Image processing device, endoscope system, and image processing method

Abstract

[Problem] To provide an image processing device 10 that selects a model suitable for a captured scene and performs processing. [Solution] An image processing device 10 is provided with: a scene determination unit 11 that determines a scene for capturing a primary image P1 acquired by a camera unit 7 of an endoscope device 9; a storage unit 12 that stores a plurality of trained models, each performing predetermined processing; a selection unit 13 that selects one model from the plurality of models on the basis of a determination result obtained by the scene determination unit 11; and an image processing unit 14 that performs processing on the primary image P1 using the selected one model.

Description

Image processing device, endoscope system, and image processing method

[0001] The present invention relates to an image processing device for processing medical images, an endoscope system, and an image processing method for processing medical images.

[0002] The images acquired by the medical device are processed, and the processed images are displayed on a monitor.

[0003] Japanese Patent Application Publication No. 2020-39851 discloses an image processing device that selects a trained model suitable for imaging conditions such as the imaging area and performs AI processing.

[0004] International Publication No. 2018 / 159083 discloses an endoscope system that determines the shooting scene of an image output by an endoscope and sets processing conditions suitable for the shooting scene.

[0005] JP 2020-39851 A International Publication No. 2018 / 159083

[0006] An embodiment of the present invention aims to provide an image processing device that selects a model suitable for a shooting scene and performs AI processing, an endoscopic system that selects a model suitable for a shooting scene and performs AI processing, and an image processing method that selects a model suitable for a shooting scene and performs AI processing.

[0007] The image processing device of the embodiment includes a scene determination unit that determines the shooting scene of a primary image acquired by a camera unit of an endoscopic device, a memory unit that stores multiple trained models, each of which performs a predetermined processing, a selection unit that selects one model from the multiple models based on the determination result by the scene determination unit, and an image processing unit that processes the primary image using the selected one model.

[0008] The image processing method of the embodiment determines the shooting scene of a primary image acquired by a camera unit of an endoscopic device, and based on the determination result, selects one model from multiple trained models, each of which performs a predetermined processing, and processes the primary image using the selected model.

[0009] According to an embodiment of the present invention, it is possible to provide an image processing device that selects a model suitable for a shooting scene and performs AI processing, an endoscopic system that selects a model suitable for a shooting scene and performs AI processing, and an image processing method that selects a model suitable for a shooting scene and performs AI processing.

[0010] Fig. 1 is a configuration diagram of an endoscope system according to a first embodiment. Fig. 2 is a flowchart of an image processing method for the endoscope system according to the first embodiment. Fig. 3 is a configuration diagram of an endoscope system according to a second embodiment. Fig. 4 is a flowchart of an image processing method for the endoscope system according to the second embodiment. Fig. 5 is a flowchart of an image processing method for the endoscope system according to a fourth embodiment. Fig. 6 is a configuration diagram of an endoscope system according to a fifth embodiment.

[0011] 1 includes an endoscope device 9, which is a medical device, an image processing device 10, and a monitor 20. In the following description, drawings based on the embodiment are schematic, and some components are not shown or labeled.

[0012] The endoscopic device 9 includes an endoscope 8, a light source 3 that illuminates a subject, and a treatment tool 2. The endoscope 8 includes a camera unit 7 and an operation unit 4 on which buttons and the like that are operated by a user are arranged. The endoscope 8 may be a flexible endoscope or a rigid endoscope.

[0013] The light source 3 generates illumination light to be supplied to the endoscope 8. For example, the illumination light is white light, narrow band light, and red light. That is, the endoscope device 9 performs white light imaging (WLI), narrow band imaging (NBI), and red dichromatic imaging (RDI).

[0014] In WLI, the subject is observed using white light. In NBI, a specific narrowband light that is easily absorbed by hemoglobin is used to highlight and clearly display capillaries in the superficial mucosa. In RDI, amber light is used in combination with green and red light to make it easier to see blood vessels deep within the mucosa.

[0015] Although not shown, the illumination light generated by the light source 3 is emitted from the tip of the elongated insertion portion of the endoscope 8 to illuminate the subject. The light source 3 may be included in the image processing device 10.

[0016] The camera unit 7 includes an optical system 5 that focuses a subject image on an image sensor 6, and an image sensor 6 that converts the subject image into an electrical signal and outputs a primary image P1. The primary image P1 is an image for one frame of a moving image, and for example, 30 primary images P1 are transmitted per second to the image processing device 10. When the light source 3 is of the frame sequential type, the primary image P1 is a set of images consisting of an R image, a G image, and a B image.

[0017] The image processing device 10 includes a scene determination unit 11, a storage unit 12, a selection unit 13, an image processing unit 14, a CPU 15, an input unit 16, an output unit 18, and an acquisition unit 19. The CPU 15 controls the entire image processing device 10 and the light source 3.

[0018] The acquisition unit 19 acquires a primary image P1 from the endoscope device 9. The scene determination unit 11 determines the captured scene of the primary image P1. The storage unit 12 stores a plurality of trained models, each of which performs a predetermined AI processing. The selection unit 13 selects one model from the plurality of models based on the determination result by the scene determination unit 11. The image processing unit 14 performs AI processing on the primary image P1 using the selected model. The output unit 18 outputs a secondary image P2, which has been subjected to AI processing by the image processing unit 14, to the monitor 20.

[0019] The storage unit 12 is a RAM, a hard disk drive (HDD), a solid state drive (SSD), an optical disk, etc. The storage unit 12 may be a so-called external device separate from the image processing device 10. Data may be transferred to the storage unit 12 from a server connected to the image processing device 10 via an internet line or the like.

[0020] At least one of the components of the image processing device 10 may be configured with an internal circuit of a semiconductor device operated by software, a dedicated hardware circuit, or may include an internal circuit of a semiconductor device and a dedicated hardware circuit. For example, the image processing unit 14 may include an FPGA (Field Programmable Gate Array) including a hardware circuit and a GPU (Graphics Processing Unit) which is a software circuit.

[0021] The monitor 20 is, for example, a liquid crystal display that displays images and the like from the image processing device 10. The monitor 20 having a touch panel function may also have the function of the input unit 16 through which the user inputs instructions.

[0022] As described above, the endoscopic system 1 of this embodiment includes an endoscopic device 9 having a camera unit 7 that acquires a primary image P1 of a subject, and an image processing device 10. The image processing device 10 includes a scene determination unit 11 that determines the captured scene of the primary image P1, a memory unit 12 that stores a plurality of trained models, each of which performs a predetermined processing, a selection unit 13 that selects one model from the plurality of models based on the determination result by the scene determination unit 11, and an image processing unit 14 that processes the primary image P1 using the one selected model.

[0023] The endoscope system 1 and the image processing device 10 can select a model suitable for the shooting scene and perform appropriate AI processing.

[0024] <Image Processing Method> The image processing method will be described with reference to the flowchart in FIG.

[0025] <Step S10> Image Input When processing by the endoscope device 9, which is a medical device, is started, a primary image P1 from the endoscope device 9 is input to the image processing device 10. The primary image P1 is, for example, a RAW image.

[0026] <Step S30> Scene Determination The scene determination unit (determination circuit) 11 determines the photographed scene of the primary image P1.

[0027] The imaging scene is either a screening scene, a detailed examination scene, or a treatment scene. During screening, the user observes the entire object of observation while making sure not to overlook any lesions. During detailed examination, the user closely observes the condition of any lesions that have been discovered. During treatment, the user uses a treatment tool to sample the lesions, etc.

[0028] The photographed scene is determined by AI using a model trained using training images that have been annotated with, for example, the presence or absence of treatment tools, the presence or absence of pigment, the magnification state, and the presence or absence of bleeding for each scene.

[0029] A teacher image that includes a treatment tool such as forceps is annotated as a treatment scene. A teacher image that has a stain applied to color the object being observed is annotated as a detailed examination scene. Images of the inside of the body that are not detailed examination or treatment scenes are annotated as a screening scene.

[0030] <Step S40> First Model Selection A plurality of trained models, each of which performs a predetermined AI process, are stored in the storage unit 12. Specifically, the selection unit 13 selects one model from three models corresponding to each of the screening scene, the examination scene, and the treatment scene, based on the determination result by the scene determination unit 11.

[0031] A selection unit (selection circuit) 13 selects the first model when the photographed scene is determined to be a screening scene.

[0032] The first model performs noise reduction processing or first high-resolution processing. The first model for performing noise reduction processing is trained using, as teacher images, a noisy original image and an image obtained by appropriately removing noise from the original image. The first model for performing noise reduction processing may also be trained using, as teacher images, an image with little noise and an image obtained by adding noise to an image with little noise. Furthermore, the first high-resolution processing is trained using, as teacher images, an image with normal resolution and a high-resolution image. The first high-resolution processing is a super-resolution processing, and differs from the second high-resolution processing described below in characteristics such as the frequency band for improving resolution.

[0033] <Step S50> Selecting a Second Model When the photographed scene is determined to be a close inspection scene, the selection unit 13 selects the second model A or the second model B.

[0034] The second model A performs the second high-resolution processing, and the second model B performs the first color conversion processing. Which processing to perform is set in advance by the user, for example.

[0035] The second high-resolution processing is a specialized super-resolution processing that makes it easier to observe detailed blood vessels, mucosal structures, etc. in the lesion area. The second model A that performs the second high-resolution processing learns using, as training images, images with normal resolution and high-resolution images in which the detailed configuration of blood vessels and mucosal structures is emphasized more than in the normal-resolution images.

[0036] The first color conversion process performed by the second model B is a specialized color conversion process that makes it easier to observe the detailed structure of the lesion area. The second model B learns using images before and after the color conversion process as training images. Note that the first color conversion process may be the same as the third color conversion process described below (a process that converts the color of an image illuminated by a WLI light source into a pseudo-color illuminated by an NBI light source).

[0037] <Step S60> Selecting a Third Model When the imaging scene is determined to be a treatment scene, the selection unit 13 selects the third model A or the third model B. The third model A performs the second high-resolution processing. The third model B performs the second color conversion processing.

[0038] The second high resolution processing and the second color conversion processing are each specialized processing for making it easier to observe the detailed structure of the area under treatment.

[0039] The second high-resolution processing is a specialized super-resolution processing that makes it easier to observe the detailed blood vessels, mucosal structures, etc., of the area under treatment. The third model A that performs the second high-resolution processing learns using as training images a normal-resolution image and a high-resolution image in which the detailed configuration of the blood vessels and mucosal structures is emphasized more than in the normal-resolution image.

[0040] The second color conversion process performs specialized color conversion to make it easier to observe the detailed structure of the area being treated. The third model B that performs the second color conversion process learns using images before and after the color conversion process as training images. The second color conversion process may be the same as the fourth color conversion process described below (a process that converts the color of an image illuminated by a WLI light source into a pseudo-color illuminated by an RDI light source).

[0041] <Step S80> AI Processing The image processing unit 14 processes the primary image P1 using the selected model and outputs the secondary image P2. For example, the image processing unit 14, which is made up of a GPU, performs AI processing using parameters of the model stored in the storage unit 12.

[0042] <Step S90> Image Display The secondary image P2 that has been processed by the image processing unit 14 is displayed on the monitor 20.

[0043] <Step S100> End The process from step S10 is repeated until the process is completed. Note that the image processing device 10 does not need to determine the photographic scene of all primary images P1. For example, the image processing device 10 may determine the photographic scene when the luminance of the primary image P1 changes significantly or at predetermined time intervals.

[0044] As described above, the image processing method of this embodiment determines the shooting scene of a primary image acquired by the camera unit of an endoscopic device, and based on the determination result, selects one model from multiple trained models, each of which performs a predetermined processing, and processes the primary image using the selected model.

[0045] The image processing method of this embodiment can select a model suitable for the photographed scene and perform appropriate AI processing.

[0046] Second Embodiment The embodiment described below is similar to the first embodiment, and therefore, components having the same functions as those in the first embodiment are denoted by the same reference numerals as those in the first embodiment, and descriptions thereof will be omitted.

[0047] In the endoscope system 1A of this embodiment shown in FIG. 3, an image processing device 10A further includes a preprocessing unit 17, as compared with the endoscope system 1 of the first embodiment.

[0048] The preprocessing unit 17 performs preprocessing on the primary image P1 of the subject captured by the camera unit 7 of the endoscope device 9, and outputs a preprocessed image P1A.

[0049] The scene determination unit 11 determines the captured scene of the preprocessed image P1A. The memory unit 12 stores a plurality of trained models, each of which performs a predetermined AI processing. The selection unit 13 selects one model from the plurality of models based on the determination result by the scene determination unit 11. The image processing unit 14 performs image processing on the preprocessed image P1A using the selected model and outputs a secondary image P2. <Image Processing Method> The image processing method of the endoscope system 1A will be described with reference to the flowchart in Figure 4.

[0050] <Step S10> Image Input When processing by the endoscope device 9, which is a medical device, is started, a primary image P1 from the endoscope device 9 is input to the image processing device 10. The primary image P1 is, for example, a RAW image.

[0051] <Step S20> Preprocessing The preprocessing unit (preprocessing circuit) 17 of the image processing device 10 performs preprocessing on the primary image P1 to generate a preprocessed image P1A.

[0052] The pre-processing is, for example, at least one of demosaicing processing, light control detection processing, white balance processing, and defective pixel correction.

[0053] <Step S30> Scene Determination The scene determination unit (determination circuit) 11 determines the photographic scene of the preprocessed image P1A. The method of determining the photographic scene is substantially the same as in the first embodiment, and determines whether the scene is a screening scene, a detailed examination scene, or a treatment scene.

[0054] The scene determination unit 11 may make the determination based on at least one of the distance between the camera unit and the subject, the zoom state of the preprocessed image P1A, and the proportion of the lesion area in the preprocessed image P1A.

[0055] The distance D between the camera unit 7 and the subject can be easily obtained, for example, when the camera unit 7 is capable of capturing stereoscopic images. For example, when the distance D is less than 15 mm, the image is determined to be a close-up scene. Regarding the zoom state of the image, when a user operates the operation unit 4 of the endoscope 8 to perform magnified photography using the zoom function of the optical system 5 of the camera unit 7, the image is determined to be a close-up scene based on the relative position data of the focus lens. Regarding the proportion of the lesion area in the image, when the color of the lesion area is different from that of the normal area, for example, when the color of the lesion area accounts for 30% or more of the entire image, the image is determined to be a close-up scene. The lesion area can be detected by AI using a model trained using training images containing the lesion area, or by detecting color differences from the surrounding area.

[0056] Images that are not screening scenes, examination scenes, or treatment scenes, i.e., in vitro images, may be annotated as teacher images and used for learning, or may be determined as unknown images by the scene determination unit 11. When the scene determination unit 11 performs AI determination, the determination may be performed using the primary image P1.

[0057] <Step S40> First Model Selection A plurality of trained models, each of which performs a predetermined AI process, are stored in the storage unit 12. Specifically, the selection unit 13 selects one model from three models corresponding to each of the screening scene, the examination scene, and the treatment scene, based on the determination result by the scene determination unit 11.

[0058] A selection unit (selection circuit) 13 selects the first model when the photographed scene is determined to be a screening scene.

[0059] The first model performs noise reduction processing or first high-resolution processing, which processing is to be performed being set in advance by the user, for example.

[0060] <Step S50> Selecting a Second Model When the photographed scene is determined to be a close inspection scene, the selection unit 13 selects a second model.

[0061] The second model performs either the second high-resolution processing or the first color conversion processing, which processing is selected in advance by the user, for example.

[0062] <Step S60> Selecting a Third Model When the imaging scene is determined to be a treatment scene, the selector 13 selects a third model. The third model performs a second high-resolution process or a second color conversion process.

[0063] <Step S70> Unknown, Extracorporeal Scene If the scene determination unit 11 is unable to determine the photographed scene, the preprocessed image P1A is output instead of the secondary image P2.

[0064] <Step S80> AI Processing The image processing unit 14 performs image processing on the primary image P1 using the selected model. For example, the image processing unit 14, which is made up of a GPU, performs AI processing using parameters of the model stored in the storage unit 12.

[0065] <Step S90> Image Display The secondary image P2 or the preprocessed image P1A is displayed on the monitor 20.

[0066] <Step S100> End The process from step S10 is repeated until the process is completed. Note that the image processing device 10 does not need to determine the photographic scene of all primary images P1. For example, the image processing device 10 may determine the photographic scene when the luminance of the primary image P1 changes significantly or at predetermined time intervals.

[0067] The embodiment described below is similar to and has the same effects as the first or second embodiment, and therefore components having the same functions as those in the first or second embodiment are denoted by the same reference numerals as those in the first or second embodiment, and descriptions thereof will be omitted.

[0068] As already explained, the light source 3 generates illumination light to be supplied to the endoscope 8. For example, the illumination light may be white light, narrow band light, or red light.

[0069] The storage unit 12 of the image processing device 10B of the endoscope system 1B of this embodiment stores nine models corresponding to three types of illumination light and three types of scenes. The selection unit 13 selects a model corresponding to the illumination light of the primary image P1.

[0070] That is, the storage unit 12 stores the following nine models: First model A: screening scene, white light First model B: screening scene, narrow band light First model C: screening scene, red light Second model A: inspection scene, white light Second model B: inspection scene, narrow band light Second model C: inspection scene, red light Third model A: treatment scene, white light Third model B: treatment scene, narrow band light Third model C: treatment scene, red light

[0071] The image processing device 10B and endoscope system 1B of this embodiment can select a model suitable for the illumination light and the shooting scene and perform appropriate AI processing.

[0072] <Fourth embodiment> The selection unit 13 of the image processing device 10C of the endoscope system 1C of this embodiment selects a model different from the already selected model based on an instruction from the user using, for example, the input unit 16 or a button on the operation unit 4.

[0073] The selection unit 13 selects a fourth model that performs the third color conversion process based on a color conversion process instruction from the user even if the scene determination unit 11 determines that the scene is an examination scene. Also, the selection unit 13 selects a fifth model that performs the fourth color conversion process based on a color conversion process instruction from the user even if the scene determination unit 11 determines that the scene is a treatment scene.

[0074] The image processing method by the image processing device 10C will be described with reference to the flowchart of Fig. 5. <Steps S10 to S50, S60, S70> Since this is the same as the image processing device 10A shown in Fig. 4, the description will be omitted.

[0075] <Steps S52, S54> Even if the scene determination unit 11 determines that the scene is a close inspection scene, if the user instructs the selection unit 13 to perform color conversion processing (S52, YES), the selection unit 13 selects a fourth model to perform the third color conversion processing based on the instruction.

[0076] Illumination with an NBI light source makes it easier to observe fine structures. The third color conversion process converts the secondary image P2 illuminated with a WLI light source into colors that are similar to those illuminated with an NBI light source.

[0077] <Steps S62, S64> Even if the scene determination unit 11 determines that the scene is a treatment scene, if there is an instruction for color conversion processing from the user (S62, YES), the selection unit 13 selects a fifth model for performing the fourth color conversion processing.

[0078] Illumination with an RDI light source makes it easier to observe blood vessels and bleeding areas deep in the mucosa, etc. The fourth color conversion process converts the secondary image P2 illuminated with a WLI light source into a color that simulates illumination with an RDI light source.

[0079] The image processing device 10C performs more appropriate AI processing based on instructions from the user.

[0080] 6, an endoscope system 1D of this embodiment includes an endoscope processor 30 installed in the same location as an endoscope device 9 and a monitor 20. On the other hand, an image processing device 10D installed in a location separate from the endoscope device 9 and the monitor 20 is a component of a server connected to the endoscope processor 30 via an internet line or the like.

[0081] The endoscope processor 30 has a transmitting unit 31 that transmits the primary image P1 input from the endoscope device 9, and a receiving unit 32 that receives the secondary image P2 or the preprocessed image P1A from the image processing device 10D.

[0082] The image processing device 10D includes an acquisition unit 19 that receives the primary image P1 from the endoscope processor 30, and an output unit 18 that transmits the secondary image P2 or the preprocessed image P1A.

[0083] As in the first embodiment, the scene determination unit 11 performs scene determination on the primary image P1 of the subject, the selection unit 13 selects one model from multiple models, and the image processing unit 14 performs image processing using the selected model.

[0084] The endoscope processor 30 may have the same input unit 16 as the image processing device 10, and the image processing device 10D may select a model different from the model that has already been selected based on an instruction from the user.

[0085] The image processing device 10D may not have the pre-processing unit 17, as in the image processing device 10. Although not shown, the endoscope processor 30 may have the pre-processing unit 17 and transmit the pre-processed image P1A to the image processing device 10D.

[0086] The endoscope processor 30 of the endoscope system 1D has a simple configuration due to its low load, is inexpensive, and has high versatility. Furthermore, it is easy for the server administrator to perform version upgrades, etc.

[0087] The ranges of the numerical values ​​described above are not limited to the ranges described above and can be increased or decreased as appropriate. Furthermore, the present invention is not limited to the above-described embodiments, and various changes and modifications can be made within the scope of the present invention.

[0088] DESCRIPTION OF SYMBOLS 1, 1A-1D... Endoscope system 2... Treatment tool 3... Light source 4... Operation unit 5... Optical system 6... Image sensor 7... Camera unit 8... Endoscope 9... Endoscope device 10, 10A-10D... Image processing device 11... Scene determination unit 12... Memory unit 13... Selection unit 14... Image processing unit 15... CPU 16... Input unit 17... Pre-processing unit 20... Monitor 30... Endoscope processor

Claims

1. An image processing device comprising: a scene determination unit that determines the captured scene of a primary image acquired by a camera unit of an endoscope device; a memory unit that stores multiple trained models, each of which performs a predetermined processing; a selection unit that selects one model from the multiple models based on the determination result by the scene determination unit; and an image processing unit that processes the primary image using the selected one model.

2. The image processing device described in claim 1, characterized in that the selection unit selects a first model that performs noise reduction processing or first high-resolution processing when the captured scene is determined to be a screening scene, and selects a second model that performs second high-resolution processing or first color conversion processing when the captured scene is determined to be a scrutiny scene.

3. The image processing device according to claim 2, characterized in that the selection unit selects a third model that performs the second high-resolution processing or the second color conversion processing when the scene is determined to be a treatment scene.

4. The image processing device described in claim 3, characterized in that the endoscopic device has a light source that generates multiple illumination lights, the memory unit stores the multiple models corresponding to each of the multiple illumination lights, and the selection unit selects the model corresponding to the illumination light of the primary image.

5. The image processing device according to claim 4, wherein the plurality of illumination lights are white light, narrow band light, and red light.

6. The image processing device according to claim 4, wherein the scene determination unit makes a determination based on at least one of the presence or absence of a treatment tool, the presence or absence of a pigment, the magnification state, and the presence or absence of bleeding.

7. The image processing device described in claim 1, further comprising a pre-processing unit that performs pre-processing on the primary image and outputs a pre-processed image, wherein the scene determination unit determines the captured scene of the pre-processed image, the memory unit stores multiple trained models, each of which performs a predetermined AI processing, and the image processing unit performs the AI ​​processing on the pre-processed image using the selected one of the models and outputs a secondary image.

8. The image processing device described in claim 7, characterized in that the scene determination unit determines that the scene is an examination scene based on at least one of the distance between the camera unit and the subject, the zoom state of the preprocessed image, and the proportion of the lesion area in the preprocessed image.

9. The image processing device according to claim 8, characterized in that, when the scene determination section is unable to determine the photographed scene, the preprocessed image is output instead of the secondary image.

10. The image processing device according to claim 1, wherein the selection unit selects a model different from the already selected model based on an instruction from the user.

11. The image processing device described in claim 5, characterized in that the selection unit selects a fourth model that performs a third color conversion process based on a color conversion process instruction from a user, even if the scene determination unit determines that the scene is the inspection scene.

12. The image processing device according to claim 11, characterized in that the third color conversion process converts the primary image captured with the white light into a secondary image captured with the narrow-band light in a pseudo manner.

13. The image processing device described in claim 5, characterized in that the selection unit selects a fifth model that performs a fourth color conversion process based on a color conversion process instruction from a user, even if the scene determination unit determines that the scene is a treatment scene.

14. The image processing device according to claim 13, characterized in that the fourth color conversion process converts the primary image captured with the white light into a secondary image captured with the red light in a pseudo manner.

15. The image processing device according to claim 1, characterized in that the endoscope device is configured as a server located at a different location.

16. An endoscope system comprising an endoscope device having a camera unit that acquires a primary image of a subject, and an image processing device, wherein the image processing device comprises: a scene determination unit that determines the captured scene of the primary image; a memory unit that stores a plurality of trained models, each of which performs a predetermined processing; a selection unit that selects one model from the plurality of models based on the determination result by the scene determination unit; and an image processing unit that processes the primary image using the one selected model.

17. An image processing method characterized by: determining the photographed scene of a primary image acquired by a camera unit of an endoscope device; selecting one model from multiple trained models, each of which performs a predetermined processing, based on the determination result; and processing the primary image using the selected model.

Images