Ultrasound diagnostic equipment

Abstract

To accurately obtain a discrimination result for a feature part.SOLUTION: An ultrasonic diagnostic device comprises a storage unit and a discrimination unit. The storage unit stores each of a plurality of learned models that corresponds to each of a plurality of stages and is used when executing each of a plurality of pieces of discrimination processing for outputting a discrimination result for a feature part of a subject drawn in input medical image data. The discrimination unit derives the discrimination result for the feature part of the subject by executing the discrimination processing corresponding to at least one learned model by using at least one learned model of the plurality of learned models. The discrimination unit executes the different discrimination processing from the discrimination processing executed in other stages in each of the plurality of stages.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Embodiments of the present invention relate to ultrasonic diagnostic apparatuses Place .

Background Art

[0002] An ultrasonic diagnostic apparatus displays, in real time, an ultrasonic image indicated by ultrasonic image data generated in real time for ultrasonic scanning, or displays an ultrasonic image indicated by ultrasonic image data obtained by past ultrasonic scanning. The real-time performance of an ultrasonic diagnostic apparatus is superior to that of other medical image diagnostic apparatuses, which is a great advantage. Also, a CAD (Computer Aided Detection) function for automatically detecting characteristic parts (characteristic parts) in medical images generated by medical image diagnostic apparatuses is known.

Prior Art Documents

Patent Documents

[0003] ''

Patent Document 1

Patent Document 2

Patent Document 3

Patent Document 4

Patent Document 5

Summary of the Invention

Problems to be Solved by the Invention

[0004] The problem to be solved by the present invention is to provide an ultrasonic diagnostic apparatus and an analysis apparatus capable of accurately obtaining a discrimination result regarding a characteristic part.

Means for Solving the Problems

[0005] The ultrasound diagnostic apparatus of this embodiment comprises a storage unit and a discrimination unit. The storage unit stores each of a plurality of trained models corresponding to each of a plurality of stages, which are used when executing each of a plurality of discrimination processes that output discrimination results regarding characteristic areas of a subject depicted in the input medical image data. The discrimination unit derives discrimination results regarding characteristic areas of a subject by executing a discrimination process corresponding to at least one of the plurality of trained models using at least one of the trained models. In each of the plurality of stages, the discrimination unit executes a discrimination process different from the discrimination process executed in the other stages. [Brief explanation of the drawing]

[0006] [Figure 1] Figure 1 shows an example of the configuration of an ultrasound diagnostic system according to the first embodiment. [Figure 2] Figure 2 is a diagram illustrating an example of the processing performed by the trained model generation function according to the first embodiment. [Figure 3] Figure 3 is a diagram illustrating an example of the processing performed by the discrimination function according to the first embodiment. [Figure 4] Figure 4 is a diagram illustrating an example of the first stage of processing during learning and the first stage of processing during operation according to the first embodiment. [Figure 5] Figure 5 is a diagram illustrating an example of the second stage of processing during learning and the second stage of processing during operation according to the first embodiment. [Figure 6] Figure 6 is a diagram illustrating an example of the third stage of processing during learning and the third stage of processing during operation according to the first embodiment. [Figure 7] Figure 7 is a flowchart showing an example of the flow of the first discrimination process according to the first embodiment. [Figure 8] Figure 8 is a diagram illustrating an example of the detection result of a feature region by the discrimination function according to the first embodiment. [Figure 9] Figure 9 is a diagram illustrating an example of the detection result of a feature region by the discrimination function according to the first embodiment. [Figure 10] Figure 10 shows an example of the results of the first detection process according to the first embodiment. [Figure 11] Figure 11 shows an example of the results of the second detection process according to the first embodiment. [Figure 12] Figure 12 is a flowchart showing an example of the process performed by the ultrasound diagnostic device according to the first embodiment. [Figure 13] Figure 13 is a diagram illustrating an example of the processing performed by the marker information generation function according to the first embodiment. [Figure 14] Figure 14 shows an example of a display according to the first embodiment. [Figure 15] Figure 15 shows an example of the configuration of an ultrasound diagnostic apparatus according to the second embodiment. [Figure 16] Figure 16 is a flowchart showing an example of the process for generating a second trained model and a second discrimination program according to the second embodiment. [Figure 17] Figure 17 is a flowchart showing an example of the process for generating a third trained model and a third discrimination program according to the second embodiment. [Figure 18] Figure 18 is a flowchart showing an example of the flow of processing performed by an ultrasound diagnostic device according to the first embodiment and a modified example 2 of the second embodiment. [Figure 19] Figure 19 is a flowchart showing an example of the flow of the first discrimination process according to the first embodiment and modification 3 of the second embodiment. [Figure 20] Figure 20 shows an example of the results of the third detection process according to the first embodiment and modification 3 of the second embodiment. [Figure 21] Figure 21 shows an example of the results of the fourth detection process according to the first embodiment and modification 3 of the second embodiment. [Figure 22]FIG. 22 is a diagram showing an example of the result of the fifth detection process according to the first embodiment and Modification 3 of the second embodiment. [Figure 23] FIG. 23 is a diagram for explaining an example of various learned models and various discrimination programs according to Modification 4 of the first embodiment and the second embodiment. [Figure 24] FIG. 24 is a diagram showing a configuration example of the medical image processing apparatus 300 according to the third embodiment.

MODE FOR CARRYING OUT THE INVENTION

[0007] Hereinafter, an ultrasonic diagnostic apparatus and a medical image processing apparatus according to an embodiment will be described with reference to the drawings. Note that the content described in one embodiment or modification may be similarly applied to other embodiments or other modifications.

[0008] (First Embodiment) First, a configuration example of the analysis system according to the first embodiment will be described. FIG. 1 is a diagram showing a configuration example of an ultrasonic diagnostic system according to the first embodiment. As illustrated in FIG. 1, the ultrasonic diagnostic system according to the first embodiment includes an ultrasonic diagnostic apparatus 1 and a learned model generation apparatus 200. The ultrasonic diagnostic apparatus 1 and the learned model generation apparatus 200 are communicably connected to each other. In the ultrasonic diagnostic system according to the first embodiment, the learned model generation apparatus 200 generates various learned models. Further, the learned model generation apparatus 200 records learning conditions such as the network structure and calculation parameters used for various learning. Then, the learned model generation apparatus 200 transmits the generated various learned models and the recorded various learning conditions to the ultrasonic diagnostic apparatus 1. Then, the ultrasonic diagnostic apparatus 1 executes various discrimination processes using the various learned models and the various learning conditions transmitted from the learned model generation apparatus 200. When the ultrasonic diagnostic apparatus 1 executes a certain discrimination process, it uses a certain learned model and the learning conditions corresponding to this learned model.

[0009] The trained model generation device 200 includes a processing circuit 201. The processing circuit 201 is implemented, for example, by a processor. The processing circuit 201 has a trained model generation function 201a.

[0010] The pre-trained model generation function 201a performs training to generate a pre-trained model used when performing discrimination processing to output discrimination results regarding the characteristic area of ​​subject P depicted in ultrasound image data. In the following explanation, the case where the characteristic area is a breast tumor is described, but the characteristic area is not limited to this.

[0011] Figure 2 is a diagram illustrating an example of the processing performed by the pre-trained model generation function 201a according to the first embodiment. As illustrated in Figure 2, in the first stage of training, the pre-trained model generation function 201a generates a first pre-trained model 11a and records the first training conditions 11b. In the second stage following the first stage of training, the pre-trained model generation function 201a generates a second pre-trained model 12a and records the second training conditions 12b. In the third stage following the second stage of training, the pre-trained model generation function 201a generates a third pre-trained model 13a and records the third training conditions 13b. The pre-trained model generation function 201a generates the first pre-trained model 11a, the second pre-trained model 12a, and the third pre-trained model 13a asynchronously. Furthermore, the pre-trained model generation function 201a asynchronously records the first training condition 11b, the second training condition 12b, and the third training condition 13b. The pre-trained model generation function 201a is an example of a generation unit.

[0012] For example, the first learning condition 11b is a set of computational parameters used when the ultrasound diagnostic device 1 performs the first discrimination process. The first trained model 11a is a set of adjusted parameters used when the ultrasound diagnostic device 1 performs the first discrimination process. Details of the first discrimination process will be described later.

[0013] Furthermore, the second learning condition 12b is a set of computational parameters used when the ultrasound diagnostic device 1 performs the second discrimination process. The second trained model 12a is a set of adjusted parameters used when the ultrasound diagnostic device 1 performs the second discrimination process. Details of the second discrimination process will be described later.

[0014] Furthermore, the third learning condition 13b is a set of computational parameters used when the ultrasound diagnostic device 1 performs the third discrimination process. The third trained model 13a is a set of adjusted parameters used when the ultrasound diagnostic device 1 performs the third discrimination process. Details of the third discrimination process will be described later.

[0015] Thus, each of the multiple pre-trained models and each of the multiple training conditions corresponds to each of the multiple stages. Furthermore, each of the multiple pre-trained models and each of the multiple training conditions is used when executing each of the multiple classification processes.

[0016] Returning to the explanation of Figure 1, the ultrasound diagnostic device 1 includes a main unit 100, an ultrasound probe 101, an input device 102, a display 103, and a velocity detector 104.

[0017] The ultrasonic probe 101 is used to collect ultrasonic image data. The ultrasonic probe 101 has, for example, a plurality of piezoelectric transducers. The plurality of piezoelectric transducers generate ultrasonic waves based on drive signals supplied from a transmission circuit 110 of the device body 100, which will be described later. The ultrasonic probe 101 also receives reflected waves from the subject P, converts them into electrical signals (reflected wave signals), and transmits the reflected wave signals to the device body 100. The ultrasonic probe 101 also has, for example, a matching layer provided on the piezoelectric transducers and a backing material to prevent the propagation of ultrasonic waves backward from the piezoelectric transducers. The ultrasonic probe 101 is detachably connected to the device body 100.

[0018] When ultrasound is transmitted from the ultrasound probe 101 to the subject P, the transmitted ultrasound is reflected one after another by discontinuities in acoustic impedance within the subject P's internal tissues, and the reflected waves are received by multiple piezoelectric transducers on the ultrasound probe 101. The amplitude of the received reflected waves depends on the difference in acoustic impedance at the discontinuities where the ultrasound is reflected. Furthermore, when the transmitted ultrasound pulse is reflected by a moving blood flow or the surface of the heart wall, the reflected waves undergo frequency shifts due to the Doppler effect, depending on the velocity component of the moving object relative to the ultrasound transmission direction.

[0019] The ultrasound probe 101 is detachably attached to the main body 100 of the device. When scanning a two-dimensional area within the subject P (two-dimensional scanning), the operator of the ultrasound diagnostic device 1 connects a 1D array probe, for example, in which multiple piezoelectric transducers are arranged in a line, to the main body 100 as the ultrasound probe 101. 1D array probes include linear ultrasound probes, convex ultrasound probes, sector ultrasound probes, etc. When scanning a three-dimensional area within the subject P (three-dimensional scanning), the operator connects a mechanical 4D probe or a 2D array probe to the main body 100 as the ultrasound probe 101. A mechanical 4D probe can perform two-dimensional scanning using multiple piezoelectric transducers arranged in a line, similar to a 1D array probe, and can also perform three-dimensional scanning by oscillating the multiple piezoelectric transducers at a predetermined angle (oscillation angle). Furthermore, the 2D array probe enables three-dimensional scanning using multiple piezoelectric transducers arranged in a matrix, and also enables two-dimensional scanning by focusing and transmitting ultrasound waves.

[0020] The input device 102 can be implemented using input means such as a mouse, keyboard, buttons, panel switches, touch command screen, foot switch, trackball, or joystick. The input device 102 receives various setting requests from the operator of the ultrasound diagnostic device 1 and forwards the received setting requests to the device body 100. For example, the input device 102 receives instructions (execution instructions) from the operator of the ultrasound diagnostic device 1 to perform CAD processing to automatically detect characteristic areas (characteristic areas) in the ultrasound image, and transmits the received execution instructions to the processing circuit 180 of the device body 100. The operator can also set the ROI (Region Of Interest), which is the search range for characteristic areas, in the ultrasound image via the input device 102.

[0021] The display 103 may, for example, display a GUI (Graphical User Interface) for the operator of the ultrasound diagnostic device 1 to input various setting requests using the input device 102, or display ultrasound images shown by ultrasound image data generated in the device body 100. The display 103 is implemented using an LCD monitor, a CRT (Cathode Ray Tube) monitor, or the like. The display 103 is an example of a display unit.

[0022] The speed detector 104 is attached to the ultrasonic probe 101 and detects the movement speed (scan speed) of the ultrasonic probe 101. The speed detector 104 then transmits a detection signal indicating the detected movement speed to the processing circuit 180 of the main unit 100. For example, the speed detector 104 can be implemented using a magnetic sensor. However, the speed detector 104 is not limited to a magnetic sensor and may be implemented using a known device other than a magnetic sensor that can detect the speed of the ultrasonic probe 101.

[0023] The device body 100 generates ultrasonic image data based on the reflected wave signal transmitted from the ultrasonic probe 101. The device body 100 can generate two-dimensional ultrasonic image data based on the reflected wave signal corresponding to the two-dimensional region of the subject P transmitted by the ultrasonic probe 101. Furthermore, the device body 100 can generate three-dimensional ultrasonic image data based on the reflected wave signal corresponding to the three-dimensional region of the subject P transmitted by the ultrasonic probe 101.

[0024] As shown in Figure 1, the main body of the device 100 includes a transmitting circuit 110, a receiving circuit 120, a B-mode processing circuit 130, a Doppler processing circuit 140, an image generation circuit 150, an image memory 160, a storage circuit 170, and a processing circuit 180. The transmitting circuit 110, the receiving circuit 120, the B-mode processing circuit 130, the Doppler processing circuit 140, the image generation circuit 150, the image memory 160, the storage circuit 170, and the processing circuit 180 are connected to each other in a way that allows them to communicate with one another.

[0025] The transmitting circuit 110, under the control of the processing circuit 180, causes the ultrasonic probe 101 to transmit ultrasound. The transmitting circuit 110 includes a rate pulser generation circuit, a transmission delay circuit, and a transmitting pulser, and supplies a drive signal to the ultrasonic probe 101. When scanning a two-dimensional region within the subject P, the transmitting circuit 110 causes the ultrasonic probe 101 to transmit an ultrasonic beam for scanning the two-dimensional region. When scanning a three-dimensional region within the subject P, the transmitting circuit 110 causes the ultrasonic probe 101 to transmit an ultrasonic beam for scanning the three-dimensional region.

[0026] The rate pulser generation circuit repeatedly generates rate pulses at a predetermined rate frequency (PRF: Pulse Repetition Frequency) to form a transmitted ultrasonic wave (transmitted beam). The rate pulses pass through a transmit delay circuit, applying a voltage to the transmit pulser with different transmit delay times. For example, the transmit delay circuit provides a transmit delay time for each piezoelectric transducer necessary to focus the ultrasonic waves generated from the ultrasonic probe 101 into a beam and determine the transmit directivity, to each rate pulse generated by the rate pulser generation circuit. The transmit pulser applies a drive signal (drive pulse) to the ultrasonic probe 101 at a timing based on these rate pulses. The transmit delay circuit can also arbitrarily adjust the transmission direction of the ultrasonic waves from the piezoelectric transducer surface by changing the transmit delay time applied to each rate pulse.

[0027] The drive pulse is transmitted from the transmitting pulser through a cable to the piezoelectric transducer in the ultrasonic probe 101, where it is converted from an electrical signal to a mechanical vibration. The ultrasound generated by this mechanical vibration is transmitted into the body of the subject P. Here, the ultrasound, which has a different transmission delay time for each piezoelectric transducer, is focused and propagates in a predetermined direction.

[0028] Furthermore, the transmitting circuit 110, under the control of the processing circuit 180, has the function of instantaneously changing the transmitting frequency, transmitting drive voltage, etc., in order to execute a predetermined scan sequence. In particular, the change in the transmitting drive voltage is achieved by a linear amplifier type oscillator circuit that can switch its value instantaneously, or by a mechanism that electrically switches multiple power supply units.

[0029] The ultrasonic waves transmitted by the ultrasonic probe 101 reach the piezoelectric transducer inside the ultrasonic probe 101, where they are converted from mechanical vibrations into electrical signals (reflected wave signals). The receiving circuit 120 then receives the reflected wave signals transmitted from the ultrasonic probe 101. Under the control of the processing circuit 180, the receiving circuit 120 performs various processes on the reflected wave signals to generate reflected wave data, and outputs the generated reflected wave data to the B-mode processing circuit 130 and the Doppler processing circuit 140. For example, each time the receiving circuit 120 receives a frame of reflected wave signal, it generates one frame of reflected wave data from the received reflected wave signal. The receiving circuit 120 generates two-dimensional reflected wave data from the two-dimensional reflected wave signals transmitted from the ultrasonic probe 101. The receiving circuit 120 also generates three-dimensional reflected wave data from the three-dimensional reflected wave signals transmitted from the ultrasonic probe 101.

[0030] The receiving circuit 120 includes a preamplifier, an A / D (Analog to Digital) converter, and a quadrature detection circuit. The preamplifier amplifies the reflected wave signal for each channel and performs gain adjustment (gain correction). The A / D converter converts the gain-corrected reflected wave signal into a digital signal by A / D conversion. The quadrature detection circuit converts the digital signal into a baseband in-phase signal (I signal, I) and a quadrature-phase signal (Q signal, Q). The quadrature detection circuit then outputs the I signal and Q signal (IQ signal) as reflected wave data to the B-mode processing circuit 130 and the Doppler processing circuit 140.

[0031] The B-mode processing circuit 130, under the control of the processing circuit 180, performs logarithmic amplification, envelope detection, and logarithmic compression on the reflected wave data output from the receiving circuit 120 to generate data (B-mode data) in which the signal strength (amplitude strength) of each sample point is represented by brightness. For example, each time the B-mode processing circuit 130 receives one frame's worth of reflected wave data, it generates one frame's worth of B-mode data from the received reflected wave data. The B-mode processing circuit 130 outputs the generated B-mode data to the image generation circuit 150. The B-mode processing circuit 130 can be implemented, for example, by a processor.

[0032] The Doppler processing circuit 140, under the control of the processing circuit 180, performs frequency analysis on the reflected wave data output from the receiving circuit 120 to extract motion information of moving objects (blood flow, tissue, contrast agent echo components, etc.) based on the Doppler effect, and generates data (Doppler data) that shows the extracted motion information. For example, the Doppler processing circuit 140 extracts average velocity, variance, power, etc., from multiple points as motion information of a moving object, and generates Doppler data that shows the extracted motion information of the moving object. For example, each time the Doppler processing circuit 140 receives one frame of reflected wave data, it generates one frame of Doppler data from the received reflected wave data. The Doppler processing circuit 140 outputs the generated Doppler data to the image generation circuit 150. The Doppler processing circuit 140 is implemented, for example, by a processor.

[0033] Furthermore, using the functions of the Doppler processing circuit 140 described above, the ultrasound diagnostic device 1 according to this embodiment can perform the color Doppler method, also known as color flow mapping (CFM). In the CFM method, ultrasound is transmitted and received multiple times on multiple scan lines. In the CFM method, an MTI (Moving Target Indicator) filter is applied to the data sequence at the same location to suppress signals (clutter signals) originating from stationary or slow-moving tissue, and to extract signals originating from blood flow. The CFM method then estimates blood flow information such as blood flow velocity, blood flow dispersion, and blood flow power from this blood flow signal. The image generation circuit 150, described later, generates ultrasound image data (color Doppler image data) that displays the distribution of the estimation results in color, for example, in two dimensions. The display 103 then displays the color Doppler image shown by the color Doppler image data.

[0034] The B-mode processing circuit 130 and the Doppler processing circuit 140 are capable of processing both two-dimensional and three-dimensional reflected wave data.

[0035] The image generation circuit 150, under the control of the processing circuit 180, generates ultrasonic image data from various data (B-mode data and Doppler data) output by the B-mode processing circuit 130 and the Doppler processing circuit 140. For example, each time the image generation circuit 150 receives various data for one frame output from the B-mode processing circuit 130 and the Doppler processing circuit 140, it generates ultrasonic image data for one frame from the received data. In other words, the image generation circuit 150 generates multiple ultrasonic image data in real time in a time-series manner at a predetermined frame rate in response to the ultrasonic scan. In this way, in this embodiment, the ultrasonic probe 101, the receiving circuit 120, the B-mode processing circuit 130, the Doppler processing circuit 140, and the image generation circuit 150 collect multiple ultrasonic image data in real time in a time-series manner at a predetermined frame rate. The ultrasonic image data is data based on reflected wave signals obtained by ultrasonic scanning performed by the ultrasonic probe 101. The ultrasound probe 101, receiving circuit 120, B-mode processing circuit 130, Doppler processing circuit 140, and image generation circuit 150 are examples of an acquisition unit. The ultrasound image data is an example of medical image data.

[0036] The image generation circuit 150 then stores multiple ultrasound image data, generated in real time and aligned with the time series, in the image memory 160. To explain with a specific example, each time the image generation circuit 150 generates ultrasound image data for one frame, it stores the generated ultrasound image data in the image memory 160. For example, among the multiple ultrasound image data obtained in a time series by ultrasound scanning, the first ultrasound image data generated is referred to as the ultrasound image data for the first frame. Similarly, the Nth (where N is an integer greater than or equal to 1)th ultrasound image data generated is referred to as the ultrasound image data for the Nth frame.

[0037] The image generation circuit 150 is implemented by a processor. Here, the image generation circuit 150 converts the scan line signal sequence of the ultrasonic scan into a scan line signal sequence of a video format, such as that used in televisions (scan conversion), and generates ultrasonic image data for display. For example, the image generation circuit 150 generates ultrasonic image data for display by performing coordinate transformation according to the ultrasonic scanning pattern of the ultrasonic probe 101. In addition to scan conversion, the image generation circuit 150 also performs various image processing, such as image processing that regenerates an average brightness image using multiple image frames after scan conversion (smoothing processing), and image processing that uses a differential filter within the image (edge ​​enhancement processing). Furthermore, the image generation circuit 150 synthesizes various parameter text information, scales, body marks, etc., with the ultrasonic image data.

[0038] Furthermore, the image generation circuit 150 generates 3D B-mode image data by performing a coordinate transformation on the 3D B-mode data generated by the B-mode processing circuit 130. The image generation circuit 150 also generates 3D Doppler image data by performing a coordinate transformation on the 3D Doppler data generated by the Doppler processing circuit 140. In other words, the image generation circuit 150 generates "3D B-mode image data and 3D Doppler image data" as "3D ultrasonic image data (volume data)". Then, the image generation circuit 150 performs various rendering processes on the volume data to generate various 2D image data for displaying the volume data on the display 103.

[0039] B-mode data and Doppler data are ultrasound image data before scan conversion processing, while the data generated by the image generation circuit 150 is ultrasound image data for display after scan conversion processing. B-mode data and Doppler data are also referred to as raw data.

[0040] The image memory 160 is a memory that stores various image data generated by the image generation circuit 150. The image memory 160 also stores data generated by the B-mode processing circuit 130 and the Doppler processing circuit 140. The B-mode data and Doppler data stored in the image memory 160 can be retrieved by the operator after a diagnosis, for example, and become ultrasound image data for display via the image generation circuit 150. For example, the image memory 160 can be implemented using semiconductor memory elements such as RAM or flash memory, a hard disk, or an optical disk.

[0041] The memory circuit 170 stores control programs for ultrasonic transmission and reception, image processing, and display processing, diagnostic information (e.g., patient ID, physician's findings, etc.), and various data such as diagnostic protocols and various body marks. The memory circuit 170 is also used, if necessary, to store data stored in the image memory 160. For example, the memory circuit 170 can be implemented using semiconductor memory elements such as flash memory, a hard disk, or an optical disk. The memory circuit 170 is an example of a memory unit.

[0042] The processing circuit 180 controls the entire processing of the ultrasound diagnostic device 1. The processing circuit 180 is implemented, for example, by a processor. The processing circuit 180 has the following processing functions: control function 181, speed detection function 182, acquisition function 183, discrimination function 184, marker information generation function 185, and display control function 186.

[0043] Here, for example, each processing function of the processing circuit 180 is stored in the memory circuit 170 in the form of a program that can be executed by a computer. The processing functions of the processing circuit 180 shown in Figure 1, namely the control function 181, speed detection function 182, acquisition function 183, discrimination function 184, marker information generation function 185, and display control function 186, are stored in the memory circuit 170 in the form of a program that can be executed by a computer. The processing circuit 180 reads each program from the memory circuit 170 and executes each read program to realize the function corresponding to each program. In other words, the processing circuit 180 in the state in which each program has been read has the functions shown in the processing circuit 180 of Figure 1.

[0044] In the above description, the term "processor" refers to circuits such as a CPU (Central Processing Unit), GPU (Graphics Processing Unit), Application Specific Integrated Circuit (ASIC), or programmable logic device (e.g., Simple Programmable Logic Device (SPLD), Complex Programmable Logic Device (CPLD), or Field Programmable Gate Array (FPGA)). The processor implements its functions by, for example, reading and executing a program stored in the memory circuit 170. Alternatively, instead of storing the program in the memory circuit 170, the program may be directly incorporated into the processor's circuit. In this case, the processor implements its functions by reading and executing the program incorporated into the circuit. In this embodiment, each processor is not limited to being configured as a single circuit; multiple independent circuits may be combined to form a single processor and implement its functions. Furthermore, multiple components shown in Figure 1 may be integrated into a single processor to implement its functions. The same applies to the term "processor" used in the following explanation.

[0045] The control function 181 controls the processing of the transmission circuit 110, reception circuit 120, B-mode processing circuit 130, Doppler processing circuit 140, and image generation circuit 150 based on various setting requests input from the operator via the input device 102 and various data read from the memory circuit 170.

[0046] The speed detection function 182 detects the movement speed of the ultrasonic probe 101 by calculating the movement speed of the ultrasonic probe 101 indicated by the detection signal transmitted from the speed detector 104. For example, the speed detection function 182 detects the movement speed of the ultrasonic probe 101 at predetermined time intervals. The speed detector 104 and the speed detection function 182 are examples of detection units.

[0047] The acquisition function 183 acquires the first trained model 11a, the second trained model 12a, and the third trained model 13a transmitted from the trained model generation device 200. The acquisition function 183 then stores the acquired first trained model 11a, the second trained model 12a, and the third trained model 13a in the memory circuit 170. The acquisition function 183 also acquires the first learning conditions 11b, the second learning conditions 12b, and the third learning conditions 13b transmitted from the trained model generation device 200. The acquisition function 183 then stores the acquired first learning conditions 11b, the second learning conditions 12b, and the third learning conditions 13b in the memory circuit 170. The various trained models and learning conditions stored in the memory circuit 170 are used by the discrimination function 184 when it performs various discrimination processes.

[0048] Furthermore, the acquisition function 183 acquires multiple ultrasound image data in chronological order when it receives an instruction (execution instruction) to execute CAD processing input from the operator via the input device 102. For example, after receiving an execution instruction, the acquisition function 183 acquires the newly generated ultrasound image data for one frame from the image memory 160 each time that newly generated ultrasound image data for one frame is stored in the image memory 160. For example, multiple ultrasound image data in chronological order are a group of data obtained by ultrasound scanning. Such ultrasound scanning is performed, for example, while the scanning area inside the subject P is moved by the movement of the ultrasound probe 101 which is moved along the surface of the subject P by the operator, or while the ultrasound probe 101 is stationary. Also, in the following description, each of the multiple ultrasound image data in chronological order is, for example, B-mode image data.

[0049] The discrimination function 184 performs operational processing to derive discrimination results regarding feature regions using various trained models and various training conditions stored in the memory circuit 170. For example, each time a newly generated frame of ultrasound image data is acquired by the acquisition function 183, the discrimination function 184 performs operational processing using the acquired frame of ultrasound image data.

[0050] Figure 3 is a diagram illustrating an example of processing performed by the discrimination function 184 according to the first embodiment. As illustrated in Figure 3, in the first stage of operation, the discrimination function 184 uses the first trained model 11a and the first training conditions 11b to perform a first discrimination process as CAD processing, which determines whether or not a feature area is included in the ultrasound image and derives a discrimination result (first discrimination result). That is, if the discrimination function 184 determines that a feature area is included in the ultrasound image, it means that the feature area has been detected. Conversely, if the discrimination function 184 determines that a feature area is not included in the ultrasound image, it means that the feature area has not been detected.

[0051] Furthermore, in the second stage of operation, the discrimination function 184 uses the second trained model 12a and the second training condition 12b to determine whether the feature region detected in the first stage of operation is benign or malignant, and performs a second discrimination process to derive a discrimination result (second discrimination result). Furthermore, in the third stage of operation, the discrimination function 184 uses the third trained model 13a and the third training condition 13b to determine the degree of malignancy of the feature region determined to be malignant in the second stage of operation, and performs a third discrimination process to derive a discrimination result (third discrimination result). For example, there are two degrees of malignancy: malignancy degree A and malignancy degree B, which is more malignant than malignancy degree A. In this case, the discrimination function 184 determines whether the degree of malignancy of the feature region determined to be malignant is malignancy degree A or malignancy degree B. The following describes a case where the discrimination function 184 distinguishes between two malignancy levels, malignancy level A and malignancy level B. However, the malignancy level is not limited to just two levels, malignancy level A and malignancy level B. The discrimination function 184 is an example of a discrimination unit.

[0052] Figure 4 is a diagram illustrating an example of the first stage of processing during learning and the first stage of processing during operation according to the first embodiment. As illustrated in Figure 4, in the first stage of learning, the trained model generation function 201a generates a first trained model 11a by learning the relationship between ultrasound image data 14 and a first label 15, and records the first learning conditions 11b. Here, the first label 15 is a label that indicates whether or not a feature area is included in the ultrasound image shown by the ultrasound image data 14. That is, the trained model generation function 201a generates a first trained model 11a by learning the association between ultrasound image data 14 and the first label 15, and records the first learning conditions 11b. The ultrasound image data 14 is, for example, B-mode image data generated from B-mode data. The ultrasound image data 14 is data generated by the ultrasound diagnostic device 1 or an ultrasound diagnostic device other than the ultrasound diagnostic device 1.

[0053] Furthermore, the ultrasound image data 14 may be B-mode image data that has undergone preprocessing such as edge enhancement and contrast enhancement. Also, the ultrasound image data 14 may be, for example, B-mode data (raw B-mode data). Also, the ultrasound image data 14 may be color Doppler image data. Also, the ultrasound image data 14 may be image data showing the FLR (Fat Lesion Ratio) obtained by strain elastography. Also, the ultrasound image data 14 may be shear wave speed image data obtained by SWE (Shear Wave Elastography), in which pixel values ​​corresponding to the shear rate at each of the multiple sample points in the scanning area are assigned to each sample point. Also, the ultrasound image data 14 may be B-mode image data obtained while strain elastography or SWE is being performed by ultrasound diagnostic device 1 or an ultrasound diagnostic device other than ultrasound diagnostic device 1.

[0054] Furthermore, the first label 15 is a label obtained, for example, by an operator determining whether or not a characteristic area is included in the ultrasound image shown by the ultrasound image data 14. The first label 15 may also be a label obtained by applying CAD processing to the ultrasound image data 14 using the ultrasound diagnostic device 1 or another device.

[0055] For example, the trained model generation function 201a performs machine learning by inputting a pair of ultrasound image data 14 and a first label 15 as training data (teaching data) into the machine learning engine.

[0056] For example, machine learning engines perform machine learning using various algorithms such as deep learning, neural networks, logistic regression analysis, nonlinear discriminant analysis, support vector machines (SVMs), random forests, and naive Bayes.

[0057] As a result of this machine learning, the trained model generation function 201a generates a first trained model 11a, which is used in a first discrimination process to derive a first discrimination result from the input ultrasound image data. Also, as a result of the machine learning described above, the trained model generation function 201a records the first learning conditions 11b, which are used in the calculations in the first discrimination process. The trained model generation function 201a then transmits the generated first trained model 11a and the recorded first learning conditions 11b to the ultrasound diagnostic device 1. As a result, the first trained model 11a and the first learning conditions 11b are stored in the memory circuit 170 of the ultrasound diagnostic device 1.

[0058] Note that the trained model generation device 200 and the ultrasound diagnostic device 1 do not need to be connected via a network. In this case, the various trained models and learning conditions generated by the trained model generation device 200 are stored in a storage medium such as a USB (Universal Serial Bus) memory. The ultrasound diagnostic device 1 then reads the various trained models and learning conditions stored in the storage medium and stores the read trained models and learning conditions in the memory circuit 170.

[0059] Next, an example of the first stage of processing during operation will be described. As illustrated in Figure 4, in the first stage of operation, the discrimination function 184 uses the first trained model 11a and the first learning conditions 11b to determine whether or not the ultrasonic image shown in the ultrasonic image data 16 contains a feature area, and executes a first discrimination process as a CAD process to derive a discrimination result (first discrimination result) 17. For example, the type of ultrasonic image data 16 is the same as the type of ultrasonic image data 14. For example, B-mode image data that has undergone the above-mentioned preprocessing, B-mode image data that has not undergone the above-mentioned preprocessing, B-mode data, color Doppler image data, image data showing FLR obtained by strain elastography, shear rate image data, and B-mode image data obtained while performing strain elastography or SWE are listed as examples of each type of ultrasonic image data.

[0060] For example, if the discrimination function 184 determines that the ultrasound image contains a characteristic area, it outputs information indicating that the ultrasound image contains a characteristic area as the first discrimination result 17. Also, if the discrimination function 184 determines that the ultrasound image does not contain a characteristic area, it outputs information indicating that the ultrasound image does not contain a characteristic area as the first discrimination result 17.

[0061] Figure 5 is a diagram illustrating an example of the second stage of processing during learning and the second stage of processing during operation according to the first embodiment. As illustrated in Figure 5, in the second stage of learning, the trained model generation function 201a generates a second trained model 12a by learning the relationship between ultrasound image data 20 and a second label 21, and records the second learning conditions 12b. Here, the second label 21 is a label that indicates whether the feature area included in the ultrasound image shown by the ultrasound image data 20 is malignant or benign. That is, the trained model generation function 201a generates a second trained model 12a by learning the association between ultrasound image data 20 and the second label 21, and records the second learning conditions 12b.

[0062] The ultrasound image data 20 used to generate the second trained model 12a and record the second training conditions 12b is image data in which feature regions are depicted. The ultrasound image data 20 is, for example, B-mode image data generated from B-mode data. The ultrasound image data 20 is data generated by ultrasound diagnostic device 1 or ultrasound diagnostic device other than ultrasound diagnostic device 1.

[0063] For example, the trained model generation function 201a performs machine learning by inputting a pair of ultrasound image data 20 and a second label 21 as training data into the machine learning engine.

[0064] As a result of this machine learning, the trained model generation function 201a generates a second trained model 12a, which is used in a second discrimination process to derive a second discrimination result from the input of ultrasound image data depicting feature regions. Also, as a result of the machine learning described above, the trained model generation function 201a records a second learning condition 12b, which is used in the calculation in the second discrimination process. The trained model generation function 201a then transmits the generated second trained model 12a and the recorded second learning condition 12b to the ultrasound diagnostic device 1. As a result, the second trained model 12a and the second learning condition 12b are stored in the memory circuit 170 of the ultrasound diagnostic device 1.

[0065] Next, an example of the second stage of processing during operation will be described. As illustrated in Figure 5, in the second stage of operation, the discrimination function 184 uses the second trained model 12a and the second training conditions 12b to determine whether the feature region depicted in the ultrasound image data 22 is benign or malignant, and performs a second discrimination process to derive a discrimination result (second discrimination result) 23.

[0066] For example, if the discrimination function 184 determines that the feature area depicted in the ultrasound image data 22 is benign, it outputs information indicating that the feature area is benign as a second discrimination result 23. Also, if the discrimination function 184 determines that the feature area depicted in the ultrasound image data 22 is malignant, it outputs information indicating that the feature area is malignant as a second discrimination result 23.

[0067] Here, the ultrasound image data 22 used in deriving the second discrimination result 23 is the ultrasound image data that was determined to contain a feature area in the first discrimination process, which is the first stage of operation. In other words, the ultrasound image data 22 used in the second discrimination process is the ultrasound image data that was determined to contain a feature area in the first discrimination process. For example, the type of ultrasound image data 22 is the same as the type of ultrasound image data 20.

[0068] Figure 6 is a diagram illustrating an example of the third stage of processing during learning and the third stage of processing during operation according to the first embodiment. As illustrated in Figure 6, in the third stage of learning, the trained model generation function 201a generates a third trained model 13a by learning the relationship between ultrasound image data 26 and a third label 27, and records the third learning conditions 13b. Here, the third label 27 is a label that indicates whether the malignancy of the malignant characteristic site included in the ultrasound image shown by the ultrasound image data 26 is malignancy grade A or malignancy grade B. That is, the trained model generation function 201a generates a third trained model 13a by learning the correspondence between ultrasound image data 26 and the third label 27, and records the third learning conditions 13b.

[0069] The ultrasound image data 26 used to generate the third trained model 13a and record the third training condition 13b is image data depicting malignant characteristic sites. The ultrasound image data 26 is, for example, B-mode image data generated from B-mode data. The ultrasound image data 26 is data generated by ultrasound diagnostic device 1 or an ultrasound diagnostic device other than ultrasound diagnostic device 1.

[0070] For example, the trained model generation function 201a performs machine learning by inputting a pair of ultrasound image data 26 and a third label 27 as training data into the machine learning engine.

[0071] As a result of this machine learning, the trained model generation function 201a generates a third trained model 13a, which is used in a third discrimination process to derive a third discrimination result from the input of ultrasound image data depicting malignant characteristic sites. Also, as a result of the machine learning described above, the trained model generation function 201a records a third learning condition 13b, which is used in the calculations in the third discrimination process. The trained model generation function 201a then transmits the generated third trained model 13a and the recorded third learning condition 13b to the ultrasound diagnostic device 1. As a result, the third trained model 13a and the third learning condition 13b are stored in the memory circuit 170 of the ultrasound diagnostic device 1.

[0072] Next, an example of the third stage of processing during operation will be described. As illustrated in Figure 6, in the third stage of operation, the discrimination function 184 uses the third trained model 13a and the third training conditions 13b to determine, in response to the input ultrasound image data 28, whether the malignancy of the malignant characteristic site depicted in the ultrasound image data 28 is malignancy grade A or malignancy grade B, and performs a third discrimination process to derive a discrimination result (third discrimination result) 29.

[0073] For example, if the discrimination function 184 determines that the malignancy of the malignant characteristic site depicted in the ultrasound image data 28 is malignancy grade A, it outputs information indicating that the malignancy of the malignant characteristic site is malignancy grade A as a third discrimination result 29. Also, if the discrimination function 184 determines that the malignancy of the malignant characteristic site depicted in the ultrasound image data 28 is malignancy grade B, it outputs information indicating that the malignancy of the malignant characteristic site is malignancy grade B as a third discrimination result 29.

[0074] Here, the ultrasound image data 28 used to derive the third discrimination result 29 is the ultrasound image data depicting the feature area that was determined to be malignant in the second discrimination process of the second stage during operation. In other words, the ultrasound image data 28 used in the third discrimination process is the ultrasound image data depicting the feature area that was determined to be malignant in the second discrimination process. For example, the type of ultrasound image data 28 is the same as the type of ultrasound image data 26.

[0075] The above describes examples of processing during training and operation with reference to Figures 2-6. As described above, the trained model generation function 201a generates various trained models and records various training conditions by learning to associate ultrasound image data with labels related to feature regions depicted in the ultrasound image data. Various discrimination processes using the various trained models and various training conditions output discrimination results regarding feature regions. The trained model generation function 201a generates such trained models at each of the multiple stages of training (stages 1 to 3 of training described above). The trained model generation function 201a also records these training conditions at each of the multiple stages of training.

[0076] The pre-trained model generation function 201a generates a pre-trained model specialized for a specific discrimination function at each stage of training and records the training conditions. Therefore, the discrimination process based on the pre-trained model generated at each stage of training and the recorded training conditions is a process that can accurately perform a specific discrimination. For example, in the first stage of training, the pre-trained model generation function 201a generates a first pre-trained model 11a specialized for the function of determining whether or not a feature area is included in an ultrasound image, and records the first training conditions 11b. Therefore, the first discrimination process based on the first pre-trained model 11a and the first training conditions 11b can accurately determine whether or not a feature area is included in an ultrasound image.

[0077] Furthermore, in the second stage of training, the pre-trained model generation function 201a generates a second pre-trained model 12a specialized in determining whether a feature region is malignant or benign, and records the second training conditions 12b. Therefore, the second discrimination process based on the second pre-trained model 12a and the second training conditions 12b can accurately determine whether a feature region is malignant or benign.

[0078] Furthermore, in the third stage of training, the trained model generation function 187 generates a third trained model 13a specialized in determining whether the malignancy of a malignant feature region is malignancy level A or malignancy level B, and records the third training conditions 13b. Therefore, the third discrimination process based on the third trained model 13a and the third training conditions 13b can accurately determine whether the malignancy of a malignant feature region is malignancy level A or malignancy level B.

[0079] In other words, the discrimination function 184 performs a unique discrimination process at each of the multiple stages of operation, which is different from the discrimination process performed at other stages. Therefore, with the ultrasound diagnostic device 1 having the first trained model 11a, the second trained model 12a, and the third trained model 13a, as well as the first learning condition 11b, the second learning condition 12b, and the third learning condition 13b, various discrimination results regarding feature areas can be obtained with high accuracy.

[0080] Furthermore, as described above, the pre-trained model generation function 201a generates a second pre-trained model 12a using ultrasound image data in which feature regions are depicted, rather than using ultrasound image data in which feature regions that are the target of determination as to whether they are malignant or benign are not depicted. Therefore, the pre-trained model generation function 201a can accelerate the convergence of machine learning when generating the second pre-trained model 12a.

[0081] Furthermore, the pre-trained model generation function 201a generates a third pre-trained model 13a using ultrasound image data depicting malignant feature regions, which are the target of discrimination, rather than using ultrasound image data depicting benign feature regions that are not the target of discrimination as malignancy grade A or malignancy grade B. Therefore, the pre-trained model generation function 201a can accelerate the convergence of machine learning when generating the third pre-trained model 13a.

[0082] Furthermore, as described above, the discrimination function 184 derives various discrimination results regarding feature regions at each of the multiple stages of operation (stages 1 to 3 of operation described above) using various trained models and various training conditions. For example, in stage 1 of operation, the discrimination function 184 derives a first discrimination result 17 regarding the feature region depicted in the input ultrasound image data 16. In stage 2 of operation, the discrimination function 184 derives a second discrimination result 23 regarding the feature region depicted in the input ultrasound image data 22. In stage 3 of operation, the discrimination function 184 derives a third discrimination result 29 regarding the feature region depicted in the input ultrasound image data 28.

[0083] Furthermore, the discrimination function 184 inputs the ultrasound image data 22 (see Figure 5) corresponding to a predetermined discrimination result derived by the first discrimination process to the second discrimination process. Here, the predetermined discrimination result derived by the first discrimination process refers to the first discrimination result 17 (see Figure 4) that indicates that a characteristic area is included in the ultrasound image.

[0084] In other words, the discrimination function 184 does not input ultrasound image data 22 in which the characteristic area to be determined as malignant or benign is depicted to the second discrimination process. Then, the discrimination function 184 derives the second discrimination result 23 derived by the second discrimination process as the discrimination result corresponding to the second stage in operation.

[0085] Furthermore, the discrimination function 184 inputs the ultrasound image data 28 (see Figure 6) corresponding to a predetermined discrimination result derived by the second discrimination process to the third discrimination process. Here, the predetermined discrimination result derived by the second discrimination process refers to the second discrimination result 23 (see Figure 5) when the characteristic site indicates malignancy.

[0086] In other words, the discrimination function 184 does not input ultrasound image data in which the malignant characteristic site, which is the target of discrimination as to whether it is malignancy grade A or malignancy grade B, is not depicted, but instead inputs ultrasound image data 28 in which the malignant characteristic site is depicted. Then, the discrimination function 184 derives the third discrimination result 29 derived by the third discrimination process as the discrimination result corresponding to the third stage.

[0087] In this way, the discrimination function 184 inputs ultrasound image data corresponding to a predetermined discrimination result derived by a discrimination process corresponding to one of the multiple stages to a discrimination process corresponding to the next stage, and derives the discrimination result derived by the discrimination process corresponding to the next stage as the discrimination result corresponding to the next stage.

[0088] As described above, the discrimination function 184 derives a first discrimination result 17 (see Figure 4) using a first trained model 11a and a first training condition 11b that can accurately determine whether or not a feature area is included in the ultrasound image. Therefore, the discrimination function 184 can derive a first discrimination result 17 with good discrimination accuracy.

[0089] Furthermore, the discrimination function 184 derives a second discrimination result 23 (see Figure 5) using a second trained model 12a and a second training condition 12b that can accurately determine whether a feature region is malignant or benign. Therefore, the discrimination function 184 can derive a second discrimination result 23 with good discrimination accuracy.

[0090] Furthermore, the discrimination function 184 uses a third pre-trained model 13a and a third training condition 13b, which can accurately determine whether the malignancy of a feature region is malignancy grade A or malignancy grade B, to derive a third discrimination result 29 (see Figure 6). Therefore, the discrimination function 184 can derive a third discrimination result 29 with good discrimination accuracy.

[0091] Based on these findings, the ultrasound diagnostic device 1 according to the first embodiment can accurately obtain discrimination results regarding characteristic areas.

[0092] Next, we will describe an example of a first discrimination process performed by the discrimination function 184 using the first trained model 11a and the first learning conditions 11b in the first stage of operation. Figure 7 is a flowchart showing the flow of an example of the first discrimination process according to the first embodiment.

[0093] As shown in Figure 7, the discrimination function 184 performs a first detection process (step S101) and then performs a second detection process (step S102).

[0094] First, the first detection process performed in step S101 will be described. The discrimination function 184 uses a search window, a first pre-trained search model, and a search algorithm to search for feature regions within a search area (ROI) set by the operator in the ultrasound image. For example, the discrimination function 184 moves the search window to multiple positions within the search area. At each position, the discrimination function 184 analyzes the image information within the search window using the first pre-trained search model and search algorithm to calculate the probability (first probability) that the image information within each search window corresponds to a feature region.

[0095] The search box is, for example, a unit of region within the search range that is compared with the first pre-trained search model. The first pre-trained search model is, for example, a model that has learned image information that serves as a sample of a feature region. The search algorithm is an algorithm for calculating the first probability described above and searching for image information corresponding to the feature region in the ultrasound image to be processed.

[0096] The first pre-trained search model is an example of the first pattern that represents feature regions.

[0097] The discrimination function 184 then detects the area enclosed by the search window as a feature area (range of the feature area) if the calculated first probability is greater than or equal to a predetermined threshold. That is, the discrimination function 184 determines that the ultrasound image contains a feature area if the first probability, which indicates the degree of similarity between the features of a part of the ultrasound image and the features of the first trained search model that represents the feature area, is greater than or equal to a threshold. The discrimination function 184 also determines that the ultrasound image does not contain a feature area if the first probability is less than the threshold. The discrimination function 184 then derives a first discrimination result, which is the result of determining whether or not the ultrasound image contains a feature area. The first detection process is an example of a feature area detection process.

[0098] The discrimination function 184 then stores the search result (detection result) for each frame of ultrasound image data in the memory circuit 170. For example, if the discrimination function 184 is able to detect a feature area from a frame of ultrasound image data, it stores position information indicating the location of the feature area as a search result in the memory circuit 170. The position of the feature area referred to here is the position in the image space of the ultrasound image data.

[0099] Here, referring to Figures 8 and 9, it will be explained that the feature regions detected by the discrimination function 184 include two types: correctly detected feature regions and feature regions that are not actually feature regions but were mistakenly detected. Figures 8 and 9 are diagrams illustrating an example of the feature region detection result by the discrimination function 184 according to the first embodiment.

[0100] Figure 8 shows a case where the discrimination function 184 correctly detects the feature. As illustrated in Figure 8, for example, the discrimination function 184 performs a first detection process on the ultrasound image and detects the area enclosed by the search window 30, which includes the actual feature area 31, as a feature area. The discrimination function 184 then stores information indicating the location of the detected feature area in the ultrasound image in the memory circuit 170.

[0101] On the other hand, Figure 9 shows a case where a feature area is incorrectly detected by the discrimination function 184. As illustrated in Figure 9, for example, the discrimination function 184 may perform a first detection process on the ultrasound image and incorrectly detect as a feature area an area enclosed by the search window 30 that does not include the actual feature area 31. This incorrect detection of a feature area is called overdetection (false positive). In such cases, the discrimination function 184 also stores information indicating the location of the detected feature area in the ultrasound image in the memory circuit 170.

[0102] Furthermore, if the discrimination function 184 is unable to detect a feature area from a single frame of ultrasound image, it stores information indicating that the feature area could not be detected as a search result in the memory circuit 170.

[0103] The discrimination function 184 has been described in an example of calculating the first probability as an indicator showing the degree of similarity between the features of the image information in the search window 30 and the features of the first pre-trained search model. However, instead of the first probability, the discrimination function 184 may calculate other indicators such as similarity, which shows the degree of similarity between the features of the image information in the search window 30 and the features of the first pre-trained search model.

[0104] The above describes an example of the first detection process performed in step S101. Figure 10 is a diagram showing an example of the result of the first detection process according to the first embodiment. The "Result determined to be a feature area" shown in Figure 10 shows an example of image information within the search window 30 at the location of the search window 30 that was determined to be a feature area by the first detection process, from a single frame of ultrasound image. The "Result determined not to be a feature area" shown in Figure 10 shows an example of image information within the search window 30 when no feature area was detected by the first detection process, from a single frame of ultrasound image. In addition, Figure 10 shows 10 pieces of image information as an example of the result of calculating a first probability at each location where the search window 30 is placed at multiple locations within the search range in the first detection process.

[0105] As shown in Figure 10, the discrimination function 184 detects feature regions at six locations in the first detection process. Specifically, as shown in Figure 10, the discrimination function 184 determines that the six image information 40a to 40f within the search window 30 are feature regions.

[0106] Image information 40a is image information depicting the actual feature region 31a. Image information 40b is image information that was over-detected. Image information 40c is image information depicting the actual feature region 31b. Image information 40d is image information depicting the actual feature region 31c. Image information 40e is image information depicting the actual feature region 31d. Image information 40f is image information that was over-detected. In other words, in the case shown in Figure 10, two of the six image pieces determined to be feature regions were over-detected.

[0107] Furthermore, as shown in Figure 10, the discrimination function 184 failed to detect feature regions at four locations during the first detection process. Specifically, as shown in Figure 10, the discrimination function 184 determined that the four image information points 40g to 40j within the search window 30 were not feature regions.

[0108] Next, the second detection process performed in step S102 will be described. The discrimination function 184 uses a trained model for over-detection discrimination and a search algorithm to calculate the probability (second probability) that the feature region detected in the first detection process (image information in the search window 30) corresponds to image information that is estimated to be over-detected in the first detection process.

[0109] The pre-trained model for overdetection discrimination is, for example, a model that has learned to distinguish between image information that is estimated to be overdetected in the first detection process and image information of a feature region. The search algorithm is an algorithm for calculating the second probability described above and searching for image information in the ultrasound image to be processed that corresponds to the image information that is estimated to be overdetected in the first detection process.

[0110] The pre-trained model for false positive detection is a model trained to extract image information that is presumed to be incorrectly identified as containing feature regions in the ultrasound image, even though the feature regions are not actually present, during the first detection process. The pre-trained model for false positive detection is an example of the second pattern.

[0111] Here, we will explain the case where six image information 40a to 40f are detected in the first detection process, as illustrated in Figure 10 above. In this case, the discrimination function 184 calculates a second probability in the second detection process for each of the six image information 40a to 40f, which is estimated to be the image information that was over-detected in the first detection process. In this way, the discrimination function 184 calculates a second probability for each of the six image information 40a to 40f.

[0112] The discrimination function 184 then detects, among the six image information pieces 40a to 40f, the image information whose calculated second probability is above a predetermined threshold as over-detected image information in the first detection process. The discrimination function 184 then stores the search result (detection result) for each frame of ultrasound image data in the memory circuit 170. For example, if over-detected image information is detected for a frame of ultrasound image, the discrimination function 184 stores location information indicating the location of the detected image information as a search result in the memory circuit 170.

[0113] In the second detection process, the discrimination function 184 may or may not correctly detect the image information that was over-detected in the first detection process. If it cannot be correctly detected, this may include incorrectly detecting image information that was not over-detected in the first detection process as the image information that was over-detected in the first detection process.

[0114] Furthermore, for example, if no over-detected image information is detected for a single frame of ultrasound image, the discrimination function 184 stores location information indicating the location of the undetected image information in the memory circuit 170 as a search result.

[0115] The discrimination function 184 has described the case in which it calculates a second probability as an example of an indicator showing the degree of similarity between the features of the image information and the features of the pre-trained model for overdetection discrimination. However, instead of the second probability, the discrimination function 184 may calculate other indicators such as similarity, which shows the degree of similarity between the features of the image information and the features of the pre-trained model for overdetection discrimination.

[0116] The above describes an example of the second detection process performed in step S102. Figure 11 is a diagram showing an example of the results of the second detection process according to the first embodiment. Figure 11 shows the results of the second detection process being performed on each of the six image information 40a to 40f shown in Figure 10. The "results determined to be false positives" in Figure 11 show an example of image information detected by the second detection process. The "results determined not to be false positives" in Figure 11 show an example of image information that was not detected by the second detection process.

[0117] As shown in Figure 11, in the second detection process, the discrimination function 184 detects two image pieces 40b and 40f as image information that was over-detected in the first detection process. Also, in the second detection process, the discrimination function 184 does not detect four image pieces 40a, 40c, 40d, and 40e. Then, in the second detection process, the discrimination function 184 detects four of the six image pieces 40a to 40f, excluding the two detected image pieces 40b and 40f, as image information that actually contains feature regions. As a result, in the second detection process, four image pieces 40a, 40c, 40d, and 40e are ultimately detected as feature regions.

[0118] As described above, the discrimination function 184 detects, in the second detection process, the feature regions that were over-detected from among the feature regions detected in the first detection process, excluding the over-detected feature regions, as the final feature regions.

[0119] In other words, if the second probability, which indicates the degree of similarity between the features of a feature region determined to be included in the ultrasound image by the first detection process and the features of the trained model for over-detection discrimination, is greater than or equal to a threshold, the discrimination function 184 corrects the determination that the feature region determined to be included in the ultrasound image in the first detection process is not included in the ultrasound image. Therefore, the discrimination function 184 can detect feature regions with high accuracy. The second detection process is an example of a correction process.

[0120] Here, if the operator wants to determine whether or not a feature area exists, rather than its details, it is likely that they will move the ultrasound probe 101 relatively quickly over a relatively wide area. On the other hand, if the operator has already determined the existence of a feature area and wants to understand its details, it is likely that they will move the ultrasound probe 101 relatively slowly over a relatively narrow area, such as the vicinity of the feature area.

[0121] Therefore, the ultrasound diagnostic apparatus 1 according to the first embodiment performs appropriate processing according to the movement speed of the ultrasound probe 101, as described below.

[0122] Figure 12 is a flowchart showing an example of the process performed by the ultrasound diagnostic device 1 according to the first embodiment. For example, the ultrasound diagnostic device 1 performs the process shown in Figure 12 at predetermined time intervals.

[0123] As shown in Figure 12, the discrimination function 184 determines whether the movement speed of the ultrasonic probe 101 is greater than a predetermined threshold Th1 (step S201). Here, if the movement speed of the ultrasonic probe 101 is greater than the predetermined threshold Th1, for example, the operator may want to know whether or not a feature area exists, rather than the details of the feature area. In step S201, the velocity detection function 182 detects the movement speed of the ultrasonic probe 101. Then, in step S201, the discrimination function 184 determines whether or not the movement speed of the ultrasonic probe 101 detected by the velocity detection function 182 is greater than the threshold Th1. Threshold Th1 is an example of a first threshold.

[0124] If it is determined that the movement speed of the ultrasonic probe 101 is greater than a predetermined threshold Th1 (step S201: Yes), then in step S202, the discrimination function 184 performs the processing described below.

[0125] For example, in step S202, the discrimination function 184 selects the first frame rate (e.g., 30 fps) from among the first frame rate (e.g., 30 fps) and the second frame rate (e.g., 15 fps) as the frame rate for deriving the discrimination result. Here, the frame rate refers to, for example, the number of ultrasound image data processed per unit time. The second frame rate is lower than the first frame rate.

[0126] Furthermore, for example, in step S202, the discrimination function 184 selects the first derivation process from among the first derivation process and the second derivation process as the derivation process for deriving the discrimination result. Here, the first derivation process is the first discrimination process of the first stage in operation as shown in Figure 4 above. The second derivation processes are the first discrimination process, the second discrimination process of the second stage in operation as shown in Figure 5 above, and the third discrimination process of the third stage in operation as shown in Figure 6 above.

[0127] The first derivation process has a lower processing load compared to the second derivation process. Therefore, the discrimination function 184 can increase the frame rate when executing the first derivation process. Accordingly, as described above, in step S202, the discrimination function 184 selects a relatively high first frame rate from among several types of frame rates (first frame rate and second frame rate).

[0128] Furthermore, if a positive determination is made in step S201 (step S201: Yes), as mentioned above, the operator is likely to want to know whether or not a feature area exists, rather than the details of the feature area. Therefore, in step S202, the discrimination function 184 selects the first derivation process (first discrimination process), which performs one type of discrimination—whether or not a feature area is included in the ultrasound image—from among several types of derivation processes (first derivation process and second derivation process), rather than the second derivation process, which performs multiple types of discrimination.

[0129] Then, in step S203, the discrimination function 184 executes the selected first derivation process. Also in step S203, the control function 181 controls the transmission circuit 110, reception circuit 120, B-mode processing circuit 130, Doppler processing circuit 140, image generation circuit 150, acquisition function 183, discrimination function 184, marker information generation function 185, and display control function 186 to perform various processes at the selected first frame rate.

[0130] Furthermore, in step S203, the marker information generation function 185 generates marker information indicating the detection result of the ultrasound image based on the first discrimination result 17 (see Figure 4) derived by the first derivation process. For example, in step S203, each time the first discrimination result 17 is derived by the first derivation process, the marker information generation function 185 generates marker information according to the derived first discrimination result 17.

[0131] For example, if the first discrimination result 17 indicates that the ultrasound image contains a feature area, the marker information generation function 185 generates marker information that shows a rectangular marker representing the area detected as a feature area.

[0132] Figure 13 is a diagram illustrating an example of the processing performed by the marker information generation function 185 according to the first embodiment. For example, if the first discrimination result 17 indicates that the ultrasound image contains a feature area, the marker information generation function 185 generates marker information indicating a rectangular marker 32 representing the area detected as a feature area, as shown in Figure 13. In Figure 13, the reference numeral "31" refers to the actual feature area.

[0133] Then, each time the marker information generation function 185 generates marker information, it superimposes the marker 32 indicated by the marker information onto the ultrasound image, as shown in Figure 13. The marker information generation function 185 then stores the ultrasound image data showing the ultrasound image with the marker 32 superimposed into the image memory 160.

[0134] Furthermore, in step S203, the display control function 186 displays the ultrasound image shown by the ultrasound image data stored in the image memory 160 on the display 103. The ultrasound image is an example of a medical image. The display control function 186 is an example of a display control unit.

[0135] For example, the display control function 186 displays the ultrasound image with the marker 32 superimposed on it on the display 103 in real time. For example, whenever ultrasound image data showing the ultrasound image with the marker 32 superimposed is stored in the image memory 160 by the marker information generation function 185, the display control function 186 retrieves the ultrasound image data from the image memory 160. Then, whenever the display control function 186 retrieves ultrasound image data showing the ultrasound image with the marker 32 superimposed on it, it displays the ultrasound image with the marker 32 superimposed on it on the display 103.

[0136] On the other hand, if the first discrimination result 17 indicates that the ultrasound image does not contain characteristic areas, the marker information generation function 185 does not generate marker information.

[0137] However, the display control function 186 also displays ultrasound images in which the marker 32 is not superimposed on the display 103 in real time. For example, if the first discrimination result 17 indicates that the ultrasound image does not contain a feature area, the display control function 186 displays the ultrasound image exemplified in Figure 14 on the display 103. Note that Figure 14 shows an ultrasound image in which the feature area 31 is actually included, but the feature area was not detected in the first derivation process. Figure 14 is a diagram showing an example of display according to the first embodiment.

[0138] Once step S203 is complete, the ultrasound diagnostic device 1 terminates the process shown in Figure 12.

[0139] On the other hand, if it is determined that the movement speed of the ultrasonic probe 101 is less than or equal to a predetermined threshold Th1 (step S201: No), the discrimination function 184 determines whether the movement speed of the ultrasonic probe 101 is less than or equal to a predetermined threshold Th2 (step S204). Here, threshold Th2 is less than threshold Th1. Also, if the movement speed of the ultrasonic probe 101 is less than threshold Th2, for example, the operator may already be aware of the presence of a feature area and want to understand the details of that feature area. Threshold Th2 is an example of a second threshold.

[0140] If it is determined that the movement speed of the ultrasonic probe 101 is less than the threshold Th2 (step S204: Yes), in step S205, the discrimination function 184 performs the processing described below.

[0141] For example, in step S205, the discrimination function 184 selects the second frame rate from the first frame rate and the second frame rate as the frame rate to use when deriving the discrimination result.

[0142] Furthermore, for example, in step S205, the discrimination function 184 selects the second derivation process from the first derivation process and the second derivation process as the derivation process for deriving the discrimination result.

[0143] Here, the second derivation process derives the first discrimination result 17, the second discrimination result 23 (see Figure 5), and the third discrimination result 29 (see Figure 6). Therefore, the second derivation process derives more discrimination results than the first derivation process. Thus, the second derivation process has a higher processing load than the first derivation process. For this reason, it is preferable to lower the frame rate when executing the second derivation process. Accordingly, as described above, in step S205, the discrimination function 184 selects the second frame rate, which is relatively lower, from among the first and second frame rates.

[0144] Furthermore, since the operator may want to understand the details of the feature area, in step S205, the discrimination function 184 selects a second derivation process from among several types of derivation processes that derives more discrimination results than the first derivation process.

[0145] Then, in step S206, the discrimination function 184 executes the first detection process (the first stage of operation) included in the selected second derivation process, and derives the first discrimination result 17.

[0146] In step S206, the marker information generation function 185 performs the same processing as in step S203. Also in step S206, the display control function 186 performs the same processing as in step S203.

[0147] In step S206, and in steps S207 and S208 described below, the control function 181 controls the transmission circuit 110, reception circuit 120, B-mode processing circuit 130, Doppler processing circuit 140, image generation circuit 150, acquisition function 183, discrimination function 184, marker information generation function 185, and display control function 186 so that various processes are performed at the selected second frame rate. In other words, the control function 181 controls each process in steps S206 to S208 so that it is performed at the second frame rate.

[0148] Then, in step S207, the discrimination function 184 executes the second discrimination process (the second stage of operation) included in the selected second derivation process, and derives the second discrimination result 23.

[0149] In step S207, the marker information generation function 185 generates the string "malignant" or the string "benign" depending on the derived second discrimination result 23. For example, if the second discrimination result 23 indicates that the feature region is malignant, the marker information generation function 185 generates the string "malignant". Also, if the second discrimination result 23 indicates that the feature region is benign, the marker information generation function 185 generates the string "benign". The marker information generation function 185 generates the string "malignant" or the string "benign" each time the second discrimination result 23 is derived.

[0150] Then, in step S207, the display control function 186, each time the string "malignant" or the string "benign" is generated, superimposes the generated string "malignant" or the string "benign" onto the ultrasound image and displays it on the display 103.

[0151] Furthermore, the marker information generation function 185 may automatically register the generated string "malignant" or string "benign" in the report information of the subject P stored in the memory circuit 170 each time it generates the string "malignant" or string "benign". For example, the report information is information such as the history of ultrasound image diagnosis results of the subject P, which is registered by an operator such as a physician via the input interface 190. The operator then displays the report screen shown by the report information on the display 103 as needed to check the history of ultrasound image diagnosis results shown on the report screen. In this embodiment, the marker information generation function 185 automatically registers the determination result of whether the characteristic site is malignant or benign in the report information of the subject P, so that the operator does not have to manually register the determination result in the report information. Therefore, according to this embodiment, the burden on the operator when registering the determination result in the report information can be reduced. Furthermore, the marker information generation function 185 may automatically register a second determination result 23 in the report information instead of the string "malignant" and the string "benign".

[0152] Then, in step S208, the discrimination function 184 executes the third discrimination process (the third stage of processing during operation) included in the selected second derivation process, and derives the third discrimination result 29.

[0153] In step S208, the marker information generation function 185 generates the string "Malignancy A" or the string "Malignancy B" according to the derived third discrimination result 29. For example, if the third discrimination result 29 indicates that the malignancy of the feature region is malignancy A, the marker information generation function 185 generates the string "Malignancy A". Also, if the third discrimination result 29 indicates that the malignancy of the feature region is malignancy B, the marker information generation function 185 generates the string "Malignancy B". The marker information generation function 185 generates the string "Malignancy A" or the string "Malignancy B" each time the third discrimination result 29 is derived.

[0154] Then, in step S208, the display control function 186, each time the string "Malignancy Grade A" or the string "Malignancy Grade B" is generated, superimposes the generated string "Malignancy Grade A" or the string "Malignancy Grade B" onto the ultrasound image and displays it on the display 103.

[0155] Furthermore, the marker information generation function 185 may automatically register the generated string "Malignancy A" or string "Malignancy B" in the report information of the subject P stored in the memory circuit 170 each time it generates the string "Malignancy A" or string "Malignancy B". In this embodiment, the marker information generation function 185 automatically registers the determination result of the malignancy of the feature site in the report information of the subject P, eliminating the need for the operator to manually register the determination result in the report information. Therefore, according to this embodiment, the burden on the operator when registering the determination result in the report information can be reduced. Furthermore, the marker information generation function 185 may automatically register a third determination result 29 in the report information instead of the strings "Malignancy A" and "Malignancy B".

[0156] Once step S208 is complete, the ultrasound diagnostic device 1 terminates the process shown in Figure 12.

[0157] On the other hand, if it is determined that the movement speed of the ultrasound probe 101 is greater than or equal to the threshold Th2 (step S204: No), then in step S209, the ultrasound diagnostic device 1 performs the processing described below.

[0158] For example, in step S209, the ultrasound diagnostic device 1 performs the same processing as in step S203. However, in step S209, the control function 181 controls the transmission circuit 110, reception circuit 120, B-mode processing circuit 130, Doppler processing circuit 140, image generation circuit 150, acquisition function 183, discrimination function 184, marker information generation function 185, and display control function 186 to perform various processing at a predetermined frame rate. This predetermined frame rate may be a frame rate other than the first frame rate and the second frame rate, or it may be the first frame rate or the second frame rate.

[0159] Once step S209 is complete, the ultrasound diagnostic device 1 terminates the process shown in Figure 12.

[0160] In the process shown in Figure 12, the discrimination function 184 performs a process to select the frame rate for deriving the discrimination result from among multiple types of frame rates of multiple ultrasound image data, according to the movement speed of the ultrasound probe 101. Such a process of selecting a frame rate is an example of the first selection process.

[0161] Furthermore, in the process shown in Figure 12, the discrimination function 184 executes a process to select a derivation process for deriving a discrimination result from among several types of derivation processes for deriving a discrimination result, according to the movement speed of the ultrasound probe 101. Such a process of selecting a derivation process is an example of a second selection process. For example, in the process shown in Figure 12, the discrimination function 184 uses at least one trained model and at least one training condition from among several trained models and multiple training conditions, according to the movement speed of the ultrasound probe 101, and executes a discrimination process corresponding to this trained model and training condition to derive a discrimination result regarding the characteristic part of the subject P.

[0162] Therefore, according to the ultrasound diagnostic apparatus 1 of the first embodiment, appropriate processing can be performed according to the movement speed of the ultrasound probe 101.

[0163] The ultrasound diagnostic apparatus 1 according to the first embodiment has been described above. As described above, the ultrasound diagnostic apparatus 1 according to the first embodiment can obtain discrimination results regarding characteristic areas with high accuracy. Furthermore, as described above, the ultrasound diagnostic apparatus 1 according to the first embodiment can perform appropriate processing according to the movement speed of the ultrasound probe 101.

[0164] (Second embodiment) In the first embodiment described above, the case was explained in which the trained model generation function 201a of the trained model generation device 200 generates various trained models and records various learning conditions. However, the ultrasound diagnostic device 1 may generate various trained models and record various learning conditions. Therefore, such an embodiment will be described as the second embodiment. In the description of the second embodiment, the differences from the first embodiment will be mainly described, and the same configurations and processes as in the first embodiment may be omitted from the description.

[0165] Figure 15 shows an example of the configuration of the ultrasound diagnostic apparatus 1 according to the second embodiment. The second embodiment differs from the first embodiment in that the processing circuit 180 has a trained model generation function 187.

[0166] In the first embodiment described above, the trained model generation function 201a generates various trained models and records various training conditions using ultrasound image data generated by the ultrasound diagnostic device 1 or another ultrasound diagnostic device.

[0167] On the other hand, the trained model generation function 187 uses ultrasonic image data (e.g., B-mode image data) generated by the image generation circuit 150 to generate various trained models, similar to the trained model generation function 201a, and records various training conditions. The trained model generation function 187 is an example of a generation unit.

[0168] Specifically, the trained model generation function 187 generates a first trained model 11a, a second trained model 12a, and a third trained model 13a. The trained model generation function 187 also records the first learning conditions 11b, the second learning conditions 12b, and the third learning conditions 13b in the memory circuit 170. In other words, the trained model generation function 187 generates a trained model used when executing a discrimination process that outputs a discrimination result regarding the feature areas of a subject depicted in the input ultrasound image data, by learning by associating ultrasound image data with discrimination results (labels) regarding the feature areas depicted in the ultrasound image data, and records the learning conditions. The trained model generation function 187 generates a trained model and records the learning conditions at each of the multiple stages of learning (the first to third stages of learning described above).

[0169] Here, the pre-trained model generation function 187, similar to the pre-trained model generation function 201a, asynchronously generates the first pre-trained model 11a, the second pre-trained model 12a, and the third pre-trained model 13a. The pre-trained model generation function 187 also asynchronously records the first learning condition 11b, the second learning condition 12b, and the third learning condition 13b.

[0170] The trained model generation function 187 then stores the first trained model 11a, the second trained model 12a, and the third trained model 13a in the memory circuit 170. The trained model generation function 187 also stores the first learning conditions 11b, the second learning conditions 12b, and the third learning conditions 13b in the memory circuit 170.

[0171] In the second embodiment, the discrimination function 184 performs the same processing as in the first embodiment using various trained models and various learning conditions stored in the memory circuit 170.

[0172] The second embodiment has been described above. According to the second embodiment, similar to the first embodiment, discrimination results regarding characteristic areas can be obtained with high accuracy. Furthermore, according to the second embodiment, similar to the first embodiment, appropriate processing can be performed according to the movement speed of the ultrasonic probe 101.

[0173] In the second embodiment, the trained model generation function 187 may generate the first trained model 11a, the second trained model 12a, and the third trained model 13a synchronously, rather than asynchronously. Similarly, the trained model generation function 187 may record the first learning condition 11b, the second learning condition 12b, and the third learning condition 13b synchronously, rather than asynchronously.

[0174] For example, first, the trained model generation function 187 generates a first trained model 11a and records the first learning conditions 11b. After generating the first trained model 11a and recording the first learning conditions 11b, the discrimination function 184 performs a first discrimination process using the first trained model 11a and the first learning conditions 11b. Then, the trained model generation function 187 uses the ultrasound image data that was determined to contain feature regions in the first discrimination process as part of the training data to generate a second trained model 12a and records the second learning conditions 12b. This makes it possible to automatically prepare a portion of the training data used when generating the second trained model 12a without the operator having to prepare it.

[0175] An example of a method for generating a second trained model 12a in such a case will be explained with reference to Figure 16. Figure 16 is a flowchart showing an example of the process for generating a second trained model 12a according to the second embodiment.

[0176] As illustrated in Figure 16, the trained model generation function 187 determines whether or not ultrasound image data depicting feature regions and a second label 21 have been input (step S301). Here, an example of ultrasound image data that is the subject of the determination of whether or not it has been input in step S301 will be described. For example, the ultrasound image data subject to determination is ultrasound image data that has been determined to contain feature regions in the first discrimination process, and that has been confirmed to contain feature regions by the operator's visual inspection. The operator inputs such ultrasound image data, along with the second label 21, to the device body 100 via the input device 102.

[0177] If it is determined that ultrasound image data and the second label 21 have not been input (step S301: No), the trained model generation function 187 performs the determination in step S301 again. On the other hand, if it is determined that ultrasound image data and the second label 21 have been input (step S301: Yes), the trained model generation function 187 uses the input ultrasound image data and the second label 21 to generate a second trained model 12a and records the second training conditions 12b (step S302). Then, the trained model generation function 187 terminates the process shown in Figure 16.

[0178] Then, after the second trained model 12a is generated and the second learning conditions 12b are recorded, the discrimination function 184 performs a second discrimination process using the second trained model 12a and the second learning conditions 12b. Then, the trained model generation function 187 generates a third trained model 13a using ultrasound image data depicting the feature region identified as malignant in the second discrimination process as part of the training data. This makes it possible to automatically prepare a portion of the training data used to generate the third trained model 13a without the operator having to prepare it.

[0179] An example of a method for generating a third trained model 13a in such a case will be explained with reference to Figure 17. Figure 17 is a flowchart showing an example of the process for generating a third trained model 13a according to the second embodiment.

[0180] As illustrated in Figure 17, the trained model generation function 187 determines whether or not the ultrasound image data depicting the feature region determined to be malignant in the second discrimination process, and the third label 27, have been input (step S401). Here, an example of the ultrasound image data that is the subject of the determination of whether or not it has been input in step S401 will be described. For example, the ultrasound image data subject to determination is ultrasound image data depicting the feature region determined to be malignant in the second discrimination process, and ultrasound image data in which the feature region has been confirmed to be malignant by the operator's visual inspection. The operator inputs such ultrasound image data, along with the third label 27, to the device body 100 via the input device 102.

[0181] If it is determined that ultrasound image data and the third label 27 have not been input (step S401: No), the trained model generation function 187 performs the determination in step S401 again. On the other hand, if it is determined that ultrasound image data and the third label 27 have been input (step S401: Yes), the trained model generation function 187 uses the input ultrasound image data and the third label 27 to generate a third trained model 13a and records the third training conditions 13b (step S402). Then, the trained model generation function 187 terminates the process shown in Figure 17.

[0182] The above describes other methods for generating pre-trained models. In these other methods for generating pre-trained models, the pre-trained model generation function 187 generates a pre-trained model corresponding to the next stage by learning by associating ultrasound image data corresponding to a predetermined discrimination result output from a discrimination process corresponding to one of the multiple stages with the discrimination result (label) corresponding to the next stage after the first stage. Therefore, with these other methods for generating pre-trained models, training data can be easily prepared.

[0183] (Modification 1 of the first and second embodiments) In the first and second embodiments described above, the case was explained in which the discrimination function 184 selects a derivation process from among several types of derivation processes when deriving the discrimination result according to the movement speed of the ultrasonic probe 101. However, the discrimination function 184 may also select a derivation process when deriving the discrimination result depending on whether the discrimination result is to be derived in real time or as a post-processing step. Therefore, such an embodiment will be described as Modification 1 of the first and second embodiments. Note that in the description of Modification 1, the differences from the first and second embodiments will be mainly described, and the same configurations and processes as in the first and second embodiments may be omitted from the description.

[0184] For example, in Modification 1, the ultrasound diagnostic device 1 receives instructions from the operator via the input device 102 to derive discrimination results in real time from the ultrasound scan (real-time processing instructions). In addition, in Modification 1, the operator also receives instructions via the input device 102 to derive discrimination results as post-processing performed after the ultrasound scan (post-processing instructions).

[0185] In the modified example 1, when a real-time processing instruction is received, the ultrasound diagnostic device 1 performs the same processing as in steps S202 and S203 shown in Figure 12.

[0186] Furthermore, when a post-processing instruction is input to the ultrasound diagnostic device 1, it performs post-processing on multiple ultrasound image data stored in the memory circuit 170, rather than in real time, similar to the processing in steps S205 to S208 shown in Figure 12.

[0187] Therefore, according to the ultrasonic diagnostic apparatus 1 according to Modification 1 of the first and second embodiments, appropriate processing can be performed depending on whether the discrimination result is derived in real time or as a post-processing step. Furthermore, according to the ultrasonic diagnostic apparatus 1 according to Modification 1, discrimination results regarding characteristic areas can be obtained with high accuracy, similar to the first and second embodiments.

[0188] (Modification 2 of the first and second embodiments) In the first and second embodiments described above, the case in which the discrimination function 184 selects whether to perform the first stage of processing during operation or the first to third stages of processing during operation, depending on the movement speed of the ultrasonic probe 101, was explained.

[0189] However, the discrimination function 184 may select whether to perform the first detection process shown in Figure 7, or to perform both the first and second detection processes shown in Figure 7, depending on the movement speed of the ultrasonic probe 101. Therefore, such an embodiment will be described as Modification 2 of the first and second embodiments. Note that in the description of Modification 2, the differences from the first and second embodiments will be mainly described, and the same configurations and processes as those in the first and second embodiments may be omitted.

[0190] Figure 18 is a flowchart showing an example of the process performed by the ultrasound diagnostic device 1 according to the first embodiment and a modified example 2 of the second embodiment. For example, the ultrasound diagnostic device 1 performs the process shown in Figure 18 at predetermined time intervals.

[0191] As shown in Figure 18, the discrimination function 184 determines whether the movement speed of the ultrasonic probe 101 is greater than a predetermined threshold Th1 (step S501). In step S501, the velocity detection function 182 detects the movement speed of the ultrasonic probe 101. Then, in step S501, the discrimination function 184 determines whether the movement speed of the ultrasonic probe 101 detected by the velocity detection function 182 is greater than the threshold Th1.

[0192] If it is determined that the movement speed of the ultrasonic probe 101 is greater than a predetermined threshold Th1 (step S501: Yes), then in step S502, the discrimination function 184 performs the processing described below.

[0193] For example, in step S502, the discrimination function 184 selects the first frame rate from the first frame rate and the second frame rate as the frame rate to be used when deriving the discrimination result.

[0194] Furthermore, for example, in step S502, the discrimination function 184 selects the first derivation process from among the first derivation process and the third derivation process as the derivation process for deriving the discrimination result. Here, the first derivation process according to the modified example 2 is the first detection process shown in Figure 7 above. Also, the third derivation process according to the modified example 2 is the first detection process and the second detection process shown in Figure 7 above.

[0195] As described above, in the second detection process, from among the feature regions detected in the first detection process, the feature regions that were over-detected in the first detection process are ultimately detected as feature regions. For this reason, in the modified example 2, the third derivation process can derive the first discrimination result, which is the result of determining whether or not feature regions are included in the ultrasound image, with greater accuracy than the first derivation process. On the other hand, the first derivation process has a lower processing load than the third derivation process.

[0196] If a positive determination is made in step S501 (step S501: Yes), it is considered that a higher frame rate takes precedence over good accuracy of the first discrimination result. Therefore, in step S502, the discrimination function 184 selects a first derivation process from among several types of derivation processes that determines whether or not a characteristic area is included in the ultrasound image with a relatively small processing load.

[0197] Furthermore, the discrimination function 184 can increase the frame rate when executing the first derivation process. Therefore, as described above, in step S502, the discrimination function 184 selects a relatively high first frame rate from among several types of frame rates.

[0198] Then, in step S503, the discrimination function 184 executes the selected first derivation process (first detection process). Also in step S503, the control function 181 controls the transmission circuit 110, reception circuit 120, B-mode processing circuit 130, Doppler processing circuit 140, image generation circuit 150, acquisition function 183, discrimination function 184, marker information generation function 185, and display control function 186 to perform various processes at the selected first frame rate.

[0199] Furthermore, in step S503, the marker information generation function 185 generates marker information indicating the detection result of the ultrasound image based on the feature area detected by the first detection process. For example, in step S503, each time a feature area is detected by the first detection process, the marker information generation function 185 generates marker information indicating a rectangular marker representing the area detected as a feature area, similar to the first embodiment.

[0200] Then, similar to the first embodiment, the marker information generation function 185 superimposes the markers indicated by the marker information onto the ultrasound image each time marker information is generated. Then, similar to the first embodiment, the marker information generation function 185 stores the ultrasound image data showing the ultrasound image with the superimposed markers in the image memory 160.

[0201] Furthermore, in step S503, the display control function 186 displays the ultrasound image shown by the ultrasound image data for display stored in the image memory 160 on the display 103, similar to the first embodiment. That is, in step S503, the ultrasound image with markers superimposed and the ultrasound image without markers superimposed are displayed on the display 103 in real time.

[0202] Once step S503 is complete, the ultrasound diagnostic device 1 terminates the process shown in Figure 18.

[0203] On the other hand, if it is determined that the movement speed of the ultrasonic probe 101 is less than or equal to a predetermined threshold Th1 (step S501: No), the discrimination function 184 determines whether or not the movement speed of the ultrasonic probe 101 is less than or equal to a predetermined threshold Th2 (step S504).

[0204] If it is determined that the movement speed of the ultrasonic probe 101 is less than the threshold Th2 (step S504: Yes), in step S505, the discrimination function 184 performs the processing described below.

[0205] If a positive determination is made in step S504 (step S504: Yes), it is considered that good accuracy of the first discrimination result takes precedence over a high frame rate. Here, the third derivation process can derive the first discrimination result, which is the result of determining whether or not a characteristic area is included in the ultrasound image, with greater accuracy than the first derivation process. Therefore, in step S505, the discrimination function 184 selects the third derivation process from among the first derivation process and the third derivation process as the derivation process for deriving the discrimination result.

[0206] Furthermore, the third derivation process has a higher processing load than the first derivation process. For this reason, it is preferable to lower the frame rate when executing the third derivation process. Therefore, in step S505, the discrimination function 184 selects the second frame rate from the first frame rate and the second frame rate as the frame rate when deriving the discrimination result.

[0207] Then, in step S506, the discrimination function 184 executes the first detection process included in the selected third derivation process.

[0208] In step S506 and S507, which will be described below, the control function 181 controls the transmission circuit 110, reception circuit 120, B-mode processing circuit 130, Doppler processing circuit 140, image generation circuit 150, acquisition function 183, discrimination function 184, marker information generation function 185, and display control function 186 so that various processes are performed at the selected second frame rate. In other words, the control function 181 controls each process in steps S506 and S507 so that it is performed at the second frame rate.

[0209] Then, in step S507, the discrimination function 184 executes the second detection process included in the selected third derivation process, and finally detects the feature region.

[0210] Furthermore, in step S507, the marker information generation function 185 generates marker information indicating the detection result of the ultrasound image based on the feature area finally detected by the second detection process. For example, in step S507, each time a feature area is finally detected by the second detection process, the marker information generation function 185 generates marker information indicating a rectangular marker representing the area detected as a feature area, similar to the first embodiment.

[0211] Then, similar to the first embodiment, the marker information generation function 185 superimposes the markers indicated by the marker information onto the ultrasound image each time marker information is generated. Then, similar to the first embodiment, the marker information generation function 185 stores the ultrasound image data showing the ultrasound image with the superimposed markers in the image memory 160.

[0212] Furthermore, in step S507, the display control function 186 displays the ultrasound image shown by the ultrasound image data for display stored in the image memory 160 on the display 103, similar to the first embodiment. That is, in step S507, the ultrasound image with markers superimposed and the ultrasound image without markers superimposed are displayed on the display 103 in real time.

[0213] Once step S507 is completed, the ultrasound diagnostic device 1 terminates the process shown in Figure 18.

[0214] On the other hand, if it is determined that the movement speed of the ultrasound probe 101 is greater than or equal to the threshold Th2 (step S504: No), then in step S508, the ultrasound diagnostic device 1 performs the processing described below.

[0215] For example, in step S508, the ultrasound diagnostic device 1 performs the same processing as in step S503. However, the control function 181 controls the transmission circuit 110, reception circuit 120, B-mode processing circuit 130, Doppler processing circuit 140, image generation circuit 150, acquisition function 183, discrimination function 184, marker information generation function 185, and display control function 186 to perform various processing at a predetermined frame rate. This predetermined frame rate may be a frame rate other than the first frame rate and the second frame rate, or it may be the first frame rate or the second frame rate.

[0216] Once step S508 is complete, the ultrasound diagnostic device 1 terminates the process shown in Figure 18.

[0217] In the process shown in Figure 18, the discrimination function 184 performs a process to select the frame rate for deriving the discrimination result from among multiple types of frame rates of multiple ultrasound image data, according to the movement speed of the ultrasound probe 101.

[0218] Furthermore, in the process shown in Figure 18, the discrimination function 184 executes a process to select a derivation process for deriving the discrimination result from among several types of derivation processes for deriving the discrimination result, according to the movement speed of the ultrasonic probe 101.

[0219] Therefore, according to the ultrasonic diagnostic apparatus 1 according to Modification 2 of the first and second embodiments, appropriate processing can be performed according to the movement speed of the ultrasonic probe 101, similar to the first and second embodiments. Furthermore, according to the ultrasonic diagnostic apparatus 1 according to Modification 2, discrimination results regarding characteristic areas can be obtained with high accuracy, similar to the first and second embodiments.

[0220] (Modification 3 of the first and second embodiments) In the first and second embodiments described above, the case in which the discrimination function 184 performs the first detection process and the second detection process was explained. However, the discrimination function 184 may perform a different process instead of the first detection process and the second detection process. Therefore, such an embodiment will be described as Modification 3 of the first and second embodiments.

[0221] In the description of Modification 3, the differences from the first and second embodiments will be mainly explained, and the same configurations and processes as those in the first and second embodiments may be omitted from the explanation.

[0222] In Modification 3, an example of the first discrimination process performed by the discrimination function 184 using the first trained model 11a in the first stage of operation will be described. Figure 19 is a flowchart showing the flow of an example of the first discrimination process according to the first embodiment and Modification 3 of the second embodiment.

[0223] As shown in Figure 19, in the first discrimination process, the discrimination function 184 executes a third detection process (step S601), and in parallel with the third detection process, executes a fourth detection process (step S602), and then executes a fifth detection process (step S603).

[0224] First, the third detection process performed in step S601 will be described. The discrimination function 184 uses the search window, the second pre-trained search model, and the search algorithm to search for feature regions within the search range set by the operator in the ultrasound image. For example, the discrimination function 184 moves the search window to multiple positions within the search range. At each position, the discrimination function 184 analyzes the image information within the search window using the second pre-trained search model and the search algorithm to calculate the probability (third probability) that the image information within each search window corresponds to a feature region.

[0225] The search box is, for example, a unit of region within the search range that is compared with the features of the second pre-trained search model. The second pre-trained search model is, for example, a model that has learned image information that serves as a sample of a feature region. The search algorithm is an algorithm for calculating the third probability mentioned above and searching for image information corresponding to the feature region in the ultrasound image to be processed.

[0226] The second pre-trained model for searching is an example of the first pattern that represents feature regions.

[0227] The discrimination function 184 then detects the area enclosed by the search window as a feature area (range of the feature area) if the calculated third probability is greater than or equal to a predetermined threshold. In other words, the discrimination function 184 determines that the ultrasound image contains a feature area if the third probability, which indicates the degree of similarity between the features of a part of the ultrasound image and the features of the second trained search model that represents the feature area, is greater than or equal to a threshold. The discrimination function 184 also determines that the ultrasound image does not contain a feature area if the third probability is less than the threshold.

[0228] The discrimination function 184 then stores the search result (detection result) from the third detection process on the ultrasound image in the memory circuit 170 for each frame of ultrasound image data. For example, if the discrimination function 184 is able to detect a feature area from a frame of ultrasound image, it stores position information indicating the location of the feature area in the memory circuit 170 as the search result from the third detection process. The position of the feature area referred to here is the position in the image space of the ultrasound image data.

[0229] Furthermore, if no feature area is detected from a single frame of ultrasound image, the discrimination function 184 stores information indicating that no feature area was detected in the memory circuit 170 as a search result in the third detection process.

[0230] The discrimination function 184 has described the case in which it calculates a third probability as an example of an indicator that shows the degree to which the features of the image information in the search box are similar to the features of the second pre-trained search model. However, instead of the third probability, the discrimination function 184 may calculate other indicators such as similarity, which shows the degree to which the features of the image information in the search box are similar to the features of the second pre-trained search model.

[0231] The above describes an example of the third detection process performed in step S601. Figure 20 shows an example of the results of the third detection process according to the first embodiment and modification 3 of the second embodiment. The "Result determined to be a feature area" shown in Figure 20 shows an example of image information in the search window 30 when a feature area is detected by the third detection process. The "Result determined not to be a feature area" shown in Figure 20 shows an example of image information in the search window 30 when a feature area is not detected by the third detection process. In addition, Figure 20 shows eight image information locations as an example of the results of calculating the third probability at each location when the search window 30 is placed at multiple locations within the search range in the third detection process.

[0232] As shown in Figure 20, the discrimination function 184 detects feature regions at four locations in the third detection process. Specifically, as shown in Figure 20, the discrimination function 184 detects four image information 50a to 50d within the search window 30.

[0233] Image information 50a is an over-detected image. Image information 50b is an image showing the actual feature region 31e. Image information 50c is an image showing the actual feature region 31f. Image information 50d is an over-detected image. In other words, in the case shown in Figure 20, two of the four image pieces determined to be feature regions are over-detected.

[0234] Furthermore, as shown in Figure 20, the discrimination function 184 failed to detect feature regions at four locations during the third detection process. Specifically, as shown in Figure 20, the discrimination function 184 failed to detect the four image information pieces 50e to 50h within the search window 30 as feature regions. In addition, the discrimination function 184 failed to detect image information 50e even though it contained feature region 31g.

[0235] Next, the fourth detection process performed in step S602 will be described. The discrimination function 184 uses the search window, the third pre-trained search model, and the search algorithm to search for feature regions within the search range set by the operator in the ultrasound image. For example, the discrimination function 184 moves the search window to multiple positions within the search range. At each position, the discrimination function 184 analyzes the image information within the search window using the third pre-trained search model and the search algorithm to calculate the probability (fourth probability) that the image information within each search window corresponds to a feature region.

[0236] The search box is, for example, a unit of region within the search range that is compared with the features of the third pre-trained search model. The third pre-trained search model is, for example, a model that has learned image information that serves as a sample of a feature region. Here, the third pre-trained search model is a different model from the second pre-trained search model. The search algorithm is an algorithm for calculating the fourth probability described above and searching for image information corresponding to the feature region in the ultrasound image to be processed.

[0237] For example, the second pre-trained search model described above is a model obtained through training using a first set of image information in which feature regions are depicted and a second set of image information in which feature regions are not depicted. On the other hand, the third pre-trained search model is a model obtained through training using a third set of image information in which feature regions are depicted and a second set of image information in which feature regions are not depicted. In other words, the second set of image information is used in common during training for both the second pre-trained search model and the third pre-trained search model.

[0238] Therefore, the second pre-trained search model and the third pre-trained search model are trained using different image information, even when the same feature area (e.g., a tumor) is depicted, resulting in different feature areas being detected. On the other hand, the second and third pre-trained search models are trained using the same set of image information (the second set of image information) as a set of image information in which feature areas are not depicted. Therefore, the second and third pre-trained search models share the same images that are correctly judged as not being feature areas, as well as the same images that are over-detected.

[0239] The third pre-trained model for retrieval is an example of a second pattern that differs from the first pattern that represents feature regions.

[0240] The discrimination function 184 then detects the area enclosed by the search window as a feature area (range of the feature area) if the calculated fourth probability is greater than or equal to a predetermined threshold. In other words, the discrimination function 184 determines that the ultrasound image contains a feature area if the fourth probability, which indicates the degree of similarity between the features of a part of the ultrasound image and the features of the third pre-trained search model that represents the feature area, is greater than or equal to a threshold. The discrimination function 184 also determines that the ultrasound image does not contain a feature area if the fourth probability is less than the threshold.

[0241] The discrimination function 184 then stores the search result (detection result) from the fourth detection process on the ultrasound image in the memory circuit 170 for each frame of ultrasound image data. For example, if a feature area is detected from a frame of ultrasound image, the discrimination function 184 stores position information indicating the location of the feature area in the memory circuit 170 as the search result from the fourth detection process. The position of the feature area referred to here is the position in the image space of the ultrasound image data.

[0242] Furthermore, if the discrimination function 184 does not detect a feature area from a single frame of ultrasound image, it stores information indicating that the feature area was not detected in the memory circuit 170 as a search result in the fourth detection process.

[0243] The discrimination function 184 has been described in an example of an indicator that shows the degree of similarity between the features of the image information in the search box and the features of the third pre-trained search model. However, instead of the fourth probability, the discrimination function 184 may calculate other indicators such as similarity, which shows the degree of similarity between the features of the image information in the search box and the features of the third pre-trained search model.

[0244] The above describes an example of the fourth detection process performed in step S602. Figure 21 is a diagram showing an example of the results of the fourth detection process according to the first embodiment and modification 3 of the second embodiment. The "Result determined to be a feature area" shown in Figure 21 shows an example of image information in the search window 30 when a feature area is detected by the fourth detection process. The "Result determined not to be a feature area" shown in Figure 21 shows an example of image information in the search window 30 when a feature area is not detected by the fourth detection process. In addition, Figure 21 shows eight image information locations as an example of the results of calculating the fourth probability at each location when the search window 30 is placed at multiple locations within the search range in the fourth detection process.

[0245] As shown in Figure 21, the discrimination function 184 detects feature regions at four locations in the fourth detection process. Specifically, as shown in Figure 21, the discrimination function 184 detects four image information points 50i to 50l within the search window 30.

[0246] Image information 50i is the image information in which the actual feature region 31h is depicted. Image information 50j is the image information that was overdetected. Image information 50k is the image information in which the actual feature region 31i is depicted. Image information 50l is the image information that was overdetected. In other words, in the case shown in Figure 21, two of the four image information pieces determined to be feature regions were overdetected.

[0247] Furthermore, as shown in Figure 21, the discrimination function 184 failed to detect feature regions at four locations during the fourth detection process. Specifically, as shown in Figure 21, the discrimination function 184 failed to detect the four image information points 50m to 50p within the search window 30 as feature regions. In addition, the discrimination function 184 failed to detect image information 50o, even though it contained feature region 31j. Similarly, the discrimination function 184 failed to detect image information 50p, even though it contained feature region 31k.

[0248] Next, the fifth detection process performed in step S603 will be described. In the fifth detection process, the discrimination function 184 uses the detection results of the third detection process and the detection results of the fourth detection process to finally detect the feature region.

[0249] Here, as described above, the image information in which the actual feature regions are depicted from the image information detected using the second pre-trained search model in the third detection process is different from the image information in which the actual feature regions are depicted from the image information detected using the third pre-trained search model in the fourth detection process.

[0250] For example, the image information 50b and 50c shown in Figure 20 are different from the image information 50i and 50k shown in Figure 20.

[0251] However, as described above, the overdetected image information among the image information detected using the second pre-trained search model in the third detection process and the overdetected image information among the image information detected using the third pre-trained search model in the fourth detection process are the same (common). In other words, the overdetected portion of the ultrasound image is the same in the third detection process and the fourth detection process.

[0252] For example, image information 50a shown in Figure 20 and image information 50j shown in Figure 21 are identical. Also, image information 50d shown in Figure 20 and image information 50l shown in Figure 21 are identical.

[0253] Therefore, in the fifth detection process, the discrimination function 184 detects feature regions other than those that were commonly detected in both the third and fourth detection processes, among the feature regions detected in the third and fourth detection processes.

[0254] For example, in the fifth detection process, the discrimination function 184 identifies the image information 50a to 50d detected in the third detection process and the image information 50i to 50l detected in the fourth detection process. Then, in the fifth detection process, the discrimination function 184 identifies the image information 50a, 50d, 50j, and 50l that were detected in common in two of the detection processes from the identified image information 50a to 50d and 50i to 50l.

[0255] Figure 22 shows an example of the results of the fifth detection process according to the first embodiment and modification 3 of the second embodiment. In the fifth detection process, the discrimination function 184 ultimately detects the image information 50b, 50c, 50i, and 50k shown in Figure 22, which are not the identified image information 50a, 50d, 50j, and 50l, as feature regions.

[0256] In this way, the discrimination function 184 ultimately detects as feature regions image information from the image information detected by the third detection process using the second pre-trained search model, and the image information detected by the fourth detection process using the third pre-trained search model, excluding the image information that was detected in common in both the third and fourth detection processes, and determines that the ultrasound image contains feature regions. The third detection process is an example of the first feature region detection process. The fourth detection process is an example of the second feature region detection process.

[0257] Thus, the discrimination function 184 according to Modification 3 detects the feature area by removing the over-detected feature area so that it is not included in the final detected feature area. Therefore, the ultrasound diagnostic device 1 according to Modification 3 can obtain discrimination results regarding the feature area with even greater accuracy.

[0258] Then, in the fifth detection process, the discrimination function 184 stores positional information indicating the location of the finally detected feature region in the memory circuit 170 as a search result.

[0259] Furthermore, the discrimination function 184 stores information indicating that a feature area was not detected in the fifth detection process in the memory circuit 170 as a search result.

[0260] The marker information generation function 185 then generates marker information indicating the detection result of the ultrasound image based on the feature region finally detected by the fifth detection process. For example, each time a feature region is finally detected by the fifth detection process, the marker information generation function 185 generates marker information indicating a rectangular marker representing the area detected as a feature region, similar to the first embodiment.

[0261] Then, similar to the first embodiment, the marker information generation function 185 superimposes the markers indicated by the marker information onto the ultrasound image each time marker information is generated. Then, similar to the first embodiment, the marker information generation function 185 stores the ultrasound image data showing the ultrasound image with the superimposed markers in the image memory 160.

[0262] Then, similar to the first embodiment, the display control function 186 causes the ultrasonic image indicated by the ultrasonic image data for display stored in the image memory 160 to be displayed on the display 103. That is, the display control function 186 displays ultrasonic images with markers superimposed and ultrasonic images without markers superimposed on the display 103 in real time.

[0263] The ultrasound diagnostic apparatus 1 according to the first embodiment and modification 3 of the second embodiment has been described above. As described above, according to the ultrasound diagnostic apparatus 1 according to modification 3, discrimination results regarding characteristic areas can be obtained with even greater accuracy.

[0264] (Modification 4 of the first and second embodiments) In the first and second embodiments described above, the case was explained in which the discrimination function 184 selects a derivation process from among several types of derivation processes when deriving the discrimination result, according to the movement speed of the ultrasonic probe 101. However, the discrimination function 184 may also select a trained model to be used when deriving the discrimination result from among several types of trained models, according to the movement speed of the ultrasonic probe 101. Therefore, such an embodiment will be described as Modification 4 of the first and second embodiments.

[0265] In the description of the 4th modified example of the first and second embodiments, the differences from the first and second embodiments will be mainly described, and the same configurations and processes as those of the first and second embodiments may be omitted from the description.

[0266] Figure 23 is a diagram illustrating examples of various trained models and various learning conditions related to the first embodiment and a modified example 4 of the second embodiment. As illustrated in Figure 23, the memory circuit 170 stores two first trained models 11a_1 and 11a_2 used in the first stage of processing during operation. The memory circuit 170 also stores two second trained models 12a_1 and 12a_2 used in the second stage of processing during operation. Furthermore, the memory circuit 170 stores two third trained models 13a_1 and 13a_2 used in the third stage of processing during operation. In this way, the memory circuit 170 stores multiple types of trained models corresponding to each of the multiple stages.

[0267] Furthermore, the memory circuit 170 stores two first learning conditions 11b_1 and 11b_2 used in the first stage of processing during operation. The memory circuit 170 also stores two second learning conditions 12b_1 and 12b_2 used in the second stage of processing during operation. Additionally, the memory circuit 170 stores two third learning conditions 13b_1 and 13b_2 used in the third stage of processing during operation. In this way, the memory circuit 170 stores multiple types of learning conditions corresponding to each of the multiple stages.

[0268] The first learning condition 11b_1 is used in the first discrimination process using the first learned model 11a_1. The first learning condition 11b_2 is used in the first discrimination process using the first learned model 11a_2.

[0269] The second learning condition 12b_1 is used in the second discrimination process using the second learned model 12a_1. The second learning condition 12b_2 is used in the second discrimination process using the second learned model 12a_2.

[0270] The third learning condition 13b_1 is used in the third discrimination process using the third learned model 13a_1. The third learning condition 13b_2 is used in the third discrimination process using the third learned model 13a_2.

[0271] Thus, one learned model corresponds to one learning condition. Here, the processing load of the discrimination process using the learned model is the load corresponding to the learning condition corresponding to this learned model.

[0272] The two first learned models 11a_1, 11a_2 are adjusted parameters used in the first discrimination process. However, the discrimination accuracy in the first discrimination process using the first learned model 11a_1 is higher than the discrimination accuracy in the first discrimination process using the first learned model 11a_2. On the other hand, the processing load of the first discrimination process using the first learned model 11a_2 is lower than the processing load of the first discrimination process using the first learned model 11a_1.

[0273] Also, the two second pre-trained models 12a_1 and 12a_2 are the adjusted parameters used for the second discrimination process. However, the discrimination accuracy in the second discrimination process using the second pre-trained model 12a_1 is higher than the discrimination accuracy in the second discrimination process using the second pre-trained model 12a_2. On the other hand, the processing load of the second discrimination process using the second pre-trained model 12a_2 is lower than the processing load of the second discrimination process using the second pre-trained model 12a_1.

[0274] Also, the two third pre-trained models 13a_1 and 13a_2 are the adjusted parameters used for the third discrimination process. However, the discrimination accuracy in the third discrimination process using the third pre-trained model 13a_1 is higher than the discrimination accuracy in the third discrimination process using the third pre-trained model 13a_2. On the other hand, the processing load of the third discrimination process using the third pre-trained model 13a_2 is lower than the processing load of the third discrimination process using the third pre-trained model 13a_1.

[0275] When the discrimination function 184 according to the fourth modification example performs the first-stage first discrimination process during operation, when the moving speed of the ultrasonic probe 101 is greater than the threshold value Th1, as the first pre-trained model used when executing the first discrimination process, the first pre-trained model 11a_2 is selected. Then, the discrimination function 184 executes the first discrimination process using the selected first pre-trained model 11a_2 and the corresponding first learning condition 11b_2.

[0276] Furthermore, when the discrimination function 184 performs the first discrimination process in the first stage of operation, if the movement speed of the ultrasonic probe 101 is less than the threshold Th2, it selects the first trained model 11a_1 as the first trained model to be used when executing the first discrimination process. In this way, when the movement speed of the ultrasonic probe 101 is less than the threshold Th2, the discrimination function 184 selects the first trained model 11a_1, which can derive the first discrimination result with better accuracy than the first trained model 11a_2. Then, the discrimination function 184 executes the first discrimination process using the selected first trained model 11a_1 and the corresponding first learning condition 11b_1.

[0277] Furthermore, when the discrimination function 184 performs the second stage of discrimination processing during operation, if the movement speed of the ultrasonic probe 101 is greater than the threshold Th1, it selects the second trained model 12a_2 as the second trained model to be used when executing the second discrimination processing. Then, the discrimination function 184 executes the second discrimination processing using the selected second trained model 12a_2 and the corresponding second training condition 12b_2.

[0278] Furthermore, when the discrimination function 184 performs the second stage of discrimination processing during operation, if the movement speed of the ultrasonic probe 101 is less than the threshold Th2, it selects the second trained model 12a_1 as the second trained model to be used when executing the second discrimination processing. In this way, when the movement speed of the ultrasonic probe 101 is less than the threshold Th2, the discrimination function 184 selects the second trained model 12a_1, which can derive the second discrimination result with better accuracy than the second trained model 12a_2. Then, the discrimination function 184 executes the second discrimination processing using the selected second trained model 12a_1 and the corresponding second learning condition 12b_1.

[0279] Furthermore, when the discrimination function 184 performs the third discrimination process in the third stage of operation, if the movement speed of the ultrasonic probe 101 is greater than the threshold Th1, it selects the third trained model 13a_2 as the third trained model to be used when executing the third discrimination process. Then, the discrimination function 184 executes the third discrimination process using the selected third trained model 13a_2 and the corresponding third learning condition 13b_2.

[0280] Furthermore, when the discrimination function 184 performs the third discrimination process in the third stage of operation, if the movement speed of the ultrasonic probe 101 is less than the threshold Th2, it selects the third trained model 13a_1 as the third trained model to be used when executing the third discrimination process. In this way, when the movement speed of the ultrasonic probe 101 is less than the threshold Th2, the discrimination function 184 selects the third trained model 13a_1, which can derive the third discrimination result with better accuracy than the third trained model 13a_2. Then, the discrimination function 184 executes the third discrimination process using the selected third trained model 13a_1 and the corresponding third learning condition 13b_1.

[0281] As described above, the discrimination function 184 performs a process to select a trained model and learning conditions to be used when deriving the discrimination result from among several types of trained models, according to the movement speed of the ultrasound probe 101. Therefore, the ultrasound diagnostic device 1 according to the modified example 4 can perform appropriate processing according to the movement speed of the ultrasound probe 101. Note that the process of selecting the trained model and learning conditions described above is an example of the third selection process.

[0282] (Modification 5 of the first and second embodiments) In the above-described modification 4, the case was explained in which the discrimination function 184 selects a trained model and learning conditions to be used when deriving the discrimination result from among several types of trained models, according to the movement speed of the ultrasonic probe 101. However, the discrimination function 184 may select a trained model to be used when deriving the discrimination result from among several types of trained models, depending on whether the discrimination result is to be derived in real time or as a post-processing step. Therefore, such an embodiment will be described as modification 5 of the first and second embodiments. In the description of modification 5, the differences from the above-described modifications 1 and 4 will be mainly explained, and the same configurations and processes as in modifications 1 and 4 may be omitted from the description.

[0283] In modified example 5, when a real-time processing instruction is input, the discrimination function 184 selects the first trained model 11a_2 as the first trained model to be used when executing the first discrimination process in the first stage of operation. Then, the discrimination function 184 executes the first discrimination process using the selected first trained model 11a_2 and the corresponding first learning condition 11b_2.

[0284] Furthermore, when a real-time processing instruction is input, the discrimination function 184 selects the second trained model 12a_2 as the second trained model to be used when executing the second discrimination process in the second stage of operation. The discrimination function 184 then executes the second discrimination process using the selected second trained model 12a_2 and the corresponding second training condition 12b_2.

[0285] Furthermore, when a real-time processing instruction is input, the discrimination function 184 selects the third pre-trained model 13a_2 as the third pre-trained model to be used when executing the third pre-trained model to perform

[0286] Furthermore, when a post-processing instruction is input, the discrimination function 184 selects the first trained model 11a_1 as the first trained model to be used when executing the first discrimination process in the first stage of operation. Then, the discrimination function 184 executes the first discrimination process using the selected first trained model 11a_1 and the corresponding first learning condition 11b_1.

[0287] Furthermore, when a post-processing instruction is input, the discrimination function 184 selects the second trained model 12a_1 as the second trained model to be used when executing the second discrimination process in the second stage of operation. The discrimination function 184 then executes the second discrimination process using the selected second trained model 12a_1 and the corresponding second training condition 12b_1.

[0288] Furthermore, when a post-processing instruction is input, the discrimination function 184 selects the third trained model 13a_1 as the third trained model to be used when executing the third discrimination process in the third stage of operation. The discrimination function 184 then executes the third discrimination process using the selected third trained model 13a_1 and the corresponding third training condition 13b_1.

[0289] Therefore, according to the ultrasound diagnostic device 1 according to the first embodiment and modification 5 of the second embodiment, appropriate processing can be performed depending on whether the discrimination result is derived in real time or as a post-processing step. Furthermore, according to the ultrasound diagnostic device 1 according to modification 5, discrimination results regarding characteristic areas can be obtained with high accuracy, similar to the first and second embodiments.

[0290] (Modification 6 of the first and second embodiments) Next, a modification to improve the discrimination accuracy of the first discrimination process using the first trained model 11a and the second discrimination process using the second trained model 12a will be described as Modification 6 of the first and second embodiments. In the description of Modification 6, the differences from the first and second embodiments will be mainly described, and the same configurations and processes as in the first and second embodiments may be omitted from the description.

[0291] For example, the second discrimination process based on the second trained model 12a and the second learning conditions 12b according to Modification 6 differs from the second discrimination process according to the first and second embodiments in that it determines whether the feature region included in the ultrasound image is malignant, benign, or neither malignant nor benign.

[0292] In Modification 6, during the second stage of training, the trained model generation function 201a or the trained model generation function 187 generates a second trained model 12a by learning the relationship between the ultrasound image data 20 and the label, and records the second training conditions 12b. In Modification 6, the label used when generating the second trained model 12a indicates whether the feature region included in the ultrasound image shown by the ultrasound image data is malignant, benign, or neither malignant nor benign.

[0293] Here, it is assumed that the ultrasound image data input to the second discrimination process includes characteristic areas. However, it is possible that over-detection may occur in the first discrimination process, and the over-detected ultrasound image data may be input to the second discrimination process.

[0294] Therefore, in modification 6, the second discrimination process determines whether the characteristic area included in the ultrasound image is malignant, benign, or neither malignant nor benign.

[0295] When it is determined by the second discrimination process that the feature site included in the ultrasonic image is neither malignant nor benign, it is conceivable that the ultrasonic image does not include the feature site. That is, when it is determined to be neither malignant nor benign, it is considered that the over-detected ultrasonic image data was input to the second discrimination process. That is, the ultrasonic image data determined to be neither malignant nor benign is considered to be the over-detected ultrasonic image data.

[0296] Therefore, the learned model generation function 201a or the learned model generation function 187 updates the first learned model 11a and the first learning condition 11b using the ultrasonic image data determined to be neither malignant nor benign. In the following description, the case where the learned model generation function 187 updates the first learned model 11a and the first learning condition 11b will be described. However, the learned model generation function 201a may execute the same process as the process executed by the learned model generation function 187 described below to update the first learned model 11a and the first learning condition 11b.

[0297] The learned model generation function 187 updates the first learned model 11a and the first learning condition 11b by associating and learning the ultrasonic image data determined to be neither malignant nor benign with the first label 15 indicating that the feature site is not included in the ultrasonic image. Thereby, the discrimination accuracy of the first discrimination process based on the first learned model 11a and the first learning condition 11b can be increased.

[0298] Also, for example, the third discrimination process based on the third learned model 13a and the third learning condition 13b according to Modification 6 is different from the third discrimination process according to the first embodiment and the second embodiment in that it discriminates whether the malignancy degree of the malignant feature site included in the ultrasonic image is malignancy degree A, malignancy degree B, or neither malignancy degree A nor malignancy degree B.

[0299] In Modification 6, during the third stage of learning, the trained model generation function 201a or the trained model generation function 187 generates a third trained model 13a by learning the relationship between the ultrasound image data 25 and the label, and records the third learning condition 13b. In Modification 6, the label used when generating the third trained model 13a indicates whether the malignancy of the malignant feature site included in the ultrasound image shown by the ultrasound image data is malignancy grade A, malignancy grade B, or neither malignancy grade A nor malignancy grade B.

[0300] Here, it is assumed that the feature regions included in the ultrasound image data input to the third discrimination process are malignant. However, in the second discrimination process, a benign feature region may be mistakenly identified as malignant, and ultrasound image data that has been mistakenly identified as malignant may be input to the third discrimination process.

[0301] Therefore, in modification 6, the third discrimination process determines whether the malignancy of the malignant characteristic site included in the ultrasound image is malignancy grade A, malignancy grade B, or neither malignancy grade A nor B.

[0302] If the third discrimination process determines that the malignancy of a feature area included in the ultrasound image is neither malignancy grade A nor malignancy grade B, then the feature area included in the ultrasound image is considered benign. In other words, if it is determined that it is neither malignancy grade A nor malignancy grade B, then it is considered that ultrasound image data in which a benign feature area was mistakenly identified as malignant was input to the third discrimination process. That is, ultrasound image data in which it is determined that it is neither malignancy grade A nor malignancy grade B is considered to be ultrasound image data in which a benign feature area was mistakenly identified as malignant.

[0303] Therefore, the trained model generation function 201a or the trained model generation function 187 updates the second trained model 12a and the second learning conditions 12b using ultrasound image data that has been determined to be neither malignancy grade A nor malignancy grade B. The following description will explain the case in which the trained model generation function 187 updates the second trained model 12a and the second learning conditions 12b. However, the trained model generation function 201a may update the second trained model 12a and the second learning conditions 12b by performing the same processing as the trained model generation function 187 described below.

[0304] The trained model generation function 187 updates the second trained model 12a and the second learning conditions 12b by learning to associate ultrasound image data that has been determined to be neither malignant grade A nor malignant grade B with labels indicating that the feature area contained in the ultrasound image is benign. This makes it possible to improve the discrimination accuracy of the second discrimination process based on the second trained model 12a and the second learning conditions 12b.

[0305] Modification 6 has been described. As described above, the trained model generation function 187 in Modification 6 generates a new trained model corresponding to one stage by associating ultrasonic image data corresponding to a predetermined discrimination result output from the discrimination process corresponding to the next stage of one stage with this predetermined discrimination result, and records the learning conditions anew. This makes it possible to increase the discrimination accuracy of the discrimination process corresponding to one stage.

[0306] (Modification 7 of the first and second embodiments) Next, other modifications for improving the discrimination accuracy of the first discrimination process using the first trained model 11a and the second discrimination process using the second trained model 12a will be described as Modification 7 of the first and second embodiments. In the description of Modification 7, the differences from the first and second embodiments will be mainly described, and the same configurations and processes as in the first and second embodiments may be omitted.

[0307] In the above-described modification 6, the case in which the second trained model 12a and the third trained model 13a are generated by a different learning method than that used in the first and second embodiments was explained. On the other hand, in modification 7, the second trained model 12a and the third trained model 13a generated by the same learning method as in the first and second embodiments are used.

[0308] However, in Modification 7, the fourth pre-trained model, the fourth training condition, the fifth pre-trained model, and the fifth training condition are used. In Modification 7, in the second stage of operation, the fourth discrimination process, which uses the fourth pre-trained model and the fourth training condition, is executed before the second discrimination process. Also, in the third stage of operation, the fifth discrimination process, which uses the fifth pre-trained model and the fifth training condition, is executed before the third discrimination process.

[0309] In the second stage of operation, the discrimination function 184 performs a fourth discrimination process as CAD processing using the fourth trained model and the fourth learning conditions. The fourth discrimination process determines whether or not a feature area is included in the ultrasound image shown in the ultrasound image data 22, and derives a discrimination result (fourth discrimination result). Thus, in the modified example 7, the discrimination function 184 performs a fourth discrimination process in the second stage of operation that is the same as the first discrimination process performed in the first stage of operation. Furthermore, the fourth trained model is generated in the same way as the first trained model 11a. Also, the fourth learning conditions are recorded in the same way as the first learning conditions 11b.

[0310] However, when generating the fourth pre-trained model, more over-detected ultrasound image data is used in the first discrimination process than when generating the first pre-trained model 11a.

[0311] Therefore, in the fourth discrimination process using the fourth trained model and the fourth training conditions, over-detection of ultrasound image data that was over-detected in the first discrimination process is suppressed.

[0312] Here, we will explain the case where the discrimination function 184, which has performed the fourth discrimination process, determines that the ultrasound image contains a feature area. In this case, in the second stage of operation, the discrimination function 184 uses the second trained model 12a and the second training condition 12b to perform the second discrimination process on the ultrasound image data in which it has been determined that the ultrasound image contains a feature area.

[0313] Next, we will describe the case where the discrimination function 184, which has performed the fourth discrimination process, determines that the ultrasound image does not contain any feature regions. In this case, the trained model generation function 201a or the trained model generation function 187 updates the first trained model 11a and the first learning conditions 11b using the ultrasound image data in which the ultrasound image has been determined not to contain any feature regions. In the following description, we will describe the case in which the trained model generation function 187 updates the first trained model 11a and the first learning conditions 11b. However, the trained model generation function 201a may update the first trained model 11a and the first learning conditions 11b by performing the same processing as the trained model generation function 187 described below.

[0314] The trained model generation function 187 updates the first trained model 11a and the first learning conditions 11b by learning by associating ultrasound image data that has been determined not to contain feature regions with a first label 15 that indicates that the ultrasound image does not contain feature regions. This makes it possible to improve the discrimination accuracy of the first discrimination process based on the first trained model 11a and the first learning conditions 11b.

[0315] Then, in the third stage of operation, the discrimination function 184 executes the fifth discrimination process using the fifth trained model and the fifth learning conditions. The fifth discrimination process is a process that determines whether the feature region, which was determined to be benign by the second discrimination process in the second stage of operation, is benign or malignant, and derives a discrimination result (the fifth discrimination result). Thus, in the modified example 7, the discrimination function 184 executes the fifth discrimination process in the third stage of operation, which is the same as the second discrimination process executed in the second stage of operation. In addition, the fifth trained model is generated in the same way as the second trained model 12a. Also, the fifth learning conditions are recorded in the same way as the second learning conditions 12b.

[0316] However, when generating the fifth pre-trained model, more ultrasound image data in which benign feature regions were mistakenly identified as malignant are used in the second discrimination process than when generating the second pre-trained model 12a.

[0317] Therefore, in the fifth discrimination process using the fifth trained model and the fifth learning conditions, feature regions included in ultrasound image data that were incorrectly identified as malignant in the second discrimination process can be correctly identified as benign.

[0318] Here, we will explain the case where the discrimination function 184, which has performed the fifth discrimination process, determines that the feature region included in the ultrasound image is malignant. In this case, in the third stage of operation, the discrimination function 184 uses the third trained model 13a and the third training condition 13b to perform the third discrimination process on the ultrasound image data in which the feature region has been determined to be malignant.

[0319] On the other hand, the case in which the discrimination function 184, which has performed the fifth discrimination process, has determined that the feature region included in the ultrasound image is benign will be described. In this case, the trained model generation function 201a or the trained model generation function 187 updates the second trained model 12a and the second learning conditions 12b using the ultrasound image data in which the feature region has been determined to be benign. The following description will explain the case in which the trained model generation function 187 updates the second trained model 12a and the second learning conditions 12b. However, the trained model generation function 201a may update the second trained model 12a and the second learning conditions 12b by performing the same processing as the trained model generation function 187 described below.

[0320] The trained model generation function 187 updates the second trained model 12a and the second learning conditions 12b by learning by associating ultrasound image data in which the feature region has been determined to be benign with a second label 21 indicating that the feature region is benign. This makes it possible to improve the discrimination accuracy of the second discrimination process based on the second trained model 12a and the second learning conditions 12b.

[0321] Modification 7 has been described. As described above, the trained model generation function 187 in Modification 7 generates a new trained model corresponding to one stage by associating ultrasonic image data corresponding to a predetermined discrimination result output from the discrimination process corresponding to the next stage of one stage with this predetermined discrimination result, and records the learning conditions anew. This makes it possible to improve the discrimination accuracy of the discrimination process corresponding to one stage.

[0322] (Third embodiment) In the embodiments and modifications described above, the case in which the ultrasound diagnostic device 1 derives the discrimination result using a trained model and training conditions was explained. However, the medical image processing device may perform the same processing as post-processing rather than in real time. Therefore, such an embodiment will be described as a third embodiment.

[0323] Figure 24 shows an example configuration of a medical image processing apparatus 300 according to a third embodiment. As illustrated in Figure 24, the medical image processing apparatus 300 comprises an input device 301, a display 302, a storage circuit 310, and a processing circuit 320. The input device 301, the display 302, the storage circuit 310, and the processing circuit 320 are connected to each other so as to be able to communicate with each other. The medical image processing apparatus 300 is an example of an analysis device.

[0324] The input device 301 can be implemented using a mouse, keyboard, buttons, panel switches, touch command screen, foot switch, trackball, joystick, etc. The input device 301 receives various setting requests from the operator of the medical image processing device 300. The input device 301 outputs the received setting requests to the processing circuit 320. For example, the input device 301 receives instructions (execution instructions) from the operator of the medical image processing device 300 to perform CAD processing and outputs the received execution instructions to the processing circuit 320. The operator can also set the ROI, which is the search range for feature areas, on the ultrasound image during CAD processing via the input device 301.

[0325] The display 302 can, for example, display medical images or a GUI for the operator to input various setting requests using the input device 301.

[0326] The memory circuit 310 stores various programs for displaying the GUI and information used by those programs. The memory circuit 310 also stores multiple ultrasound image data generated by the ultrasound diagnostic device 1 in a time-series format. Furthermore, the memory circuit 310 stores various trained models and training conditions from one of the first embodiment, the second embodiment, or modifications 1 to 7 described above. These trained models and training conditions are used in the processing executed by the processing circuit 320.

[0327] The processing circuit 320 controls the entire processing of the medical image processing device 300. The processing circuit 320 is implemented, for example, by a processor. For example, as shown in Figure 24, the processing circuit 320 has the following processing functions: control function 321, velocity detection function 322, acquisition function 323, trained model generation function 324, discrimination function 325, marker information generation function 326, and display control function 327. Here, for example, the processing functions of the processing circuit 320 shown in Figure 24, namely the control function 321, velocity detection function 322, acquisition function 323, trained model generation function 324, discrimination function 325, marker information generation function 326, and display control function 327, are recorded in the memory circuit 310 in the form of a program that can be executed by a computer. The processing circuit 320 is a processor that reads each program from the memory circuit 310 and executes it to realize the function corresponding to each program. In other words, the processing circuit 320, when each program has been loaded, will have the functions shown in the processing circuit 320 in Figure 24.

[0328] The medical image processing device 300 acquires, for example, multiple ultrasound image data in a time series generated by the ultrasound diagnostic device 1. However, this ultrasound image data is accompanied by a detection signal indicating the movement speed of the ultrasound probe 101 detected by the velocity detector 104. The medical image processing device 300 then processes the ultrasound image data in the same manner as in the first embodiment, the second embodiment, or any of the modifications 1 to 7 described above, and displays the ultrasound image and information such as detected feature areas on the display 302.

[0329] The control function 321 controls the entire processing of the medical image processing device 300. The velocity detection function 322 has the same function as the velocity detection function 182 described above. For example, the velocity detection function 322 detects the movement speed of the ultrasound probe 101 based on a detection signal added to the ultrasound image data. The acquisition function 323 has the same function as the acquisition function 183 described above. The trained model generation function 324 has the same function as the trained model generation function 187 described above. The discrimination function 325 has the same function as the discrimination function 184 described above. The marker information generation function 326 performs the same function as the marker information generation function 185 described above. The display control function 327 performs the same function as the display control function 186 described above.

[0330] The medical image processing apparatus 300 according to the third embodiment has been described above. According to the medical image processing apparatus 300 according to the third embodiment, similar to the first embodiment, discrimination results regarding feature areas can be obtained with high accuracy.

[0331] In addition, in at least one embodiment or modification described above, the ultrasound diagnostic device 1 may treat each function of the processing circuit 180, various trained models, and various learning conditions as a single software package. Similarly, the medical image processing device 300 may treat each function of the processing circuit 320, various trained models, and various learning conditions as a single software package.

[0332] According to at least one embodiment or modification described above, it is possible to obtain a discrimination result regarding characteristic parts with high accuracy.

[0333] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents. [Explanation of Symbols]

[0334] 1. Ultrasound diagnostic equipment 184 Discrimination function 187 Pre-trained model generation function

Claims

1. A movement speed detection unit detects the movement speed of the ultrasound probe moving within the subject, A control unit performs a first detection process to detect feature regions from an ultrasonic image based on ultrasound received by the ultrasonic probe when the movement speed of the ultrasonic probe is greater than a predetermined speed, and performs the first detection process when the movement speed of the ultrasonic probe is less than the predetermined speed, and also detects over-detected feature regions from among the feature regions detected by the first detection process, and detects the feature regions excluding the over-detected feature regions from among the feature regions detected in the first detection process as the final feature regions. A display unit that displays identification information for identifying the characteristic area together with the ultrasound image, Equipped with, The control unit performs a first detection process to detect a portion corresponding to one of a plurality of patterns at each position in the ultrasound image, detects a portion that is commonly detected as a portion corresponding to one of the plurality of patterns by the first detection process at each position in the ultrasound image as the over-detected feature portion, and performs a second detection process to detect the feature portions excluding the over-detected feature portions from among the feature portions detected in the first detection process as the final feature portion. Ultrasound diagnostic equipment.

2. The control unit acquires the ultrasound image at a frame rate predetermined for the movement speed of the ultrasound probe. The ultrasound diagnostic apparatus according to claim 1.

3. The frame rate is set to be higher as the movement speed of the ultrasonic probe increases. The ultrasound diagnostic apparatus according to claim 2.

Images