JP2007037906A - Abnormal shadow candidate detecting system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a system improving a detection accuracy by suppressing the number of false positives when detecting an abnormal shadow candidate of tumor shadow, shortening a processing time when detecting the abnormal shadow candidate, and allowing a doctor to temporally understand the progress state (curing trend) of an affected part (tumor). <P>SOLUTION: An image processing device sets a first smoothing filter and a second smoothing filter to be applied to a newly inputted medical image based on information indicating the dimension of the tumor shadow, being the detection object when previously detecting the abnormal shadow candidate, and the abnormal shadow candidate detecting result stored in a detection result storage part 251 of a storage part 25 associated with patient information of a same patient to the inputted medical image, extracts the detection object region of the abnormal shadow candidate from the inputted medical image using the set first smoothing filer and second smoothing filter, and detects the abnormal shadow candidate of the extracted detection object region in the medical image. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an abnormal shadow candidate detection system that detects an abnormal shadow candidate from a medical image.

  In the medical field, digitalization of medical images is realized, medical image data generated by a CR (Computed Radiography) device or the like is displayed on a monitor, and a doctor interprets the medical image displayed on the monitor, Diagnosis is made by observing the state of the lesion and changes over time.

  Conventionally, for the purpose of reducing the burden on doctors' interpretation, computer diagnostic support apparatus that automatically detects the shadow of a lesion appearing on an image as an abnormal shadow candidate by performing image processing on the medical image data ( An abnormal shadow candidate detection device called Computer-Aided Diagnosis (hereinafter referred to as CAD) has been developed.

  The shadow of a lesion often has a characteristic density distribution, and the CAD detects an image area estimated as a lesion based on such density characteristics as an abnormal shadow candidate area.

  In the CAD, various detection algorithms have been developed according to the type of lesion to be detected, and a method using an iris filter has been proposed as an optimal algorithm for detecting a shadow of a tumor. In addition, a method using a morphological filter has been proposed as an optimal algorithm for detecting a microcalcification cluster shadow.

  By the way, the detection of the abnormal shadow candidate has a problem that it takes a very long time because the entire image is processed for each pixel. In addition, there is a problem that noise or non-lesional areas are detected as abnormal shadows.

Therefore, for example, in Patent Document 1, when a plurality of types of abnormal shadow detection purposes exist in the same imaging region, an algorithm selection unit that selects one or a plurality of abnormal shadow candidate detection algorithms according to the detection purpose is provided. Describes a technique for shortening the processing time by making it possible to select only the algorithms required by. Further, in Patent Document 2, when a doctor has a certain degree of prediction regarding the existence region of an abnormal shadow candidate, or when an abnormal shadow candidate is detected in another image such as an image acquired in the past, the region is used. A technique is described in which an abnormal shadow candidate detection algorithm is applied by designating only.
JP 2002-112986 A JP 2001-346787 A

  However, in Patent Document 1, since the entire image is calculated for each pixel using the selected abnormal shadow candidate detection algorithm, the processing time is shorter than when a plurality of unnecessary abnormal shadow candidate detection algorithms are used. Although it can be shortened, it still takes a long time to search the entire image, and there is a problem that a non-lesional noise or a normal tissue region is detected as an abnormal shadow. Moreover, in patent document 2, even if there was an abnormal shadow in the area | regions other than the area | region which a doctor estimated, or the area | region detected in the past, there existed a problem that it was not detected. In addition, since calculation processing is performed for each pixel in the designated area, there is a possibility that noise or a normal tissue area that is not a lesion is detected as an abnormal shadow even if the search area is limited.

  An object of the present invention is to suppress the number of false positives when detecting an abnormal shadow candidate for a mass shadow and improve the detection accuracy, reduce the processing time when detecting an abnormal shadow candidate, and It is to provide a system that can quickly grasp the progress (healing trend) of a lesion (tumor) of a patient.

In order to solve the above problem, the abnormal shadow candidate detection system of the invention according to claim 1 comprises:
Filter setting means for setting a first smoothing filter and a second smoothing filter corresponding to the size of the mass shadow to be detected from the input medical image;
Extraction using the set first smoothing filter and second smoothing filter to extract a detection target region of an abnormal shadow candidate having the size of the tumor shadow to be detected from the input medical image Means,
An abnormal shadow candidate detecting means for detecting an abnormal shadow candidate for the extracted detection target area in the input medical image;
Storage means for associating and storing patient information of a patient who is a subject of the medical image, information regarding the size of a tumor shadow that is the detection target, and an abnormal shadow candidate detection result of the medical image,
The filter setting means is stored in association with the patient information of the patient in the storage means when a medical image having the same patient as the subject as the subject of the medical image is newly input. The first smoothing filter to be applied to the newly input medical image based on the information about the size of the tumor shadow that was detected at the time of the previous abnormal shadow candidate detection and the abnormal shadow candidate detection result; The second smoothing filter is set.

The invention according to claim 2 is the invention according to claim 1,
The input medical image is composed of pixel values indicating density,
The first smoothing filter has a mask size corresponding to a minimum size of the lesion to be detected;
The second smoothing filter has a mask size corresponding to a maximum size of the lesion to be detected;
The extraction means uses the first smoothing filter to smooth a region having a density lower than the mass shadow to be detected in the input medical image and having a lower density than the surrounding area. A smoothed image is generated, and the second smoothing filter is used to smooth a region having a tumor shadow size to be detected in the first smoothed image and having a lower density than the surrounding area. Generating a second smoothed image, and extracting a detection target region of the abnormal shadow candidate by calculating a difference between pixel values of corresponding pixels of the first smoothed image and the second smoothed image. It is a feature.

The invention according to claim 3 is the invention according to claim 1 or 2,
When the abnormal shadow candidate has not been detected as a result of the previous abnormal shadow candidate detection, the filter setting means includes the first smoothing filter having the same mask size as the previous setting and the same mask size as the previous setting. The second smoothing filter is set.

The invention according to claim 4 is the invention according to claim 1 or 2,
When a tumor shadow is detected as a result of the previous abnormal shadow candidate detection, the filter setting means includes a first smoothing filter having a mask size larger than the previous setting and a second smoothing filter having a mask size larger than the previous setting. A smoothing filter is set.

  According to the first and second aspects of the present invention, the detection target region of the abnormal shadow candidate in the medical image can be limited according to the size of the tumor shadow to be detected. The number of false positives (FP (fault positive) number) at the time of candidate detection can be suppressed, the detection accuracy can be improved, and the processing time at the time of detecting an abnormal shadow candidate for a tumor shadow can be shortened. It becomes possible. In addition, if an abnormal shadow candidate has been previously detected for a tumor shadow with the same patient as a subject, the first smoothing applied to a new medical image according to the previous abnormal shadow candidate detection result. Since the normalization filter and the second smoothing filter are automatically set, it is possible to perform efficient abnormal shadow candidate detection according to the previous abnormal shadow candidate detection result, and the doctor can change the lesion part over time. It is possible to grasp the progress (healing trend) of the (tumor) at an early stage.

  According to the third aspect of the present invention, when the abnormal shadow candidate is not detected as a result of the previous abnormal shadow candidate detection, the first smoothing filter having the same mask size as that set last time and the previous one are set. Since the second smoothing filter with the same mask size as the one is automatically set, the doctor can compare the current abnormal shadow candidate detection result with the previous one so that the pathology of the tumor progresses from the previous detection. It is possible to easily recognize whether or not

  According to the fourth aspect of the present invention, when an abnormal shadow candidate is detected as a result of the previous abnormal shadow candidate detection, the first smoothing filter having a mask size larger than the previous setting and the previous setting are used. Since the second smoothing filter with a large mask size is automatically set, the doctor compares the abnormal shadow candidate detection result of this time with the previous one, so that the pathology of the tumor has progressed since the previous detection. It becomes possible to easily recognize whether or not.

First, the configuration of the present embodiment will be described.
FIG. 1 shows an overall configuration of a medical image system 100 according to the present embodiment. As shown in FIG. 1, a medical image system 100 is a abnormal shadow candidate detection system that generates a medical image and detects an abnormal shadow candidate from the generated medical image. The image generation device 1, the image processing device 2, and the like are connected to a network. N is connected so as to be able to transmit and receive data to and from each other.

  In the present embodiment, an example in which the image generation apparatus 1 and the image processing apparatus 2 are connected via a network will be described. However, the present invention is not limited to this, and a system configuration in which a direct wired connection may be used. Further, the number and installation location of each device are not particularly limited. The medical image system 100 further includes a server that stores and manages the image data of the medical image generated by the image generation apparatus 1, and a monitor that performs display output of the detection result of the abnormal shadow candidate and the processed image in the image processing apparatus 2. Further, a detection result of abnormal shadow candidates in the image processing device 2 and a film output device that outputs a film of the processed image may be connected.

  As the network N, various line forms such as a LAN (Local Area Network), a WAN (Wide Area Network), and the Internet can be applied. Note that wireless communication or infrared communication may be used as long as it is permitted within a medical institution such as a hospital. However, since important patient information is included, it is preferable to encrypt information to be transmitted and received.

  The image generation apparatus 1 is composed of modalities such as CR (Computed Radiography), FPD (Flat Panel Detector), CT (Computed Tomography), MRI (Magnetic Resonance Imaging), and the like. It is an apparatus that generates image data of medical images after conversion. In the present embodiment, description will be made assuming that the image generation apparatus 1 performs radiation imaging of a breast and generates image data of a breast image. Note that the image generation apparatus 1 can input or automatically generate image supplementary information (hereinafter referred to as supplementary information) for a breast image. The image generation apparatus 1 outputs the accompanying information along with the generated image data of the breast image to the image processing apparatus 2 via the network N as header information of the image data.

  The incidental information of the breast image includes, for example, the patient name, patient ID, age, sex, etc. of the patient who has been imaged, patient information about the patient, imaging date, examination ID, imaging location, imaging conditions (distinguish left and right breasts, imaging Shooting information such as image generation device (modality type) information.

  The image processing apparatus 2 is an abnormal shadow candidate detection apparatus that performs an abnormal shadow candidate detection process on the image data supplied from the image generation apparatus 1.

Hereinafter, the internal configuration of the image processing apparatus 2 will be described.
FIG. 2 shows a functional configuration of the image processing apparatus 2. As shown in FIG. 2, the image processing apparatus 2 includes a CPU 21, an operation unit 22, a display unit 23, a RAM 24, a storage unit 25, a communication control unit 26, and the like, and each unit is connected by a bus 27.

  The CPU 21 reads a system program stored in the storage unit 25, develops it in a work area formed in the RAM 24, and controls each unit according to the system program. Further, the CPU 21 reads out various processing programs including the abnormal shadow candidate detection processing program stored in the storage unit 25 and develops them in the work area, and starts the abnormal shadow candidate detection processing (see FIG. 3) described later. Various processes to be executed are executed.

  The operation unit 22 includes a keyboard having cursor keys, numeric input keys, various function keys, and the like, and a pointing device such as a mouse. The operation signal is input to the CPU 21 through key operations on the keyboard and mouse operations. Output. The operation unit 22 may include a touch panel on the display screen of the display unit 23. In this case, the operation unit 22 outputs an instruction signal input via the touch panel to the CPU 21.

  The display unit 23 includes a monitor such as an LCD (Liquid Crystal Display) or a CRT, and displays an image in accordance with an instruction of a display signal input from the CPU 21.

  The RAM 24 forms a work area for temporarily storing various programs, input or output data, parameters, and the like that can be executed by the CPU 21 read from the storage unit 25 in various processes that are executed and controlled by the CPU 21.

The storage unit 25 includes an HDD (Hard Disc Drive), a nonvolatile semiconductor memory, and the like, and includes various programs including a system program executed by the CPU 21, an abnormal shadow candidate detection processing program corresponding to the system program, Store data etc. These various programs are stored in the form of readable program codes, and the CPU 21 sequentially executes operations according to the program codes.
In addition, the storage unit 25 includes a detection result storage unit 251 as a storage unit. The storage unit 25 includes data of a medical image that has been taken and abnormal shadow candidate detection data for the medical image (accompanying information of the medical image (including patient information). ), Information on the size of the tumor shadow as the detection target in the medical image (for example, the range of the size of the tumor shadow as the detection target, the first smoothing filter used for extracting the detection target region, and the second Smoothing filter mask size and the like) and abnormal shadow candidate detection results (detection number and the like) in the medical image are stored in association with each other.

  The communication control unit 26 includes a LAN adapter, a router, a TA (Terminal Adapter), and the like, and controls communication with each device connected to the network N.

Next, the operation in this embodiment will be described.
FIG. 3 is a flowchart showing an abnormal shadow candidate detection process executed by the CPU 21 of the image processing apparatus 2. CPU21 performs the said process by the software process by cooperation with the abnormal shadow candidate detection process program memorize | stored in the memory | storage part 25. FIG.

  Here, in the abnormal shadow candidate detection process of FIG. 3, a tumor shadow is detected from the breast image. The mass shadow is a mass having a certain size, and is displayed as a whitish round shadow close to a Gaussian distribution on the breast image. That is, on the breast image, a drop in density value is seen in the area of the shadow of the tumor. In the following abnormal shadow candidate detection processing, after detecting the detection target area of the abnormal shadow candidate based on the characteristics of the tumor shadow on the breast image, the abnormal shadow candidate is detected for the detection target area. .

  First, image data D of a breast image generated by photographing a breast in the image generation apparatus 1 is input via the communication control unit 26 and stored in the work area of the RAM 24 (step S1).

  Next, the first smoothing filter and the second smoothing filter corresponding to the size of the tumor shadow to be detected are set (step S2; filter setting means). The first smoothing filter and the second smoothing filter, which will be described in detail later, extract a low-density area having the same size as the mass shadow to be detected from the breast image as an abnormal shadow candidate detection target area. It is a filter used when doing. The minimum size of the detection target area to be extracted is determined by the mask size of the first smoothing filter and the sampling pitch of the image data, and is extracted by the mask size of the second smoothing filter and the sampling pitch of the image data. The maximum size of the detection target area is determined. Therefore, in step S2, the first smoothing filter whose mask size corresponds to the minimum size (lower limit) of the tumor shadow to be detected is set, and the second smoothing filter has its mask size. Is set corresponding to the maximum size (upper limit) of the tumor shadow to be detected.

  The mass shadow generally becomes larger as the disease progresses. In the image processing apparatus 2 of the present embodiment, for example, the size of the mass shadow of the first (initial) stage of the disease state is about 5 mm to 15 mm, and the size of the tumor shadow of the second stage of the disease state is about 15 mm to 30 mm. As described above, the size of the mass shadow is defined in a plurality of stages according to the progress stage of the disease state, and the detection target area corresponding to the size of the mass shadow at each stage can be extracted from the input image data D. Therefore, a plurality of combinations of the first smoothing filter and the second smoothing filter having different mask sizes are prepared. In the first medical examination patient, first, it is necessary to set the first smoothing filter and the second smoothing filter having a mask size corresponding to the size of the mass shadow in the initial stage. In addition, for patients who have previously detected a mass shadow, the mass to be detected last time is available so that the doctor can grasp the progress (healing trend) of the lesion (tumor) over time. It is preferable to limit the size of the detection target according to the size of the shadow and the abnormal shadow candidate detection result. For example, if no tumor shadow is detected at the time of the previous detection, a tumor shadow having the same size as that of the previous detection target is set as the detection target, and the first smoothing filter and the second smoothing filter having the same mask size as the previous detection target are detected. By setting the conversion filter, the doctor can easily recognize whether or not the disease state of the tumor has progressed since the previous detection. If a tumor shadow is detected at the time of the previous detection, a tumor shadow that is one step larger than the previous one is detected, and by setting the first smoothing filter and the second smoothing filter with a mask size larger than the previous time, It becomes possible for the doctor to easily recognize whether the disease state of the tumor has progressed since the previous detection.

  Therefore, in step S2, the detection result storage unit 251 of the storage unit 25 is searched, and there is no abnormal shadow candidate detection data including the same patient information as the patient information included in the incidental information of the input image data D. In other words, in the case of the first visit, the first and second smoothing filters having mask sizes corresponding to the size of the tumor shadow (about 5 mm to 15 mm) at the initial stage are automatically set. As a result of the search, when there is abnormal shadow candidate detection data including the same patient information as the patient information included in the incidental information of the input image data D, the abnormal shadow candidate detection result is referred to, and the detection result is 0. In some cases, the information about the size of the tumor shadow that is the detection target at the previous detection is referred to, and the first smoothing filter and the second smoothing filter having the same mask size used at the previous detection are automatically set. Is done. On the other hand, when the abnormal shadow candidate detection result included in the detected abnormal shadow candidate detection data exceeds 0, that is, when a tumor shadow is detected in the previous detection of the abnormal shadow candidate, it is detected at the previous detection. Information on the size of the target tumor shadow is referred to, and the first smoothing filter and the second smoothing filter having a mask size larger than that used at the previous detection are automatically set.

  Next, region extraction processing is performed on the image data D using the first smoothing filter and the second smoothing filter set in step S2 (step S3; extraction means). The area extraction process is a process of extracting a detection target area corresponding to the size of the tumor shadow to be detected from the entire area of the image data D.

Hereinafter, the region extraction processing will be described in detail with reference to FIGS.
FIG. 4 is a flowchart showing the region extraction process executed by the CPU 21 in step S3.
FIG. 5 is a diagram schematically illustrating an image before the region extraction process of FIG. 4 and a processing result in each step of the region extraction process of the image. 5A to 5D, the horizontal axis represents the pixel position in a certain column of image data (image data D1 to D4) (FIGS. 5A to 5D all indicate the same column). The pixel value (density value) is shown as the vertical axis.

  In the region extraction processing of FIG. 4, first, reduction processing is performed on the image data D, and image data D1 having a sampling pitch of about 1.6 mm is generated (step S11). For example, if the sampling pitch of the image data D is 100 μm, the vertical and horizontal directions are each 1/16. The reduction processing algorithm may be any method, such as taking the average of neighboring pixel values or thinning out at a fixed pixel interval. Here, by reducing the image data D, the processing time in the subsequent processing can be shortened.

  In the breast image, as shown in FIG. 5 (a), it has the same size as the mass shadow to be detected, and has a lower density area (detection target area) than the surrounding area, smaller than the mass shadow to be detected, In addition, a region having a concentration lower than the surroundings (a minute region), a region larger than the shadow of the tumor to be detected, and a region having a concentration lower than the surroundings (a region larger than the target) are included. In the processing steps subsequent to the region extraction processing, a low-concentration region having the same size as the tumor shadow to be detected is extracted from the reduced image data D1 as a detection target region.

  Next, the first smoothing process is performed on the image data D1 using the first smoothing filter set in step S2 of FIG. 3 to generate image data D2 (first smoothed image). (Step S12).

  The first smoothing filter is a median filter having the pixel of interest in the image data D1 as the center, the pixel values in the square area (mask) arranged in descending order, and the median as the pixel value of the pixel of interest. FIG. 6 shows an example of a first smoothing filter having a mask size of 3 (pixels) × 3 (pixels). For example, when the first smoothing process is performed using a median filter having a mask size of 3 × 3, the pixel values of the regions 1 to 9 shown in FIG. 6 are arranged in descending order, and the median value is set as the pixel value of the region 5. . This is repeated for all the pixels of the image data D1 by shifting the position of the mask pixel by pixel. As a result, as shown in FIG. 5 (b), the vertical and horizontal widths are each smoothed about the mask width, that is, a minute region having a size up to about 4.8 mm (1.6 mm (sampling pitch) × 3). Is done.

  Next, the second smoothing process is performed on the image data D2 using the second smoothing filter set in step S2 of FIG. 3 to generate image data D3 (second smoothed image). (Step S13).

  The second smoothing filter includes a maximum value filter having the maximum value among the pixel values in the mask as the value of the central target pixel, and a minimum value among the pixel values in the mask as the value of the central target pixel. And applying a minimum value filter to the image data D2, and then applying a minimum value filter to fill in a drop in a pixel value having a size of about a mask. is there. In general, the mass shadow has a feature that the X-ray transmission density decreases toward the center thereof, and has the same size as the tumor shadow to be detected for the image data D2 by the second smoothing process. By applying a second smoothing filter having a mask size corresponding to the above, it is possible to fill a low-concentration region of the same extent as the tumor shadow to be detected.

Here, with reference to FIG. 7, the principle of the second smoothing filter will be described by taking a one-dimensional data string as an example. In FIG. 7, the horizontal axis indicates the pixel position in the one-dimensional data string, and the vertical axis indicates the pixel value (density value) in the one-dimensional data string.
L1 in FIG. 7 is a data string of the original image. For example, a maximum value filter of mask size 1 × vertical 7 is set with the target pixel as the center in order from the pixel located on the left of the data row of the original image, and the maximum value within the mask range is focused. The pixel value of the pixel. By shifting this to the right by one pixel, a data string indicated by L2 in FIG. 7 can be obtained. The data string indicated by L2 is input, the pixel of interest is set in order from the pixel located on the left, a minimum value filter of mask size 1 × vertical 7 is set with the pixel of interest at the center, and the minimum value within the mask range is noted. The pixel value of the pixel. As a result, as shown by L3 in FIG. 7, a data string in which the drop of the density value of the original data string L1 up to about 11.2 mm (sampling pitch 1.6 mm × 7) is smoothed is obtained. be able to.

  When the second smoothing process ends, a difference image generation process is executed, and a difference between pixel values at the same pixel position in the image data D3 shown in FIG. 5C and the image data D2 shown in FIG. 5B is obtained. As a result, a difference image (image data D4) shown in FIG. 5D is generated (step S14). Then, threshold processing is performed on the image data D4 with a preset threshold, and only data having a pixel value exceeding the threshold is extracted (step S15), and a low density having the same size as the abnormal shadow candidate region to be detected. Image data D5 of the area is generated.

  In addition, as shown in FIG. 8, it classify | categorizes according to magnitude | size by changing the mask size of a 1st smoothing filter, and the mask size of a 2nd smoothing filter, and performing the above-mentioned area | region extraction process in multiple times. It is possible to extract the detection target region. For example, first, image data D2 is generated using a first smoothing filter having a mask size of 3 × 3, and the image data D3 is subjected to a second smoothing filter having a mask size of 7 × 7. , And a threshold process is performed by taking the difference between the image data D3 and the image data D2, thereby extracting a detection target region of about 5 mm to 15 mm. Next, a first smoothing filter having a mask size of 7 × 7 is further applied to the image data D2 to generate image data D2 ′, and a second smoothing having a mask size of 11 × 11 is performed on the image data D2 ′. The detection target region of about 15 mm to 30 mm can be extracted by generating the image data D3 ′ by performing the filtering and taking the difference between the image data D3 ′ and the image data D2 ′ and performing threshold processing. In this way, it is possible to extract a detection target region of about 5 mm to 15 mm and a detection target region of about 15 mm to 30 mm, respectively.

  Since the tumor shadow is substantially circular, the same processing may be performed in both the vertical and horizontal directions of the image data, and the extracted region and the tumor shadow are easily associated with each other. Therefore, it is particularly preferable to perform the region extraction processing.

  In FIG. 3, when the region extraction process is completed, an abnormal shadow candidate is detected for the detection target region of the extracted image data D5 in the image data D (step S4; abnormal shadow candidate detection means).

  Various algorithms for detecting abnormal shadow candidates have been developed according to the type of lesion to be detected, and methods that use iris filters and curvature methods are the best algorithms for detecting tumor shadows. Proposed. In the present embodiment, a case will be described in which a technique for detecting a tumor shadow candidate using an iris filter is applied as an abnormal shadow candidate detection algorithm.

  It is known that a mass shadow in a radiographic image of a breast generally has a lower density value than the surrounding image part, and the density value of the distribution of density values decreases from the periphery of the circular shape toward the center. It has a gradient of concentration values. Therefore, a local density value gradient is recognized in the mass shadow, and the gradient line is concentrated in the central direction of the mass. The iris filter calculates the gradient of the image signal typified by this density value as a gradient vector, and outputs the degree of concentration of the gradient vector. Based on this concentration and other feature quantities, candidates for tumor shadows A region is detected.

  First, an arbitrary pixel of interest is set in a detection target region of an abnormal shadow candidate in the image data D (image region extracted as the image data D5). Next, an image feature amount is calculated at the set target pixel. First, a feature quantity such as a concentration gradient direction component and intensity component concentration around a pixel of interest (for example, a region within a predetermined mask size) is obtained. As other feature amounts, feature amounts such as contrast, standard deviation, and fractal dimension around the target pixel are obtained. When various feature amounts are calculated, they are compared with threshold values for detecting abnormal shadow candidates that are set in advance for each feature amount. Based on the comparison result, the region around the target pixel may be an abnormal shadow Whether or not is high is determined. When it is determined that there is a high possibility of an abnormal shadow, the region around the target pixel is detected as a candidate region for the abnormal shadow. A pixel of interest is set in all detection target areas, and detection of abnormal shadow candidates is repeated.

  Thus, in detecting an abnormal shadow candidate, each pixel in the detection target area is searched for as a target pixel, and it is determined whether or not the surrounding area is likely to be an abnormal shadow.

  After detection of the abnormal shadow candidate, the process proceeds to step S5 in FIG. 3, and the abnormal shadow candidate detection result is displayed on the display unit 23 (step S5). For example, a breast image based on the image data D is displayed on the display unit 23, and a candidate area detected as an abnormal shadow candidate (tumor shadow candidate) is indicated by an arrow or displayed in color on the breast image. And so on. Further, the feature amount in the abnormal shadow candidate may be output. Then, the data of the medical image that has been taken and the abnormal shadow candidate detection data (the accompanying information of the medical image, the information about the size of the tumor shadow that is the detection target in the medical image, the detection result of the abnormal shadow candidate in the medical image Are associated with each other and stored in the detection result storage unit 251 of the storage unit 25 (step S6), and the process ends.

  As described above, according to the image processing apparatus 2, abnormal shadow candidate detection data having the same patient information as the patient information included in the incidental information of the medical image input to the detection result storage unit 251 of the storage unit 25. If the abnormal shadow candidate detection data having the same patient information as the patient information included in the incidental information of the input medical image is not stored, the size of the tumor shadow at the initial stage is searched. First and second smoothing filters having a mask size corresponding to the height (about 5 mm to 15 mm) are automatically set. As a result of the search, when there is abnormal shadow candidate detection data including the same patient information as the patient information included in the incidental information of the input medical image, the abnormal shadow candidate detection result is referred to and the detection result is 0. In this case, the first smoothing filter and the second smoothing filter having the same mask size as those used in the previous detection are automatically set with reference to the information on the size of the tumor shadow to be detected in the previous detection. . On the other hand, when the abnormal shadow candidate detection result included in the detected abnormal shadow candidate detection data exceeds 0, that is, when a tumor shadow is detected in the previous detection of the abnormal shadow candidate, it is detected at the previous detection. The first smoothing filter and the second smoothing filter having a mask size larger than that used at the previous detection are automatically set with reference to information on the size of the target tumor shadow. Then, using the set first smoothing filter and second smoothing filter, the detection target region of the abnormal shadow candidate having the size of the tumor shadow to be detected is extracted from the input medical image and extracted. An abnormal shadow candidate is detected for the detected region.

  Therefore, the search area for detecting abnormal shadow candidates in the breast image generated by the image generating apparatus 1 is limited according to the size of the tumor shadow to be detected, so that it is not a size corresponding to the abnormal shadow candidate to be detected. Low density regions, for example, small regions such as noise and larger normal tissue regions can be excluded from the search target in advance, and detection of abnormal shadow candidates targeting the tumor shadow for the entire breast image as in the past Compared with the case where processing is performed, the number of false positives can be suppressed, and the processing time can be significantly shortened. In addition, if an abnormal shadow candidate has been previously detected for a tumor shadow with the same patient as a subject, a search area for detecting an abnormal shadow candidate in a breast image according to the previous detection result of the abnormal shadow candidate. Since it is possible to perform efficient detection according to the detection result of the previous abnormal shadow candidate, the doctor can check the progress of the lesion (tumor) over time (healing trend) It becomes possible to grasp at an early stage.

  In addition, the description content in the said embodiment is a suitable example of the medical image system 100 which concerns on this invention, and is not limited to this.

  For example, the medical image data and the abnormal shadow candidate detection data have been described as being stored in the image processing apparatus 2, but may be stored in a server connected to the network N.

  In addition, the detailed configuration and detailed operation of each apparatus constituting the medical image system 100 can be changed as appropriate without departing from the spirit of the present invention.

1 is a diagram illustrating an overall configuration of a medical image system 100 according to the present invention. It is a block diagram which shows the functional structure of the image processing apparatus 2 of FIG. It is a flowchart which shows the abnormal shadow candidate detection process performed by CPU21 of FIG. It is a flowchart which shows the area | region extraction process of FIG. It is a figure which shows typically the image data produced | generated at each step of the area | region extraction process of FIG. It is a figure for demonstrating a 1st smoothing filter. It is a figure for demonstrating a 2nd smoothing filter. The figure which shows the process sequence in the case of classifying according to magnitude | size and extracting a detection object area | region by changing the mask size of a 1st smoothing filter, and the mask size of a 2nd smoothing filter, and performing a process in multiple times. It is.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Breast image generation system 1 Image generation device 2 Image processing device 21 CPU
22 Operation part 23 Display part 24 RAM
25 Storage Unit 26 Communication Control Unit 27 Bus

Claims (4)

Filter setting means for setting a first smoothing filter and a second smoothing filter corresponding to the size of the mass shadow to be detected from the input medical image;
Extraction using the set first smoothing filter and second smoothing filter to extract a detection target region of an abnormal shadow candidate having the size of the tumor shadow to be detected from the input medical image Means,
An abnormal shadow candidate detecting means for detecting an abnormal shadow candidate for the extracted detection target area in the input medical image;
Storage means for associating and storing patient information of a patient who is a subject of the medical image, information regarding the size of a tumor shadow that is the detection target, and an abnormal shadow candidate detection result of the medical image,
The filter setting means is stored in association with the patient information of the patient in the storage means when a medical image having the same patient as the subject as the subject of the medical image is newly input. The first smoothing filter to be applied to the newly input medical image based on the information about the size of the tumor shadow that was detected at the time of the previous abnormal shadow candidate detection and the abnormal shadow candidate detection result; An abnormal shadow candidate detection system, wherein the second smoothing filter is set. The input medical image is composed of pixel values indicating density,
The first smoothing filter has a mask size corresponding to a minimum size of the lesion to be detected;
The second smoothing filter has a mask size corresponding to a maximum size of the lesion to be detected;
The extraction means uses the first smoothing filter to smooth a region having a density lower than the mass shadow to be detected in the input medical image and having a lower density than the surrounding area. A smoothed image is generated, and the second smoothing filter is used to smooth a region having a tumor shadow size to be detected in the first smoothed image and having a lower density than the surrounding area. Generating a second smoothed image, and extracting a detection target region of the abnormal shadow candidate by calculating a difference between pixel values of corresponding pixels of the first smoothed image and the second smoothed image. The abnormal shadow candidate detection system according to claim 1, wherein:   When the abnormal shadow candidate has not been detected as a result of the previous abnormal shadow candidate detection, the filter setting means includes the first smoothing filter having the same mask size as the previous setting and the same mask size as the previous setting. The abnormal shadow candidate detection system according to claim 1, wherein the second smoothing filter is set.   When an abnormal shadow candidate is detected as a result of the previous abnormal shadow candidate detection, the filter setting means includes a first smoothing filter having a mask size larger than the previous setting and a mask having a mask size larger than the previous setting. The abnormal shadow candidate detection system according to claim 1, wherein two smoothing filters are set.