissue having anatomical characteristics such as breast, scapula, and pectoralis major muscles and scattered rays, the outside of the lung field and the vicinity of the center The nature of is different.

  Therefore, as described above, in the case where an abnormal shadow in a medical image is automatically detected using a computer, the influence of the difference in properties on the image depending on the anatomical position deteriorates the detection performance of the abnormal shadow. This is considered to be one of the causes.

Therefore, as a technique for solving the above problem, for example, in a chest X-ray image, the lung field is divided into a plurality of small areas such as the spine, the clavicle, the lung field margin, the lung field center, and the like. A method of performing appropriate abnormal shadow detection processing (see Patent Document 1), a method of changing the detection sensitivity of abnormal shadows for each of the small areas, or a method of changing the number of abnormal shadow candidates picked up for each of the small areas Has been proposed.
US Pat. No. 6,549,646

  However, a technique proposed in Patent Document 1 that divides the lung field into a plurality of small regions and performs appropriate abnormal shadow detection processing on each small region, or abnormal shadows for each of the small regions described above. Even with a method of changing the detection sensitivity or changing the number of abnormal shadow candidates picked up for each of the small areas, a detection performance at a practical level has not yet been obtained. One reason for this is that in the above method, the medical image itself used to detect abnormal shadows has not changed, and the properties on the image due to the anatomical structure of the subject have not changed. It is considered that the image plane does not sufficiently consider the anatomical features of the image.

  In view of the above circumstances, an object of the present invention is to provide an abnormal shadow detection method and apparatus, and a program therefor, in which the anatomical characteristics of the subject are more considered on the image plane and the abnormal shadow detection performance is further improved. Is.

  The abnormal shadow detection method of the present invention is a medical type of a kind determined for each divided area corresponding to each of a plurality of divided areas obtained by dividing the area of the subject on an image according to a predetermined depiction form of the subject. Each medical image data representing an image is acquired, and an abnormal shadow in the medical image of each divided region represented by each medical image data is detected based on the acquired medical image data. Method (first abnormal shadow detection method).

  In addition, the abnormal shadow detection method of the present invention provides, for each of the plurality of divided areas obtained by dividing the area of the subject on an image according to a predetermined depiction form of the subject, for each divided area. The type of the medical image used when detecting an abnormal shadow in the medical image corresponding to the medical image is set, medical image data representing the medical image of the subject for each set type is acquired, and the acquired The anatomical information of the subject in each type of acquired medical image represented by each medical image data is acquired, and the divided areas are set for the divided areas based on the anatomical information. Based on the data representing the images of the respective detection target areas of the respective types of medical image data, extracted on the acquired medical images of the types, setting the extracted divided areas as the detection target areas of the abnormal shadows. There are a method and detecting the abnormal shadow in an image of the detection target region (second abnormal shadow detecting method).

  The abnormal shadow detection apparatus of the present invention is a medical device of a type determined for each divided area corresponding to each of a plurality of divided areas obtained by dividing the area of the subject on an image according to a predetermined depiction form of the subject. Image data acquisition means for acquiring each medical image data representing an image, and detecting an abnormal shadow in the medical image of each divided region represented by each medical image data based on each acquired medical image data An abnormal shadow detecting means is provided (first abnormal shadow detecting device).

  In addition, the abnormal shadow detection apparatus of the present invention provides, for each of the plurality of divided areas obtained by dividing the area of the subject on the image according to the predetermined depiction form of the subject, the divided area. An image type setting means for setting the type of the medical image used when detecting an abnormal shadow in the medical image corresponding to the image, and medical image data representing the medical image of the subject for each set type is acquired. Image data acquisition means, anatomical information acquisition means for acquiring the anatomical information of the subject in each type of acquired medical image represented by each of the acquired medical image data, and Based on the region extraction means for extracting the respective divided regions on the acquired medical image of the type set for each divided region, and the extracted divided regions as abnormal shadow detection target regions Detection target area setting means for setting, and abnormal shadow detection means for detecting an abnormal shadow in the image of each detection target area based on data representing the image of each detection target area among the respective types of medical image data (2nd abnormal shadow detection apparatus).

  The second abnormal shadow detection apparatus of the present invention further includes image data storage means for storing a plurality of types of medical image data of the subject, wherein the image type setting means is preset for each of the divided regions. Based on the priority order of the types of medical images, the types of the medical images corresponding to the respective divided regions are determined from the types of the medical images represented by the medical image data stored in the image data storage unit. It may be set to the type with the highest priority order.

  In this case, the abnormal shadow detecting means may detect an abnormal shadow by performing a predetermined abnormal shadow detection process corresponding to the type of the image on the image of the detection target region. Good.

  Further, in the second abnormal shadow detection apparatus, an image display for displaying information indicating the position of the abnormal shadow detected by the abnormal shadow detection means on one screen together with any one of the acquired medical images. You may make it further provide a means.

  Alternatively, information indicating the position of the abnormal shadow detected by the abnormal shadow detection means together with the image arranged so that the image of each detection target area has a positional relationship equivalent to the positional relationship of each divided region Further, an image display means for displaying on one screen may be further provided.

  In the second abnormal shadow detection device, the subject is a lung field of a human chest, and one of the divided regions includes a region including a lung apex, a region including a lung field central portion, and an outer side. The medical image may be any one of regions including a peripheral portion, and the type of the medical image may include any one of a temporal subtraction image, an energy subtraction image, and a simple radiation image.

  Here, the “subject” may be the whole or a part of a living organism such as an animal including a human body, and may be, for example, a lung field of a human chest.

  As the “image according to the predetermined drawing form”, for example, a projection image obtained by projecting the subject in one direction, a tomographic image obtained by slicing the subject imagewise, and the like can be considered.

  Regarding “the type of medical image defined for each divided region”, the types of medical images defined in the plurality of divided regions may overlap, but the same type of medical image is defined in all the divided regions. Does not include cases. Also, different types of medical images need not be defined for each divided area.

  The “image type setting means” may automatically set the type of medical image for each divided area by itself, or may be set based on an input from the operator. .

  As “predetermined divided regions”, for example, regions divided by anatomical features based on the anatomical structure of the subject can be considered.

  As the “type of medical image”, in a projected image, for example, a “simple radiation image” obtained by detecting radiation that has passed through the subject in almost one direction, or a radiographic image of the same subject at different imaging times is overlapped with each other. Paying attention to the fact that the absorption spectrum of radiation differs depending on the structure that constitutes the subject, and the two types of radiation with different energy distributions transmitted through the subject. “Energy subtraction image” as an image in which the shadow of a predetermined structure is emphasized or suppressed, obtained by overlapping and subtracting the subject between the images obtained by detection, a plurality of subjects obtained by the CT scanning device A set of tomographic image data is taken as three-dimensional data, and the image is projected when the subject is projected in a predetermined direction. For example, a “CT image” of a subject obtained by a CT scanning device or an “MRI image” obtained by an MRI device can be considered. ”,“ RI image ”obtained by the RI apparatus,“ temporal subtraction image ”as a difference image obtained by overlapping and subtracting subjects of the same kind of images of the same subject with different shooting times, etc. it can.

  “Anatomical information” refers to information relating to the anatomical structure of a subject represented on a medical image, and refers to information that can specify at least the position / contour shape of the subject. It is possible to consider information on positions that change sharply and the positional relationship between these positions.

  “Abnormal shadow” refers to a shadow such as a mass or calcification in a chest X-ray image that is not found in a standard shadow. In the present invention, when extracting abnormal shadow candidates, it is not necessary to extract all of these abnormal shadows. For example, only a tumor may be extracted as an abnormal shadow candidate.

  As a technique for “detecting an abnormal shadow”, for example, a lesion is suspected by using iris filter processing or the like based on image data representing a medical image as disclosed in Japanese Patent No. 2998733. A region is extracted as a temporary candidate for an abnormal shadow, and a plurality of feature quantities that reflect the probability as an abnormal shadow are calculated for each candidate based on image data in the vicinity of the extracted candidate. Based on a plurality of feature quantities, a new index value (score) is calculated, and whether or not the candidate is an abnormal shadow is determined based on whether or not the index value exceeds a predetermined reference value. Can be used to extract abnormal shadows from the above candidates by using a neural network or the like that outputs an amount representing the degree of probability that the candidate is an abnormal shadow

  “Feature amount” is a collective term for various amounts that reflect the probability of each candidate as an abnormal shadow. For example, “improvement of the ability to discriminate malignant tumor shadows on a mammogram and selection criteria As described in "Evaluation" (The Institute of Electronics, Information and Communication Engineers Journal D-IIVol86-DIINo5, pp587-597), each candidate's brightness, area, edge, shape (roundness, etc.), etc. are considered. More specifically, as for luminance, the QL value (pixel signal value) average, variance, second moment, and area are surrounded by candidate areas such as a dynamic contour model. As for the area and edge, the features and shape calculated from the co-occurrence matrix for the points obtained by the area division process are used for the figure enclosed by the area division process. Spreadness, flatness, etc. can be considered.

  The program of the present invention is a medical program of a type determined for each divided area corresponding to each of a plurality of divided areas obtained by dividing the area of the subject on an image according to a predetermined depiction form of the subject. Image data acquisition means for acquiring each medical image data representing an image, and detecting an abnormal shadow in the medical image of each divided region represented by each medical image data based on each acquired medical image data It is a program for functioning as an abnormal shadow detection means.

  Further, the program according to the present invention provides, for each of the plurality of divided areas obtained by dividing the area of the subject on the image in a predetermined depiction form of the subject, the divided area for each divided area. An image type setting means for setting the type of the medical image used when detecting an abnormal shadow in the medical image corresponding to the image, and medical image data representing the medical image of the subject for each set type is acquired. Image data acquisition means, anatomical information acquisition means for acquiring the anatomical information of the subject in each type of acquired medical image represented by each of the acquired medical image data, and A region extracting means for extracting each divided region on the acquired medical image of the type set for each divided region, and detecting each of the extracted divided regions for an abnormal shadow. A detection target region setting means for setting as a region, and an abnormal shadow for detecting an abnormal shadow in the image of each detection target region based on data representing the image of each detection target region among the respective types of medical image data It is a program for functioning as detection means.

  According to the abnormal shadow detection method and apparatus and the program therefor according to the present invention, when detecting an abnormal shadow, the area to be detected is divided, and the medical image used for detecting the abnormal shadow is divided for each divided area. Since the type is determined, for each divided area, the abnormal shadow is detected using a medical image of a type that improves the detection ability of the abnormal shadow in consideration of the anatomical characteristics of the subject in the divided area. It is possible to improve the performance of detecting abnormal shadows overall.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a schematic block diagram showing the configuration of the abnormal shadow detection apparatus according to this embodiment. The abnormal shadow detection apparatus 1 according to the present embodiment detects an abnormal shadow in a medical image of a chest and displays the result. As shown in FIG. 1, an image data storage unit 10 and an image type setting unit 12 are displayed. , Image data acquisition unit 14, lung field region recognition unit 16, lung field region segmentation unit 18, detection target region setting unit 20, abnormal shadow candidate detection unit 22, feature amount calculation unit 24, discrimination processing unit 26, and display control unit 28. The display unit 30 is provided. The lung field region recognition unit 16 functions as an anatomical information acquisition unit in the present invention, and the lung field region division unit 18 functions as a region extraction unit in the present invention, and an abnormal shadow candidate detection processing unit 22, a feature amount calculation The unit 24 and the discrimination processing unit 26 function as an abnormal shadow detection unit in the present invention, and the display control unit 28 and the display unit 30 function as an image display unit in the present invention.

  The image data storage unit 10 stores medical image data representing a plurality of types of medical images of each patient's chest. The multiple types of medical images include simple radiation images, temporal subtraction images, energy subtraction images, and the like. Each medical image data includes, as supplementary information, information identifying the patient ID information represented by the medical image and the type of the medical image. The image type setting unit 12 and the image data acquisition unit 14 described later are Using this incidental information, medical image data representing a specific type of medical image of a specific patient can be searched.

  The image type setting unit 12 determines the type of medical image used for detecting an abnormal shadow for each divided region obtained by dividing the lung field region of the chest on the image into predetermined small regions separately for left and right. Set. In each divided area, the priority order of the types of medical images to be set (priority of which image to use next when there is no type of medical image to be used) is determined in advance. The type with the highest priority among the medical image data stored in the image data storage unit 10 is set for the region.

  The image data acquisition unit 14 searches and acquires medical image data representing each type of medical image set in each divided region by the image type setting unit 12 from the image data acquisition unit 10 by type.

  Based on the medical image data acquired by the image data acquisition unit 14, the lung field region recognition unit 16 uses a known image analysis technique to analyze the anatomy of the chest lung field in each type of medical image represented by each medical image data. Target information is acquired and the lung region is recognized.

  The lung field region dividing unit 18 performs a normalization process to unify the lung field regions in the various types of medical images recognized by the lung field region recognition unit 16 into the same shape and size, and each normalized type. The lung field region in the medical image is divided into the predetermined small regions for the left and right lungs.

  The detection target region setting unit 20 abnormally shades a region corresponding to the divided region and the type of the medical image set by the image type setting unit 12 among the divided regions divided on each medical image. Is set as the detection target area.

  The abnormal shadow candidate detection unit 22 includes an image of each detection target region set by the detection target region setting unit 20 in the medical image data acquired by the image data acquisition unit 14 (a type corresponding to each division region and its division region). Based on the data corresponding to the image cut out on the medical image), an abnormal shadow candidate as an abnormal shadow candidate is detected by a known method in the image of each detection target region.

  For each abnormal shadow candidate detected by the abnormal shadow candidate detection unit 22, the feature amount calculation unit 24 uses the image data representing the image of the candidate area and the vicinity area of the candidate to determine whether the candidate has an abnormal shadow likelihood. A plurality of reflected feature quantities are calculated.

  For each abnormal shadow candidate, the discrimination processing unit 26 has an individual feature amount calculated for each candidate or a new index value calculated using these individual feature amounts exceeding a predetermined reference. Is determined as an abnormal shadow.

  The display control unit 28 converts the medical image represented by any one of the medical image data acquired by the image data acquisition unit 14 or the partial medical image in which each detection target region is set into each divided region. Information indicating the position of a shadow determined to be an abnormal shadow by the determination processing unit 26 is displayed on the screen of the display unit 30 with an image arranged so as to have a positional relationship equivalent to the positional relationship of

  The display unit 30 includes a CRT, a liquid crystal panel, a projector, or the like, and displays the detection result of the abnormal shadow on the screen under the control of the display control unit 28.

  Next, the operation of the abnormal shadow detection apparatus 1 according to this embodiment will be described. FIG. 2 is a flowchart showing an operation flow in the abnormal shadow detection apparatus 1.

  As shown in FIG. 3, the image type setting unit 12 divides the chest lung region on the medical image into a first region R1 including the lung apex and a second region R2 including the center of the lung field separately for the left and right. The type of medical image used for detecting an abnormal shadow is set for each of these divided regions when the region is divided into three regions R3 including the hilar region. by.

  First, medical image data representing a medical image of the chest of the patient is searched from medical image data stored in the image data storage unit 10 based on information for specifying a patient for which an abnormal shadow is to be detected. Then, it is examined which type of medical image data representing the medical image exists (step # 1).

Here, as shown in Table 1, the priority order of the types of medical images to be set is determined in advance in each of the divided regions R1 to R3.

That is, assuming that the temporal subtraction image is Pt, the energy subtraction image is Pe, and the simple radiation image is Px, the region R1 is (1) Pt (or Pe), (2) Pe (or Pt), (3) Px. In order, the region 2 is in the order of (1) Px, (2) Pt (or Pe), (3) Pe (or Pt), and the region R3 is (1) Pt (or Pe), (2) Pe (or Pt) , (3) Priorities are determined in the order of Px.

  These priorities are determined in an order suitable for detecting abnormal shadows in the images of the respective regions. For example, for the region R1 including the apex of the lung, the region includes normal structures similar to actual lesions such as overlap of the clavicle and ribs, and these may be erroneously detected as lesions. It is determined to preferentially use the temporal subtraction image and the energy subtraction image from which the structure is removed. Further, for example, regarding the region R2 including the central portion of the lung field, the region has few images that obstruct observation, and it is often easy for a doctor to find a lesion. Here, the temporal subtraction image Pt and energy subtraction are used here. If the image Pe is used, a part similar to the lesion is created in the part where the lesion does not exist, and the result may be a mysterious abnormal shadow detection process that does not match the human sense. In order to obtain an abnormal shadow detection result, it is determined that the simple radiation image is preferentially used. In addition, for example, for the region R3 including the hilar portion, as in the region R1 including the lung apex, the region includes normal structures such as pulmonary blood vessels, and these may be erroneously detected as lesions. It is determined to preferentially use a temporal subtraction image or energy subtraction image from which these normal structures have been removed.

  Therefore, the image type setting unit 12 can be known by the previous search as the type of medical image used for detecting abnormal shadows for the divided regions R1 to R3 of the lung field region according to the priority order of Table 1. Among the types of medical images, the type with the highest priority is set (step # 2). For example, when the simple radiation image Px and the temporal subtraction image Pt exist, as shown in FIG. 4, the temporal subtraction image Pt is set in the regions R1 and R3, and the simple radiation image Px is set in the region R2. For example, when the simple radiation image Px and the energy subtraction image Pe exist, as shown in FIG. 5, the energy subtraction image Pe is set in the regions R1 and R3, and the simple radiation image Px is set in the region R2.

  When the type of medical image is set for each divided region by the image type setting unit 12, the image data acquisition unit 14 selects the type of medical image data representing the set type of medical image from the image data storage unit 10. Separately read out and obtain (step # 3).

  When the image data acquisition unit 14 acquires medical image data, the lung field region recognition unit 16 performs, for example, Japanese Patent No. 3433928 (a rib cage boundary detection method for a rib cage chest image and a digital chest image diagnostic apparatus) based on each acquired medical image data. ), The landmark information is detected from the digital chest image, and the rib cage boundary is detected from the landmark information, so that each type of medical image represented by each medical image data A lung field region is recognized (step # 4).

  When the lung field region recognition unit 16 recognizes a lung field region on each medical image, the lung field region division unit 18 first performs processing such as enlargement / reduction, parallel movement, warping, and the like on each type of medical image. Thus, a normalization process is performed to unify the lung field regions in each type of medical image recognized by the lung field region recognition unit 16 into the same shape and size (step # 5). The lung field region in the medical image is divided into the regions R1 to R3 according to the left and right lungs (step # 6). Here, as shown in FIG. 6 (1), when a straight line L1 is drawn horizontally at a position that is lowered by 30% of the maximum vertical width of the lung field from the uppermost end of the lung field, the contour of the lung field is drawn. The upper region surrounded by the straight line L1 is R1, and is 60% of the lateral width of the lung field from the outer edge of the lung field outline (the side edge extending in the vertical direction away from the spinal cord). When the curve L2 is drawn parallel to the outer edge at the position moved inward, the lung field contour including the outer edge, the straight line L1, the area surrounded by the curve L2 is R2, and the remaining area is R3. To divide.

  In addition to the above, the lung field region may be divided as shown in FIG. That is, when a straight line L1 is drawn horizontally at a position that is lowered by 30% of the maximum vertical width of the lung field from the uppermost end of the lung field, the upper region surrounded by the lung field outline and the straight line L1 R1 and the innermost and lowermost position of the region R1 as the point M1, the outer edge of the lung field outline on the lower edge of the lung field outline (the side edge extending in the vertical direction away from the spinal cord) ) From the point M2 to the inner side by 30% of the lateral width of the lung field, and when a straight line L3 is drawn from the point M1 to the point M2, the lung field outline including the outer edge, The area surrounded by the straight lines L1 and L3 is divided into R2, and the remaining area is divided into R3.

  Further, as a method of dividing the lung field region, in addition to the above, a neural network is used in the vicinity of the longitudinal section, etc. described in Patent Document 1, and other regions are processed using, for example, erosion processing. A technique such as division can be used.

  When the lung field region on each medical image is divided by the lung field region dividing unit 18, the detection target region setting unit 20 includes the divided regions in the divided regions as shown in FIGS. And a region corresponding to the type of medical image set by the image type setting unit 12 is set as an abnormal shadow detection target region (step # 7). However, when the images used in each of the regions R1 to R3 are different, for example, when it is determined that the temporal subtraction image Pt is used in the region R1 and the simple radiation image Px is used in the region R2, the respective regions are near the boundary of the region. For example, the detection target region is set so as to protrude about one tumor. By doing in this way, even if a tumor exists on the boundary line, it can be detected without any problem.

  When the detection target region setting unit 20 sets the detection target region, the abnormal shadow candidate detection unit 22 includes each detection target region set by the detection target region setting unit 20 among the medical image data acquired by the image data acquisition unit 14. On the basis of data corresponding to the image (an image obtained by cutting each divided region on a medical image of a type corresponding to the divided region), an abnormal shadow candidate that is a candidate for an abnormal shadow is detected in each detection target region image. Detect (step # 8). As a technique for detecting a candidate for an abnormal shadow, for example, a density gradient (or luminance gradient) in an image, which is disclosed in Japanese Patent Laid-Open No. 2002-109510, is represented as a density gradient vector. Iris filter processing that extracts a high image part as a candidate (tumor shadow) and multiple structural elements according to the size of the abnormal shadow to be extracted, and density in a spatial range narrower than this multiple structural element A technique using morphological filter processing for extracting a fluctuating image portion as a candidate (calcified shadow) can be considered.

  When the abnormal shadow candidate is detected by the abnormal shadow candidate detection unit 22, the feature amount calculation unit 24, for each abnormal shadow candidate detected by the abnormal shadow candidate detection unit 22, the candidate area and the vicinity area of the candidate. Using the image data representing the image, a plurality of feature amounts reflecting the candidate abnormal shadowiness are calculated (step # 9). As the feature amount, for example, described in “Improvement of ability to discriminate malignant mass on mammogram by feature amount selection and evaluation of selection criteria” (Electronic Information and Communication Society paper D-IIVol86-DIINo5, pp587-597). Specifically, the candidate area, circularity, irregularity, average luminance value, contrast, luminance value distribution, luminance value maximum / minimum values, density histogram Skewness, kurtosis, etc. can be employed.

  When the feature amount calculation unit 24 calculates the feature amount of each abnormal shadow candidate, the discrimination processing unit 26 uses, for each abnormal shadow candidate, the individual feature amount calculated for each candidate or these individual feature amounts. When the new index value calculated in this way exceeds a predetermined standard, it is determined as an abnormal shadow (step # 10), and the display control unit 28 selects any one of the medical image data acquired by the image data acquisition unit 14 In the background, a medical image represented by one medical image data or an image in which a partial medical image in which each detection target region is set is arranged so as to have a positional relationship equivalent to the positional relationship of each divided region. As information indicating the position of the shadow determined as an abnormal shadow by the determination processing unit 26, for example, an arrow mark or the like indicating the shadow is displayed on the screen of the display unit 30 (step # 11). FIG. 7 shows an example in which the detection result is displayed with the simple radiation image Px as the background, and FIG. 8 shows an example in which the detection result is displayed with the image of each detection target area as the background.

  As described above, according to the abnormal shadow detection apparatus 1 according to the present embodiment, when detecting an abnormal shadow, the region to be detected is divided, and the medical image used for detecting the abnormal shadow is divided for each divided region. Since the types are determined, abnormal shadows are used for each divided region by using a medical image of a type that provides better detection of abnormal shadows considering the anatomical characteristics of the subject (lung field) in each divided region. Can be detected, and the overall abnormal shadow detection performance can be further improved. The improvement in detection performance means that a shadow (FP) in which a shadow that is not truly an abnormal shadow is erroneously detected as an abnormal shadow, or a true abnormal shadow (FN) that should be detected but not detected is reduced. Say that.

  In the calculation of the feature amount in the present embodiment, it is preferable to change the combination of the feature amount types to be calculated depending on the type of medical image. This is because the detection performance of abnormal shadows can be further improved by using a combination of feature quantities that more accurately reflect abnormal shadows for each type of medical image. This type of combination may be determined based on a theoretical basis, but may be determined statistically. For example, abnormal shadow detection processing is performed on a large number of simple radiological images, temporal subtraction images, energy subtraction images, etc. whose correct answers are known (where the lesions are known), Calculating feature quantities and, for each combination of these feature quantities, determine the combination of feature quantities that provides the best result of discrimination processing to discriminate true abnormal shadows for each type of medical image. be able to.

  Further, the aspect of division of the chest lung field region and the priority order of the types of medical images used for abnormal shadow detection for each divided region in this embodiment are merely examples, and the present invention is limited to these specific contents. It is not a thing.

The figure which shows the structure by one Embodiment of the abnormal shadow detection apparatus of this invention. The figure which shows the flowchart showing the flow of operation | movement in the abnormal shadow detection apparatus 1. The figure which shows each division area of the lung field area in a chest medical image The figure which shows an example of the correspondence of each division area and the kind of medical image used for every area (the 1) The figure which shows an example of the correspondence of each division area and the kind of medical image used for every area (the 2) The figure for demonstrating the method of division | segmentation of the lung field area | region in a medical image The figure which shows the example of the screen where the detection result of the abnormal shadow is displayed with the simple radiographic image as the background The figure which shows the example of the screen on which the detection result of the abnormal shadow was displayed against each image of the detection target area

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Abnormal shadow detection apparatus 10 Image data storage part 12 Image type setting part 14 Image data acquisition part 16 Lung and area | region recognition part 18 Lung field area division part 20 Detection object area | region setting part 22 Abnormal shadow candidate detection part 24 Feature-value calculation part 26 Discrimination processing unit 28 Display control unit 30 Display unit

Claims (11)

Acquire each medical image data representing a medical image of a type defined for each divided area corresponding to each of a plurality of divided areas obtained by dividing the area of the subject on an image in a predetermined depiction form of the subject. And
An abnormal shadow detection method, comprising: detecting an abnormal shadow in a medical image of each divided region represented by each medical image data based on each acquired medical image data. For each of a plurality of divided regions obtained by dividing the subject region on an image in a predetermined depiction form of the subject, an abnormal shadow in the medical image corresponding to the divided region is detected for each divided region. Set the type of the medical image used when
Obtaining medical image data representing a medical image of the subject for each of the set types;
Acquiring the anatomical information of the subject in each type of acquired medical image represented by each acquired medical image data,
Based on the anatomical information, each segmented area is extracted on the acquired medical image of the type set for each segmented area,
Each extracted divided area is set as a detection target area of an abnormal shadow,
An abnormal shadow detection method, comprising: detecting an abnormal shadow in an image of each detection target region based on data representing an image of each detection target region among the various types of medical image data. Acquire each medical image data representing a medical image of a type defined for each divided area corresponding to each of a plurality of divided areas obtained by dividing the area of the subject on an image in a predetermined depiction form of the subject. Image data acquisition means for
Abnormal shadow detection, comprising: an abnormal shadow detection means for detecting an abnormal shadow in the medical image of each divided region represented by each medical image data based on each acquired medical image data apparatus. For each of a plurality of divided regions obtained by dividing the subject region on an image in a predetermined depiction form of the subject, an abnormal shadow in the medical image corresponding to the divided region is detected for each divided region. Image type setting means for setting the type of the medical image used when
Image data acquisition means for acquiring medical image data representing a medical image of the subject for each of the set types;
Anatomical information acquisition means for acquiring anatomical information of the subject in each type of acquired medical image represented by each acquired medical image data;
Area extracting means for extracting the divided areas on the acquired medical image of the type set for each divided area based on the anatomical information;
Detection target area setting means for setting each of the extracted divided areas as a detection target area of an abnormal shadow;
An abnormal shadow detecting unit that detects an abnormal shadow in the image of each detection target area based on data representing the image of each detection target area among the various types of medical image data. Abnormal shadow detection device. Image data storage means for storing medical image data of a plurality of types of subjects,
The image type setting unit stores the type of the medical image for each divided region in the image data storage unit based on the priority order of the type of the medical image set in advance for each divided region. The abnormal shadow detection apparatus according to claim 4, wherein among the types of medical images represented by each medical image data being set, the type having the highest priority order is set.   The abnormal shadow detection unit detects an abnormal shadow by performing a predetermined abnormal shadow detection process corresponding to the type of the image on the image of the detection target region. Item 6. The abnormal shadow detection apparatus according to Item 5.   An image display means for displaying on one screen information indicating the position of the abnormal shadow detected by the abnormal shadow detection means together with any one of the acquired medical images. Item 7. An abnormal shadow detection apparatus according to item 5, 5 or 6.   Information indicating the position of the abnormal shadow detected by the abnormal shadow detection means together with an image arranged so that the image of each detection target area has a positional relationship equivalent to the positional relationship of each of the divided regions. 7. The abnormal shadow detection apparatus according to claim 4, further comprising image display means for displaying on a screen. The subject is a lung field of a human chest,
One of each of the divided regions is any one of a region including a lung apex, a region including a central lung field, and a region including an outer edge,
The abnormal shadow detection apparatus according to any one of claims 2 to 6, wherein the type of medical image includes any one of a temporal subtraction image, an energy subtraction image, and a simple radiation image. Computer
Acquire each medical image data representing a medical image of a type defined for each divided area corresponding to each of a plurality of divided areas obtained by dividing the area of the subject on an image in a predetermined depiction form of the subject. Image data acquisition means for
A program for functioning as an abnormal shadow detecting means for detecting an abnormal shadow in a medical image of each divided region represented by each medical image data based on each acquired medical image data. Computer
For each of a plurality of divided regions obtained by dividing the subject region on an image in a predetermined depiction form of the subject, an abnormal shadow in the medical image corresponding to the divided region is detected for each divided region. Image type setting means for setting the type of the medical image used when
Image data acquisition means for acquiring medical image data representing a medical image of the subject for each of the set types;
Anatomical information acquisition means for acquiring anatomical information of the subject in each type of acquired medical image represented by each acquired medical image data;
Area extracting means for extracting the divided areas on the acquired medical image of the type set for each divided area based on the anatomical information;
Detection target area setting means for setting each of the extracted divided areas as a detection target area of an abnormal shadow;
A program for functioning as an abnormal shadow detecting means for detecting an abnormal shadow in an image of each detection target area based on data representing an image of each detection target area among the various types of medical image data.

Images