Brain medical image processing device and brain medical image processing system

Abstract

A brain medical image processing device according to an embodiment comprises a calculation unit (150d) and a determination unit (150e). The calculation unit (150d) calculates, for each of a plurality of functional areas of the brain, an indicator related to the amount of a predetermined substance distributed in each of the functional areas. The determination unit (150e) determines the presence or absence of the predetermined substance related to a lesion in the brain, or calculates a determination value related to the presence or absence of the predetermined substance, on the basis of the indicator calculated for each of the plurality of functional areas.

Description

Brain medical image processing device and brain medical image processing system

[0001] The embodiments disclosed in the present specification and drawings relate to a brain medical image processing device and a brain medical image processing system.

[0002] Computer-aided diagnosis (CAD) refers to standalone software or equipment with this software built in that has the function of not only detecting areas suspected of being a lesion, but also differentiating between benign and malignant lesion candidates, and outputting quantitative data such as the degree of disease progression and probability of malignancy as numerical values ​​and graphs. CAD is currently widely used in the diagnosis of solid cancers such as lung cancer and breast cancer, and many products are already in clinical use.

[0003] For organs other than the brain, such as the solid cancers mentioned above, there are definitive diagnostic methods that can directly convert suspected lesions into lesions and even determine whether they are benign or malignant by cytology or histology. The results of the definitive diagnosis, such as the presence or absence of a lesion or whether it is malignant or benign, can be used as true values ​​for evaluating the diagnostic performance of CAD.

[0004] On the other hand, in the case of brain tissue, cytology or histology that requires skull puncture for diagnostic purposes is not permitted, which means that a definitive diagnosis cannot be made for dementia, particularly Alzheimer's disease (AD), which cannot be definitively diagnosed using conventional medical brain imaging diagnostic devices.

[0005] Amyloid PET imaging is the only diagnostic imaging device for AD that is clinically applied as the gold standard. Amyloid PET imaging detects the accumulation of amyloid beta protein, the protein that causes AD. In amyloid PET imaging, the diagnostic results are presented as "amyloid PET positive / negative" based on a visual evaluation of the PET images. This expression essentially sets a threshold for quantitative data on amyloid beta protein, the protein that causes AD, and determines whether the result is positive if it is above the threshold or negative if it is below the threshold, resulting in a qualitative assessment of the data in two categories: positive / negative or present / absent.

[0006] Therefore, it can be said that a unique feature of brain medical imaging diagnosis is that the actual diagnostic result of amyloid PET imaging, which is the gold standard for diagnosing AD, - positive / negative for amyloid beta protein - is not a definitive diagnosis.

[0007] Visual evaluation of amyloid PET imaging is performed while paying attention to the amount of amyloid β protein at sites of high accumulation where accumulation of amyloid β protein is likely to be observed clinically.

[0008] Currently, there is an analytical technology that supports the diagnosis of dementia based on the relationship between the volume of the hippocampus and the magnetic susceptibility of the entorhinal cortex. This technology observes the hippocampus and its surrounding areas, which are the brain functional regions most closely related to cognitive decline due to Alzheimer's disease. However, in the very early stages of Alzheimer's disease, it is necessary to observe the spatiotemporal changes in the accumulation of amyloid beta protein in areas other than the hippocampus, so analyzing only a localized area may not be sufficient.

[0009] U.S. Patent No. 6,501,272 U.S. Patent No. 6,658,280

[0010] One of the problems to be solved by the embodiments disclosed in the present specification and drawings is to determine the presence or absence of a substance that causes a pathology in Alzheimer's disease or the like.

[0011] A brain medical image processing apparatus according to an embodiment includes a calculation unit and a determination unit. The calculation unit calculates, for each of a plurality of functional regions of the brain, an index related to the amount of a predetermined substance distributed in the functional region. The determination unit determines the presence or absence of a substance causing a lesion in the brain, or calculates a determination value related to the amount of the predetermined substance, based on the index calculated for each of the functional regions.

[0012] FIG. 1 is a diagram illustrating an example of the configuration of a magnetic resonance imaging apparatus 100 according to a first embodiment. FIG. 2 is a diagram illustrating a background according to the embodiment. FIG. 3 is a diagram illustrating a background according to the embodiment. FIG. 4 is a diagram illustrating processing performed by an image processing apparatus according to the first embodiment. FIG. 5 is a diagram illustrating processing performed by an image processing apparatus according to the first embodiment. FIG. 6 is a flowchart illustrating the flow of processing performed by the magnetic resonance imaging apparatus 100 according to the first embodiment. FIG. 7 is a flowchart illustrating the processing of step S100 in FIG. 6 in more detail. FIG. 8 is a flowchart illustrating the processing of steps S110 and S120 in FIG. 7 in more detail. FIG. 9 is a flowchart illustrating the processing of steps S150 and S160 in FIG. 7 in more detail. FIG. 10 is a diagram illustrating an example of a display of indices calculated in the first embodiment. FIG. 11A is a diagram illustrating a first example of processing performed by the magnetic resonance imaging apparatus according to the first embodiment. FIG. 11B is a diagram illustrating a second example of processing performed by the magnetic resonance imaging apparatus according to the first embodiment. FIG. 11C is a diagram showing a third example of processing performed by the magnetic resonance imaging apparatus according to the first embodiment. FIG. 12 is a diagram showing an example of processing performed by the magnetic resonance imaging apparatus according to the first embodiment. FIG. 13 is a diagram illustrating a background according to the second embodiment. FIG. 14 is a diagram illustrating a background according to the second embodiment. FIG. 15 is a diagram illustrating a background according to the second embodiment. FIG. 16 is a diagram illustrating processing performed by the magnetic resonance imaging apparatus 100 according to the second embodiment. FIG. 17 is a diagram illustrating processing performed by the magnetic resonance imaging apparatus 100 according to the third embodiment. FIG. 18 is a diagram illustrating processing performed by the magnetic resonance imaging apparatus 100 according to the fourth embodiment. FIG. 19 is a diagram showing an example of a display screen according to an embodiment.

[0013] First Embodiment A brain medical image processing apparatus according to an embodiment will be described in detail below with reference to the drawings. Note that Fig. 1 shows an example of a configuration in which a magnetic resonance imaging apparatus 100 includes an image processing apparatus 130.

[0014] FIG. 1 shows an example of the configuration of a magnetic resonance imaging apparatus 100 corresponding to the brain medical image processing system of the present invention. The magnetic resonance imaging apparatus 100 includes a static magnetic field magnet 101, a static magnetic field power supply (not shown), a gradient magnetic field coil 103, a gradient magnetic field power supply 104, a bed 105, a bed control circuit 106, a transmission coil 107, a transmission circuit 108, a reception coil 109, a reception circuit 110, a sequence control unit 120, and an image processing apparatus 130. Note that the magnetic resonance imaging apparatus 100 does not include a subject P (e.g., a human body). The configuration shown in FIG. 1 is merely an example. For example, the components of the sequence control unit 120 and the image processing apparatus 130 may be integrated or separated as appropriate.

[0015] The static magnetic field magnet 101 is a magnet formed in a hollow, approximately cylindrical shape, and generates a static magnetic field in the space inside the cylinder in the direction of its central axis (Z-axis). The static magnetic field magnet 101 is, for example, a superconducting magnet, and is excited by receiving a current from a static magnetic field power supply. The static magnetic field power supply supplies a current to the static magnetic field magnet 101. As another example, the static magnetic field magnet 101 may be a permanent magnet, in which case the magnetic resonance imaging apparatus 100 does not need to include a static magnetic field power supply. Furthermore, the static magnetic field power supply may be provided separately from the magnetic resonance imaging apparatus 100.

[0016] The gradient magnetic field coil 103 is a hollow, approximately cylindrical coil and is disposed inside the static magnetic field magnet 101. The gradient magnetic field coil 103 is formed by combining three coils corresponding to the mutually orthogonal X, Y, and Z axes. These three coils are individually supplied with current from a gradient magnetic field power supply 104 to generate gradient magnetic fields along the X, Y, and Z axes, whose magnetic field strength in the Z direction varies depending on the distance from the center of each axis. The gradient magnetic fields of the X, Y, and Z axes generated by the gradient magnetic field coil 103 are, for example, a slice gradient magnetic field Gs, a phase encoding gradient magnetic field Ge, and a readout gradient magnetic field Gr. The gradient magnetic field power supply 104 supplies current to the gradient magnetic field coil 103.

[0017] The bed 105 includes a top plate 105a on which the subject P is placed, and under the control of a bed control circuit 106, the top plate 105a is inserted into the cavity (imaging port) of the gradient magnetic field coil 103 with the subject P placed thereon. The bed 105 is usually installed so that its longitudinal direction is parallel to the central axis of the static magnetic field magnet 101. Under the control of the image processing device 130, the bed control circuit 106 drives the bed 105 to move the top plate 105a in the longitudinal direction and up and down.

[0018] The transmitting coil 107 is disposed inside the gradient magnetic field coil 103, and generates a radio frequency magnetic field upon receiving RF (Radio Frequency) pulses from a transmitting circuit 108. The transmitting circuit 108 supplies the transmitting coil 107 with RF pulses corresponding to a Larmor frequency determined by the type of atom of interest and the magnetic field strength.

[0019] The receiving coil 109 is disposed inside the gradient magnetic field coil 103, and receives magnetic resonance signals (hereinafter referred to as "MR signals" as necessary) emitted from the subject P by excitation with a radio frequency magnetic field. Upon receiving the magnetic resonance signals, the receiving coil 109 outputs the received magnetic resonance signals to the receiving circuitry 110.

[0020] The above-described transmitting coil 107 and receiving coil 109 are merely examples. The coil may be configured by combining one or more of a coil having only a transmitting function, a coil having only a receiving function, or a coil having both a transmitting and receiving function.

[0021] The receiving circuitry 110 detects magnetic resonance signals output from the receiving coil 109 and generates magnetic resonance data based on the detected magnetic resonance signals. Specifically, the receiving circuitry 110 generates magnetic resonance data by digitally converting the magnetic resonance signals output from the receiving coil 109. The receiving circuitry 110 also transmits the generated magnetic resonance data to the sequence control unit 120. The receiving circuitry 110 may be provided on the gantry side that includes the static magnetic field magnet 101, the gradient magnetic field coil 103, etc.

[0022] The sequence control unit 120 performs imaging of the subject P by driving the gradient magnetic field power supply 104, the transmission circuitry 108, and the reception circuitry 110 based on sequence information transmitted from the image processing device 130. Here, the sequence information is information that defines a procedure for performing imaging. The sequence information defines the strength of the current that the gradient magnetic field power supply 104 supplies to the gradient magnetic field coil 103 and the timing of supplying the current, the strength of the RF pulse that the transmission circuitry 108 supplies to the transmission coil 107 and the timing of applying the RF pulse, and the timing of detecting a magnetic resonance signal by the reception circuitry 110. For example, the sequence control unit 120 is an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array), or an electronic circuit such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit).

[0023] The image processing device 130 corresponds to the brain medical image processing device of the present invention, and performs processing such as image generation. In addition, the image processing device 130 performs overall control of the magnetic resonance imaging apparatus 100. The image processing device 130 includes a memory 132, an input device 134, a display 135, and a processing circuit 150. The processing circuit 150 includes a control unit 150a, a generation unit 150b, a display control unit 150c, a calculation unit 150d, and a determination unit 150e.

[0024] The control unit 150a, the generation unit 150b, the display control unit 150c, the calculation unit 150d, and the determination unit 150e are realized by a processing circuit that executes the corresponding functions. That is, the control unit 150a, the generation unit 150b, the display control unit 150c, the calculation unit 150d, and the determination unit 150e are realized by the processing circuit 150 having a processor that reads out from the memory 132 a program having the function corresponding to each unit and executes the program.

[0025] Here, the term "processor" refers to, for example, a CPU (Central Processing Unit), a GPU (Graphical Processing Unit), an Application Specific Integrated Circuit (ASIC), a programmable logic device (for example, a Simple Programmable Logic Device (SPLD), a Complex Programmable Logic Device (CPLD), and a Field Programmable Gate Array (FPGA)).

[0026] The control unit 150a transmits sequence information to the sequence control unit 120 and receives magnetic resonance data from the sequence control unit 120. The control unit 150a also stores the received magnetic resonance data in the memory 132.

[0027] The control unit 150a arranges the magnetic resonance data stored in the memory 132 in the k-space. As a result, the memory 132 stores the k-space data.

[0028] The generator 150b generates a magnetic resonance image by performing image reconstruction on the magnetic resonance data acquired from the sequence controller 120. The generator 150b also reads out k-space data from the memory 132 and performs reconstruction processing such as Fourier transform on the read out k-space data to generate a magnetic resonance image.

[0029] The functions of the display control unit 150c, the calculation unit 150d, and the determination unit 150e will be described later.

[0030] The memory 132 stores magnetic resonance data received by the control unit 150 a, k-space data arranged in k-space, image data generated by the generation unit 150 b, etc. For example, the memory 132 is a semiconductor memory element such as a RAM (Random Access Memory), a flash memory, a hard disk, an optical disk, etc.

[0031] The input device 134 accepts various instructions and information input from an operator. The input device 134 is, for example, a pointing device such as a mouse or a trackball, a selection device such as a mode switch, or an input device such as a keyboard. The display 135, under the control of the control unit 150a, displays a GUI (Graphical User Interface) for accepting input of imaging conditions, images generated by the generation unit 150b, etc. The display 135 is, for example, a display device such as a liquid crystal display.

[0032] The control unit 150a performs overall control of the magnetic resonance imaging apparatus 100, and controls imaging, image generation, image display, etc. For example, the control unit 150a accepts input of imaging conditions (imaging parameters, etc.) on a GUI and generates sequence information according to the accepted imaging conditions. The control unit 150a also transmits the generated sequence information to the sequence control unit 120.

[0033] Next, the background of the embodiment will be described.

[0034] Computer-aided diagnosis (CAD) is a well-known technology that uses computers to assist in diagnosis. CAD can be broadly divided into two categories: computer-aided detection (CADe) and computer-aided diagnosis (CADx). Computer-aided detection (CADe) refers to standalone software (medical device programs) or devices (programmed medical devices) that automatically detect and mark suspected lesions on images. Computer-aided diagnosis (CADx) refers to standalone software or devices that incorporate such software. These functions not only detect suspected lesions but also output quantitative data, such as the differentiation of benign or malignant lesions, the progression of disease, and the probability of malignancy, as numerical values ​​or graphs. CAD is currently widely used in the diagnosis of solid tumors, such as lung cancer and breast cancer, and many products are already in clinical use.

[0035] Amyloid PET imaging detects the accumulation of amyloid β protein, the causative protein of Alzheimer's disease (AD), and is the gold standard for diagnosing Alzheimer's disease. In amyloid PET imaging, the positive / negative status of amyloid β protein accumulation in the brain is determined by visual evaluation of the PET image. In this case, the image is interpreted while paying attention to the amount of amyloid β protein at the preferred accumulation site where amyloid β protein accumulation is easily observed.

[0036] Next, we will explain amyloid MRI as an alternative method to amyloid PET, that is, the search for a method for calculating the amount of amyloid β protein using MRI.

[0037] We analyzed histograms of voxel slope values ​​in specific brain regions on MRI phase slope maps from 48 patients with mild cognitive impairment (MCI) to dementia. We found that the histogram shapes differed between amyloid-β protein-positive data (e.g., curve 1, shown by the solid line) and amyloid-β protein-negative data (e.g., curve 2, shown by the dotted line) as shown in Figure 2. This likely reflects differences in the spatial distribution of amyloid-β protein, which is iron deposits, in the brain. For example, as shown in Figure 3, the histogram shape of amyloid-MRI phase slope map values ​​from amyloid-β protein-positive individuals resembles curve 1, whereas the histogram shape of amyloid-β protein-negative individuals resembles curve 2. Statistics such as mean m, standard deviation σ, and entropy e also differ between curves 1 and 2.

[0038] Therefore, we focused on texture (appearance), a quantity that quantitatively represents the spatial distribution pattern, structure, roughness, and fineness, and devised a quantitative index that reflects these differences. This quantitative index, which indicates the amount of amyloid beta protein distributed in a specific functional region, is named the Paramagnetic Accumulation Index (PAI) because the iron component (paramagnetic substance) that accumulates in amyloid beta protein is the object directly observed by amyloid MRI, and this index represents its accumulation. As an example, if the mean value of the shape of the histogram of the MRI phase slope map is m, the standard deviation is σ, and the entropy is e, then m / e / σ is a quantitative index indicating the amount of amyloid beta protein distributed in a specific functional region. PAI is an index that characterizes the amyloid beta protein-like distribution in a certain region. Details of this index will be described later.

[0039] Next, we will explain the spatiotemporal distribution of amyloid β protein accumulation. As mentioned above, in the early stages of Alzheimer's disease, analyzing PAI solely through localized analysis may be insufficient, and it may be necessary to observe the spatiotemporal changes in amyloid β protein accumulation. Here, amyloid β protein is known to spread to other functional areas over time, as shown in Figures 4 and 5, for example. This spreading phenomenon is called propagation. As an example, consider a case where, at time t1, amyloid β protein is present only in the parietal lobe, which is functional area 3a, a preferred site of amyloid β protein accumulation. In this case, for example, at time t2, amyloid β protein is present in the parietal lobe, which is functional area 3a, and the posterior cingulate cortex / precuneus, which is functional area 3b. Furthermore, for example, at time t3, amyloid β protein spreads to the parietal lobe, which is functional area 3a, the posterior cingulate cortex / precuneus, which is functional area 3b, and the frontal lobe, which is functional area 3c, and at time t4, amyloid β protein spreads to the parietal lobe, which is functional area 3a, the posterior cingulate cortex / precuneus, which is functional area 3b, the frontal lobe, which is functional area 3c, and the temporal lobe, which is functional area 3d. In this way, by making a judgment taking into account the spatiotemporal distribution of amyloid β protein accumulation, improved accuracy of the judgment can be expected.

[0040] It should be noted that functional area 1 is not defined as the parietal lobe, which is a specific functional area, as in this example.

[0041] In the above example, functional area 3a, which is "functional area 1," is the parietal lobe, but it may also be the posterior cingulate cortex / precuneus, the frontal lobe, the temporal lobe, or another functional area. That is, the functional area that first detects amyloid beta protein is designated "functional area 1," the functional area that secondly detects amyloid beta protein by propagation is designated "functional area 2," the functional area that thirdly detects amyloid beta protein by propagation is designated "functional area 3," and so on. The above content corresponds to the "likelihood of the order of propagation" described in the claims. The number of target functional areas may be any number greater than or equal to two.

[0042] In other words, the image processing device 130 of the embodiment calculates an amyloid beta protein index, which is an index related to the amount of substance distributed in each of multiple functional areas of the brain, and makes a comprehensive judgment based on the index calculated for each functional area, taking into account the spatiotemporal distribution of amyloid beta protein accumulation, to determine whether or not there is an amount (mainly the amount of accumulation) of a specified substance related to a lesion in the brain.

[0043] Specifically, the image processing device 130 according to the embodiment includes a calculation unit 150d and a determination unit 150e. The calculation unit 150d calculates, for each of a plurality of functional regions of the brain, an index related to a predetermined substance distributed in the functional region. The determination unit 150e determines the presence or absence of a predetermined substance related to a lesion in the brain based on the index calculated for each functional region.

[0044] The magnetic resonance imaging apparatus 100 according to the embodiment also includes a sequence control unit 120, a calculation unit 150d, and a determination unit 150e. The sequence control unit 120 executes a pulse sequence to collect data from an imaging target. The calculation unit 150d calculates, for each of a plurality of functional regions of the brain, an index related to the amount of a predetermined substance distributed in the functional region based on the collected data. The determination unit 150e determines the presence or absence of a predetermined substance related to a lesion in the brain based on the index calculated for each functional region.

[0045] The processing performed by the image processing device 130 according to the embodiment will be described in detail below with reference to Figures 6 to 12. Figure 6 is a flowchart showing the flow of processing performed by the image processing device 130 according to the embodiment. Figure 7 is a flowchart illustrating in more detail the processing in step S100 in Figure 6. Figure 8 is a flowchart illustrating in more detail the processing in steps S100 and S120 in Figure 7. Figure 9 is a flowchart illustrating in more detail the processing in steps S150 and S160 in Figure 7.

[0046] Referring to FIG. 6, first, in step S100, the calculation unit 150d calculates, for each of a plurality of functional areas of the brain, an index related to the amount of a predetermined substance distributed in the functional area. Here, the predetermined substance is, for example, amyloid β protein. FIG. 7 shows details of the processing of step S100 in FIG. 6. That is, the processing of steps S110 to S160 in FIG. 7 corresponds to the processing of step S100 in FIG. 6. The processing of step S100 will be described in detail below using FIG. 7. Note that the processing of steps S110 and S120 is shown in more detail in FIG. 8, and the processing of steps S150 and S160 is shown in more detail in FIG. 9.

[0047] The above embodiment has been described with reference to a case where an index relating to the amount of a predetermined substance distributed in a functional region is calculated using data obtained by the magnetic resonance imaging apparatus 100. However, the embodiment is not limited to this, and is also applicable to images obtained using an apparatus other than a magnetic resonance imaging apparatus, such as CT, PET, or SPECT, or images obtained using a magnetic resonance imaging apparatus when the magnetic resonance imaging apparatus is capable of imaging a compound introduced from outside the body that specifically accumulates in a lesion.

[0048] First, in step S110, the sequence controller 120 executes a pulse sequence to collect data from an imaging target. Here, the pulse sequence executed by the sequence controller 120 in step S110 is, for example, a sequence that generates multiple echoes by a single excitation using a radio-frequency magnetic field. As an example, as shown in step S111 of FIG. 9 , the sequence controller 120 executes a pulse sequence of the Multi-GRE method, which is a pulse sequence of the GRE (Gradient Echo) method, to perform T2*-weighted imaging. The sequence controller 120, for example, collects data of multiple echoes with different echo times.

[0049] Subsequently, in step S120, the calculation unit 150d calculates a first map indicating the magnitude of a quantity corresponding to the volume magnetic susceptibility based on the magnetic resonance data collected in step S110. Details of the process of step S120 are shown in FIG.

[0050] In step S121, the calculation unit 150d performs a phase unwrapping process on the plurality of T2* phase images obtained by the pulse sequence executed in step S111.

[0051] Next, in step S122, the calculation unit 150d applies a high-pass filter to each phase image that has undergone the phase unwrapping process in step S121. This makes it possible to remove inhomogeneity in the static magnetic field caused by biological tissue in the static magnetic field magnet. If the inhomogeneity of the static magnetic field is sufficiently guaranteed, this step may be omitted.

[0052] Subsequently, in step S123, in order to reduce phase noise, the calculation unit 150d applies an adaptive noise filter such as a Wiener filter or a Kalman filter to the phase image for each echo to which the high-pass filter has been applied in step S122.

[0053] Subsequently, in step S124, the calculation unit 150d calculates a value (Δθ / ΔTE) by dividing the value of the phase change (Δθ) between each echo by the value of the change in TE (ΔTE) between each echo. As an example, in the case of four echoes in the Multi-GRE method, the calculation unit 150d calculates, for each voxel of the phase image, phase values ​​for four different echo times TE as points P, Q, R, and S, and the calculation unit 150d calculates a straight line that best fits the data points of points P, Q, R, and S using, for example, the least squares method, and calculates the phase gradient (Δθ / ΔTE) from the slope of the straight line.

[0054] Here, if γ represents the gyromagnetic ratio, B 0 represents the magnitude of the static magnetic field, and χ represents the volume magnetic susceptibility, the following equation (1) holds:

[0055]

[0056] Here, the phase gradient (Δθ / ΔTE) with respect to TE is a physical quantity with a unit of rad / sec and a dimension of angular frequency. From equation (1), the phase gradient (Δθ / ΔTE) with respect to TE is proportional to the volume magnetic susceptibility χ, and therefore the phase gradient (Δθ / ΔTE) with respect to TE can be interpreted as a quantity that roughly represents the volume magnetic susceptibility χ.

[0057] Subsequently, in step S125, the calculation unit 150d performs a predetermined linear transformation on the phase gradient (Δθ / ΔTE) with respect to TE calculated in step S124.

[0058] In step S126, the calculation unit 150d generates an echo intensity image in the first imaging coordinate system based on the data obtained from the pulse sequence executed in step S110. Subsequently, in step S127, the sequence control unit 120 performs T1-weighted imaging in a second imaging coordinate system different from the first imaging coordinate system in which imaging was executed in step S110.

[0059] Subsequently, in step S128, the calculation unit 150d generates a T1-weighted image after deformation and registration based on the T1-weighted imaging performed in the second imaging coordinate system in step S127. This deformation and registration makes the T1-weighted image equivalent to an image captured in the first imaging coordinate system. The T1-weighted image generated in step S128 is used to generate a vascular component image mask in step S255, generate a CSF (Cerebrospinal Fluid) mask in step S129, create a standard parcel atlas in step S161, and so on.

[0060] In step S129, the calculation unit 150d generates a vascular component image mask based on the phase image that has been subjected to the unwrapping process generated in step S121 and the T1-weighted image that has been subjected to the process of step S128. In addition, in step S130, the calculation unit 150d generates a CSF mask based on the T1-weighted image that has been subjected to the process of step S128.

[0061] In step S131, the calculation unit 150d generates a slope map, which is a first map, based on the phase gradient (Δθ / ΔTE) map after linear transformation in step S125, the blood vessel component image mask generated in step S129, and the CSF mask generated in step S130.

[0062] That is, the calculation unit 150d removes the vascular components and CSF components from the phase gradient (Δθ / ΔTE) map after linear transformation using the vascular component image mask and CSF mask, and creates a first map as a map of the brain parenchyma that is the target of feature analysis in step S150.

[0063] As described above, in step S120, the calculator 150d creates a phase gradient (Δθ / ΔTE) map based on the data collected in step S110, and calculates a first map indicating the magnitude of a quantity corresponding to the volume magnetic susceptibility of only the brain parenchyma by removing the vascular components and CSF components. Note that the embodiment is not limited to the case where the first map indicating the magnitude of a quantity corresponding to the volume magnetic susceptibility is created by creating a phase gradient (Δθ / ΔTE) map. The calculator 150d may also create a first map indicating the magnitude of a quantity corresponding to the volume magnetic susceptibility of only the brain parenchyma by removing the vascular components and CSF components from an image obtained by a QSM (Quantitative Susceptibility Map) method, a SWI (Susceptibility Weighted Imaging) method, or the like as the phase gradient (Δθ / ΔTE) map.

[0064] 7, in step S150, the calculation unit 150d calculates a second map indicating feature quantities of the signal values ​​in the first map based on the first map. FIG. 9 shows the details of steps S150 and S160.

[0065] 9, in step S153, the calculation unit 150d calculates texture features based on the first map calculated in step S131, and generates a second map representing the texture features. Here, the texture features may be, for example, the mean value m and variance σ of multiple voxel values ​​included in a specific three-dimensional structure of the first map. 2 , standard deviation σ, entropy e, Z-value, skewness or kurtosis, etc. Here, entropy e is, for example, Shannon's information entropy. Furthermore, the Z-value is calculated using a population of normal individuals, with its mean value μ and standard deviation σ n Then, Z = (m - μ) / σ n means the value given by

[0066] Furthermore, a particularly promising texture feature may be m / e / σ, which is a combination of texture values. Here, m, e, and σ represent the average value, entropy, and standard deviation of multiple voxel values ​​included in a specific three-dimensional structure of the first map, respectively. That is, the calculation unit 150d may employ a texture feature, such as m / e / σ, whose value increases as m increases, e decreases, and σ decreases, as an index representing the amount of amyloid beta protein.

[0067] If the texture feature in step S153 includes entropy, kurtosis, skewness, etc., the process in step S150 proceeds to step S151. In step S151, the calculation unit 150d calculates a histogram of the first map (hereinafter referred to as the "slope map") from multiple voxel values ​​included in a specific three-dimensional region of the first map. Subsequently, in step S152, the calculation unit 150d calculates statistics of the first map, which is the slope map, based on the histogram calculated in step S151. Based on this, the calculation unit 150d calculates texture feature in step S153. That is, if the texture feature includes entropy e, kurtosis, skewness, etc., the feature calculated in step S153 is calculated from the histogram of voxels created in step S151, excluding masked blood vessel or CSF regions.

[0068] In this way, in step S150, the calculation unit 150d calculates the second map representing texture feature amounts such as the mean, standard deviation, and entropy, and combination feature amounts of texture feature amounts such as m / e / σ.

[0069] Subsequently, in step S160, the calculation unit 150d calculates a third map by aggregating the second map for each functional unit, that is, functional region, of the imaging target, known as a parcel, for example.

[0070] Here, the standard parcel atlas will be described. A parcel refers to a functional unit of an imaging target, and an atlas refers to a map that represents location information. When the imaging target is the brain, for example, there is a standard parcel atlas called AAL (Automated Anatomical Labeling), which is a global standard that divides brain tissue into 116 functional regions. In addition to AAL, there are standard parcel atlases for gray matter and white matter. In step S161, the calculation unit 150d selects at least one standard parcel atlas according to the purpose of analysis. Furthermore, as necessary, characteristic parcels are selected from each standard parcel atlas. The calculation unit 150d acquires each parcel, i.e., functional unit, in association with its location information.

[0071] Next, in step S162, the calculation unit 150d performs deformation and registration of the standard parcel atlas selected in step S161 on the T1-weighted image after deformation and registration generated in step S128, and creates a parcel region for analysis and display.

[0072] Next, in step S163, the calculation unit 150d calculates a third map by aggregating the second map for each parcel based on the second map generated in step S153 and the parcel region for analysis and display created in step S162, as an index representing the amount of amyloid beta protein distributed in the functional region. Thus, in step S160, the calculation unit 150d calculates a third map by aggregating the second map for each functional unit of the imaging target. As an example, the calculation unit 150d calculates an average value of the values ​​of the second map, which are texture features, for each functional unit of the imaging target, and sets the calculated average value as the value in the third map for all voxels in that functional unit. For example, if m / e / σ is selected as the texture feature in the second map in step S153, the calculation unit 150d calculates the value of m / e / σ, which is a combination of texture features, for each parcel that is a functional region, and sets the calculated value as the value of the third map for that functional region. In other words, the calculation unit 150d generates the third map so that, for example, all voxel values ​​belonging to the same parcel are the same.

[0073] 10 visualizes the indices representing the amounts of amyloid β protein distributed in the functional regions calculated in this manner in step S100. In Fig. 10, for example, regions 41, 42, and 43 belong to different functional regions, and the indices representing the amounts of amyloid β protein calculated for each functional region are displayed in different colors depending on the value of the indices.

[0074] Next, the selection of texture features in step S153 will be explained again. In order to consider what texture features should be adopted, a first map was created for 48 cases, and slope map analysis was performed. These 48 cases were classified into cases without Aβ protein (PET negative) and cases with Aβ protein (PET positive) based on the image findings using amyloid PET images. As a result, 18 cases were PET negative and 30 cases were PET positive.

[0075] In this study, the DK atlas, which is a type of standard parcel atlas for the cortex, was adopted in step S161. The number of parcels in the DK atlas is 70 (35 for the right brain and 35 for the left brain).

[0076] In addition, the mean, standard deviation, and entropy were adopted as texture feature quantities to be calculated in step S153. These texture feature quantities were calculated in step S152 from the histogram of all voxels calculated in step S151, excluding the blood vessels and CSF regions masked by tissue within the parcel.

[0077] Regarding the texture feature quantities of the mean, standard deviation, and entropy described above, we analyzed how the texture feature quantities change between PET-negative cases and PET-positive cases.

[0078] When the mean slope map m was used as a texture feature, a comparison of the values ​​between the two groups showed that the PET-negative cases were lower than the PET-positive cases, and the mean value of the slope map histogram tended to be higher on the PET-positive side. This is thought to be due to an increase in the paramagnetic component caused by an increase in the iron component in the PET-positive cases.

[0079] Furthermore, when the standard deviation σ of the slope map was used as the texture feature, a comparison of the values ​​between the two groups showed that the PET-negative cases were greater than the PET-positive cases, and the spread of the slope map histogram tended to be lower on the PET-positive side. This is thought to be because, in the absence of proteins such as Aβ to which iron is bound, the slope map values ​​are approximately normally distributed, but as the amount of iron-bound proteins increases, the number of voxels with corresponding slope map values ​​increases, and the area near the peak of the normal distribution of the histogram of slope map values ​​increases, causing the standard deviation representing the spread of the normal distribution to become relatively lower.

[0080] Furthermore, when the entropy e of the slope map was used as the texture feature, a comparison of the values ​​between the two groups showed that the PET-negative cases were higher than the PET-positive cases. In other words, a tendency was observed for the entropy of the slope map histogram to be lower on the PET-positive side. This is thought to be because, in the absence of highly specific proteins such as amyloid β protein bound to iron components, the spatial distribution of slope values ​​is highly disordered. However, as the amount of proteins such as amyloid β protein bound to iron components increases, the number of voxels with corresponding slope values ​​increases, and the specific granularity increases, thereby reducing the disorder of the spatial distribution of slope values.

[0081] Based on the above considerations, for example, when the calculation unit 150d uses m / e / σ, obtained by dividing the mean m, which increases in PET-positive cases, by the entropy e and standard deviation σ, which decrease in PET-positive cases, as the texture feature in step S153, the PET-negative cases are less than the PET-positive cases. In other words, it can be seen that using the value obtained by dividing the mean m of the voxel values ​​in the three-dimensional structure of the first map by the entropy e and then by the standard deviation σ, is a promising option. A high value of this feature can be considered to indicate a high level of amyloid-β protein accumulation. Furthermore, apart from the accumulation of amyloid-β protein, pathological increases in iron components can occur with aging. In such cases, similar to the interpretation of amyloid-β protein, it can be considered that the accumulation of age-related iron, a causative substance of specific age-related pathologies, has increased.

[0082] To summarize the processing of step S100, in step S100, the calculation unit 150d calculates a PAI (Paramagnetic Accumulation Index), which is an index related to the amount of amyloid beta protein, a predetermined substance distributed in each of multiple functional areas of the brain, for each of the functional areas.

[0083] More specifically, in step S110, the sequence control circuit 120 executes a pulse sequence that generates multiple echoes to collect data from the imaging target. In step S120, the calculation unit 150d generates a first map based on the data collected in step S110, by dividing the phase change value between the multiple echoes by the change value of TE between the echoes. The first map indicates the magnitude of a quantity corresponding to volume magnetic susceptibility and is also called a slope map. The first map indicates a quantity related to the magnitude of a predetermined substance, for example, amyloid beta protein, present in the voxel.

[0084] Next, in step S150, the calculation unit 150d calculates a second map indicating feature quantities of the signal values ​​of the first map based on the first map. The second map is also called a texture feature. As the texture feature, m / e / σ is selected, where m is the mean of the first map, σ is the standard deviation, and e is the entropy. Next, in step S160, the calculation unit 150d generates a third map by aggregating the second map for each functional region, as an index (PAI) indicating the amount of amyloid beta protein distributed in the functional region. That is, the index is, for example, a value (m / e / σ) obtained by dividing the mean value m of the voxel values ​​of the first map by the entropy e of the voxel values ​​of the first map, divided by the standard deviation σ of the voxel values ​​of the first map.

[0085] In the above example, the m / e / σ of the first map (slope map) is used as the index (PAI) indicating the amount of amyloid β protein distributed in the functional region, but the embodiment is not limited to this. For example, the calculation unit 150d may use a feature quantity such as m, m / e, or m / σ of the first map as the index (PAI) indicating the amount of amyloid β protein distributed in the functional region.

[0086] As another example, the calculation unit 150d may further perform an inverse scaling operation on the above-mentioned value (m / e / σ). In other words, the index calculated in step S100 is a value obtained by multiplying the average value m of the voxel values ​​of the first map by the value obtained after the inverse scaling operation of the entropy of the voxel values ​​of the first map, and by multiplying the standard deviation of the voxel values ​​of the first map by the value obtained after the inverse scaling operation.

[0087] Here, the meaning of performing the inverse scaling operation will be explained. The above-mentioned value m / e / σ can be interpreted as the mean value m being the main parameter, multiplied by 1 / e and 1 / σ as correction terms. However, since these correction terms are functions in the form of y = 1 / x, the coefficients 1 / Δe and 1 / Δσ become nonlinear with respect to the changes Δe and Δσ in e and σ, respectively, which tends to increase the error. Furthermore, the value of m / e / σ may fluctuate greatly with a slight change Δe and Δσ. Therefore, as shown in FIG. 11A, the calculation unit 150d calculates 1 / e ｍａｘ (1 / σ ｍａｘ ) to 0, 1 / e ｍｉｎ (1 / σ ｍｉｎ ) to 1, and an index (PAI) indicating the amount of amyloid β protein distributed in the functional region may be calculated. mａｘ (σ mａｘ ) and e mｉｎ (σ min ) are the maximum and minimum values ​​that the entropy e (standard deviation σ) can take. This inverse scaling transformation can improve the accuracy of the index that indicates the amount of amyloid β protein distributed in the functional region.

[0088] In the above inverse scaling transformation, as shown in FIG. 11B, e mａｘ (σ mａｘ ) and e mｉｎ (σ min ) is set to the maximum and minimum values ​​of the entropy e (standard deviation σ) by setting margins a and b, respectively. mａｘ (σ mａｘ ) + a, e mｉｎ (σ min)-b. Inverse scaling conversion is not limited to the above. For example, as shown in FIG. 11C, ｍａｘ (1 / σ ｍａｘ ) to 1, 1 / e ｍｉｎ (1 / σ ｍｉｎ ) to 2.

[0089] Furthermore, as mentioned above, in the examples described so far, the embodiments have been described as cases in which an index related to the amount of a predetermined substance distributed in a functional region is calculated using data obtained by the magnetic resonance imaging device 100, but the embodiments are not limited to this, and the embodiments are also applicable to images obtained using devices other than magnetic resonance imaging devices, such as CT, PET, SPECT, FDG-PET, or images obtained using a magnetic resonance imaging device if the magnetic resonance imaging device is capable of imaging a compound that specifically accumulates in a lesion introduced from outside the body.

[0090] For example, in step S100, the calculation unit 150d may calculate, for each of a plurality of functional regions of the brain, an index (PAI) indicating the amount of amyloid beta protein distributed in the functional region based on data obtained by amyloid PET. As an example, in step S100, the calculation unit 150d calculates, as a first map, SUVR (Standardized Uptake Value Radius) data obtained by amyloid PET imaging divided (or normalized) by the SUV of the cerebellum. Since it is known that there is almost no accumulation of amyloid beta protein in the cerebellum in mild cognitive impairment (MCI) and early dementia of Alzheimer's disease, the SUV of the cerebellum is used as a coefficient for normalization. The calculation unit 150d may calculate, for example, m, m / e, m / σ, m / e / σ, etc., using m as the mean value, e as the entropy, and σ as the standard deviation in the histogram of the first map, and may calculate the calculated values ​​as an index (PAI) representing the amount of amyloid beta protein distributed in the functional region.

[0091] In calculating the index (PAI) that indicates the amount of amyloid β protein distributed in functional areas using the slope map of this patent, since it is known that there is almost no accumulation of amyloid β protein in the cerebellum in mild cognitive impairment of Alzheimer's disease or early dementia, the index value (PAI) in the cerebellum is calculated. ｃｅｒｅｂｅｌｌｉｕｍ ) may be used for normalization. That is, the index calculated in step S100 may be an index normalized using the value of the index in the cerebellum. As an example, the calculation unit 150d may normalize the PAI * = PAI / PAI ｃｅｒｅｂｅｌｌｉｕｍ Normalized by the normalized PAI * may be calculated as an index showing the amount of amyloid β protein distributed in the functional region. * = PAI-PAI ｃｅｒｅｂｅｌｌｉｕｍ Normalized by the normalized PAI * may be calculated as an index showing the amount of amyloid β protein distributed in the functional region.

[0092] Returning to FIG. 6, in steps S200 to S400, the determining unit 150e determines the presence or absence of a predetermined substance associated with a lesion in the brain based on the index calculated for each functional region in step S100.

[0093] This processing will be described below with reference to Fig. 12. Fig. 12 is a diagram illustrating a method for determining the presence or absence of a predetermined substance associated with a brain lesion, in this case the predetermined substance being amyloid β protein, based on the index calculated in step S100.

[0094] In FIG. 12 , each row of data indicates a patient number. Column 4 shows the results of the determination of the presence or absence of amyloid beta protein in Alzheimer's disease based on amyloid PET images. Here, "p" indicates positive, i.e., the presence of amyloid beta protein, and "n" indicates negative, i.e., the absence of amyloid beta protein. In contrast, column 11 shows the results of the determination of the presence or absence of amyloid beta protein in Alzheimer's disease by MRI using the method according to the embodiment. The validity of the method according to the embodiment can be verified by comparing the determination results using the method according to the embodiment with the results of the amyloid PET image interpretation.

[0095] Columns 5a, 5b, 5c, and 5d show the values ​​of the paramagnetic accumulation index (PAI) representing the amount of amyloid beta protein calculated in step S100 for each patient in different functional regions. As an example, columns 5a, 5b, 5c, and 5d show the values ​​of the index representing the amount of amyloid beta protein calculated in step S200 for the parietal lobe, posterior cingulate cortex / precuneus, parietal lobe, and temporal lobe, respectively. Cutoff values ​​6a, 6b, 6c, and 6d correspond to thresholds that serve as references for calculating the first determination value in each functional region. As an example, cutoff values ​​6a, 6b, 6c, and 6d serve as reference cutoff values ​​for calculating the first determination value in the parietal lobe, posterior cingulate cortex / precuneus, parietal lobe, and temporal lobe, respectively. These cutoff values ​​are obtained as optimal cutoff values ​​by performing ROC (Receiver Operating Characteristic) analysis using the logistic method, which is applied to a diagnostic method for binary determination such as negative / positive.

[0096] Furthermore, columns 7a, 7b, 7c, and 7d show values ​​obtained by multiplying the first judgment value of each patient in each functional area by the weight corresponding to each functional area, which will be described later. As an example, columns 7a, 7b, 7c, and 7d show values ​​obtained by multiplying the first judgment value in the parietal lobe, posterior cingulate cortex / precuneus, parietal lobe, and temporal lobe by the weight corresponding to each functional area. In step S200, the judgment unit 150e calculates a first judgment value based on the index calculated for each functional area. Specifically, if the index calculated for each functional area is greater than the cutoff value calculated for that functional area, the judgment unit 150e determines the first judgment value for that functional area to be 1. If the index calculated for that functional area is equal to or less than the cutoff value calculated for that functional area, the judgment unit 150e determines the first judgment value for that functional area to be 0. As an example, in patient 1, the values ​​of the index representing the amount of amyloid β protein in functional areas 1, 2, and 4 are equal to or less than the cutoff values ​​6a, 6b, and 6d for those functional areas, so the first judgment value for patient 1 is 0 in those functional areas. On the other hand, in patient 1, the value of the index representing the amount of amyloid β protein in functional area 3 is greater than the cutoff value 6c for functional area 3, so the first judgment value for patient 1 is 1 in that functional area. Furthermore, in patient 7, the values ​​of the index representing the amount of amyloid β protein in each functional area are greater than the cutoff values ​​6a to 6d for that functional area, so the first judgment value for patient 7 is 1 in all functional areas.

[0097] Cells 8a, 8b, 8c, and 8d represent weights assigned to each functional region. In other words, the larger the weight, the greater the contribution of that functional region to the Alzheimer's disease assessment result. The weights assigned to each functional region are determined based on the likelihood of the order of propagation of a specific substance, such as amyloid beta protein, within the brain over time. For example, the weights assigned to each functional region are set so that the more upstream the functional region is in the propagation of the specific substance, the greater the weight, and the more downstream the functional region is in the propagation of the specific substance. For example, the weights assigned to the parietal lobe, posterior superior cortex / precuneus, parietal lobe, and temporal lobe are "4," "3," "2," and "1," respectively. In this way, by assigning a large weight to the functional region upstream of the propagation of amyloid beta protein and a certain weight to the functional region downstream of the propagation, the accuracy of the assessment result can be maintained.

[0098] Next, in step S300, the determination unit 150e calculates a second determination value by weighting and adding the first determination value using the weights determined for each functional area. First, the determination unit 150e calculates a value for each functional area by multiplying the first determination value by the weight of the functional area. As an example, for cells 9a, 9b, and 9d, the first determination value for the functional area for patient 1 is 0, so the value of these cells is "0." On the other hand, for cell 9c, the first determination value for functional area 3c is 1, so when this is multiplied by the weight of the functional area, "2," the value of the cell becomes "2." Similarly, for cells 12a, 12b, 12c, and 12d, the first determination value for these functional areas for patient 7 is 1, so when this is multiplied by the weights of the functional areas, "4," "3," "2," and "1," the values ​​of these cells become "4," "3," "2," and "1," respectively.

[0099] Next, the determination unit 150e sums the products of these first determination values ​​and weights. Column 10 is a column showing these sums, and illustrates the second determination value. For example, for patient 1, the value of cell 9e indicating the second determination value is the sum of cells 9a, 9b, 9c, and 9d, so the second determination value is "2." Similarly, for patient 7, the value of cell 12e indicating the second determination value is the sum of cells 12a, 12b, 12c, and 12d, so the second determination value is "10."

[0100] Next, in step S400, the determination unit 150e determines whether or not amyloid β protein is present in the brain based on the second determination value. For example, in step S400, the determination unit 150e determines that amyloid β protein is positive if the second determination value exceeds a predetermined threshold, and determines that amyloid β protein is negative if the second determination value is equal to or less than the predetermined threshold. Column 11 shows the determination results, with "p" indicating that amyloid β protein is positive and "n" indicating that amyloid β protein is negative. For example, as shown in cell 9f, for patient 1, the second determination value "2" shown in cell 9e is equal to or less than the overall determination threshold "5," so the overall determination is "n." On the other hand, as shown in cell 12f, for patient 7, the second determination value "10" shown in cell 12e is greater than the overall determination threshold "5," so the overall determination is "p."

[0101] We studied 48 patients with mild cognitive impairment (MCI) and dementia stages, and calculated an index (PAI) that represents the amount of amyloid beta protein based on MRI slope maps for each of four functional regions where amyloid beta protein is likely to accumulate: the parietal lobe, the posterior superior cortex / precuneus, the parietal lobe, and the temporal lobe.We then determined the presence or absence of amyloid beta protein based on this index and compared it with the results of an amyloid PET reading.

[0102] In the parietal lobe, posterior superior cortex / precuneus, parietal lobe, and temporal lobe, the index values ​​for the amyloid-beta-negative group were 12.701, 8.236, 6.344, and 6.337, respectively, while the index values ​​for the amyloid-beta-positive group were 26.432, 16.328, 11.394, and 11.372, respectively. The index values ​​for the amyloid-beta-positive group were approximately 1.8 to 2.1 times higher than those for the amyloid-beta-negative group, demonstrating a significant difference between the two groups (p-values ​​for the t-test were 0.0004 to 0.02). The cutoff values ​​for distinguishing between positive and negative scores in each functional area were 19.64, 10.37, 6.27, and 7.96, respectively. When the interpretation result determined by amyloid PET is taken as the true value, the determination method of this embodiment had a sensitivity of 83.3%, a specificity of 88.9%, a positive predictive value of 92.6%, a negative predictive value of 76.2%, and an accuracy of 85.4%. Considering that the 95% CIs were 79.3% to 100.7% for sensitivity, 38.6% to 83.6% for specificity, 65.8% to 93.0% for positive predictive value, and 57.1% to 100.1% for negative predictive value, it is suggested that the method for determining the presence or absence of amyloid beta protein using MRI according to this embodiment is promising as an alternative to the interpretation determination method using amyloid PET.

[0103] As described above, in the image processing device 130 according to the first embodiment, the calculation unit 150d calculates, for each of a plurality of functional areas of the brain, an index related to the amount of a predetermined substance distributed in the functional area, and the determination unit 150e determines the presence or absence of a predetermined substance related to a brain lesion based on the index calculated for each of the functional areas. This makes it possible to analyze, for example, the accumulation of amyloid beta protein, which changes over time and space, and to more accurately determine Alzheimer's disease.

[0104] Second Embodiment Next, a second embodiment will be described. In the second embodiment, attention is focused on the site of amyloid β-protein accumulation in the brain. FIG. 13 shows a schematic diagram of the structure of a neuron. A neuron's cell body 61, which has a dendrite 60, is connected to other neurons via an axon 63, which is covered with a myelin sheath 62. For example, FIG. 14 shows a cross section 70 in FIG. 13. The cross section 70 is composed of the axon 63, which is a gray matter region, and the myelin sheath 62, which is a white matter region and covers the neuron. The myelin sheath 62 is composed of oligodendrocytes. It has been suggested that amyloid β-protein accumulation may primarily begin in the white matter region in the early stages of Alzheimer's disease.

[0105] Here, myelin sheath 62, which is white matter, has a lipid-rich structure and is therefore diamagnetic, making it magnetically distinguishable from gray matter. For example, as shown in Figure 15, in the histogram of phase slope values, the histogram 31 of the white matter region and the histogram 30 of the gray matter region are distinguishable.

[0106] In view of this, in the second embodiment, each functional region is divided into multiple sub-functional regions, such as white matter regions and gray matter regions, and a determination is made for each sub-functional region using the same process as in the first embodiment. In other words, the calculation unit 150d divides each of the multiple functional regions into multiple sub-functional regions, such as white matter regions and gray matter regions. The calculation unit 150d calculates an index representing the amount of amyloid beta protein distributed in each functional region and sub-functional region. For example, as shown in FIG. 16 , the calculation unit 150d distinguishes between functional regions 50a, 50b, 50c, and 50d in the white matter region and functional regions 51a, 51b, 51c, and 52d in the gray matter region, and calculates an index representing the amount of amyloid beta protein for each functional region and sub-functional region (gray matter / white matter). Furthermore, the determination unit 150e determines the presence or absence of amyloid beta protein in the brain based on the indexes calculated for each functional region and sub-functional region. As an example, the determination unit 150e determines the presence or absence of amyloid beta protein in the brain by performing weighted addition on the indices calculated for the functional areas 50a, 50b, 50c, and 50d in the white matter region and the functional areas 51a, 51b, 51c, and 52d in the gray matter region, using different weights for each functional area and sub-functional area.

[0107] Note that new functional areas may be created by regrouping a plurality of sub-functional areas thus divided to optimize sensitivity, specificity, accuracy, etc. In this case, the calculation unit 150d and the determination unit 150e calculate the index and determine the presence or absence of amyloid β protein for each newly created functional area.

[0108] As described above, in the second embodiment, each functional region is divided into a plurality of sub-functional regions, such as a white matter region and a gray matter region, and a judgment is made for each sub-functional region. According to the second embodiment, for example, by dividing into sub-functional regions and making a judgment, a more accurate judgment can be made.

[0109] (Third embodiment) In the second embodiment, the determination unit 150e divides each of a plurality of functional areas into a plurality of sub-functional areas, the calculation unit 150d calculates an index (PAI) for each sub-functional area, and adds up the indexes for each of the plurality of functional areas using different cutoff thresholds and weights for each sub-functional area, and the determination unit 150c determines the presence or absence of a predetermined substance related to a lesion in the brain based on the index calculated for each functional area.

[0110] In this case, for example, as shown in Figure 17, if functional area 80 is divided into multiple sub-functional areas 81 to 84, calculation unit 150d calculates an index (PAI) for each of the sub-functional areas 81 to 84, and calculates the index (PAI) for functional area 80 by adding up the indexes using different cutoff thresholds and weights for each of the sub-functional areas 81 to 84 using a method similar to the method shown in Figure 12.

[0111] Here, when the functional region 80 is considered to be the whole brain, the determination unit 150e can determine the presence or absence of amyloid β protein by treating each of all functional regions in the DK atlas as sub-functional regions 81 to 84, etc. In the third embodiment, an example of such a case will be described.

[0112] In the third embodiment, a case will be described in which the functional region 80 is the parietal lobe, which is a preferred site of amyloid β-protein accumulation. In this case, the parietal lobe, which is the functional region 80, is composed of six sub-functional regions, namely, the superior parietal lobule, the inferior parietal lobule, and the supramarginal gyrus, located on the left and right sides of the head. As in FIG. 12 , the calculation unit 150d and the determination unit 150e perform weighted addition using different cutoff values ​​and weights for each sub-functional region to determine whether amyloid β-protein has accumulated in the parietal lobe, which is the functional region 80.

[0113] Specifically, the calculation unit 150d calculates an index (PAI) representing the amount of amyloid β protein distributed in each of six sub-functional areas constituting the parietal lobe, which is the functional area 80. Next, the determination unit 150e sets the weight of each sub-functional area to the weight determined for that sub-functional area if the index for that sub-functional area is equal to or greater than the cutoff value determined for that sub-functional area, or sets the weight to 0 if the index for that sub-functional area is less than the cutoff value determined for that sub-functional area, and sets the weighted sum of the weights for the sub-functional areas to the value of the index in the functional area 80. The determination unit 150e determines whether or not amyloid β protein has accumulated in the functional area 80 based on the values ​​of the index in the functional area 80.

[0114] Fourth Embodiment In the fourth embodiment, a case of time-series testing will be described. Here, we consider a case where multiple tests are performed, such as when tests are performed periodically. First, we consider a case where the method used in the first embodiment is used as is to determine whether or not amyloid β protein has accumulated. Such a case is shown in the upper part of Figure 18. In this case, for example, at times t = t1, t2, t3, and t4, the calculation unit 150d calculates an index representing the amount of amyloid β protein in each functional region. In this case, the amount of amyloid β protein tends to increase overall with age, indicating gradual progression of Alzheimer's disease.

[0115] As described in the first embodiment, in step S200, the judgment unit 150e calculates a first judgment value for each functional area based on the index calculated in step S100, in step S300, it calculates a second judgment value by weighting and adding the first judgment value, and in step S400, it judges the presence or absence of amyloid beta protein in the brain based on the second judgment value.

[0116] As shown in FIG. 18 , the cutoff values ​​for calculating the first judgment value are 23 for functional area 1, 18 for functional area 2, 12 for functional area 3, and 9 for functional area 4. Here, at t = t2, when the index value is slightly below 23, which is the cutoff value for functional area 1, the first judgment value for functional area 1 is 0. As a result, the first judgment value for functional area 1 is slightly underestimated, and the final judgment value is "n." In this example, the true judgment value is "p," resulting in a false negative judgment. Similarly, at t = t3, when the index value is slightly below 12, which is the cutoff value for functional area 3, the first judgment value for functional area 3 is 0. As a result, the first judgment value for functional area 3 is slightly underestimated, and the final judgment value is "n." In this example, the true judgment value is "p," resulting in a false negative judgment.

[0117] That is, in the method shown in the first embodiment in which judgment is made independently at each time, if the calculated index value is slightly below the cutoff value, it may be difficult to obtain a correct value.

[0118] Therefore, in the fourth embodiment, the calculation unit 150d calculates an index related to the amount of a predetermined substance distributed in the functional region for each of the multiple imaging operations, and the determination unit 150e makes a determination taking into account changes in the index between imaging operations. As an example, the determination unit 150e makes a determination based on the difference between the index calculated for the latest imaging operation and the index calculated for the imaging operation immediately before the latest imaging operation.

[0119] 18, for example, the determination unit 150e calculates the difference between the index calculated for the latest imaging (t=t2) and the index calculated for the imaging immediately preceding the latest imaging (t=t1) with respect to imaging at t=t2. Also, the determination unit 150e calculates the difference between the index calculated for the latest imaging (t=t3) and the index calculated for the imaging immediately preceding the latest imaging (t=t2) with respect to imaging at t=t3.

[0120] Next, the judgment unit 150e compares the calculated index difference for each functional area with the cutoff value for the index difference defined for that functional area to calculate a third judgment value, and then performs weighted addition on the third judgment value using the weight defined for that functional area to calculate a fourth judgment value indicated by "Sum (2)" in FIG. 18 . As an example, when t = t3, the differences from the index at t = t2 are "10," "10," "9," and "5" for the respective functional areas. Here, the cutoff values ​​for the index difference are "9," "7," "5," and "3" for the respective functional areas, so that the index difference is equal to or greater than the cutoff value for functional areas 1, 2, and 3, and less than the cutoff value for functional area 4. In this case, the third judgment values ​​are "1," "1," "1," and "0," respectively. Here, the weights defined for the respective functional areas are "4," "3," "2," and "1," respectively. Therefore, when these weights are multiplied, the results are "4," "3," "2," and "0," respectively. Next, the determination unit 150e adds these values ​​together to calculate a fourth determination value of "9", which is indicated by "Sum (2)" in FIG.

[0121] Next, the determination unit 150e calculates a fifth determination value by adding the second determination value indicated by "Sum (1)" to the fourth determination value indicated by "Sum (2)" in FIG. 18 . The determination unit 150e then determines the presence or absence of a predetermined substance associated with the lesion based on whether the fifth determination value is equal to or greater than the final threshold value serving as a reference. For example, at t=t3, the second determination value is "7" and the fourth determination value is "9," so these are added together to obtain the fifth determination value of "16." Since this exceeds the final threshold value of "7," the determination unit 150e determines the determination result at t=t3 to be "p." This reduces the number of false negative determinations compared to the upper row of FIG. 17 .

[0122] As described above, in the fourth embodiment, the determination unit 150e performs imaging while taking into account changes in the indices between imaging sessions, which can reduce false negative results, for example, and can provide a determination method suitable for determining the presence or absence of amyloid beta protein at an earlier stage, for example.

[0123] In the first to fourth embodiments described above, the presence or absence of amyloid β protein (mainly accumulation amount) is determined as a predetermined substance associated with Alzheimer's disease. However, the embodiments are not limited to this, and the predetermined substance associated with Alzheimer's disease may be, for example, tau protein instead of amyloid β protein. Furthermore, instead of (or in addition to) this configuration, the presence or absence of a predetermined substance associated with other brain lesions (e.g., demyelination) may be determined. Alternatively, the presence or absence of a predetermined substance associated with a lesion in a human body other than the brain may be determined based on data (e.g., images) obtained by medical devices such as CT, PET, SPECT, etc. Either configuration contributes to ultra-early detection of lesions.

[0124] In the above-described embodiment, the determination unit 150e determines the presence or absence of a predetermined substance in the brain based on the second determination value, as shown in step S400 of FIG. 6 , and the display control unit 150c displays, for example, whether the disease is positive or negative on the display 135 based on the determination result. However, the embodiment is not limited to this. In an embodiment, the determination unit 150e may not perform a determination but may simply calculate some quantitative value. As an example, the display control unit 150c may calculate a quantitative value reflecting the amount of amyloid, such as m, m / e, m / σ, or m / e / σ, where m is the mean value in a histogram, e is the entropy, and σ is the standard deviation, instead of the determination result, and display an index (PAI) representing the amount of amyloid beta protein distributed in the functional region on the display screen 90 of the display 135 shown in FIG. 19 as a determination value. In this case, in FIG. 6 , the processing of step S400 does not need to be executed, and in step S300, the judgment unit 150e calculates a judgment value regarding the presence or absence of a specified substance related to a lesion in the brain based on the indexes calculated for each of the multiple functional areas, and then the display control unit 150c displays the second judgment value calculated in step S300 on the display 135.

[0125] In the above embodiment, the process performed by the determination unit 150e is described as a case in which the determination unit 150e determines the presence or absence of a predetermined substance based on predetermined criteria, such as a predetermined mathematical formula, but the embodiment is not limited to this. For example, the determination unit 150e may make the determination using a trained model. As an example, the determination unit 150e may make an estimation using a trained model that uses machine learning such as a neural network, regression, decision tree, or gradient boosting, and determine the presence or absence of a lesion.

[0126] In this case, in the model learning stage, quantitative value data for each functional region and information on the determination result as ground truth are provided to the model. For example, a trained model with updated weight values ​​is generated by updating the weights of neurons in the neural network through back error propagation or through machine learning. In the model inference stage, when quantitative value data for each functional region is input to the trained model thus generated, the determination unit 150e outputs a determination result on the presence or absence of a predetermined substance.

[0127] According to at least one of the embodiments described above, it is possible to determine the presence or absence of amyloid β protein in Alzheimer's disease and the like.

[0128] REFERENCE SIGNS LIST 100 Magnetic resonance imaging apparatus 130 Image processing device 132 Memory 134 Input device 135 Display 150 Processing circuit 150a Control unit 150b Generation unit 150c Display control unit 150d Calculation unit 150e Determination unit

Claims

1. A brain medical image processing device comprising: a calculation unit that calculates, for each of a plurality of functional areas of the brain, an index related to the amount of a predetermined substance distributed in the functional area; and a determination unit that determines the presence or absence of the predetermined substance related to a lesion in the brain based on the index calculated for each of the plurality of functional areas, or calculates a determination value related to the amount of the predetermined substance present.

2. The brain medical image processing device described in claim 1, wherein the judgment unit calculates a first judgment value based on the index calculated for each of the multiple functional areas, calculates a second judgment value by weighting and adding the first judgment value using weights determined for each functional area, and makes the judgment based on the second judgment value.

3. The brain medical image processing device according to claim 1, further comprising a display control unit that displays the judgment value on a display unit.

4. The brain medical image processing device according to claim 1, wherein the predetermined substance is a substance associated with Alzheimer's disease.

5. The brain medical image processing device according to claim 1, wherein the predetermined substance is amyloid β protein or tau protein.

6. A brain medical image processing device as described in claim 2, wherein the first judgment value is determined by a cutoff value calculated from ROC analysis based on the index calculated for each of the multiple functional areas.

7. The brain medical image processing device according to claim 2, wherein the weights are determined based on the likelihood of the order of propagation of the predetermined substance within the brain over time.

8. The brain medical image processing device according to claim 2, wherein the weight is larger for a functional region located upstream of the propagation of the predetermined substance and smaller for a functional region located downstream.

9. The brain medical image processing device according to claim 1, wherein the calculation unit calculates a first map indicating the magnitude of the amount of the specified substance present in the voxel, and calculates the index based on the first map.

10. A brain medical image processing device as described in claim 9, wherein the index is the average value of the voxel values ​​of the first map divided by the entropy of the voxel values ​​of the first map divided by the standard deviation of the voxel values ​​of the first map.

11. A brain medical image processing device as described in claim 9, wherein the index is a value obtained by multiplying the average value of the voxel values ​​of the first map by the value obtained after an inverse scaling operation is performed on the entropy of the voxel values ​​of the first map, and multiplying the standard deviation of the voxel values ​​of the first map by the value obtained after an inverse scaling operation is performed.

12. A brain medical image processing device as described in claim 9, wherein the calculation unit calculates, as the first map, a map indicating the magnitude of a quantity corresponding to volume magnetic susceptibility based on data collected from the imaging subject by executing a pulse sequence.

13. The brain medical image processing device according to claim 12, wherein the pulse sequence is a sequence that generates a plurality of echoes, and the calculation unit calculates the first map based on a value obtained by dividing the value of the phase change between the plurality of echoes by the value of the change in TE between the echoes.

14. A brain medical image processing device as described in claim 9, wherein the calculation unit calculates the index by performing a normalization process by subtracting the average value of the voxel values ​​of the first map from the average value of the voxel values ​​of the lesion-free region.

15. The brain medical image processing device according to claim 1, wherein the index is an index normalized using an index in a lesion-free region.

16. A brain medical image processing device as described in claim 1, wherein each of the plurality of functional areas consists of a plurality of sub-functional areas, the calculation unit calculates the index for each of the functional area and each of the sub-functional areas, and the judgment unit makes the judgment based on the index calculated for each of the functional area and each of the sub-functional areas.

17. The brain medical image processing device according to claim 16, wherein the plurality of sub-functional regions are white matter regions and gray matter regions.

18. A brain medical image processing device as described in claim 1, wherein the calculation unit calculates the index for each of multiple imaging sessions, and the determination unit makes the determination taking into account changes in the index between imaging sessions.

19. A brain medical image processing device as described in claim 18, wherein the determination unit makes the determination based on the difference between the index calculated for the latest imaging and the index calculated for the imaging immediately preceding the latest imaging.

20. A brain medical image processing system comprising: a sequence control unit that executes a pulse sequence to collect data from an imaging target; a calculation unit that calculates, for each of a plurality of functional areas of the brain, an index related to the amount of a predetermined substance distributed in the functional area based on the data; and a determination unit that determines the amount of the predetermined substance related to a lesion in the brain based on the index calculated for each of the plurality of functional areas, or calculates a determination value related to the presence or absence of the predetermined substance.

Images