JP2020031810A - Medical image processing apparatus, medical image processing system, and medical image processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the efficiency of a workflow using an image processing application. A medical image processing apparatus according to an embodiment includes an acquisition unit, a storage unit, and an estimation unit. The acquisition unit acquires the extraction area in the medical image extracted by the image processing and the correction area in which the correction person has corrected the extraction area. The storage unit is a trained model obtained by learning the difference between the extraction area and the correction area and the relationship with the extraction area, and is correction for the extraction area extracted from the medical image by image processing. Stores a trained model that is functionalized to estimate the modified region after it has been subjected to. The estimation unit controls to estimate the modified region by inputting the extracted region extracted from the medical image by the image processing into the trained model. [Selection diagram] Fig. 2

Description

  An embodiment of the present invention relates to a medical image processing device, a medical image processing system, and a medical image processing program.

  Conventionally, an image processing algorithm installed in an image processing application performed on a medical image targets subjective images of a typical image of the medical image, a subject such as a developer or an evaluator of the image processing application, and statistics. It may be optimized by the idea. In such an image processing application, what the user (operator) expects for the output of the image processing application does not always match the policy of optimizing the image processing application. In addition, even when a typical image referred to at the time of optimization is significantly different from an image used by the user, the user's expectation may differ from the content of the optimization. Therefore, when such an image processing application is used in a clinical setting, the result output by the image processing algorithm may be different from the output expected by the user. The user may modify the output result so that the result output from the image processing algorithm is in accordance with his / her expectations.

JP 2014-30556 A JP 2014-212820 A

  An object of the present invention is to improve the efficiency of a workflow using an image processing application.

  A medical image processing device according to an embodiment includes an acquisition unit, a storage unit, and an estimation unit. The obtaining unit obtains an extraction region in the medical image extracted by the image processing and a correction region in which the corrector has corrected the extraction region. The storage unit is a trained model obtained by learning a relationship between the difference between the extraction region and the correction region and the extraction region, and the extraction region extracted from a medical image by the image processing. And stores a learned model that is provided with a function of estimating a corrected area after correction has been performed on. The estimating unit performs control so as to estimate a corrected area by inputting an extracted area extracted from a medical image by the image processing into the learned model.

FIG. 1 is a diagram illustrating an example of a configuration of a medical image report processing system according to the present embodiment. FIG. 2 is a diagram illustrating processing during learning and during operation performed by the medical image processing apparatus according to the present embodiment. FIG. 3 is a diagram for explaining an example of processing at the time of learning of the medical image processing apparatus according to the present embodiment. FIG. 4A is a diagram for describing an example of the analysis processing by the analysis function according to the present embodiment. FIG. 4B is a diagram schematically illustrating texture parameters used for generating a learned model according to the present embodiment. FIG. 5 is a diagram for explaining an example of a process at the time of operation of the medical image processing apparatus according to the present embodiment. FIG. 6 is a flowchart illustrating a procedure of a learning process performed by the medical image processing apparatus according to the present embodiment. FIG. 7 is a flowchart illustrating a procedure of processing during operation by the medical image processing apparatus according to the present embodiment. FIG. 8 is a diagram schematically illustrating texture parameters used for generating a learned model according to the present embodiment.

  Hereinafter, embodiments of a medical image processing apparatus, a medical image processing system, and a medical image processing program will be described in detail with reference to the drawings. Note that the medical image processing apparatus, the medical image processing system, and the medical image processing program according to the present application are not limited to the embodiments described below.

(Embodiment)
FIG. 1 is a diagram illustrating a configuration example of a medical image processing system 100 according to the present embodiment. As shown in FIG. 1, the medical image processing system 100 according to the present embodiment includes a medical image diagnostic device 110, a terminal device 120, and a medical image processing device 130. Here, each device is communicably connected via the network 200.

  The medical image diagnostic apparatus 110 images a subject and generates a medical image. Then, the medical image diagnostic apparatus 110 transmits the generated medical image to the terminal apparatus 120 and the medical image processing apparatus 130. For example, the medical image diagnostic apparatus 110 includes an X-ray diagnostic apparatus, an X-ray CT (Computed Tomography) apparatus, an MRI (Magnetic Resonance Imaging) apparatus, an ultrasonic diagnostic apparatus, a SPECT (Single Photon Emission Computed Tomography) apparatus, and a PET (Positron Emission). computed tomography) device.

  The terminal device 120 is a device for allowing a doctor or a laboratory technician who works in a hospital to browse medical images. For example, the terminal device 140 is realized by a personal computer (PC), a tablet PC, a PDA (Personal Digital Assistant), a mobile phone, or the like operated by a doctor or a laboratory technician who works in a hospital. For example, the terminal device 120 displays a medical image received from the medical image diagnostic device 110 or the medical image processing device 130 on its own display, and receives various operations on the medical image via its own input interface.

  The medical image processing device 130 acquires various information from the medical image diagnostic device 110 and the terminal device 120, and performs various information processing using the acquired information. For example, the medical image processing apparatus 130 is realized by a computer device such as a server, a workstation, a personal computer, and a tablet terminal.

  As shown in FIG. 1, the medical image processing apparatus 130 includes a communication interface 131, a storage circuit 132, an input interface 133, a display 134, and a processing circuit 135.

  The communication interface 131 is connected to the processing circuit 135 and controls communication performed between the medical image processing apparatus 130 and each apparatus. Specifically, the communication interface 131 receives various information from each device, and outputs the received information to the processing circuit 135. For example, the communication interface 131 is realized by a network card, a network adapter, a NIC (Network Interface Controller), or the like.

  The storage circuit 132 is connected to the processing circuit 135 and stores various data. For example, the storage circuit 132 stores a medical image received from the medical image diagnostic apparatus 110, a medical image acquired from the terminal device 120, and the like. The storage circuit 132 stores various programs for realizing various functions by being read and executed by the processing circuit 135. For example, the storage circuit 132 is realized by a semiconductor memory element such as a random access memory (RAM) or a flash memory, a hard disk, an optical disk, or the like. Note that the storage circuit 132 is an example of a storage unit.

  The input interface 133 is connected to the processing circuit 135, and receives various instructions and information input operations from the operator. Specifically, the input interface 133 converts the input operation received from the operator into an electric signal and outputs the electric signal to the processing circuit 135. For example, the input interface 133 includes a track ball, a switch button, a mouse, a keyboard, a touch pad for performing an input operation by touching an operation surface, a touch screen in which a display screen and a touch pad are integrated, and a non-display using an optical sensor. This is realized by a contact input circuit, a voice input circuit, and the like. In the present specification, the input interface 133 is not limited to one having physical operation components such as a mouse and a keyboard. For example, an example of the input interface 133 includes an electric signal processing circuit that receives an electric signal corresponding to an input operation from an external input device provided separately from the device and outputs the electric signal to a control circuit.

  The display 134 is connected to the processing circuit 135 and displays various information and images. Specifically, the display 134 converts the information and the image data sent from the processing circuit 135 into electric signals for display and outputs them. For example, the display 134 is realized by a liquid crystal monitor, a CRT (Cathode Ray Tube) monitor, a touch panel, or the like.

  The processing circuit 135 controls the operation of the medical image processing apparatus 130 according to an input operation received from the operator via the input interface 133. For example, the processing circuit 135 is realized by a processor.

  The configuration of the medical image processing system 100 according to the present embodiment has been described above. For example, the medical image processing system 100 according to the present embodiment is installed in a medical institution such as a hospital or a clinic, and a medical image generated by the medical image diagnostic apparatus 110 with a patient or the like who is admitted or visited a medical institution as a subject. It is used for various types of image diagnosis using.

  Here, in the image diagnosis using the medical image, an image processing application may be applied to the medical image collected by the medical image diagnostic apparatus 110, and the processed medical image may be used for the diagnosis. For example, an image processing application for extracting a high-brightness region of brain white matter from a FLAIR (Fluid-Attenuated Inversion Recovery) image collected by an MRI apparatus is performed, and the processed FLAIR image is used for diagnosis of multiple sclerosis. Can be

  However, as described above, the image processing algorithm installed in such an image processing application targets a typical image of a medical image, and subjects the image such as a subject of an image processing application developer or an evaluator, or a statistical image. Optimized by the idea. That is, in such an image processing algorithm, the output result is uniquely determined according to the input medical image. Therefore, when the user (operator) is not satisfied with the output result, the user may correct the output result by manual operation. In addition, such a user is likely to make corrections to the output result with high frequency, and such corrections hinder the diagnostic workflow. Furthermore, the tendency of correction may differ for each user, and it is difficult to output an output result appropriate for each user from the image application.

  Therefore, in the present embodiment, the medical image processing apparatus 130 in the medical image processing system 100 learns the relationship between the output result of the image processing application and the correction result of the user (corrector) for the output result, thereby performing image processing. Build a trained model that outputs an output result that reflects the area where the input of the output result by the application will be modified. Then, the medical image processing apparatus 130 estimates the corrected output result by inputting the output result of the newly acquired image processing application to the learned model. Thereby, the medical image processing apparatus 130 can automatically obtain an output result according to the correction tendency of the user, and can provide an appropriate output result without manual correction by the user himself. As a result, the medical image processing apparatus 130 can improve the efficiency of the workflow using the image processing application.

  Hereinafter, details of the medical image processing apparatus 130 according to the present embodiment will be described. As shown in FIG. 1, the processing circuit 135 of the medical image processing apparatus 130 executes a control function 135a, an image processing function 135b, an analysis function 135c, a learning function 135d, and an estimation function 135e. Here, the control function 135a is an example of an acquisition unit and a display control unit. The image processing function 135b is an example of an obtaining unit. The learning function 135d is an example of a learning unit. The estimating function 135e is an example of an estimating unit.

  The control function 135a controls so as to execute a process according to various requests input via the input interface 133. For example, the control function 135a includes transmission and reception of medical images and the like via the communication interface 131, storage of information in the storage circuit 132, information on the display 134 (for example, display images, processing results of the image processing function 135b, analysis functions The display of the analysis result of 135c, the estimation result of the estimation function 135e, and the like are controlled.

  For example, the control function 135a acquires a medical image from the medical image diagnostic apparatus 110 and stores the medical image in the storage circuit 132. Further, for example, the control function 135a controls the display 134 to display a GUI for executing a process on a medical image acquired from the medical image diagnostic apparatus 110 and a processing result by each function. The details of the display content displayed by the control function 135a will be described later.

  The image processing function 135b executes predetermined image processing on a medical image acquired from the medical image diagnostic apparatus 110. Specifically, the image processing function 135b executes image processing for extracting a predetermined area included in the medical image. For example, the image processing function 135b extracts a predetermined area from a medical image using a normal (not a learning algorithm) image processing algorithm installed in the image processing application. As an example, the image processing function 135b is an image processing application that extracts a high-brightness region of brain white matter included in an MRI image collected for the head.

  Note that the image processing function 135b can execute various image processing algorithms according to the content of the image processing, and extract an area from the medical image. For example, the image processing function 135b may execute an image processing algorithm for extracting a tumor region included in a medical image.

  The analysis function 135c performs image analysis on a difference area between the area extracted by the image processing function 135b and the area where the operator has modified the area. The analysis function 135c performs image analysis on the region extracted by the image processing function 135b. The details of the processing by the analysis function 135c will be described later.

  The learning function 135d learns the relationship between the difference between the area extracted by the image processing function 135b and the area where the operator has modified the area and the area extracted by the image processing function 135b. Then, a learned model which is provided with a function of estimating a corrected area after correcting the area extracted from the medical image by the image processing function 135b is generated. The details of the process by the learning function 135d will be described later.

  The estimating function 135e controls to estimate the corrected area by inputting the area extracted from the medical image by the image processing function 135b into the learned model. The details of the processing by the estimation function 135e will be described later.

  In the present embodiment, each function of the above-described processing circuit 135 performs a process at the time of learning for generating a learned model and a process at the time of operation using the generated learned model. More specifically, the control function 135a, the image processing function 135b, the analysis function 135c, and the learning function 135d perform the image processing application on the medical image collected by the medical image diagnostic apparatus 110 and output the result. On the other hand, when the user makes a correction, or when an instruction to start learning is received from the operator, processing at the time of learning is performed. When the control function 135a, the image processing function 135b, the analysis function 135c, and the estimation function 135e execute an image processing application on a medical image collected by the medical image diagnostic apparatus 110 and output an output result, Perform the operation process.

  FIG. 2 is a diagram illustrating processing during learning and during operation performed by the medical image processing apparatus 130 according to the present embodiment.

  For example, as shown in the upper part of FIG. 2, at the time of learning, the learning function 135d learns the relationship between the region extraction result obtained by the image processing function 135b and the analysis result obtained by correcting the extraction result by the analysis function 135c. Based on the region extracted from the medical image, a learned model that outputs the corrected region is generated.

  In such a case, first, the control function 135a acquires a medical image from the medical image diagnostic apparatus 110 and stores it in the storage circuit 132. The image processing function 135b extracts an area by executing image processing by an image processing algorithm on the medical image acquired by the control function 135a, and stores the extracted result in the storage circuit 132 in association with the medical image. I do. Here, the image processing function 135b extracts an area by an image processing algorithm specified by an operator, for example. That is, the image processing function 135b performs the area extraction by the image processing algorithm selected when the operator performs the image diagnosis.

  Then, the control function 135a causes the display 134 to display the extraction result obtained by the image processing function 135b. The operator executes a correction operation for correcting the extraction result of the area displayed on the display 134 via the input interface. Specifically, the operator performs a correction operation for correcting the region extracted by the image processing function 135b. Here, the information on the area corrected by the operator is stored in the storage circuit 132 in association with the medical image and the extraction result by the image processing function 135b.

  The analysis function 135c extracts a difference area between the extraction area extracted by the image processing function 135b and the correction area corrected by the operator, and performs image analysis on the extracted difference area. Specifically, the analysis function 135c performs an image analysis for acquiring an image feature amount in the difference area. For example, the analysis function 135c acquires the image feature amount in the difference area by performing texture analysis on the difference area. The analysis function 135c stores the acquired image feature amount in the storage circuit 132 in association with the medical image, the extraction result by the image processing function 135b, and the correction area.

  The learning function 135d generates a learned model by learning the relationship between the extraction region extracted by the image processing function 135b and the result of the image analysis in the difference region. Specifically, the learning function 135d refers to the storage circuit 132, and acquires the result of the image analysis in the extraction area and the difference area extracted by the image processing function 135b. Then, the learning function 135d performs machine learning by inputting the obtained extraction region and the result of the image analysis to the machine learning engine as learning data. Here, as information on the extraction region, for example, luminance information of the extraction region is used.

  The machine learning engine refers to the DB of the storage circuit 142 based on, for example, the input standard image and the image feature amount, and compares the image correlation and the image feature amount, thereby obtaining a clinically optimal imaging parameter. To determine. For example, machine learning engines include Deep Learning, Neural Network, Logistic regression analysis, nonlinear discriminant analysis, Support Vector Machine (SVM), and Random Forest. The optimal parameters are determined by using various algorithms such as Naive Bayes.

  As a result of such machine learning, the learning function 135d generates a learned model that outputs an output result reflecting a region where the input of the output result by the image processing function 135b will be corrected. Then, the learning function 135d stores the generated learned model in the storage circuit 142. At this time, if the previously created learned model has already been stored in the storage circuit 132, the learning function 135d replaces the stored learned model with the newly created learned model.

  Further, the learning function 135d can also generate a learned model for each operator (corrector) who makes a correction and store the generated model in the storage circuit 132. For example, the correction to the region extracted by the image processing function 135b may have a different tendency for each operator. Therefore, the learning function 135d causes the storage circuit 132 to store an identifier (for example, a user ID) for uniquely specifying the operator in association with the learned algorithm. When replacing the already-stored learned model with a newly generated learned model, the learning function 135d identifies the operator based on a user ID or the like input when performing image diagnosis, and responds. Identify the trained model to be trained.

  Hereinafter, an example of a process at the time of learning will be described with reference to FIGS. FIG. 3 is a diagram for explaining an example of processing at the time of learning of the medical image processing apparatus 130 according to the present embodiment. Here, FIG. 3 shows a case where correction is applied to a high-luminance area extracted from brain white matter included in an MRI image collected for the head.

  For example, the image processing function 135b applies the image processing algorithm to the original image (MRI image collected for the head) shown on the left end of FIG. 3 to obtain the second image from the left of FIG. As shown in (1), a high brightness region of the brain white matter is extracted. When the high-brightness area is extracted by the image processing function 135b, the control function 135a displays the extraction result (the second image from the left in FIG. 3) on the display 134.

  The operator corrects the extraction result of the high-brightness region of the brain white matter with reference to the extraction result displayed on the display 134. That is, as shown in the third image from the left in FIG. 3, the operator corrects the high luminance area extracted by the image processing function 135b.

  The analysis function 135c performs a texture analysis on a difference area between the correction area corrected by the operator and the extraction area extracted by the image processing function 135b. For example, the analysis function 135c performs a texture analysis on a difference area between the extraction area shown in the second image from the left in FIG. 3 and the correction area shown in the third image from the left in FIG. I do.

  FIG. 4A is a diagram for describing an example of an analysis process by the analysis function 135c according to the present embodiment. Here, in FIG. 4A, a correction region in which an unnecessary region has been deleted by the operator correcting the extraction region extracted by the image processing function 135b is shown in the region R1. Further, the two images shown in the lower part of FIG. 4A are the same images in which the vicinity of the region R1 is enlarged, respectively, and show the 3 × 3 kernel position shifted.

  For example, the analysis function 135c extracts a difference area between the extraction area extracted by the image processing function 135b and a correction area after the extraction area has been corrected by the operator. That is, the analysis function 135c extracts an area other than the correction area shown in the area R1 in FIG. 4A from the extraction area. Then, the analysis function 135c calculates a pixel value at each position in the difference area by applying, for example, a 3 × 3 9-pixel kernel to the extracted difference area. As an example, as shown in FIG. 4A, the analysis function 135c calculates the pixel value at the central pixel position of the kernel from the peripheral eight pixel values while changing the kernel position with respect to the difference region. . Then, the analysis function 135c calculates various texture parameters based on the pixel values at each position in the difference area.

  To give an example, the analysis function 135c calculates the pixel values “Mean”, “SD”, “Skewness”, “Kurtosis”, “Energy”, “Entropy”, “dissimilarity”, “Homogeneity” at each position in the difference area. , “Size” and the like are calculated. That is, the analysis function 135c calculates a plurality of texture parameters for one difference area. The analysis function 135c extracts a difference region from each of the correction regions corrected by the operator, and calculates a plurality of texture parameters for each difference region. Then, the analysis function 135c causes the storage circuit 132 to store the calculated plurality of texture parameters in association with the difference area. The above-described texture parameters are merely examples, and the analysis function 135c can calculate other various texture parameters.

  Returning to FIG. 3, the learning function 135 d generates a learned model using the result of the texture analysis of the extracted area and the difference area, and stores the generated model in the database of the learned model in the storage circuit 132. For example, the learning function 135d learns the relationship between the luminance value of the extraction area and a plurality of texture parameters in the corresponding difference area, and provides the operator with a texture parameter corresponding to the correction tendency based on the luminance value of the extraction area. Generate a trained model that outputs.

  FIG. 4B is a diagram schematically illustrating texture parameters used for generating a learned model according to the present embodiment. FIG. 4B shows an example in which three texture parameters are calculated for each difference area. The texture parameters used for generating the learned model according to the present embodiment include, for example, as shown in FIG. 4B, a first parameter (a value of a horizontal axis) for each difference area (for each point), It has three-dimensional information of two parameters (vertical axis values) and a third parameter (depth axis values).

  The learning function 135d generates a learned model from information on the luminance value of the extraction region acquired from many medical images (MRI images of the brain) and information on the texture parameters as shown in FIG. 4B. In FIG. 4B, since three texture parameters are calculated, three-dimensional information is provided for each difference region. For example, when the above-described nine texture parameters are calculated, nine-dimensional information is provided for each difference region. Will be provided. For example, the learning function 135d generates a learned model in this way, and stores the generated learned model in the storage circuit 132.

  On the other hand, for example, as shown in the lower part of FIG. 2, during operation, the control function 135a acquires a medical image from the medical image diagnostic apparatus 110 and stores it in the storage circuit 132. The image processing function 135b extracts an extraction region by executing image processing by an image processing algorithm on the medical image acquired by the control function 135a, and stores the extracted region in the storage circuit 132 in association with the medical image.

  After that, the estimation function 135e estimates the corrected area by causing the extracted area to be a learned model. Specifically, the estimation function 135e outputs the texture parameter by inputting the information on the luminance value of the extraction region extracted by the image processing function 135b to the learned model. Then, the estimation function 135e estimates a region corresponding to the output texture parameter as a corrected region.

  For example, the estimation function 135e refers to the result of the texture analysis performed on the extraction region by the analysis function 135c, and specifies the region corresponding to the texture parameter output from the learned model in the extraction region. The estimation function 135e estimates the specified area as a corrected area.

  FIG. 5 is a diagram for explaining an example of a process at the time of operation of the medical image processing apparatus according to the present embodiment. Here, FIG. 5 shows a case where a corrected region in a high-brightness region extracted from brain white matter included in an MRI image collected for the head is estimated.

  For example, the image processing function 135b applies the image processing algorithm to the original image (MRI image collected for the head) shown on the left end of FIG. 5 to obtain the second image from the left in FIG. As shown in (1), a high brightness region of the brain white matter is extracted. The analysis function 135c performs a texture analysis in the high luminance area.

  The estimation function 135e acquires the value of the texture parameter by inputting the luminance value of the high luminance area into the learned model. Further, the estimation function 135e refers to the result of the texture analysis performed on the high-luminance area by the analysis function 135c, and in the high-luminance area, the area corresponding to the texture parameter output from the learned model (the corrected area). ).

  The control function 135a causes the display 134 to display a display image (the image on the right end in FIG. 5) in which the corrected area estimated by the estimation function 135e is reflected on the medical image. Here, the control function 135a can display the extraction area extracted by the image processing function 135b and the corrected area estimated by the estimation function 135e in a distinguishable manner. For example, the control function 135a can display the extracted area and the corrected area in different colors so as to be identifiable.

  Here, in the medical image processing apparatus 130, the corrected area estimated by the estimation function 135e can be further modified, and the result can be reflected in the learned model. For example, the operator refers to the display image displayed on the display 134 and determines whether or not to re-correct the corrected area. Then, when performing the re-correction, the operator corrects the post-correction area via the input interface 133.

  The analysis function 135c performs a texture analysis on a difference area between the correction area corrected by the operator and the corrected area. Then, the learning function 135d newly generates a learned model using the result of the texture analysis, and replaces the generated model with the learned model stored in the storage circuit 132.

  The processing functions of the processing circuit 135 of the medical image processing apparatus 130 have been described above. Here, as described above, when the processing circuit 135 is realized by a processor, each processing function of the processing circuit 135 is stored in the storage circuit 132 in the form of a program executable by a computer. Then, the processing circuit 135 reads out each program from the storage circuit 132 and executes it to realize a function corresponding to each program. In other words, the processing circuit 135 in a state where each program is read has each function shown in the processing circuit 135 in FIG. Although FIG. 1 illustrates that each processing function is realized by a single processor, a plurality of independent processors are combined to form a processing circuit, and each processor executes a program to realize a function. It does not matter. Further, the processing functions of the processing circuit 135 may be implemented by being appropriately distributed or integrated in a single or a plurality of processing circuits. Further, in the example shown in FIG. 1, a single storage circuit 132 has been described as storing a program corresponding to each processing function. However, a plurality of storage circuits are arranged in a distributed manner, and the processing circuits are individually stored. A configuration in which a corresponding program is read from a circuit may be employed.

  Next, a procedure of processing by the medical image processing apparatus 130 will be described. FIG. 6 is a flowchart illustrating a procedure of processing at the time of learning by the medical image processing apparatus 130 according to the present embodiment. Here, steps S101 and S103 in FIG. 6 are realized by the processing circuit 135 calling and executing a program corresponding to the control function 135a from the storage circuit 132. Step S102 in FIG. 6 is a step realized by the processing circuit 135 calling and executing a program corresponding to the image processing function 135b from the storage circuit 132. Steps S104 and S105 in FIG. 6 are realized by the processing circuit 135 calling and executing a program corresponding to the analysis function 135c from the storage circuit 132. Steps S106 and S107 in FIG. 6 are realized by the processing circuit 135 calling and executing a program corresponding to the learning function 135d from the storage circuit 132.

  As shown in FIG. 6, in the medical image processing apparatus 130, the processing circuit 135 first determines whether image processing has been started (step S101). Here, when the image processing is started (Yes at Step S101), the processing circuit 135 executes the image processing on the designated original image (Step S102), and displays the processing result (Step S103). . Note that the processing circuit 135 is in a standby state until the image processing is started (No at Step S101).

  Then, the processing circuit 135 determines whether a correction has been received for the processing result (step S104). Here, when the correction has not been received (No at Step S104), the processing circuit 135 ends the processing. On the other hand, when the correction has been received (Yes at Step S104), the processing circuit 135 performs the texture analysis of the corrected area (Step S105).

  Thereafter, the processing circuit 135 creates a learned model for correction using the analysis result (step S106), and stores the created learned model in the storage circuit 132 (step S107).

  FIG. 7 is a flowchart illustrating a procedure of processing during operation by the medical image processing apparatus 130 according to the present embodiment. Here, steps S201 and S204 in FIG. 7 are realized by the processing circuit 135 calling and executing a program corresponding to the control function 135a from the storage circuit 132. Step S202 in FIG. 7 is realized by the processing circuit 135 calling and executing a program corresponding to the image processing function 135b from the storage circuit 132. Step S203 in FIG. 7 is a step realized by the processing circuit 135 calling and executing a program corresponding to the estimation function 135e and the analysis function 135c from the storage circuit 132. Steps S205 and S206 in FIG. 7 are realized by the processing circuit 135 calling and executing a program corresponding to the analysis function 135c from the storage circuit 132. Step S207 in FIG. 7 is a step realized by the processing circuit 135 calling and executing a program corresponding to the learning function 135d from the storage circuit 132.

  As shown in FIG. 7, in the medical image processing apparatus 130, the processing circuit 135 first determines whether image processing has been started (step S201). Here, when the image processing is started (Yes at Step S201), the processing circuit 135 executes the image processing on the designated original image (Step S202). Note that the processing circuit 135 is in a standby state until the image processing is started (No at Step S201).

  Then, the processing circuit 135 inputs the result of the image processing to the learned model (step S203), and displays a display image based on the output result of the learned model (step S204). Thereafter, the processing circuit 135 determines whether a correction has been received for the display image (step S205).

  Here, when the correction has not been received (No at Step S205), the processing circuit 135 ends the processing. On the other hand, when the correction has been received (Yes at Step S205), the processing circuit 135 executes the texture analysis of the corrected area (Step S206). Thereafter, the processing circuit 135 updates the learned model (step S107).

  As described above, according to the present embodiment, the control function 135a and the image processing function 135b allow the operator (corrector) to correct the extraction region in the medical image extracted by the image processing and the extraction region. Acquire the corrected area that has been applied. The storage circuit 132 is a learned model obtained by learning the relationship between the difference between the extraction region and the correction region and the extraction region, and stores a learned model extracted from a medical image by image processing. A learned model that is provided with a function of estimating the corrected area after the correction is stored. The estimating function 135e controls to estimate the corrected area by inputting the extracted area extracted from the medical image by the image processing into the learned model. Therefore, the medical image processing apparatus 130 according to the present embodiment can automatically correct the processing result of the image processing application according to the correction tendency of the corrector, and improve the efficiency of the workflow using the image processing application. Enable. For example, the medical image processing apparatus 130 may automatically correct the output of the image processing application to the content expected by the user even if the optimization of the image processing application is different from the optimization expected by the user. And improve the efficiency of the workflow using the image processing application. Further, for example, the medical image processing apparatus 130 may perform image processing even if the medical image input by the user to the image processing application is an image that is significantly different from a typical image used when developing the image processing application. Can be automatically corrected to the content expected by the user, and the efficiency of the workflow using the image processing application can be improved. Further, the medical image processing apparatus 130 according to the present embodiment makes it possible to suppress excessive detection by an image processing algorithm.

  Further, according to the present embodiment, the storage circuit 132 stores a learned model obtained by learning the relationship between the result of image analysis on the difference region between the extraction region and the correction region and the extraction region. The estimation function 135e controls the estimation of the area to be corrected in the extraction area by inputting the result of the image analysis of the extraction area extracted from the medical image by the image processing into the learned model. Therefore, the medical image processing apparatus 130 according to the present embodiment can easily perform appropriate correction.

  Further, according to the present embodiment, the image analysis is a texture analysis. Therefore, the medical image processing apparatus 130 according to the present embodiment can generate a trained model using various image feature amounts, and automatically performs a correction that more accurately reflects the correction tendency of the corrector. Enable.

  Further, according to the present embodiment, the storage circuit 132 stores a learned model for each corrector who has made a correction to an extraction region extracted from a medical image by image processing. The estimating function 135e performs control so as to estimate an after-correction area for each corrector by inputting an extraction area extracted from the medical image to a learned model for each corrector. Therefore, the medical image processing apparatus 130 according to the present embodiment can automatically perform the correction according to the correction tendency for each corrector, and improve the efficiency of the workflow even when the correction is performed by a plurality of correctors. Enable.

  Further, according to the present embodiment, the learning function 135d generates a learned model. Therefore, the medical image processing apparatus 130 according to the present embodiment can perform stand-alone processing.

  Further, according to the present embodiment, the control function 135a controls the medical image so that the extracted area and the corrected area in the medical image extracted by the image processing are displayed in a distinguishable manner. Therefore, the medical image processing apparatus 130 according to the present embodiment can present the result before the correction together with the result after the correction performed automatically to the operator.

(Other embodiments)
The above-described embodiment can be appropriately modified and implemented by partially changing the configuration or function of the medical image processing apparatus 130. Therefore, hereinafter, some modified examples according to the above-described embodiment will be described as other embodiments. In the following, points different from the above-described embodiment will be mainly described, and detailed description of points common to the already described contents will be omitted. Further, each embodiment described below may be implemented individually or may be implemented in combination as appropriate.

(Other Embodiment-1)
For example, in the above-described embodiment, the case where the post-correction area is displayed for each corrector has been described. However, the embodiment is not limited to this. For example, a case may be used in which a plurality of corrected regions are displayed in one medical image using a learned model corresponding to a plurality of correctors.

  In such a case, for example, the estimation function 135e causes the information of the extraction region extracted by the image processing function 135b to be input to each of the plurality of learned models, and to generate the plurality of corrected regions based on the output results. Estimate each. The control function 135a displays the estimated plurality of post-correction areas on one medical image. Here, the control function 135a controls so that the post-correction area for each corrector is identifiably displayed. Accordingly, the medical image processing apparatus 130 allows the operator to select an optimum correction result from the plurality of post-correction areas.

(Other Embodiment-2)
Further, for example, in the above-described embodiment, a case has been described in which a learned model for each corrector is generated and stored. However, the embodiment is not limited to this, and, for example, a case where a learned model in which correction results are integrated by a plurality of correctors may be generated.

  In such a case, for example, the storage circuit 132 integrates a plurality of image analysis results corresponding to correction results performed by a plurality of correctors on an extraction region extracted from a medical image by image processing. And a learned model obtained by learning the relationship between the model and the extraction region. The estimation function 135e controls the estimation of the area to be corrected in the extraction area by inputting the result of the image analysis of the extraction area extracted from the medical image by the image processing into the learned model.

  FIG. 8 is a diagram schematically illustrating texture parameters used for generating a learned model according to the present embodiment. FIG. 8 shows an example in which three texture parameters are calculated for each difference area. In FIG. 8, texture parameters for each difference area obtained from the correction results by the two correctors are indicated by solid circles and solid triangles, respectively. The texture parameters used for generating the learned model according to the present embodiment are, for example, as shown in FIG. 8, a first parameter (a value of a horizontal axis) and a second parameter (a vertical axis) for each difference area. Direction value), and three-dimensional information of a third parameter (depth direction axis value).

  The learning function 135d generates a trained model using only the texture parameter information that is similar among the results of two correctors acquired from a large number of medical images (MRI images of the brain). For example, the learning function 135d generates a learned model using the texture parameters included in the region R2 in FIG. 8 and the extraction region corresponding to each texture parameter, and stores the generated model in the storage circuit 132. Thus, the medical image processing apparatus 130 according to the present embodiment can automatically perform a more accurate correction.

(Other embodiment-3)
In the above-described embodiment, an example has been described in which a learned model is generated and used for each medical institution such as a hospital or clinic where the medical image processing system 100 is installed. However, the embodiment is not limited to this. For example, a case where the generated learned model is used in a medical institution different from the generated medical institution may be used.

  In such a case, the medical image processing apparatus 130 is connected to another medical institution via the network 200, and the learned model (trained model for each corrector and a plurality of correctors) generated at the other medical institution. A corrected model is acquired, and a corrected area is estimated using the acquired learned model.

  In the above-described embodiment, an example has been described in which a learned model is generated for each medical institution such as a hospital or a clinic in which the medical image processing system 100 is installed. However, the embodiment is not limited to this. For example, a comprehensive trained model in which learning data of a plurality of medical institutions and a plurality of countries are aggregated may be generated.

(Other Embodiment-4)
Further, in the above-described embodiment, an example in which the medical image processing apparatus 130 is installed in a medical institution such as a hospital or a clinic has been described, but the embodiment is not limited to this. For example, the medical image processing apparatus 130 is installed at a location different from the medical institution, and is communicably connected to the medical image diagnostic apparatuses 110 installed at one or more medical institutions via the network 200. Is also good.

  In this case, for example, the medical image processing apparatus 130 collects medical images from the medical image diagnostic apparatuses 110 installed in each medical institution via the network 200, and generates or updates a learned model. Then, the medical image processing apparatus 130 receives an image processing request for area extraction from the terminal apparatus 120 installed in each medical institution via the network 200, and estimates the image processing request to the terminal apparatus 120 that transmitted the request. Send the modified area.

  In this case, for example, the medical image processing system 100 is realized as a client-server system in which the terminal device 120 is a client and the medical image processing device 130 is a server. More specifically, for example, the medical image processing system 100 is realized as a cloud system in which the terminal device 120 and the medical image processing device 130 are connected via the Internet or the like.

(Other Embodiment-5)
Further, in the above-described embodiment, an example in which the medical image processing apparatus 130 performs both the processing at the time of learning and the processing at the time of operation has been described, but the embodiment is not limited to this. For example, among the functions of the medical image processing apparatus 130, a function of generating a learned model is provided by another apparatus (hereinafter referred to as a model generation apparatus) that is installed at a location different from the medical institution and is connected via the network 200. ).

  In this case, for example, the model generation apparatus periodically collects medical images from the medical image diagnostic apparatuses 110 installed in each medical institution via the network 200, and generates or updates a learned model. . Then, the medical image processing apparatus 130 acquires the latest learned model from the model generation apparatus and executes the estimation processing.

  In this case, for example, the medical image processing system 100 is realized as a client-server system in which the medical image processing apparatus 130 is a client and the model generation apparatus is a server. More specifically, for example, the medical image processing system 100 is realized as a cloud system in which the model generation device and the medical image processing device 130 are connected via the Internet or the like. In this case, for example, the medical image processing system 100 causes the medical image processing apparatus 130, which is a client, to execute the minimum necessary processing, and causes the model generation apparatus, which is a server, to execute most of the processing. Client).

(Other Embodiment-6)
Further, in each of the above-described embodiments, the display control unit, the acquisition unit, the learning unit, and the estimation unit in this specification correspond to the control function 135a, the control function 135a, the image processing function 135b, and the learning function 135d of the processing circuit 145, respectively. , And an example of the case of realization by the estimation function 135e, but the embodiment is not limited thereto. For example, the display control unit, the acquisition unit, the learning unit, and the estimation unit in this specification are realized by the control function 135a, the control function 135a, the image processing function 135b, the learning function 135d, and the estimation function 135e described in the embodiment. In addition, the same function may be realized only by hardware or a mixture of hardware and software.

  The term “processor” used in the above description may be, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an application specific integrated circuit (ASIC), or a programmable logic device. (For example, a circuit such as a Simple Programmable Logic Device (SPLD), a Complex Programmable Logic Device (CPLD), and a Field Programmable Gate Array (FPGA)). The processor realizes a function by reading and executing the program stored in the storage circuit 132. Instead of storing the program in the storage circuit 132, the program may be directly incorporated in the circuit of the processor. In this case, the processor realizes a function by reading and executing a program incorporated in the circuit. Further, the processor according to the present embodiment is not limited to being configured as a single circuit, but may be configured as a single processor by combining a plurality of independent circuits to realize its function.

  Here, the program (medical image processing program) executed by the processor is provided by being incorporated in a ROM (Read Only Memory) or a storage circuit in advance. This program is a file in a format that can be installed or executed in these devices, such as a CD (Compact Disk) -ROM, an FD (Flexible Disk), a CD-R (Recordable), and a DVD (Digital Versatile Disk). May be provided by being recorded on a computer-readable storage medium. The program may be stored on a computer connected to a network such as the Internet, and provided or distributed by being downloaded via the network. For example, this program is configured by modules including the above-described functional units. As actual hardware, the CPU reads a program from a storage medium such as a ROM and executes the program, whereby each module is loaded on the main storage device and generated on the main storage device.

  According to at least one embodiment described above, the efficiency of a workflow using an image processing application can be improved.

  Although several embodiments of the present invention have been described, these embodiments are provided by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in other various forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and their equivalents.

REFERENCE SIGNS LIST 100 medical image processing system 130 medical image processing device 132 storage circuit 135 processing circuit 135 a control function 135 b image processing function 135 c analysis function 135 d learning function 135 e estimation function

Claims (9)

An extraction unit for acquiring an extraction region in the medical image extracted by the image processing, and a correction region in which a corrector has corrected the extraction region;
A learned model obtained by learning the relationship between the difference between the extracted region and the modified region and the extracted region, wherein the extracted region is modified from the medical image by the image processing. A storage unit that stores a learned model that is functioned to estimate the corrected area after performing
By inputting the extracted region extracted from within the medical image by the image processing to the learned model, an estimating unit that controls to estimate the corrected region,
A medical image processing apparatus comprising: The storage unit stores a result of image analysis on a difference region between the extraction region and the correction region, and a learned model obtained by learning a relationship between the extraction region and
The estimating unit is configured to input a result of image analysis of an extraction region extracted from within a medical image by the image processing to the learned model, thereby performing control to estimate a region to be corrected in the extraction region. The medical image processing apparatus according to claim 1, wherein: The storage unit stores a learned model for each corrector that has made a correction to an extraction region extracted from within a medical image by the image processing,
The said estimation part controls so that the after-correction area | region for each said corrector may be estimated by inputting the extraction area | region extracted from the inside of the medical image with respect to the learned model for every said corrector. Or the medical image processing apparatus according to 2. The storage unit integrates a plurality of image analysis results corresponding to correction results performed by a plurality of correctors on an extraction region extracted from within a medical image by the image processing, and the extraction region Memorized the learned model obtained by learning the relationship of
The estimating unit is configured to input a result of image analysis of an extraction region extracted from within a medical image by the image processing to the learned model, thereby performing control to estimate a region to be corrected in the extraction region. The medical image processing apparatus according to claim 2, which performs the processing.   The medical image processing apparatus according to claim 1, further comprising a learning unit configured to generate the learned model.   The medical image according to any one of claims 1 to 5, further comprising: a display control unit configured to control to display an extraction area in the medical image extracted by the image processing and the corrected area in a distinguishable manner. The medical image processing device according to one of the above.   The medical image processing apparatus according to claim 3, further comprising a display control unit that controls the medical image so that the corrected area for each of the correctors is identifiably displayed. An extraction unit for acquiring an extraction region in the medical image extracted by the image processing, and a correction region in which a corrector has corrected the extraction region;
A learned model obtained by learning the relationship between the difference between the extracted region and the modified region and the extracted region, wherein the extracted region is modified from the medical image by the image processing. A storage unit that stores a learned model that is functioned to estimate the corrected area after performing
By inputting the extracted region extracted from within the medical image by the image processing to the learned model, an estimating unit that controls to estimate the corrected region,
A medical image processing system comprising: An acquisition function of acquiring an extraction area in the medical image extracted by the image processing, and a correction area in which a corrector has corrected the extraction area;
A learned model obtained by learning the relationship between the difference between the extracted region and the modified region and the extracted region, wherein the extracted region is modified from the medical image by the image processing. The modified region is estimated by inputting the extracted region extracted from the medical image by the image processing to the learned model that is provided with a function of estimating the modified region after performing the correction. Estimating function to control
Medical image processing program for causing a computer to realize.

Images