JP2006280682A - Method of supporting diagnostic image provided with noise detection function - Google Patents

Method of supporting diagnostic image provided with noise detection function Download PDF

Info

Publication number
JP2006280682A
JP2006280682A JP2005105595A JP2005105595A JP2006280682A JP 2006280682 A JP2006280682 A JP 2006280682A JP 2005105595 A JP2005105595 A JP 2005105595A JP 2005105595 A JP2005105595 A JP 2005105595A JP 2006280682 A JP2006280682 A JP 2006280682A
Authority
JP
Japan
Prior art keywords
pixel
image
noise
images
lesion candidate
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.)
Pending
Application number
JP2005105595A
Other languages
Japanese (ja)
Inventor
Mitsuhisa Himaga
Yoshitaka Hiramatsu
Tatsuhiko Kagehiro
Hirotaka Nagayoshi
Yutaka Sako
義崇 平松
達彦 影広
充寿 日間賀
洋登 永吉
裕 酒匂
Original Assignee
Hitachi Omron Terminal Solutions Corp
日立オムロンターミナルソリューションズ株式会社
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 Hitachi Omron Terminal Solutions Corp, 日立オムロンターミナルソリューションズ株式会社 filed Critical Hitachi Omron Terminal Solutions Corp
Priority to JP2005105595A priority Critical patent/JP2006280682A/en
Publication of JP2006280682A publication Critical patent/JP2006280682A/en
Pending legal-status Critical Current

Links

Images

Abstract

<P>PROBLEM TO BE SOLVED: To precisely detect noise caused by a soil etc. on a lens from a medical image such as a fundus oculi image with a reduced burden on a patient, and to precisely extract a candidate area of a lesion from the medical image. <P>SOLUTION: In each color component image in a plurality of sheets of images obtained by photographing a pluraity of subjects, pixels having pixel values not smaller than the pixel value of neighboring pixels are detected as lesion pixels, and pixels detected as the lesion pixels continuously in images of sheets not smaller than a prescribed value are detected as noise pixel candidates. By searching timing of noise occurrence going back to a plurality of sheets of images in a time-series order of photographing timing, from an image next to the newest image where the position of a noise pixel is neither the lesion candidate pixel nor a part of an internal organ which has to be photographed, presence of noise at the position of the noise pixel is judged. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a method for detecting noise from a medical image obtained by photographing a subject, and an image diagnosis for detecting a lesion present in the image and displaying the lesion on a display device to improve diagnosis efficiency. It is about support method.

  When dirt such as dust adheres to optical components such as a lens of a camera or an image sensor of a digital camera has a defect, noise such as black spots or white spots appears in an image taken by the camera. In the image diagnosis support apparatus, this noise may be misunderstood as a lesion. Therefore, it is necessary to distinguish the noise from the lesion, and it is necessary to remove the noise depending on the application.

  As a method for detecting noise generated in an image due to dirt or the like adhering to a lens or the like of a camera, many methods for detecting or removing noise using information related to the operation of the imaging apparatus have been devised. In Japanese Patent Application Laid-Open No. 2004-16486 (Patent Document 1), noise is removed using a correction pattern image previously captured by an imaging device. The correction pattern image creating method stores in advance the image obtained by changing the focal position and the focal position information, and calculates the ratio between the focal position at the time of photographing and the focal position stored in advance. A correction pattern corresponding to the focal position at the time of photographing the subject is created by calculating the ratio to an image stored in advance. Japanese Patent Laid-Open No. 2004-16486 also discloses a method in which an image obtained by photographing an optical member having a uniform reflecting surface in advance is used as a correction pattern image.

  In Japanese Patent Laid-Open No. 2004-153422 (Patent Document 2), noise is detected by taking a difference between a reference image captured in advance and an image obtained by capturing a subject. Japanese Patent Laid-Open No. 2004-153422 also discloses a method for detecting noise by taking a difference between an image obtained by photographing a subject and an image obtained by photographing the subject after moving or rotating the camera back and forth.

JP 2004-16486 A

JP 2004-153422 A

  In the above prior art, in the technique using an image obtained by photographing an optical member composed of a uniform reflecting surface in Patent Document 1, it is necessary to place the optical member in front of the camera every time in order to detect noise. Increase the labor of the person. Considering that the technique using the reference image taken in advance in Patent Document 2 is used for detecting noise in the fundus image for assisting image diagnosis, it is difficult to obtain the reference image because there are large individual differences in the fundus image. . Further, with these techniques, noise cannot be detected from a fundus image that has already been taken. Furthermore, if a technique for removing noise by repeatedly moving or rotating an image obtained by capturing an image of a subject and the subject of Patent Document 2 to remove noise by repeatedly photographing the subject is required for medical images, the subject needs to be photographed multiple times. There are significant disadvantages in photographing fundus images and the like that are burdensome to the user.

  Accordingly, an object of the present invention is to provide a method for detecting noise caused by dirt on a lens from a medical image such as a fundus image without increasing the number of times of shooting by shooting a correction pattern image or shooting a subject multiple times. By providing, a method for accurately extracting a lesion from a medical image is realized.

  The image diagnosis support method of the present invention includes (1): a first step of accepting input of a plurality of images to an image diagnosis support processing unit connected to the image input device, and each of the color component images of the plurality of images. A second step of detecting a lesion candidate pixel based on a pixel value of each pixel, a third step of determining a pixel that is repeatedly detected as a lesion candidate pixel in the plurality of images as a noise pixel, A fourth step in which a lesion candidate pixel detected in a plurality of images other than the position of the noise pixel is determined as a lesion candidate region; and information on the lesion candidate region and the noise pixel is associated with each of the plurality of images. And storing the fifth step.

  (2): In the image diagnosis support method (1), in the second step, the detection results for the plurality of images are accumulated for each pixel, and the accumulation result is referred to in the third step. A pixel whose cumulative value detected as a lesion candidate pixel is a threshold value or more is determined as a noise pixel.

  (3): Image diagnosis support method (2), characterized in that, in the second step, a lesion candidate pixel is detected on the condition that the pixel is not a part of an organ to be imaged. .

  (4): The diagnostic imaging support method according to any one of (1) to (3), wherein in the third step, the noise generation timing is searched retrospectively in the time series of the imaging timing, The noise pixel position is not a lesion candidate pixel, and it is determined that there is noise at the noise pixel position from the next image after the latest image that is not a part of an organ to be imaged. To do.

  (5): A diagnostic imaging support method according to any one of (1) to (4), wherein a noise region is created by connecting the pixels detected as the noise, and the noise region is displayed on a display device. These steps are executed.

  (6): The image diagnosis support method according to any one of (1) to (4), wherein a pixel group that is not determined to be a noise pixel among pixel groups in the vicinity of the noise pixel detected as the noise. An average value is calculated, an image in which the pixel value of the noise pixel is replaced with the previous average value is created, and a seventh step of displaying the created image on a display device is executed.

  (7): The image diagnosis support method according to any one of (1) to (6), wherein the eighth step of displaying information on the candidate lesion area in association with the image on the display device is executed. It is characterized by that.

  ADVANTAGE OF THE INVENTION According to this invention, a test subject's burden can be reduced and a noise can be detected correctly, without taking the manual labor. By this noise detection method, it is possible to provide an image diagnosis support method in which a false positive rate, which is a rate at which a healthy person is determined to be sick, is reduced.

Hereinafter, the best mode for carrying out the present invention will be described with reference to the drawings by examples of fundus images.
A schematic diagram of a fundus image without a lesion is shown in FIG. 7A, and a schematic diagram of a fundus image of a diabetic retinopathy patient is shown in FIG. 7B. A blood vessel (703), an optic disc (704), and a macula (705) are imaged in a fundus image without a lesion. On the other hand, in the fundus image of a diabetic retinopathy patient, in addition to the blood vessels, the optic disc, and the macula, lesions such as light-colored vitiligo (701) and dark-colored bleeding (702) are imaged. One method of diagnosis support for a fundus image is to detect a lesion candidate area from the fundus image and present the lesion candidate area to a doctor.

  If the fundus image has a light-colored region or dark-colored region due to noise, the noise may be mistaken for white spots of light-colored lesions or bleeding of dark-colored lesions. Therefore, processing is performed such as detecting noise in the fundus image and displaying the noise region on a display device or interpolating the noise region. The noise detection method utilizes the fact that a pixel that is frequently detected as a lesion candidate pixel is likely to be imaged in dark or light tones due to noise.

Hereinafter, an image diagnosis support method for a plurality of fundus images obtained by photographing in advance using an imaging unit such as a digital fundus camera will be described.
FIG. 1 is a processing flow diagram illustrating an image diagnosis support method for detecting a lesion present in a fundus image and displaying the detected lesion on a display device. In the image diagnosis support processing, a plurality of fundus images obtained by imaging means such as a digital fundus camera are captured as image input (101), and the pixel value of the green component image of the image input is predetermined as compared to surrounding pixels. A candidate region of a lesion composed of pixels higher or lower than the threshold value is extracted (102), and the extraction result is written in each detection result image (108). The detection result image (108) is data including a label indicating that each pixel value is a lesion or noise, and is stored in the detection result storage unit 109 in the memory of the image diagnosis support apparatus described later. Next, the detection result image 108 stored in the detection result storage unit 109 is referred to, a pixel repeatedly detected as a lesion candidate pixel is determined as a noise pixel, and the determination result is written in the detection result image 108 (103) ). Next, for each of the noise pixels, a plurality of images are traced back in chronological order to determine an image in which noise has occurred, and noise information is written in the detection result image 108 corresponding to the determined image (104). ). After determining the noise generation image, when displaying the fundus image of the newest image in time, a warning is displayed near the area determined to be noise, and an image that emphasizes the periphery of the area determined to be a lesion is created. Display on the display device (105). After that, for each of the past images before the most recent image in time, a warning is issued near the area determined to be noise, and an image highlighting the periphery of the area determined to be a lesion is created, if necessary The images may be displayed on the display device one by one.

  The process of issuing a warning near the area determined to be noise may be performed only when instructed by a doctor. Further, instead of issuing a warning in the vicinity of the area determined as noise, it is possible to correct the area determined as noise.

Hereinafter, the details of each part of the image diagnosis support apparatus that executes the image diagnosis support process and the noise detection process and the process of FIG. 1 will be described with reference to the drawings. Note that the image input 101 is the same as in the prior art.
FIG. 2 shows the configuration of an image diagnosis support apparatus that executes image diagnosis support processing and noise detection processing. In FIG. 2, 201 is a camera for photographing the fundus of the subject, and 209 is a detection device that performs steps 101 to 105 in FIG. The camera 201 and the detection device 209 are connected by a cable 208. Here, instead of the camera 201, a scanner that captures an already filmed film as an image, a digital image reading device, or the like may be used.

  The detection device 209 includes a bus 206 for connecting each unit in the device, an image input unit 202 that performs image input 101 from the camera 201, a CPU 203 that performs overall control of the detection device 209, noise detection processing, and image diagnosis support processing, Display unit 204 for displaying noise detection results and lesion candidate region detection results, operation unit 205 for performing operations at startup and writing of diagnosis results, and image data (detection for noise detection processing and image diagnosis support processing) A memory 207 for storing the result image 108 and the like, a program, and the like. Steps 101 to 105 in FIG. 1 are realized by the CPU 203 executing a program stored in the memory 207.

  FIG. 3 shows a lesion candidate region extraction processing flow (102). The lesion candidate region extraction processing 102 detects a lesion candidate pixel based on the pixel value of each pixel for each of the color component images of the plurality of images, and the lesion candidate is provided on the condition that the pixel is not a part of the fundus tissue. This is a process of extracting pixels. The lesions to be extracted are light-colored lesions and dark-colored lesions. In the lesion candidate region extraction process, first, i representing the order from the oldest in time is set to 1 (301), and the fundus tissue region is detected. In the fundus tissue region detection process, a blood vessel region, an optic disc region, and a macular region are detected from an image input, and the detection result is written in the detection result image 108. As a method for detecting a blood vessel region, for example, a known technique described in the Journal of the Institute of Electronics, Information and Communication Engineers, Vol. E87-D, No. 1, pp. 155-163 is applied. As a method for detecting the optic nerve head area, for example, a known technique described in IEEE Transactions on Medical Imaging, Vol. 21, No. 10, October 2002 is applied. As a method for detecting the macular region, for example, a known technique described in JP-A-7-136123 is applied. As a result, detection result image data as shown in FIG. 9A-1 is generated for each input image. Here, a flag of 9 is set for pixels in the blood vessel region, a flag of 8 is set for pixels in the optic disc region, and a flag of 7 is set for pixels in the macular region.

  Next, in the detection result image data of the i-th image, the i-th pixel among the pixels at positions excluding the blood vessel region pixel (flag 9), the optic disc region pixel (flag 8), and the macular region pixel (flag 7). Dark pixels that are equal to or greater than a predetermined threshold value than the pixel values of neighboring pixels of the green component image of the image are detected as dark tone lesion candidate pixels (303), and the detection result is written in the detection result image. Further, a pixel that is brighter than a pixel value of a neighboring pixel of the green component image of the i-th image by a predetermined threshold or more is detected as a light tone lesion candidate pixel (304), and the detection result is written in the detection result image 108. Here, instead of the green component image, a color component image in which the lesion and other contrast are high may be used. Alternatively, a green component image or a color component image with high contrast may be used after image correction. The pixel values of the neighboring pixels may be calculated by including or excluding the pixels of the blood vessel region, the optic disc region, and the macular region. The detection method of the lesion candidate pixel is not limited to this method, and known techniques described in, for example, David Benjamin Usher, “Image Analysis for the Screening of Diabetic Retinopathy”, Doctor of Philosophy Thesis, King's College, University of London. Applicable. As a result, for each input image, the detection result image data is changed as shown in FIG. Here, a flag of 1 is set for pixels that are darker than neighboring pixels, and a flag of 2 is set for pixels that are brighter than neighboring pixels. If i is not w (305), i = i + 1 (306), and the processing from 302 is repeated. As described above, 302, 303, and 304 are repeated from the oldest image (i = 1) to the latest image (i = w).

  FIG. 4 shows a noise pixel detection processing flow (103). In the noise pixel detection process 103, detection results for a plurality of images are accumulated for each pixel, and the accumulated result detected as a lesion candidate pixel is determined as a noise pixel by referring to the accumulation result. Hereinafter, processing for detecting black noise pixels will be described. The lesion in the above description refers to a dark-colored lesion. The cumulative frequency of lesion candidate pixels (flag 1) is calculated by referring to each pixel of the detection result image 108 of the w-th image from the first image stored in the detection result storage unit 109, and each pixel is a lesion candidate pixel. An accumulated frequency image that is the accumulated frequency is created and stored in the memory (401). FIG. 9B shows an example of cumulative frequency image data. Here, for each pixel, the cumulative frequency of how many pixels from the first image to the w-th image have been detected as lesion candidate pixels is stored. Next, a pixel having a pixel value equal to or greater than a threshold in the cumulative frequency image is detected as a noise pixel (402), and the result is written in the noise detection result image 110. This is because a pixel that is frequently detected as a lesion candidate pixel is considered to have a high possibility of being captured in a dark tone due to noise. For example, if the threshold value is 5, for example, a pixel whose cumulative frequency of detection as a lesion candidate pixel is 9 is determined as a noise pixel, and a pixel whose cumulative frequency is 1 is not a noise pixel. It is determined that the pixel is correctly detected as a lesion pixel. The noise detection result image (110) is stored in the detection result storage means 109 in the memory of the diagnostic imaging support apparatus.

  FIG. 5 shows a determination processing flow (104) of a noise occurrence image. In the noise generation image determination processing 104, the noise generation timing is searched by going back the plurality of images in time series of imaging timing, and the position of the noise pixel is not a lesion candidate pixel and is not a part of the fundus tissue. It is determined that there is noise at the position of the noise pixel from the next image of the latest image. In the noise generation image determination processing flow, the position of each noise pixel is referred to from the detection result image 110 and stored in the processing target position list (501). K representing the order from the newest image in time is set to w (502), and the pixel at the processing target position is referred to from the detection result image 108 in the k-th image and the (k + 1) -th image among the plurality of fundus images (503). ). When the reference result of the kth image is a lesion candidate pixel (flag 1) (504), the pixel is determined as noise (507), and the result is written in the detection result image 108 of the kth image. When the reference result of the k-th image is not a lesion pixel (flag 1) (504), the reference result of the k + 1-th image is a noise pixel (flag 3), and the reference result of the k-th image is a pixel of the blood vessel region (flag 9). Alternatively, if it is a pixel in the macular region (flag 7), the pixel is determined as noise (507), and the result is written in the detection result image 108 of the kth image. After the writing, it is checked whether or not the processing from 502 to 507 in the k-th image has been executed for all processing target positions (509).

  When the reference result of the k-th image is not a lesion pixel (flag 1) (504), if the reference result of the k + 1-th image is not a noise pixel (flag 3), the processes from 502 to 507 in the k-th image are all processed. It is investigated whether or not it has been executed for the target position (509).

  When the reference result of the kth image is not a lesion pixel (flag 1) (504), the reference result of the k + 1 image is a noise pixel (flag 3), and the reference result of the kth image is a pixel of the blood vessel region (flag 9). Alternatively, if it is not a pixel (flag 7) in the macular region, it is investigated whether or not the processing from 502 to 507 in the kth image has been executed for all the processing target positions (509).

  When there are processing target positions in which the processing from 502 to 507 has not been executed in the k-th image (509), the processing from 502 to 507 is repeated for the pixels at the processing target positions of the k-th image. When the processing from 502 to 507 has been executed at all the processing target positions in the kth image, if k = 1 is not set (510), k = k−1 is set (511), and the processing target position of the kth image is set. The processing from 502 to 507 is repeated for the pixel. If k = 1 (510), the noise generation image determination processing is terminated.

  When the reference result of the kth image is not a lesion pixel (flag 1) (504), the reference result of the k + 1 image is a noise pixel (flag 3), and the reference result of the kth image is a pixel of the blood vessel region (flag). 9) or when it is not a pixel (flag 7) in the macular region, it is presumed that the noise is generated after the k-th image is taken and before the k + 1-th image is taken, and the noise detection process before the k-th image It is also possible to end the noise detection process for the past image without performing.

  FIG. 6 shows a processing flowchart of display (105). The display processing flow refers to the detection result image 108 (601), and when a noise pixel (flag 3) exists in the detection result image (602), the noise pixels (flag 3) are connected to each other to define a noise region. Generate (603), and create a new display image with a warning message added at a position near the noise region in the input image (604). Next, when there is a lesion candidate pixel (flag 1) in the detection result image (605), the lesion candidate pixel (flag 1) in the detection result image is connected to generate a lesion candidate region (606), and the display An image in which the lesion candidate area is emphasized is added to the image for use (607). The display image is displayed on the display unit 204 (609) to inform the doctor of the presence of noise.

  An example in which the noise warning message is displayed on the display unit 204 is shown in FIG. Reference numeral 807 denotes a schematic diagram of a fundus image. 701 is a light-colored lesion, 702 is a dark-colored lesion, 703 is a blood vessel, 704 is an optic disc, 705 is a macula, 806 is a warning message, and 808 is noise. As shown in FIG. 8, at the same time as displaying a warning of noise, a display that emphasizes the area around the area determined to be a lesion candidate is performed, thereby notifying the doctor of the presence of a lesion candidate. By displaying the location that is judged to be noise together with the lesion candidate warning, the detection result of the lesion candidate area with higher detection accuracy can be presented to the doctor, so that the person who is ill is determined to be ill It helps to improve the degree.

  Further, it is possible to display an image obtained by performing interpolation processing on the pixels determined to be noise in 602. As an example of the interpolation processing, among pixels around the pixel determined to be noise, a method of setting an average of pixel values not determined to be noise to a pixel value of pixels determined to be noise, a nearest neighbor method, There are methods for determining pixel values of pixels determined to be noise by a bilinear method, a bicubic method, or the like. Displaying these interpolated images to a doctor helps to reduce the false positive rate, which is the rate at which a healthy person is determined to be ill.

  In the above, the noise pixel detection 103 and the noise generation image determination 104 have been described for black noise pixels that appear when dirt such as dust adheres to optical components such as a camera lens. Some white noise pixels appear due to defects in the image sensor. In the case of a white noise pixel, the noise pixel detection 103 and the noise generation image determination 104 can be realized by using a part of the optic disc instead of a part of the blood vessel or the macula in 506.

It is a processing flowchart which shows the image diagnosis assistance method in an Example. 1 is a configuration diagram of an image diagnosis support apparatus that performs noise detection and image diagnosis support according to the present invention. FIG. It is a processing flowchart which shows the method of extracting a lesion candidate area | region. It is a processing flowchart which shows the method of detecting a noise pixel. It is a processing flowchart which shows the method of determining a noise generation image. It is a processing flowchart which shows the warning display method of a noise, and the method of displaying a lesion candidate area | region. It is a schematic diagram of a fundus image of a healthy person and a fundus image of a patient with diabetic retinopathy. It is a figure which shows the example of the display screen which displayed the warning of a noise and a lesion candidate area | region. It is a figure which shows the example (A) of a detection result image, and the example (B) of a cumulative frequency image.

Explanation of symbols

  102 ... Extraction of lesion candidate area, 103 ... Detection of noise pixel, 104 ... Determination of noise generation image, 105 ... Noise warning display and lesion candidate area display.

Claims (9)

  1. An image diagnosis support program for detecting and displaying a lesion candidate area from an image, the image diagnosis support processing unit connected to the image input device,
    A first step for accepting input of a plurality of images;
    A second step of extracting lesion candidate pixels based on the pixel value of each pixel for each of the color component images of the plurality of images;
    A third step of determining a pixel detected as a repeated lesion candidate pixel as a noise pixel in the plurality of images;
    A fourth step of determining in the plurality of images a lesion candidate pixel detected other than the position of the noise pixel as a lesion candidate region;
    And a fifth step of storing the lesion candidate area and noise pixel information in association with each of the plurality of images.
  2. The diagnostic imaging support program according to claim 1,
    In the second step, the detection results for the plurality of images are accumulated for each pixel,
    In the third step, referring to the accumulation result, a pixel whose cumulative value detected as a lesion candidate pixel is a noise pixel or more is determined as a noise pixel.
  3. The diagnostic imaging support program according to claim 2,
    A diagnostic imaging support program characterized in that in the second step, a lesion candidate pixel is detected on condition that the pixel is not a part of an organ to be imaged.
  4. The diagnostic imaging support program according to any one of claims 1 to 3,
    In the third step, noise generation timing is searched by tracing back the plurality of images in time series of imaging timing, and the position of the noise pixel is not a lesion candidate pixel and even if it is a part of an organ to be imaged An image diagnosis support program characterized in that it is determined that there is noise at the position of the noise pixel from an image next to the latest one among the images that are not present.
  5. The diagnostic imaging support program according to any one of claims 1 to 4,
    An image diagnosis support program characterized in that a sixth step of creating a noise region by connecting pixels detected as noise and displaying the noise region on a display device is executed.
  6. The diagnostic imaging support program according to any one of claims 1 to 4,
    Of the pixel groups in the vicinity of the noise pixel detected as the noise, calculate the average value of the pixel group that is not determined as the noise pixel, create an image in which the pixel value of the noise pixel is replaced with the previous average value, An image diagnosis support program for executing a seventh step of displaying the created image on a display device.
  7. The diagnostic imaging support program according to any one of claims 1 to 6,
    An image diagnosis support program for executing an eighth step of displaying information on the candidate lesion area in association with the image on a display device.
  8. An image diagnosis support device for detecting and displaying a lesion area candidate from an image,
    An image input unit that receives input of an image, a calculation unit that detects a lesion candidate region from the input image, a display unit that displays information on the detected lesion candidate region, and information on the input image A memory for storing,
    The computing unit is
    For each of the color component images of the plurality of images input from the image input unit, a lesion candidate pixel is detected based on the pixel value of each pixel,
    In the plurality of images, a pixel that is repeatedly detected as a lesion candidate pixel is determined as a noise pixel,
    In the plurality of images, a lesion candidate pixel detected other than the position of the noise pixel is determined as a lesion candidate region,
    An image diagnosis support apparatus, wherein information on the lesion candidate area and noise pixel is stored in the memory in association with each of the plurality of images.
  9. An image diagnosis support method for detecting and displaying a lesion area candidate from an image,
    A first step of receiving input of a plurality of images in the image input unit;
    Executed in the arithmetic unit,
    A second step of extracting lesion candidate pixels based on the pixel value of each pixel for each of the color component images of the plurality of images;
    A third step of determining a pixel detected as a repeated lesion candidate pixel as a noise pixel in the plurality of images;
    A fourth step of determining in the plurality of images a lesion candidate pixel detected other than the position of the noise pixel as a lesion candidate region;
    And a fifth step of storing the lesion candidate area and noise pixel information in association with each of the plurality of images.
JP2005105595A 2005-04-01 2005-04-01 Method of supporting diagnostic image provided with noise detection function Pending JP2006280682A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2005105595A JP2006280682A (en) 2005-04-01 2005-04-01 Method of supporting diagnostic image provided with noise detection function

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2005105595A JP2006280682A (en) 2005-04-01 2005-04-01 Method of supporting diagnostic image provided with noise detection function

Publications (1)

Publication Number Publication Date
JP2006280682A true JP2006280682A (en) 2006-10-19

Family

ID=37403195

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2005105595A Pending JP2006280682A (en) 2005-04-01 2005-04-01 Method of supporting diagnostic image provided with noise detection function

Country Status (1)

Country Link
JP (1) JP2006280682A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008062528A1 (en) * 2006-11-24 2008-05-29 Nidek Co., Ltd. Fundus image analyzer
JP2011092702A (en) * 2009-09-30 2011-05-12 Nidek Co Ltd Eye fundus observation apparatus
JP2015232824A (en) * 2014-06-10 2015-12-24 株式会社デンソー Detector
JPWO2014112611A1 (en) * 2013-01-21 2017-01-19 興和株式会社 Image processing apparatus, image processing method, image processing program, and recording medium storing the program
WO2018074459A1 (en) * 2016-10-18 2018-04-26 興和株式会社 Image processing device, image processing method, and image processing program

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008062528A1 (en) * 2006-11-24 2008-05-29 Nidek Co., Ltd. Fundus image analyzer
JP2011092702A (en) * 2009-09-30 2011-05-12 Nidek Co Ltd Eye fundus observation apparatus
JPWO2014112611A1 (en) * 2013-01-21 2017-01-19 興和株式会社 Image processing apparatus, image processing method, image processing program, and recording medium storing the program
JP2015232824A (en) * 2014-06-10 2015-12-24 株式会社デンソー Detector
WO2018074459A1 (en) * 2016-10-18 2018-04-26 興和株式会社 Image processing device, image processing method, and image processing program

Similar Documents

Publication Publication Date Title
JP5323275B2 (en) Ultrasonic diagnostic apparatus and control program for ultrasonic diagnostic apparatus
CN102626305B (en) Image processing equipment and image processing method
US10441163B2 (en) Ophthalmic diagnosis support apparatus and ophthalmic diagnosis support method
US10383511B2 (en) Image processing apparatus, image processing method, and program
Gagnon et al. Procedure to detect anatomical structures in optical fundus images
JP4745080B2 (en) X-ray diagnostic apparatus, image processing apparatus, and program
JP4819851B2 (en) Diagnosis support apparatus and method, program, and recording medium
JP5665768B2 (en) Image processing apparatus, image processing method, imaging system, and program
KR101267755B1 (en) Image processing apparatus, image processing method, tomogram capturing apparatus, and program recording medium
JP5161427B2 (en) Image photographing apparatus, image processing apparatus, and program
KR101478598B1 (en) Image processing apparatus, control method thereof, and computer program
US8687863B2 (en) Image processing apparatus, control method thereof and computer program
EP1994879B1 (en) Image analysis device and image analysis method
EP2352124B1 (en) Image processing apparatus and image processing method
EP2085019B1 (en) Image processing device, image processing method, and image processing program
JP5281826B2 (en) Image processing apparatus, image processing program, and image processing method
JP4434705B2 (en) Image analysis method
JP5209213B2 (en) Ultrasonic diagnostic apparatus and ultrasonic image generation program
Macgillivray et al. Fractal analysis of the retinal vascular network in fundus images
EP1994878B9 (en) Medical image processing device and medical image processing method
EP1361820B1 (en) Method for the non-invasive imaging of anatomic tissue structures
US7940975B2 (en) Medical image diagnosing support method and apparatus, and image processing program configured to extract and correct lesion candidate region
JP5113841B2 (en) Computer-aided analysis using video from an endoscope
JP5080944B2 (en) Panorama fundus image synthesis apparatus and method
JP5800468B2 (en) image processing apparatus, image processing method, and image processing program