Medical information processor and medical information processing method

Abstract

To improve analysis accuracy using a medical image.SOLUTION: A medical information processor includes a calculation part and an identifying part. The calculation part calculates a first feature amount every position in a first area of the pancreas included in a medical image and a second feature amount every position in a second area different from the first area in the pancreas. The identifying part identifies a position to correct of the calculated feature amount.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The embodiments disclosed in this specification and the drawings relate to a medical information processing apparatus and a medical information processing method. [Background technology]

[0002] Conventionally, when analyzing diseases using medical images, segmentation is performed to detect target organs from medical images. However, detection accuracy can be low depending on the organ being segmented, which affects the accuracy of analysis using medical images. For example, the pancreas may be segmented inaccurately due to its structure. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-54179 [Patent Document 2] Japanese Patent Application Publication No. 2023-59149 [Patent Document 3] Japanese Patent Application Laid-Open No. 2017-29461 Summary of the Invention [Problem to be solved by the invention]

[0004] One of the problems to be solved by the embodiments disclosed in this specification and the drawings is to improve the accuracy of analysis using medical images. However, the problems to be solved by the embodiments disclosed in this specification and the drawings are not limited to the above problem. Problems corresponding to the effects of each configuration shown in the embodiments described below can also be positioned as other problems. [Means for solving the problem]

[0005] A medical image processing apparatus according to an embodiment includes a calculation unit and an identification unit. The calculation unit calculates a first feature amount, which is a feature amount for each position in a first region of a pancreas included in a medical image, and a second feature amount, which is a feature amount for each position in a second region of the pancreas that is different from the first region. The identification unit identifies a correction target position for the calculated feature amount. [Brief explanation of the drawings]

[0006] [Figure 1] FIG. 1 is a diagram showing an example of the configuration of a medical image processing apparatus according to the first embodiment. [Figure 2] FIG. 2 is a flowchart showing the processing procedure of the processing performed by each processing function of the processing circuitry of the medical image processing apparatus according to the first embodiment. [Figure 3] FIG. 3 is a diagram for explaining processing by the estimation function according to the first embodiment. [Figure 4A] FIG. 4A is a diagram for explaining an example of correction processing according to the first embodiment. [Figure 4B] FIG. 4B is a diagram for explaining an example of the correction process according to the first embodiment. [Figure 4C] FIG. 4C is a diagram for explaining an example of the correction process according to the first embodiment. [Figure 4D] FIG. 4D is a diagram for explaining an example of the correction process according to the first embodiment. [Figure 5A] FIG. 5A is a diagram showing an example of display information according to the first embodiment. [Figure 5B] FIG. 5B is a diagram showing an example of display information according to the first embodiment. [Figure 5C] FIG. 5C is a diagram showing an example of display information according to the first embodiment. [Figure 5D] FIG. 5D is a diagram showing an example of display information according to the first embodiment. [Figure 6A] FIG. 6A is a diagram showing an example of displaying an abnormal region according to the first embodiment. [Figure 6B]FIG. 6B is a diagram showing an example of displaying an abnormal region according to the first embodiment. [Figure 6C] FIG. 6C is a diagram showing an example of displaying an abnormal region according to the first embodiment. [Figure 6D] FIG. 6D is a diagram showing an example of displaying an abnormal region according to the first embodiment. [Figure 6E] FIG. 6E is a diagram showing an example of displaying an abnormal region according to the first embodiment. [Figure 6F] FIG. 6F is a diagram showing an example of displaying an abnormal region according to the first embodiment. [Figure 7] FIG. 7 is a diagram showing an example of display information according to the first embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0007] Hereinafter, embodiments of a medical information processing apparatus and a medical information processing method will be described in detail with reference to the drawings. Note that the medical information processing apparatus and the medical information processing method according to the present application are not limited to the embodiments described below.

[0008] (First embodiment) Fig. 1 is a diagram showing an example of the configuration of a medical information processing device according to the first embodiment. For example, as shown in Fig. 1, a medical information processing device 3 according to this embodiment is communicably connected to a medical image diagnostic device 1 and a medical image storage device 2 via a network. Note that various other devices and systems may also be connected to the network shown in Fig. 1.

[0009] The medical image diagnostic device 1 captures an image of a subject to generate a medical image. The medical image diagnostic device 1 then transmits the generated medical image to various devices on a network. For example, the medical image diagnostic device 1 is an X-ray diagnostic device, an X-ray CT (Computed Tomography) device, an MRI (Magnetic Resonance Imaging) device, an ultrasound diagnostic device, a SPECT (Single Photon Emission Computed Tomography) device, a PET (Positron Emission Computed Tomography) device, etc.

[0010] The medical image storage device 2 stores various medical images related to subjects. Specifically, the medical image storage device 2 receives medical images from the medical image diagnostic device 1 via a network, and stores the medical images in a memory circuit within the device. For example, the medical image storage device 2 is realized by a computer device such as a server or a workstation. Furthermore, for example, the medical image storage device 2 is realized by a PACS (Picture Archiving and Communication System) or the like, and stores medical images in a format compliant with DICOM (Digital Imaging and Communications in Medicine).

[0011] The medical information processing device 3 performs various information processing on medical images collected from a subject. Specifically, the medical information processing device 3 receives medical images from the medical image diagnostic device 1 or the medical image storage device 2 via a network, and performs various information processing using the medical images. For example, the medical information processing device 3 is realized by computer equipment such as a server or a workstation.

[0012] For example, the medical information processing device 3 includes a communication interface 31, an input interface 32, a display 33, a memory circuit 34, and a processing circuit 35.

[0013] The communication interface 31 controls the transmission and communication of various data sent and received between the medical information processing device 3 and other devices connected via a network. Specifically, the communication interface 31 is connected to the processing circuitry 35, and transmits data received from other devices to the processing circuitry 35, or transmits data received from the processing circuitry 35 to other devices. For example, the communication interface 31 is realized by a network card, a network adapter, a NIC (Network Interface Controller), or the like.

[0014] The input interface 32 accepts input operations of various instructions and information from a user. Specifically, the input interface 32 is connected to the processing circuit 35, converts the input operations received from the user into electrical signals, and transmits the electrical signals to the processing circuit 35. For example, the input interface 32 may be realized by a trackball, a switch button, a mouse, a keyboard, a touchpad that performs input operations by touching the operation surface, a touchscreen that integrates a display screen and a touchpad, a non-contact input interface using an optical sensor, a voice input interface, etc. Note that in this specification, the input interface 32 is not limited to those that have physical operation components such as a mouse and a keyboard. For example, an electrical signal processing circuit that receives electrical signals corresponding to input operations from an external input device provided separately from the device and transmits the electrical signals to a control circuit is also included as an example of the input interface 32.

[0015] The display 33 displays various types of information and data. Specifically, the display 33 is connected to the processing circuit 35 and displays various types of information and data received from the processing circuit 35. For example, the display 33 is realized by a liquid crystal display, a CRT (Cathode Ray Tube) display, a touch panel, or the like.

[0016] The memory circuitry 34 stores various data and programs. Specifically, the memory circuitry 34 is connected to the processing circuitry 35 and stores data received from the processing circuitry 35, or reads out stored data and transmits it to the processing circuitry 35. For example, the memory circuitry 34 is realized by a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, a hard disk, an optical disk, or the like.

[0017] The processing circuitry 35 controls the entire medical information processing device 3. For example, the processing circuitry 35 performs various processes in response to input operations received from a user via the input interface 32. For example, the processing circuitry 35 receives data transmitted by another device via the communication interface 31 and stores the received data in the storage circuitry 34. Also, for example, the processing circuitry 35 transmits data received from the storage circuitry 34 to the communication interface 31, thereby transmitting the data to another device. Also, for example, the processing circuitry 35 displays the data received from the storage circuitry 34 on the display 33.

[0018] The above describes an example of the configuration of the medical information processing device 3 according to this embodiment. For example, the medical information processing device 3 according to this embodiment is installed in a medical facility such as a hospital or a clinic, and supports various diagnoses and the formulation of treatment plans performed by users such as doctors. Specifically, the medical information processing device 3 identifies inaccurate results in calculating pancreatic feature amounts performed on medical images, and corrects the identified results.

[0019] In early-stage pancreatic cancer, where the tumor itself is difficult to observe in medical images, findings such as pancreatic atrophy and pancreatic duct dilation may be detected. Detecting such findings requires accurate segmentation of the pancreas itself and the pancreatic duct within the pancreas. However, due to the structure of the pancreas, segmentation may be inaccurate. In such cases, feature values ​​related to the pancreas and pancreatic duct cannot be calculated as expected, resulting in inaccurate detection of abnormalities using the feature values. For example, if the pancreatic duct segmentation is inaccurate in a specific area of ​​the pancreas, the pancreatic duct diameter may also be inaccurate, potentially resulting in overlooking findings such as pancreatic duct dilation.

[0020] Therefore, in this embodiment, in order to prevent overlooking abnormalities in areas where segmentation is poor or, conversely, detecting them as false positives, the accuracy of analysis using medical images is improved by identifying positions where the feature values ​​calculated for the pancreas may be inaccurate. Below, a medical information processing device 3 having such a configuration will be described in detail.

[0021] For example, as shown in FIG. 1 , in this embodiment, the processing circuitry 35 of the medical image processing device 3 executes a control function 351, an image acquisition function 352, an estimation function 353, a feature amount calculation function 354, a correction target position identification function 355, a feature amount correction function 356, and an abnormality detection function 357. Here, the control function 351 is an example of a display control unit. The feature amount calculation function 354 is an example of a calculation unit. The correction target position identification function 355 is an example of an identification unit. The feature amount correction function 356 is an example of a correction unit. The abnormality detection function 357 is an example of a detection unit.

[0022] The control function 351 generates various GUIs (Graphical User Interfaces) and various display information in response to operations performed via the input interface 32, and controls the display 33 to display them. For example, the control function 351 causes the display 33 to display the results of processing by each function. The control function 351 can also generate and display various display images based on medical images acquired by the image acquisition function 352. The display information displayed by the control function 351 will be described in detail later.

[0023] The image acquisition function 352 acquires medical images of the subject from the medical image diagnostic device 1 or the medical image storage device 2 via the communication interface 31. Specifically, the image acquisition function 352 acquires medical images including three-dimensional or two-dimensional morphological information of the pancreas.

[0024] The image acquisition function 352 acquires, as the above-mentioned medical images, CT images, ultrasound images, MRI images, X-ray images (for example, endoscopic retrograde cholangiopancreatography (ERCP) images, etc.) etc. By executing the above-mentioned image acquisition function 352, the processing circuitry 35 receives medical images of the subject from the medical image diagnostic device 1 or the medical image storage device 2 and stores the received medical images in the memory circuitry 34.

[0025] The estimation function 353 estimates the centerline from the head to the tail of the pancreas from the medical image acquired by the image acquisition function 352. The processing by the estimation function 353 will be described in detail later.

[0026] The feature calculation function 354 calculates a first feature, which is a feature for each position in a first region of the pancreas included in the medical image, and a second feature, which is a feature for each position in a second region of the pancreas that is different from the first region. Specifically, the feature calculation function 354 calculates the first feature and the second feature for each position based on a standard set for the pancreas. The processing by the feature calculation function 354 will be described in detail later.

[0027] The correction target position specifying function 355 specifies the correction target position of the calculated feature amount. Specifically, the correction target position specifying function 355 specifies the correction target position based on the feature amount calculated by the feature amount calculation function 354 or information on the anatomical structure of the pancreas. The processing by the correction target position specifying function 355 will be described in detail later.

[0028] The feature amount correction function 356 corrects the feature amount at the correction target position. The processing by the feature amount correction function 356 will be described in detail later.

[0029] The abnormality detection function 357 detects an abnormality in the pancreas using at least the feature amount at the correction target position after correction, out of the first feature amount and the second feature amount. Note that the processing by the abnormality detection function 357 will be described in detail later.

[0030] The processing circuitry 35 described above is realized by, for example, a processor. In this case, each of the processing functions described above is stored in the storage circuitry 34 in the form of a program executable by a computer. The processing circuitry 35 then reads and executes each program stored in the storage circuitry 34 to realize the function corresponding to each program. In other words, the processing circuitry 35 has each of the processing functions shown in FIG. 1 when each program has been read.

[0031] Next, the processing procedure by the medical information processing device 3 will be explained using Fig. 2, and then each process will be explained in detail. Fig. 2 is a flowchart showing the processing procedure of the processes performed by each processing function of the processing circuitry 35 of the medical information processing device 3 according to the first embodiment. Note that Fig. 2 shows a case where the center line from the head to the tail of the pancreas is used as the reference set for the pancreas, but the embodiment is not limited to this, and other references may be set.

[0032] 2, in this embodiment, the image acquisition function 352 acquires a medical image of a subject from the medical image diagnostic apparatus 1 or the medical image storage apparatus 2 (step S101). For example, the image acquisition function 352 acquires a medical image including morphological information of the three-dimensional anatomical structure of the pancreas in response to an operation for acquiring a medical image via the input interface 32. This process is realized, for example, by the processing circuitry 35 calling up a program corresponding to the image acquisition function 352 from the storage circuitry 34 and executing it.

[0033] Next, the estimation function 353 estimates the centerline of the pancreas included in the acquired medical image (step S102). This process is realized, for example, by the processing circuitry 35 calling up and executing a program corresponding to the estimation function 353 from the storage circuitry 34.

[0034] Next, the feature calculation function 354 calculates features related to the pancreas based on the estimated core line (step S103). This process is realized, for example, by the processing circuitry 35 calling up a program corresponding to the feature calculation function 354 from the storage circuitry 34 and executing it.

[0035] Next, the correction target position specifying function 355 specifies the correction target position (step S104). This process is realized, for example, by the processing circuitry 35 calling up a program corresponding to the correction target position specifying function 355 from the storage circuitry 34 and executing the program.

[0036] Next, the feature amount correction function 356 corrects the feature amount at the correction target position identified in step S104 (step S105). This process is realized, for example, by the processing circuitry 35 calling up a program corresponding to the feature amount correction function 356 from the storage circuitry 34 and executing it.

[0037] Subsequently, the abnormality detection function 357 detects an abnormality related to the pancreas based on the corrected feature amount (step S106). This process is realized, for example, by the processing circuitry 35 calling up and executing a program corresponding to the abnormality detection function 357 from the storage circuitry 34.

[0038] Next, the control function 351 displays the display information (step S107). This process is realized, for example, by the processing circuitry 35 calling up a program corresponding to the control function 351 from the storage circuitry 34 and executing the program.

[0039] The following describes in detail each process executed by the medical information processing device 3. Note that, although the following describes a case where the reference set for the pancreas is a centerline, the reference is not limited to this. For example, a reference line or a reference plane set by a user may also be used. In addition, the following describes an example where a CT image captured by an X-ray CT device is acquired as a three-dimensional medical image.

[0040] (Medical image acquisition processing) As described in step S101 of FIG. 2, the image acquisition function 352 acquires a medical image including three-dimensional morphological information of the pancreas in response to a medical image acquisition operation via the input interface 32. For example, the image acquisition function 352 acquires a CT image of the pancreas captured in three dimensions. Note that, if the feature amount calculated in the subsequent processing is of a type that can be calculated from a two-dimensional image, the medical image acquired by the image acquisition function 352 may be an image of the pancreas captured in two dimensions. For example, an ultrasound image, an ERCP image, or the like can be used as the image of the pancreas captured in two dimensions.

[0041] (Core line estimation process) As described in step S102 of FIG. 2, the estimation function 353 estimates a centerline of the pancreas from the head to the tail of the pancreas for the medical image acquired by the image acquisition function 352. Specifically, the estimation function 353 acquires coordinate information of pixels representing the pancreas in the three-dimensional CT image and acquires coordinate information of the centerline from the acquired coordinate information. Here, the estimation function 353 can acquire the above coordinate information using various methods. For example, the estimation function 353 can acquire a region specified on the CT image via the input interface 32 as the pancreas, and further acquire a reference line specified on the region as the centerline. In other words, the estimation function 353 acquires the region and reference line manually specified by the user as the pancreas and the centerline, respectively.

[0042] Furthermore, for example, the estimation function 353 can extract the pancreas based on the anatomical structure depicted in the CT image using a known region extraction technique and estimate the centerline of the extracted pancreas. For example, the estimation function 353 extracts the pancreas from the CT image using Otsu's binarization method based on the CT value, a region growing method, a snake algorithm, a graph cut algorithm, a mean shift algorithm, or the like. Then, the estimation function 353 estimates the centerline of the extracted pancreas using a known thinning process, a Dijkstra algorithm, or the like.

[0043] Furthermore, for example, the estimation function 353 can estimate the region (coordinate information) and centerline (coordinate information) of the pancreas in a CT image using a trained model of the pancreas constructed based on training data prepared in advance using machine learning technology (including deep learning). Note that the above-described estimation process of the centerline of the pancreas does not have to target the entire image. For example, a region that includes the pancreas but is smaller than the entire image may be identified, and the above-described known region extraction technology or machine learning technology may be applied to only the identified region.

[0044] An example of estimating the centerline of the pancreas will now be described with reference to Fig. 3. Fig. 3 is a diagram for explaining processing by the estimation function 353 according to the first embodiment. Note that in Fig. 3, the left side of the figure indicates the head side of the pancreas, and the right side indicates the tail side of the pancreas. Also, in Fig. 3, the centerline of the pancreas is illustrated from the head to the tail of the pancreas, but this is a representation for the purpose of illustration, and the actual centerline is a three-dimensional curve and does not necessarily exist on the same plane.

[0045] For example, the estimation function 353 extracts a region R1 representing the pancreas in a CT image, and estimates a centerline L1 from the head to the tail of the extracted pancreas, as shown in Fig. 3. Here, the centerline L1 passes through the inside of a region R2 representing the main pancreatic duct, as shown in Fig. 3. The centerline L1 is also called a running line or a center line.

[0046] The estimation function 353 can estimate any reference (reference line, reference plane, etc.) for the pancreas other than the above-mentioned center line.

[0047] (Feature calculation process) 2, feature amount calculation function 354 calculates feature amounts related to the pancreas extracted by estimation function 353. Specifically, feature amount calculation function 354 calculates feature amounts (first feature amounts) for each position in a first region of the pancreas and feature amounts (second feature amounts) for each position in a second region of the pancreas. For example, feature amount calculation function 354 calculates feature amounts from the entire pancreas as the first region and the pancreatic duct as the second region.

[0048] Here, the feature calculation function 354 calculates the first feature and the second feature based on a criterion set for the pancreas. For example, the feature calculation function 354 calculates a feature for the entire pancreas and a feature for the pancreatic duct at each position along the center line of the pancreas estimated by the estimation function 353. As an example, the feature calculation function 354 calculates the cross-sectional area of ​​the pancreas and the cross-sectional area of ​​the pancreatic duct in a cross section perpendicular to the center line at each position along the center line from the head to the tail of the pancreas.

[0049] 3, the centerline L1 estimated by the estimation function 353 passes through the inside of the region R2 of the main pancreatic duct. Therefore, each cross section (short-axis cross section of the pancreas) perpendicular to the centerline L1 includes a cross section of the entire pancreas and a cross section of the pancreatic duct within the cross section of the entire pancreas. The feature calculation function 354 calculates the feature of the entire pancreas and the feature of the pancreatic duct for each of these cross sections.

[0050] Here, the feature calculation function 354 can calculate various information as features. For example, the feature calculation function 354 can calculate first-order statistics (average, standard deviation, variance, entropy, kurtosis, skewness, percentile value, median, mode, etc. of pixel values ​​or signal values), shape features (area, volume, diameter, circumference, circularity, sphericity, volume / surface area ratio), texture features (calculating first-order statistics from matrix information such as GLCM, Runlength, NGTDM), composite features (such as feature ratios) obtained by combining features through arithmetic operations, etc.

[0051] For example, the feature calculation function 354 sets a cross section perpendicular to the centerline at each coordinate on the centerline, and calculates the cross-sectional area of ​​the pancreas from the coordinates indicating the outline of the pancreas for each set cross section. Similarly, the feature calculation function 354 calculates the cross-sectional area of ​​the pancreas from the coordinates indicating the outline of the pancreatic duct for each set cross section. Hereinafter, a case will be described in which the cross-sectional area of ​​the pancreas and the cross-sectional area of ​​the pancreatic duct are calculated as feature amounts. However, the embodiment is not limited to this, and other feature amounts may be calculated. For example, feature amounts may be calculated for different regions of the pancreatic parenchyma (e.g., a small region in the parenchyma on the head side of the pancreas and a small region in the parenchyma on the tail side of the pancreas), or for different regions of the pancreatic duct. Also, feature amounts may be calculated for a small region including the pancreatic parenchyma and the pancreatic duct. Also, feature amounts may be calculated for a three-dimensional region.

[0052] Furthermore, for example, the feature calculation function 354 can calculate a third feature based on the first feature and the second feature, as in the above-described composite feature. For example, the feature calculation function 354 calculates the third feature by calculating the ratio between the first feature and the second feature. That is, when the first feature and the second feature are the cross-sectional area of ​​the pancreas and the cross-sectional area of ​​the pancreatic duct, respectively, the feature calculation function 354 calculates the pancreatic duct-to-pancreas cross-sectional area ratio (hereinafter also referred to as D / P ratio), which is the ratio between the cross-sectional area of ​​the pancreas duct and the cross-sectional area of ​​the pancreas (or the cross-sectional area of ​​the pancreatic parenchyma), as the third feature. Hereinafter, the first feature, the second feature, and the third feature may be collectively referred to as features.

[0053] The feature calculation function 354 calculates the various feature amounts described above using the coordinate information of the pancreas and the coordinate information of the centerline acquired by the estimation function 353. Note that feature extraction may also be performed by an autoencoder constructed by a convolutional neural network (CNN). In such a case, the feature calculation function 354 acquires feature amounts by inputting the medical image acquired by the image acquisition function 352 into the autoencoder.

[0054] Here, the medical image (e.g., CT image) for which the feature amount is to be calculated may be subjected to filtering before the feature amount is calculated. For example, after being acquired by the image acquisition function 352 or after being processed by the estimation function 353, the medical image may be subjected to filtering using a Gaussian filter, a low-pass filter, a high-pass filter, a low-pass emphasis filter, a high-pass emphasis filter, a notch filter, a differential filter, a Laplacian filter, a Laplacian of Gaussian filter, a statistical filter (average filter, variance filter, maximum value filter, minimum value filter, median filter), or the like. This may further enhance the features related to the analysis target.

[0055] Furthermore, the medical images acquired by the image acquisition function 352 may be resampled to make pixel intervals or voxel intervals isotropic, to equalize between data sets, or to normalize or standardize pixel values ​​or signal values. Resampling or normalization of signal values ​​may reduce bias between medical images, stabilize learning, and increase the accuracy of inference.

[0056] (Processing for identifying the position to be corrected) As described in step S104 of FIG. 2, the correction target position identifying function 355 identifies a correction target position where the calculated feature amount may be defective. Specifically, the correction target position identifying function 355 identifies a correction target position where the first feature amount and / or the second feature amount calculated by the feature amount calculating function 354 may not be calculated as expected. For example, the correction target position identifying function 355 identifies a correction target position where the calculated cross-sectional area may be defective for at least one of the cross-sectional area of ​​the pancreas and the cross-sectional area of ​​the pancreatic duct. The correction target position identifying function 355 also identifies a correction target position based on a third feature amount calculated by the feature amount calculating function 354. That is, the correction target position identifying function 355 identifies a correction target position based on a third feature amount calculated based on the first feature amount and the second feature amount (for example, a D / P ratio calculated from the ratio between the cross-sectional area of ​​the pancreatic duct and the cross-sectional area of ​​the pancreas). Here, the correction target position identifying function 355 identifies a correction target position based on various conditions. These will be described below.

[0057] For example, the correction target position identifying function 355 can identify the correction target position based on a comparison of the feature amounts (first feature amount, second feature amount, and third feature amount) calculated by the feature amount calculating function 354 with a reference value. In this case, the correction target position identifying function 355 identifies the correction target position by performing threshold processing using the reference value as a threshold. For example, the correction target position identifying function 355 compares the cross-sectional area of ​​the pancreas with a threshold for each coordinate on the core line, and identifies a position (coordinate) where the cross-sectional area of ​​the pancreas does not reach the reference value or a position (coordinate) where the cross-sectional area of ​​the pancreas exceeds the reference value as the correction target position of the pancreas. Similarly, the correction target position identifying function 355 compares the cross-sectional area of ​​the pancreatic duct with a threshold for each coordinate on the core line, and identifies a position (coordinate) where the cross-sectional area of ​​the pancreatic duct does not reach the reference value or a position (coordinate) where the cross-sectional area of ​​the pancreatic duct exceeds the reference value as the correction target position of the pancreas.

[0058] Here, the above-mentioned reference values ​​can be set individually for the cross-sectional area of ​​the pancreas and the cross-sectional area of ​​the pancreatic duct. For example, the reference values ​​may be determined by a designer based on literature values ​​or the possible ranges of feature quantities (e.g., cross-sectional areas) of healthy or abnormal cases. Alternatively, the reference values ​​may be determined based on the values ​​of feature quantities of specific anatomical structures in the image. Furthermore, the reference values ​​may be changed based on image conditions and clinical information of the patient.

[0059] For example, when the correction target position specifying function 355 specifies the correction target position of the pancreatic duct based on the cross-sectional area, if the cross-sectional area is “0 mm 2 " or "0mm 2 The position corresponding to "0 mm" is specified as the position to be corrected. 2 " equivalent to, for example, "1 mm 2 " means the following:

[0060] Furthermore, for example, when the correction target position identification function 355 identifies the correction target position of the pancreatic duct based on the cross-sectional area, it identifies as the correction target position a position where the cross-sectional area of ​​the pancreatic duct is smaller than the average cross-sectional area of ​​a certain section of the pancreatic duct of the patient.

[0061] Furthermore, for example, when the correction target position specifying function 355 specifies the correction target position of the pancreatic duct based on the cross-sectional area, it specifies a position where the cross-sectional area of ​​the pancreatic duct is larger than the cross-sectional area of ​​the pancreas as the correction target position. Because the pancreatic duct is located inside the pancreas, the cross-sectional area of ​​the pancreatic duct will never be larger than the cross-sectional area of ​​the entire pancreas. However, for example, when the pancreas and the pancreatic duct are extracted separately from a medical image, the above-mentioned result may occur due to an extraction error.

[0062] Furthermore, for example, when the correction target position specifying function 355 specifies the correction target position of the pancreas based on the cross-sectional area, it specifies as the correction target position a position where the cross-sectional area of ​​the pancreas is less than a first reference value (threshold) or a position where the cross-sectional area of ​​the pancreas exceeds a second reference value (threshold). The pancreas may shrink with age or enlarge due to pancreatitis or the like. Therefore, it is preferable to set the first reference value to be smaller as the patient ages, and to set the second reference value to be higher if there is an underlying disease that may lead to pancreatic enlargement.

[0063] Furthermore, the reference value may be set to a range in comparison with the reference value. For example, a range of values ​​that can be assumed in the normal case is set for each position based on feature values ​​obtained from healthy subjects, and a range of values ​​that can be assumed in the abnormal case is set for each position based on feature values ​​obtained from subjects with a disease or medical condition. The correction target position identification function 355 compares the feature values ​​for each position calculated by the feature value calculation function 354 with the normal numerical range for the same position. Similarly, the correction target position identification function 355 compares the feature values ​​for each position calculated by the feature value calculation function 354 with the abnormal numerical range for the same position. Here, if the compared feature values ​​are not included in either the normal numerical range or the abnormal numerical range, the correction target position identification function 355 identifies the position of the compared feature value as the correction target position. The normal numerical range and the abnormal numerical range can be estimated using the feature values ​​for each position obtained from healthy subjects and the feature values ​​for each position obtained from subjects with a disease or medical condition.

[0064] In the above example, a case where a single feature value for each coordinate is compared with a reference value has been described, but the embodiment is not limited to this, and for example, a case where a feature value within a region including multiple positions (coordinates) on the core line is compared with a reference value may also be used. In such a case, for example, a case where an average value of multiple feature values ​​within the region is compared with a reference value may also be used.

[0065] The correction target position identifying function 355 can also identify correction target positions based on changes in the feature amounts (first feature amount, second feature amount, and third feature amount) for each position. In this case, the correction target position identifying function 355 compares the feature amount values ​​at each position with the surrounding feature amount values ​​to identify positions where the feature amount values ​​suddenly increase or decrease, or other positions where the feature amount changes anatomically unnatural, as correction target positions. The correction target position identifying function 355 can identify positions where the value changes suddenly by determining whether the difference in the feature amount values ​​between adjacent positions is greater than or equal to a certain level, or by determining whether the rate of change or slope of the feature amount is greater than or equal to a certain level. The correction target position identifying function 355 can also identify a region as a correction target if the standard deviation, variance, or coefficient of variation of the feature amounts around the position is greater than or equal to a certain level. For example, if the pancreatic duct cross-sectional area is used as a feature, the correction target position identifying function 355 can also identify a region where the pancreatic duct cross-sectional area decreases at a rate greater than or equal to a certain level when tracing and searching the pancreatic duct cross-sectional area from the head to the tail of the pancreas.

[0066] For example, when the correction target position specifying function 355 specifies the correction target position of the pancreatic duct based on the cross-sectional area, the cross-sectional area of ​​the pancreatic duct is determined to be 30 mm when the pancreatic duct moves 1 mm from the head side to the tail side of the pancreas. 2 The portion that changes by more than this is identified as the position to be corrected.

[0067] Furthermore, for example, when the correction target position specifying function 355 specifies the correction target position of the pancreatic duct based on the cross-sectional area, it specifies a portion of the cross-sectional area of ​​the pancreatic duct where the coefficient of variation is equal to or greater than a certain value as the correction target position.

[0068] Furthermore, the correction target position specifying function 355 can specify the position to be corrected based on the shape of the pancreas. In this case, the correction target position specifying function 355 specifies the position to be corrected by analyzing the morphological state (coordinate information) of the pancreas acquired by the estimation function 353. For example, the correction target position specifying function 355 acquires information on the extension and running direction of the anatomical structure and the shape of the surface from the morphological state (coordinate information) of the pancreas, and compares the acquired information with functional information on the anatomical structure of the pancreas to specify an area that is progressing in an incorrect direction or has an incorrect shape as the position to be corrected.

[0069] Here, the correction target position specifying function 355 can also specify the correction target position using a trained model generated by deep learning using, as training data, a medical image including morphological information of the pancreas and information on whether the morphological information is invalid. That is, the correction target position specifying function 355 specifies the correction target position by inputting the medical image acquired by the image acquisition function 352 into the trained model.

[0070] For example, the correction target position specifying function 355 obtains the degree of meandering of the centerline by calculating the sum of changes in the orientation of the centerline of the pancreas, and determines a region where the obtained degree of meandering exceeds a threshold as a region where the centerline of the pancreas meanders abnormally, and specifies that region as a position to be corrected. In a region where the pancreas meanders, the calculation of the cross-sectional area of ​​the pancreatic or pancreatic duct may not be as expected, but by specifying the region where the pancreas meanders as a position to be corrected as described above, it is possible to correct the feature amount of that region.

[0071] Furthermore, for example, the correction target position specifying function 355 acquires the contour line of the pancreas and the contour line of the pancreatic duct based on the morphological information (coordinate information) of the pancreas and the morphological information (coordinate information) of the pancreas and the pancreatic duct acquired by the estimation function 353. Then, the correction target position specifying function 355 specifies an area with many irregularities on each acquired contour line as a correction target position.

[0072] The correction target position identifying function 355 also identifies the correction target position based on the positional relationship between the pancreas and surrounding structures. In this case, the correction target position identifying function 355 acquires information on structures other than the anatomical structures (pancreas, pancreatic duct) targeted for feature calculation, and identifies the correction target position based on their positional relationship. Here, structures other than the anatomical structures (pancreas, pancreatic duct) targeted for feature calculation include, for example, organ tissues other than the target organ tissue, anatomical structures such as blood vessels, vasculature, bone, cartilage, tendons, muscles, and visceral fat, as well as structures resulting from lesions such as tumors and calcifications, and artificial structures such as catheters and stents. The correction target position identifying function 355 identifies, as the correction target position, a position (area) near a structure that may adversely affect the results of obtaining morphological information (coordinate information) about the liver and the calculation of features.

[0073] For example, in a region where a blood vessel is adjacent to the pancreas, the blood vessel may be mistaken for the pancreas, resulting in an incorrect segmentation of the pancreas. Therefore, the correction target position specifying function 355 specifies, for example, a region where the aorta, the superior mesenteric artery, or the superior mesenteric vein is adjacent to the pancreas as a correction target position.

[0074] Furthermore, for example, the correction target position specifying function 355 specifies an area included within a range of a certain distance from the bile duct stent as a correction target position.

[0075] Furthermore, for example, the correction target position specifying function 355 specifies an area at a certain distance from the pancreatic stone as the correction target position.

[0076] Furthermore, when collecting CT images, the patient's arm is usually raised, but sometimes the patient's arm is lowered due to shoulder problems or other reasons. In such cases, the CT value of the organ tissue between the spine and the arm may change. Therefore, the correction target position identification function 355 identifies the area of ​​the pancreas between the arm and the spine as the correction target position.

[0077] (Feature correction processing) As explained in step S105 of FIG. 2, the feature amount correction function 356 corrects the feature amount of the correction target position identified by the correction target position identification function 355. Here, the feature amount correction function 356 can correct the feature amount of the correction target position using various methods. These will be explained below with reference to FIGS. 4A to 4D. FIGS. 4A to 4D are diagrams for explaining an example of the correction process according to the first embodiment. Here, in FIGS. 4A to 4D, the cross-sectional area value of the pancreatic duct for each position is indicated by numbers in multiple rectangles arranged horizontally. That is, each rectangle indicates the position where the cross-sectional area was acquired. Furthermore, the curves shown on the rectangles are schematic representations of the feature amount values.

[0078] For example, the feature amount correction function 356 can correct the feature amounts (first feature amount, second feature amount, and third feature amount) at the correction target position by replacing the feature amounts (first feature amount, second feature amount, and third feature amount) at the correction target position with reference values. In this case, the feature amount correction function 356 replaces the feature amounts at the correction target position with predetermined values. Here, the predetermined values ​​include, for example, typical values ​​of feature amounts for healthy cases (normal cases) and typical values ​​of feature amounts for abnormal cases (cases with disease or findings).

[0079] For example, as shown in FIG. 4A, the feature amount correction function 356 sets the cross-sectional area of ​​the pancreatic duct at each position to “0 mm 2 ", the values ​​of the correction target position (area including the three correction target positions) are set to "10mm 2 Here, the cross-sectional area of ​​the pancreatic duct is replaced with "10 mm 2 " is a typical value for the cross-sectional area of ​​the pancreatic duct in a healthy individual.

[0080] Furthermore, the feature amount correction function 356 can correct the feature amounts (first feature amount, second feature amount, and third feature amount) of the correction target position by predicting the feature amounts (first feature amount, second feature amount, and third feature amount) of the correction target position based on the feature amounts around the correction target position. In this case, the feature amount correction function 356 predicts the value of the feature amount at the correction target position using linear interpolation, spline interpolation, polynomial interpolation, or Bezier curve based on the values ​​around the correction target position. Note that to prevent the value of the feature amount obtained by interpolation or the like from becoming an anatomically incorrect value, upper and lower limits may be set for the corrected value. Furthermore, a typical value of the feature amount may be added as a limiting term to the interpolation formula, and the predicted value may be controlled to approach the typical value.

[0081] For example, as shown in FIG. 4B, the feature amount correction function 356 sets the cross-sectional area of ​​the pancreatic duct at each position to “0 mm 2 " is the value obtained by linearly interpolating each value of the correction target position (the area including the three correction target positions) corresponding to "13mm 2 ", "14mm 2 ", "15mm 2 " is replaced with ".

[0082] Furthermore, for example, the feature amount correction function 356 can correct the feature amounts (first feature amount, second feature amount, and third feature amount) at the correction target position by acquiring heterogeneous feature amounts different from the feature amount for the correction target region and predicting the feature amounts (first feature amount, second feature amount, and third feature amount) at the correction target position based on the acquired heterogeneous feature amounts. In this case, the feature amount correction function 356 may predict the value of the feature amount at the correction target position using another feature amount (a feature amount other than the cross-sectional area) calculated at the correction target position. For example, if the cross-sectional area of ​​the pancreatic duct is "0 mm 2 In the region of the correction target position corresponding to ", the feature amount correction function 356 predicts the cross-sectional area of ​​the pancreatic duct using a weighted sum of the pixel values ​​of the pancreas in the same region and the circularity of the pancreas. The weights at this time can be obtained by regressing the cross-sectional area of ​​the pancreatic duct using a large number of training cases.

[0083] Furthermore, for example, the feature amount correction function 356 can correct the feature amounts (first feature amount, second feature amount, and third feature amount) of the correction target position by replacing the feature amounts (first feature amount, second feature amount, and third feature amount) of the correction target position with a set numerical range. In this case, the feature amount correction function 356 replaces the feature amounts of the correction target position with a numerical range for normal cases or a numerical range for abnormal cases.

[0084] For example, as shown in FIG. 4C, the feature amount correction function 356 sets the cross-sectional area of ​​the pancreatic duct at each position to “0 mm 2 The values ​​of the correction target positions (areas including the three correction target positions) corresponding to the cross-sectional area of ​​the pancreatic duct in normal cases were set to the range of 5 to 30 mm 2 " is replaced with ".

[0085] Furthermore, for example, the feature amount correction function 356 can correct the feature amounts (first feature amount, second feature amount, and third feature amount) at the correction target position by replacing the feature amounts (first feature amount, second feature amount, and third feature amount) at the correction target position with frequency distributions corresponding to the feature amounts. In this case, first, a frequency distribution of actual values ​​is obtained for each feature amount value calculated by the feature amount calculation function 354. Specifically, correct values ​​of the feature amounts calculated by the feature amount calculation function 354 are obtained from multiple subjects, and a frequency distribution of the correct values ​​is generated. Here, frequency distributions are generated for healthy subjects and subjects with diseases or findings, respectively. The feature amount correction function 356 replaces the feature amounts at the correction target position with frequency distributions.

[0086] For example, if the cross-sectional area of ​​the pancreatic duct calculated by the feature amount calculation function 354 is “0 mm 2 ", the correct value of the cross-sectional area of ​​the pancreatic duct is "0 mm 2 "," 1mm 2 "," 2mm 2 For example, as shown in the histogram in Figure 4D, the frequency of the cross-sectional area of ​​the pancreatic duct is 0 mm 2A histogram is generated showing the distribution of true values ​​of the cross-sectional area of ​​the pancreatic duct when the cross-sectional area corresponds to "." Here, as shown in FIG. 4D, this histogram includes a histogram L2 of a healthy subject and a histogram L3 of a subject with a disease or condition.

[0087] The feature amount correction function 356 corrects the cross-sectional area of ​​the pancreatic duct at each position by "0 mm 2 The correction is performed by replacing each value of the correction target position (area including three correction target positions) corresponding to " with the above-mentioned histogram. Note that the histogram may be a probability density function.

[0088] (Abnormality detection process) As described in step S106 of FIG. 2, the abnormality detection function 357 uses the features corrected by the feature correction function 356 to detect abnormalities in the analysis target (tendencies or distributions of features not seen in healthy cases).

[0089] Here, the abnormality detection function 357 detects at least one of the presence of a lesion, the presence of a tumor, the presence of a finding, and the presence of a pathological condition as an abnormality. For example, the abnormality detection function 357 can detect pancreatic malignant tumors, pancreatic duct dilation, pancreatic atrophy, chronic pancreatitis, etc. as abnormalities in the pancreas.

[0090] Furthermore, the anomaly detection function 357 detects at least one of the presence or absence of an anomaly, the degree of the anomaly, the classification of the anomaly, and the probability of the anomaly as the anomaly detection result. That is, the anomaly detection function 357 can output anomaly determination results in various variations. For example, the anomaly detection function 357 can perform a binary determination such as "abnormal" or "no abnormality," a three-level or more determination such as "no abnormality" or "suspected abnormality" or "clearly abnormality," a multi-class anomaly classification determination such as "no abnormality" or "abnormality (Type A)" or "abnormality (Type B)" or "abnormality (Type Z)," or a determination based on a continuous value such as the probability value of an anomaly.

[0091] For example, in the case of the pancreas, the abnormality detection function 357 may output the presence or absence of a pancreatic malignant tumor, the type of pancreatic tumor (PDAC, IPMN, MCN, SPN, SCN, other cystic masses, other solid masses), the probability of the presence of a pancreatic malignant tumor, the probability of the presence of each type of pancreatic tumor (PDAC 10%, IPMN 20%, etc.), etc.

[0092] For example, the abnormality detection function 357 can detect an abnormality in the pancreas based on a comparison between the feature amount at the correction target position after it has been corrected to a single value and a threshold value. In this case, the abnormality detection function 357 compares the feature amount calculated by the feature amount calculation function 354 and the feature amount corrected by the feature amount correction function 356 with the threshold value, and detects an abnormality when the value of the feature amount exceeds a certain threshold value or when it is below a certain threshold value.

[0093] For example, the anomaly detection function 357 detects a pancreatic duct whose cross-sectional area is 30 mm 2 " or more (or a region including a plurality of positions) is determined to be abnormal. 2 " or a region including multiple locations) is normal, and the cross-sectional area of ​​the pancreatic duct is "30 mm 2 " or "49mm 2 " or a region including multiple locations is suspected of being abnormal, and the cross-sectional area of ​​the pancreatic duct is "50mm 2 " or more (or an area including multiple positions) is determined to be abnormal.

[0094] Furthermore, for example, the abnormality detection function 357 can detect an abnormality in the pancreas based on a combination of the first feature amount and the second feature amount after the feature amount at the correction target position for at least one of the first feature amount and the second feature amount has been corrected to a single value. In this case, the abnormality detection function 357 detects an abnormality by combining the multiple feature amounts using arithmetic operations (for example, determining the ratio between the first feature amount and the second feature amount) and performing threshold processing on the value of the combined feature amount.

[0095] For example, the abnormality detection function 357 calculates the ratio of the cross-sectional area of ​​the pancreatic duct to the cross-sectional area of ​​the pancreas (or the cross-sectional area of ​​the pancreatic parenchyma) for each position on the core line to obtain the pancreatic duct-to-pancreas cross-sectional area ratio (D / P ratio). The abnormality detection function 357 detects a portion where the pancreatic duct-to-pancreas cross-sectional area ratio is a certain value (for example, 0.1 or more) as an abnormal region.

[0096] Furthermore, for example, the abnormality detection function 357 can detect an abnormality in the pancreas based on a combination of a feature amount obtained after the feature amount at the correction target position of at least one of the first feature amount and the second feature amount has been corrected to a single value, and a heterogeneous feature amount different from the first feature amount and the second feature amount. In this case, the abnormality detection function 357 acquires multiple types of feature amounts (for example, cross-sectional area and perimeter) for each position, and detects an abnormality using the acquired multiple types of feature amounts.

[0097] For example, the abnormality detection function 357 detects abnormalities using multiple types of feature amounts calculated on the cross section at each position on the core line. Also, for example, the abnormality detection function 357 can detect abnormalities by combining non-image information (clinical information such as age and gender, results of blood tests, biochemical tests, and genetic tests, family history, medical history, etc.) in addition to the feature amounts calculated on the cross section at each position on the core line.

[0098] When combining multiple types of feature quantities as described above, the anomaly detection function 357 detects anomalies by, for example, inputting the feature quantities into a trained model that has been trained to output information about anomalies in response to the input of the feature quantities. Examples of machine learning models include regression models (linear regression, least-squares regression, logistic regression, Lasso regression, etc.), support vector machines, decision trees, XGBoost, random forests, clustering, etc. The regression coefficients and weights of the feature quantities may be determined based on the values ​​of the feature quantities of a large amount of training data, or may be determined by optimizing the operation of the medical image processing device according to this embodiment, as in reinforcement learning.

[0099] Furthermore, for example, when the feature amount at the correction target position is corrected to fall within a set numerical range, the abnormality detection function 357 can detect an abnormality in the pancreas based on the upper or lower limit value of the set numerical range. In this case, the abnormality detection function 357 detects an abnormality based on the criterion set for the numerical range. For example, the abnormality detection function 357 determines whether an abnormality exists based on whether the corrected numerical range corresponds to a normal example.

[0100] Furthermore, for example, when the feature amount at the correction target position is corrected to a frequency distribution corresponding to the feature amount, the abnormality detection function 357 can detect an abnormality in the pancreas based on the frequency of the abnormality in the frequency distribution. In this case, the abnormality detection function 357 calculates the frequency at which an abnormality is determined to exist in the replaced frequency distribution, and detects an abnormality in the pancreas based on the calculated frequency.

[0101] For example, referring to Fig. 4D, the abnormality detection function 357 calculates the ratio of the frequency of histogram L3 of a subject with a disease or finding to the overall frequency, and detects an abnormality if the calculated ratio is equal to or greater than a threshold. For example, in Fig. 4D, since the ratio of the frequency of histogram L3 to the overall frequency is low, the abnormality detection function 357 determines that there is no abnormality.

[0102] (Display information display processing) As explained in step S107 of FIG. 2, the control function 351 causes the display 33 to display display information indicating the first feature amount and the second feature amount of the pancreas. Here, the control function 351 can display various information as the display information. These will be explained below with reference to FIGS. 5A to 5D. FIGS. 5A to 5D are diagrams showing examples of display information according to the first embodiment.

[0103] For example, as shown in FIG. 5A, the control function 351 displays display information in which a stretched multi-planar reconstruction (SPR) image of the pancreas and a graph of feature quantities are arranged side by side. Here, the SPR image of the pancreas is generated from a cross section passing through the centerline, and includes the pancreatic parenchyma R3 and the pancreatic duct R4. The feature quantity graph indicates the position of the pancreas's running line (centerline) on the horizontal axis, and the cross-sectional area value and the D / P ratio value on the vertical axis. For example, as shown in FIG. 5A, the control function 351 generates a graph including a curve L4 indicating the cross-sectional area value of the pancreas, a curve L5 indicating the cross-sectional area value of the pancreatic duct, and a curve L6 indicating the D / P ratio value, and displays the graph on the display 33.

[0104] Here, the control function 351 can display a marker M1 at the position on the SPR image where the D / P ratio is maximum, as shown in Fig. 5A. The control function 351 can also display an image showing the detection results of the pancreatic cancer body below the graph, as shown in Fig. 5A. By displaying in this manner, the user can check the position where the D / P ratio is high on the SPR image, and can also check surrounding structures, such as the presence or absence of a tumor.

[0105] Furthermore, the control function 351 can change the display form of the SPR image of the pancreas depending on the presence or absence of pancreatic duct dilation and pancreatic atrophy. For example, after calculating the D / P ratio, the control function 351 determines the presence or absence of pancreatic duct dilation and pancreatic atrophy, and colors the SPR image according to the determination result. For example, the control function 351 displays the pancreatic parenchyma in blue when pancreatic duct dilation is occurring, displays the pancreatic parenchyma in yellow when pancreatic atrophy is occurring, and displays the pancreatic parenchyma in green when both are occurring.

[0106] For example, in FIG. 5B, the control function 351 displays the pancreatic parenchyma R3 on the SPR image in yellow because the cross-sectional area of ​​the pancreatic duct is below the threshold at the position where the D / P ratio is maximum (the position of the line L7 in the figure). The display format can be changed by coloring the entire region or by changing the color of the outline. By displaying in this manner, the user can immediately confirm whether pancreatic atrophy or pancreatic duct dilation is occurring by observing the pancreatic parenchyma on the SPR image.

[0107] Furthermore, the control function 351 can change and display the projection plane of the SPR image so that the thickness of the pancreatic parenchyma matches the appearance on the SPR image and the appearance on the graph. In this case, the control function 351 calculates the diameter of the pancreas from the graph of the cross-sectional area of ​​the pancreas and rotates the projection plane at each position on the centerline of the pancreas to calculate the diameter of the pancreas. Then, the control function 351 generates an SPR image projected on the projection plane with the angle closest to the diameter of the pancreas for each position. For example, for a position where the pancreatic diameter of "60 mm" is calculated on the graph, the control function 351 selects the projection plane that corresponds to the diameter of the pancreas of "60 mm" in the orthogonal cross section, and generates an SPR image. By performing such a display, the apparent thickness of the pancreas on the SPR image can be matched to the thickness of the pancreas shown on the graph.

[0108] Furthermore, for example, the control function 351 can display display information including the position detected as abnormal. As an example, as shown in Fig. 5D, the control function 351 indicates the position on the SPR image and the position on the graph where the D / P ratio exceeds the threshold as an abnormal region, and further displays a cross-sectional image orthogonal to the center line at the position corresponding to the abnormal region.

[0109] Here, the method for displaying an abnormal region on an image (for example, an SPR image) is not limited to the method shown in Fig. 5D, and various other methods may be used to display the abnormal region. Hereinafter, a method for displaying an abnormal region will be described with reference to Figs. 6A to 6F. Figs. 6A to 6F are diagrams showing examples of displaying an abnormal region according to the first embodiment.

[0110] For example, the control function 351 can display an abnormal region with a marker attached, as shown in FIG. 6A. Furthermore, for example, the control function 351 can display a heat map of the abnormal region, as shown in FIG. 6B. Furthermore, for example, the control function 351 can display a heat map of the abnormal region in multiple colors, as shown in FIG. 6C. Furthermore, for example, the control function 351 can display a bounding box of the abnormal region, as shown in FIG. 6D. Furthermore, for example, the control function 351 can display the outline of the heat map of the abnormal region, as shown in FIG. 6E. Furthermore, for example, the control function 351 can display an image in which only the abnormal region is cropped, as shown in FIG. 6F.

[0111] The control function 351 can also display an area using the feature amount calculated by the feature amount calculation function 354 and an area using the feature amount corrected by the feature amount correction function 356 in a distinguishable manner. In this case, the control function 351 distinguishes the areas by, for example, changing the display format of the SPR image and the graph. Fig. 7 is a diagram showing an example of display information according to the first embodiment.

[0112] For example, as shown in Figure 7, in an SPR image showing the cross-sectional area of ​​the pancreatic duct, the control function 351 displays the color of the pancreatic parenchyma corresponding to the part that uses the feature calculated by the feature calculation function 354 as is (the color of area R3 in the figure) in different colors from the color of the pancreatic parenchyma corresponding to the part that uses the feature corrected by the feature correction function 356 (area R5).

[0113] 7, the control function 351 displays a graph of the cross-sectional area of ​​the pancreatic duct, in which the values ​​of the feature amounts calculated by the feature amount calculation function 354 are indicated by dotted lines, and the values ​​of the feature amounts corrected by the feature amount correction function 356 are indicated by solid lines. Similarly, as shown in FIG. 7, the control function 351 displays a graph of the D / P ratio, in which the values ​​of the D / P ratio calculated using the values ​​of the feature amounts calculated by the feature amount calculation function 354 are indicated by dotted lines, and the values ​​of the D / P ratio calculated using the values ​​of the feature amounts corrected by the feature amount correction function 356 are indicated by solid lines.

[0114] As described above, according to the first embodiment, the feature calculation function 354 calculates a first feature, which is a feature for each position in a first region of the pancreas included in the medical image, and a second feature, which is a feature for each position in a second region of the pancreas different from the first region. The correction target position identification function 355 identifies a correction target position for the calculated feature. Therefore, the medical image processing device 3 according to the first embodiment can identify a position where a feature may be defective (a position where a feature has not been calculated as expected), and by avoiding analysis using a feature that may be defective, it is possible to improve the accuracy of analysis using medical images.

[0115] Furthermore, according to the first embodiment, the feature amount correction function 356 corrects the feature amount at the correction target position. Therefore, the medical image processing device 3 according to the first embodiment can correctly detect abnormalities even at positions where feature amount calculation is poor, thereby enabling improvement in the accuracy of analysis using medical images.

[0116] Furthermore, according to the first embodiment, the abnormality detection function 357 detects an abnormality in the pancreas using at least the feature amount at the correction target position after correction, out of the first feature amount and the second feature amount. Therefore, the medical information processing device 3 according to the first embodiment can correctly detect an abnormality, and can prevent an abnormality from being overlooked or, conversely, detected as a false positive.

[0117] Furthermore, according to the first embodiment, the correction target position specifying function 355 specifies the correction target position based on the feature amount calculated by the feature amount calculating function 354 or information on the anatomical structure of the pancreas. Therefore, the medical information processing device 3 according to the first embodiment makes it possible to easily specify the correction target position.

[0118] Furthermore, according to the first embodiment, the feature calculation function 354 calculates a third feature based on the first feature and the second feature. The correction target position identification function 355 identifies the correction target position based on the third feature. Therefore, the medical image processing apparatus 3 according to the first embodiment can identify the correction target position under various conditions, and can perform identification processing on the correction target position according to the condition of the subject.

[0119] According to the first embodiment, the first feature amount and the second feature amount are the cross-sectional area of ​​the pancreas and the cross-sectional area of ​​the pancreatic duct, respectively. The feature amount calculation function 354 calculates the ratio between the first feature amount and the second feature amount to calculate the third feature amount. Therefore, the medical information processing device 3 according to the first embodiment can appropriately analyze the pancreas.

[0120] Furthermore, according to the first embodiment, the correction target position specifying function 355 specifies the correction target position based on a comparison between the feature amount calculated by the feature amount calculating function 354 and a reference value. Furthermore, the correction target position specifying function 355 specifies the correction target position based on a change in the feature amount for each position. Furthermore, the correction target position specifying function 355 specifies the correction target position based on the shape of the pancreas. Furthermore, the correction target position specifying function 355 specifies the correction target position based on the positional relationship between the pancreas and surrounding structures. Therefore, the medical image processing device 3 according to the first embodiment makes it possible to appropriately specify the correction target position in the pancreas.

[0121] According to the first embodiment, the feature amount correction function 356 corrects the feature amount of the correction target position by replacing the feature amount of the correction target position with a reference value. The feature amount correction function 356 corrects the feature amount of the correction target position by predicting the feature amount of the correction target position based on the feature amounts of the surrounding area of ​​the correction target position. The feature amount correction function 356 corrects the feature amount of the correction target position by acquiring heterogeneous feature amounts different from the feature amount for the correction target region and predicting the feature amount of the correction target position based on the acquired heterogeneous feature amounts. The feature amount correction function 356 corrects the feature amount of the correction target position by replacing the feature amount of the correction target position with a set numerical range. The feature amount correction function 356 corrects the feature amount of the correction target position by replacing the feature amount of the correction target position with a frequency distribution corresponding to the feature amount. Therefore, the medical image processing device 3 according to the first embodiment enables appropriate correction of the feature amount of the correction target position in the pancreas.

[0122] According to the first embodiment, the anomaly detection function 357 detects a pancreatic abnormality based on a comparison between the feature amount at the correction target position after the feature amount has been corrected to a single value and a threshold value. The anomaly detection function 357 detects a pancreatic abnormality based on a combination of the first feature amount and the second feature amount after the feature amount at the correction target position has been corrected to a single value for at least one of the first feature amount and the second feature amount. The anomaly detection function 357 detects a pancreatic abnormality based on a combination of the feature amount at the correction target position after the feature amount at the correction target position has been corrected to a single value for at least one of the first feature amount and the second feature amount, and a heterogeneous feature amount different from the first feature amount and the second feature amount. When the feature amount at the correction target position is corrected to a set numerical range, the anomaly detection function 357 detects a pancreatic abnormality based on an upper limit or lower limit of the set numerical range. When the feature amount at the correction target position is corrected to a frequency distribution corresponding to the feature amount, the anomaly detection function 357 detects a pancreatic abnormality based on the frequency of abnormalities in the frequency distribution. Therefore, the medical information processing apparatus 3 according to the first embodiment makes it possible to appropriately detect abnormalities in the pancreas.

[0123] Furthermore, according to the first embodiment, the abnormality detection function 357 detects at least one of the presence of a lesion, the presence of a tumor, the presence of a finding, and the presence of a pathological condition as an abnormality. Furthermore, the abnormality detection function 357 detects at least one of the presence or absence of an abnormality, the degree of the abnormality, the classification of the abnormality, and the probability of the abnormality as an abnormality detection result. Therefore, the medical information processing device 3 according to the first embodiment is capable of detecting a wide variety of abnormalities.

[0124] Furthermore, according to the first embodiment, the feature calculation function 354 calculates the first feature and the second feature for each position based on the criteria set for the pancreas. Therefore, the medical information processing device 3 according to the first embodiment can calculate appropriate feature for the pancreas.

[0125] Furthermore, according to the first embodiment, the control function 351 displays display information indicating the first feature amount and the second feature amount of the pancreas. Therefore, the medical information processing device 3 according to the first embodiment can display feature amounts with high accuracy.

[0126] Furthermore, according to the first embodiment, the control function 351 can display various display information about an abnormal region where an abnormality has been detected. Therefore, the medical information processing device 3 according to the first embodiment can display only the parts of the region where the pancreatic duct has not been extracted that have a high probability of abnormality on a graph or image, thereby allowing the user to check only the parts that are likely to be abnormal, thereby reducing the amount of work required. Furthermore, by checking images of the parts that have not been extracted / areas with a high probability of abnormality, the user can gain a more accurate understanding of the D / P ratio, which may improve the quality of medical care.

[0127] (Other embodiments)

[0128] The processing circuit described in each of the above embodiments may be configured by combining multiple independent processors, and each processor may execute a program to realize each processing function. Furthermore, each processing function of a processing circuit may be appropriately distributed or integrated among a single or multiple processing circuits. Furthermore, each processing function of a processing circuit may be realized by a combination of hardware and software, such as a circuit. While the above description illustrates an example in which programs corresponding to each processing function are stored in a single storage circuit 34, the embodiments are not limited to this. For example, programs corresponding to each processing function may be distributed and stored among multiple storage circuits, and the processing circuit may read and execute each program from each storage circuit.

[0129] In the above-described embodiment, an example has been described in which each unit in this specification is realized by each function of a processing circuit, but the embodiment is not limited to this. For example, each unit in this specification may be realized by the function described in the embodiment, or may be realized by hardware only, software only, or a combination of hardware and software.

[0130] Furthermore, the term "processor" used in the description of the above-mentioned embodiments refers to a circuit such as a central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)). Here, instead of storing a program in a memory circuit, the program may be directly embedded in the processor circuit. In this case, the processor realizes its function by reading and executing the program embedded in the circuit. Furthermore, each processor in the present embodiment is not limited to being configured as a single circuit, but may be configured as a single processor by combining multiple independent circuits to realize its function.

[0131] Here, the medical image processing program executed by the processor is provided by being pre-installed in a read-only memory (ROM), a storage circuit, or the like. The medical image processing program may be provided by being recorded on a computer-readable, non-transitory storage medium such as a compact disk (CD)-ROM, a flexible disk (FD), a recordable CD-R (CD-R), or a digital versatile disk (DVD) in a format that can be installed or executed by these devices. The medical image processing program may also be provided or distributed by being stored on a computer connected to a network such as the Internet and downloaded via the network. For example, the medical image processing program may be composed of modules including each of the processing functions described above. In terms of actual hardware, a CPU reads and executes the medical image processing program from a storage medium such as a ROM, whereby each module is loaded into a main memory device and generated on the main memory device.

[0132] In addition, in the above-described embodiments and modifications, the components of each device shown in the drawings are functional concepts and do not necessarily have to be physically configured as shown. In other words, the specific form of distribution or integration of each device is not limited to that shown in the drawings, and all or part of the devices can be functionally or physically distributed or integrated in any unit depending on various loads, usage conditions, etc. Furthermore, all or any part of the processing functions performed by each device can be realized by a CPU and a program analyzed and executed by the CPU, or can be realized as hardware using wired logic.

[0133] Furthermore, among the processes described in the above-mentioned embodiments and modifications, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using a known method.In addition, the information including the processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be changed as desired unless otherwise specified.

[0134] According to at least one of the embodiments described above, it is possible to improve the accuracy of analysis using medical images.

[0135] Although several embodiments have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, as well as within the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0136] 3 Medical information processing equipment 351 Control Functions 352 Image Acquisition Function 353 Estimation Function 354 Feature Calculation Function 355 Correction target position identification function 356 Feature Correction Function 357 Anomaly Detection Function

Claims

1. a calculation unit that calculates a first feature amount, which is a feature amount for each position in a first region of a pancreas included in a medical image, and a second feature amount, which is a feature amount for each position in a second region of the pancreas that is different from the first region; an identification unit that identifies a correction target position of the calculated feature amount; A medical information processing device comprising:

2. The medical image processing apparatus according to claim 1 , further comprising a correction unit that corrects the feature amount at the correction target position.

3. The medical image processing apparatus according to claim 2 , further comprising a detection unit that detects an abnormality in the pancreas using at least the feature amount at the correction target position after correction, out of the first feature amount and the second feature amount.

4. The medical image processing apparatus according to claim 1 , wherein the specifying unit specifies the correction target position based on the feature amount calculated by the calculating unit or information on the anatomical structure of the pancreas.

5. the calculation unit calculates a third feature amount based on the first feature amount and the second feature amount; The medical image processing apparatus according to claim 4 , wherein the specifying unit specifies the correction target position based on the third feature amount.

6. The medical image processing apparatus according to claim 5 , wherein the first feature amount and the second feature amount are a cross-sectional area of ​​the pancreas and a cross-sectional area of ​​the pancreatic duct, respectively.

7. The medical image processing apparatus according to claim 5 , wherein the calculation unit calculates the third feature amount by calculating a ratio between the first feature amount and the second feature amount.

8. The medical image processing apparatus according to claim 4 , wherein the specifying unit specifies the correction target position based on a comparison between the feature amount calculated by the calculating unit and a reference value.

9. The medical image processing apparatus according to claim 4 , wherein the specifying unit specifies the correction target position based on a change in the feature amount for each position.

10. The medical image processing apparatus according to claim 4 , wherein the specifying unit specifies the correction target position based on a shape of the pancreas.

11. The medical image processing apparatus according to claim 4 , wherein the specifying unit specifies the correction target position based on a positional relationship between the pancreas and surrounding structures.

12. The medical image processing apparatus according to claim 2 , wherein the correction unit corrects the feature amount at the correction target position by replacing the feature amount at the correction target position with a reference value.

13. The medical image processing apparatus according to claim 2 , wherein the correction unit corrects the feature amount of the correction target position by predicting the feature amount of the correction target position based on feature amounts around the correction target position.

14. 3. The medical image processing device according to claim 2, wherein the correction unit corrects the feature amount of the correction target position by acquiring a heterogeneous feature amount different from the feature amount for the correction target position and predicting the feature amount of the correction target position based on the acquired heterogeneous feature amount.

15. The medical image processing apparatus according to claim 2 , wherein the correction unit corrects the feature amount at the correction target position by replacing the feature amount at the correction target position with a value within a set range.

16. The medical image processing apparatus according to claim 2 , wherein the correction unit corrects the feature amount of the correction target position by replacing the feature amount of the correction target position with a frequency distribution corresponding to the feature amount.

17. The medical image processing apparatus according to claim 3 , wherein the detection unit detects the abnormality of the pancreas based on a comparison between the feature amount after the feature amount at the correction target position has been corrected to a single value and a threshold value.

18. 4. The medical information processing device according to claim 3, wherein the detection unit detects an abnormality in the pancreas based on a combination of the first feature amount and the second feature amount after the feature amount at the correction target position is corrected to a single value for at least one of the first feature amount and the second feature amount.

19. 4. The medical information processing device according to claim 3, wherein the detection unit detects the abnormality of the pancreas based on a combination of a feature amount obtained after the feature amount at the correction target position is corrected to a single value for at least one of the first feature amount and the second feature amount, and a heterogeneous feature amount different from the first feature amount and the second feature amount.

20. The medical information processing device according to claim 3 , wherein when the feature amount at the correction target position is corrected to fall within a set numerical range, the detection unit detects an abnormality in the pancreas based on an upper limit value or a lower limit value in the set numerical range.

21. The medical image processing device according to claim 3 , wherein when the feature amount at the correction target position is corrected to a frequency distribution corresponding to the feature amount, the detection unit detects the abnormality of the pancreas based on the frequency of the abnormality in the frequency distribution.

22. The medical information processing apparatus according to claim 3 , wherein the detection unit detects, as the abnormality, at least one of the presence of a lesion, the presence of a tumor, the presence of a finding, and the presence of a pathological condition.

23. The medical image processing apparatus according to claim 3 , wherein the detection unit detects at least one of the presence or absence of the abnormality, the degree of the abnormality, the classification of the abnormality, and the probability of the abnormality as the detection result of the abnormality.

24. The medical image processing apparatus according to claim 1 , wherein the calculation unit calculates the first feature amount and the second feature amount for each position based on a criterion set for the pancreas.

25. The medical image processing apparatus according to claim 1 , further comprising a display control unit that displays display information indicating the first feature amount and the second feature amount of the pancreas.

26. 7. The medical information processing device according to claim 6, wherein the calculation unit calculates a cross-sectional area of ​​the pancreas and a cross-sectional area of ​​the pancreatic duct in a cross section perpendicular to the centerline for each position along the centerline from the head to the tail of the pancreas.

27. calculating a first feature amount, which is a feature amount for each position in a first region of a pancreas included in a medical image, and a second feature amount, which is a feature amount for each position in a second region of the pancreas that is different from the first region; Identifying a position to be corrected for the calculated feature amount; A medical information processing method, comprising:

Images