IMAGE ANALYSIS DEVICE, IMAGE ANALYSIS METHOD, AND IMAGE ANALYSIS PROGRAM

Abstract

To provide an image analysis device and method capable of performing automatic measurement of tissue precisely even for images different in imaging conditions and machine types.SOLUTION: An image including contrast is standardized and an area is divided. A measurement cross section that most matches a measurement cross section determined to be appropriate for measuring a predetermined index is determined based on a feature of a notable tissue in the divided area, and a measurement value serving as an index is calculated in the measurement cross section.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to an image analysis device that presents diagnostic indices for diseases and the like based on images acquired by a medical imaging device, and more particularly to a technique for presenting indices of brain abnormalities that are important in diagnosing dementia. [Background technology]

[0002] Medical imaging devices such as X-ray imaging devices, CT devices, ultrasound imaging devices, and MRI devices are widely used as effective means in the diagnosis of various diseases. A doctor checks images of the test subject obtained from these medical imaging devices to make a diagnosis, but in some cases, the doctor may measure changes in the shape of the tissue shown in the image and use the measured values ​​as indicators to make a diagnosis. For this reason, a technology has been developed for performing measurements and outputting diagnostic indicators in the medical imaging device or in an image processing device that receives image data from the medical imaging device. When performing measurements with the device, for example, an image including the target tissue is displayed on a display device, and a doctor, technician, or the like (hereinafter referred to as a user) inputs information such as the position to be measured for the displayed image via an input means, and the device then calculates the measurement values ​​and presents the results.

[0003] A technology has also been proposed that eliminates the need for user input, automatically determines the region of interest (ROI) to be measured from multiple cross sections, and calculates the measurement values ​​(Patent Document 1, etc.). In the technology described in Patent Document 1, when the ROI is automatically determined, anatomical standardization is performed on the target tissue (here, the brain) in order to eliminate the variation in measurements due to individual differences in the imaging subject, and data on the anatomical regions allocated on the standard brain are presented as candidates for the ROI. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2019-74343 A DISCLOSURE OF THEINVENTION [Problem to be solved by the invention]

[0005] However, diagnostic images vary significantly in image resolution and contrast depending on, for example, the modality of the imaging device, or, in the case of an MRI device, the imaging conditions (imaging parameters). Therefore, it is difficult to maintain the accuracy of determining the measurement cross section and measurement position for various images of different image quality using anatomical standardization alone.

[0006] In recent years, research on brain findings of dementia and brain diseases with similar symptoms has progressed, and the procedure for diagnosing dementia from various brain findings is becoming clearer. One of these is normal pressure hydrocephalus. It is difficult to distinguish and diagnose secondary normal pressure hydrocephalus (sNPH) and idiopathic normal pressure hydrocephalus (iNPH) from dementia, and guidelines for diagnosing it by imaging diagnosis have been created. The iNPH guidelines list indicators such as the Evans index and the corpus callosum angle, as well as asymptomatic ventricular enlargement with DESH findings. The Evans index is a diagnostic indicator for iNPH, and the ratio of the maximum width between the anterior horns of the lateral ventricles on both sides to the width of the intracranial cavity at that site is greater than 0.3. Regarding the corpus callosum angle, a steep corpus callosum angle (90 degrees or less) in an MRI coronal section passing through the posterior commissure and perpendicular to the anterior commissure-posterior commissure is a diagnostic indicator that can distinguish it from Alzheimer's dementia. DESH findings are characterized by minimization of the higher fornix accompanied by disproportionate expansion of the subarachnoid space, and are a diagnostic indicator that can distinguish Alzheimer's disease from dementia due to their high sensitivity and specificity.

[0007] When a doctor measures and calculates such diagnostic indices from images, he or she uses an image of a suitable modality or contrast depending on the diagnostic index. For example, in the case of MR images, the Evans Index is preferably obtained using T2-weighted images or FLAIR images in which cerebrospinal fluid is depicted as the highest or lowest signal area, while the corpus callosum angle is preferably obtained using T1-weighted images. Therefore, if an automatic measurement is to be performed from input diagnostic images, it is necessary to prepare diagnostic images according to the index, which is a huge burden.

[0008] In addition, because these indices are calculated from measurements of the fine shape of the brain, slight differences in the measurement section (slice position) can lead to large differences in values, so it is extremely important to select an appropriate slice position. However, it is difficult to select an appropriate measurement section (slice position), and even when a doctor performs measurements, it is difficult to determine an appropriate section for measuring the corpus callosum angle using only one coronal section (COR section). Similarly, for other indices, it is difficult to determine an appropriate section for subjects with different ages, symptoms, etc.

[0009] An object of the present invention is to provide an image analysis device and method that are capable of measuring tissue with high accuracy even for images obtained using different modalities, models, or under different imaging conditions. [Means for solving the problem]

[0010] In order to solve the above problems, the present invention standardizes an image including contrast and divides the image into regions, and determines a measurement cross section that best matches a measurement cross section that has been confirmed to be appropriate for measuring a specified indicator based on the characteristics of the tissue of interest in the divided regions, and calculates a measurement value that serves as the indicator in the measurement cross section.

[0011] That is, the image analysis device of the present invention includes an image receiving unit that inputs a diagnostic image and imaging conditions of the diagnostic image, an image standardization unit that standardizes the diagnostic image, a region dividing unit that extracts a region of a tissue of interest from the diagnostic image standardized by the image standardization unit, a slice position determination unit that determines a slice position in the standardized diagnostic image based on characteristics of the tissue of interest divided into regions, a measurement unit that measures the tissue of interest at the slice position determined by the slice position determination unit, and an index calculation unit that calculates an index using the measurement value of the tissue of interest measured by the measurement unit. The region dividing unit divides the region so that the pixel values ​​of the divided region become predetermined standard pixel values ​​(labels) based on the imaging conditions of the diagnostic image.

[0012] In addition, the image analysis method of the present invention is an image analysis method for inputting a diagnostic image and calculating an index of a specific disease, and includes the steps of inputting the diagnostic image and the imaging conditions of the diagnostic image, an image standardization step for standardizing the diagnostic image, a region division step for extracting a region of tissue of interest from the standardized diagnostic image, a step for determining a slice position in the standardized diagnostic image based on the characteristics of the region-divided tissue of interest, a step for measuring a specified tissue of interest at the determined slice position, and a step for calculating an index using the measured measurement value of the tissue of interest, and is characterized in that in the region division step, the region is divided so that the pixel values ​​of the divided region become predetermined standard pixel values ​​(labels) based on the imaging conditions.

[0013] Furthermore, the image analysis program of the present invention is a program for causing a computer to execute the above steps. In this specification, the term "section" includes both sections such as an axial section, a coronal section, a sagittal section, and the like, and a section at a predetermined slice position. Effect of the Invention

[0014] According to the present invention, by standardizing the image including the image quality using the imaging conditions, it is possible to perform accurate measurement regardless of the difference in modality or imaging conditions. In addition, by determining the slice position based on the characteristics of the target tissue standardized in this way, it is possible to determine the most appropriate slice position for calculating the index, and it is possible to prevent erroneous measurement of the index due to a slight difference in the slice position. [Brief description of the drawings]

[0015] [Figure 1] FIG. 1 shows an overview of the configuration of a medical imaging device and an image analysis device. [Diagram 2] A functional block diagram showing an image processing unit of an image analysis device. [Diagram 3] FIG. 2 shows a process flow of the image analysis device. [Figure 4]A diagram explaining the indices measured from brain images. (A) is an image of an axial section used to measure the Evans Index, and (B) is an image of a coronal section used to measure the corpus callosum angle. [Diagram 5] Diagram showing the steps of image standardization processing [Figure 6] A diagram showing the procedure for determining slice positions [Figure 7] FIG. 1 is a diagram for explaining the details of one procedure for determining slice positions. [Figure 8] FIG. 1 is a diagram for explaining slice position determination. [Figure 9] 13A and 13B are diagrams for explaining index calculation in a modified example, in which (A) shows an axial cross section and (B) shows a coronal cross section. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] Hereinafter, an embodiment of an image analyzing apparatus according to the present invention will be described with reference to the drawings.

[0017] <Embodiment 1> As shown in FIG. 1, the image analysis device 100 has an image capture unit (image receiving unit) 10 that captures images captured by a medical imaging device 300, an image processing unit 20 that performs various processes using the captured images, and a statistical analysis unit 30 that performs statistical processing using a large number of images.It also has a memory device 40 that stores data necessary for processing and processing results, and a UI unit 50 that has a display unit and input means for displaying images and measurement values ​​that are the processing results, as well as a GUI, etc.

[0018] The image analysis device 100 may be connected to a medical imaging device 300 and a database 500 of multiple images. The medical imaging device 300 is not limited to a single device, but may be multiple devices installed in different locations, or multiple types of devices with different modalities. Examples of devices with different modalities include an MRI device, an X-ray imaging device, an ultrasound imaging device, a CT device, and a PET device.

[0019] The image acquisition unit 10 corresponds to an image receiving unit that inputs a diagnostic image and the imaging conditions of the diagnostic image, and acquires the image data to be processed and its DICOM information (various information including subject information and imaging conditions attached to the image) via wired, wireless, portable media, or a network, etc., and passes it to the image processing unit 20 or the statistical analysis unit 30.

[0020] The image processing unit 20 mainly performs processes such as image standardization, tissue segmentation, slice position determination, predetermined measurement, and index calculation for the image to be measured (measurement image), as shown in Fig. 2. For this purpose, the image processing unit 20 has an image information extraction unit 210 that extracts image information from the data captured by the image capture unit 10, an image standardization unit 220 that standardizes the diagnostic image, a region division unit 230 that extracts the region of the tissue of interest for the standardized diagnostic image, a slice position determination unit 240 that determines the slice position in the standardized diagnostic image based on the characteristics of the tissue of interest divided into regions, a measurement unit (here, a tissue of interest extraction unit 250 and a line segment calculation unit 260) that measures the tissue of interest at the slice position determined by the slice position determination unit, and a diagnostic index calculation unit (hereinafter, referred to as the index calculation unit) 270 that calculates an index using the measurement value of the tissue of interest measured by the measurement unit.

[0021] The statistical analysis unit 30 performs statistical processing of indices and creates standard images for a large number of images including images processed by the image processing unit 20. In this embodiment, the statistical analysis unit 30 selects or creates a standard image to be referred to when the image processing unit 20 selects a slice position for a measurement image.

[0022] The main elements of the image analysis device 100 described above can be constructed in a computer (workstation) equipped with a memory and a CPU or GPU, and the functions of the main elements are realized by the computer reading a program. However, some of the functions of the image analysis device 100 can also be realized by hardware such as a PLD (programmable logic device).

[0023] Next, an overview of the operation of the image analyzing device 100 having the above configuration will be described.

[0024] First, the image import unit 10 imports DICOM information along with image data, and the image information extraction unit 210 extracts information such as modality (MR / CT), 2D / 3D, imaging section, contrast, FOV, and resolution from the DICOM information (S1). Next, the image standardization unit 220 performs processing to align image parameters and positions such as FOV and resolution, which differ depending on the image, to predetermined values ​​and positions (standard values ​​or standard positions) (S2).

[0025] Next, the region dividing unit 230 extracts regions of tissues of interest by segmentation from the standardized image (S3). Segmentation divides regions based on differences in pixel values ​​due to tissues from modality (MR image, CT image, etc.) and contrast information extracted from DICOM information. At this time, in this region division (region extraction), the region dividing unit 230 assigns predetermined pixel values ​​to each tissue in the segmented image, regardless of the modality, image type (difference in contrast), etc. of the original image. This allows for the acquisition of a segmented image with contrast that is independent of the contrast of the original image.

[0026] Next, the slice position determination unit 240 uses the segmentation image to determine the slice section to be measured (S4). The slice position determination unit 240 creates a binary image of a predetermined tissue of interest from the segmentation image, and aligns it with a pre-standard image in which a slice position suitable for calculating an index is empirically known, to determine the slice position. The pre-standard image (hereinafter simply referred to as a standard image) is an image in which the statistical analysis unit 30 previously averages images acquired from a large number of subjects, and the slice position used for calculating the index is determined, and the image standardization unit 220 has standardized the standard values ​​of FOV and resolution used for standardization, and the standard position of the subject.

[0027] When the slice position is determined, the tissue of interest extraction unit 250 extracts the tissue of interest to be measured from the segmentation image at the slice position (S5). The tissue of interest to be extracted may be the same as the tissue of interest (region) used when determining the slice, or it may be different. The tissue of interest to be measured is determined by the index to be calculated, and the region division unit 230 assigns a predetermined pixel value to each divided region. Therefore, if the index is determined, the tissue of interest can be automatically extracted. In other words, the region of the tissue of interest is determined based on the position, shape, and size of the tissue of interest.

[0028] The line segment calculation unit 260 calculates a predetermined line segment for the extracted tissue of interest (S6). The line segment differs depending on the tissue of interest and the index calculated by the index calculation unit 270, and a specific method will be described later, but it corresponds to the doctor measuring the width and angle of a predetermined position of the tissue of interest based on the image, and calculates a line segment at the position of the maximum width, a line segment at the position of the tangent, etc. based on the shape characteristics of the tissue of interest.

[0029] Finally, the index calculation unit 270 calculates an index using the calculated line segment, and outputs the result to an output device such as a display device (S7).

[0030] According to this embodiment, the image is standardized, and then the pixel values ​​of each region are unified to perform region division, thereby minimizing the modality dependency, imaging condition dependency, and imaging position dependency when calculating the index. Furthermore, according to this embodiment, the slice is determined based on the shape characteristics of the tissue of interest, and the tissue of interest is determined based on the position, shape, and size of the tissue of interest, so that the subject dependency of the slice position and the tissue of interest can be minimized. As described above, the index can be automatically calculated with higher accuracy.

[0031] <Embodiment 2> Next, an embodiment will be described in which the image is a brain image and the indices to be calculated are the Evans Index and the corpus callosum angle.

[0032] The configuration of the device and the outline of the processing (FIG. 3) are the same as those in the first embodiment, so the following mainly describes the details of the function and operation of each unit of the image processing unit 20.

[0033] Figures 4(A) and (B) show the axial and coronal planes of brain images. As shown in Figure 4(A), the Evans Index is an indicator of iNPH diagnosis when the ratio (W401 / W402) between the width W401 of the anterior horn of the lateral ventricle and the width W402 of the intracranial cavity exceeds 0.3. As for the corpus callosum angle, an indicator of iNPH diagnosis is that the corpus callosum angle in the MRI coronal section (a section perpendicular to the line connecting the anterior commissure and the posterior commissure) passing through the posterior commissure is steep, specifically, 90 degrees or less. Therefore, in the Evans Index, the slice position is determined in the axial section, and the ventricles and brain parenchyma are processed as tissues of interest, and in the corpus callosum angle, the slice position is determined in the coronal section, and the ventricles are extracted.

[0034] In this embodiment, the image capture unit 10 captures DICOM information together with image data, and the image information extraction unit 210 extracts information such as modality (MR / CT), 2D / 3D, imaging section, contrast, FOV, and resolution from the DICOM information (S1), similar to the first embodiment. Below, the details of the processing from S2 in Fig. 3 onwards in the case of targeting brain images will be described.

[0035] <Image standardization process: S2> As shown in FIG. 5, the image standardization unit 220 adjusts the FOV (S21), the resolution (S22), and the subject position (S23). The standard parameter value is set in advance to a value most suitable for measuring the index to be measured. If the measurement target is, for example, the Evans Index, the standard values ​​are set to FOV: 300 mm, resolution: 1 mm, etc., and the FOV and resolution of the measurement image are adjusted to the standard values. The FOV is adjusted to the standard value by cutting out unnecessary areas or zero-filling the periphery (S21). As for the resolution, for example, if the resolution is larger than the standard value (for example, 2 mm), the gap between pixels is filled by linear interpolation or the like to adjust it to the standard value (S22). The above processing is either two-dimensional processing or three-dimensional processing depending on whether the image data is a 2D multi-slice image or a 3D image.

[0036] The image standardization unit 210 further adjusts the subject position in the image (S23). The position is adjusted by, for example, translating or correcting the angle of the brain center position so that the brain center becomes the image center. The brain center position can be determined by creating a head outline mask (an image in which the inside and outside of the head outline are binarized) from the subject's image (diagnostic image) and determining it as the center of gravity of the mask, or by approximating the head outline mask on the axial plane to an ellipse and determining the center of the ellipse as the brain center position, and any of these methods may be used. For the coronal plane, the brain center may be determined by utilizing the symmetry of the head with landmarks such as the parietal part and the eyeballs.

[0037] In the case of 3D image data, the center positions of multiple cross sections (e.g., axial and coronal planes) may be determined, and the intersection point of a line passing through the center positions and perpendicular to the cross sections may be determined as the center of the 3D image data. Through the above processing, the measurement images are standardized to images with a unified FOV and resolution and with the center of the brain as the image center.

[0038] <Area division process: S3> When performing region division by segmentation, the region division unit 230 first obtains modality (MR image or CT image, etc.) and contrast information from the DICOM information, and divides the region based on the difference in pixel value due to tissue. In the case of MR images, the contrast information differs depending on the type of image determined by the imaging conditions, and in the case of T2 weighted images, cerebrospinal fluid is depicted as the highest signal region, and bone is depicted as the lowest signal region. Therefore, the highest signal region can be divided into cerebrospinal fluid, the lowest signal region into bone, and the rest into brain parenchyma. Similarly, in the case of T1 weighted images, cerebrospinal fluid is depicted as the central signal region, and bone is depicted as the lowest signal region, and in the case of FLAIR images, cerebrospinal fluid and bone are depicted as black, so that they can be segmented as the lowest signal region. In the case of CT images, contrast is determined by the CT value of each tissue, so segmentation can be performed based on the CT value.

[0039] The segmentation method can be a known method such as the k-means method, region growing method, or semantic segmentation, which is a type of deep learning that associates a category with all pixels in an image, and thereby an image (segmented image) is obtained in which brain tissues such as cerebrospinal fluid, bone, and brain parenchyma (white matter, gray matter) are each represented by a single pixel value (brightness value). At this time, the region division unit 230 assigns a predetermined pixel value (label) to the pixel value of each divided tissue in the segmented image, regardless of the modality of the original image, image type (difference in contrast), etc. In other words, although the resolution of the pixel value differs depending on the image, it is divided into a predetermined number of stages (for example, the number of tissues of interest), and regardless of the original image, the pixel value is set to a predetermined pixel value such as pixel value 1 for white matter, pixel value 2 for gray matter, pixel value 3 for cerebrospinal fluid, etc. This allows a segmented image with uniform contrast to be obtained, and the following processing can be performed regardless of the image type.

[0040] <Slice position determination: S4> The slice position determination unit 240 uses the segmentation image to determine the slice section to be measured. In the case of 2D image data, since it is composed of multiple slice images, the slice image most suitable for measurement is selected from these slice images. In the case of 3D image data, multiple slices are cut out at a predetermined slice interval for a section according to the measurement target, and the slice position most suitable for measurement is determined. Whether the image data is 2D or 3D can be known from the DICOM information extracted by the image information extraction unit 210, and the above processing is performed accordingly. Then, for example, in the case of Evans Index, the slice position of the axial section is determined, and in the case of the corpus callosum angle, the slice position of the coronal section is determined.

[0041] For this reason, as shown in FIG. 6, the slice position determining unit 240 reads a standard image (S41), generates a binary image (S42), compares it with the standard binary image (S43), and determines the slice position that minimizes the error (S44).

[0042] As described in the first embodiment, the standard image is an image created from a large number of images captured in the past, with known slice positions for calculating a predetermined index, and is an image segmented such that the pixel values ​​of each region are the reference pixel values, similar to the region segmentation process of a diagnostic image. That is, this segmentation image is an image in which the position and pixel values ​​of each region are the average values ​​of a large number of images. The standard image or its segmentation image is created in advance in the statistical analysis unit 30, and stored in the storage device 40 or database 500. Standard images may be created for each age group, such as children / adults / elderly people, and a standard image for the corresponding age group may be read based on the DICOM subject information (S41).

[0043] Next, the slice position determination unit 240 creates a reference image from the read standard image, and creates a binary image of the tissue of interest (an image in which the pixel value of the tissue of interest is 1 and the rest is 0) from the segmentation image obtained by region division (S3) (S42). The tissue of interest is a tissue whose shape changes clearly for each slice position and whose slice position is easy to determine, such as a cerebrospinal fluid region or eyeball tissue. A binary image is obtained by expressing the tissue of interest and other tissues in the segmentation image as binary values. Note that the reference image may be created in advance by the statistical analysis unit 30 for each predetermined tissue of interest based on the standard image, and then read in S41.

[0044] The binary image is an image in which the shape characteristics of the tissue of interest are extracted, and the slice position determining unit 240 determines the slice position using these characteristics. For this purpose, first, a reference image created from a standard image is compared with a binary image of the diagnostic image to be measured, and an error is calculated (S43). Here, an example of a specific method will be described using the case in which the tissue of interest is the cerebrospinal fluid region as an example.

[0045] The upper side of FIG. 7 is a reference image created by averaging a number of images of the cerebrospinal fluid region captured in the past, and the lower side of FIG. 7 is a binary image of the cerebrospinal fluid region of the image to be measured. Images A1, A2, and A3 at two or more slice positions (here, 3) from the reference image are compared with binary images I1, I2, and I3 of two or more measurement images with the same slice interval, and the difference (In-An) (n is any of 1 to 3) is calculated. From the difference between the corresponding images (binary images of the reference image and the diagnostic image), the root mean square error (RMSE) shown in formula (1) or the normalized RMSE shown in (2) is calculated. Note that normalization is not essential in calculating the RMSE, but normalization can eliminate the influence of the change in error due to the size of the tissue of interest, that is, the influence on the calculation result caused by the difference in the set to be compared.

[0046] (Number 1) √(Σ(In-An) 2 ) (1) (Number 2) √(Σ(In-An) 2 / ΣAn 2 ) (2)

[0047] FIG. 7 shows a case where three images are compared, one of the reference images being an image at slice position S0 used for measuring the index, and the other two images being images at slice positions spaced a predetermined distance from slice position S0. The binary images of the diagnostic image are three images at slice positions spaced the same distance as the reference images. The image spacing may be the same or different from the middle image, and may be adjacent or spaced apart to some extent. However, the spacing between the reference image and the binary images of the diagnostic image is the same.

[0048] The slice position of the diagnostic image is shifted successively while maintaining the same slice interval as the previous three images, changing the set of three images to be subtracted from the three reference images, and the root mean square error is calculated for each set. By calculating the root mean square error for multiple sets, the set of three images with the smallest value is found, as shown in Figure 8. Since the central slice position in this set corresponds to the slice position S0 of the reference image, this slice position is set as the slice position of the diagnostic image (S44).

[0049] In the example shown in Fig. 7, the set of reference images includes a slice position having a slice position S0 at the center, but in this process, it is only necessary to know the positional relationship between the reference image with a determined slice position S0 and the binary image of the diagnostic image, so it is not necessary to set the set so that the slice position S0 is at the center. In addition, the positional relationship can be found more accurately by using images of three slice positions, but it is also possible to find it with two, and it is not limited to three.

[0050] In addition, since it is only necessary to find the position where the difference is the smallest, it is not necessary to perform the above-mentioned repeated calculation for all images of the diagnostic image, and it is sufficient that the slice positions included in the entire set are discretely selected from slice positions on both sides of the center of the image. Furthermore, regarding the difference, although an example of calculating the root mean square error has been described here, other indexes of difference, such as the root mean square error, may also be used.

[0051] Furthermore, in the above description, the slice position is determined using the error of the tissue of interest from the average shape (standard image). However, it is also possible to use indices representing features other than the shape, such as area or contour length.

[0052] The same method can be used to set a predetermined slice position whether the measurement plane is an axial plane or a coronal plane. In addition, for 3D image data, the slice position may be determined by comparing binary images of both the axial plane and the coronal plane.

[0053] <Extraction of tissue of interest: S5> Once the slice position determination unit 240 has determined the slice position as described above, the tissue of interest extraction unit 250 extracts tissue of interest for measurement from the image at the determined slice position. For example, in the case of Evans Index measurement, the ventricle width and the intracranial cavity width are measured, so the ventricle and brain parenchyma (intracranial cavity) are extracted. The tissue of interest is extracted based on the signal values ​​of each region divided by segmentation. For the corpus callosum angle, only the ventricle is extracted.

[0054] The tissue of interest may be extracted from only the image of one determined slice position. However, in this embodiment, in order to improve the accuracy of index calculation, it is preferable to also extract the tissue of interest from images of neighboring slice positions, for example, the slice positions before and after it.

[0055] <Line calculation: S6> Next, the line segment calculation unit 260 performs a predetermined measurement on the extracted tissue of interest (a predetermined slice position and its neighboring slice positions).

[0056] In the case of the Evans Index, as shown in Figure 4(A), since it is known that the maximum width between the anterior horns of the lateral ventricles to be measured is located in the posterior half of the brain from the center position, the line segment where the ventricle width is maximum is calculated for the posterior half (upper half) of the axial section of the extracted ventricle, and this is taken as the maximum width between the anterior horns of the lateral ventricles. Also, since it is known that the width of the intracranial cavity is located in the anterior half, the width of the brain parenchyma is measured for the lower half of the axial section, and the line segment where this width is maximum is calculated and taken as the width of the intracranial cavity.

[0057] When the tissue of interest is also extracted from the neighboring slices, the maximum width between the anterior horns of the lateral ventricles on both sides is calculated for those slices as well, and the slice with the largest maximum width among the predetermined slice and the neighboring slices is determined as the measurement target, and the intracranial cavity width at that slice position is calculated. Usually, the slices of an MR image are spaced 6 to 7 mm apart. Therefore, the maximum width between the anterior horns of the lateral ventricles on both sides may be slightly shifted from the slice position determined in the slice position determination step S4. By calculating the maximum width between the anterior horns of the lateral ventricles on both sides for the neighboring slices, the slice position to be measured can be determined among the multiple slices including the neighboring slices, and the maximum width between the anterior horns of the lateral ventricles on both sides and the intracranial cavity width can be measured at that slice position.

[0058] In the case of the corpus callosum angle, the center line along the sulcus (dotted line L in FIG. 4(B)) and the closest point between the left and right ventricles (point P in FIG. 4(B)) are determined for the ventricles extracted in the coronal section, and the left and right tangents of the ventricle shape from the point toward the vertex are obtained, and two line segments are calculated. For example, the tangent can be determined as the line segment where the line segment and the contour of the ventricle are in contact when the angle of the dotted line L is changed by first setting the point P. Note that, although the two line segments are calculated using only the coronal section here, if the image data read by the image processing unit 20 is a 3D image or includes image data of a sagittal section, it is also possible to use the image of the sagittal plane, for example, to determine the point P, such as to determine the center line. In the above explanation, the tangents are obtained, but other methods can be used, such as determining the line segment by connecting one point on the left and right sides of the ventricle shape at a predetermined distance from the point P to the point P.

[0059] <Indicator calculation: S7> Finally, the index calculation unit 270 calculates an index using the calculation result of the line segment calculation unit 260. That is, the maximum width between the anterior horns of both lateral ventricles is divided by the width of the intracranial cavity to obtain the Evans Index. In addition, the angle between the two line segments calculated based on the point P is calculated to obtain the corpus callosum angle.

[0060] The index calculated by the index calculation unit 270 may be output as a numerical value via the UI unit 50 associated with the image analysis device 100, or may be transmitted to another image processing device within the image analysis device 100 and output as a medical finding integrated with other findings of iNPH, etc. Furthermore, the statistical analysis unit 30 may display a map of the Evans Index and the corpus callosum angle by age together with the results calculated from a large number of diagnostic images in the past, and display the position of the subject on the map in an identifiable manner.

[0061] The above process completes the calculation of the index for the diagnostic image to be measured, and data for this diagnostic image (segmentation image or binary image, and information on the determined slice position) may be transmitted to the statistical analysis unit 30. The statistical analysis unit 30 can update an already created standard image (averaged segmentation image or binary image of a specified region) using the transmitted data.

[0062] According to this embodiment, the slice position and tissue of interest for measuring the Evans Index and the corpus callosum angle, which are diagnostic indices for iNPH, can be determined without depending on the type of diagnostic image input or individual differences between subjects, thereby making it possible to automatically calculate and present diagnostic indices with high accuracy.

[0063] Furthermore, after a slice is determined, the accuracy of determining the maximum width can be improved by also measuring neighboring slices. Furthermore, depending on the facility or examination flow, it is not always the case that sagittal section image data is acquired; however, in this embodiment, since the index can be calculated from a single section, it is also possible to deal with situations where all data sets are not available. In this embodiment, the case where two diagnostic indexes for iNPH are calculated has been described, but the present invention naturally encompasses the case where only one of them is calculated.

[0064] <Modification of the second embodiment> In the second embodiment, the case of obtaining the Evans Index and the corpus callosum angle, which are numerically quantified indices, has been described, but it is also possible to obtain information related to DESH findings by applying a similar method. The acquisition of information related to DESH findings will be described below.

[0065] The way in which the brain shrinks differs between Alzheimer's disease and iNPH, with the former being accompanied by atrophy of the entire brain parenchyma, while the latter being characterized by disproportionate expansion of the subarachnoid space, and such findings are called DESH findings. Although the guidelines do not provide numerical indicators for DESH findings, when viewed in images, the parietal brain parenchyma is sparse (sparse) in Alzheimer's disease, whereas such findings are not seen in iNPH. In this embodiment, in order to find the characteristics of such DESH findings, the ventricles and brain parenchyma (intracerebral cavity) are extracted as tissues of interest, and the area of ​​the brain parenchyma is calculated. The cross section for extracting the tissue of interest is preferably a coronal cross section, but information from an axial cross section or both cross sections may be used together.

[0066] In this modified example, the basic processing flow (image information extraction (S1), image standardization (S2), region division (S3), slice position determination (S4), and tissue of interest extraction (S5)) is similar to the processing flow of the second embodiment shown in FIG. 3. Hereinafter, this modified example will be described, focusing on the points that differ from the second embodiment.

[0067] After determining the slice position using a binary image created from the segmentation image (S1 to S4), the tissue of interest extraction unit 250 extracts the ventricles and brain parenchyma using the slice image (S5). The same process may be performed on the neighboring slice images.

[0068] Next, the line segment calculation unit 260 calculates lines for further dividing the brain parenchyma of the extracted cross section. In the axial cross section, as shown in FIG. 9(A), for example, four lines are determined to divide the brain parenchyma area surrounded by the outer shape of the ventricle and a shape that is a reduced shape of the outer shape and is at a predetermined distance from the center of the brain into four small areas. The four lines may be lines that are equally spaced from the center of the brain, or may be determined based on a standard of another shape. In the coronal cross section, as shown in FIG. 9(B), the brain parenchyma area surrounded by the outer shape of the brain parenchyma and the brain parenchyma surrounded by a similar shape at a predetermined distance from the outer shape of the brain parenchyma is divided into three small areas by four lines (lines that divide the area and lines that determine the ends of the area).

[0069] Furthermore, in this modification, the line segment calculation unit 260 (measurement unit) calculates the area (volume) of each divided small region. The area can be calculated as the sum of pixel values. When the area is calculated for a specific slice and its neighboring slices, the average value may be used as the area of ​​the small region.

[0070] The index calculation unit 270 provides information related to the DESH findings based on the area (volume) of each small region. The information related to the DESH findings may be calculated as an index, such as the area ratio or volume ratio of the small regions, or a comparison with the average value of normal subjects or patients diagnosed with Alzheimer's disease.

[0071] This modified example may be a process specialized for DESH findings, but by implementing it in conjunction with the process of embodiment 2 and presenting it together with the Evans Index and the corpus callosum angle index, it can contribute to the diagnosis of iNPH.

[0072] Furthermore, in the second embodiment and its modified example, an example of calculating an index of iNPH from a brain image has been described; however, the present invention can also be applied to brain diseases other than iNPH and areas other than the brain, as long as there is a technology for determining a cross section and automatically measuring an area of ​​interest, thereby making it possible to perform measurements while avoiding modality dependency, image type dependency, subject dependency, and the like. [Explanation of symbols]

[0073] 100: image analysis device, 10: image acquisition section, 20: image processing section, 30: statistical analysis section, 40: storage device, 50: UI section, 210: image information extraction section, 220: image standardization section, 230: region division section, 240: slice position determination section, 250: tissue of interest extraction section, 260: line segment calculation section, 270: diagnostic index calculation section, 300: medical imaging device, 500: medical image database

Claims

1. An image receiving unit for inputting a diagnostic image consisting of a two-dimensional multiple slice image or a three-dimensional image and imaging conditions for the diagnostic image; an image standardization unit for standardizing the diagnostic image; a region segmentation unit that extracts a region of a tissue of interest from the diagnostic image standardized by the image standardization unit; a slice position determination unit that determines a slice position in the standardized diagnostic image based on the characteristics of the region-divided tissue of interest; a measurement unit that determines a tissue of interest to be measured based on at least one of a position, a shape, a length, and a size of the tissue at the slice position determined by the slice position determination unit, and measures the tissue; an index calculation unit that calculates an index using a measurement value of the tissue of interest measured by the measurement unit; The image analysis device according to the present invention, characterized in that the region dividing unit extracts regions based on the imaging conditions so as to assign predetermined standard pixel values ​​to pixels of the divided regions.

2. 2. The image analysis device according to claim 1, The image analysis device, characterized in that the imaging conditions include the type of imaging device that acquired the diagnostic image, the image dimension, the image contrast, the FOV, and the resolution.

3. 2. The image analysis device according to claim 1, The image analysis device according to the present invention, wherein the image standardization unit performs standardization to align the FOV, resolution, and subject position included in the imaging conditions with predetermined standard values ​​and standard positions.

4. 2. The image analysis device according to claim 1, The slice position determination unit is an image analysis device characterized in that, in a pre-standard image consisting of a two-dimensional multiple slice image or a three-dimensional image, in which a reference slice position for measuring the index has been previously determined, the slice position of the diagnostic image that corresponds to the reference slice position of the pre-standard image is determined as the slice position based on evaluation values ​​obtained from the pre-standard image at multiple slice positions and the diagnostic image at multiple slice positions.

5. 5. The image analysis device according to claim 4, An image analyzing apparatus, wherein the plurality of slice positions are different slice positions on the same cross section.

6. 5. The image analysis device according to claim 4, An image analysis device, wherein the plurality of slice positions are slice positions of different cross sections.

7. 5. The image analysis device according to claim 4, An image analysis device further comprising a statistical analysis unit that creates the preliminary standard image using a large number of diagnostic images.

8. 2. The image analysis device according to claim 1, The image analyzing device according to the present invention, wherein the measurement unit sets a line segment for calculating an index for the tissue of interest.

9. 9. The image analysis device according to claim 8, An image analysis device characterized in that the tissue of interest includes at least a portion of a ventricle and brain parenchyma, and the measurement unit calculates a line segment that is the maximum width of the anterior horn of the ventricle and a line segment that is the maximum width of the brain parenchyma.

10. 9. The image analysis device according to claim 8, An image analysis device characterized in that the tissue of interest includes at least a portion of a ventricle, and the measurement unit calculates a tangent to the ventricle as two line segments that determine the corpus callosum angle.

11. 2. The image analysis device according to claim 1, The image analyzing device is characterized in that the measuring unit divides the tissue of interest into a plurality of small regions and calculates the volume of each small region.

12. The image analysis device according to claim 11, The image analysis device is characterized in that the index calculation unit calculates a qualitative or quantitative index related to DESH findings of idiopathic normal pressure hydrocephalus based on the volume of the small region.

13. 2. The image analysis device according to claim 1, The diagnostic image is a brain image, and the indices calculated by the index calculation unit include at least one of the Evans Index and the corpus callosum angle, which are diagnostic indices for idiopathic normal pressure hydrocephalus.

14. An image analysis method for inputting a diagnostic image consisting of a two-dimensional multiple slice image or a three-dimensional image and calculating an index of a specific disease, comprising: inputting the diagnostic image and imaging conditions of the diagnostic image; an image standardization step of standardizing the diagnostic image; A region segmentation step of extracting a region of a tissue of interest from the standardized diagnostic image; determining a slice position in the standardized diagnostic image based on the features of the segmented tissue of interest; A step of measuring a predetermined tissue of interest at the determined slice position; Calculating an index using the measured value of the tissue of interest; An image analysis method, characterized in that in the region dividing step, regions are extracted so that pixels of the divided regions are assigned predetermined standard pixel values ​​based on the imaging conditions.

15. 15. The image analysis method according to claim 14, The diagnostic image is an image acquired by an MRI device, and the imaging conditions include a type of imaging device that acquired the diagnostic image, an image dimension, an image contrast, an FOV, and a resolution; The image analysis method according to the present invention, wherein the image standardization step performs standardization to align the FOV, resolution, and subject position with predetermined standard values ​​and standard positions.

16. 15. The image analysis method according to claim 14, The diagnostic image is a brain image, An image analysis method characterized in that the indices calculated in the step of calculating the indices include at least one of the Evans Index, which is a diagnostic indices for idiopathic normal pressure hydrocephalus, the corpus callosum angle, and a qualitative or quantitative indices related to DESH findings.

17. On the computer, receiving a diagnostic image consisting of a two-dimensional multi-slice image or a three-dimensional image and imaging conditions for the diagnostic image; an image standardization step of standardizing the diagnostic image; a region division step of performing region division processing on the standardized diagnostic image, and at that time, standardizing contrast by assigning predetermined standard pixel values ​​to pixels of the divided regions based on the imaging conditions; determining a slice position in the standardized image based on the characteristics of at least one tissue of interest segmented; A step of measuring a predetermined tissue of interest at the determined slice position; Calculating an index using the measured value of the tissue of interest; The image analysis program to be run.

Images