A construction method for an osteoporosis fracture risk prediction model

Abstract

The present invention relates to the technical field of medical informatics, and specifically relates to a method for constructing an osteoporosis fracture risk prediction model. The method includes: obtaining the angle between the foot and the ground, the walking speed, and the CT image during the monitoring of the target person; dividing the pixel points in the CT image into multiple bone tissue regions; obtaining the bone mass characteristic values of the bone structure according to the shape distribution characteristics of the edge lines in the bone tissue regions; obtaining the posture control factor of each gait according to the angle between the left and right feet of the target person and the ground in each gait; combining the posture control factor and the corresponding walking speed to determine the posture abnormality index of each gait; obtaining the control disorder coefficient of each time period according to the numerical distribution characteristics and change characteristics of all the posture control factors within each time period; and determining the osteoporosis evaluation value by comprehensively considering the bone mass characteristic values of the bone structure and the control disorder coefficient. The present invention improves the accuracy of the comprehensive bone mass evaluation result of the target person.

Description

Technical Field

[0001] The present invention relates to the technical field of medical informatics, and in particular to a method for constructing an osteoporotic fracture risk prediction model. Background Art

[0002] Osteoporosis is a disease characterized by decreased bone density, microarchitectural disruption, and reduced bone strength, significantly increasing the risk of fractures. The osteoporotic fracture risk prediction model integrates multiple variables, including bone density, clinical risk factors, and lifestyle, using statistical or machine learning algorithms to assess an individual's probability of osteoporotic fracture within a specific timeframe. The core goal of this model is to accurately identify high-risk individuals, thereby guiding early intervention and reducing fracture incidence.

[0003] Traditional fracture risk prediction methods are limited by the characterization capabilities of single-modal data, such as bone density or static images. Existing multimodal fusion models lack the ability to quantify the high-frequency dynamic evolution characteristics of the bone metabolic microenvironment due to the heterogeneity of cross-source data (such as dimensional differences in images, biochemical indicators and mechanical parameters) and the lack of time-series dynamic correlation modeling capabilities. For example, hidden risks such as nonlinear attenuation of trabecular topological structure and entropy increase in stress distribution are difficult to detect, resulting in a decrease in model sensitivity and specificity. It is impossible to obtain an accurate comprehensive indicator to reflect the comprehensive bone condition of the tested person, and it cannot provide doctors with an effective reference basis. Summary of the invention

[0004] In order to solve the problem that existing methods cannot accurately obtain comprehensive indicators that reflect the comprehensive bone condition of the tested person, the purpose of the present invention is to provide a method for constructing an osteoporotic fracture risk prediction model. The technical solution adopted is as follows:

[0005] The present invention provides a method for constructing an osteoporotic fracture risk prediction model, the method comprising the following steps:

[0006] Obtain the angle between the target person's feet and the ground, walking speed, and CT images during monitoring;

[0007] Based on the grayscale distribution of pixels in the CT image, the pixels are divided into several bone tissue areas; according to the shape distribution characteristics of the edge lines in the bone tissue area, the bone characteristic value of the bone structure is obtained;

[0008] The posture control factor of each gait is obtained based on the angle between the left and right feet of the target person and the ground in each gait; the posture abnormality index of each gait is determined by combining the posture control factor and the corresponding walking speed; and the control disorder coefficient of each time period is obtained based on the numerical distribution characteristics and change characteristics of all posture control factors in each time period;

[0009] Determine the osteoporosis evaluation value of the target person based on the bone mass characteristic value of the comprehensive bone structure, the control disorder coefficient in the current time period, the overall distribution of all posture abnormality indicators in the current time period, and the correlation between the bone mass characteristic value and the control disorder coefficient in all time periods during the monitoring process.

[0010] Preferably, dividing the pixel points into several bone tissue regions based on the gray-scale distribution of the pixel points in the CT image, including:

[0011] Select the pixel points with gray-scale values greater than the preset gray-scale threshold in the CT image as seed points, and perform region growing on the pixel points in the CT image to obtain several bone tissue regions. Among them, the rule of region growing is: the gray-scale value difference between the neighboring pixel points and the seed points is less than the preset gray-scale difference threshold.

[0012] Preferably, obtaining the bone mass characteristic value of the bone structure according to the shape distribution characteristics of the inner edge lines in the bone tissue region, including:

[0013] For any bone tissue region:

[0014] Calculate the mean Euclidean distance between all edge pixel points in the any bone tissue region;

[0015] Perform linear fitting on each edge line in the any bone tissue region respectively to obtain the corresponding goodness of fit, and calculate the mean value of the goodness of fit corresponding to all edge lines in the any bone tissue region;

[0016] According to the mean Euclidean distance and the mean value of the goodness of fit, obtain the shape irregularity value of any bone tissue region. The mean Euclidean distance is positively correlated with the shape irregularity value, and the mean value of the goodness of fit is negatively correlated with the shape irregularity value;

[0017] Integrate the gray-scale mean value of the pixel points in all bone tissue regions and the shape irregularity value to obtain the bone mass characteristic value of the bone structure.

[0018] Preferably, integrating the gray-scale mean value of the pixel points in all bone tissue regions and the shape irregularity value to obtain the bone mass characteristic value of the bone structure, including:

[0019] Calculate the first product of the gray-scale mean value of the pixel points in each bone tissue region and the shape irregularity value;

[0020] Determine the average value of the first products of all bone tissue regions as the bone mass characteristic value of the bone structure.

[0021] Preferably, obtaining the posture control factor of each gait according to the angle between the left and right feet of the target person and the ground for each gait, including:

[0022] For any gait:

[0023] Take the difference between the driving indices of the left and right feet as the driving asymmetry coefficient; the driving index is the normalized value of the maximum angle between the sole of the foot and the ground when the toe leaves the ground;

[0024] Take the difference between the braking indices of the left and right feet as the braking asymmetry coefficient, and the braking index is the normalized value of the maximum angle between the sole of the foot and the ground when the heel touches the ground;

[0025] Take the sum of the driving asymmetry coefficient and the braking asymmetry coefficient as the posture control factor of any one gait.

[0026] Preferably, combine the posture control factor and the corresponding walking speed to determine the posture abnormality index of each gait, including:

[0027] Take the ratio between the posture control factor of any one gait and the corresponding walking speed as the posture abnormality index of any one gait.

[0028] Preferably, obtain the control disorder coefficient of each time period according to the numerical distribution characteristics and change characteristics of all posture control factors in each time period, including:

[0029] For any time period:

[0030] Perform curve fitting on all posture control factors in the any time period to obtain the corresponding first fitting curve;

[0031] Obtain the window corresponding to each point on the first fitting curve; calculate the frequency of the data value of each point in the window corresponding to each point appearing in the first fitting curve and the slope of each point on the first fitting curve respectively;

[0032] Integrate the frequency of the data value of each point in the window corresponding to each point appearing in the first fitting curve and the slope to obtain the control disorder coefficient of the any time period.

[0033] Preferably, integrate the frequency of the data value of each point in the window corresponding to each point appearing in the first fitting curve and the slope to obtain the control disorder coefficient of the any time period, including:

[0034] Take the ratio between the slope of each point in the window corresponding to each point and the frequency of the data value of each point in the window corresponding to each point appearing in the first fitting curve as the first ratio of each point in the window corresponding to each point;

[0035] Integrate the first ratios of all points in the windows corresponding to all points on the first fitting curve to obtain the control disorder coefficient of the any time period.

[0036] Preferably, obtaining the osteoporosis evaluation value of the target person includes:

[0037] Calculate the first sum of the average value of all abnormal posture indicators in the current time period and the control disorder coefficient in the current time period;

[0038] Calculating a second product of the bone characteristic value of the current time period and the first sum value;

[0039] The sum of the second product and the correlation between the bone characteristic values and the control disorder coefficients in all time periods during the monitoring process is determined as the osteoporosis evaluation value of the target person.

[0040] Preferably, obtaining the correlation between the bone characteristic values and the control disorder coefficient in all time periods during the monitoring process includes:

[0041] The Pearson correlation coefficient between the bone characteristic value sequence and the control disorder coefficient sequence was used as the correlation between the bone characteristic value and the control disorder coefficient at all time periods during the monitoring process;

[0042] The bone characteristic value sequence is composed of the bone characteristic values of all time periods during the monitoring process, and the control disorder coefficient sequence is composed of the control disorder coefficients of all time periods during the monitoring process.

[0043] The present invention has at least the following beneficial effects:

[0044] The present invention evaluates the skeletal structure of the target person by integrating the grayscale distribution of pixel points and the morphological distribution characteristics of edge lines in the static CT image of the target person, quantifies the sparsity, irregularity and edge shape distribution of the bone microstructure in the CT image of the target person, and obtains the bone characteristic value of the skeletal structure. Then, combined with the dynamic gait data, the gait control of the target person is evaluated according to the angle between the left and right feet and the ground and the walking speed of each gait of the target person, dynamically captures the nonlinear evolution characteristics of bone degeneration, and obtains the control disorder coefficient. Finally, the dynamic and static characteristics of the target person are fused to obtain the osteoporosis evaluation value of the target person. The osteoporosis evaluation value can accurately reflect the comprehensive condition of the target person's bone quality, overcomes the problems of insufficient representation ability of traditional single-modal data and difficulty in fusing cross-source heterogeneous data, improves the accuracy of the evaluation results, and can provide an effective reference basis for doctors. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] 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.

[0046] Figure 1 A flowchart of a method for constructing an osteoporotic fracture risk prediction model provided by an embodiment of the present invention;

[0047] Figure 2 This is a structural block diagram of a system for constructing an osteoporotic fracture risk prediction model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] 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 is a detailed description of the method for constructing an osteoporotic fracture risk prediction model proposed by the present invention in combination with the accompanying drawings and preferred embodiments.

[0049] 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.

[0050] The specific scheme of the method for constructing an osteoporotic fracture risk prediction model provided by the present invention is described in detail below with reference to the accompanying drawings.

[0051] An embodiment of a method for constructing an osteoporotic fracture risk prediction model:

[0052] The specific scenario targeted by this embodiment is: when analyzing the comprehensive bone condition of the target person, by integrating the bone density, trabecular structure and morphological characteristics in the static CT images of the target person, combined with the time series analysis of the dynamic gait data, a multimodal fusion model for determining the comprehensive bone index is constructed to obtain an accurate osteoporosis evaluation value, providing an effective reference basis for doctors.

[0053] This embodiment proposes a method for constructing an osteoporotic fracture risk prediction model, such as Figure 1 As shown, the method for constructing an osteoporotic fracture risk prediction model in this embodiment includes the following steps:

[0054] Step S1, obtaining the angle between the feet and the ground, the walking speed and the CT image of the target person during monitoring.

[0055] First, for the target person, a standardized abdominal CT and chest CT scanning protocol is adopted to collect the lumbar CT (L1 to L4) images of the target person. The tube voltage is set to 120 KV, the tube current and scanning time are automatically matched, the scanning field of view is greater than 400 mm, the slice thickness is 0.5 mm, and the slice interval is 0.5 mm. For the L1 to L4 vertebral bodies in the CT images, a standard bone window reconstruction algorithm is used to reconstruct the slice thickness of 0.1 mm or 0.5 mm and the images are transferred to the picture archiving.

[0056] Meanwhile, an intelligent insole integrated with an accelerometer and a gyroscope is worn by the target person, and the walking speed data of the patient during daily activities and the angle between the foot and the ground for each gait are collected in real time through the built-in sensors. It should be noted that: in this embodiment, a complete gait includes the process that both feet have completed leaving the ground and returning to the ground.

[0057] So far, in this embodiment, the angle between the foot and the ground, the walking speed, and the CT images of the target person are obtained during the monitoring process. Among them, the monitoring process is a set composed of all historical moments whose time interval from the current moment is less than or equal to the preset duration. In this embodiment, the preset duration is 1 week. In specific applications, the implementer can set it according to the specific situation.

[0058] Step S2: Divide the pixel points into several bone tissue regions based on the gray-scale distribution of the pixel points in the lumbar CT images; according to the shape distribution characteristics of the edge lines within the bone tissue regions, obtain the bone quality characteristic values of the bone structure.

[0059] Considering that there are certain differences among different individuals, such as differences in basic physical conditions, the bone structures of patients show different osteoporosis characteristics. Therefore, next, bone density, trabecular bone structure, morphological characteristics, etc. will be extracted from the CT images of the target person as the static input features of the model.

[0060] In the CT images of the target person, within the same bone tissue or structure region, due to the stability of the structure or tissue density and the scanning consistency, the pixel value distribution of the pixel points within the region is usually relatively consistent.

[0061] Pixels with grayscale values greater than a preset grayscale threshold in the CT image of the target person are selected as seed points. These seed points are usually located in the bone tissue area because the X-ray absorption value of the bone is high and the grayscale value is more significant; a region growing algorithm is used to perform region growing on the pixel points in the CT image, and each region obtained after region growing is used as the bone tissue area, that is, multiple bone tissue areas are obtained, wherein the rule of region growing is: when the grayscale value difference between the neighborhood pixel point and the seed point is less than the preset grayscale difference threshold, the neighborhood pixel point and the seed point are merged. In this embodiment, the preset grayscale threshold is 80, and the preset grayscale difference threshold is 3. In specific applications, the implementer can set them according to specific circumstances. The neighborhood size in this embodiment is four neighborhoods, and the region growing algorithm is a prior art and will not be described in detail here.

[0062] As the symptoms of osteoporosis increase, long-term osteoporosis can cause the patient's bone morphology to change, and cause degenerative changes in the articular surface, such as widening of the joint space and irregularity of the joint surface. Trabecular bone, as the supporting network of skeleton, will appear thinning, rarefaction, or even fracture in the process of osteoporosis. In CT images, the patient's trabecular bone structure shows obvious rarefaction and irregularity, and the imaging characteristics of trabecular bone fracture or structural damage occur, which further affect the mechanical properties and stability of skeleton. Therefore, the present embodiment will evaluate the irregularity of skeletal tissue structure.

[0063] Next, this embodiment is described by taking a bone tissue region in a CT image of a target person as an example. Other bone tissue regions in the CT image of the target person can be processed using the method provided in this embodiment.

[0064] Specifically, for any bone tissue region:

[0065] A deep learning-based segmentation model, such as a U-Net network model, is used to extract a binary edge map of the skeletal region and all edge lines within the region. The mean Euclidean distance between all edge pixels within the region is calculated based on the Euclidean distance between each pair of edge pixels.

[0066] The least squares method and linear fitting are used to perform linear fitting on each edge line in the bone tissue area to obtain the corresponding goodness of fit. Each edge line in the bone tissue area corresponds to a goodness of fit; the mean of the goodness of fit corresponding to all edge lines in the bone tissue area is calculated.

[0067] The smaller the mean goodness of fit of all edge lines within the skeletal tissue region, the greater the difference between the edge lines and their corresponding fitted straight lines within the skeletal tissue region. This indicates that the edge curve has more inflection points and the edge structure is more complex and irregular. The larger the mean Euclidean distance between all edge pixels within the skeletal tissue region, the sparser the bone edge structure within the skeletal tissue region, reflecting a more abnormal support network for the regional skeletal structure.

[0068] Based on the above characteristics, this embodiment will obtain the shape irregularity value of the bone tissue area according to the mean Euclidean distance between all edge pixel points in the bone tissue area and the mean goodness of fit corresponding to all edge lines in the bone tissue area. The shape irregularity value is used to reflect the complexity and irregularity of the bone structure. The mean Euclidean distance is positively correlated with the shape irregularity value, and the mean goodness of fit is negatively correlated with the shape irregularity value.

[0069] Among them, a positive correlation relationship indicates that the dependent variable will increase as the independent variable increases, and the dependent variable will decrease as the independent variable decreases. It can be an additive relationship, a multiplicative relationship, etc., which is determined by actual application; a negative correlation relationship indicates that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. It can be a subtractive relationship, a division relationship, etc., which is determined by actual application.

[0070] In this embodiment, a specific calculation formula for the shape irregularity value is given. The shape irregularity value of the mth bone tissue region in the lumbar spine CT image of the target person can be expressed as:

[0071]

[0072] in, Indicates the shape irregularity value of the mth bone tissue area in the lumbar spine CT image of the target person, represents the mean Euclidean distance between all edge pixels in the mth bone tissue area, Represents the mean goodness of fit corresponding to all edge lines in the mth bone tissue region.

[0073] In this embodiment, 0.01 is added to the denominator of the calculation formula for the shape irregularity value to prevent the denominator from being zero. In specific applications, the implementer can set this value based on specific circumstances. When the mean Euclidean distance between all edge pixels in the mth skeletal tissue region is larger and the mean goodness of fit corresponding to all edge lines in the mth skeletal tissue region is smaller, it indicates that the edges of the skeletal structure in the mth skeletal tissue region are more dispersed and irregular, reflecting a more severe degree of osteoporosis and potentially poor bone stability and mechanical properties.

[0074] By adopting the above method, the shape irregularity value of each bone tissue region can be obtained.

[0075] During the course of osteoporosis, bone density in the skeletal structure is typically decreased, leading to a decrease in the grayscale values of pixels in the skeletal region image. Therefore, this embodiment calculates the mean grayscale value of all pixels within each skeletal tissue region based on the grayscale value of each pixel within the region, and records this as the grayscale mean of the pixels within each skeletal tissue region. Each skeletal tissue region has a corresponding grayscale mean. The smaller the grayscale mean, the more significant degeneration of the skeletal structure in the skeletal tissue region, resulting in a decrease in bone density, which may lead to the occurrence of osteoporosis and affect the mechanical properties and stability of the bone.

[0076] Next, the bone quality characteristic value of the skeletal structure is obtained by combining the grayscale mean of the pixels in all skeletal tissue regions and the shape irregularity value. Specifically, the product of the grayscale mean of the pixels in each skeletal tissue region and the shape irregularity value is calculated, and this product is recorded as a first product. Each skeletal tissue region has a corresponding first product, and the average of the first products of all skeletal tissue regions is determined as the bone quality characteristic value of the skeletal structure.

[0077] In this embodiment, a specific calculation formula for the bone characteristic value is given, which can be expressed as:

[0078]

[0079] in, The bone characteristic value representing the bone structure, Indicates the number of bone tissue areas in the CT image of the target person, represents the grayscale mean of the pixel points in the mth bone tissue area in the CT image of the target person, Indicates the shape irregularity value of the mth bone tissue region in the CT image of the target person. Represents the first product of the mth bone tissue area.

[0080] The bone quality characteristic value of the skeletal structure is used to reflect the stability and mechanical properties of the target person's skeletal structure. When the grayscale mean value of the pixels in all bone tissue areas in the CT image of the target person is larger and the shape irregularity value is also larger, it means that the mechanical properties and stability of the target person's bones are poor and the risk of fracture is higher, that is, the bone quality characteristic value of the skeletal structure is larger.

[0081] So far, the bone characteristic values of the skeletal structure have been obtained using the above method.

[0082] Step S3, obtaining the posture control factor of each gait based on the angle between the left and right feet and the ground in each gait of the target person; determining the posture abnormality index of each gait by combining the posture control factor and the corresponding walking speed; and obtaining the control disorder coefficient of each time period based on the numerical distribution characteristics and change characteristics of all posture control factors in each time period.

[0083] By analyzing gait parameters, we can evaluate the patient's motor function and coordination, and then infer the nonlinear decline of bone structure with time and lifestyle changes, and further capture the high-frequency dynamic change characteristics of bone density and stress distribution, thereby assisting in predicting the potential risk of fractures.

[0084] Changes in the patient's gait reflect the development trend of osteoporosis. As bone density decreases and bone fragility increases, the patient's motor ability and activity level are affected, causing the gait to become slow or unstable; specifically, the pain caused by osteoporosis, especially pain in the spine and joints, limits the patient's normal motor function, which in turn causes gait changes. Patients with pain and limited motor function often exhibit irregular gait, such as shortened stride length, slowed gait, etc., which increases the risk of fractures. Based on this, this embodiment will next analyze each gait of the target person separately.

[0085] This embodiment is described by taking one gait as an example, and other gaits can be processed using the method provided in this embodiment.

[0086] Specifically, for any gait:

[0087] The difference in the left and right foot actuation indices is used as the actuation asymmetry coefficient. The actuation index is the normalized value of the maximum angle between the sole of the foot and the ground during toe-off. In this embodiment, the difference in the left and right foot actuation indices is obtained by calculating the absolute value of the difference between the left and right foot actuation indices and using this absolute value as the actuation asymmetry coefficient. A larger actuation asymmetry coefficient indicates more pronounced asymmetry between the left and right foot actuation movements, reflecting actuation instability in the patient's gait.

[0088] The difference in the braking indexes of the left and right feet is used as the braking asymmetry coefficient. The braking index is the normalized value of the maximum angle between the sole of the foot and the ground when the heel strikes. In this embodiment, the difference in the braking indexes of the left and right feet is obtained by calculating the absolute value of the difference between the braking indexes of the left and right feet, and using this absolute value as the braking index difference of the left and right feet. In this embodiment, there are many methods for normalizing the angle. In specific applications, the implementer can select a method based on the specific situation, so that the normalized result is in the range [0, 1]. This will not be detailed here.

[0089] Take the sum of the driving asymmetry coefficient and the braking asymmetry coefficient as the posture control factor for this gait.

[0090] Considering individual differences, gait complexity, and minor gait variations, there may be certain errors in measuring gait stability solely based on the driving asymmetry coefficient and the braking asymmetry coefficient; healthy individuals usually exhibit slight gait asymmetry, but it does not significantly affect the overall gait stability. Osteoporosis patients, especially those with more severe conditions, show significant gait asymmetry, which leads to gait instability, accompanied by manifestations such as a slow walking speed. Therefore, next, the degree of postural abnormality is evaluated by combining the posture control factor and the walking speed to obtain a postural abnormality index.

[0091] Specifically, for any gait: the ratio between the posture control factor of this gait and the corresponding walking speed is determined as the postural abnormality index for this gait; when the posture control factor of this gait is larger and the walking speed is slower, it indicates that the gait of the target person exhibits stronger disorder characteristics, that is, the asymmetry and instability in the gait of the target person are more obvious, that is, the postural abnormality index is larger.

[0092] Using the above method, the postural abnormality index of each gait of the target person can be obtained.

[0093] The patient's bone structure undergoes non-linear degradation over time and lifestyle changes, and the gait pattern will show disorder. Specifically, the gait stability of the patient during daily activities gradually decreases, which is reflected in the reduced stability of posture control in each time period. This change indicates that the gait behavior pattern of the patient is affected by bone decline, resulting in impaired posture control function, which in turn affects their overall gait performance.

[0094] Divide the monitoring process into time periods, each time period being 1 hour long. In specific applications, the implementer can set it according to the specific situation.

[0095] Next, this embodiment separately analyzes each time period during the monitoring process.

[0096] For any time period:

[0097] In chronological order, curve fitting is performed on all the posture control factors within this time period to obtain the corresponding fitting curve, and this fitting curve is denoted as the first fitting curve; curve fitting is a prior art and will not be elaborated here.

[0098] Taking each point on this fitting curve as the window center respectively, a window with a preset length is obtained, and this window is used as the window corresponding to each point on this fitting curve. In this embodiment, the preset length is 21. In specific applications, the implementer can set it according to the specific situation.

[0099] Obtain the frequency of the data values of each point in the window corresponding to each point appearing in the first fitting curve and the slope of each point on the first fitting curve respectively. Denote the ratio between the slope of each point in the window corresponding to each point and the frequency of the data values of each point in the window corresponding to each point appearing in the first fitting curve as the first ratio of each point in the window corresponding to each point. Synthesize the first ratios of all points in the windows corresponding to all points on the first fitting curve to obtain the control disorder coefficient for any time period.

[0100] In this embodiment, a specific calculation formula for the control disorder coefficient is given. The control disorder coefficient for the r-th time period can be expressed as:

[0101]

[0102] where, represents the control disorder coefficient for the r-th time period, represents the number of points on the first fitting curve corresponding to the r-th time period, represents the number of points in the window corresponding to the n-th point on the first fitting curve corresponding to the r-th time period, represents the slope of the -th point in the window corresponding to the n-th point on the first fitting curve corresponding to the r-th time period, represents the frequency of the data value of the -th point in the window corresponding to the n-th point on the first fitting curve corresponding to the r-th time period appearing in its corresponding first fitting curve.

[0103] represents the first ratio of the -th point in the window corresponding to the n-th point on the first fitting curve corresponding to the r-th time period. When the first ratios of all points in the windows corresponding to each point on the first fitting curve are larger, it indicates that during daily activities, the target person becomes more unstable in gait control, showing greater gait disorder, meaning that the patient's bone degeneration is severe and the non-linear attenuation of the bone structure is more intense, resulting in a decline in their gait posture control ability.

[0104] By using the above method, the control disorder coefficient for each time period of the target person during the monitoring process can be obtained.

[0105] Step S4, determine the osteoporosis evaluation value of the target person by synthesizing the bone mass characteristic value of the bone structure, the control disorder coefficient of the current time period, the overall distribution of all posture abnormality indicators in the current time period, and the correlation between the bone mass characteristic values and the control disorder coefficients of all time periods during the monitoring process.

[0106] Arrange the bone mass characteristic values for all time periods during the monitoring process in chronological order to obtain a bone mass characteristic value sequence; similarly, arrange the control disorder coefficients for all time periods during the monitoring process in chronological order to obtain a control disorder coefficient sequence.

[0107] Calculate the Pearson correlation coefficient between the bone mass characteristic value sequence and the control disorder coefficient sequence. The calculation method of the Pearson correlation coefficient is a prior art and will not be elaborated here. Take the Pearson correlation coefficient between the bone mass characteristic value sequence and the control disorder coefficient sequence as the correlation between the bone mass characteristic values and the control disorder coefficients for all time periods during the monitoring process. The value range of the correlation is [-1, 1]. The closer its value is to 1, the more positive the correlation between the degree of osteoporosis and gait instability is, indicating that the gait control disorder of the target person may be exacerbated due to the fragile bones caused by their own osteoporosis, thus increasing the risk of the target person falling and fracturing.

[0108] Further, calculate the average value of all posture abnormality indicators for the current time period according to the posture abnormality indicators of each gait in the current time period; then, calculate the sum of the average value of all posture abnormality indicators for the current time period and the control disorder coefficient for the current time period, and record this sum as the first sum; calculate the product between the bone mass characteristic value for the current time period and the first sum, and record this product as the second product; then, determine the osteoporosis evaluation value of the target person by adding the second product and the sum of the correlations between the bone mass characteristic values and the control disorder coefficients for all time periods during the monitoring process.

[0109] In this embodiment, a specific calculation formula for the osteoporosis evaluation value of the target person is given. The osteoporosis evaluation value of the target person can be expressed as:

[0110]

[0111] Among them, represents the osteoporosis evaluation value of the target person, represents the bone mass characteristic value of the bone structure, represents the average value of all posture abnormality indicators for the current time period, represents the control disorder coefficient for the current time period, represents the bone mass characteristic value sequence, represents the control disorder coefficient sequence, represents the correlation between the bone mass characteristic values and the control disorder coefficients for all time periods during the monitoring process.

[0112] represents the first sum, reflecting the degree of gait disorder of the target person; It represents the second product, which reflects the combined effect of osteoporosis and gait disorder. The correlation between the bone mass characteristic values and the control disorder coefficient during all time periods of the monitoring process quantifies the strength of the mutual association between the two. This comprehensive evaluation reveals the deficiencies of the target person in skeletal fragility and motor coordination, which may be accompanied by a relatively high degree of physiological degradation, indicating a relatively high risk of fracture. Therefore, continuous monitoring and intervention of their bone health and motor function are required.

[0113] So far, the method provided in this embodiment has completed the analysis of the comprehensive bone mass condition of the target person, and obtained the osteoporosis evaluation value of the target person. The larger the osteoporosis evaluation value of the target person, the greater the severity of the gait control disorder of the target patient, and higher attention should be given to this patient subsequently.

[0114] In this embodiment, by integrating the gray-scale distribution of pixel points and the morphological distribution characteristics of edge lines in the static CT image of the target person, the bone structure of the target person is evaluated, and the sparsity, irregularity and edge shape distribution of the bone microstructure in the CT image of the target person are quantified to obtain the bone mass characteristic value of the bone structure. Then, combined with the dynamic gait data, according to the angle between the left and right feet of the target person and the ground and the walking speed in each gait, the gait control situation of the target person is evaluated, and the non-linear evolution characteristics of bone degradation are dynamically captured to obtain the control disorder coefficient. Finally, by fusing the dynamic characteristics and static characteristics of the target person, the osteoporosis evaluation value of the target person is obtained. The osteoporosis evaluation value can accurately reflect the comprehensive bone mass condition of the target person, overcomes the problems of insufficient characterization ability of traditional single-modal data and difficulty in fusing cross-source heterogeneous data, improves the accuracy of the evaluation result, and can provide an effective reference basis for doctors.

[0115] An embodiment of a system for constructing an osteoporosis fracture risk prediction model:

[0116] Refer to Figure 2 , which shows the structural block diagram of a system for constructing an osteoporosis fracture risk prediction model provided by an embodiment of the present invention. The system may include a data acquisition module, a static analysis module, a dynamic analysis module and a comprehensive evaluation module;

[0117] Among them, the data acquisition module is used to obtain the angle between the feet and the ground, the walking speed and the CT image during the monitoring process of the target person;

[0118] The static analysis module is used to divide the pixel points into several bone tissue regions based on the gray-scale distribution of the pixel points in the CT image; according to the shape distribution characteristics of the edge lines in the bone tissue region, obtain the bone mass characteristic value of the bone structure;

[0119] A dynamic analysis module is used to obtain a posture control factor for each gait based on the angle between the left and right feet of the target person and the ground during each gait; determine a posture abnormality index for each gait based on the posture control factor and the corresponding gait speed; and obtain a control disorder coefficient for each time period based on the numerical distribution and variation characteristics of all posture control factors within each time period;

[0120] The comprehensive evaluation module is used to comprehensively consider the bone characteristic values of the skeletal structure, the control disorder coefficient of the current time period, the overall distribution of all abnormal posture indicators in the current time period, and the correlation between the bone characteristic values and the control disorder coefficient of all time periods during the monitoring process to determine the osteoporosis evaluation value of the target person.

[0121] It should be understood that Figure 2 The structural block diagram of a system for building an osteoporotic fracture risk prediction model and its modules can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented using hardware, software, or a combination of software and hardware. The hardware portion can be implemented using dedicated logic, while the software portion can be stored in memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated hardware. Those skilled in the art will appreciate that the above-described methods and systems can be implemented using computer-executable instructions and / or contained in processor control code, for example, provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The systems and their modules described herein can be implemented not only using hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips or transistors, or programmable hardware devices such as field-programmable gate arrays or programmable logic devices, but can also be implemented using software executed by various types of processors, or a combination of such hardware circuits and software (e.g., firmware).

[0122] For more details about the above modules, please refer to other places in this manual and will not be repeated here.

[0123] In other embodiments, a medium is also provided, which stores at least one program executable by a computer. When the at least one program is executed by a computer, the computer executes the steps in the method for constructing an osteoporotic fracture risk prediction model in the above embodiment. The medium can be a computer-readable storage medium.

[0124] Among them, the provided systems and media are used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0125] It should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for constructing an osteoporosis fracture risk prediction model, characterized in that, The method includes the following steps: Obtain the angle between the foot and the ground, the walking speed, and the CT image during the monitoring of the target person; Divide the pixel points into several bone tissue regions based on the gray-scale distribution of the pixel points in the CT image; obtain the bone quality characteristic values of the bone structure according to the shape distribution characteristics of the inner edge lines in the bone tissue regions; Obtain the posture control factor for each gait according to the angle between the left and right feet and the ground of the target person for each gait; determine the posture abnormality index for each gait by combining the posture control factor and the corresponding walking speed; obtain the control disorder coefficient for each time period according to the numerical distribution characteristics and change characteristics of all the posture control factors within each time period; Comprehensively determine the osteoporosis evaluation value of the target person based on the bone quality characteristic values of the bone structure, the control disorder coefficient of the current time period, the overall distribution of all the posture abnormality indexes of the current time period, and the correlation between the bone quality characteristic values and the control disorder coefficients of all the time periods during the monitoring process; The obtaining of the control disorder coefficient for each time period includes: For any time period: Perform curve fitting on all the posture control factors within the any time period to obtain the corresponding first fitting curve; Obtain the window corresponding to each point on the first fitting curve; calculate the frequency of the data value of each point within the window corresponding to each point appearing in the first fitting curve and the slope of each point on the first fitting curve respectively; Denote the ratio between the slope of each point within the window corresponding to each point and the frequency of the data value of each point within the window corresponding to each point appearing in the first fitting curve as the first ratio of each point within the window corresponding to each point; Obtain the control disorder coefficient of the any time period by comprehensively considering the first ratios of all the points within the windows corresponding to all the points on the first fitting curve.

2. The method for constructing an osteoporosis fracture risk prediction model according to claim 1, characterized in that The dividing of the pixel points into several bone tissue regions based on the gray-scale distribution of the pixel points in the CT image includes: Select the pixel points with gray-scale values greater than the preset gray-scale threshold in the CT image as seed points, and perform region growing on the pixel points in the CT image to obtain several bone tissue regions, where the rule of region growing is: the gray-scale difference between the neighborhood pixel points and the seed points is less than the preset gray-scale difference threshold.

3. The method for constructing an osteoporosis fracture risk prediction model according to claim 1, characterized in that, The obtaining of the bone quality characteristic values of the bone structure according to the shape distribution characteristics of the inner edge lines in the bone tissue regions includes: For any bone tissue region: Calculate the mean Euclidean distance between all the edge pixel points within the any bone tissue region; Perform linear fitting on each edge line within the any bone tissue region respectively to obtain the corresponding goodness of fit, and calculate the mean value of the goodness of fit corresponding to all the edge lines within the any bone tissue region; Obtain the shape irregularity value of any bone tissue region according to the mean Euclidean distance and the mean value of the goodness of fit, where the mean Euclidean distance is positively correlated with the shape irregularity value, and the mean value of the goodness of fit is negatively correlated with the shape irregularity value; Obtain the bone quality characteristic values of the bone structure by comprehensively considering the mean gray-scale value of the pixel points within all the bone tissue regions and the shape irregularity value.

4. The method for constructing an osteoporosis fracture risk prediction model according to claim 3, characterized in that, Combining the gray - scale mean value of pixel points within all bone tissue regions and the said shape irregularity value, the bone mass characteristic value of the bone structure is obtained, including: Calculating the first product of the gray - scale mean value of pixel points within each bone tissue region and the said shape irregularity value; Determining the average value of the said first products of all bone tissue regions as the bone mass characteristic value of the bone structure.

5. The method for constructing an osteoporosis fracture risk prediction model according to claim 1, wherein, The obtaining of the posture control factor for each gait of the target person according to the angle between the left and right feet and the ground for each gait includes: For any one gait: Taking the difference between the driving indices of the left and right feet as the driving asymmetry coefficient; the driving index is the normalized value of the maximum angle between the sole of the foot and the ground when the toe leaves the ground; Taking the difference between the braking indices of the left and right feet as the braking asymmetry coefficient, the braking index is the normalized value of the maximum angle between the sole of the foot and the ground when the heel touches the ground; Taking the sum of the said driving asymmetry coefficient and the said braking asymmetry coefficient as the posture control factor for any one gait.

6. The method for constructing an osteoporosis fracture risk prediction model according to claim 5, wherein Combining the said posture control factor and the corresponding walking speed to determine the posture abnormality index for each gait, including: Taking the ratio between the posture control factor of any one gait and the corresponding walking speed as the posture abnormality index for any one gait.

7. The method for constructing an osteoporosis fracture risk prediction model according to claim 1, characterized in that The obtaining of the osteoporosis evaluation value of the target person includes: Calculating the first sum value of the average value of all posture abnormality indices in the current time period and the control disorder coefficient in the current time period; Calculating the second product of the bone mass characteristic value in the current time period and the said first sum value; Determining the sum of the said second product and the correlation between the bone mass characteristic values and the control disorder coefficient in all time periods during the monitoring process as the osteoporosis evaluation value of the target person.

8. The method for constructing an osteoporosis fracture risk prediction model according to claim 1, wherein, The obtaining of the correlation between the bone mass characteristic values and the control disorder coefficient in all time periods during the monitoring process includes: Taking the Pearson correlation coefficient between the bone mass characteristic value sequence and the control disorder coefficient sequence as the correlation between the bone mass characteristic values and the control disorder coefficient in all time periods during the monitoring process; The bone mass characteristic value sequence is composed of the bone mass characteristic values in all time periods during the monitoring process, and the control disorder coefficient sequence is composed of the control disorder coefficients in all time periods during the monitoring process.

Images