JP7144292B2 - MEDICAL IMAGE PROCESSING APPARATUS AND MEDICAL IMAGE PROCESSING METHOD - Google Patents

Description

TECHNICAL FIELD Embodiments of the present invention relate to a medical image processing apparatus and a medical image processing method.

Conventionally, in an image diagnostic apparatus such as an X-ray CT apparatus, a gain that determines an amplification factor when outputting data collected by a detector is adjusted based on an estimated thickness distribution of a subject. Therefore, a technique for optimizing the settings of the detector so that the maximum count value does not overflow has been disclosed.

However, in the above method, although the reduction of artifacts caused by circuit noise was considered, when imaging high-absorption areas such as the shoulder and pelvis, the maximum count value at the edge of the body surface where absorption is low tends to overflow, and there is a possibility that artifacts caused by overflow cannot be reduced.

JP 2018-050664 A

The problem to be solved by the present invention is to reduce artifacts caused by overflow.

A medical image processing apparatus according to an embodiment includes an acquisition unit, a first determination unit, and a correction unit. The acquisition unit acquires first medical data. The first determination unit determines whether the acquired first medical data includes an overflow artifact. The correcting unit corrects the first medical data determined to include the overflow artifact by the first determining unit for the correction model that outputs the medical data in which the overflow artifact is corrected when the medical data is input. By inputting, medical data in which the overflow artifact of the first medical data is corrected is generated.

1 is a configuration diagram of an X-ray CT apparatus 1 according to a first embodiment; FIG. 4 is a diagram showing an example of data stored in a memory 41; FIG. FIG. 4 is a configuration diagram of an image processing function 54; 4 is a configuration diagram of a first learning function 57; FIG. FIG. 4 is a diagram for explaining the details of processing by a judgment model 41-6; FIG. 11 is a diagram for explaining judgment model generation processing by a judgment model generation function 57-3; FIG. 4 is a configuration diagram of a second learning function 58; FIG. 4 is a diagram for explaining the details of processing by a correction model 41-7; FIG. 11 is a diagram for explaining correction model generation processing of a correction model generation function 58-3; FIG. 4 is a diagram for explaining a method of generating a corrected image 41-5; 4 is a flowchart showing an example of the flow of imaging processing by the X-ray CT apparatus 1; 4 is a flowchart showing an example of the flow of first learning processing by a first learning function 57; 4 is a flowchart showing an example of the flow of second learning processing by a second learning function 58; The block diagram of 57 A of 1st learning functions which concern on 2nd Embodiment. FIG. 4 is a diagram for explaining the details of processing by a type determination model 41-6A; FIG. 11 is a diagram for explaining the type determination model generation processing of the type determination model generation function 57-3A; 4 is a flow chart showing an example of the flow of imaging processing by the X-ray CT apparatus 1A. 5 is a flowchart showing an example of the flow of third learning processing by the first learning function 57A;

A medical image processing apparatus and a medical image processing method according to embodiments will be described below with reference to the drawings. A medical image processing apparatus is an apparatus, such as an X-ray CT (Computed Tomography) apparatus, for diagnosing a subject by processing a medical image.

(First embodiment)
FIG. 1 is a configuration diagram of an X-ray CT apparatus 1 according to the first embodiment. The X-ray CT apparatus 1 has a gantry device 10, a bed device 30, and a console device 40, for example. For convenience of explanation, FIG. 1 shows both a view of the gantry device 10 viewed from the Z-axis direction and a view of the gantry device 10 viewed from the X-axis direction. In the embodiment, the rotation axis of the rotating frame 17 in the non-tilt state or the longitudinal direction of the top plate 33 of the bed device 30 is the Z-axis direction, and the axis perpendicular to the Z-axis direction and horizontal to the floor surface is the X-axis. A direction perpendicular to the Z-axis direction and perpendicular to the floor surface is defined as a Y-axis direction.

The gantry device 10 includes, for example, an X-ray tube 11, a wedge 12, a collimator 13, an X-ray high voltage device 14, an X-ray detector 15, and a data acquisition system (DAS: Data Acquisition System) 16. , a rotating frame 17 and a control device 18 .

The X-ray tube 11 generates X-rays by applying a high voltage from the X-ray high voltage device 14 and irradiating thermal electrons from a cathode (filament) toward an anode (target). X-ray tube 11 includes a vacuum tube. For example, the X-ray tube 11 is a rotating anode type X-ray tube that generates X-rays by irradiating a rotating anode with thermal electrons.

The wedge 12 is a filter for adjusting the dose of X-rays irradiated from the X-ray tube 11 to the subject P. As shown in FIG. The wedge 12 attenuates the X-rays passing through itself so that the X-ray dose distribution irradiated from the X-ray tube 11 to the subject P becomes a predetermined distribution. The wedge 12 is also called a wedge filter or a bow-tie filter. The wedge 12 is, for example, processed aluminum so as to have a predetermined target angle and a predetermined thickness.

The collimator 13 is a mechanism for narrowing down the irradiation range of X-rays transmitted through the wedge 12 . The collimator 13 narrows down the X-ray irradiation range by, for example, forming a slit by combining a plurality of lead plates. The collimator 13 is sometimes called an X-ray diaphragm.

The X-ray high voltage device 14 has, for example, a high voltage generator and an X-ray controller. The high voltage generator has an electric circuit including a transformer, a rectifier, etc., and generates a high voltage to be applied to the X-ray tube 11 . The X-ray controller controls the output voltage of the high voltage generator in accordance with the amount of X-rays to be generated by the X-ray tube 11 . The high voltage generator may be one that boosts the voltage with the transformer described above, or one that boosts the voltage with an inverter.
The X-ray high voltage device 14 may be provided on the rotating frame 17 or may be provided on the fixed frame (not shown) side of the gantry device 10 .

The X-ray detector 15 detects the intensity of X-rays generated by the X-ray tube 11 and incident through the subject P. FIG. The X-ray detector 15 outputs to the DAS 18 an electrical signal (which may be an optical signal or the like) corresponding to the intensity of the detected X-rays. The X-ray detector 15 has, for example, multiple X-ray detection element arrays. Each of the plurality of X-ray detection element arrays has a plurality of X-ray detection elements arranged in the channel direction along an arc centered on the focal point of the X-ray tube 11 . A plurality of X-ray detection element arrays are arranged in a slice direction (column direction, row direction).

X-ray detector 15 is, for example, an indirect detector having a grid, a scintillator array, and a photosensor array. The scintillator array has a plurality of scintillators. Each scintillator has a scintillator crystal. The scintillator crystal emits an amount of light corresponding to the intensity of incident X-rays. The grid has an X-ray shielding plate arranged on the surface of the scintillator array on which X-rays are incident and having a function of absorbing scattered X-rays. Note that the grid may also be called a collimator (one-dimensional collimator or two-dimensional collimator). The photosensor array has photosensors such as, for example, photomultiplier tubes (photomultipliers: PMTs). The photosensor array outputs an electrical signal corresponding to the amount of light emitted by the scintillator. The X-ray detector 15 may be a direct conversion detector having a semiconductor element that converts incident X-rays into electrical signals.

DAS 16 has, for example, an amplifier, an integrator, and an A/D converter. The amplifier amplifies the electrical signal output from each X-ray detection element of the X-ray detector 15 . The integrator integrates the amplified electrical signal over a view period (described later). The A/D converter converts an electrical signal representing the result of integration into a digital signal. The DAS 16 outputs detection data based on digital signals to the console device 40 . Detected data is a digital value of x-ray intensity identified by the channel number of the x-ray detector element from which it was generated, the row number, and the view number indicating the acquired view. The view number is a number that changes according to the rotation of the rotating frame 17, and is a number that is incremented according to the rotation of the rotating frame 17, for example. Therefore, the view number is information indicating the rotation angle of the X-ray tube 11 . A view period is a period that falls between the rotation angle corresponding to a certain view number and the rotation angle corresponding to the next view number. The DAS 16 may detect the switching of the view by a timing signal input from the control device 18, by an internal timer, or by a signal obtained from a sensor (not shown). . When X-rays are continuously emitted from the X-ray tube 11 when performing a full scan, the DAS 16 collects detection data groups for the entire circumference (for 360 degrees). When X-rays are continuously emitted from the X-ray tube 11 when half scanning is performed, the DAS 16 collects detection data for a half circumference (180 degrees).

The rotating frame 17 is an annular member that supports the X-ray tube 11, the wedge 12, the collimator 13, and the X-ray detector 15 so as to face each other. The rotating frame 17 is rotatably supported by the stationary frame around the subject P introduced therein. Rotating frame 17 also supports DAS 16 . Detection data output by the DAS 16 is transmitted by optical communication from a transmitter having a light emitting diode (LED) mounted on the rotating frame 17 to a photodiode mounted on a non-rotating portion (e.g., fixed frame) of the gantry 10. It is transmitted to the receiver and transferred to the console device 40 by the receiver. The method of transmitting the detection data from the rotating frame 17 to the non-rotating portion is not limited to the above-described method using optical communication, and any non-contact transmission method may be employed. The rotating frame 17 is not limited to an annular member, and may be a member such as an arm as long as it can support and rotate the X-ray tube 11 or the like.

The X-ray CT apparatus 1 is, for example, a Rotate/Rotate-type X-ray CT apparatus (third Generation CT), but not limited to this, Stationary/Rotate-Type in which a plurality of annularly arranged X-ray detection elements are fixed to a fixed frame, and the X-ray tube 11 rotates around the subject P. It may be an X-ray CT device (fourth generation CT).

The control device 18 has, for example, a processing circuit having a processor such as a CPU (Central Processing Unit), and a driving mechanism including a motor, an actuator, and the like. The control device 18 receives an input signal from an input interface 43 attached to the console device 40 or the gantry device 10 and controls the operations of the gantry device 10 and the bed device 30 .

The control device 18 rotates the rotating frame 17, tilts the gantry device 10, and moves the top board 33 of the bed device 30, for example. When tilting the gantry device 10 , the control device 18 rotates the rotating frame 17 about an axis parallel to the Z-axis direction based on the tilt angle (tilt angle) input to the input interface 43 . The control device 18 grasps the rotation angle of the rotating frame 17 from the output of a sensor (not shown) or the like. The control device 18 also provides the rotation angle of the rotating frame 17 to the processing circuit 50 at any time. The control device 18 may be provided in the gantry device 10 or may be provided in the console device 40 .

The bed device 30 is a device on which the subject P to be scanned is placed, moved, and introduced into the rotating frame 17 of the gantry device 10 . The bed device 30 has, for example, a base 31 , a bed driving device 32 , a top board 33 and a support frame 34 . The base 31 includes a housing that supports the support frame 34 so as to be movable in the vertical direction (Y-axis direction). The bed driving device 32 includes motors and actuators. The bed driving device 32 moves the tabletop 33 on which the subject P is placed in the longitudinal direction (Z-axis direction) of the tabletop 33 along the support frame 34 . The top plate 33 is a plate-like member on which the subject P is placed.

The bed drive device 32 may move not only the top plate 33 but also the support frame 34 in the longitudinal direction of the top plate 33 . Contrary to the above, the gantry device 10 may be movable in the Z-axis direction, and the rotation frame 17 may be controlled to come around the subject P by moving the gantry device 10 . Moreover, both the gantry device 10 and the top plate 33 may be configured to be movable.

The console device 40 has, for example, a memory 41 , a display 42 , an input interface 43 , a memory 41 , a network connection circuit 44 and a processing circuit 50 . In the embodiment, the console device 40 is described as being separate from the gantry device 10 , but the gantry device 10 may include some or all of the components of the console device 40 .

The memory 41 is implemented by, for example, a RAM (Random Access Memory), a semiconductor memory device such as a flash memory, a hard disk, an optical disk, or the like. The memory 41 stores, for example, detection data, projection data, reconstructed images, CT images, and the like. These data may be stored in an external memory with which the X-ray CT apparatus 1 can communicate, instead of the memory 41 (or in addition to the memory 41). The external memory is controlled by the cloud server, for example, when the cloud server that manages the external memory receives a read/write request.

The display 42 displays various information. For example, the display 42 displays a medical image (CT image) generated by the processing circuit, a GUI (Graphical User Interface) image for receiving various operations by the operator, and the like. The display 42 is, for example, a liquid crystal display, a CRT (Cathode Ray Tube), an organic EL (Electroluminescence) display, or the like. The display 42 may be provided on the gantry device 10 . The display 42 may be of a desktop type, or may be a display device (for example, a tablet terminal) capable of wireless communication with the main body of the console device 40 .

The input interface 43 accepts various input operations by the operator and outputs an electrical signal indicating the content of the accepted input operation to the processing circuit 50 . For example, the input interface 43 accepts acquisition conditions for acquiring detection data or projection data (described later), reconstruction conditions for reconstructing CT images, image processing conditions for generating post-processed images from CT images, and the like. accepts the input operation of For example, the input interface 43 is implemented by a mouse, keyboard, touch panel, drag ball, switch, button, joystick, foot pedal, camera, infrared sensor, microphone, and the like. The input interface 43 may be provided on the gantry device 10 . Also, the input interface 43 may be realized by a display device (for example, a tablet terminal) capable of wireless communication with the main body of the console device 40 .

The network connection circuit 44 includes, for example, a network card having a printed circuit board, or a wireless communication module. The network connection circuit 44 implements an information communication protocol according to the form of the network to be connected. Networks include, for example, LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, cellular networks, dedicated lines, and the like.

A processing circuit 50 controls the overall operation of the X-ray CT apparatus 1 . The processing circuit 50 includes, for example, a system control function 51, a preprocessing function 52, a reconstruction processing function 53, an image processing function 54, a scan control function 55, a display control function 56, a first learning function 57, a second learning function 58, and the like. to run. The processing circuit 50 implements these functions by executing a program stored in the memory 41 by a hardware processor, for example.

A hardware processor is, for example, a CPU, a GPU (Graphics Processing Unit), an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (Simple Programmable Logic Device; SPLD) or Circuitry such as a Complex Programmable Logic Device (CPLD) or a Field Programmable Gate Array (FPGA). Instead of storing the program in the memory 41, the program may be directly embedded in the circuitry of the hardware processor. In this case, the hardware processor realizes its function by reading and executing the program embedded in the circuit. The hardware processor is not limited to being configured as a single circuit, and may be configured as one hardware processor by combining a plurality of independent circuits to implement each function. Also, a plurality of components may be integrated into one hardware processor to realize each function.

Each component of the console device 40 or the processing circuit 50 may be distributed and realized by a plurality of pieces of hardware. The processing circuit 50 may be realized by a processing device that can communicate with the console device 40 instead of the configuration that the console device 40 has. The processing device is, for example, a workstation connected to one X-ray CT device, or a device connected to a plurality of X-ray CT devices and collectively executing processing equivalent to the processing circuit 50 described below (for example, cloud server).

The system control function 51 controls various functions of the processing circuit 50 based on input operations received by the input interface 43 .

A preprocessing function 52 performs preprocessing such as logarithmic conversion processing, offset correction processing, inter-channel sensitivity correction processing, and beam hardening correction on the detection data output from the DAS 16 to generate projection data. The projection data thus obtained are stored in the memory 41 .

The reconstruction processing function 53 performs reconstruction processing on the projection data generated by the preprocessing function 52 using a filtered back projection method, an iterative reconstruction method, or the like, generates a CT image, and generates a CT image. The image is stored in memory 41 .

Based on the input operation received by the input interface 43, the image processing function 54 converts the CT image into a three-dimensional image or cross-sectional image data of an arbitrary cross-section by a known method. Conversion to a three-dimensional image may be performed by preprocessing function 52 .

Further, the image processing function 54 performs artifact correction processing for reducing or removing artifacts from the CT image in which the artifacts are detected, based on the input operation received by the input interface 43 . Artifact correction processing will be described later. In the following, reducing or eliminating artifacts is simply referred to as "mitigating."

The scan control function 55 controls detection data collection processing in the gantry device 10 by instructing the X-ray high-voltage device 14 , the DAS 16 , the control device 18 , and the bed driving device 32 . The scan control function 55 controls the operation of each unit when capturing a scanogram image and when capturing an image used for diagnosis.

The display control function 56 controls the display mode of the display 42 .

The first learning function 57 learns whether or not the CT image processed by the image processing function 54 contains artifacts, and generates a judgment model 41-6, which will be described later. The first learning function 57 is an example of a "judgment model generator".

The second learning function 58 learns the relationship between subject information, a corrected image obtained by reducing artifacts from the CT image processed by the image processing function 54, and the original CT image, and uses a correction model 41-7 to be described later. to generate The second learning function 58 is an example of a "correction model generator".

With the above configuration, the X-ray CT apparatus 1 scans the subject P in a manner such as helical scanning, conventional scanning, and step-and-shoot. Helical scanning is a mode in which the subject P is helically scanned by rotating the rotating frame 17 while moving the top plate 33 . Conventional scanning is a mode in which the subject P is scanned in a circular orbit by rotating the rotation frame 17 while the tabletop 33 is stationary. Run a conventional scan. Step-and-shoot is a mode in which the position of the top plate 33 is moved at regular intervals to perform conventional scanning in a plurality of scan areas.

FIG. 2 is a diagram showing an example of data stored in the memory 41. As shown in FIG. As shown in FIG. 2, the memory 41 stores subject information 41-1 generated by the processing circuit 50, detection data 41-2, projection data 41-3, a reconstructed image 41-4, and a corrected image 41, for example. -5, determination model 41-6, correction model 41-7, etc. are stored.

[Image processing function]
FIG. 3 is a configuration diagram of the image processing function 54. As shown in FIG. The image processing function 54 includes, for example, a reconstructed image acquisition function 54-1, an artifact determination function 54-2, and an artifact correction processing function 54-3.

The reconstructed image acquisition function 54-1 acquires the captured reconstructed image 41-4 and subject information 41-1 associated with the subject P from the memory 41. FIG. The subject information 41-1 includes, for example, information such as ID information for identifying the subject P, information indicating characteristics of the subject P such as the sex, height, and weight of the subject P, and information on the imaging region. The reconstructed image acquisition function 54-1 is an example of an "acquisition unit".

The artifact determination function 54-2 determines whether or not the reconstructed image 41-4 acquired by the reconstructed image acquisition function 54-1 contains an overflow artifact. Overflow artifacts will now be described. Artifacts are noise that appears in reconstructed images. An overflow artifact is noise that appears on a reconstructed image when the maximum count that can be handled by the X-ray detector 15 is exceeded. The artifact determination function 54-2 may receive a determination input by the operator, may use the output result of the determination model 41-6 described later as the determination result, or may use the determination result of another known determination method. may be The artifact determination function 54-2 is an example of the "first determination section".

The artifact correction processing function 54-3 performs artifact correction processing for reducing overflow artifacts in the reconstructed image 41-4 determined to include overflow artifacts by the artifact determination function 54-2. The artifact correction processing function 54-3 is an example of a "corrector".

The reconstructed image 41-4 applied to the judgment model 41-6 and the correction model 41-7 is an example of "first medical data".

Note that the artifact determination function 54-2 may perform artifact determination processing based on the output obtained by inputting the reconstructed image 41-4 to the determination model 41-6, which will be described later, or the input operation by the operator. Artifact determination processing may be performed according to .

The artifact correction processing function 54-3 may perform the artifact correction processing by inputting the reconstructed image 41-4 into the correction model 41-7 described later, or may perform the artifact correction processing in accordance with the operator's input operation. processing may be performed. The artifact correction processing is realized by, for example, application of an artifact correction filter and software processing such as further beam hardening correction. Artifact correction processing using the correction model 41-7 will be described later.

Note that the artifact correction processing function 54-3 selects a suitable correction model 41-7 for the subject information 41-1 acquired by the reconstructed image acquisition function 54-1, and the selected correction model 41-7 , the reconstructed image 41-4 is applied to . A suitable correction model 41-7 is, for example, a learned model obtained by machine-learning the reconstructed image 41-4 of only the subject P. FIG.

Note that the artifact correction processing function 54-3 does not have a correction model 41-7 obtained by machine-learning the reconstructed image 41-4 of only the subject P, and the reconstructed image 41-4 of the subject P and other images. A trained model reflecting the reconstructed image 41-4 of the subject may be selected, or the subject information 41-1 is similar (for example, sex, height, weight, part of the imaged part, etc.) may be selected. Alternatively, a trained model in which the reconstructed image 41-4 of another subject (all similar) is reflected may be selected.

[First learning function]
Processing by the first learning function 57 will be described below. FIG. 4 is a configuration diagram of the first learning function 57. As shown in FIG. The first learning function 57 includes, for example, a reconstructed image acquisition function 57-1, an artifact determination result reception function 57-2, and a determination model generation function 57-3.

The reconstructed image acquisition function 57-1 acquires subject information 41-1 and reconstructed image 41-4, which are learning data, from the memory 41. FIG.

The artifact determination result reception function 57-2 receives the determination result regarding whether or not the reconstructed image 41-4 includes artifacts, which is determined by the artifact determination function 54-2.

The determination model generation function 57-3 reconstructs images based on the reconstructed image 41-4 acquired by the reconstructed image acquisition function 57-1 and the determination result received by the artifact determination result reception function 57-2. A determination model 41-6 is generated that outputs information indicating whether or not the reconstructed image 41-4 contains an overflow artifact when the image 41-4 is input.

[Judgment model]
The judgment model generation function 57-3 uses the reconstructed image 41-4 acquired by the reconstructed image acquisition function 57-1 as learning data, and the reconstructed image 41-4 associated with the reconstructed image 41-4. 4 (determination result determined by the artifact determination result receiving function 57-2) is used as teaching data to perform machine learning to generate a determination model 41-6. Note that the judgment model generation function 57-3 may be realized by an external device.

FIG. 5 is a diagram for explaining the details of processing by the judgment model 41-6. The artifact determination function 54-2 inputs the reconstructed image 41-4 as a parameter to the determination model 41-6, thereby outputting information indicating whether or not the reconstructed image 41-4 contains an overflow artifact.

Note that in the processing shown in FIG. 5, the projection data 41-3 may be used as an input parameter instead of the reconstructed image 41-4.

FIG. 6 is a diagram for explaining judgment model generation processing by the judgment model generation function 57-3. The judgment model generation function 57-3 generates a machine learning model (hereinafter sometimes referred to as a first machine learning model) in which connection information etc. are defined in advance and parameters such as connection coefficients are provisionally set. Then, a plurality of sets of learning data are input, and the parameters in the machine learning model are adjusted so that the result approaches the teacher data corresponding to the learning data. The judgment model generation function 57-3 uses, for example, a reconstructed image 41-4 of a certain subject P included in the subject information 41-1 as learning data, and an overflow artifact in the reconstructed image 41-4 at the time of imaging. A machine learning model is generated that generates a trained model using information indicating whether or not is included as teacher data.

The judgment model generation function 57-3 adjusts the parameters of the machine learning model by, for example, back propagation (back propagation method). The machine learning model is, for example, a DNN (Deep Neural Network) using a CNN (Convolution Neural Network). Note that the machine learning model may perform weighting setting for each learning data so that newer learning data is more likely to be reflected as an output.

The judgment model generation function 57-3 ends the process after performing back propagation on a predetermined number of sets of learning data and the corresponding teacher data. The machine learning model at that time is stored as the judgment model 41-6. Note that the judgment model generation function 57-3 uses a set of the reconstructed image 41-4 of a subject P and information indicating whether or not the reconstructed image 41-4 contains an overflow artifact. A judgment model 41-6 may be generated.

[Second learning function]
Processing by the second learning function 58 will be described below. FIG. 7 is a configuration diagram of the second learning function 58. As shown in FIG. The second learning function 58 includes, for example, a reconstructed image acquisition function 58-1, an artifact correction result reception function 58-2, and a correction model generation function 58-3.

The reconstructed image acquisition function 58-1 acquires from the memory 41 the object information 41-1 and reconstructed image 41-4, which are learning data, and the corrected image 41-5, which is teacher data.

The artifact correction result reception function 58-2 receives information indicating whether or not the artifact correction processing function 54-3 has performed correction processing on the reconstructed image 41-4, which is the learning data.

The correction model generation function 58-3 determines whether or not the correction processing accepted by the artifact correction result acceptance function 58-2 has been performed with the reconstructed image 41-4 acquired by the reconstructed image acquisition function 58-1. A correction model 41-7 is generated based on the indicated information.

[Correction model]
FIG. 8 is a diagram for explaining the contents of processing by the correction model 41-7. The artifact correction processing function 54-3 corrects overflow artifacts contained in the reconstructed image 41-4 by inputting the subject information 41-1 and the reconstructed image 41-4 as parameters into the correction model 41-7. A corrected image 41-5 is output.

In the processing shown in FIG. 8, it is preferable to use only the reconstructed image 41-4 of the same subject P as learning data to improve the accuracy of the subsequent processing of artifact correction. However, at the stage where the required number of learning data cannot be collected, the reconstructed image 41-4 associated with the subject information 41-1 similar to the subject P may be included in the learning data.

FIG. 9 is a diagram for explaining the correction model generation processing of the correction model generation function 58-3. The correction model generation function 58-3 generates a machine learning model in which connection information etc. are defined in advance and parameters such as connection coefficients are provisionally set (hereinafter sometimes referred to as a second machine learning model). Then, a plurality of sets of learning data are input, and the parameters in the machine learning model are adjusted so that the result approaches the teacher data corresponding to the learning data. The correction model generation function 58-3 uses, for example, a set of object information 41-1 and a reconstructed image 41-4 of a certain object P, which are included in the object information 41-1, as learning data, and performs imaging thereof. A machine learning model is generated for generating a learned model using a corrected image 41-5 obtained by correcting the reconstructed image 41-4 at the time as teacher data.

The correction model generation function 58-3, for example, similarly to the judgment model generation function 57-3, adjusts the parameters of the machine learning model such as DNN by back propagation. Note that the machine learning model may perform weighting setting for each learning data so that newer learning data is more likely to be reflected as an output.

The correction model generation function 58-3 ends the process after performing back propagation on a predetermined number of sets of learning data and the corresponding teacher data. The machine learning model at that time is stored as a correction model 41-7. Note that the correction model generation function 58-3 generates a set of the object information 41-1 of a certain object P and the reconstructed image 41-4, and the corrected image 41-5 associated with the reconstructed image 41-4. and a set of , may be used to generate the correction model 41-7.

Note that the correction model generation function 58-3 generates the reconstructed image 41-4 of only the subject P of the learning data of the number of sets equal to or more than the predetermined number. Generate a machine-learned trained model. The correction model generating function 58-3 generates the reconstructed image 41-4 of the subject P and the reconstructed images 41-4 of the other subjects when the predetermined number of sets of the reconstructed image 41-4 of only the subject P does not exist. A correction model 41-7 that reflects the configuration image 41-4 may be generated, or other models having similar subject information 41-1 (for example, similar sex, height, weight, and radiographed parts) may be generated. A correction model 41-7 may be generated in which the reconstructed image 41-4 of the subject is reflected.

[artificial artifact]
A correction method using artificial artifacts will be described below. FIG. 10 is a diagram for explaining the method of generating the corrected image 41-5. In the area a1+a2 of the X-ray detector 15 that detects the amount of X-rays irradiated from the X-ray tube 11 to the subject P, the area a1 is the case where the reconstructed image 41-4 can be acquired as expected (no overflow). However, the maximum value of the count value overflows in the area a2 that detects X-rays that have passed through the imaging site that is likely to cause overflow (the edge of the body surface where absorption is low when measuring the shoulder or pelvis). There is

When an overflow of the count value of the detection data 41-2 associated with the area a2 is detected, the artifact correction processing function 54-3 makes the detection data 41-2 associated with the areas a1 and a2 to be corrected. The artifact correction processing function 54-3 does not only correct the count value of the detection data 41-2 associated with the area a2.

The artifact correction processing function 54-3, for example, among the reconstructed images 41-4 acquired in the past, detects the detected data 41-2 associated with the image that the operator has confirmed does not contain the overflow artifact. is subjected to a count gain that reproduces the reconstructed image 41-4 in which the overflow artifact is detected. The count gain is, for example, to increase the count value associated with area a2 to a value that reproduces the overflow state. The artifact correction processing function 54-3 thus generates a reconstructed image to which artificial overflow artifacts based on the imaging region are added. The artifact correction processing function 54-3 performs artifact correction processing by performing the above-described inverse calculation of the count gain on the reconstructed image 41-4 in which overflow artifacts are detected.

Further, the correction model generation function 58-3 may use the reconstructed image to which the above artificial overflow artifact is added and the object information thereof as learning data, or may use the original reconstructed image as learning data.

[Processing flow 1]
FIG. 11 is a flowchart showing an example of the flow of imaging processing by the X-ray CT apparatus 1. As shown in FIG.

First, the reconstructed image acquisition function 54-1 acquires the reconstructed image 41-4 and subject information 41-1 associated with the reconstructed image 41-4 (step S100). Next, the artifact determination function 54-2 determines whether or not the reconstructed image 41-4 contains an overflow artifact (step S102). When the artifact determination function 54-2 determines that the overflow artifact is not included, the processing ends.

When the artifact determination function 54-2 determines that an overflow artifact is included in step S102, the correction model 41-7 stored in the memory 41 includes only the reconstructed image of the same subject as learning data. 41-7 exists (step S104).

If the artifact determination function 54-2 determines in step S104 that only the reconstructed image of the same subject is present in the correction model 41-7, the artifact correction processing function 54-3 reconstructs the image using the correction model 41-7. By inputting the image 41-4 and the associated object information 41-1, a corrected image 41-5 in which the overflow artifact is corrected is obtained (step S106).

The artifact determination function 54-2 determines whether or not there is another suitable correction model 41-7 if it is not determined in step S104 that there is a trained model for only the reconstructed image of the same subject. (Step S108). When the artifact determination function 54-2 determines that there is another suitable learned model, it inputs the reconstructed image 41-4 and the associated subject information 41-1 to the selected correction model 41-7. to obtain a corrected image 41-5 in which artifacts have been corrected (step S110). If the artifact determination function 54-2 does not determine that there is another suitable trained model, the process proceeds to step S112, which will be described later.

After the process of step S106 or S110, the first learning function 57 and the second learning function each save learning data (step S112). Note that in the processing of step S112, the artifact correction processing function 54-3 may accept an operation for artifact correction (re-correction or fine adjustment) by the operator. This completes the description of the processing of this flowchart.

[Processing flow 2]
FIG. 12 is a flow chart showing an example of the flow of the first learning process by the first learning function 57. As shown in FIG. The flowchart of FIG. 12 is performed, for example, when a predetermined number of sets or more of learning data are stored in step S114 of FIG.

First, the first learning function 57 acquires one set of learning data (step S200). Next, the first learning function 57 inputs the one set of learning data acquired in step S200 to the first machine learning model (step S202), and back-propagates the error from the teacher data corresponding to the one set of learning data. (step S204).

Next, the first learning function 57 determines whether or not the processing of steps S202 and S204 has been performed for a predetermined number of sets of learning data (step S206). If the processing of steps S202 and S204 has not been performed for the predetermined number of sets of learning data, the first learning function 57 returns the processing to step S200. When the processing of steps S202 and S204 is performed for a predetermined number of sets of learning data, the first learning function 57 determines the judgment model 41-6 using the parameters at that time (step S208), and performs the processing of this flowchart. finish.

[Processing flow 3]
FIG. 13 is a flow chart showing an example of the flow of the second learning process by the second learning function 58. As shown in FIG. The flowchart of FIG. 13 is performed, for example, when a predetermined number of sets or more of learning data are saved in step S114 of FIG.

First, the second learning function 58 acquires one set of learning data (step S300). Next, the second learning function 58 inputs the set of learning data acquired in step S300 to the second machine learning model (step S302), and back-propagates the error from the teacher data corresponding to the set of learning data. (step S304).

Next, the second learning function 58 determines whether or not the processing of steps S302 and S304 has been performed for a predetermined number of sets of learning data (step S306). If the processing of steps S302 and S304 has not been performed for the predetermined number of sets of learning data, the second learning function 58 returns the processing to step S300. When the processing of steps S302 and S304 is performed for a predetermined number of sets of learning data, the second learning function 58 determines the correction model 41-7 using the parameters at that time (step S308), and performs the processing of this flowchart. finish.

The reconstructed image 41-4 processed by the image processing function 54 may be a sinogram generated based on the projection data 41-3. A sinogram is a plot of x-rays passing through the body as a function of position along x-ray detector 15 for various projection measurements, or as a function of projection angle between x-ray tube 11 and x-ray detector 15. Data showing attenuation.

According to the X-ray CT apparatus 1 of the first embodiment described above, the reconstructed image acquisition function 54-1 for acquiring the reconstructed image 41-4 and the subject information 41-1, and the reconstructed image 41- 4 includes an overflow artifact. An artifact determination function 54-2 that outputs information indicating whether or not the image 41-4 contains an overflow artifact, and based on the reconstructed image 41-4, the overflow artifact contained in the reconstructed image 41-4 is corrected. By inputting the reconstructed image 41-4 acquired by the reconstructed image acquisition function 54-1 to the correction model 41-7 that outputs the corrected image 41-5, the reconstructed image acquisition function 54- and the artifact determination function 54-2 for generating the corrected image 41-5 in which the overflow artifact of 1 is corrected, it is possible to suppress the occurrence of the overflow artifact.

(Second embodiment)
The X-ray CT apparatus 1A of the second embodiment will be described below. In the following description, the same reference numerals as in the first embodiment are given to the same configurations and functions as in the first embodiment, and detailed description thereof will be omitted. Further, "A" is added to the end of the reference numerals for components that have the same names as those of the first embodiment but have different configurations or functions.

FIG. 14 is a configuration diagram of the first learning function 57A according to the second embodiment. The first learning function 57A includes, for example, a reconstructed image acquisition function 57-1, an artifact type reception function 57-2A, and a type determination model generation function 57-3A.

The artifact type receiving function 57-2A receives the type of artifact included in the reconstructed image 41-4 acquired by the reconstructed image acquiring function 57-1. The artifact type reception function 57-2A is an example of the "second determination unit".

The artifact type receiving function 57-2A may, for example, receive input by the operator of the type of artifact included in the reconstructed image 41-4 acquired by the reconstructed image acquiring function 57-1. Information indicating the type of artifact output by the type determination model 41-6A may be accepted. Further, the artifact type receiving function 57-2A may receive the type of artifact indirectly input by the operator. Indirect input means, for example, inferring the type of artifact from the content of the correction processing performed by the artifact correction processing function 54-3, and using the result of the analogy as an input.

The type determination model generation function 57-3A generates a type determination model 41-6A for determining the type of artifact included in the reconstructed image 41-4 acquired by the reconstructed image acquisition function 57-1. The type determination model generation function 57-3A is an example of a "type determination model generation unit".

The types of artifacts include, for example, in addition to overflow artifacts, artifacts indicating an abnormal output of the X-ray detection element (ring artifact), artifacts indicating an abnormality in the X-ray detector 15 and discharge of the X-ray tube 11 (shower artifact), Artifacts generated by signal transmission noise (moire artifacts), artifacts caused by a large decrease or lack of detected photons (dark band artifacts), and the like. It should be noted that it is difficult to correct images such as artifacts (motion artifacts) caused by the movement of the subject P during imaging and artifacts (metal artifacts) caused by metals in the body of the subject P such as implants and artificial joints. Artifacts that normally require recapture may also be included in the above types of artifacts. The type determination model 41-6A, for example, may be generated for each type of artifact described above and may determine the type of a single artifact. It may also determine the type of artifact.

The type determination model generation function 57-3A generates a type determination model 41-6A. The type determination model 41-6A will be described later.

[Type determination model]
The type determination model generation function 57-3A uses the reconstructed image 41-4 acquired by the reconstructed image acquisition function 57-1 as learning data, and determines the artifact type associated with the reconstructed image 41-4. By performing machine learning using the indicated information as teacher data, the type determination model 41-6A is generated. The type determination model generation function 57-3A may be realized by an external device.

FIG. 15 is a diagram for explaining the details of processing by the type determination model 41-6A. The artifact determination function 54-2A inputs the reconstructed image 41-4 as a parameter to the type determination model 41-6A, thereby outputting information indicating the type of artifact included in the reconstructed image 41-4.

The artifact determination function 54-2A selects a suitable type determination model 41-6A, and determines the presence or absence and type of artifact based on the output obtained by inputting the reconstructed image 41-4 to the type determination model 41-6A. may be performed. Further, the artifact determination function 54-2A receives an input operation indicating selection of the type determination model 41-6A by the operator, and inputs the reconstructed image 41-4 to the selected type determination model 41-6A. The presence or absence and type of artifact may be determined based on the output obtained.

Note that in the processing shown in FIG. 15, the projection data 41-3 may be input as an input parameter instead of the reconstructed image 41-4.

The artifact correction processing function 54-3 may receive an operator's input regarding correction of the reconstructed image 41-4 and generate a corrected image 41-5. Further, the artifact correction processing function 54-3 selects a suitable correction model 41-7 based on the determination result of the artifact determination function 54-2A, and converts the selected correction model 41-7 into the reconstructed image 41-4. may be input to generate the corrected image 41-5.

FIG. 16 is a diagram for explaining the type determination model generation processing of the type determination model generation function 57-3A. The type determination model generation function 57-3A creates a machine learning model (hereinafter sometimes referred to as a third machine learning model) in which connection information etc. are defined in advance and parameters such as connection coefficients are provisionally set. On the other hand, a plurality of sets of learning data are input, and the parameters in the third machine learning model are adjusted so that the result approaches teacher data corresponding to the learning data. The judgment model generation function 57-3 generates a trained model using, for example, the reconstructed image 41-4 as learning data and information indicating the type of artifact contained in the reconstructed image 41-4 as training data. Generate 3 machine learning models.

The type determination model generation function 57-3A, for example, similarly to the determination model generation function 57-3 of the first embodiment, uses back propagation (back propagation method) such as DNN to generate the parameters of the third machine learning model. to adjust. Note that the third machine learning model may perform weighting setting for each learning data so that newer learning data is more likely to be reflected as an output.

When the type determination model generation function 57-3A performs back propagation on a predetermined number of sets of learning data and the corresponding teacher data, the process ends. The third machine learning model at that time is stored as the type determination model 41-6A. The type determination model generation function 57-3A generates a type determination model 41-6A using a set of a reconstructed image 41-4 and information indicating the type of artifact included in the reconstructed image 41-4. may be generated.

[Processing flow 4]
FIG. 17 is a flow chart showing an example of the flow of imaging processing by the X-ray CT apparatus 1A.

First, the reconstructed image acquisition function 54-1 acquires the reconstructed image 41-4 (step S400). Next, the artifact determination function 54-2A inputs the reconstructed image 41-4 to the type determination model 41-6A and acquires information indicating the type of artifact (step S402). Next, the image processing function 54 instructs the display control function 56 to output the identified artifact type to the display 42 (step S404), and receives an artifact correction operation by the operator (step S406). In step S406, similarly to the first embodiment, the corrected image 41-5 may be generated by inputting the reconstructed image 41-4 into the corrected model 41-7. Next, the first learning function 57A stores the processing result of step S406 as information indicating the type of artifact, which is learning data (step S408). This completes the description of the processing of this flowchart.

[Processing flow 5]
FIG. 18 is a flow chart showing an example of the flow of the third learning process by the first learning function 57A. The flowchart of FIG. 18 is executed, for example, when a predetermined number of sets or more of learning data are saved in step S408 of FIG. will be

First, the first learning function 57A acquires one set of learning data (step S500). Next, the first learning function 57A inputs the one set of learning data acquired in step S200 to the third machine learning model (step S502), and back-propagates the error from the teacher data corresponding to the one set of learning data. (step S504).

Next, the first learning function 57A determines whether or not the processing of steps S502 and S504 has been performed for a predetermined number of sets of learning data (step S506). If the processing of steps S502 and S504 has not been performed for the predetermined number of sets of learning data, the first learning function 57A returns the processing to step S500. When the processing of steps S502 and S504 is performed for a predetermined number of sets of learning data, the first learning function 57A determines the type determination model 41-6A using the parameters at that time (step S508), and performs the processing of this flowchart. exit.

According to the X-ray CT apparatus 1A of the second embodiment described above, the reconstructed image acquisition function 54-1 for acquiring the reconstructed image 41-4 and the type of artifact included in the reconstructed image 41-4 are By providing an artifact determination function 54-2A that determines the type of artifact by inputting the reconstructed image 41-4 to the output type determination model 41-6A, the load required for correcting the artifact ( In particular, the load required for determining the type of artifact and the load required for selecting a correction method according to the type of artifact can be reduced.

The embodiment described above can be expressed as follows.
a hardware processor;
a storage device storing a program,
obtaining first medical data;
determining whether the acquired first medical data includes an overflow artifact;
By inputting the first medical data determined to include the overflow artifact to a correction model that outputs medical data in which the overflow artifact is corrected when the medical data is input, generating medical data corrected for medical data overflow artifacts;
A medical image processing apparatus configured as follows.

According to the X-ray CT apparatus of at least one embodiment described above, the acquisition unit (54-1) for acquiring the first medical data (41-4) and the acquired first medical data overflow For a first determination unit (54-2) that determines whether or not an artifact is included, and a correction model (41-7) that outputs medical data in which overflow artifacts are corrected when medical data is input, Correction for generating medical data (41-5) in which the overflow artifact of the first medical data is corrected by inputting the first medical data determined to include the overflow artifact by the first determining unit. By providing the part (54-3), it is possible to reduce artifacts caused by overflow.

While several embodiments of the invention have been described, these embodiments have been presented by way of example and are not intended to limit the scope of the invention. These embodiments can be embodied in various other forms, and various reductions, substitutions, and alterations can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and spirit of the invention, as well as the scope of the invention described in the claims and equivalents thereof.

DESCRIPTION OF SYMBOLS 1... X-ray CT apparatus 10... Mounting apparatus 11... X-ray tube 12... Wedge 13... Collimator 14... X-ray high-voltage apparatus 15... X-ray detector 16... Data collection system 17... Rotating frame 18... Control device 30... Bed Device 31 Base 32 Bed drive device 33 Top plate 34 Support frame 40 Console device 50 Processing circuit 54-1 Reconstructed image acquisition functions 54-2, 54-2A Artifact determination function 54-3 Artifact correction processing functions 57, 57A First learning function 57-3 Judgment model generation function 57-3A Type judgment model generation function 58 Second learning function 58-3 Correction model generation function

Claims (9)

an acquisition unit that acquires first medical data and subject information associated with the first medical data ;
a first determination unit that determines whether the acquired first medical data includes an overflow artifact;
When the medical data and the subject information associated with the medical data are input, the first determination unit determines that the overflow artifact is included in the correction model that outputs the medical data in which the overflow artifact is corrected. a correction unit that generates medical data in which overflow artifacts of the first medical data are corrected by inputting the first medical data and the subject information associated with the first medical data; ,
A medical image processing apparatus comprising: The first determination unit inputs the obtained first medical data to a determination model that outputs information indicating whether an overflow artifact is included when the first medical data is input. determining whether the obtained first medical data includes an overflow artifact;
The medical image processing apparatus according to claim 1. When there is a correction model that reflects medical data corrected for overflow artifacts of the same subject as the subject identified by the subject information associated with the first medical data, the correction unit correcting the medical data using a correction model that reflects the medical data in which the overflow artifact of the same subject is corrected;
The medical image processing apparatus according to claim 1 or 2. further comprising a correction model generation unit that generates the correction model,
The correction model generation unit generates the correction model using medical data obtained by adding an artificial artifact based on an imaging region of a subject in medical data not including overflow artifacts to the medical data not including overflow artifacts. ,
The medical image processing apparatus according to any one of claims 1 to 3. The correction model generation unit adds the artificial artifact to the medical data that does not contain the overflow artifact by increasing the count value for the medical data that does not contain the overflow artifact.
The medical image processing apparatus according to claim 4. a determination model generation unit that generates a determination model using the first medical data as learning data and a determination result of the first medical data by the first determination unit as teacher data;
The medical image processing apparatus according to any one of claims 1 to 5, further comprising: the computer
obtaining first medical data and subject information associated with the first medical data ;
determining whether the acquired first medical data includes an overflow artifact;
The first method determined to include the overflow artifact with respect to a correction model that outputs medical data in which the overflow artifact is corrected when medical data and subject information associated with the medical data are input. generating medical data in which an overflow artifact of the first medical data is corrected by inputting medical data and the subject information associated with the first medical data;
Medical image processing method. By inputting the acquired first medical data to a type determination model that outputs information indicating the type of artifact contained in the medical data based on the medical data, the acquired first medical data includes Further comprising a second determination unit that determines the type of artifact,
The medical image processing apparatus according to any one of claims 1 to 6 . further comprising a type determination model generation unit that generates the type determination model,
The type determination model generation unit generates the type determination model by performing machine learning using the first medical data as learning data and information indicating the type of the artifact included in the first medical data as training data. generate,
The medical image processing apparatus according to claim 8.

Images