Method for optimizing and enhancing big dada in medical health physical examination

Abstract

A method for optimizing and enhancing big data in medical examination is provided. All closed edges and all feature corner points in an anteroposterior spinal X-ray image are obtained and an internal

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to Chinese Patent Application No. 202410185725.6, filed on February 20, 2024, which is herein incorporated by reference in its entirety. TECHNICAL FIELD

[0002] The disclosure relates to the field of image local extraction and enhancement technology, and more particularly to a method for optimizing and enhancing big data in medical health physical examination. BACKGROUND

[0003] Medical health physical examinations may help people detect potential health problems early through a series of medical tests and evaluations, so that timely measures may be taken to prevent the occurrence and development of diseases. Spinal examination is an important examination item in the medical health physical examinations. A spine is a supporting structure of a human body, and is responsible for protecting spinal cord and nerve roots. Spinal problems may lead to pain, functional impairment and other health problems. It is very important to detect and treat spinal problems early.

[0004] During the spinal examination. X-ray imaging of the spine is commonly used to assist medical personnel in evaluating spinal health. However, imaging effect of the spine may deteriorate due to the overlapping of complex tissue structures such as internal organs and bones in the human body, and structural features of the spine cannot be clearly observed, which in turn affects the health assessment. Therefore, local enhancement of a spinal region is very important. Even though deep learning algorithms may be used for the extraction of the spinal structural featuresl, annotating large-scale datasets is a laborious and time-consuming process, and is easily affected by the quality of X-ray image, resulting in low quality and efficiency in the extraction of the spinal region, thereby affecting the subsequent enhancement effect and reducing the accuracy of spinal examination evaluation. Consequently, in the process of optimizing and enhancing big data in medical health physical examinations, the accurate extraction of the spinal region is a technical problem that needs to be addressed. SUMMARY

[0005] In order to solve the technical problem of poor extraction quality of a spinal region in an optimization and enhancement process of big data in medical health physical examination, an objective of the disclosure is to provide a method for optimizing and enhancing big data in medical health physical examination. The technical solution adopted is specifically as follows.

[0006] The disclosure provides a method for optimizing and enhancing big data in the medical health physical examination, including: obtaining all closed edges and all feature comer points in an anteroposterior spinal X-ray image of a health examinee; obtaining an internal difference coefficient of each closed edge relative to each other closed edge based on a positional distribution difference coefficient of all feature comer points within each closed edge relative to all feature comer points within each other closed edge, obtaining a vertebral body edge probability of each closed edge being a spinal vertebral body based on a rectangularity degree of each closed edge and the internal difference coefficient of each closed edge relative to each other closed edge, and obtaining a vertebral body edge confidence probability of each closed edge being the spinal vertebral body based on a number of all feature comer points within each closed edge and the vertebral body edge probability of each closed edge; and selecting a reference vertebral body edge from the all closed edges based on the vertebral body edge confidence probability of each closed edge, obtaining all spinal vertebral bodies by a feature matching algorithm based on feature comer points within a region enclosed by the reference vertebral body edge and feature corner points within a region enclosed by all other closed edges, and performing grayscale enhancement on the all spinal vertebral bodies.

[0007] In an embodiment, the obtaining an internal difference coefficient of each closed edge includes: with any one of the all closed edges as a target closed edge and any one of all feature corner points within the target closed edge as a target feature comer point, obtaining a positional distribution difference coefficient of the target feature comer point relative to each feature corner point within all other closed edges except the target closed edge, and using a feature corner point with a smallest positional distribution difference coefficient relative to the target feature corner point within each closed edge except the target closed edge as a suspected matching corner point for the target feature corner point in each closed edge. A suspected matching corner point within each closed edge except the target closed edge corresponding to each feature comer point within the target closed edge is obtained, a mean value of positional distribution difference coefficients between the all feature corner points within the target closed edge and corresponding suspected matching corner points within a corresponding closed edge except the target closed edge is taken as the internal difference coefficient between the target closed edge and the corresponding closed edge.

[0008] In an embodiment, a calculation formula of the positional distribution difference coefficient is as follows: Fus,vr 7 Oyj + 7 X bus bvr , where Fus,vr represents a positional distribution difference coefficient between an 5-th target feature comer point within a z / -th target closed edge and a / -th feature corner point within a v-th closed edge, Lu represents a number of edge pixel points of the / / -th target closed edge, Lv represents a number of edge pixel points of the v-th closed edge, aus represents a vector from the 5-th target feature corner point within the / / -th target closed edge to a nearest edge pixel point from the 5-th target feature corner point on a closed region enclosed by the / / -th target closed edge, avr represents a vector from the r-th target feature corner point within the v-th closed edge to a nearest edge pixel point from the r-th target feature comer point on a closed region enclosed by the v-th closed edge, bus represents a vector from the 5-th target feature corner point within the ? / -th target closed edge to a farthest edge pixel point from the s-th target feature comer point on the closed region enclosed by the / / -th target closed edge, and bvr represents a vector from the r-th target feature corner point within the v-th closed edge to a farthest edge pixel point from the r-th target feature corner point on the closed region enclosed by the v-th closed edge.

[0009] In an embodiment, the obtaining a rectangularity degree includes following steps: with any one of the edge pixel points on the closed edge is taken as a starting point, an 8-chain code corresponding to the closed edge is obtained, an occurrence frequency of each of chain code values in the 8-chain code corresponding to the closed edge is counted, the occurrence frequencies of the chain code values are sorted in a descending order, a first sorted chain code value and a second sorted chain code value are taken as a first group of chain code values, a third sorted chain code value and a fourth chain code value are taken as a second group of chain code values, a confidence coefficient for each of the first group of chain code values and the second group of chain code values being chain code values corresponding to a pair of rectangular opposite sides is obtained based on a chain code value difference in each of the first group of chain code values and the second group of chain code values and an edge pixel point number difference corresponding to chain code values of each of the first group of chain code values and the second group of chain code values on the closed edge, and the confidence coefficients for the first group of chain code values and the second group of chain code values are multiplied to obtain the rectangularity degree of the closed edge.

[0010] In an embodiment, a calculation formula of the confidence coefficient is as follows: (lg _g l\ -1............. r...............-) X exp (-|DZ1 - Dz2 - a|), where Au.z represents a confidence coefficient of a z-th group of chain code values being chain code values corresponding to a pair of the rectangular opposite sides within a w-th closed edge, Bzi represents a number of edge pixel points corresponding to a chain code value with a highest occurrence frequency in the z-th group of chain code values, BZ2 represents a number of edge pixel points corresponding to a chain code value with a lowest occurrence frequency in the z-th group of chain code values, C: represents a number of all edge pixel points corresponding to the z-th group of chain code values, Dzi represents the chain code value with the highest occurrence frequency in the z-th group of chain code values, DZ2 represents the chain code value with the lowest occurrence frequency in the z-th group of chain code values, a represents a preset positive integer, and exp () represents an exponential function with a natural constant e as a base.

[0011] In an embodiment, a calculation formula of the vertebral body edge probability is as follows: — * exp ^uv)L where Hu represents a vertebral body edge probability of the / / -th closed edge, Eu represents a rectangularity degree of the w-th closed edge, Ev represents a rectangularity degree of the v-th closed edge, Muv represents an internal difference coefficient between the w-th closed edge and the v-th closed edge, N represents a number of all other closed edges except the w-th closed edge in the anteroposterior spinal X-ray image, and exp () represents an exponential function with the natural constant e as the base.

[0012] In an embodiment, the obtaining a vertebral body edge confidence probability of each closed edge includes following steps: normalizing the number of the all feature corner points within the closed edge in a positive correlation to obtain a normalized value, and the normalized value is multiplied by the vertebral body edge probability of the closed edge to obtain the vertebral body edge confidence probability of the closed edge.

[0013] In an embodiment, the obtaining all closed edges includes following steps: performing double-threshold edge detection on the anteroposterior spinal X-ray image to obtain all edges, and obtaining the all closed edges in all edges based on an edge tracking algorithm.

[0014] In an embodiment, the feature matching algorithm is a scale-invariant feature transform (SIFT) algorithm.

[0015] In an embodiment, the performing the grayscale enhancement on the all spinal vertebral bodies includes: performing histogram equalization on each spinal vertebral body individually.

[0016] The disclosure has following beneficial effects.

[0017] In the disclosure, all closed edges and all feature corner points in the anteroposterior spinal X-ray image of the health examinee are firstly obtained, so as to facilitate subsequent analysis of the structural similarity of each closed edge based on the structural features of the spinal vertebral bodies. Then, the internal difference coefficient of each closed edge relative to each other closed edge is obtained based on the positional distribution difference coefficient of each feature comer point within each closed edge relative to each feature corner point within each other closed edge, and subsequently the vertebral body edge probability of each closed edge being the spinal vertebral body is obtained by combining the rectangularity degree of each closed edge. The vertebral body edge probability fully considers a rectangularity appearance of each spinal vertebral body in the anteroposterior spinal X-ray image and similar gradient changes in its internal feature structure caused by the tissue structure. The internal difference coefficient between closed edges is obtained based on the rectangularity degree of each closed edge and the positional distribution difference coefficients of all feature comer points which reflect gradient changes within the internal regions, and the probability that a closed edge corresponds to a spinal vertebral body is comprehensively evaluated. Furthermore, credibility of the internal difference coefficient is evaluated by combining the number of the all feature corner points within each closed edge to obtain the vertebral body edge confidence probability, thereby selecting the reference vertebral body edge. Then, the all spinal vertebral bodies are obtained by the feature matching algorithm and the grayscale enhancement is performed on the all spinal vertebral bodies. In the disclosure, structural feature information of the all spinal vertebral bodies is fully utilized, the possibility of each closed edge being the vertebral body edge in the anteroposterior spinal X-ray image is analyzed, the most likely reference vertebral body is selected, and all other spinal vertebral bodies are obtained based on the feature matching algorithm. The quality of spinal region extraction for precise local enhancement and optimization is improved, which improves the optimization and enhancement effect of big data in the medical health physical examination. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the disclosure or the related art, the accompanying drawings required in a description of the embodiments or the related art are briefly introduced below. It is apparent that the accompanying drawings in the following description are only some embodiments of the disclosure. For those skilled in the art, other accompanying drawings may be obtained based on these drawings without creative labor.

[0019] FIG. 1 illustrates a method flowchart of a method for optimizing and enhancing big data in medical health physical examination according to an embodiment of the disclosure.

[0020] FIG. 2 illustrates an anteroposterior spinal X-ray image according to an embodiment of the disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] In order to further elaborate on the technical means and effects adopted by the disclosure to achieve the intended objectives, the specific implementations, structures, features, and effects of a method for optimizing and enhancing big data in medical health physical examination provided by the disclosure are described in detail as follows, combined with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" may not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art of the disclosure.

[0023] The technical solutions of the method for optimizing and enhancing big data in the medical health physical examination of the disclosure are specifically described below in conjunction with the accompanying drawings.

[0024] Referring to FIG. 1, FIG. 1 illustrates a flowchart of the method for optimizing and enhancing big data in the medical health physical examination according to an embodiment of the disclosure. The method includes following steps SI-S3.

[0025] The disclosure is aimed at optimizing and enhancing anteroposterior spinal X-ray images of examinees. Therefore, in the embodiment of the disclosure, all closed edges and feature corner points in an anteroposterior spinal X-ray image of a health examinee are firstly obtained; then, a rectangularity degree and an internal difference coefficient of each closed edge are analyzed and a vertebral edge confidence probability for each closed edge is obtained by combining approximately rectangular structural characteristics of spinal vertebral bodies and similar tissue structure characteristics of each vertebral body. And thus, a reference vertebral body is obtained, and all spinal vertebral bodies are obtained by feature matching. Local enhancement is performed on the all spinal vertebral bodies in the anteroposterior spinal X-ray image, making the spinal structural features clearer, thereby assisting medical personnel in accurate assessment.

[0026] In step SI, all closed edges and the all feature corner points are obtained in the anteroposterior spinal X-ray image of the health examinee.

[0027] The anteroposterior spinal X-ray image in the embodiment of the disclosure is a spinal X-ray image collected from a back of the examinee. No further grayscale processing is required for considering that the spinal X-ray image itself is already a grayscale image. A method for acquiring the anteroposterior spinal X-ray image is already a prior art well-known to those skilled in the art and is not elaborated here.

[0028] To facilitate understanding of the solutions in the embodiments of the disclosure, referring to FIG. 2, FIG. 2 illustrates an anteroposterior spinal X-ray image provided by an embodiment of the disclosure. The nodular structure in FIG. 2 is a human spine. The human spine includes spinal vertebral bodies and intervertebral discs, and is composed of the vertebral bodies connected by the intervertebral discs. The spinal vertebral bodies are mainly composed of high-density bone tissue, and the intervertebral discs are composed of lower-density cartilage tissue and fluid, and the imaging principle of X-ray images is based on tissue density, the higher the density, the higher the gray value in the grayscale image. Therefore, in the anteroposterior spinal X-ray image, the spinal vertebral bodies appear as individual, rectangle-like nodular structures, while the intervertebral discs are darker areas between adjacent spinal vertebral bodies. The spinal vertebral bodies also include spinous processes, vertebral arch pedicles, and marrow cavities. The marrow cavities usually appear as darker areas relative to the spinous processes and the vertebral arch pedicles in the anteroposterior spinal X-ray image, and thus there is a significant gradient at a junction of a marrow cavity and the spinous processes and the vertebral arch pedicles within a corresponding rectangle-like area corresponding to each spinal vertebral body.

[0029] Each spinal vertebral body mainly appears as a closed edge which approximates a rectangle in the anteroposterior spinal X-ray image. There is a significant gradient change at the junction of the marrow cavity with the spinous process and the vertebral arch pedicles, a tissue structure between spinal vertebral bodies in a spinal region has similar gradient change features, and the feature comer points may provide stable distinctive feature information under different scale changes, which may accurately match the similar characteristic gradient changes of images at different scales. Therefore, the embodiment of the disclosure obtains all closed edges and all feature corner points in the anteroposterior spinal X-ray image of the health examinee, to combine the rectangularity degree of each closed edge and the distribution of the feature corner points within each closed edge to judge the structural similarity of each closed edge, and then to judge the possibility of each closed edge being a vertebral body edge. Further, the feature comer points also facilitate subsequent selection of the reference vertebral body for feature matching to obtain all the spinal vertebral bodies in the entire spinal region. In one embodiment of the disclosure, all feature comer points are specifically obtained through a scaleinvariant feature transform (SIFT) algorithm.

[0030] In an embodiment of the disclosure, the obtaining the closed edges includes: performing double-threshold edge detection on the anteroposterior spinal X-ray image to obtain all edges, and obtaining the all closed edges based on an edge tracking algorithm. The anteroposterior spinal X-ray image is subjected to double-threshold edge detection by a Canny edge detection operator , where a low threshold is set to 3, and a high threshold is set to 5. Setting double thresholds may better capture texture features in the image while suppressing the weak edges and noise . Then, the edge tracking algorithm is used to determine whether the edges are closed, so as to obtain all closed edges that may be spinal vertebral bodies.

[0031] It should be noted that, the SIFT algorithm, the double-threshold edge detection and the edge tracking algorithm are all prior art knownto those skilled in the art and will not be repeated here. In other embodiments of the disclosure, an implementer may use other methods to obtain the closed edges and the feature comer points according to specific implementation situations.

[0032] In step S2, the internal difference coefficient of each closed edge relative to each other closed edge is obtained based on a positional distribution difference coefficient of all feature corner points within each closed edge relative to all feature comer points within each other closed edge, a vertebral body edge probability that each closed edge is the spinal vertebral body is obtained based on the rectangularity degree of each closed edge and the internal difference coefficient of each closed edge relative to each other closed edge, and a vertebral body edge confidence probability that each closed edge is the spinal vertebral body is obtained based on a number of all feature corner points within each closed edge and the vertebral body edge probability of each closed edge.

[0033] There is the significant gradient change at the junction of the marrow cavity with the spinous processes and the vertebral arch pedicles, the tissue structure between spinal vertebral bodies in the spinal region has the similar gradient change features, and the all feature comer points obtained in the step SI may greatly reflect gradient change features at the junction of the marrow cavity with the spinous processes and the vertebral arch pedicles. Since each spinal vertebral body hass basically the same structure features regardless of size, and the gradient change feature at the junction of the spinal vertebral body with the spinous processes and the vertebral arch pedicles is also basically similar, and within the rectangle-like closed edge corresponding to each spinal vertebral body in the anteroposterior spinal X-ray image, the feature comer points should exhibit similar positional distribution features. Therefore, in the embodiment of the disclosure, the internal difference coefficient of each closed edge relative to each other closed edge is obtained based on the positional distribution difference coefficient of each feature comer point within each closed edge relative to each feature comer point within each other closed edge. The internal difference coefficient indirectly reflects the similarity between closed edges, which is convenient for subsequent judgment of the possibility of each closed edge being the spinal vertebral body based on the r ectangularity degree of each closed edge.

[0034] In an embodiment of the disclosure, it is considered that the feature corner points are pixel points reflecting distinctive feature information within the closed edges and the positional distribution of the feature corner points within two closed edges may indirectly reflect internal differences between the two closed edges. The obtaining the internal difference coefficient of each closed edge includes: with any one of the all closed edges taken as a target closed edge and any one of all feature corner points within the target closed edge as a target feature comer point, obtaining a positional distribution difference coefficient of the target feature comer point relative to each feature corner point within all other closed edges except the target closed edge, a feature corner point with a smallest positional distribution difference coefficient relative to the target feature comer point within each closed edge except the target closed edge is taken as a suspected matching corner point for the target feature corner point in each closed edge, a suspected matching comer point within each closed edge except the target closed edge corresponding to each feature comer point within the target closed edge is obtained, a mean value of positional distribution difference coefficients between the all feature comer points within the target closed edge and corresponding suspected matching corner points within a corresponding closed edge except the target closed edge is taken as the internal difference coefficient between the target closed edge and the corresponding closed edge.

[0035] It should be noted that the suspected matching comer point is selected based on the position distribution difference coefficients between the pixel points. There is a possibility that suspected matching comer points in other closed edges corresponding to different feature corner points within the target closed edge are a same feature corner point, i.e., the suspected matching corner points are just a group of feature comer points that may be a pair of matching corner points with small positional distribution difference, and are not actual matching comer points.

[0036] In an embodiment of the disclosure, considering that there are corresponding feature corner points within a closed edge of each candidate vertebral body, the orientation and distance of each feature comer relative to edge pixel points on its corresponding closed edge may well reflect the positional distribution features of each feature corner point, and vectors may reflect orientation and distance. Since calculating the orientation and distance of each feature corner point relative to each pixel point on its corresponding closed edge would consume a lot of computational resources, two edge pixel points on the corresponding closed edge that are closest and farthest from the feature corner point are selected. The difference between the vectors from the feature comer point to the two edge pixel points is calculated to reflect the positional distribution features of the feature comer point. Furthermore,considering that different spinal vertebral bodies have similar structures but have different sizes and contours, and the circumference of each closed edge may approximately reflect its size and contour, the number of the edge pixel points on each closed edge is obtained as a corresponding size and contour ratio. This avoids inaccurate calculations of the positional distribution differences due to inconsistent vector magnitudes caused by different sizes of spinal vertebral bodies. Accordingly, a calculation formula of the positional distribution difference coefficient is as follows: C — y n _ n y h — h rus,vr >^us uvr uus uvr h ■ *Lu I where Fus,vr represents a positional distribution difference coefficient between a 5-th target feature comer point within a w-th target closed edge and a r-th feature corner point within a v-th closed edge, Lu represents a number of the edge pixel points of the w-th target closed edge, Lv represents a number of the edge pixel points of the v-th closed edge, aus represents a vector from the 5-th target feature comer point within the z / -th target closed edge to a nearest edge pixel point from the 5-th target feature comer point on a closed region enclosed by the w-th target closed edge, avr represents a vector from the r-th target feature corner point within the v-th closed edge to a nearest edge pixel point from the r-th target feature corner point on a closed region enclosed by the v-th closed edge, bus represents a vector from the 5-th target feature comer point within the w-th target closed edge to a farthest edge pixel point from the 5-th target feature comer point on the closed region enclosed by the w-th target closed edge, and bvr represents a vector from the r-th target feature corner point within the v-th closed edge to a farthest edge pixel point from the r-th target feature corner point on the closed region enclosed by the v-th closed edge.

[0037] In the calculation formula of the positional distribution difference coefficient, — reflects a size and contour ratio between two closed edges, the vector is adjusted by the ratio of the number of the edge pixel points on the corresponding closed edges. Vector difference modulus of the 5-th target feature comer point in the / / -th target closed edge and the r-th feature corner point in the v-th closed edge to the nearest edge pixel point and the farthest edge pixel point on the corresponding closed edge respectively are calculated. When the vectors of the two feature corner points to the nearest edge pixel point and the farthest edge pixel point on the corresponding closed edge respectively in the two closed edges are more similar, the sum of the vector difference modulus is smaller, indicating that the position distribution features of the two feature comer points in the corresponding closed edges are more similar, and the two feature corner points are more likely to be a pair of matched comer points in the two closed edges.

[0038] It should be noted that, the vector calculation application is prior art well known to those skilled in the art, and will not be repeated here.

[0039] It is considered that the spinal vertebral body mainly appears as an approximately rectangular closed edge in the anteroposterior spinal X-ray image, thus, after obtaining the internal difference coefficient of each closed edge relative to other closed edges, the embodiment of the disclosure further obtains the vertebral body edge probability of each closed edge based on the rectangularity degree of each closed edge and the internal difference coefficient of each closed edge relative to all other closed edges.

[0040] Before obtaining the vertebral body edge probability of each closed edge, the rectangularity degree of each closed edge is first analyzed.

[0041] In an embodiment of the disclosure, considering that chain codes are commonly used to describe curves or boundary contours, in a rectangle, starting from any boundary point on any side of the rectangle, each boundary point on each side corresponds to a chain code value. The direction of the boundary points on each side is consistent, and the directions of the opposite sides are opposite, which is reflected in certain differences in the chain code values, and the number of the boundary points corresponding to the chain code values on the opposite sides should be consistent. Accordingly, the obtaining the rectangularity degree includes: with any one of the edge pixel points on the closed as a starting point, obtaining an 8-chain code corresponding to the closed edge, counting an occurrence frequency of each chain code value in the 8-chain code corresponding to the closed edge, sorting the occurrence frequencies in a descending order, with a first sorted chain code value and a second sorted chain code value as a first group of chain code values, a third sorted chain code value and a fourth chain code value as a second group of chain code values, obtaining a confidence coefficient that chain code values of each group are chain code values corresponding to a pair of rectangular opposite sides based on a chain code value difference in each of the first group of chain code values and the second group of chain code values and an edge pixel point number difference corresponding to chain code values of each of the first group of chain code values and the second group of chain code values on the closed edge, and multiplying the confidence coefficients for the first group of chain code values and the second group of chain code values to obtain the rectangularity degree of the closed edge. On the closed edge, each edge pixel point corresponds to a chain code value of 0-7, the top four chain code values are most likely to be suspected to be the chain code values corresponding to the pixel points on four sides respectively of the rectangle. After obtaining the first group of chain code values suspected to be the corresponding chain code values of long sides of the rectangle and the second group of chain code values suspected to be the corresponding chain code values of short sides of the rectangle, the confidence coefficient for each group of chain code values corresponding to the opposite sides of the rectangle may be calculated individually by combining relevant feature information of the 8-chain code of the rectangle, to determine its rectangularity degree. The 8-chain code is well-known in the related art, which is not elaborated here.

[0042] To facilitate understanding the analysis of the rectangularity degree of each closed edge, a rectangle is taken as an example, two longer sides in a horizontal direction are the long sides of the rectangle and other two are the short sides, the corresponding 8-chain code for the rectangle is obtained by starting from any point on the long sides of the rectangle and moving clockwise, and ultimately four types of the chain code values are statistically obtained, namely 0, 6, 4, and 2. The chain code values 0 and 4 correspond to the long sides, and the chain code values 6 and 2 correspond to the short sides. Each group of the opposite sides has a chain code value difference of 4 to represent two opposite directions, and the number of the edge pixel points corresponding to each group of the chain code values is consistent, i ,e„ lengths of the opposite sides are the same. Therefore, in the process of analyzing the rectangularity degree, the top two frequently occurring chain code values may be a group of the chain code values corresponding to the long sides of the rectangle, and the third and fourth frequently occurring chain code values may be a group of the chain code values corresponding to the short sides of the rectangle. By analyzing the difference in the chain code values corresponding to each group of opposite sides and the difference in the number of corresponding pixel points, the rectangularity degree of the corresponding closed edge may be judged.

[0043] In an embodiment of the disclosure, considering that in the 8-chain code representation, each pair of opposite sides of the rectangle corresponds to the chain code values in opposite direction, i.e., the chain code values differ by 4, the lengths of each pair of the opposite sides of the rectangle are the same, and the occurrence frequency of each group of chain code values, which corresponds to the number of pixel points, should be roughly the same, a calculation formula of the confidence coefficient is as follows: Au,z = exp (- x exp (-|DZ1 - Dz2 - a|) , where Ju,z represents a confidence coefficient of a z-th group of chain code values being chain code values corresponding to a pair of the rectangular opposite sides within a w-th closed edge, Bzi represents a number of edge pixel points corresponding to a chain code value with a highest occurrence frequency in the z-th group of chain code values, BZ2 represents the number of the edge pixel points corresponding to a chain code value with a lowest occurrence frequency in the z-th group of chain code values, C- represents a number of all edge pixel points corresponding to the z-th group of chain code values, D : represents the chain code value with the highest occurrence frequency in the z-th group of chain code values, DZ2 represents the chain code value with the lowest occurrence frequency in the z-th group of chain code values, a represents a preset positive integer, and exp () represents an exponential function with a natural constant e as a base. In the embodiment of the disclosure, the preset positive integer is 4, due to a fact that the direction of the chain code values for each pair of opposite sides of the rectangle is opposite, i.e., the chain code values differ by 4.

[0044] In the calculation formula of the confidence coefficient, \Bzi~BZ2\ represents the difference in the occurrence frequency of the z-th group of chain code values, i.e., difference in the number of pixel points corresponding to the z-th group of the chain code values, which is divided by the number of all edge pixel points Cz corresponding to the z-th group of chain code values for normalization. The smaller the difference after normalization, the more similar the candidate rectangle sides corresponding to the z-th group of chain code values in length, which indicates that the z-th group of chain code values is more likely to be the chain code values corresponding to a pair of opposite sides of the candidate rectangle. \Dzi~DZ2~a\ reflects the direction features of the z-th group of chain code values, the closer the difference between the chain code values in each group is to 4, the z-th group of chain code values corresponding to the candidate rectangle sides is parallel but opposite in direction, which indicates that the z-th group of the chain code values is also more likely to be the chain code values corresponding to a pair of the opposite sides of the candidate rectangle. Both the \Bzi~BZ2\ and the \Dzi-DZ2~a\ are negatively correlatedly mapped into the exponential function for normalization, and then multiplied and merged to obtain the confidence coefficient. The larger the confidence coefficient ofchain code values in the z-th group within the w-th closed edge, the higher the possibility that chain code values in the z-th group are the chain code values corresponding to a pair of the opposite sides of the candidate rectangle in the w-th closed edge, and the higher the rectangularity degree of the w-th closed edge.

[0045] After obtaining the internal difference coefficient of each closed edge relative to other closed edges and the rectangularity degree of each closed edge, the vertebral body edge probability of each closed edge may be obtained by further combining rectangularity features and features similar to combined structure features of external vertebral body edges.

[0046] In an embodiment of the disclosure, considering that the more similar the internal differences between each closed edge and other closed edges, the more likely it is that the closed edge and other closed edges correspond to the spinal vertebral body, and the greater the rectangularity degree of the closed edge, the greater the likelihood that the closed edge corresponds to the spinal vertebral body, based on this, a calculation formula of the vertebral body edge probability is as follows: Hu= Eux Yv=itEv x exp (—Muv)], where Hu represents the vertebral body edge probability of the w-th closed edge, Eu represents the rectangularity degree of the w-th closed edge, Ev represents the rectangularity degree of the v-th closed edge, Muv represents the internal difference coefficient between the u-th closed edge and the v-th closed edge, N represents a number of all other closed edges except the w-th closed edge in the anteroposterior spinal X-ray image, and exp () represents an exponential function with the natural constant e as the base.

[0047] In the calculation formula of the vertebral body edge probability, the internal difference coefficient between the zz-th closed edge and other non-w closed edges are negatively correlatedly mapped into the exponential function for normalization. Then, the rectangularity degree of each other closed edge is used as a weight of the corresponding internal difference coefficient. The structural similarity between the w-th closed edge and all other closed edges is comprehensively evaluated, and the higher the structural similarity, the more likely the w-th closed edge is to be the vertebral body edge. Meanwhile, the larger the rectangularity degree of the w-th closed edge, the more likely it is that the w-th closed edge is the vertebral body edge.

[0048] It is considered that the internal difference coefficient of each closed edge relative to each other closed edge is obtained based on the position distribution difference coefficients of the all feature corner points within each closed edge, there is a possibility that the internal difference coefficient of the closed edge relative to each other closed edge is small due to fewer feature corner points within the closed edge, which in turn makes evaluation of the vertebral body edge probability of the closed edge inaccurate. Therefore, in the embodiment of the disclosure, the vertebral body edge confidence probability that each closed edge is the spinal vertebral body is obtained based on the number of the feature corner points within each closed edge and the corresponding vertebral edge probability.

[0049] In an embodiment of the disclosure, it is considered that the vertebral body edge probability obtained only based on the internal difference coefficient and the rectangularity degree of the closed edge is not accurate, and it is further considered that the higher the number of feature comer points within the closed edge, the more credible the evaluation of the internal difference coefficient, and thus the more credible the evaluation of the vertebral body edge probability. Based on this, the number of the feature comer points within each closed edge is normalized in a positive correlation, and the normalized value is multiplied by the vertebral body edge probability of the corresponding closed edge to obtain the vertebral body edge confidence probability of each closed edge. A calculation formula of the vertebral body edge confidence probability is as follows: Pu = [1 - exp(-Ku)] x Hu, where Pu represents the vertebral body edge confidence probability of the w-th closed edge, Hu represents the vertebral body edge probability of the w-th closed edge, Ku represents the number of the feature corner points within the w-th closed edge, and exp () represents an exponential function with the natural constant e as the base.

[0050] In the calculation formula of the vertebral body edge confidence probability, first, the number of the feature comer points within the w-th closed edge is negatively correlatedly mapped into the exponential function for normalization. Then, the corresponding logical relationship is adjusted by subtracting the normalized value from 1 and used as a confidence weight, so that the number of the feature corner points is positively correlated with the vertebral body edge confidence probability and normalized. The more the number of the feature comer points, the greater the corresponding confidence weight. Then, the vertebral body edge confidence probability is obtained by multiplying the confidence weight with the vertebral body edge probability. In other embodiments of the disclosure, other positive correlation mapping methods may also be adopted, which are not limited or elaborated here.

[0051] In step S3, a reference vertebral body edge is selected from the all closed edges based on the vertebral body edge confidence probability, all spinal vertebral bodies are obtained by a feature matching algorithm based on all feature comer points in the region enclosed by the reference vertebral body edges and in the region enclosed by all other closed edges, and grayscale enhancement is performed on the all vertebral bodies.

[0052] In the step S2, the confidence probability that all closed edges in the anteroposterior spinal X-ray image are vertebral body edges is obtained by combining the structural features of the spinal vertebral body. Then, the reference vertebral body edge may be selected among all closed edges based on the vertebral body edge confidence probability, thereby obtaining all the spinal vertebral bodies on the spine for local enhancement and optimization.

[0053] In an embodiment of the disclosure, it is considered that the higher the vertebral body edge confidence probability, the more likely the corresponding closed edge is to be the spinal vertebral body. Therefore, the closed edge with the highest vertebral body edge confidence probability among all closed edges is taken as the reference vertebral body edge, and a region enclosed by the reference vertebral body edge is taken as the corresponding region of the reference vertebral body.

[0054] After obtaining the reference vertebral body in the anteroposterior spinal X-ray image, it is considered that different spinal vertebral bodies may be regarded as a same structure under different scale perspectives. It is further considered that there are corresponding feature corner points on the reference vertebral body edge and within its internal area, and the feature comer points may provide stable distinguishing feature information under different scale changes, and accurately match the similar features of structures at different scales. Therefore, in the disclosure, the all spinal vertebral bodies are obtained by the feature matching algorithm based on the feature comer points in the region enclosed by the reference vertebral body edge and the feature comer points in the region enclosed by all other closed edges.

[0055] In an embodiment of the disclosure, the SIFT algorithm is used to obtain all other spinal vertebral bodies. Since the SIFT algorithm is a matching algorithm well known to those skilled in the art, only its general implementation steps are described herein: obtaining feature descriptors of all feature comer points within the corresponding region of the reference vertebral body and corresponding regions of all other closed edges respectively, and obtaining all other spinal vertebral bodies by calculating the similarity between the feature descriptors to match similar feature structures between the region enclosed by the reference vertebral body edge and the regions enclosed by other closed edges. In other embodiments of the disclosure, implementers may also use other feature matching algorithms to obtain all other spinal vertebral bodies.

[0056] The reference vertebral body and all other spinal vertebral bodies obtained through the feature matching algorithm are considered as all of the spinal vertebral bodies in the spinal region. After obtaining the all spinal vertebral bodies in the anteroposterior spinal X-ray image, the grayscale enhancement may be performed on the all spinal vertebral bodies. In an embodiment of the disclosure, histogram equalization is specifically performed on the corresponding region of each spinal vertebral body to improve contrast. The histogram equalization is a prior art well-known to those skilled in the art and is not further elaborated here.

[0057] During a the medical health examination of the examinee, the clarity of the structural features of the vertebral body edges and internal regions may be improved by performing the local enhancement optimization on all spinal vertebral bodies in the anteroposterior spinal X-ray image, thereby assisting medical personnel in judging the health status of the spine of the examinee based on the distance between the spinal vertebral bodies and edge information of areas such as the vertebral arch pedicles.

[0058] In summary, in the disclosure, all closed edges and all feature comer points in the anteroposterior spinal X-ray image of the health examinee are first obtained. Then, the internal difference coefficient of each closed edge relative to each other closed edge is obtained based on the positional distribution difference coefficient of each feature corner points within each closed edge relative to each feature corner point within each other closed edge, and subsequently the vertebral body edge probability of each closed edge being the spinal vertebral body is obtained by combining the rectangularity degree of each closed edge. Furthermore, the vertebral body edge confidence probability is obtained based on the number of the feature corner points within each closed edge and the rectangularity degree of each closed edge, and the reference vertebral body edge is selected. Then, the all spinal vertebral bodies are obtained by the feature matching algorithm and the grayscale enhancement is performed on all spinal vertebral bodies. In the disclosure, structural feature information of the all spinal vertebral bodies is fully utilized, the possibility of each closed edge being the vertebral body edge in the anteroposterior spinal X-ray image is analyzed, the most likely reference vertebral body is selected, and all other spinal vertebral bodies based on the feature matching algorithm are obtained. The quality of spinal region extraction for precise local enhancement and optimization is improved, which improves the optimization and enhancement effect of big data in the medical health physical examination.

[0059] It should be noted that the order of the above embodiments of the disclosure is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings does not necessarily require a specific or continuous sequence to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0060] The various embodiments in the specification are described in a progressive manner, and the same and similar parts between embodiments may be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for optimizing and enhancing big data in medical health physical examination, comprising:obtaining all closed edges and all feature corner points in an anteroposterior spinal X-ray image of a health examinee;obtaining an internal difference coefficient of each closed edge relative to each other closed edge based on a positional distribution difference coefficient of all feature comer points within each closed edge relative to all feature comer points within each other closed edge, obtaining a vertebral body edge probability of each closed edge being a spinal vertebral body based on a rectangularity degree of each closed edge and the internal difference coefficient of each closed edge relative to each other closed edge, and obtaining a vertebral body edge confidence probability of each closed edge being the spinal vertebral body based on a number of all feature corner points within each closed edge and the vertebral body edge probability of each closed edge; andselecting a reference vertebral body edge from the all closed edges based on the vertebral body edge confidence probability of each closed edge, obtaining all spinal vertebral bodies by a feature matching algorithm based on feature corner points within a region enclosed by the reference vertebral body edge and feature corner points within a region enclosed by all other closed edges, and performing grayscale enhancement on the all spinal vertebral bodies;wherein the obtaining an internal difference coefficient of each closed edge comprises:with any one of the all closed edges as a target closed edge and any one of all feature corner points within the target closed edge as a target feature comer point, obtaining a positional distribution difference coefficient of the target feature corner point relative to each feature corner point within each other closed edge except the target closed edge, and using a feature corner point with a smallest positional distribution difference coefficient relative to the target feature corner point within each closed edge except the target closed edge as a suspectedmatching corner point for the target feature corner point within each closed edge;obtaining a suspected matching comer point corresponding to each feature corner point within the target closed edge within each other closed edge except the target closed edge, and using a mean value of positional distribution difference coefficients between all feature corner points within the target closed edge and corresponding suspected matching corner points within a corresponding closed edge except the target closed edge as the internal difference coefficient between the target closed edge and the corresponding closed edge;wherein a calculation formula of the positional distribution difference coefficient is as follows:^us,vr h + 7 X bus bvr ,I |LW Iwhere w represents a positional distribution difference coefficient between an 5-th target feature corner point within a w-th target closed edge and a / -th feature comer point within a v-th closed edge, Lu represents a number of edge pixel points of the / / -th target closed edge, Lv represents a number of edge pixel points of the v-th closed edge, aus represents a vector from the 5-th target feature corner point within the w-th target closed edge to a nearest edge pixel point from the 5-th target feature comer point on a closed region enclosed by the / / -th target closed edge, avr represents a vector from the r-th target feature corner point within the v-th closed edge to a nearest edge pixel point from the r-th target feature comer point on a closed region enclosed by the v-th closed edge, bus represents a vector from the 5-th target feature corner point within the / / -th target closed edge to a farthest edge pixel point from the 5-th target feature comer point on the closed region enclosed by the / / -th target closed edge, and bvr represents a vector from the r-th target feature comer point within the v-th closed edge to a farthest edge pixel point from the r-th target feature corner point on the closed region enclosed by the v-th closed edge;wherein the obtaining a rectangularity degree of each closed edge comprises:with any one of edge pixel points on the closed edge as a starting point, obtaining an8-chain code corresponding to the closed edge, counting an occurrence frequency of each chain code values in the 8-chain code corresponding to the closed edge, sorting occurrence frequencies of the chain code values in a descending order, with a first sorted chain code value and a second sorted chain code value as a first group of chain code values, with a third sorted chain code value and a fourth chain code value as a second group of chain code values, obtaining a confidence coefficient for each group of chain code values that chain code values of each group are chain code values corresponding to a pair of rectangular opposite sides based on a chain code value difference in each group of chain code values and an edge pixel point number difference corresponding to chain code values of each group of chain code values on the closed edge, and multiplying confidence coefficients for the first group of chain code values and the second group of chain code values to obtain the rectangul arity degree of the closed edge;wherein a calculation formula of the confidence coefficient is as follows:At,z = exp x exp (-|DZ1 - Dz2 - a|),where Au,z represents a confidence coefficient of a z-th group of chain code values being chain code values corresponding to a pair of the rectangular opposite sides within a w-th closed edge, Bzi represents a number of edge pixel points corresponding to a chain code value with a highest occurrence frequency in the z-th group of chain code values, Bz? represents a number of edge pixel points corresponding to a chain code value with a lowest occurrence frequency in the z-th group of chain code values, Cz represents a number of all edge pixel points corresponding to the z-th group of chain code values, Dzj represents the chain code value with highest occurrence frequency in the z-th group of chain code values, DZ2 represents the chain code value with the lowest occurrence frequency in the z-th group of chain code values, a represents a preset positive integer, and exp () represents an exponential function with a natural constant e as a base;wherein a calculation formula of the vertebral body edge probability is as follows:Hu — Eu x [Ev x exp (—Mw)],where Hu represents a vertebral body edge probability of the w-th closed edge, Eu represents a rectangularity degree of the w-th closed edge, Ev represents a rectangularity degree of the v-th closed edge, Muv represents an internal difference coefficient between the w-th closed edge and the v-th closed edge, N represents a number of all other closed edges except the / / -th closed edge in the anteroposterior spinal X-ray image, and exp () represents an exponential function with the natural constant e as the base;wherein the obtaining a vertebral body edge confidence probability of each closed edge comprises:normalizing the number of all feature comer points within the closed edge in a positive correlation to obtain a normalized value, and multiplying the normalized value by the vertebral body edge probability of the closed edge to obtain the vertebral body edge confidence probability of the closed edge.

2. The method for optimizing and enhancing big data in medical health physical examination according to claim 1, wherein the obtaining all closed edges comprises:performing double-threshold edge detection on the anteroposterior spinal X-ray image to obtain all edges, and obtaining all closed edges in the all edges based on an edge tracking algorithm.

3. The method for optimizing and enhancing big data in medical health physical examination according to claim 1, wherein the feature matching algorithm is a scale-invariant feature transform (SIFT) algorithm.

4. The method for optimizing and enhancing big data in medical health physical examination according to claim 1, wherein the performing grayscale enhancement on all spinal vertebral bodies comprises:performing histogram equalization on each spinal vertebral body individually.

Images