Image evaluation method, image evaluation device, computer program, and non-transitory computer-readable medium

Abstract

Image data corresponding to an ultrasonic image showing a pelvic region of a subject is acquired. A plurality of feature points relating to a prescribed section are extracted by applying image processing to the image data. An index value associated with at least one of lower urinary tract function and pelvic floor function is calculated on the basis of the plurality of feature points.

Description

Image evaluation method, image evaluation device, computer program, and non-transitory computer-readable medium

[0001] The present disclosure relates to a method and apparatus for assessing indicators related to at least one of lower urinary tract function and pelvic floor function based on ultrasound images of a subject's pelvic region, as well as a computer program executable by a processor installed in the apparatus and a non-transitory computer-readable medium on which the computer program is stored.

[0002] Patent Literature 1 discloses a device capable of acquiring ultrasound images that capture a subject's pelvic region. The ultrasound images can be used, for example, to evaluate lower urinary tract function after radical prostatectomy, or for pelvic floor muscle training to suppress incontinence due to pelvic floor dysfunction that can occur after childbirth. However, evaluation of the ultrasound images depends on the observer's subjectivity and experience.

[0003] Japanese Patent Application Publication No. 2018-504185

[0004] There is a need to provide a method that can objectively evaluate indicators related to at least one of lower urinary tract function and pelvic floor function based on ultrasound images that capture the subject's pelvic region.

[0005] One example embodiment provided by the present disclosure is an image evaluation method executed by at least one computing device, which acquires image data corresponding to an ultrasound image that captures the pelvic region of a subject, applies image processing to the image data to extract multiple feature points related to a predetermined area, and calculates an index value associated with at least one of lower urinary tract function and pelvic floor function based on the multiple feature points.

[0006] One example embodiment provided by the present disclosure is an image evaluation device comprising: an interface that accepts image data corresponding to an ultrasound image that captures the pelvic region of a subject; and a processor that applies image processing to the image data to extract a plurality of feature points related to a predetermined region, and calculates an index value associated with at least one of lower urinary tract function and pelvic floor function based on the plurality of feature points.

[0007] One example aspect provided by the present disclosure is a computer program executable by a processor installed in an image evaluation device, which, when executed, causes the image evaluation device to accept image data corresponding to an ultrasound image that captures the pelvic region of a subject, apply image processing to the image data to extract multiple feature points related to a specified area, and calculate an index value associated with at least one of lower urinary tract function and pelvic floor function based on the multiple feature points.

[0008] A non-transitory computer-readable medium on which the above computer program is stored is also one example of an aspect provided by the present disclosure.

[0009] According to the configurations of the above-described embodiments, by inputting an ultrasound image showing the pelvic region of a subject, it is possible to automate the process of calculating an index value associated with at least one of the subject's lower urinary tract function and pelvic floor function. Because the index value can be obtained without relying on the subjectivity or experience of the person observing the ultrasound image, an objective evaluation environment can be provided.

[0010] 9 is a diagram illustrating a schematic diagram of organs included in the male pelvic region. It illustrates a state after radical prostatectomy. It illustrates an image evaluation method according to an embodiment. It illustrates a detailed flow of image processing in FIG. 3. It illustrates an ultrasound image in which the pelvic region of a subject is reflected. It illustrates a state in which a region of interest is identified in the ultrasound image of FIG. 5. It illustrates an image in which only the region of interest of FIG. 6 is extracted. It illustrates an image in which binarization processing is applied to the image of FIG. 7. It illustrates an image in which edge extraction processing is applied to the image of FIG. 8. It illustrates an image in which Hough transform is applied to the image of FIG. 9. It is a diagram for explaining the process of calculating the angle of the bladder neck ... illustrates a state in which the angle of the bladder neck is displayed. It is a diagram for explaining the angle of the bulbar urethra. It illustrates an ultrasound image in which the urethra of a subject is reflected. It is a diagram for explaining the process of calculating the angle of the bulbar urethra. It is a diagram for explaining the process of calculating the angle of the bulbar urethra. It is a diagram for explaining the process of calculating the angle of the bulbar urethra. 1 is a diagram for explaining the distance between the pubic lower edge and the anorectal angle; FIG. 2 is a diagram for explaining a process for calculating the distance between the pubic lower edge and the anorectal angle; and FIG. 3 is a diagram for explaining the process for calculating the distance between the pubic lower edge and the anorectal angle.

[0011] With reference to the accompanying drawings, exemplary embodiments of a method for assessing indicators related to at least one of lower urinary tract function and pelvic floor function based on ultrasound images of a subject's pelvic region are described in detail below.

[0012] Figure 1 illustrates a schematic representation of the organs contained in the male pelvic region, and Figure 2 illustrates the state after radical prostatectomy, a known radical treatment for prostate cancer.

[0013] In this surgery, the prostate gland 1 is removed and the bladder 2 and urethra 3 are reconnected. As a result of this procedure, the bladder 2 is pulled downward, which may cause funneling, in which the shape of the bladder neck 4 expands. This change in shape makes it easier for urine in the bladder 2 to flow into the urethra 3, and it has been suggested that the occurrence of funneling may be a mechanism for urinary incontinence. Urinary incontinence is an example of a disease associated with decreased function of the lower urinary tract.

[0014] The occurrence of funneling is characterized by the bladder neck angle θ becoming smaller than the preoperative value (180°). When funneling is recognized to be occurring (when the funneling angle is non-zero), the bladder neck angle θ corresponds to the degree of funneling. That is, the larger the angle θ, the greater the degree of funneling and the greater the severity of the impairment of the lower urinary tract function is estimated to be. Lower urinary tract function is evaluated by medical professionals visually checking ultrasound images that capture the subject's pelvic region.

[0015] An image evaluation method according to one embodiment will be described with reference to Figure 3. This method is based on the idea that the bladder neck angle θ can be an index value for quantitatively evaluating the function of the lower urinary tract. In this method, this index value is automatically calculated based on an ultrasound image that captures the subject's pelvic region.

[0016] First, image data corresponding to an ultrasound image that captures the subject's pelvic region is acquired (STEP 1). The ultrasound image can be acquired, for example, by placing an ultrasound probe between the subject's legs.

[0017] Next, the image data is subjected to image processing (STEP 2). The image processing is a preliminary process to enable automatic calculation of the bladder neck angle θ. The detailed flow of the image processing will be described with reference to FIG. 4.

[0018] First, a region including at least the bladder neck is extracted as a region of interest from the ultrasound image capturing the pelvic region (STEP 21). This process is performed by applying a known object detection algorithm to the image data acquired in STEP 1.

[0019] In this embodiment, YOLO (You Only Look Once) is used as the object detection algorithm. A region where YOLO determines that the probability that the bladder neck is reflected exceeds a predetermined threshold is identified as a region of interest. To create a deep learning model that forms the basis of YOLO, training data in which the position of the bladder neck is annotated for ultrasound images that reflect the pelvic region is used.

[0020] FIG. 5 illustrates an ultrasound image 11 that includes the subject's pelvic region. Note that the up-down direction in this image is opposite to that in FIG. 2 . That is, the bladder neck is reflected above the bladder. FIG. 6 illustrates an example in which a region of interest A is identified in the ultrasound image 11 by applying YOLO. FIG. 7 illustrates an extracted image 12 in which only the region of interest A is extracted from the ultrasound image 11.

[0021] Next, the extracted image 12 is subjected to a binarization process (STEP 22 in FIG. 4). Each of the pixels constituting the extracted image 12 has one of a number of gradation values ​​capable of expressing different shades of gray. In the binarization process, the gradation value of each pixel is converted to either a value corresponding to white or a value corresponding to black.

[0022] For example, if the gradation value of a pixel is less than a threshold, it is converted to a value corresponding to black, and if it is greater than or equal to the threshold, it is converted to a value corresponding to white. The threshold may be a predetermined fixed value or a variable value determined through a discriminant analysis method. Figure 8 shows an example of a binarized image 13 obtained by applying binarization processing to the extracted image 12.

[0023] Next, edge extraction processing is performed on the binarized image 13 (STEP 23 in FIG. 4). Edge extraction is a process that emphasizes areas where there is a large difference in grayscale values ​​between adjacent pixels. As is clear from the binarized image 13, grayscale values ​​change significantly at the boundary between the lumen and inner wall of the bladder. Therefore, the edge extraction processing emphasizes the boundary that forms part of the bladder neck.

[0024] Known edge extraction methods include the Sobel method and the Laplacian method. In this example, the Laplacian method is used, which can extract edges more smoothly while suppressing the influence of noise. Figure 9 shows an example of an edge-extracted image 14 obtained by applying edge extraction processing to a binarized image 13.

[0025] Next, the edge-extracted image 14 is subjected to Hough transform processing (STEP 24 in FIG. 4). Hough transform is a process for detecting straight lines in an input image. FIG. 10 shows an example of a straight-line-detected image 15 obtained by applying the Hough transform to the edge-extracted image 14. A large number of straight lines are displayed along the boundary between the bladder lumen and the inner wall.

[0026] Note that at least the binarized image 13, the edge extracted image 14, and the line detected image 15 are shown as examples for the sake of convenience, and do not necessarily need to be visualized for the user. Each process can be explained as a conversion process performed on data corresponding to each image without visualization.

[0027] 3, an index value calculation process is then performed (STEP 3). Based on the line detection image 15, the angle θ of the subject's bladder neck is calculated as an index value.

[0028] 11 shows an example of a graph obtained by plotting the positions of the representative points in the line detection image 15, each of which is a midpoint of a large number of lines detected in the line detection image 15. The horizontal axis represents the horizontal coordinate in the line detection image 15, and the vertical axis represents the vertical coordinate in the same image.

[0029] Next, clustering is applied to the large number of plots obtained as described above. Specifically, the gradient values ​​of the lines that form the basis of each plot are referenced in the line detection image 15. As illustrated in FIG. 12 , plots obtained from lines with positive gradient values ​​are classified into cluster C1. Plots obtained from lines with negative gradient values ​​are classified into cluster C2. In FIG. 12 , the vertical axis represents the horizontal coordinate in the line detection image 15, and the vertical axis represents the gradient value.

[0030] Additionally, among the plots classified into cluster C1, those whose absolute value of the gradient is equal to or greater than the threshold are classified into cluster C11, and those whose absolute value is less than the threshold are classified into cluster C12. Similarly, among the plots classified into cluster C2, those whose absolute value of the gradient is equal to or greater than the threshold are classified into cluster C21, and those whose absolute value is less than the threshold are classified into cluster C22.

[0031] Fig. 13 illustrates a state in which the plots classified into clusters C12 and C22 have been removed from the multiple plots illustrated in Fig. 11. In other words, the plots whose absolute values ​​of the gradients are equal to or greater than the threshold remain.

[0032] Linear regression is applied to the multiple plots belonging to cluster C12, weighting the lines according to their lengths obtained by applying a Hough transform, and then the line L1 is obtained. Linear regression corresponds to a line approximating the inner wall surface of the bladder neck, which appears to have a positive gradient in the extracted image 12 of FIG. 5 . Similarly, linear regression is applied to the multiple plots belonging to cluster C22, and line L2 is obtained. Linear regression corresponds to a line approximating the inner wall surface of the bladder neck, which appears to have a negative gradient in the extracted image 12 of FIG. 5 . The angle θ formed by line L1 and line L2 is identified as the angle of the bladder neck.

[0033] The above clustering approach, which focuses not only on the positive or negative gradient but also on the absolute value of the gradient, is based on the finding that the bladder neck where funneling occurs has a shape that includes areas with relatively large gradients and areas with relatively small gradients.

[0034] 11 is obtained by applying linear regression to all plots belonging to cluster C1. Similarly, line L2' is obtained by applying linear regression to all plots belonging to cluster C2. These lines do not adequately approximate the shape of the actual bladder neck, and the angle θ' between lines L1' and L2' also does not correspond to the angle of the actual bladder neck.

[0035] A comparison between FIG. 11 and FIG. 13 reveals that by performing processing that focuses on the portion of the bladder neck where the gradient is relatively large, it is possible to calculate a value that is closer to the actual angle of the bladder neck.

[0036] The above clustering can be achieved by applying a well-known method at least once. Examples of well-known methods include the k-means method and a Gaussian mixture distribution. Note that it is not necessary to perform clustering by focusing on both the gradient and magnitude of the line. For example, clustering may be performed at least once by focusing only on the magnitude of the gradient.

[0037] As shown in Fig. 3, the result data is then output (STEP 4). The result data is configured to include at least data corresponding to the bladder neck angle calculated as described above.

[0038] 14 shows an example of the output mode of the result data. In this example, the calculated bladder neck angle is displayed on the display device. That is, the result data is configured to cause the display device to display the value of the angle. In this example, an image 16 showing the basis for the calculation of the angle is also displayed on the display device. Image 16 in this example corresponds to the line detection image 15 shown in FIG. 10 with the lines that did not contribute to the calculation of the bladder neck angle removed.

[0039] The output form of the result data is not limited to display on a display device, but may be configured to correspond to the format of a report that is provided for printing or data transmission.

[0040] As described above, according to the configuration of this embodiment, by inputting an ultrasound image that captures the pelvic region of a subject, the processing steps up to calculating the angle of the subject's bladder neck can be automated. This makes it possible to provide an index value associated with the function of the subject's lower urinary tract without relying on the subjectivity or experience of the person observing the ultrasound image. In other words, it is possible to provide an environment in which an index value associated with the function of the subject's lower urinary tract can be objectively evaluated.

[0041] Another example of an index that can be used to evaluate the function of the lower urinary tract is the angle of the bulbar urethra. As shown in Figure 15, the bulbar urethra 3a is a part of the urethra 3 that curves greatly near the bladder neck 4. The angle φ of the bulbar urethra 3a is defined as the angle between two approximate curves extending along both ends of the bulbar urethra 3a.

[0042] 16 shows an example of an ultrasound image 17 that includes the subject's urethra 3 (the position of the urethra 3 is highlighted by a dashed line) and corresponds to the image data acquired in STEP 1 of Fig. 3. Because the urethra 3 is a very long, tubular organ, it is not easy to visually determine the position of the urethra 3 from the ultrasound image 17.

[0043] 17 shows the ultrasound image 17 of FIG. 16 upside down. That is, the urethra is reflected above the bladder neck 4. The corpus cavernosum muscle 8 and soft tissue 9 are adjacent to the urethra and appear as low-brightness areas in the ultrasound image 17. Therefore, the boundary B between the low-brightness area and the adjacent high-brightness area can be considered to be the urethra 3.

[0044] FIG. 18 illustrates an example of a state in which the contours of the regions of interest corresponding to the above-mentioned low-brightness and high-brightness regions are extracted by applying image processing to image data corresponding to the ultrasound image 17 (STEP 2 in FIG. 3).

[0045] In this example, YOLO is also used as the object detection algorithm. YOLO extracts the contours of areas where the probability that a region of interest is reflected exceeds a predetermined threshold. The contours are acquired as a set of point clouds. To create the deep learning model that forms the basis of YOLO in this example, training data is used that is annotated by tracing the contours using images of regions of interest marked by experts.

[0046] Next, the angle of the bulbar urethra is calculated as an index value (STEP 3 in FIG. 3). Specifically, a first endpoint P1 and a second endpoint P2 are identified to extract a portion corresponding to the bulbar urethra 3a from the extracted contour. The first endpoint P1 can be identified as the extreme point when the above-mentioned point group is regarded as a curve. The second endpoint P2 can be, for example, the coordinates of the point identified as the vertex of the bladder neck 4.

[0047] In Figure 19, a first approximate line L11 and a second approximate line L12 are set for the portion corresponding to the extracted bulbar urethra 3a. The first approximate line L11 is set by applying linear regression to a group of points in an appropriate range including the first endpoint P1. The second approximate line L12 is set by applying linear regression to a group of points in an appropriate range including the second endpoint P2. The angle φ formed by the first approximate line L11 and the second approximate line L12 is identified as the angle of the bulbar urethra.

[0048] Next, a process for outputting the result data is performed (STEP 4 in FIG. 3). The result data in this example is configured to include at least data corresponding to the angle φ calculated as described above. The result data can be output in the form of a printed report or data transmission, in addition to or instead of being displayed on the display device described with reference to FIG. 14.

[0049] According to the configuration of this example, by inputting an ultrasound image showing the subject's pelvic region, the processing steps up to calculating the angle of the subject's bulbar urethra can be automated. This makes it possible to provide an index value associated with the subject's lower urinary tract function without relying on the subjective opinion or experience of the person observing the ultrasound image. In other words, it is possible to provide an environment in which the index value associated with the subject's lower urinary tract function can be objectively evaluated.

[0050] The pelvic floor muscles 5 illustrated in Figure 20 are multiple muscles that make up the pelvic floor. The pelvic floor is a collective term for the muscles, fascia, and ligaments that support the pelvic organs (uterus, bladder, rectum, etc.). Training to strengthen the pelvic floor muscles 5 is performed for the purposes of preventing incontinence due to decreased pelvic floor function that can occur after childbirth, and restoring lower urinary tract function that has decreased following the aforementioned radical prostatectomy. In the following explanation, this training will be referred to as "pelvic floor muscle training."

[0051] Pelvic floor muscle training is performed by having the subject consciously contract the pelvic floor muscles 5. Contracting the pelvic floor muscles 5 narrows the distance D between the inferior pubic border 6 and the anorectal angle 7. A healthcare professional visually checks the distance D in an ultrasound image of the subject's pelvic region to assess the subject's ability to appropriately contract the pelvic floor muscles 5. In other words, the distance D can be used as an index to quantitatively evaluate pelvic floor function.

[0052] The image evaluation method described with reference to Fig. 3 can also be used to calculate the index value. That is, the distance D between the pubic margin 6 and the anorectal angle 7 can be automatically calculated based on an ultrasound image that captures the subject's pelvic region.

[0053] Fig. 21 illustrates an ultrasound image 18 that includes the subject's pelvic region and corresponds to the image data acquired in STEP 1 of Fig. 3. By applying image processing to the image data (STEP 2 of Fig. 3), a region including at least the pubic margin and a region including at least the anorectal angle are extracted as regions of interest.

[0054] In this example, YOLO is also used as the object detection algorithm. A region in which YOLO determines that the probability that the pubic margin is reflected exceeds a predetermined threshold is identified as a region of interest A1. Similarly, a region in which YOLO determines that the probability that the anorectal angle is reflected exceeds a predetermined threshold is identified as a region of interest A2. To create the deep learning model that forms the basis of YOLO in this example, training data is used in which annotations are added to ultrasound images that reflect the pelvic region, with the positions of the pubic margin and the anorectal angle.

[0055] Next, distance D is calculated as an index value (STEP 3 in FIG. 3). In this example, the amount corresponding to the length of a line connecting the center point of region of interest A1 and the center point of region of interest A2 is specified as distance D. However, appropriate representative points may be determined for each region of interest, and the amount corresponding to the length of a line connecting the representative points may be specified as distance D.

[0056] Next, a process of outputting the result data is performed (STEP 4 in FIG. 3). The result data in this example is configured to include at least data corresponding to the distance D calculated as described above. The result data can be output in the form of a printed report or data transmission, in addition to or instead of being displayed on the display device described with reference to FIG. 14.

[0057] With the configuration of this example, by inputting an ultrasound image showing the subject's pelvic region, the processing steps up to calculating the distance between the subject's pubic margin and the anorectal angle can be automated. This makes it possible to provide an index value associated with the subject's pelvic floor function without relying on the subjective opinion or experience of the person observing the ultrasound image. In other words, it is possible to provide an environment in which an index value associated with the subject's pelvic floor function can be objectively evaluated.

[0058] As mentioned above, pelvic floor muscle training is also performed with the aim of restoring lower urinary tract function. Therefore, the distance between the pubic border and the anorectal angle can be an example of an index related to lower urinary tract function.

[0059] When the ultrasound images 18 are continuously acquired over time, the change in the distance D over time can be automatically acquired by applying a process for tracking the positions of the pubic margin and the anorectal angle identified through image processing. Examples of such a process include a kernelized correction filter (KCF), multiple instance learning (MIL), tracking learning detection (TLD), boosting, and median flow.

[0060] By obtaining the change in distance D over time, it is possible to quantitatively evaluate index values ​​such as the contraction speed, contraction acceleration, contraction frequency, duration of the contraction state, ability to repeat contractions, and delay time from the muscle contraction command (self-intention to start muscle contraction) of the pelvic floor muscles.

[0061] In addition, by displaying ultrasound images with the region of interest displayed continuously over time on a display device, as shown in Figure 21, the state of contraction of the pelvic floor muscles can be visualized more clearly. This provides feedback to the subject of pelvic floor muscle training, helping them understand the success or failure of their exercises and any necessary corrections.

[0062] The result data output in STEP 4 of Figure 3 may include an estimate of the severity of the decline in at least one of the lower urinary tract function and the pelvic floor function based on the various index values ​​calculated as described above.

[0063] For example, the greater the degree of funneling, the greater the bladder neck angle θ, which is estimated to indicate a higher severity of dysfunction of the lower urinary tract. As another example, the smaller the change in the distance between the pubic border and the anorectal angle obtained during pelvic floor muscle training, the greater the severity of dysfunction of the pelvic floor.

[0064] Data showing the correspondence between index values ​​and severity levels is stored in advance in a storage device in table format or as the result of machine learning. The data is referenced by a computing device that receives index values ​​as input, and the severity level is automatically estimated.

[0065] The output result data is configured to visualize the estimated result of severity through at least one of text, color, and symbol.

[0066] With this configuration, it is possible to automate the process of estimating the severity of functional impairment of at least one of the subject's lower urinary tract and pelvic floor based on an ultrasound image that captures the subject's pelvic region, thereby providing an evaluation environment that is not dependent on the subjectivity or experience of the observer of the ultrasound image.

[0067] The result data output in STEP 4 of Figure 3 may include the selection result of a pelvic floor muscle training method based on at least one of the various index values ​​calculated as described above and the severity estimated as described above.

[0068] The selection of a training method based on the evaluation results of the contractile ability of the pelvic floor muscles is a common practice. Therefore, by defining a correspondence relationship between the evaluation results of the contractile ability of the pelvic floor muscles and at least one of the index value and severity estimation results obtained as described above, it is possible to automate the selection of a training method appropriate for the subject. Data indicating the correspondence relationship is stored in advance in a storage device in table format or as the result of machine learning. The data is referenced by a computing device that receives at least one of the index value and severity estimation results as input, thereby automatically selecting a training method.

[0069] The output result data is configured to visualize the results of the training method selection through at least one of text, color, and symbol.

[0070] With this configuration, it is possible to automate the process of selecting a pelvic floor muscle training method appropriate for a subject based on an ultrasound image showing the subject's pelvic region, providing a teaching environment that is not dependent on the subjectivity or experience of the person observing the ultrasound image.

[0071] As illustrated in FIG. 3, the image evaluation method according to the present disclosure can be configured to enable comparison of index values ​​calculated at multiple points in time.

[0072] Specifically, after the result data at a certain point in time is output, it is determined whether a comparison of the index values ​​is necessary (STEP 5). If it is determined that a comparison is not necessary (NO in STEP 5), the processing according to this method ends.

[0073] If it is determined that a comparison of index values ​​is necessary (YES in STEP 5), it is determined whether additional index values ​​to be used for comparison have been acquired (STEP 6). If it is determined that additional index values ​​have not been acquired (NO in STEP 6), the process returns to STEP 1, and the process for acquiring additional index values ​​is repeated.

[0074] If it is determined that an additional index value has been acquired (YES in STEP 6), it is compared with the previously acquired index value, and data indicating the comparison result is output (STEP 7). The comparison result may be presented in a form that lists index values ​​acquired at multiple points in time, or a new index value obtained based on the multiple index values ​​may be presented. Examples of the new index value include a difference value, an average value, etc.

[0075] This configuration makes it possible to automate the monitoring of the function of at least one of the subject's lower urinary tract and pelvic floor, as well as the evaluation of the results of pelvic floor muscle training, thereby providing an evaluation environment that is not dependent on the subjectivity or experience of the ultrasound image observer.

[0076] The image evaluation method illustrated in FIG. 3 can be executed by at least one computing device. Each processing step may be executed by a separate computing device, or multiple processing steps may be executed by a single computing device. When the image evaluation method is executed by multiple computing devices, the multiple computing devices may be located remotely from each other. In this case, data is exchanged between the computing devices via a public or private communication network. The communication network may be realized by a wired connection, a wireless connection, or a combination thereof.

[0077] 22 illustrates the functional configuration of an image evaluation device 30 according to an embodiment. The image evaluation device 30 is an example of the above-mentioned arithmetic device. The image evaluation device 30 may be a device that is installed in a specific location, or may be a device that can be carried by a user.

[0078] The image evaluation device 30 includes an input interface 31 , a processor 32 , and an output interface 33 .

[0079] The input interface 31 is configured as a hardware interface that receives image data IM from the input device 40. The image data IM corresponds to an ultrasound image that captures the subject's pelvic region. The input device 40 may be an ultrasound probe that is placed against the subject's body, or a device that visualizes ultrasound images. The image data IM may be in the form of analog data or digital data, depending on the specifications of the input device 40. If the image data IM is in the form of analog data, the input interface 31 includes an appropriate conversion circuit including an A / D converter.

[0080] The processor 32 is configured to execute the processes described with reference to FIG. 3 based on the image data IM received by the input interface 31. The processor 32 having such functions may be realized by a general-purpose microprocessor operating in cooperation with general-purpose memory. Examples of the general-purpose microprocessor include a CPU, an MPU, and a GPU. Examples of the general-purpose memory include a ROM and a RAM. In this case, a computer program for implementing the functions may be stored in the ROM. The general-purpose microprocessor specifies at least a portion of the program stored in the ROM, expands it on the RAM, and executes the above-described processes in cooperation with the RAM. In this case, the general-purpose memory is an example of a non-transitory computer-readable medium on which a computer program is stored.

[0081] The processor 32 may be realized by a dedicated integrated circuit such as a microcontroller, ASIC, or FPGA having a storage element in which a computer program for implementing the processor's functions is pre-installed. In this case, the storage element is an example of a non-transitory computer-readable medium in which a computer program is stored. The processor 32 may also be realized by a combination of a general-purpose microprocessor and a dedicated integrated circuit.

[0082] The output interface 33 is configured as a hardware interface that outputs the result data RS generated by the processor 32 to the output device 50. Examples of the output device 50 include a display device, a speaker, a printer, and a data transmission device. The output device 50 may be a part of the image evaluation device 30, or may be a device independent of the image evaluation device 30. The result data RS may be in the form of analog data or digital data, depending on the specifications of the output device 50. When the result data RS is in the form of analog data, the output interface 33 is equipped with an appropriate conversion circuit including a D / A converter.

[0083] The configurations described above are merely examples to facilitate understanding of the present disclosure. Each configuration example can be appropriately modified and combined with other configuration examples without departing from the spirit of the present disclosure.

[0084] The index values ​​associated with at least one of the subject's lower urinary tract function and pelvic floor function are not limited to the bladder neck angle, the distance between the pubic margin and the anorectal angle, and the bulbar urethra angle. Other index values ​​may be calculated as long as they are identifiable based on ultrasound images of the subject's pelvic organs and are associated with at least one of the lower urinary tract function and pelvic floor function. Examples of such organs include the bladder, uterus, urethra, and vagina. Examples of such index values ​​include the bladder neck travel distance and the contraction rate of the levator hiatus area.

[0085] The contents of Japanese Patent Application No. 2023-090404 filed on May 31, 2023 are incorporated by reference as part of this disclosure.

Claims

1. 1. A method of image evaluation executed by at least one computing device, comprising: acquiring image data corresponding to an ultrasound image capturing the subject's pelvic region; extracting a plurality of feature points relating to a predetermined region by applying image processing to the image data; calculating an index value associated with at least one of a lower urinary tract function and a pelvic floor function based on the plurality of feature points; Image evaluation methods.

2. each of the plurality of feature points corresponds to a portion of a bladder neck; the index value corresponds to the angle of the bladder neck; The image evaluation method according to claim 1 .

3. Each of the plurality of feature points is a part of a straight line extracted by the image processing, and clustering is performed focusing on the gradient of the straight line. The image evaluation method according to claim 2 .

4. the plurality of feature points correspond to the pubic margin and the anorectal angle; the index value corresponds to at least one of the distance between the pubic lower edge and the anorectal angle and a change in the distance over time; The image evaluation method according to claim 1 .

5. the index value is calculated by applying an image tracking technique to the plurality of feature points; The image evaluation method according to claim 4.

6. estimating the severity of a decline in at least one of the lower urinary tract function and the pelvic floor function of the subject based on the index value; outputting data corresponding to the result of said estimation; The image evaluation method according to claim 1 .

7. selecting a pelvic floor muscle training method suitable for the subject based on the index value; outputting data corresponding to the results of said selection; The image evaluation method according to claim 1 .

8. comparing the index values ​​at multiple time points; outputting data corresponding to the results of said comparison; The image evaluation method according to claim 1 .

9. an interface that accepts image data corresponding to an ultrasound image that includes a pelvic region of the subject; a processor that applies image processing to the image data to extract a plurality of feature points relating to a predetermined region, and calculates an index value associated with at least one of a lower urinary tract function and a pelvic floor function based on the plurality of feature points; Equipped with Image evaluation device.

10. A computer program executable by a processor installed in an image evaluation device, By executing the above, the image evaluation device receiving image data corresponding to an ultrasound image capturing a pelvic region of the subject; extracting a plurality of feature points relating to a predetermined region by applying image processing to the image data; calculating an index value associated with at least one of a lower urinary tract function and a pelvic floor function based on the plurality of feature points; Computer program.

11. A non-transitory computer-readable medium having stored thereon the computer program of claim 10.