JP2019213785A - Medical image processor, method and program - Google Patents

Abstract

To provide a medical image processor, a method and a program that are capable of accurately extracting a region with possible subarachnoid hemorrhage using a medical image such as a CT image of a head.SOLUTION: An image acquisition unit 21 acquires a brain image B0 of a subject. A candidate region extraction unit 22 extracts a bleeding onset candidate region in the brain image B0. A region identification unit 23 identifies a first region 51 inside of a boundary 44 of the bleeding onset candidate region 42, a second region 52 other than the first region 51 in the bleeding onset candidate region 42, and a third region 53 outside the boundary 44 of the bleeding onset candidate region 42. A determination unit 24 compares a signal value Q1 for the first region 51, a signal value Q2 for the second region 52, and a signal value Q3 for the third region 53 to determine whether or not the bleeding onset candidate region 42 is a bleeding region.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a medical image processing apparatus, method, and program for determining a bleeding region in which a subarachnoid hemorrhage is suspected in a brain image such as a CT (Computed Tomography) image of a head.

  2. Description of the Related Art In recent years, advances in medical equipment such as a CT apparatus and an MRI (Magnetic Resonance Imaging) apparatus have enabled image diagnosis using higher-quality, high-resolution medical images. In particular, in the case where the target region is the brain, a region where vascular disorders such as cerebral infarction and cerebral hemorrhage can be identified by image diagnosis using CT images and MRI images, etc. And appropriate treatment is being provided.

  By the way, subarachnoid hemorrhage is a disease with a high rebleeding rate, and even if it is mild bleeding, the rebleeding rate is said to be about 50% within one month if appropriate treatment is not performed . Furthermore, since the probability of seriousness is higher at the time of rebleeding, it is important to prevent oversight of the onset and prevent rebleeding. Here, for patients suspected of having subarachnoid hemorrhage, image diagnosis using a CT image of the head is often the first choice. Therefore, it is very important to accurately diagnose a disease state using a CT image in order to realize early treatment.

  Generally, a bleeding area shows a higher CT value on a CT image than a surrounding area. In addition, subarachnoid hemorrhage occurs in places where cerebrospinal fluid is perfused, such as the cisternal and ventricles. Here, the region where the cerebrospinal fluid exists has a low CT value in the CT image. For this reason, when a region with a high CT value exists in a region of the cerebral cistern and the ventricle that originally has a low CT value, it can be diagnosed that subarachnoid hemorrhage has developed.

  On the other hand, one of the cases that is easily overlooked in subarachnoid hemorrhage is "a case in which blood has been washed out several days after the onset of subarachnoid hemorrhage". Washout is the washing of blood mixed into the cerebrospinal fluid by the cerebrospinal fluid circulation over time. Subarachnoid hemorrhage is a hemorrhage that occurs in a place where the cerebrospinal fluid is perfused, such as the cistern and the ventricle, and therefore has a high probability of occurrence of a washout.

  In the case of a washout patient, the development of subarachnoid hemorrhage can be confirmed by finding blood remaining in the cistern and ventricle on the CT image. If only a small portion remains in the peripheral part of the ventricle or the like, the probability of overlooking subarachnoid hemorrhage is higher than in the early stage of onset when most of the cerebral cistern has a high CT value.

  For this reason, in the CT image, it is determined whether or not the region is a cerebral hemorrhage site based on the CT value and the size of the region that may include the cerebral hemorrhage site. Of the CT images, the difference between the CT value of the bleeding-onset candidate area and the surrounding area or the CT value of the bleeding-onset candidate area for the area that may be a cerebral bleeding site in the CT image (hereinafter, referred to as the bleeding-onset candidate area) A method of determining whether a bleeding onset candidate region is a cerebral bleeding site based on the bleeding occurrence candidate region has been proposed (see Patent Document 1). Since the washout occurs in the cerebral cistern and the ventricle, it is possible to determine whether or not the washout has occurred by using the method described in Patent Document 1.

JP-T-2009-539510

  However, when a washout occurs in a bleeding area that has developed subarachnoid hemorrhage, most of the bleeding area has returned to a normal structure. Therefore, in a CT image, it is difficult to distinguish a region where a washout has occurred from a normal region. The method described in Patent Document 1 determines whether or not a bleeding area is based on signal values of a bleeding occurrence candidate area and a surrounding area. For this reason, when most of the bleeding area returns to a normal structure due to the washout, the technique described in Patent Document 1 cannot accurately determine whether or not the area is a bleeding area.

  The present invention has been made in view of the above circumstances, and has as its object to accurately determine whether or not a bleeding-onset candidate area is a bleeding area.

A medical image processing apparatus according to the present invention includes: an image acquisition unit configured to acquire a three-dimensional brain image including a brain of a subject;
A candidate region extraction unit for extracting a bleeding onset candidate region in the brain image,
An area specifying unit that specifies a first area inside the boundary of the bleeding onset candidate area, a second area other than the first area in the bleeding onset candidate area, and a third area outside the boundary of the bleeding onset candidate area When,
A determination unit that compares the signal value of the first region, the signal value of the second region, and the signal value of the third region to determine whether the bleeding-onset candidate region is a bleeding region;

  In the medical image processing device according to the present invention, the determination unit includes, in the second area, an area having a signal value higher than the signal value of the first area and higher than the signal value of the third area. In this case, the bleeding-onset candidate area may be determined as a bleeding area.

  In the medical image processing apparatus according to the present invention, the candidate region extracting unit may extract a bleeding-onset candidate region by aligning a standard brain image, which is a standard brain image, with the brain image. .

  The standard brain image is an image representing an average brain structure. The brain image is obtained from a plurality of medical images obtained by photographing the heads of a plurality of healthy persons with an imaging device such as a CT device and an MRI device. It is created by extracting regions and averaging a plurality of extracted brain regions. Further, the standard brain image may be an image created by computer graphics or the like. Further, a brain image of one healthy person may be used as a standard brain image.

  Further, in the medical image processing apparatus according to the present invention, the bleeding-onset candidate region may be at least one of a cistern and a ventricle in the brain.

  In the medical image processing device according to the present invention, the brain image may be a CT image acquired by a CT device.

  Further, the medical image processing apparatus according to the present invention may further include a display control unit that displays bleeding area information indicating that the bleeding area has been specified on the display unit.

The medical image processing method according to the present invention acquires a three-dimensional brain image including a brain of a subject,
Extract the bleeding onset candidate region in the brain image,
A first area inside the boundary of the bleeding onset candidate area, a second area other than the first area in the bleeding onset candidate area, and a third area outside the boundary of the bleeding onset candidate area,
The signal value of the first area, the signal value of the second area, and the signal value of the third area are compared to determine whether the bleeding-onset candidate area is a bleeding area.

  The medical image processing method according to the present invention may be provided as a program for causing a computer to execute the method.

Another medical image processing apparatus according to the present invention includes: a memory that stores instructions to be executed by a computer;
A processor configured to execute the stored instructions, the processor comprising:
Acquire a three-dimensional brain image including the subject's brain,
Extract the bleeding onset candidate region in the brain image,
A first area inside the boundary of the bleeding onset candidate area, a second area other than the first area in the bleeding onset candidate area, and a third area outside the boundary of the bleeding onset candidate area,
A process of comparing the signal value of the first area, the signal value of the second area, and the signal value of the third area to determine whether the bleeding-onset candidate area is a bleeding area is executed.

  According to the present invention, a three-dimensional brain image including the brain of the subject is obtained, and a candidate region for bleeding occurrence in the brain image is extracted. Then, the first region inside the boundary of the bleeding onset candidate region, the second region other than the first region in the bleeding onset candidate region, and the third region outside the boundary of the bleeding onset candidate region are specified. . Here, when a washout occurs in the bleeding area, most of the bleeding area has returned to a normal structure. However, bleeding has occurred when the signal value of the region inside the boundary of the bleeding region, the signal value of the region outside the inner region within the bleeding region, and the signal value of the region outside the bleeding region are compared. There is a difference in the signal value of each area between the case where there is no bleeding and the case where bleeding occurs and washout occurs. Therefore, by comparing the signal value of the first area, the signal value of the second area, and the signal value of the third area, it is determined whether the bleeding-onset candidate area is a bleeding area. Can be. Therefore, according to the present invention, even if a washout has occurred, it is possible to accurately determine whether or not the bleeding-onset candidate area is a bleeding area.

Hardware configuration diagram showing an outline of a diagnosis support system to which a medical image processing device according to an embodiment of the present invention is applied FIG. 1 is a diagram illustrating a schematic configuration of a medical image processing apparatus according to an embodiment of the present invention. Diagram showing a standard brain image Diagram showing brain image of subject Diagram showing candidate bleeding episodes extracted from brain images FIG. 9 is a diagram for explaining the specification of the first to third regions. Diagram showing structural elements of erosion processing and dilation processing Diagram showing brain image displayed on display Flow chart showing processing performed in the present embodiment Diagram showing another example of a bleeding onset candidate area FIG. 9 is a diagram for explaining the specification of the first to third regions.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a hardware configuration diagram showing an outline of a diagnosis support system to which a medical image processing apparatus according to an embodiment of the present invention is applied. As shown in FIG. 1, in the diagnosis support system, a medical image processing apparatus 1, a three-dimensional image photographing apparatus 2, and an image storage server 3 according to the present embodiment are communicably connected via a network 4. I have.

  The three-dimensional image capturing device 2 is a device that captures a part to be diagnosed of a subject to generate a three-dimensional image representing the part, and specifically, a CT device, an MRI device, and a PET ( Positron Emission Tomography) device. The three-dimensional image generated by the three-dimensional image capturing device 2 is transmitted to the image storage server 3 and stored. In the present embodiment, the part to be diagnosed of the patient who is the subject is the brain, the three-dimensional image capturing device 2 is a CT device, and the CT image of the head including the subject's brain is converted into a three-dimensional brain. Generated as image B0.

  The image storage server 3 is a computer that stores and manages various data, and includes a large-capacity external storage device and database management software. The image storage server 3 communicates with other devices via a wired or wireless network 4 to transmit and receive image data and the like. Specifically, various data including the image data of the brain image B0 generated by the three-dimensional image capturing device 2 is acquired via a network, and stored and managed in a recording medium such as a large-capacity external storage device. The storage format of the image data and communication between the devices via the network 4 are based on a protocol such as DICOM (Digital Imaging and Communication in Medicine).

  The medical image processing apparatus 1 has the medical image processing program of the present invention installed in one computer. The computer may be a workstation or personal computer directly operated by a physician performing the diagnosis, or a server computer connected to them via a network. The medical image processing program is recorded and distributed on a recording medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory), and is installed in a computer from the recording medium. Alternatively, the information is stored in a storage device of a server computer connected to a network or a network storage in a state where it can be accessed from the outside, and is downloaded and installed on a computer used by a doctor in response to a request.

  FIG. 2 is a diagram illustrating a schematic configuration of a medical image processing apparatus realized by installing a medical image processing program in a computer. As shown in FIG. 2, the medical image processing apparatus 1 includes a CPU (Central Processing Unit) 11, a memory 12, and a storage 13 as a standard workstation configuration. Further, a display 14 and an input unit 15 such as a keyboard and a mouse are connected to the medical image processing apparatus 1. The display 14 corresponds to a display unit.

  The storage 13 is composed of a hard disk drive or the like, and stores brain images of the subject and various types of information including information necessary for processing obtained from the image storage server 3 via the network 4.

  The memory 12 stores a medical image processing program. The medical image processing program includes, as processes to be executed by the CPU 11, an image acquiring process for acquiring a brain image B0 of a subject, a candidate region extracting process for extracting a bleeding onset candidate region in the brain image B0, an inner side of a boundary of the bleeding onset candidate region. Area specifying processing for specifying the first area, the second area other than the first area in the bleeding onset candidate area, and the third area outside the boundary of the bleeding onset candidate area, the signal value of the first area A determination process of comparing the signal value of the second region and the signal value of the third region to determine whether the bleeding-onset candidate region is a bleeding region, and bleeding indicating that the bleeding region has been identified. A display control process for displaying the area information on the display 14 is defined.

  When the CPU 11 executes these processes according to the program, the computer functions as the image acquisition unit 21, the candidate region extraction unit 22, the region identification unit 23, the determination unit 24, and the display control unit 25. In the present embodiment, the CPU 11 executes the function of each unit by the medical image processing program. However, in addition to the CPU 11, a general-purpose processor that executes software and functions as various processing units is described below. A programmable logic device (PLD), which is a processor whose circuit configuration can be changed after manufacturing an FPGA (Field Programmable Gate Array) or the like, can be used. Further, the processing of each unit may be executed by a dedicated electric circuit or the like, which is a processor having a circuit configuration specifically designed to execute a specific processing such as an ASIC (Application Specific Integrated Circuit).

  One processing unit may be configured by one of these various processors, or may be a combination of two or more processors of the same type or different types (for example, a plurality of FPGAs or a combination of a CPU and an FPGA). It may be configured. Further, the plurality of processing units may be configured by one processor. As an example in which a plurality of processing units are configured by one processor, first, as represented by computers such as clients and servers, one processor is configured by a combination of one or more CPUs and software, There is a form in which this processor functions as a plurality of processing units. Second, as represented by a system-on-chip (System On Chip: SoC), a form using a processor that realizes the functions of the entire system including a plurality of processing units with one integrated circuit (IC) chip is used. is there. As described above, various processing units are configured using one or more of the various processors as a hardware structure.

  Furthermore, the hardware structure of these various processors is more specifically an electric circuit (circuitry) in which circuit elements such as semiconductor elements are combined.

  The image acquisition unit 21 acquires a brain image B0 of the subject's brain from the image storage server 3. When the brain image B0 is already stored in the storage 13, the image acquisition unit 21 may acquire the brain image B0 from the storage 13.

  The candidate area extraction unit 22 extracts a bleeding onset candidate area in the brain image B0. Here, the bleeding onset candidate region includes a cerebral cistern and a ventricle. In the present embodiment, the cerebral ventricle is set as the bleeding onset candidate region. For this purpose, the candidate region extraction unit 22 extracts a ventricular region in the brain image B0 as a bleeding onset candidate region. Specifically, the candidate area extracting unit 22 extracts a bleeding onset candidate area by aligning the standard brain image and the brain image B0. The standard brain image is a brain generated by extracting brain regions from a plurality of CT images obtained from a plurality of healthy persons by a CT apparatus, generating a plurality of brain images, and averaging the plurality of brain images. It is an image. In the present embodiment, the brain region includes the brain parenchyma and the skull, but may include only the brain parenchyma. FIG. 3 is a diagram showing a standard brain image. Although the standard brain image Bs is a three-dimensional image representing the brain, FIG. 3 shows a tomographic image of the brain in a certain axial section for explanation. As shown in FIG. 3, the standard brain image Bs includes a skull 30 and a brain parenchyma 31, and the brain parenchyma 31 includes a ventricle region 32. It should be noted that although the brain parenchyma 31 actually includes other anatomical regions such as the cerebral cistern, FIG. 3 shows only the ventricular region 32. The image data of the standard brain image Bs is stored in the image storage server 3 or the storage 13, and the candidate area extracting unit 22 acquires the image data from the image storage server 3 or the storage 13.

  Note that the standard brain image Bs may be one created by computer graphics or the like. Further, a brain image of one healthy person may be used as the standard brain image Bs.

  As shown in FIG. 3, in the standard brain image Bs, the skull 30 has a very high CT value and is shown in white with high brightness, and the brain parenchyma 31 is shown in gray. Further, the ventricular region 32 is filled with cerebrospinal fluid and has a very low CT value, so that it is shown in black and with low brightness.

  Here, the shape and size of the head differ depending on the subject. For example, as shown in FIG. 4, the shape and size of the brain included in the brain image B0 of the subject are different from the shape and size of the brain included in the standard brain image Bs shown in FIG. In FIG. 4, only the skull 40 and the brain parenchyma 41 are shown, and the ventricles are omitted. Although the brain image B0 is a three-dimensional image representing the brain, FIG. 4 shows a tomographic image of the brain in a certain axial cross section for explanation. In the following description, the brain image B0 is shown as a tomographic image.

  The candidate region extraction unit 22 performs positioning between the brain image B0 and the standard brain image Bs in order to extract a bleeding occurrence candidate region from the brain image B0. Positioning is performed between the three-dimensional brain image B0 and the three-dimensional standard brain image Bs. As a positioning method, a first positioning using landmarks between the standard brain image Bs and the brain image B0 is first performed. Then, after performing the first positioning, a second positioning using the entire region between the standard brain image Bs and the brain image B0 is performed. In addition, specifically, at least one of characteristic regions such as a cerebral sulcus and a ventricle included in the brain can be used as the landmark.

  The candidate region extraction unit 22 performs the first alignment between the standard brain image Bs and the brain image B0 so that the corresponding landmarks match. In the present embodiment, the first alignment is alignment by similarity conversion. Specifically, alignment is performed by translating, rotating, and scaling the brain image B0 similarly. The candidate region extraction unit 22 converts the brain image B0 such that the correlation between the landmark included in the standard brain image Bs and the landmark corresponding to the landmark of the standard brain image Bs included in the brain image B0 is maximized. By performing similarity conversion, first alignment is performed.

  After performing the first alignment using the landmark in this way, the candidate area extraction unit 22 performs the second alignment using the entire area between the standard brain image Bs and the brain image B0. In the present embodiment, the second alignment is an alignment by non-linear conversion. The alignment by non-linear conversion includes, for example, alignment by non-linearly converting pixel positions using a function such as a B-spline and a thin plate spline. The candidate region extracting unit 22 performs the second alignment by performing a non-linear conversion of each pixel position of the brain image B0 after the first alignment into a corresponding pixel position included in the standard brain image Bs.

  By performing the positioning in this manner, the ventricle region in the brain image B0 can be associated with the ventricle region 32 in the standard brain image Bs. Then, the candidate region extracting unit 22 extracts a region associated with the ventricle region 32 of the standard brain image Bs in the brain image B0 as the bleeding onset candidate region 42 of the brain image B0. FIG. 5 is a diagram showing a bleeding onset candidate region extracted from the brain image B0. As shown in FIG. 5, in the brain image B0, a bleeding onset candidate region 42 is extracted. In FIG. 5, the bleeding-onset candidate area 42 is shown in black. Further, in FIG. 5, a region 43 having a high CT value is included in the bleeding occurrence candidate region 42.

  The region specifying unit 23 includes a first region inside the boundary of the bleeding onset candidate region 42, a second region other than the first region in the bleeding onset candidate region 42, and a second region outside the boundary of the bleeding onset candidate region 42. Region 3 is specified. FIG. 6 is a diagram for explaining the specification of the area. In FIG. 6, only the bleeding-onset candidate area 42 on the right side in FIG. 5 is shown two-dimensionally. First, the region specifying unit 23 specifies the boundary 44 of the bleeding onset candidate region 42 as shown in FIG. Then, the erosion process is performed on the boundary 44 of the bleeding onset candidate area 42 to reduce the bleeding onset candidate area 42 to the inside. The erosion process is a process of searching for a minimum value within a predetermined width centered on the boundary 44 of the bleeding occurrence candidate area 42 using a structural element as shown in FIG. Although the structural elements are shown two-dimensionally in FIG. 7, since the brain image B0 is a three-dimensional image, a three-dimensional structural element is actually used. By performing the erosion process using the structural element shown in FIG. 7 once, the bleeding occurrence candidate area 42 is reduced by one pixel inward. Then, the region identification unit 23 generates the first region 51 inside the boundary 44 of the bleeding occurrence candidate region 42 by performing the erosion process a predetermined number of times. The first area 51 is an area surrounded by the reduced boundary 45 in the bleeding onset candidate area 42.

  When generating the first area 51, the area specifying unit 23 generates a second area 52 in which the first area 51 is excluded from the bleeding occurrence candidate area 42. The second area 52 is an area surrounded by the boundary 44 and the reduced boundary 45 of the bleeding occurrence candidate area 42.

  On the other hand, the region identification unit 23 enlarges the bleeding onset candidate region 42 to the outside by performing dilation processing on the boundary 44 of the bleeding onset candidate region 42. The dilation process is a process of searching for the maximum value within a predetermined width centered on the boundary 44 of the bleeding onset candidate region 42 using a structural element as shown in FIG. By performing the dilation process using the structural element shown in FIG. 7 once, the bleeding occurrence candidate area 42 is enlarged by one pixel outward. Then, the region specifying unit 23 generates the third region 53 outside the boundary 44 of the bleeding occurrence candidate region 42 by performing the dilation process a predetermined number of times. The third region 53 is a region surrounded by the boundary 44 of the bleeding occurrence candidate region 42 and the enlarged boundary 46.

  The determination unit 24 compares the signal value (that is, CT value) Q1 of the first area 51, the signal value Q2 of the second area 52, and the signal value Q3 of the third area 53, and determines that the bleeding-onset candidate area is bleeding. It is determined whether it is an area. In the present embodiment, the determination unit 24 determines the average of the CT values included in each of the first area 51, the second area 52, and the third area 53 as the signal values Q1, Q2, and Q3 of each area. Calculate the value.

  Here, as shown in FIG. 5, if a washout occurs after bleeding occurs in the ventricle, the entire ventricle has a low CT value. However, even when washout occurs, a region with a high CT value due to hemorrhage appears near the boundary of the ventricle. The region with a high CT value resulting from such bleeding has a higher CT value than the region of the brain parenchyma. Therefore, when the signal value Q2 of the second area 52> the signal value Q1 of the first area 51 and the signal value Q2 of the second area 52> the signal value Q3 of the third area 53, bleeding occurs. Will be present in the ventricle of the brain. Therefore, in the present embodiment, the determination unit 24 determines that the signal value Q2 of the second area 52> the signal value Q1 of the first area 51 and the signal value Q2 of the second area 52> the third area It is determined whether or not the signal value Q3 is 53. If this determination is affirmative, the determination unit 24 determines that the bleeding-onset candidate area 42 is a bleeding area. If this determination is negative, the determination unit 24 determines that the bleeding-onset candidate area 42 is not a bleeding area.

  The display control unit 25 displays on the display 14 bleeding area information indicating that the bleeding area has been specified. FIG. 8 is a diagram showing the brain image B0 displayed on the display. As shown in FIG. 8, the display 14 displays a brain image B0 and a text 50 indicating that the left ventricle is a bleeding region as bleeding region information. In FIG. 8, the tomographic image shown in FIG. 5 is shown as a brain image B0 for explanation. By displaying the text 50 representing the bleeding area information on the display 14 in this manner, the doctor can carefully observe the bleeding onset candidate area 42 in the brain image B0 of the subject, so that the bleeding onset candidate area is Whether or not the area is a bleeding area can be easily recognized.

  The information indicating that the bleeding region has been specified is not limited to the text 50, but may be information other than text such as symbols and icons.

  Next, processing performed in the present embodiment will be described. FIG. 9 is a flowchart showing processing performed in the present embodiment. First, the image acquiring unit 21 acquires a brain image B0 of the subject (step ST1). Next, the candidate area extraction unit 22 extracts a bleeding onset candidate area 42 in the brain image B0 (step ST2). Subsequently, the region specifying unit 23 determines a first region 51 inside the boundary 44 of the bleeding onset candidate region 42, a second region 52 other than the first region 51 in the bleeding onset candidate region 42, and a bleeding onset candidate region. The third area 53 outside the boundary 44 of the area 42 is specified (area specification; step ST3). Then, the determination unit 24 compares the signal value Q1 of the first area 51, the signal value Q2 of the second area 52, and the signal value Q3 of the third area 53, and determines that the bleeding-onset candidate area 42 is a bleeding area. It is determined whether or not there is (step ST4). Further, the display control unit 25 displays bleeding area information indicating that the bleeding area has been specified on the display 14 (step ST5), and ends the processing.

  As described above, in the present embodiment, the brain image B0 is acquired, the bleeding occurrence candidate region 42 in the brain image B0 is extracted, and the first to third regions 51 to 53 are specified. Here, when a washout occurs in the bleeding area, most of the bleeding area has returned to a normal structure. However, bleeding has occurred when the signal value of the region inside the boundary of the bleeding region, the signal value of the region outside the inner region within the bleeding region, and the signal value of the region outside the bleeding region are compared. There is a difference in the signal value of each area between the case where there is no bleeding and the case where bleeding occurs and washout occurs. Therefore, by comparing the signal value Q1 of the first area 51, the signal value Q2 of the second area 52, and the signal value Q3 of the third area 53, the bleeding-onset candidate area 42 is a bleeding area. Can be determined. Therefore, according to the present embodiment, even if a washout has occurred, it is possible to accurately determine whether or not the bleeding-onset candidate area 42 is a bleeding area.

  In the above embodiment, the bleeding onset candidate region 42 is a ventricle, but the candidate region extracting unit 22 may extract a cerebral cistern as a bleeding onset candidate region. FIG. 10 is a diagram for explaining the extraction of the cisternal region in the brain image B0. As shown in FIG. 10, the skull 60 and the brain parenchyma 61 are included in the brain image B0. As for the cisternal region, the cerebral cistern region can be extracted as a hemorrhage-onset candidate region by aligning the brain image B0 with the standard brain image Bs. In FIG. 10, three cerebral cistern regions in the brain parenchyma 61 are extracted from the brain image B0 as candidate bleeding-onset regions 62, 63, and 64. In the bleeding occurrence candidate area 63, the outline shown by a broken line corresponds to the outline of the actual cistern area, but the area with a low CT value is smaller than the outline of the actual cisternal area. This is because washout has occurred after bleeding.

  In this case, the area specifying unit 23 specifies the first area, the second area, and the third area for all the bleeding occurrence candidate areas 62, 63, and 64. FIG. 11 shows the first to third regions specified for the bleeding occurrence candidate region 63. As shown in FIG. 11, a first region 71, a second region 72, and a third region 73 are specified for a bleeding-onset candidate region 63, which is a cisternal region, as in the case of the ventricle. .

  Here, as shown in FIG. 10, when a washout occurs after bleeding, the signal value Q2 of the second region 72> the first region 71, as in the case where the ventricular region is extracted as the bleeding occurrence candidate region 42. And the signal value Q2 of the second area 72> the signal value Q3 of the third area 73. For this reason, when the cisternal region is extracted as the bleeding-onset candidate regions 62, 63, and 64, the bleeding-onset candidate regions 62, 63, and 64 are also set as the bleeding-onset candidate region, similarly to the case where the ventricular region is extracted as the bleeding-onset candidate region. Can be determined.

  Further, in the above embodiment, it is a matter of course that both the ventricles and the cistern may be extracted as the bleeding onset candidate regions.

  In the above embodiment, the candidate region extracting unit 22 extracts the bleeding onset candidate region 42 from the brain image B0 using the standard brain image Bs. However, the brain image B0 is used without using the standard brain image Bs. The bleeding onset candidate area 42 may be extracted from the. For example, the absolute positions of the ventricles and the cistern are almost fixed inside the brain with respect to the skull. For this reason, the bleeding onset candidate region 42 may be extracted from the brain image B0 using information on the absolute positions of the ventricles and the cistern with reference to the skull. Further, the bleeding onset candidate region 42 may be extracted by performing a threshold process on the brain image B0. Alternatively, an arbitrary point in the bleeding onset candidate area 42 may be specified as a seed point, and the bleeding onset candidate area 42 may be extracted using a graph cut method.

  In the above-described embodiment, the signal value Q2 of the second area> the signal value Q1 of the first area, and the signal value Q2 of the second area> the signal value of the third area In the case of Q3, it is determined that the bleeding-onset candidate area is a bleeding area, but the present invention is not limited to this. For example, the bleeding-onset candidate region is a bleeding region based on the signal value distribution of the first region (hereinafter referred to as signal value distribution), the signal value distribution of the second region, and the signal value distribution of the third region. Or not.

  Specifically, the determination unit 24 determines the first signal value distribution that is the signal value distribution of the first area, the second signal value distribution that is the signal value distribution of the second area, and the signal of the third area. A third signal value distribution, which is a value distribution, is calculated, and the first to third signal value distributions are fitted to a probability density function of a normal distribution. Note that the first signal value distribution is a histogram of signal values in the first region, the second signal value distribution is a histogram of signal values in the second region, and the third signal value distribution is a histogram in the third region. Is used.

  Here, when the signal values of all the pixels in the first to third regions are totaled as samples, the average value and the variance of the signal values in the first to third regions can be calculated. If the average value and the variance value are obtained, the probability density function of the normal distribution can be uniquely defined. For this reason, the determination unit 24 determines the probability density functions for the first to third regions (hereinafter, referred to as first to third probability density functions) based on the signal values in the first to third regions. Is calculated.

  Then, the determination unit 24 fits the first signal value distribution to the first probability density function, fits the second signal value distribution to the second probability density function, and fits the third signal value distribution to the third probability density function. Fitting to the probability density function of. Further, the determination unit 24 determines an index value (hereinafter, referred to as a first index value) indicating a difference between a normal distribution of the fitted first signal value distribution and a fitted second signal value distribution, and An index value (referred to as a second index value) indicating a difference between the normal distribution of the second signal value distribution and the fitted third signal value distribution is calculated. As the index value indicating the difference, for example, KL (Kullback-Leibler) -divergence (Kulback-Leibler information amount) can be used. Then, when the first index value exceeds a predetermined threshold value Th1 and the second index value exceeds a predetermined threshold value Th2, the bleeding-onset candidate area is determined to be a bleeding area. .

  In addition, the determination unit 24 may include a discriminator that receives a histogram of signal values in the first to third regions and performs machine learning so as to determine whether or not the region is a bleeding region. In addition, as a method of machine learning, a support vector machine (SVM (Support Vector Machine)), a deep neural network (DNN (Deep Neural Network)), a convolutional neural network (CNN (Convolutional Neural Network)), and a recurrent neural network ( RNN (Recurrent Neural Network) or the like can be used.

  Further, in the above embodiment, the CT image of the subject is used as the brain image B0. However, in medical images other than CT images such as an MRI image and a PET image, a normal cistern region and a bleeding cistern region are also used. Will have different pixel values. Therefore, a medical image other than the CT image may be used as the brain image B0.

DESCRIPTION OF SYMBOLS 1 Medical image processing apparatus 2 3D image photographing apparatus 3 Image storage server 4 Network 11 CPU
Reference Signs List 12 memory 13 storage 14 display 15 input unit 21 image acquisition unit 22 candidate region extraction unit 23 region identification unit 24 determination unit 25 display control unit 30, 40, 60 skull 31, 41, 61 brain parenchyma 32 ventricular region 42 bleeding onset candidate Area 43 High-intensity area 44, 45, 46 Boundary 51, 71 First area 52, 72 Second area 53, 73 Third area 61, 62, 63 Cerebral cistern area B0 Brain image Bs Standard brain image

Claims (8)

An image acquisition unit that acquires a three-dimensional brain image including the brain of the subject;
A candidate region extraction unit for extracting a bleeding-onset candidate region in the brain image,
Identify a first area inside the boundary of the bleeding onset candidate area, a second area other than the first area in the bleeding onset candidate area, and a third area outside the boundary of the bleeding onset candidate area An area specifying unit to be
A determination unit that compares the signal value of the first region, the signal value of the second region, and the signal value of the third region to determine whether the bleeding-onset candidate region is a bleeding region; Medical image processing apparatus comprising:   In the second region, when there is a region having a signal value higher than the signal value of the first region and a signal value higher than the signal value of the third region, the bleeding-onset candidate region is determined. The medical image processing device according to claim 1, wherein the medical image processing device determines the bleeding area.   3. The medical image processing apparatus according to claim 1, wherein the candidate region extracting unit extracts the bleeding-onset candidate region by aligning a standard brain image, which is a standard brain image, with the brain image. 4.   4. The medical image processing apparatus according to claim 1, wherein the bleeding-onset candidate region is at least one of a cistern and a ventricle in a brain. 5.   The medical image processing device according to claim 1, wherein the brain image is a CT image acquired by a CT device.   The medical image processing apparatus according to claim 1, further comprising a display control unit configured to display bleeding area information indicating that the bleeding area has been specified on a display unit. Acquire a three-dimensional brain image including the subject's brain,
Bleeding onset candidate region in the brain image is extracted,
Identify a first area inside the boundary of the bleeding onset candidate area, a second area other than the first area in the bleeding onset candidate area, and a third area outside the boundary of the bleeding onset candidate area And
Medical image processing for comparing the signal value of the first area, the signal value of the second area, and the signal value of the third area to determine whether the bleeding-onset candidate area is a bleeding area Method. Acquiring a three-dimensional brain image including the subject's brain;
A procedure for extracting a bleeding-onset candidate region in the brain image,
Identify a first area inside the boundary of the bleeding onset candidate area, a second area other than the first area in the bleeding onset candidate area, and a third area outside the boundary of the bleeding onset candidate area Steps to
Comparing the signal value of the first area, the signal value of the second area, and the signal value of the third area to determine whether the bleeding-onset candidate area is a bleeding area. A medical image processing program to be executed by a computer.