JP2024058897A - Brain lesion analysis device and brain lesion analysis program - Google Patents

Abstract

[Problem] To provide a brain lesion analysis device that identifies brain functional areas that may be affected by brain lesions and notifies a user of the identified areas. [Solution] A positional relationship calculation unit 54 extracts a tumor area 74 from an analysis target brain image 44, which is a three-dimensional brain image showing the brain of a subject of interest, by image processing of the analysis target brain image 44. The positional relationship calculation unit 54 calculates a group of positional relationships between the tumor area 74 and each brain functional area 70 based on an individual brain atlas 72, which is a three-dimensional image in which the brain is partitioned into a plurality of brain functional areas 70 and is obtained by deforming a standard brain atlas 46 to fit the shape of the brain of the subject of interest, and the extracted tumor area 74. A display control unit 56 causes an affected brain functional area 76 that may be affected by a tumor corresponding to the tumor area 74 to be displayed on a display 24 of a user terminal 14 based on the group of positional relationships. [Selected Figure] Figure 3

Description

The present invention relates to a brain lesion analysis device and a brain lesion analysis program.

Conventionally, data acquired by medical imaging diagnostic devices such as MRI (Magnetic Resonance Imaging) devices and CT (Computed Tomography) devices is reconstructed to generate a three-dimensional brain image representing the subject's brain.

It is known that the brain exerts different functions depending on the part of the brain (brain functional area). Therefore, for the purpose of analysis using brain images, a technique has been proposed to compartmentalize the brain in brain images into multiple brain functional areas.

For example, Patent Document 1 discloses a medical image processing device that applies an exclusion mask to areas including abnormal regions in a brain image of a subject generated by a medical image diagnostic device, obtains a comparison portion that is the area outside the exclusion mask of the brain image, applies the same exclusion mask to an atlas image showing a standard brain in which functional brain areas have been pre-partitioned, obtains a reference partial image that is the area outside the exclusion mask of the atlas image, and compares the comparison portion with the reference partial image to partition the subject's brain in the brain image of the subject.

JP 2019-154744 A

When a subject has a brain lesion, the brain image of the subject generated by a medical imaging diagnostic device will contain a lesion area, which is an image area that corresponds to the lesion. Conventionally, when a brain image contains a lesion area, a human being such as a doctor has determined which brain functional area is affected by the lesion corresponding to the lesion area. For example, a doctor has determined which brain functional area is affected by the lesion corresponding to the lesion area by comparing the lesion area in the brain image with an atlas image in which brain functional areas are pre-divided.

In this way, when judged by humans, different judgements may be made depending on the person. More specifically, for the same brain image containing a lesion area, the brain functional area that is judged to be affected by the lesion corresponding to the lesion area may differ from person to person. In particular, because the shape of the brain differs slightly from subject to subject, it may be difficult to make an accurate judgement.

The objective of the present invention is to provide a brain lesion analysis device that identifies brain functional areas that may be affected by brain lesions and notifies the user.

The present invention is a brain lesion analysis device characterized by comprising: a positional relationship calculation unit that calculates a group of positional relationships between a lesion area extracted from a brain image of a subject of interest generated by a medical image diagnostic device and each of the multiple brain functional areas based on an atlas image showing multiple brain functional areas corresponding to parts of the brain; and a display control unit that displays on a display unit an affected brain functional area, which is a brain functional area that may be affected by a lesion corresponding to the lesion area, based on the group of positional relationships.

The present invention provides a brain lesion analysis device that identifies brain functional areas that may be affected by brain lesions and notifies the user.

1 is a schematic diagram illustrating the configuration of a brain lesion analysis system according to an embodiment of the present invention. FIG. 2 is a schematic diagram of the configuration of a user terminal. FIG. 2 is a schematic diagram of the configuration of an analysis server. FIG. 2 is a diagram showing an example of a brain image to be analyzed. FIG. 1 is a diagram showing an example of a standard brain atlas. FIG. 1 is a conceptual diagram showing the process of generating an individual brain atlas. FIG. 13 is a conceptual diagram showing the computational processing of a group of positional relationships between a tumor region and each brain functional region. FIG. 13 is a conceptual diagram showing an example of a positional relationship group. FIG. 1 is a conceptual diagram showing feature amounts of a tumor region. FIG. 13 is a diagram showing a display example of an affected brain functional region. FIG. 2 is a conceptual diagram showing the contents of a case DB. FIG. 13 is a diagram illustrating a display example of similar case data. 1 is a map showing each case data in a feature space. 13 is a flowchart showing a process flow of the analysis server.

Figure 1 is a schematic diagram of a brain lesion analysis system 10 according to this embodiment. The brain lesion analysis system 10 includes one or more medical image diagnostic devices 12, a user terminal 14, and an analysis server 16 as a brain lesion analysis device. The medical image diagnostic devices 12, the user terminal 14, and the analysis server 16 are connected to each other so that they can communicate with each other via a communication line 18, which may include, for example, an Internet line or a LAN (Local Area Network). Note that the medical image diagnostic device 12 does not need to be able to communicate with the analysis server 16, as long as it is possible to send images from the medical image diagnostic device 12 to the analysis server 16 in a unidirectional manner. Alternatively, the medical image diagnostic device 12 does not need to be connected to the analysis server 16 so that it can communicate with each other. In this case, the images generated by the medical image diagnostic device 12 are imported into the analysis server 16 via a storage medium or the like.

The medical imaging diagnostic device 12 is a device that scans a subject using radio waves, radiation, etc., and reconstructs the data obtained thereby to generate a medical image, which is a three-dimensional image that represents the subject's internal structure. In particular, the medical imaging diagnostic device 12 according to this embodiment generates a three-dimensional brain image that represents the subject's brain.

In this embodiment, an MRI device is used as the medical image diagnostic device 12. The MRI device places a subject in a static magnetic field to longitudinally magnetize the protons (protons) that constitute the nuclei of hydrogen atoms in the subject's body, irradiates the subject with radio waves (RF pulses) to transversely magnetize the subject's protons, stops irradiating the radio waves and detects the energy released when the protons return (relax) to their original state (before the RF pulse irradiation), and performs reconstruction processing based on the detection signal to generate a medical image. The MRI device can generate a T1-weighted image, which is a medical image based on longitudinal relaxation in which the axis of the precession of the protons (spins) returns from the transverse magnetized state to the static magnetic field direction, and a T2-weighted image, which is a medical image based on transverse relaxation in which the phases of each proton (spin) return from a state in which the phases of each proton (spin) are aligned due to transverse magnetization to different phases. T1-weighted images and T2-weighted images differ in the tissues that become high signals (e.g., high brightness on the image) or low signals (e.g., low brightness on the image). For example, in a T1 weighted image, fat and other substances have a high signal intensity and water and other substances have a low signal intensity, while in a T2 weighted image, water and other substances have a high signal intensity and fat and other substances have a low signal intensity.

The medical imaging diagnostic device 12 is not limited to an MRI device, so long as it can acquire a three-dimensional brain image of the subject. For example, it may be a CT device. A CT device irradiates the subject with X-rays from all directions, detects data on the X-rays that pass through the subject, and performs reconstruction processing based on the detected X-ray attenuation data to generate a medical image.

Figure 2 is a schematic diagram of the configuration of the user terminal 14. The user terminal 14 is a computer used by a user of the brain lesion analysis system 10, such as a personal computer. In this specification, a user is someone who uses brain images generated by the medical image diagnostic device 12, and is typically a doctor.

The user terminal 14 includes a communication interface 20, an input interface 22, a display 24, a memory 26, and a processor 28. The communication interface 20 includes, for example, a network adapter, and communicates with other devices such as the analysis server 16 via the communication line 18. The input interface 22 includes, for example, a keyboard, a mouse, a touch panel, and receives user instructions. The display 24 includes, for example, a liquid crystal panel, and displays various screens. The memory 26 includes, for example, a hard disk drive (HDD), a solid state drive (SSD), an embedded multi media card (eMMC), a read only memory (ROM), or a random access memory (RAM), and stores various data. The processor 28 includes, for example, a central processing unit (CPU), and controls each part of the user terminal 14.

Figure 3 is a schematic diagram of the configuration of the analysis server 16. The analysis server 16 is configured, for example, by a server computer. Note that the analysis server 16 may be configured by multiple computers. In other words, the functions performed by the analysis server 16 described below may be realized by the cooperation of multiple computers.

The communication interface 40 includes, for example, a network adapter. The communication interface 40 performs the function of communicating with other devices, including the medical image diagnostic device 12 and the user terminal 14, via the communication line 18.

The memory 42 includes, for example, a HDD, SSD, eMMC, ROM, or RAM. The memory 42 stores a brain lesion analysis program for operating each part of the analysis server 16. As shown in FIG. 3, the memory 42 also stores a brain image to be analyzed 44, a standard brain atlas 46, a case DB 48, and a learning model 50.

Figure 4 is a diagram showing an example of a brain image 44 to be analyzed. The brain image 44 to be analyzed is an image of the subject's brain that is generated by the medical image diagnostic device 12 and transmitted to the analysis server 16. Although a tomographic image is shown in Figure 4, the brain image 44 to be analyzed is a three-dimensional image. A plurality of brain images are sequentially transmitted from one or a plurality of medical image diagnostic devices 12 to the analysis server 16 and stored in the memory 42, and the brain image that the analysis server 16 is currently analyzing is called the brain image 44 to be analyzed. In this specification, the subject corresponding to the brain image 44 to be analyzed is called the "subject of interest." As described above, the brain image generated by the medical image diagnostic device 12 may be imported into the analysis server 16 via a storage medium or the like.

In this embodiment, the subject's brain is assumed to have a lesion, and the analysis target brain image 44 is assumed to have a lesion area, which is an image area corresponding to the lesion. A lesion is typically a tumor, and in this embodiment, an example in which the lesion is a tumor will be described. However, the lesion in the subject's brain may be something other than a tumor, for example, a nodule or other lesion. As will be described in more detail below, in the analysis target brain image 44 in Figure 4, the area indicated by diagonal lines on the right side is the tumor area, which is the lesion area. In an actual brain image, a lesion is represented by, for example, high-luminance pixels (white pixels).

Figure 5 is a diagram showing an example of a standard brain atlas 46. Like the analysis target brain image 44, the standard brain atlas 46 is a three-dimensional image of the brain, and is an image showing multiple brain functional areas 70 corresponding to parts of the brain. While the size and shape of the brain vary slightly depending on the subject, the standard brain atlas 46 is an image showing a standard (general) brain, and is an image in which the standard brain is compartmentalized into multiple brain functional areas 70. Examples of brain functional areas 70 include "Amygdala_R (right amygdala)", "Olfactory_R (right olfactory bulb)", and "OFCpost_R (right orbitofrontal cortex)".

The standard brain atlas 46 is composed of a group of pixels arranged in three dimensions, and for example, information that identifies each brain functional area 70 is associated as attribute information with each group of pixels that make up each brain functional area 70. For example, among the pixels in the standard brain atlas 46, information indicating "Amygdala_R (right amygdala)" is associated with a group of pixels that make up "Amygdala_R (right amygdala)."

As with Figure 4, Figure 5 shows cross-sectional images, but as mentioned above, the standard brain atlas 46 is also a three-dimensional image.

Details about the case DB 48 and learning model 50 and how to use them will be described later.

The processor 52 includes, for example, a CPU, a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), etc. The processor 52 may not be a single processing device, but may be configured by the cooperation of multiple processing devices located at physically separate locations. As shown in FIG. 3, the processor 52 performs the functions of a positional relationship calculation unit 54, a display control unit 56, a similar case identification unit 58, and a learning processing unit 60 according to the brain lesion analysis program stored in the memory 42.

The positional relationship calculation unit 54 first extracts the tumor region from the analysis target brain image 44 by performing image processing on the analysis target brain image 44. A known method can be used to extract the tumor region. For example, an edge extraction method, a region growing method, or a combination of these methods can be used.

Next, the positional relationship calculation unit 54 generates an individual brain atlas by deforming the standard brain atlas 46 to match the shape of the brain of the subject of interest based on the analysis target brain image 44. FIG. 6 is a conceptual diagram showing the process of generating the individual brain atlas 72. A known method can also be used to generate the individual brain atlas 72. For example, a method can be used in which multiple feature points are extracted from each of the analysis target brain image 44 and the standard brain atlas 46, and the standard brain atlas 46 is deformed so that the feature points of the standard brain atlas 46 match the corresponding feature points of the analysis target brain image 44. The individual brain atlas 72 can be said to be an image in which a brain of the same shape (or very similar shape) as that of the subject of interest is partitioned into multiple brain functional areas 70.

Figure 7 is a conceptual diagram showing the computation process of a group of positional relationships between a tumor region 74 and each brain function region 70. The positional relationship computation unit 54 computes the positional relationships between the tumor region 74 and each brain function region 70 based on the tumor region 74 extracted from the analysis target brain image 44 and the generated individual brain atlas 72. As described above, since multiple brain function regions 70 are defined in the individual brain atlas 72, the positional relationship computation unit 54 computes multiple positional relationships between the tumor region 74 and each of the multiple brain function regions 70. In this specification, the multiple positional relationships are referred to as a group of positional relationships.

Here, the positional relationship is a concept that includes the distance between the tumor region 74 and the brain function region 70, and the amount of overlap between the tumor region 74 and the brain function region 70.

The distance between the tumor region 74 and the brain function region 70 is, for example, the distance between the center of gravity of the tumor region 74 and the center of gravity of the brain function region 70. The positional relationship calculation unit 54 aligns the analysis target brain image 44 with the individual brain atlas 72, identifies the position (coordinates) of the center of gravity of the tumor region 74 in the analysis target brain image 44, obtains the position (coordinates) of the center of gravity of each brain function region 70 in the individual brain atlas 72, and calculates the distance between the center of gravity coordinates of the tumor region 74 and the center of gravity coordinates of each brain function region 70. Here, the distance may be a Euclidean distance or a Mahalanobis distance that reflects the variability of the case data. In addition, the distance between the tumor region 74 and the brain function region 70 may be the shortest distance between the tumor region 74 and the brain function region 70, more specifically, the shortest distance between a point in the tumor region 74 and a point in the brain function region 70.

As a method for calculating the amount of overlap between the tumor region 74 and the brain function region 70, for example, the positional relationship calculation unit 54 aligns the analysis target brain image 44 with the individual brain atlas 72 and then superimposes them, and calculates the amount of overlap of the tumor region 74 superimposed on each brain function region 70 in the individual brain atlas 72. As the amount of overlap, the volume of the overlapping part may be calculated, or the overlap ratio (with respect to the tumor region) which is the ratio of a certain brain function region 70 to the tumor region 74 may be calculated, or the overlap ratio (with respect to the brain function region) which is the ratio of the tumor region 74 to a certain brain function region 70 may be calculated. Note that a known method may be used as a method for calculating the volume of a specific part based on a three-dimensional medical image acquired by the medical image diagnostic device 12.

Figure 8 is a table showing an example of a group of positional relationships between the tumor region 74 and each of multiple brain functional regions 70, calculated by the positional relationship calculation unit 54. In the example of Figure 8, the positional relationship calculation unit 54 calculates the distance (Euclidean distance and Mahalanobis distance) between the tumor region 74 and each brain functional region 70 (e.g., "Precentral_L (left precentral prefrontal region)", "Precentral_R (right precentral prefrontal region)", etc.), as well as the overlap rate (with respect to the tumor region) and overlap rate (with respect to the brain functional region) of the tumor region 74.

Here, the positional relationship calculation unit 54 may also acquire the feature amount of the tumor region 74 by image processing of the analysis target brain image 44. The feature amount of the tumor region 74 may be, but is not limited to, the volume of the tumor region 74 (tumor volume) and the signal value of each pixel contained in the tumor region 74. The signal value of each pixel contained in the tumor region 74 may be, for example, the maximum value, minimum value, average value, variance, standard deviation, entropy, etc. of the signal value of the pixel in the tumor region 74. FIG. 9 shows an example of the feature amount of the tumor region 74 acquired by the positional relationship calculation unit 54. In FIG. 9, T1 means the signal value in the T1 weighted image, and T2 means the signal value in the T2 weighted image. In this way, the positional relationship calculation unit 54 may acquire the feature amount of the tumor region 74 in multiple types of medical images. The method of using the feature amount of the tumor region 74 will be described later.

The positional relationship calculation unit 54 calculates the group of positional relationships between the tumor region 74 and each of the multiple brain function regions 70, which means that it calculates the magnitude of the influence of a tumor as a lesion corresponding to the tumor region 74 on each brain function region 70. Specifically, it can be said that the smaller the distance between the tumor region 74 and the brain function region 70, or the greater the amount of overlap between the tumor region 74 and the brain function region 70, the greater the influence of the tumor corresponding to the tumor region 74 on the brain function region 70. Conversely, it can be said that the greater the distance between the tumor region 74 and the brain function region 70, or the smaller the amount of overlap between the tumor region 74 and the brain function region 70, the smaller the influence of the tumor corresponding to the tumor region 74 on the brain function region 70. In this way, at least one of the distance between the tumor region 74 and the brain function region 70 or the amount of overlap between the tumor region 74 and the brain function region 70, i.e., the positional relationship between the tumor region 74 and the brain function region 70, is a parameter indicating the magnitude of the influence of the tumor corresponding to the tumor region 74 on the brain function region 70 (in this specification, the parameter is referred to as an "influence parameter"). In addition, the positional relationship calculation unit 54 may calculate a parameter that takes into account both the distance and the amount of overlap between the tumor region 74 and the brain function region 70 (e.g., a parameter that takes into account the weighting of both) as the influence parameter.

The display control unit 56 displays on the display unit the brain functional area 70 that may be affected by the tumor corresponding to the tumor region 74 (in this specification, the brain functional area 70 is referred to as the "affected brain functional area") based on the group of positional relationships between the tumor region 74 and each of the multiple brain functional areas 70 calculated by the positional relationship calculation unit 54. In this embodiment, the display control unit 56 displays the affected brain functional area on the display 24 of the user terminal 14. The affected brain functional area may be, for example, a brain functional area 70 whose influence parameter calculated by the positional relationship calculation unit 54 is equal to or greater than a predetermined influence threshold, or a predetermined number of brain functional areas 70 in order of the largest influence parameter. The influence that the tumor corresponding to the tumor region 74 may have on the brain functional area 70 is a concept that includes the influence of aftereffects caused when the tumor is treated (such as resection) by surgery, and the influence of the tumor itself.

Figure 10 is a diagram showing an example of the display of affected brain functional regions 76. As shown in Figure 10, the display control unit 56 displays the affected brain functional regions 76 on the display 24 of the user terminal 14 in response to a request from the user terminal 14. In the example of Figure 10, four affected brain functional regions 76 are displayed in descending order of influence parameters. As shown in Figure 10, the display control unit 56 may display multiple affected brain functional regions 76 on the display 24 in a manner that allows identification of the order of the magnitude of the influence parameters calculated by the positional relationship calculation unit 54 (in other words, the order of the magnitude of the influence of the tumor corresponding to the tumor region 74).

In the example of FIG. 10, the affected brain functional areas 76 are displayed in text and the order of the magnitude of the influence parameters is expressed in the form of a ranking (〇th place), but the display control unit 56 may display the affected brain functional areas 76 and the order of the magnitude of the influence parameters in other ways. For example, the display control unit 56 may display the individual brain atlas 72 on the display 24 and then indicate the affected brain functional areas 76 in the individual brain atlas 72, for example, by coloring. The display control unit 56 may also display the order of the magnitude of the influence parameters in a manner that makes them identifiable by color. Note that the display control unit 56 can also display the influence parameters themselves, i.e., the positional relationship group (see FIG. 8), on the display 24.

The display control unit 56 may display, on the display 24, the analysis target brain image 44 showing the tumor region 74, together with the affected brain functional region 76, and the individual brain atlas 72, as shown in FIG. 10.

As described above, in this embodiment, the affected brain functional regions 76 affected by the tumor corresponding to the tumor region 74 are identified and displayed. This allows the user to easily grasp the affected brain functional regions 76, and also prevents the occurrence of variability in the judgment of the affected brain functional regions 76 depending on the user. Also, in this embodiment, as described above, the influence parameter of each brain functional region 70 indicating the magnitude of the influence of the tumor corresponding to the tumor region 74 on each brain functional region 70 is calculated in the form of a numerical value, so that the order of the magnitude of the influence can be identified. Then, by displaying the order, the user can easily grasp the order of the magnitude of the influence of the tumor corresponding to the tumor region 74 on each brain functional region 70.

Before describing the similar case identification unit 58, the case DB 48 will be described below. The case DB 48 is a database that stores multiple case data related to multiple subjects (in this specification, these subjects are referred to as "past subjects") whose brain images were analyzed by the analysis server 16 before the subject of interest.

Figure 11 is a diagram showing an example of the contents of case DB48. As shown in Figure 11, the case data of each past subject includes a case number, medical information, tumor area features, a positional relationship group, affected brain functional areas, and brain images.

The case number is an identifier that uniquely identifies the case data. The case number is assigned by the processor 52 or the user when the case data is stored in the case DB 48. The medical information is attribute information of the past subject. Examples of the attribute information of the past subject include the sex, age (age at the time of diagnosis), race, and grade (malignancy of the tumor) of the past subject. The medical information is input by a user such as a doctor.

As described above, the tumor region features are obtained by the positional relationship calculation unit 54 performing image analysis of the brain images of the past subject. An example of the contents of the tumor region feature items is shown in FIG. 9. As described above, the positional relationship group is a group of positional relationships between the tumor region and multiple brain functional regions calculated by the positional relationship calculation unit 54 based on the tumor region extracted from the brain image of the past subject and the personal brain atlas for that past subject, which is obtained by modifying the standard brain atlas 46 based on the brain image of the past subject. An example of the contents of the positional relationship group is shown in FIG. 8.

The affected brain functional area is the brain functional area affected by the brain tumor of the past subject, identified by the positional relationship calculation unit 54 according to the positional relationship group of the past subject. In the case DB 48, multiple affected brain functional areas may be associated with one past subject, and in that case, information indicating the order of the magnitude of the effect of the tumor on the multiple affected brain functional areas may also be stored.

The similar case identification unit 58 refers to the case data DB 48 and identifies case data similar to the case of the subject of interest based on the set of positional relationships between the tumor region 74 and each of the multiple brain function regions 70 for the subject of interest calculated by the positional relationship calculation unit 54 and the set of positional relationships included in each case data (i.e., the set of positional relationships between the tumor region and each of the multiple brain function regions for past subjects). In this specification, case data similar to the case of the subject of interest is referred to as "similar case data".

In detail, first, the similar case identification unit 58 calculates the similarity between positional relationships between the group of positional relationships for the subject of interest and the group of positional relationships for each past subject. A known method can be used to calculate the similarity between positional relationships. For example, the similar case identification unit 58 defines a multidimensional vector having as elements the distance and the amount of overlap between the tumor area and each brain function area, and calculates the similarity (e.g., cosine similarity) between the multidimensional vector for the subject of interest and the multidimensional vector for each past subject. As a result, the similarity between positional relationships is calculated as a numerical value between the group of positional relationships for the subject of interest and the group of positional relationships for each past subject.

The similarity between positional relationships is one index of inter-case similarity, which is the similarity between the case of the subject of interest and the case data. As described below, the similarity between cases can also be calculated based on multiple indexes including the similarity between positional relationships, but here we will explain an example in which the similarity between positional relationships is directly used as the similarity between cases (in this case, similarity between positional relationships and similarity between cases are synonymous).

The similar case identification unit 58 identifies similar case data based on the calculated inter-case similarity (here, positional relationship similarity). For example, the similar case data can be case data whose inter-case similarity is equal to or exceeds a predetermined similarity threshold, or a predetermined number of case data in order of greatest inter-case similarity.

The display control unit 56 displays the similar case data identified by the similar case identification unit 58 on the display unit. FIG. 12 is a diagram showing an example of display of similar case data. As shown in FIG. 12, the display control unit 56 displays similar case data 78 on the display 24 of the user terminal 14 together with the analysis target brain image 44 and the affected brain functional region 76 of the subject of interest. In the example of FIG. 12, the medical information, tumor region feature amount, affected brain functional region, and brain image of the past subject are displayed as the similar case data 78, but the items displayed as the similar case data 78 are not limited to these. In the example of FIG. 12, the display control unit 56 also displays the medical information and tumor region feature amount of the subject of interest in correspondence with the similar case data 78. The medical information of the subject of interest is input in advance to the analysis server 16 by a user such as a doctor.

When the similar case identification unit 58 has identified a plurality of similar case data 78, the display control unit 56 may display the plurality of similar case data 78 on the display 24 in a manner that allows the order of the magnitude of the inter-case similarity between the case of the subject of interest to be identified. For example, as shown in FIG. 12, the display control unit 56 can display the order of the magnitude of the inter-case similarity between the plurality of similar case data 78 by text. In the example of FIG. 12, it is shown by text that the inter-case similarity of the similar case data 78a is first (the inter-case similarity between the case of the subject of interest is the highest among the plurality of case data stored in the case DB 48), the inter-case similarity of the similar case data 78b is second, and the inter-case similarity of the similar case data 78c is third. Note that the display manner of the order of the inter-case similarity between the plurality of similar case data 78 is not limited to text. For example, the order of the inter-case similarity between the plurality of similar case data 78 may be shown by the display order of the plurality of similar case data 78 (for example, the higher the inter-case similarity of the similar case data 78, the higher the similar case data 78 is displayed).

By displaying similar case data 78, the user can easily obtain similar case data 78 without having to search case DB 48 himself/herself. By obtaining similar case data 78, the user can refer to the diagnosis, treatment, or prognosis of past subjects (i.e., similar cases) related to similar case data 78 when deciding the treatment content for the subject of interest. This information can potentially be used in considering the treatment content, IC (Informed Consent), conferences, etc.

In the above embodiment, the similarity between positional relationships is directly used as the similarity between cases, but the similar case identification unit 58 may identify similar case data or calculate the similarity between cases based on the features of the tumor region 74 of the subject of interest and the features of the tumor regions of past subjects included in the case data. For example, the similar case identification unit 58 defines a multidimensional vector having not only a group of positional relationships between the tumor region and each brain functional region, but also features of the tumor region (e.g., tumor volume, see FIG. 9) as elements, and calculates the similarity between the multidimensional vector for the subject of interest and the multidimensional vector for each past subject. This allows the similarity between cases to be calculated taking into account the features of the tumor region.

The similar case identification unit 58 may also identify similar case data or calculate the similarity between cases based on the medical information of the subject of interest and the medical information of past subjects included in the case data. For example, the similar case identification unit 58 defines a multidimensional vector having elements of medical information (e.g., gender, age, grade, etc.) as well as a group of positional relationships between the tumor region and each functional brain region, and calculates the similarity between the multidimensional vector for the subject of interest and the multidimensional vector for each past subject. This allows the similarity between cases to be calculated taking into account the medical information.

Of course, in addition to the positional relationship similarity, the inter-case similarity may be calculated based on both the features of the tumor region and the medical information. For example, the similar case identification unit 58 defines a multidimensional vector having as its elements the set of positional relationships between the tumor region and each functional brain region, the features of the tumor region, and the medical information, and calculates the similarity between the multidimensional vector for the subject of interest and the multidimensional vector for each past subject.

In the above embodiment, in order for the display control unit 56 to display similar case data on the display 24, the similar case identification unit 58 calculates the inter-case similarity between all case data stored in the case DB 48 and the subject of interest. However, if a huge number of case data are stored in the case DB 48, the amount of calculations that the similar case identification unit 58 must perform to calculate the inter-case similarity becomes enormous.

In view of this, the similar case identification unit 58 may classify the multiple case data stored in the case DB 48 into multiple clusters, which are collections of similar case data, identify the cluster to which the case of the subject of interest belongs from among the multiple clusters (referred to as the "belonging cluster"), and calculate the inter-case similarity between the case of the subject of interest and each case data included in the belonging cluster, respectively. This allows the similar case identification unit 58 to identify similar case data without the need to calculate the inter-case similarity between the case of the subject of interest and case data not included in the belonging cluster. Details of this process are described below.

First, the similar case identification unit 58 classifies the multiple case data stored in the case DB 48 into multiple clusters. For example, K-means clustering can be used as a classification method. In detail, the similar case identification unit 58 randomly classifies each case data stored in the case DB 48 into multiple clusters. Then, in each cluster, the center of gravity of the cluster is calculated, and the case data closest (similar) to the center of gravity is added to the cluster. Here, when calculating the center of gravity and identifying the case data closest to the center of gravity, a positional relationship group included in the case data is used. For example, for multiple multidimensional vectors having as elements the positional relationship group (distance or overlap between the tumor area and each brain function area) of the multiple case data included in each cluster, a multidimensional vector obtained by taking the average for each element can be used as the center of gravity. In addition, the case data most similar to the center of gravity can be identified based on the similarity between the multidimensional vector as the center of gravity and the multidimensional vector of the case data. The elements of the multidimensional vector may include not only the positional relationship group but also the feature amount of the tumor area and the medical information of the past subject. By repeatedly calculating the center of gravity of the cluster and adding the case data closest to the center of gravity to that cluster, multiple case data are classified into multiple clusters, which are collections of similar case data.

Figure 13 is a conceptual diagram showing how multiple case data CA are classified into multiple clusters CL in a feature space. The feature space can have four or more dimensions (the number of elements in the multidimensional vector), but for convenience, the feature space is represented as a three-dimensional space in the example of Figure 13.

Next, the similar case identification unit 58 identifies the cluster to which the case of the subject of interest belongs from among the multiple clusters. In this embodiment, the learning model 50 is used for the process of identifying the cluster to which the case belongs. Here, the learning model 50 and the learning processing unit 60 will be described.

The learning model 50 is a learning device such as a neural network, which learns using learning data and is capable of outputting output data corresponding to the input data with high accuracy based on the input data. In this embodiment, the learning model 50 receives case data as input data and outputs the cluster to which the case data belongs as output data. The content of the case data as input data may be only a positional relationship group, but may also include features of the tumor area and past medical information of the subject.

The learning processing unit 60 uses a combination of case data stored in the case DB 48 and the cluster (teacher data) to which the case data belongs as learning data, and trains the learning model 50 to predict and output the cluster of the case data when the case data is input.

The similar case identification unit 58 can input the positional relationships of the subject of interest (or further the features of the tumor region 74 or the medical information of the subject of interest) to the fully trained learning model 50, and obtain the cluster to which the case of the subject of interest belongs as output data of the learning model 50. In the example of Figure 13, it is shown that the cluster to which the case CAref of the subject of interest belongs is cluster CL1.

Then, the similar case identification unit 58 calculates the inter-case similarity between the case CAref of the subject of interest and each case data CA included in the cluster CL1 to which the subject belongs. The method of calculating the inter-case similarity is as described above. As a result, similar case data similar to the case CAref of the subject of interest is identified from the case data CA included in cluster CL1.

The processing flow of the analysis server 16 will be explained below according to the flowchart shown in Figure 14.

In step S10, the positional relationship calculation unit 54 extracts the tumor region 74 from the analysis target brain image 44 by image processing of the analysis target brain image 44.

In step S12, the positional relationship calculation unit 54 performs image processing on the analysis target brain image 44 to obtain features of the tumor region 74, such as the volume of the tumor region 74 (tumor volume) and the signal values of each pixel contained in the tumor region 74.

In step S14, the positional relationship calculation unit 54 generates an individual brain atlas 72 by deforming the standard brain atlas 46 to fit the shape of the brain of the subject of interest based on the analysis target brain image 44.

In step S16, the positional relationship calculation unit 54 calculates a group of positional relationships between the tumor region 74 and each brain functional region 70 based on the tumor region 74 extracted from the analysis target brain image 44 in step S10 and the individual brain atlas 72 generated in step S14. The group of positional relationships between the tumor region 74 and each brain functional region 70 represents the magnitude of the influence that the tumor corresponding to the tumor region 74 has on each brain functional region 70. Here, the display control unit 56 may display, based on the group of positional relationships, an affected brain functional region 76 that may be influenced by the tumor corresponding to the tumor region 74 on the display 24 of the user terminal 14 (see FIG. 10).

In step S18, the similar case identification unit 58 identifies similar case data that is similar to the case of the subject of interest from among the multiple case data stored in the case data DB 48.

In step S20, the display control unit 56 displays the affected brain functional area 76 of the subject of interest identified in step S16 and the similar case data identified in step S18 on the display 24 of the user terminal 14 (see FIG. 12).

Although the embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various modifications are possible without departing from the spirit of the present invention.

For example, in this embodiment, the analysis target brain image 44, the standard brain atlas 46, the case DB 48, and the learning model 50 are stored in the memory 42 of the analysis server 16, but these may be stored in the memory of another device accessible from the analysis server 16. Also, in this embodiment, the learning model 50 is learned by the processor 52 (more specifically, the learning processing unit 60) of the analysis server 16, but the learned learning model 50 may be stored in the memory 42, in which case the processor 52 may not have the function of the learning processing unit 60.

10 Brain lesion analysis system, 12 Medical imaging diagnostic device, 14 User terminal, 16 Analysis server, 18 Communication line, 20, 40 Communication interface, 22 Input interface, 24 Display, 26, 42 Memory, 28, 52 Processor, 44 Brain image to be analyzed, 46 Standard brain atlas, 48 Case DB, 50 Learning model, 54 Positional relationship calculation unit, 56 Display control unit, 58 Similar case identification unit, 60 Learning processing unit, 70 Brain functional area, 72 Individual brain atlas, 74 Tumor area, 76 Affected brain functional area, 78 Similar case data.

Claims (8)

a positional relationship calculation unit that calculates a group of positional relationships between a lesion area extracted from a brain image of a subject of interest generated by a medical image diagnostic device and each of the plurality of brain functional areas based on an atlas image showing the plurality of brain functional areas corresponding to parts of the brain;
a display control unit that displays on a display unit an affected brain functional area that is a brain functional area that may be affected by a lesion corresponding to the lesion area, based on the positional relationship group;
A brain lesion analysis device comprising: The positional relationship calculation unit determines, based on the group of positional relationships, a magnitude of an effect that a lesion corresponding to the lesion area may have on the affected brain functional areas;
The display control unit causes the display unit to display the plurality of affected brain functional regions in a manner that allows the determined order of the magnitude of the influence to be identified.
2. The brain lesion analysis device according to claim 1 . a similar case identification unit that refers to a case database that stores a plurality of case data regarding a plurality of past subjects, the case data including a group of positional relationships between a lesion area extracted from a brain image of the past subject and each of the plurality of brain functional areas, and identifies similar case data that is the case data similar to the case of the subject of interest, based on the group of positional relationships regarding the subject of interest and the group of positional relationships included in each case data;
Further equipped with
the display control unit causes the similar case data to be displayed on a display unit.
3. The brain lesion analysis device according to claim 1 or 2. the similar case identifying unit calculates an inter-case similarity between the case of the subject of interest and each of the case data based on the positional relationship group for the subject of interest and the positional relationship group included in each of the case data;
the display control unit causes the display unit to display the plurality of similar case data in a manner that enables identification of the order of magnitude of the inter-case similarity.
4. The brain lesion analysis device according to claim 3. The similar case identifying unit,
classifying the case data into a plurality of clusters based on inter-case similarities between the case data based on the group of positional relationships between the plurality of case data;
Identifying a cluster to which a case of the subject of interest belongs from among the plurality of clusters based on the group of positional relationships for the subject of interest;
calculating inter-case similarities between the cases of the subject of interest and each case data included in the cluster based on the positional relationship group for the subject of interest and the positional relationship group included in each case data included in the cluster;
5. The brain lesion analysis device according to claim 4. the similar case identifying unit identifies the similar case data or calculates the inter-case similarity based on the feature amount of the lesion area of the subject of interest and the feature amount of the lesion area of the past subjects.
6. The brain lesion analysis device according to claim 3, the similar case identifying unit identifies the similar case data or calculates the inter-case similarity based on the medical information of the subject of interest and the medical information of the past subjects;
7. The brain lesion analysis device according to claim 3, wherein the brain lesion analysis device is a computer-readable storage medium. Computer,
a positional relationship calculation unit that calculates a group of positional relationships between a lesion area extracted from a brain image of a subject of interest generated by a medical image diagnostic device and each of the plurality of brain functional areas based on an atlas image showing the plurality of brain functional areas corresponding to parts of the brain;
a display control unit that displays on a display unit an affected brain functional area that is a brain functional area that may be affected by a lesion corresponding to the lesion area, based on the positional relationship group;
A brain lesion analysis program characterized by functioning as a