A method for localizing peripheral nerve injuries in the hand based on medical imaging

Abstract

The present invention relates to the field of image segmentation technology, and more particularly to a method for locating hand peripheral nerve injuries based on medical images. The present application first divides various tissue regions for further screening based on the relatively uniform grayscale characteristics of the ulnar nerve region; then, based on the characteristics of the long strip echo structure of the ulnar nerve, the probability of the ulnar nerve is determined by calculating the echo expression levels of each sub-block in the tissue region, thereby preliminarily screening the ulnar nerve region; further, based on the characteristic that the width of the compressed region is suppressed under compression, a more accurate and true compressed region is screened based on the relative distribution of the neighborhood widths of the sub-blocks in the ulnar nerve region, thereby achieving higher accuracy in locating hand peripheral nerve injuries based on the true compressed region.

Description

Technical Field

[0001] The present invention relates to the technical field of image segmentation, and in particular to a method for locating hand peripheral nerve damage based on medical images. Background Art

[0002] The peripheral nerves of the hand refer to the nervous system responsible for controlling hand muscle movement and transmitting sensory signals. It includes multiple nerves, such as the ulnar nerve, median nerve, and radial nerve. These nerves extend from the spinal cord to the hand, responsible for transmitting sensations such as touch, temperature, and pain, while also controlling fine movements of the hand. Among them, the ulnar nerve plays a key role in hand control and sensory transmission, and its compression often causes more complex symptoms, such as hand weakness and numbness, necessitating the localization of ulnar nerve damage.

[0003] Common types of ulnar nerve injury include compression injury, mechanical injury, traction injury, and chronic strain. Among them, the most common type of injury is compression injury. Compression injury refers to damage to the peripheral nerve caused by chronic compression, friction, or traction in anatomically narrow or fibrous bony channels. It is common in areas such as the carpal tunnel, cubital tunnel, or Guyon's tunnel. Typical manifestations include paresthesia in the area innervated by the compressed nerve, motor dysfunction, and nerve swelling, blurred fascicular structure, or restricted sliding under dynamic ultrasound. It is mostly caused by anatomical variation, repeated strain, or local space-occupying injuries.

[0004] Existing technologies typically use semantic segmentation methods to input long-axis nerve images collected by ultrasound imaging technology into a trained semantic segmentation network to segment out the compressed areas affected by compression injuries. However, manual labeling is required when training the semantic segmentation network. The peripheral nerve structure of the hand is complex, the image data volume is large, and the information is complex. Manual labeling may have errors, resulting in low accuracy of the compressed areas segmented by the semantic segmentation network. Summary of the Invention

[0005] To address the technical problem of low accuracy in segmenting the entrapment area of ​​a compression injury using semantic segmentation methods in the prior art, the present application aims to provide a method for locating peripheral nerve injuries in the hand based on medical images. The technical solution employed is as follows:

[0006] The first aspect of the present application provides a method for locating hand peripheral nerve damage based on medical imaging, comprising:

[0007] Acquiring a long-axis grayscale image of the ulnar nerve through medical ultrasound imaging; determining each edge tissue sub-block in each tissue region based on a local grayscale distribution in the long-axis grayscale image of the ulnar nerve;

[0008] The echogenicity of each edge tissue sub-block is determined based on the similarity of the echogenicity directions between each edge tissue sub-block and other edge tissue sub-blocks with similar edge extension trends; the corresponding ulnar nerve probability is determined based on the width of each edge tissue sub-block and the overall size of the echogenicity in each tissue region;

[0009] The ulnar nerve region is screened out according to the ulnar nerve probability; the true entrapment region is screened out according to the relative distribution of the neighborhood widths of each edge tissue sub-block in the ulnar nerve region and the corresponding ulnar nerve probability; and the peripheral nerve injury of the hand is located according to the true entrapment region.

[0010] Furthermore, the process of obtaining the edge tissue sub-block includes:

[0011] Performing superpixel segmentation on the long axis grayscale image of the ulnar nerve to obtain at least two segmented tissue sub-blocks; determining the corresponding overall grayscale value according to the mean grayscale value of all pixels in each segmented tissue sub-block;

[0012] The other segmented tissue sub-blocks outside each segmented tissue sub-block are used as corresponding contrast tissue sub-blocks; the corresponding cluster distance is determined according to the relative distance between each segmented tissue sub-block and each corresponding contrast tissue sub-block and the overall grayscale deviation;

[0013] Perform cluster analysis based on the cluster distances between all segmented tissue sub-blocks to obtain at least two tissue sub-block clusters;

[0014] The area corresponding to all the segmented tissue sub-blocks corresponding to each tissue sub-block cluster is taken as the corresponding tissue area; and each segmented tissue sub-block in each tissue area is taken as an edge tissue sub-block.

[0015] Furthermore, the process of obtaining the cluster distance includes:

[0016] Determine a corresponding grayscale difference value according to a difference between the overall grayscale value of each segmented tissue sub-block and the overall grayscale value of each corresponding contrasted tissue sub-block;

[0017] Determine the corresponding basic physical distance according to the Euclidean distance between the centroid of each segmented tissue sub-block and the centroid of each corresponding comparison tissue sub-block;

[0018] The clustering distance between each segmented tissue sub-block and each corresponding contrasted tissue sub-block is determined according to a normalized value of the sum of the squares of the grayscale difference value and the basic physical distance.

[0019] Furthermore, the process of obtaining the echo performance level includes:

[0020] In each edge tissue sub-block, a least squares straight line fitting is performed on each boundary pixel point and a preset number of adjacent edge pixel points to determine a reference fitting straight line for each boundary pixel point;

[0021] Taking the counterclockwise direction as the positive direction, determine the corresponding local direction angle according to the minimum angle between the reference fitting line corresponding to each boundary pixel point and the horizontal right direction; determine the corresponding global direction angle according to the minimum angle between the line where the major axis of the minimum circumscribed rectangle of each edge tissue sub-block lies and the horizontal right direction; the absolute value of the local direction angle and the absolute value of the global direction angle are both less than 90 degrees;

[0022] The reference direction angle of each boundary pixel point is determined by taking the average of the local direction angle of each boundary pixel point and the overall direction angle; the sub-block extension angle of each edge organization sub-block is determined according to the average of the reference direction angles of all boundary pixels in each edge organization sub-block; and the sub-block extension direction of each edge organization sub-block is determined according to the sub-block extension angle;

[0023] In each tissue region, the other edge tissue sub-blocks that are passed through a straight line that passes through the centroid of each edge tissue sub-block and is parallel to the major axis of the corresponding minimum circumscribed rectangle are used as reference tissue sub-blocks for each edge tissue sub-block. The local echo direction of each tissue sub-block is determined by structural tensor estimation. The corresponding echo performance level is determined based on the overall deviation distribution between the sub-block extension direction of each edge tissue sub-block and the local echo directions of each reference tissue sub-block.

[0024] Furthermore, the process of determining the corresponding echo performance level according to the overall deviation distribution between the sub-block extension direction of each edge tissue sub-block and the local echo direction of each reference tissue sub-block includes:

[0025] The angle between the sub-block extension direction of each edge tissue sub-block and the local echo direction of each reference tissue sub-block is used as the echo direction deviation value of each reference tissue sub-block; and the overall echo deviation value is determined according to the average of the echo direction deviation values ​​of the reference tissue sub-blocks;

[0026] Determining the corresponding echo direction disorder according to the variance of the echo direction deviation values ​​of all reference tissue sub-blocks;

[0027] A negative correlation mapping is performed on the product of the echo direction disorder degree and the overall echo deviation value to determine the echo performance degree of each edge tissue sub-block.

[0028] Furthermore, the process of obtaining the ulnar nerve probability includes:

[0029] In each edge tissue sub-block, a straight line passing through each boundary pixel and perpendicular to the straight line corresponding to the reference direction angle of the boundary pixel is used as the width direction straight line of each boundary pixel;

[0030] The minimum distance between the pixel points where the width-direction straight line of each boundary pixel intersects the boundary of the tissue region is used as the local measured width of each boundary pixel; the corresponding local tissue width is determined based on the average of the local measured widths of all boundary pixels in each edge tissue sub-block;

[0031] determining a local weighted probability of each edge tissue sub-block according to a product of a normalized value of the local tissue width and the echo performance degree;

[0032] The corresponding ulnar nerve probability is determined according to the accumulated value of the local weighted probabilities of all edge tissue sub-blocks in each tissue region.

[0033] Furthermore, the process of obtaining the ulnar nerve area includes:

[0034] The tissue area whose corresponding ulnar nerve probability is greater than the preset ulnar nerve threshold is regarded as the ulnar nerve area.

[0035] Furthermore, the process of obtaining the actual compression area includes:

[0036] In each ulnar nerve region, a k-means cluster analysis was performed on the local tissue widths of all edge tissue sub-blocks to obtain at least two tissue sub-block clusters. The corresponding width eigenvalue was determined based on the mean of the local tissue widths of all edge tissue sub-blocks in each tissue sub-block cluster. Adjacent edge tissue sub-blocks in the tissue sub-block cluster with the largest width eigenvalue were merged to determine all suspected swelling areas. Adjacent edge tissue sub-blocks in the tissue sub-block cluster with the smallest width eigenvalue were merged to determine all suspected compression areas.

[0037] In each ulnar nerve region, the nearest suspected swelling region corresponding to each suspected compression region is used as the corresponding reference impact region; the number of edge tissue sub-blocks in the reference impact region is normalized to determine the corresponding swelling persistence characteristic value; the minimum distance between each suspected compression region and the reference impact region is negatively correlated to determine the corresponding distance impact characteristic value; the mean of the local tissue width of all edge tissue sub-blocks in each suspected compression region is negatively correlated to determine the corresponding reference compression degree; the compression characteristic degree of each suspected compression region is determined based on the normalized value of the product of the swelling persistence characteristic value, the distance impact characteristic value and the reference compression degree; and the true compression region is screened out based on the compression characteristic degree.

[0038] Furthermore, the process of screening out the real compression area according to the compression feature degree includes:

[0039] The suspected pressed area whose corresponding pressed feature degree is greater than the preset pressed threshold is regarded as the real pressed area.

[0040] Furthermore, the process of locating hand peripheral nerve injury according to the actual compression area includes:

[0041] In the ulnar nerve long axis grayscale image, the minimum circumscribed rectangle of each real compression area is marked as the nerve compression injury area.

[0042] In a second aspect, the present application provides a hand peripheral nerve injury localization system based on medical imaging, the system comprising:

[0043] a data acquisition and preprocessing module configured to acquire a long-axis grayscale image of the ulnar nerve through medical ultrasound imaging; and determine each edge tissue sub-block in each tissue region based on the local grayscale distribution in the long-axis grayscale image of the ulnar nerve;

[0044] The ulnar nerve probability determination module is used to determine the echogenicity of each edge tissue sub-block based on the similarity of the echo directions between each edge tissue sub-block and other edge tissue sub-blocks with similar edge extension trends; and to determine the corresponding ulnar nerve probability based on the width of each edge tissue sub-block and the overall echogenicity in each tissue region;

[0045] The hand peripheral nerve injury localization module is used to filter out the ulnar nerve area based on the ulnar nerve probability; filter out the real entrapment area based on the relative distribution of the neighborhood width of each edge tissue sub-block in the ulnar nerve area and the corresponding ulnar nerve probability; and locate the hand peripheral nerve injury based on the real entrapment area.

[0046] In a third aspect, the present application provides a computer device comprising a memory and a processor. The memory is configured to store computer program code, and the processor is configured to call and execute the computer program code from the memory to perform the method of the first aspect or any embodiment of the first aspect of the present application.

[0047] In a fourth aspect, the present application provides a computer program product, comprising a computer program code. When the computer program code is executed, the method of the first aspect or any embodiment of the first aspect of the present application is performed.

[0048] In a fifth aspect, the present application provides a computer-readable storage medium, which stores computer program code. When the computer program code is executed, it performs the method of the first aspect of the present application or any embodiment of the first aspect.

[0049] This application has the following beneficial effects:

[0050] This application first divides the various tissue regions for further screening based on the relatively uniform grayscale characteristics of the ulnar nerve region; then, based on the characteristics of the long strip echo structure of the ulnar nerve, the probability of the ulnar nerve is determined by calculating the echo performance of each sub-block in the tissue region, thereby preliminarily screening the ulnar nerve region; further, based on the characteristic that the width of the entrapment area will be suppressed under compression, a more accurate and true entrapment area is screened according to the relative distribution of the neighborhood widths of the sub-blocks in the ulnar nerve region, so that the accuracy of locating the peripheral nerve injury of the hand based on the true entrapment area is higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A flowchart of a method for locating hand peripheral nerve damage based on medical imaging provided by one embodiment of the present invention;

[0053] Figure 2 A structural diagram of a hand peripheral nerve injury localization system based on medical imaging provided by one embodiment of the present invention;

[0054] Figure 3 The present invention provides a schematic diagram of a computer device structure according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following, in combination with the accompanying drawings and preferred embodiments, a method for locating peripheral nerve injuries of the hand based on medical imaging proposed by the present invention, its specific implementation method, structure, characteristics and effects are described in detail as follows. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment, and the specific features, structures or characteristics in one or more embodiments may be combined in any suitable form. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as implying or suggesting relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features.

[0056] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0057] The following describes in detail a specific solution of a method for locating hand peripheral nerve damage based on medical imaging provided by the present invention with reference to the accompanying drawings.

[0058] This application embodiment provides a method for locating hand peripheral nerve damage based on medical imaging. Figure 1 , which shows a flow chart of a method for locating hand peripheral nerve damage based on medical imaging provided by one embodiment of the present invention, the method comprising:

[0059] Step S101: acquiring a long-axis grayscale image of the ulnar nerve through medical ultrasound imaging; and determining each edge tissue sub-block in each tissue region according to the local grayscale distribution in the long-axis grayscale image of the ulnar nerve.

[0060] In a specific implementation of an embodiment of the present invention, a 12MHz high-frequency linear array probe is placed on the inner side of the elbow and scanned along the elbow area to obtain a long-axis ultrasound image of the ulnar nerve; the long-axis ultrasound image of the ulnar nerve is grayscaled to obtain the long-axis grayscale image of the ulnar nerve required by this application.

[0061] In long-axis images of the ulnar nerve, the nerve typically appears as a long, hypoechoic, linear structure running along the inner side of the forearm or hand. The nerve typically appears as a hypoechoic region, particularly when compared to the surrounding soft tissue. Specifically, in long-axis images of the ulnar nerve, the nerve's grayscale value is lower than that of the surrounding tissue, and the grayscale distribution is relatively uniform, making it easier to distinguish on ultrasound images. To fully segment the ulnar nerve in long-axis images, tissues in these images are segmented based on grayscale.

[0062] Preferably, in some possible implementations of the embodiments of the present invention, the process of acquiring the edge tissue sub-block includes:

[0063] Performing superpixel segmentation on the long axis grayscale image of the ulnar nerve to obtain at least two segmented tissue sub-blocks; determining the corresponding overall grayscale value based on the mean grayscale value of all pixels in each segmented tissue sub-block;

[0064] The other segmented tissue sub-blocks outside each segmented tissue sub-block are used as corresponding contrast tissue sub-blocks; the corresponding cluster distance is determined based on the relative distance between each segmented tissue sub-block and each corresponding contrast tissue sub-block and the overall grayscale deviation; wherein the cluster distance acquisition process includes: determining the corresponding grayscale difference value based on the difference between the overall grayscale value of each segmented tissue sub-block and the overall grayscale value of each corresponding contrast tissue sub-block; determining the corresponding basic physical distance based on the Euclidean distance between the centroid of each segmented tissue sub-block and the centroid of each corresponding contrast tissue sub-block; and determining the cluster distance between each segmented tissue sub-block and each corresponding contrast tissue sub-block based on the normalized value of the sum of the squares of the grayscale difference value and the basic physical distance.

[0065] The grayscale in the ulnar nerve region is relatively uniform, so by merging the segmented tissue sub-blocks that are relatively close and have similar grayscales, the ulnar nerve region can be more accurately aligned with one of the clusters obtained after clustering; therefore, in order to further screen out the ulnar nerve region, the cluster distance is characterized by the physical distance and overall difference between the segmented tissue sub-blocks. In a specific implementation of an embodiment of the present invention, the process of obtaining the cluster distance is expressed by the formula: ;in, To split tissue sub-blocks Compare with the corresponding tissue sub-block The cluster distance between To segment tissue sub-blocks The centroid of the corresponding contrast tissue block The Euclidean distance between the centers of mass, that is, the basic physical distance; To segment tissue sub-blocks The overall gray value and the corresponding contrast tissue sub-block The difference between the overall gray values, that is, the gray difference value; To split tissue sub-blocks Compare with the corresponding tissue sub-block The sum of the squares of the grayscale difference value and the basic physical distance; It should be noted that the difference in the embodiment of the present invention represents the absolute value of the difference, which will not be further elaborated later.

[0066] Cluster analysis is further performed based on the cluster distances between all segmented tissue sub-blocks to obtain at least two tissue sub-block clusters. In a specific implementation of an embodiment of the present invention, the cluster analysis method uses DBSCAN cluster analysis, which can be adjusted according to the specific implementation environment and is not further limited or elaborated herein. Furthermore, based on the cluster analysis results, the area corresponding to all segmented tissue sub-blocks corresponding to each tissue sub-block cluster is used as the corresponding tissue region; and each segmented tissue sub-block in each tissue region is used as an edge tissue sub-block. Among the obtained tissue regions, some tissue regions belong to the ulnar nerve region, and therefore require further screening.

[0067] Step S102: Determine the echogenicity of each edge tissue sub-block based on the similarity of the echo directions between each edge tissue sub-block and other edge tissue sub-blocks with similar edge extension trends; determine the corresponding ulnar nerve probability based on the width of each edge tissue sub-block in each tissue region and the overall size of the echogenicity.

[0068] In ultrasound images, the perineurium and epineurium inside the nerve appear as linear hyperechoic bands on the long-axis section, forming the hyperechoic fascicular structure of the ulnar nerve. These structures separate and wrap the nerve fiber bundles, forming a characteristic "fascicular" or "cable-like" appearance. The linear hyperechoic bands cause multiple linear parallel strong echoes inside the ulnar nerve, and the extension direction of the parallel strong echoes is the same as that of the ulnar nerve. Therefore, the linear hyperechoic bands are used as features to distinguish the ulnar nerve from other tissues. Therefore, the echo intensity of each marginal tissue sub-block in each tissue area is first measured so that the echo intensity can characterize the distribution pattern of the linear hyperechoic bands of the ulnar nerve.

[0069] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the echo performance level includes:

[0070] In each edge organization sub-block, a least squares straight line fit is performed on each boundary pixel and a preset number of adjacent edge pixels to determine a reference fitting line for each boundary pixel. In one specific implementation of the present invention, the preset number is set to 20, which can be adjusted according to the specific implementation environment.

[0071] Taking the counterclockwise direction as the positive direction, the corresponding local direction angle is determined based on the minimum angle between the reference fitting line corresponding to each boundary pixel point and the horizontal right direction; the corresponding global direction angle is determined based on the minimum angle between the line containing the major axis of the minimum circumscribed rectangle of each edge tissue sub-block and the horizontal right direction; the absolute value of the local direction angle and the absolute value of the global direction angle are both less than 90 degrees. That is, the value range of the line is -90 degrees to 90 degrees, excluding the boundary value, and for the vertically upward line, its corresponding local direction angle or global direction angle is set to 90 degrees. By specifying the positive direction and the horizontal right direction as the angle reference standard, each line can uniquely correspond to a local direction angle or global direction angle, making the subsequent analysis process more robust.

[0072] The reference direction angle of each boundary pixel point is determined by taking the average of the local direction angle and the overall direction angle of each boundary pixel point; the sub-block extension angle of each edge organization sub-block is determined according to the average of the reference direction angles of all boundary pixels in each edge organization sub-block; and the sub-block extension direction of each edge organization sub-block is determined according to the sub-block extension angle. Considering that determining the extension direction of the edge organization sub-block only from the overall perspective will ignore the directional influence of local details, and integrating the extension direction of the edge organization sub-block only from the local perspective may result in a low accuracy of the final sub-block extension direction due to some accidental extreme structures; therefore, the present application combines the local and the overall perspectives, and uses the overall direction angle to correct the local direction angle of each boundary pixel point, thereby determining a more accurate sub-block extension direction of each edge organization sub-block according to the corrected reference direction angle of each boundary pixel point; the sub-block extension direction is also the direction corresponding to the straight line with the minimum angle between the counterclockwise direction and the horizontal right direction as the sub-block extension angle when the positive direction is the counterclockwise direction.

[0073] In each tissue region, the other edge tissue sub-blocks that are passed through a straight line that passes through the centroid of each edge tissue sub-block and is parallel to the major axis of the corresponding minimum circumscribed rectangle are used as reference tissue sub-blocks for each edge tissue sub-block; the local echo direction of each tissue sub-block is determined through structural tensor estimation; and the corresponding echo performance level is determined based on the overall deviation distribution between the sub-block extension direction of each edge tissue sub-block and the local echo direction of each reference tissue sub-block. Specifically, the angle between the sub-block extension direction of each edge tissue sub-block and the local echo direction of each reference tissue sub-block is used as the echo direction deviation value of each reference tissue sub-block; the overall echo deviation value is determined based on the mean of the echo direction deviation values ​​of the reference tissue sub-blocks; the corresponding echo direction disorder is determined based on the variance of the echo direction deviation values ​​of all reference tissue sub-blocks; and the product between the echo direction disorder and the overall echo deviation value is negatively correlated to determine the echo performance level of each edge tissue sub-block. It should be noted that structural tensor estimation is a technical means well known to those skilled in the art and will not be further described here.

[0074] For the ulnar nerve, its linear hyperechoic band must be consistent with the direction of nerve extension. Therefore, if the local echo direction of the reference sub-block is consistent with the sub-block extension direction, it indicates that the area conforms to the normal nerve structure; if it is inconsistent, it may be affected by abnormal compression. Therefore, the lower the echo direction disorder obtained according to the variance of the echo direction deviation value, and the smaller the overall deviation value of each echo direction, the more the tissue area conforms to the distribution pattern of the linear hyperechoic band of the ulnar nerve in the image, the stronger the corresponding echo performance, and the higher the possibility of belonging to the ulnar nerve area.

[0075] In a specific implementation of the embodiment of the present invention, the process of obtaining the echo performance level is expressed by the formula: ;in, Edge tissue sub-block The degree of echogenicity; Edge tissue sub-block The variance of the echo direction deviation values ​​of all reference tissue sub-blocks, that is, the echo direction disorder; Edge tissue sub-block The average of the echo direction deviation values ​​of all reference tissue sub-blocks is the overall echo deviation value.

[0076] In addition to the echogenicity, it's also important to consider the ulnar nerve's "hyperechoic fasciculations," essentially the ultrasound manifestation of the connective tissue of the perineurium and epineurium. This characteristic is the core basis for ultrasound identification of nerves and differentiation from other structures. However, when locating ulnar nerve compression injuries, the ulnar nerve may undergo compression, and when the ulnar nerve undergoes compression, the hyperechoic fasciculations at the localized point of compression disappear. Therefore, when determining whether a tissue region belongs to the ulnar nerve, it is necessary to reduce the weight of locations with smaller local tissue widths to more accurately calculate the probability of ulnar nerve involvement.

[0077] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the ulnar nerve probability includes:

[0078] In each edge tissue sub-block, a straight line passing through each boundary pixel point and perpendicular to the straight line corresponding to the reference direction angle of the boundary pixel point is used as the width direction straight line of each boundary pixel point; the minimum distance between the pixel points where the width direction straight line of each boundary pixel point intersects with the boundary of the tissue area is used as the local measured width of each boundary pixel point; and the corresponding local tissue width is determined based on the average of the local measured widths of all boundary pixel points in each edge tissue sub-block.

[0079] For each edge tissue subblock, since the reference direction angle represents the overall extension direction of the edge tissue subblock, the overall extension direction is used as a criterion for extracting the width value. This allows the resulting local measured width to more accurately represent the width of the local location. Furthermore, the overall width of the local measured widths of all boundary pixels is combined to determine the overall width of each edge tissue subblock, i.e., the local tissue width, by taking the mean. Based on the principle that smaller widths have a smaller weight when determining the ulnar nerve, the local weighted probability of each edge tissue subblock is determined by multiplying the normalized value of the local tissue width by the echogenicity level. It should be noted that the local tissue width is normalized using softmax normalization, so that the sum of the normalized values ​​of all local tissue widths in each tissue region is 1. Furthermore, the local weighted probabilities of all edge tissue subblocks are integrated, and the corresponding more accurate ulnar nerve probability is determined based on the cumulative value of the local weighted probabilities of all edge tissue subblocks in each tissue region.

[0080] In a specific implementation of the embodiment of the present invention, the process of obtaining the ulnar nerve probability is expressed by the formula: ;in, For the ulnar nerve probability per tissue region; For the The number of marginal tissue sub-blocks in each tissue region; For the The first The average of the local measurement widths of all boundary pixels in the edge tissue sub-block is the corresponding local tissue width; For the The first The echogenicity of the marginal tissue sub-blocks; is the softmax normalization function; For the The first The local weighted probability of the edge tissue sub-block.

[0081] Step S103: Filter out the ulnar nerve region based on the ulnar nerve probability; filter out the true entrapment region based on the relative distribution of the neighborhood widths of each edge tissue sub-block in the ulnar nerve region and the corresponding ulnar nerve probability; and locate the peripheral nerve injury in the hand based on the true entrapment region.

[0082] The tissue region corresponding to the ulnar nerve is further screened based on the ulnar nerve probability to accurately screen the actual entrapment area. The ulnar nerve region acquisition process includes: The ulnar nerve region acquisition process includes: The tissue region corresponding to the ulnar nerve probability greater than a preset ulnar nerve threshold is defined as the ulnar nerve region. In a specific implementation of an embodiment of the present invention, the preset ulnar nerve threshold is set to 0.6, which can be adjusted according to the specific implementation environment and is not further described here.

[0083] After determining the ulnar nerve region, the actual compression area needs to be determined based on the presence of ulnar nerve compression. Considering that when compression occurs, the ulnar nerve tissue area will narrow locally at the compression point, while the proximal nerve will swell due to obstructed axoplasmic transport and venous return. Therefore, after the ulnar nerve is compressed, the overall width of the compressed area will be relatively small, while the width of the area adjacent to the compressed area will be relatively large due to swelling. Therefore, based on this characteristic, the compression area can be located.

[0084] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the actual pressed area includes:

[0085] In each ulnar nerve region, k-means cluster analysis was performed on the local tissue widths of all edge tissue sub-blocks to obtain at least two tissue sub-block clusters. The corresponding width characteristic value was determined based on the mean of the local tissue widths of all edge tissue sub-blocks in each tissue sub-block cluster. When compression occurred, the width distribution had three clustering characteristics, including: the edge tissue sub-block directly affected by compression with the smallest overall width, the edge tissue block with the largest overall width due to compression and swelling, and the edge tissue block with normal width that was not affected by compression or was less affected by compression. In the present embodiment, the k-means cluster analysis uses a K value of 3, and analysis is performed based on width eigenvalues, distinguishing three regions with different width characteristics. Because the overall width of the actual compression region is relatively small, while the overall width of the region swollen due to compression is relatively large, and the compression effect is typically continuous, the adjacent edge tissue sub-blocks within the cluster with the largest width eigenvalue are merged to identify all suspected swollen regions. The adjacent edge tissue sub-blocks within the cluster with the smallest width eigenvalue are merged to identify all suspected compression regions. This allows for further screening based on the suspected compression regions, combined with the distribution of width during compression.

[0086] For the compression area, its compression will cause swelling in the adjacent area, which usually manifests as a suspected swelling area. In each ulnar nerve area, the nearest suspected swelling area corresponding to each suspected compression area is used as the corresponding reference influence area. The closer the suspected compression area is to the corresponding reference influence area, the more consistent it is with the situation that compression will cause swelling in the adjacent area for the suspected compression area, and the higher the possibility that it belongs to the real compression area. Therefore, the minimum distance between each suspected compression area and the reference influence area is further negatively correlated to determine the corresponding distance influence eigenvalue. The larger the distance influence eigenvalue, the higher the possibility that the corresponding suspected compression area is compressed, that is, the more likely it is to belong to the real compression area.

[0087] Based on the distance influence eigenvalue, the more edge tissue sub-blocks there are in the corresponding reference influence area, the more obvious the swelling characteristics of the reference influence area are under the corresponding compression influence; therefore, the embodiment of the present invention further normalizes the number of edge tissue sub-blocks in the reference influence area to determine the corresponding swelling persistence eigenvalue; so that the larger the swelling persistence eigenvalue, the higher the possibility that the corresponding suspected compression area is a real compression area.

[0088] For each suspected compression area, the greater the impact of compression, the smaller the overall tissue width of the corresponding suspected compression area will be, and the more likely it is to belong to the real compression area; therefore, the mean of the local tissue width of all edge tissue sub-blocks in each suspected compression area is further negatively correlated to determine the corresponding reference compression degree; so that the larger the reference compression degree, the higher the possibility that the corresponding suspected compression area belongs to the real compression area.

[0089] Finally, based on the correlation, the compression characteristic degree of each suspected compression area was determined according to the normalized value of the product between the swelling persistence characteristic value, the distance influence characteristic value and the reference compression degree.

[0090] In a specific implementation of the embodiment of the present invention, the process of obtaining the degree of the pressing feature is expressed by the formula: ;in, For the The degree of compression characteristics of the suspected compression area; For the The minimum distance between a suspected entrapment area and the corresponding reference impact area; is an exponential function with a natural constant as its base; For the The distance corresponding to the suspected compression area affects the characteristic value; For the The number of marginal tissue sub-blocks in the reference affected area corresponding to the suspected entrapment area; For the The swelling persistence characteristic value corresponding to the suspected compression area; For the The mean local tissue width of all marginal tissue sub-blocks in the suspected compression area; For the The reference compression degree corresponding to the suspected compression area.

[0091] The greater the degree of the pinching feature, the more likely the suspected pinching area is a true pinching area. Therefore, suspected pinching areas with a pinching feature greater than a preset pinching threshold are considered true pinching areas. In one specific implementation of this embodiment of the present invention, the preset pinching threshold is set to 0.8. This threshold can be adjusted based on the specific implementation environment and is not further described here.

[0092] After determining the true entrapment area, the nerve entrapment injury can be located by further marking the true entrapment area. In one specific implementation of the present invention, the minimum bounding rectangle of each true entrapment area in the ulnar nerve long-axis grayscale image is annotated as the nerve entrapment injury area. This minimum bounding rectangle allows for a more complete and standardized display of the included nerve entrapment injury area. Depending on the specific implementation environment, the implementer can also directly mark the true entrapment area as the nerve entrapment injury area. This will not be further explained here.

[0093] In summary, the hand peripheral nerve injury localization method based on medical imaging first divides the tissue areas for further screening according to the relatively uniform grayscale characteristics of the ulnar nerve area; then, based on the characteristics of the long strip echo structure of the ulnar nerve, the probability of the ulnar nerve is determined by calculating the echo performance of each sub-block in the tissue area, thereby preliminarily screening the ulnar nerve area; further, based on the characteristic that the width of the entrapment area will be suppressed under compression, a more accurate and real entrapment area is screened according to the relative distribution of the neighborhood width of the sub-blocks in the ulnar nerve area, so that the accuracy of hand peripheral nerve injury localization based on the real entrapment area is higher.

[0094] This application also provides a hand peripheral nerve injury positioning system based on medical imaging, please refer to Figure 2 , which shows a structural diagram of a hand peripheral nerve injury localization system based on medical imaging provided by an embodiment of the present invention. The system includes: a data acquisition and preprocessing module 201, an ulnar nerve probability determination module 202 and a hand peripheral nerve injury localization module 203.

[0095] The data acquisition and preprocessing module 201 is configured to acquire a long-axis grayscale image of the ulnar nerve through medical ultrasound imaging; and determine each edge tissue sub-block in each tissue region based on the local grayscale distribution in the long-axis grayscale image of the ulnar nerve;

[0096] The ulnar nerve probability determination module 202 is configured to determine the echogenicity of each edge tissue sub-block based on the echogenicity between each edge tissue sub-block and other edge tissue sub-blocks with similar edge extension trends; and to determine the corresponding ulnar nerve probability based on the width of each edge tissue sub-block and the overall echogenicity in each tissue region.

[0097] The hand peripheral nerve injury localization module 203 is used to filter out the ulnar nerve area based on the ulnar nerve probability; filter out the real entrapment area based on the relative distribution of the neighborhood width of each edge tissue sub-block in the ulnar nerve area and the corresponding ulnar nerve probability; and locate the hand peripheral nerve injury based on the real entrapment area.

[0098] It should be noted that the system provided in the above embodiment is merely an example of the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the above embodiment provides a hand peripheral nerve injury localization system based on medical imaging and a hand peripheral nerve injury localization method based on medical imaging. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0099] The present application also provides a computer device. Figure 3 , which shows a schematic diagram of the structure of a computer device provided by an embodiment of the present invention, the computer device includes a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302, wherein when the processor 302 executes the computer program 303, the computer device can execute any one of the medical image-based hand peripheral nerve injury localization methods introduced above.

[0100] An embodiment of the present application also provides a computer program product. When the computer program product is run on a computer device, the computer device can execute any one of the medical image-based hand peripheral nerve injury localization methods introduced above.

[0101] An embodiment of the present application also provides a computer-readable storage medium, which stores computer program code. When the computer program code runs on a computer device, the computer device can execute any of the medical imaging-based hand peripheral nerve injury localization methods introduced above.

[0102] In the embodiments provided in the present application, it should be understood that the provided computer devices, computer program products and computer-readable storage media are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the methods provided above and will not be repeated here.

[0103] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0104] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for locating peripheral nerve damage in the hand based on medical imaging, characterized in that: The method comprises: Acquiring a long-axis grayscale image of the ulnar nerve through medical ultrasound imaging; determining each edge tissue sub-block in each tissue region based on a local grayscale distribution in the long-axis grayscale image of the ulnar nerve; The echogenicity of each edge tissue sub-block is determined based on the similarity of the echogenicity directions between each edge tissue sub-block and other edge tissue sub-blocks with similar edge extension trends; the corresponding ulnar nerve probability is determined based on the width of each edge tissue sub-block and the overall size of the echogenicity in each tissue region; Filtering the ulnar nerve region based on the ulnar nerve probability; filtering the true entrapment region based on the relative distribution of the neighborhood widths of each edge tissue sub-block in the ulnar nerve region and the corresponding ulnar nerve probability; and locating the peripheral nerve injury in the hand based on the true entrapment region; The process of obtaining the real compression area includes: In each ulnar nerve region, a k-means cluster analysis was performed on the local tissue widths of all edge tissue sub-blocks to obtain at least two tissue sub-block clusters. The corresponding width eigenvalue was determined based on the mean of the local tissue widths of all edge tissue sub-blocks in each tissue sub-block cluster. Adjacent edge tissue sub-blocks in the tissue sub-block cluster with the largest width eigenvalue were merged to determine all suspected swelling areas. Adjacent edge tissue sub-blocks in the tissue sub-block cluster with the smallest width eigenvalue were merged to determine all suspected compression areas. In each ulnar nerve region, the nearest suspected swelling region corresponding to each suspected compression region is used as the corresponding reference impact region; the number of edge tissue sub-blocks in the reference impact region is normalized to determine the corresponding swelling persistence characteristic value; the minimum distance between each suspected compression region and the reference impact region is negatively correlated to determine the corresponding distance impact characteristic value; the mean of the local tissue width of all edge tissue sub-blocks in each suspected compression region is negatively correlated to determine the corresponding reference compression degree; the compression characteristic degree of each suspected compression region is determined based on the normalized value of the product of the swelling persistence characteristic value, the distance impact characteristic value and the reference compression degree; and the true compression region is screened out based on the compression characteristic degree.

2. The method for locating hand peripheral nerve damage based on medical imaging according to claim 1, characterized in that: The process of obtaining the edge tissue sub-block includes: Performing superpixel segmentation on the long axis grayscale image of the ulnar nerve to obtain at least two segmented tissue sub-blocks; determining the corresponding overall grayscale value according to the mean grayscale value of all pixels in each segmented tissue sub-block; The other segmented tissue sub-blocks outside each segmented tissue sub-block are used as corresponding contrast tissue sub-blocks; the corresponding cluster distance is determined according to the relative distance between each segmented tissue sub-block and each corresponding contrast tissue sub-block and the overall grayscale deviation; Perform cluster analysis based on the cluster distances between all segmented tissue sub-blocks to obtain at least two tissue sub-block clusters; The area corresponding to all the segmented tissue sub-blocks corresponding to each tissue sub-block cluster is taken as the corresponding tissue area; and each segmented tissue sub-block in each tissue area is taken as an edge tissue sub-block.

3. The method for locating hand peripheral nerve damage based on medical imaging according to claim 2, characterized in that: The process of obtaining the cluster distance includes: Determine a corresponding grayscale difference value according to a difference between the overall grayscale value of each segmented tissue sub-block and the overall grayscale value of each corresponding contrasted tissue sub-block; Determine the corresponding basic physical distance according to the Euclidean distance between the centroid of each segmented tissue sub-block and the centroid of each corresponding comparison tissue sub-block; The clustering distance between each segmented tissue sub-block and each corresponding contrasted tissue sub-block is determined according to a normalized value of the sum of the squares of the grayscale difference value and the basic physical distance.

4. The method for locating hand peripheral nerve damage based on medical imaging according to claim 1, characterized in that: The process of obtaining the echo performance level includes: In each edge tissue sub-block, a least squares straight line fitting is performed on each boundary pixel point and a preset number of adjacent edge pixel points to determine a reference fitting straight line for each boundary pixel point; Taking the counterclockwise direction as the positive direction, determine the corresponding local direction angle according to the minimum angle between the reference fitting line corresponding to each boundary pixel point and the horizontal right direction; determine the corresponding global direction angle according to the minimum angle between the line where the major axis of the minimum circumscribed rectangle of each edge tissue sub-block lies and the horizontal right direction; the absolute value of the local direction angle and the absolute value of the global direction angle are both less than 90 degrees; The reference direction angle of each boundary pixel point is determined by taking the average of the local direction angle of each boundary pixel point and the overall direction angle; the sub-block extension angle of each edge organization sub-block is determined according to the average of the reference direction angles of all boundary pixels in each edge organization sub-block; and the sub-block extension direction of each edge organization sub-block is determined according to the sub-block extension angle; In each tissue region, the other edge tissue sub-blocks that are passed through a straight line that passes through the centroid of each edge tissue sub-block and is parallel to the major axis of the corresponding minimum circumscribed rectangle are used as reference tissue sub-blocks for each edge tissue sub-block. The local echo direction of each tissue sub-block is determined by structural tensor estimation. The corresponding echo performance level is determined based on the overall deviation distribution between the sub-block extension direction of each edge tissue sub-block and the local echo directions of each reference tissue sub-block.

5. The method for locating hand peripheral nerve damage based on medical imaging according to claim 4, characterized in that: The process of determining the corresponding echo performance level according to the overall deviation distribution between the sub-block extension direction of each edge tissue sub-block and the local echo direction of each reference tissue sub-block includes: The angle between the sub-block extension direction of each edge tissue sub-block and the local echo direction of each reference tissue sub-block is used as the echo direction deviation value of each reference tissue sub-block; and the overall echo deviation value is determined according to the average of the echo direction deviation values ​​of the reference tissue sub-blocks; Determining the corresponding echo direction disorder according to the variance of the echo direction deviation values ​​of all reference tissue sub-blocks; A negative correlation mapping is performed on the product of the echo direction disorder degree and the overall echo deviation value to determine the echo performance degree of each edge tissue sub-block.

6. The method for locating hand peripheral nerve damage based on medical imaging according to claim 4, characterized in that: The process of obtaining the ulnar nerve probability includes: In each edge tissue sub-block, a straight line passing through each boundary pixel and perpendicular to the straight line corresponding to the reference direction angle of the boundary pixel is used as the width direction straight line of each boundary pixel; The minimum distance between the pixel points where the width-direction straight line of each boundary pixel intersects the boundary of the tissue region is used as the local measured width of each boundary pixel; the corresponding local tissue width is determined based on the average of the local measured widths of all boundary pixels in each edge tissue sub-block; determining a local weighted probability of each edge tissue sub-block according to a product of a normalized value of the local tissue width and the echo performance degree; The corresponding ulnar nerve probability is determined according to the accumulated value of the local weighted probabilities of all edge tissue sub-blocks in each tissue region.

7. The method for locating hand peripheral nerve damage based on medical imaging according to claim 1, characterized in that: The process of obtaining the ulnar nerve area includes: The tissue area whose corresponding ulnar nerve probability is greater than the preset ulnar nerve threshold is regarded as the ulnar nerve area.

8. The method for locating hand peripheral nerve damage based on medical imaging according to claim 1, characterized in that: The process of screening out the real compression area according to the compression feature degree includes: The suspected pressed area whose corresponding pressed feature degree is greater than the preset pressed threshold is regarded as the real pressed area.

9. The method for locating hand peripheral nerve damage based on medical imaging according to claim 1, characterized in that: The process of locating hand peripheral nerve injury according to the actual compression area includes: In the ulnar nerve long axis grayscale image, the minimum circumscribed rectangle of each real compression area is marked as the nerve compression injury area.

Images