A preoperative femoral head collapse risk prediction system based on biomechanical analysis

Abstract

The present application relates to the technical field of medical data processing, in particular to a preoperative femoral head collapse risk prediction system based on biomechanical analysis, the execution steps of the system comprising: obtaining the pressure monitoring set and the marker position coordinates of each monitoring point in the necrotic area of the femoral head of a plurality of patient samples; determining the daily wear performance degree and the local potential risk degree of each patient sample according to the pressure monitoring set and the marker position coordinates; and performing data fusion analysis on the daily wear performance degree and the local potential risk degree of the same patient sample to determine the multi-dimensional comprehensive risk assessment value of each patient sample. By determining the multi-dimensional comprehensive risk assessment value, the present application ensures that the prediction model can make accurate judgments based on the mechanical function of the skeleton rather than just the appearance, thereby improving the prediction accuracy of the model.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, specifically to a preoperative femoral head collapse risk prediction system based on biomechanical analysis. Background Technology

[0002] Existing preoperative femoral head collapse risk prediction models typically use only imaging data as training data. While imaging data primarily provides anatomical and morphological information, it cannot directly quantify the biomechanical behavior of the femoral head under load. For example, imaging data cannot accurately reflect the stress distribution, stress concentration, or fatigue accumulation effect of the necrotic area during daily activities, causing the model to fail to capture the fundamental mechanical mechanism of collapse and thus limiting prediction accuracy. Furthermore, relying solely on imaging data, the model may focus on morphological abnormalities, but femoral head collapse is often related to functional mechanical failure. For instance, even if the imaging shows a small necrotic area, if that area is under high stress concentration, the risk of collapse may be high. In such cases, the model cannot recognize this disconnect between morphology and function, leading to biased risk assessment. In summary, the training data of existing models used to predict preoperative femoral head collapse risk exhibits incomplete femoral head collapse characteristics, resulting in low prediction accuracy. Summary of the Invention

[0003] To address the technical problem of incomplete femoral head collapse features in the training data of existing models used to predict preoperative femoral head collapse risk, leading to low prediction accuracy, the present invention aims to provide a preoperative femoral head collapse risk prediction system based on biomechanical analysis. The specific technical solution adopted is as follows:

[0004] One embodiment of the present invention provides a preoperative femoral head collapse risk prediction system based on biomechanical analysis, comprising:

[0005] The data acquisition module is used to acquire the pressure monitoring set and marked position coordinates of each monitoring point in the femoral head necrosis area of ​​several patient samples. The pressure monitoring set includes pressure monitoring data under several load pressures.

[0006] The first determining module is used to analyze the relative abnormalities of the patient samples based on the pressure monitoring set and marker location coordinates of each monitoring point corresponding to each patient sample, and to determine the daily wear performance of each patient sample.

[0007] The second determining module is used to analyze the high stress concentration of the patient sample based on the pressure monitoring set and marker location coordinates of each monitoring point corresponding to each patient sample, and to determine the local potential risk level of each patient sample.

[0008] The third determination module is used to determine the multidimensional comprehensive risk assessment value of each patient sample based on the daily wear performance and local potential risk of each patient sample. The multidimensional comprehensive risk assessment value is one of the input variables for training the prediction model.

[0009] Furthermore, the data acquisition module is also used for:

[0010] Obtain physiological feature vectors from several patient samples, wherein the physiological feature vectors include at least age, weight, and bone density;

[0011] Based on the differences between the physiological feature vectors of every two patient samples, all patient samples are divided into several patient sample groups.

[0012] Further, the first determining module includes:

[0013] The data determination unit is used to determine the pressure baseline value of each monitoring point in the current patient sample population under each load pressure, and to obtain the target location marker sequence of the subchondral bone region of the current patient sample, wherein the current patient sample is any patient sample.

[0014] The first determining unit is used to determine the degree of pressure abnormality of each monitoring point of the current patient sample under each load pressure based on the difference between the pressure monitoring data of each monitoring point under each load pressure and the pressure reference value.

[0015] The second determining unit is used to determine the performance weight of each monitoring point in the current patient sample based on the distance between the marker position coordinates of each monitoring point corresponding to the current patient sample and each position marker in the target position marker sequence.

[0016] The third determining unit is used to perform weighted fusion processing based on the pressure abnormality performance of each monitoring point of the current patient sample under each load pressure and the performance weight of each monitoring point to determine the daily wear performance of the current patient sample.

[0017] Further, determining the pressure baseline value for each monitoring point in the current patient sample population under each load pressure includes:

[0018] Within the current patient sample population, the average pressure monitoring data of all patient samples at the same monitoring point under the same load pressure is calculated and used as the pressure benchmark value of the corresponding monitoring point under the corresponding load pressure in the current patient sample population.

[0019] Further, the execution steps of the second determining unit include:

[0020] Calculate the distance between the marker position of each monitoring point and each marker position in the target position marker sequence, and then perform data fusion processing on all distance values ​​of the same monitoring point to obtain the fusion value of each monitoring point of the current patient sample;

[0021] The fusion value of each monitoring point in the current patient sample is inversely normalized, and the normalized value is used as the performance weight of the corresponding monitoring point in the current patient sample.

[0022] Furthermore, the second determining module includes:

[0023] The fourth determining unit is used to determine the relative stress stability performance and high stress performance coefficient of each monitoring point based on the pressure monitoring data of each monitoring point corresponding to each patient sample under each load pressure.

[0024] The fifth determining unit is used to determine the potential risk of the relatively high stress concentration area of ​​each patient sample based on the relative stress stability performance and high stress performance coefficient of each monitoring point corresponding to each patient sample.

[0025] The sixth determining unit is used to determine the attention weight of the relatively high stress concentration area of ​​each patient sample based on the distance between each monitoring point in the relatively high stress concentration area and each marker position in the target position marker sequence.

[0026] The data fusion unit is used to perform weighted fusion processing on the potential risk level and attention weight of relatively high stress concentration areas of the same patient sample to obtain the local potential risk level of each patient sample.

[0027] Furthermore, the execution steps of the fourth determining unit include:

[0028] Based on the pressure monitoring data of each monitoring point corresponding to each patient sample under each load pressure, determine the relative pressure performance value of each monitoring point under each load pressure;

[0029] Analyze the pressure fluctuations at each monitoring point under each load pressure to determine the relative stress stability of each monitoring point.

[0030] Based on the pressure monitoring data of each monitoring point corresponding to each patient sample under each load pressure, the high stress performance coefficient of each monitoring point is determined.

[0031] Further, determining the high-stress performance coefficient for each monitoring point based on the pressure monitoring data of each monitoring point corresponding to each patient sample under each load pressure includes:

[0032] For each patient sample, the average value of the pressure monitoring data at the same monitoring point for each load pressure is calculated and denoted as the high stress performance factor.

[0033] The high-stress performance factor for each monitoring point corresponding to the patient sample is normalized, and the normalized value is used as the high-stress performance coefficient for the corresponding monitoring point.

[0034] Furthermore, the execution steps of the fifth determining unit include:

[0035] Based on the relative stress stability performance and high stress performance coefficient of each monitoring point corresponding to each patient sample, the sustained relative high stress performance of each monitoring point corresponding to each patient sample is determined.

[0036] By utilizing the sustained relatively high stress performance of each monitoring point corresponding to each patient sample, the regions corresponding to all monitoring points of the patient sample are divided to obtain the relatively high stress concentration area.

[0037] The potential risk level of the relatively high stress concentration area for each patient sample is determined based on the sustained relatively high stress performance at each monitoring point within the relatively high stress concentration area.

[0038] Furthermore, the execution steps of the third determining module include:

[0039] The number of femoral head collapses in the medical records of each patient sample group was counted, and the risk weight of each patient sample in each patient sample group was determined based on the number of femoral head collapses.

[0040] By using the risk weight of each patient sample, the daily wear and tear performance and local potential risk of each patient sample are weighted and fused to obtain a multidimensional comprehensive risk assessment value for each patient sample.

[0041] The present invention has the following beneficial effects:

[0042] Existing preoperative femoral head collapse risk prediction models rely excessively on static imaging data, failing to reflect the biomechanical failures leading to collapse, resulting in insufficient prediction accuracy and difficulty in supporting precise clinical decision-making. Therefore, this invention provides a preoperative femoral head collapse risk prediction system based on biomechanical analysis. In this system, the data acquisition module addresses the problem of a single data source; the first and second determination modules transform raw data into clinically significant biomechanical indicators from temporal and spatial dimensions, respectively; the third determination module, through information fusion, creates a super-feature that comprehensively characterizes collapse risk. Ultimately, these features work synergistically to ensure that the prediction model makes accurate judgments based on the mechanical function of the bone rather than merely its morphological appearance, thereby improving the model's prediction accuracy. Attached Figure Description

[0043] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 A structural diagram of a preoperative femoral head collapse risk prediction system based on biomechanical analysis, provided as an embodiment of the present invention;

[0045] Figure 2 This is a flowchart illustrating the execution of a preoperative femoral head collapse risk prediction system based on biomechanical analysis, as described in this invention.

[0046] Figure 3 This is a schematic diagram of a pressure sensor placed in the area of ​​femoral head necrosis in an embodiment of the present invention;

[0047] Figure 4 This is a flowchart illustrating the implementation of step S2 in an embodiment of the present invention;

[0048] Figure 5 This is a flowchart illustrating the implementation of step S3 in an embodiment of the present invention. Detailed Implementation

[0049] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the specific implementation methods, structures, features, and effects of the technical solution proposed according to the present invention are described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0050] Unless otherwise defined, 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 pertains.

[0051] This invention provides a preoperative femoral head collapse risk prediction system based on biomechanical analysis, such as... Figure 1 As shown, it includes:

[0052] The data acquisition module is used to acquire the pressure monitoring set and marked position coordinates of each monitoring point in the femoral head necrosis area of ​​several patient samples. The pressure monitoring set includes pressure monitoring data under several load pressures.

[0053] The first determining module is used to analyze the relative abnormalities of the patient samples based on the pressure monitoring set and marker location coordinates of each monitoring point corresponding to each patient sample, and to determine the daily wear performance of each patient sample.

[0054] The second determining module is used to analyze the high stress concentration of the patient sample based on the pressure monitoring set and marker location coordinates of each monitoring point corresponding to each patient sample, and to determine the local potential risk level of each patient sample.

[0055] The third determination module is used to determine the multidimensional comprehensive risk assessment value of each patient sample based on the daily wear performance and local potential risk of each patient sample. The multidimensional comprehensive risk assessment value is one of the input variables for training the prediction model.

[0056] Furthermore, the first determining module includes: a data determining unit, a first determining unit, a second determining unit, and a third determining unit;

[0057] The second determining module includes: a fourth determining unit, a fifth determining unit, a sixth determining unit, and a data fusion unit.

[0058] refer to Figure 2 The flowchart of the preoperative femoral head collapse risk prediction system based on biomechanical analysis of the present invention is shown, including:

[0059] S1, obtain the pressure monitoring set and marked location coordinates of each monitoring point in the femoral head necrosis area of ​​several patient samples.

[0060] Here, the patient sample consists of patients at risk of femoral head collapse. The number of patient samples can be set by the implementer according to actual training needs. The more patient samples, the higher the accuracy of model training. The number of patient samples can exceed 30,000. Generally speaking, the root cause of femoral head collapse is the cumulative fatigue damage caused by cyclic stress generated by daily activities in the fragile necrotic bone and surrounding areas. Therefore, it is necessary to monitor the pressure in the necrotic area of ​​the femoral head to assess the femoral head collapse status of the patient samples. The pressure monitoring set can include pressure monitoring data under several load pressures. The number and magnitude of the load pressures can be set by the implementer according to specific data collection requirements and are not specifically limited. Since the subchondral bone region at the junction of the necrotic area of ​​the femoral head and normal bone bears the greatest stress value, it is necessary to obtain the marked position of each monitoring point in the necrotic area of ​​the femoral head to facilitate subsequent analysis of the distance between the necrotic area of ​​the femoral head and the subchondral bone region.

[0061] As an exemplary implementation, a pressure monitoring set is obtained for each monitoring point within the avascular necrosis region of several patient samples, including:

[0062] In this embodiment, a real-time pressure sensor located in the lower channel of an "8"-shaped dual-channel system is used to acquire pressure monitoring data for each monitoring point within the femoral head necrosis area of ​​the patient sample under various load pressures. The real-time pressure sensor is placed within the femoral head necrosis area, and the monitoring points are evenly distributed throughout the area. For example, a grid method can be used to achieve a uniform distribution of monitoring points, with a one-to-one correspondence between the pressure sensor and the preset monitoring point distribution. A series of known static loads, such as 500N, 800N, and 1200N, are applied to the femoral head to acquire pressure monitoring data under different load pressures. The process of determining the femoral head necrosis area of ​​the patient sample is existing technology and will not be described in detail here.

[0063] The figure-eight dual-channel configuration includes an upper channel and a lower channel, such as... Figure 3 As shown, the upper channel is empirically defined as 5mm, used for filling with bone graft materials such as autologous bone, allogeneic bone, or artificial bone particles. These materials must possess biocompatibility and osteoconductivity. The lower channel is empirically defined as 2.5mm, used for pressure monitoring and negative pressure drainage in the femoral head necrosis area. Furthermore, the lower channel integrates a suitable conduit and pressure transmission medium, facilitating the connection between the pressure sensor and external monitoring equipment to establish a pressure transmission channel. The pressure sensor converts the detected pressure signal into an electrical signal, which is transmitted in real-time to the external monitoring equipment via a wired or wireless transmission module to provide real-time feedback on pressure monitoring data in the femoral head necrosis area.

[0064] It should be noted that, in order to facilitate data analysis and obtain pressure data for the preset monitoring period, the average value of all pressure data within the preset monitoring period is used as the representative pressure monitoring data under the corresponding load pressure. In other words, the representative pressure monitoring data is used as the data to be analyzed and participates in subsequent data processing steps. Both the preset monitoring period and the pressure data collection frequency can be set by the implementer according to specific circumstances, such as setting it to 1 hour and collecting one data point every 10 seconds.

[0065] As an exemplary implementation, the coordinates of the marked location of each monitoring point within the avascular necrosis region of several patient samples are obtained, including:

[0066] In this embodiment, virtual markers are placed on the surface or inside the necrotic area of ​​the three-dimensional model according to the distribution of monitoring points, and each marker is assigned a unique identifier; all marked points are selected, and the point set export function is executed to obtain the coordinate values ​​of each monitoring point in three-dimensional space.

[0067] Thus, this embodiment has obtained pressure monitoring data and marker location coordinates for each monitoring point in the femoral head necrosis area of ​​the patient sample under each load pressure.

[0068] It should be noted that the pressure monitoring data and the marked location coordinates are the data foundation for the entire system, enabling a leap from single morphological data to multimodal biomechanical data. The pressure monitoring dataset contains data under several different load pressures, making it possible to analyze the dynamic mechanical response of bones under different physiological activities (such as walking and running), thus overcoming the shortcomings of traditional imaging data, which are merely static snapshots. The marked location coordinates bind each pressure data point to a precise anatomical location within the femoral head, allowing subsequent analysis to move from overall averaging to regional specificity, and enabling the identification of mechanical abnormalities at specific locations.

[0069] After completing the data acquisition steps, in order to avoid ignoring the differences in physiological characteristics among patient samples during the risk prediction process, all patient samples are clustered. The clustering results can then be used to analyze the relevant data.

[0070] For patients at risk of femoral head collapse, their basic physiological characteristics, such as age, weight, and bone mineral density, significantly influence the risk differences. Risk assessment for each patient sample relies not only on the extent of femoral head necrosis but also on a comprehensive consideration of their physiological characteristics to more accurately predict the damage and collapse process. For example, under the same necrosis conditions, younger patients may have a significantly lower risk of femoral head collapse than older patients due to the better elasticity and repair capacity of their bones. Furthermore, differences in bone mineral density also affect collapse risk; patients with severe osteoporosis experience higher stress concentration under the same load, leading to a much faster rate of damage and collapse accumulation compared to femoral heads with normal bone mineral density. These differences in physiological characteristics determine that different patient samples will exhibit significantly different femoral head collapse risks when facing similar pathological conditions. Therefore, clinical assessment of femoral head collapse risk must comprehensively consider patients' age, weight, bone mineral density, and other physiological characteristics to accurately differentiate the risk levels of different groups.

[0071] As an exemplary implementation, several patient sample groups are obtained, including:

[0072] In this embodiment, physiological feature vectors of several patient samples are obtained, and the physiological feature vectors include at least age, weight, and bone density; based on the differences between the physiological feature vectors of every two patient samples, all patient samples are divided into groups to obtain several patient sample groups.

[0073] Furthermore, the Euclidean distance and modulo operation are used to determine the Euclidean distance between the physiological feature vectors of two patient samples, thereby determining the measurement result of the physiological feature distance between the two patient samples. Based on the Euclidean distance between each pair of patient samples, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering method is used to divide all patient samples into groups, resulting in several patient sample groups.

[0074] The implementation processes of Euclidean distance and DBSCAN clustering methods are existing technologies and will not be elaborated here.

[0075] It should be noted that patient samples from the same patient sample group can exhibit a high degree of similarity in physiological characteristics, which provides a data foundation for subsequent analysis and research. This can help identify unique risk patterns hidden under different physiological characteristics and avoid the neglect of special individuals due to uniformity analysis.

[0076] S2, based on the pressure monitoring set and marker location coordinates of each monitoring point corresponding to each patient sample, analyze the relative abnormalities of the patient samples and determine the daily wear performance of each patient sample.

[0077] Here, the daily wear performance is used at least to characterize the cumulative fatigue loss experienced by the femoral head of the patient sample during daily activities.

[0078] Femoral head collapse is generally caused by the cumulative fatigue damage resulting from cyclic stress generated during daily activities on the fragile necrotic bone and surrounding areas. The subchondral bone region at the junction of the necrotic area and normal bone bears the greatest stress, making it the most direct predictor of collapse. If the maximum pressure exceeds the bone's yield strength, it signifies the onset and accumulation of peripheral fractures, making subsequent macroscopic collapse inevitable.

[0079] As an exemplary implementation, step S2 described above can be achieved through... Figure 4 The steps shown are to be implemented as follows:

[0080] S21, determine the pressure baseline value of each monitoring point in the current patient sample population under each load pressure, and obtain the target location marker sequence of the subchondral bone region of the current patient sample.

[0081] Here, the current patient sample is any single patient sample, and the pressure baseline value can at least reflect the overall pressure magnitude of each load pressure at each monitoring point for the patient sample group to which the current patient sample belongs.

[0082] As an exemplary implementation, determining the pressure baseline value for each monitoring point in the current patient sample population under each load pressure includes:

[0083] In this embodiment, within the population to which the current patient sample belongs, the average value of the pressure monitoring data of all patient samples at the same monitoring point under the same load pressure is calculated, and used as the pressure benchmark value of the corresponding monitoring point of the current patient sample population under the corresponding load pressure.

[0084] As an example, the formula for calculating the baseline pressure value of the j-th monitoring point in the i-th patient sample group under the m-th load pressure can be:

[0085] In the formula, Let N represent the baseline pressure value at the j-th monitoring point of the i-th patient sample group under the m-th load pressure, and let N represent the number of patient samples in the i-th patient sample group. This represents the pressure monitoring data of the j-th monitoring point in the n-th patient sample within the i-th patient sample group under the m-th load pressure.

[0086] Of course, implementers can also use other calculation methods to determine the pressure monitoring data for each monitoring point of each patient sample under each load pressure.

[0087] As an exemplary implementation, obtaining a target location marker sequence for the subchondral bone region of the current patient sample includes:

[0088] In this embodiment, the subchondral bone region at the junction of the necrotic area and normal bone is marked to obtain a target location marking sequence for the subchondral bone region. The femoral head necrosis region and the subchondral bone region can be obtained using existing technologies such as imaging analysis, image segmentation, and manual annotation. The target location marking sequence consists of the coordinate positions corresponding to multiple location marking points within the subchondral bone region.

[0089] It should be noted that the pressure baseline value is used for comparative analysis with each pressure monitoring data to determine the abnormal pressure performance of each monitoring point under each load pressure, while the target location marker sequence is used to measure the distance from the femoral head necrosis area to determine the importance of the abnormal pressure performance of different monitoring points relative to the performance of the patient sample.

[0090] S22, Based on the difference between the pressure monitoring data and the pressure baseline value of each monitoring point in the current patient sample under each load pressure, determine the degree of pressure abnormality of each monitoring point in the current patient sample under each load pressure.

[0091] While the physiological characteristics of different patients within the overall patient sample remain largely similar, significant differences may exist in the risk of femoral head collapse and stress performance among different patients within the same patient population. This is because the degree of femoral head necrosis, the extent of lesions, and the effects of long-term fatigue and wear on the bone structure and load-bearing capacity of individual patients vary. Therefore, pressure monitoring data from the patient sample can characterize the response of the patient's femoral head necrosis to external loads. By analyzing the pressure differences between individual patients and their respective populations, the degree of abnormal pressure performance can be determined.

[0092] As an example, the formula for calculating the degree of pressure abnormality at the j-th monitoring point of the n-th patient sample under the m-th load pressure can be:

[0093] In the formula, This represents the degree of pressure abnormality at the j-th monitoring point of the n-th patient sample under the m-th load pressure. This represents the baseline pressure value at the j-th monitoring point of the i-th patient sample group under the m-th load pressure. This represents the pressure monitoring data of the j-th monitoring point in the n-th patient sample within the i-th patient sample group under the m-th load pressure.

[0094] Formula for calculating abnormal stress performance. Greater than This indicates that the more abnormal the pressure performance of the j-th monitoring point of the n-th patient sample under the m-th load pressure, the greater the degree of pressure abnormality.

[0095] S23, determine the performance weight of each monitoring point in the current patient sample based on the distance between the marker coordinates of each monitoring point corresponding to the current patient sample and each marker in the target location marker sequence.

[0096] The stress manifestations caused by daily wear and fatigue accumulation are mainly based on the bone micro-damage caused by long-term weight-bearing activities and daily exercise. The stress accumulated by these wear and tear is often concentrated in the area close to the subchondral bone. Therefore, the closer the monitoring point in the femoral head necrosis area is to the target location marker sequence in the subchondral bone area, the more reliable the pressure abnormality manifestation of the corresponding monitoring point is, and the greater the manifestation weight of the monitoring point is.

[0097] As an exemplary implementation, determining the performance weight of each monitoring point in the current patient sample includes:

[0098] The first step is to calculate the distance between the marker position of each monitoring point and each marker position in the target location marker sequence, and then perform data fusion processing on all distance values ​​of the same monitoring point to obtain the fusion value of each monitoring point of the current patient sample.

[0099] The second step is to perform inverse proportional normalization on the fusion value of each monitoring point in the current patient sample, and use the obtained normalized value as the performance weight of the corresponding monitoring point in the current patient sample.

[0100] As an example, the formula for calculating the performance weight of the j-th monitoring point of the n-th patient sample can be:

[0101] In the formula, Let represent the performance weight of the j-th monitoring point in the n-th patient sample, exp represent the exponential function with base e, and K represent the number of labeled positions in the target location labeled sequence. This indicates the marked position of the j-th monitoring point within the femoral head necrosis area of ​​the n-th patient sample. This represents the k-th marker position in the target location marker sequence for the subchondral bone region of the nth patient sample.

[0102] In the formula for calculating performance weight, This indicates the calculation of the Euclidean distance between two marked locations. Let represent the fusion value of the j-th monitoring point of the n-th patient sample, and exp(-) is used to perform inverse proportional normalization on the fusion value; A smaller value indicates that the j-th monitoring point within the femoral head necrosis area is closer to the subchondral bone region. Therefore, the reliability of the pressure abnormality at the j-th monitoring point within the femoral head necrosis area is greater, and the performance weight of the j-th monitoring point in the n-th patient sample is also greater. Furthermore, the performance weight of the same monitoring point in each patient sample is equal under each load pressure.

[0103] S24. Based on the pressure abnormality performance of each monitoring point of the current patient sample under each load pressure and the performance weight of each monitoring point, a weighted fusion process is performed to determine the daily wear performance of the current patient sample.

[0104] The greater the degree of abnormal pressure manifestation, the more obvious the collapse characteristics of bone micro-damage caused by weight-bearing activities and daily movements in the corresponding monitoring point within the femoral head necrosis area. Therefore, the degree of abnormal pressure manifestation is positively correlated with the degree of daily wear and tear manifestation. The greater the manifestation weight, the closer the corresponding monitoring point within the femoral head necrosis area is to the subchondral bone region. Since the stress accumulated by wear and tear is often concentrated near the subchondral bone region, the manifestation weight is also positively correlated with the degree of daily wear and tear manifestation.

[0105] As an example, the formula for calculating the daily wear and tear performance of the nth patient sample can be:

[0106] In the formula, This represents the daily wear and tear performance of the nth patient sample, J represents the number of monitoring points within the femoral head necrosis area, and M represents the number of different types of load pressures. This represents the performance weight of the j-th monitoring point in the n-th patient sample. This represents the degree of pressure abnormality at the j-th monitoring point of the n-th patient sample under the m-th load pressure.

[0107] In the formula for calculating the daily wear performance, the greater the performance weight of the j-th monitoring point of the n-th patient sample, the more attention should be paid to the abnormal pressure performance of the j-th monitoring point of the n-th patient sample under each load pressure. The daily wear performance determined by combining the abnormal pressure performance and the performance weight can reflect the abnormal performance of the n-th patient sample under long-term daily wear fatigue accumulation in preoperative monitoring.

[0108] By referring to the calculation process of the daily wear and tear performance of the nth patient sample, the daily wear and tear performance of each patient sample can be obtained.

[0109] It should be noted that step S2 above transforms the raw data into a quantitative indicator reflecting long-term, chronic damage, solving the problem that traditional methods cannot assess bone fatigue. The daily wear performance essentially simulates the fatigue effect of bones due to repeated loading during millions of gait cycles. This indicator can identify patient samples who, although currently not at acute risk, have deteriorated physical condition due to a long-term poor biomechanical environment. Furthermore, it extends the time dimension of risk prediction from the present moment to a continuous process, capturing the characteristic of collapse as a progressive disease, which helps in the early identification of high-risk individuals whose condition deteriorates slowly.

[0110] S3. Based on the pressure monitoring set and marker location coordinates of each monitoring point corresponding to each patient sample, analyze the high stress concentration of the patient sample and determine the local potential risk level of each patient sample.

[0111] In addition to analyzing the aforementioned daily wear and tear characteristics, clinical attention must be paid to potential risks. This refers to situations where, although imaging examinations or clinical assessments may not show particularly significant abnormalities in the necrotic area of ​​the femoral head, concentrated high-stress conditions may exist in certain small, localized areas. While these high-stress areas may not exhibit obvious signs of collapse or damage, their potential risks cannot be ignored. Over time, these small areas of stress concentration may lead to microstructural damage in the local bone, thereby accelerating the osteonecrosis process. Especially without timely detection and intervention, localized high-stress areas may become a hidden danger for disease progression, ultimately leading to femoral head collapse or other serious complications. Therefore, it is necessary to analyze the high-stress concentration of each patient sample based on the pressure monitoring set and marker location coordinates for each monitoring point, and determine the local potential risk level for each patient sample.

[0112] As an exemplary implementation, step S3 described above can be achieved through... Figure 5 The steps shown are to be implemented as follows:

[0113] S31. Based on the pressure monitoring data of each monitoring point corresponding to each patient sample under each load pressure, determine the relative stress stability performance and high stress performance coefficient of each monitoring point.

[0114] Here, both relative stress stability performance and high stress performance coefficient are used to determine the final high stress concentration region. For a high stress concentration region, "high" means that the pressure value in the region is large, while "concentrated" means that the pressure values ​​in the region are similar. In other words, a high stress concentration region refers to a region where the internal stress values ​​are large and the values ​​are concentrated (similar).

[0115] As an exemplary implementation, determining the relative stress stability performance at each monitoring point includes:

[0116] The first step is to determine the relative pressure performance value of each monitoring point under each load pressure based on the pressure monitoring data of each monitoring point corresponding to each patient sample.

[0117] In this embodiment, for each monitoring point, the pressure monitoring data under each load pressure is compared with the maximum pressure monitoring data under the same load pressure. The ratio is used as the relative pressure performance value, and the relative pressure performance value of each monitoring point under each load pressure can be obtained.

[0118] The second step is to analyze the pressure fluctuation based on the relative pressure performance value of each monitoring point under each load pressure, and determine the relative stress stability performance of each monitoring point.

[0119] In this embodiment, the variance or standard deviation of all relative pressure performance values ​​corresponding to the same monitoring point is calculated. The variance or standard deviation of all relative pressure performance values ​​is then inversely normalized, and the normalized value is used as the relative stress stability performance of the corresponding monitoring point. The inverse normalization process can be achieved using the exponential function exp(-), although other inverse normalization methods can also be employed.

[0120] It should be noted that the relative stress stability of the monitoring point may be caused by a continuous low pressure or a continuous high pressure. Therefore, in order to determine the final high stress concentration area, it is also necessary to analyze the pressure of each monitoring point under multiple loads.

[0121] As an exemplary implementation, determining the high-stress performance coefficient for each monitoring point includes:

[0122] The first step is to calculate the average value of the pressure monitoring data at the same monitoring point for each patient sample under each load pressure, and record it as the high stress performance factor.

[0123] The second step is to normalize the high-stress performance factor for each monitoring point corresponding to the patient sample, and use the normalized value as the high-stress performance coefficient for the corresponding monitoring point.

[0124] In this embodiment, the normalization function can be used to normalize the high stress performance factor of each monitoring point. Of course, the implementer can also use other normalization methods, which are not specifically limited here.

[0125] S32, based on the relative stress stability performance and high stress performance coefficient of each monitoring point corresponding to each patient sample, determine the potential risk of the relatively high stress concentration area of ​​each patient sample.

[0126] Here, the potential risk level is used to characterize the likelihood of potential collapse in areas of relatively high stress concentration. The higher the potential risk level, the more likely there is a potential risk of femoral head collapse in areas of relatively high stress concentration.

[0127] As an exemplary implementation, obtaining the local potential risk level for each patient sample includes:

[0128] The first step is to determine the sustained relative high stress performance of each monitoring point for each patient sample based on the relative stress stability performance and high stress performance coefficient of each monitoring point.

[0129] Using the high stress performance coefficient as a weight, the more prominent the high stress performance, the more likely the relative stress stability performance of the corresponding monitoring point is to be a continuous high pressure performance, and the greater the continuous relative high stress performance.

[0130] In this embodiment, for each patient sample, the product of the relative stress stability performance and the high stress performance coefficient at the same monitoring point is calculated as the continuous relative high stress performance of the corresponding monitoring point.

[0131] The second step involves using the sustained relatively high stress performance of each monitoring point corresponding to each patient sample to divide all monitoring points corresponding to the patient sample into regions, thereby obtaining relatively high stress concentration areas.

[0132] In this embodiment, for each patient sample, based on the continuous relative high stress performance of each monitoring point corresponding to the patient sample, the monitoring points in the femoral head necrosis area are divided into two categories by the Otsu's method: one category is high stress concentration points and the other category is non-high stress concentration points. The connected region formed by the high stress concentration points is taken as the relative high stress concentration region of the patient sample.

[0133] In this embodiment, isolated points are not analyzed, meaning they are not included in the process of determining relatively high stress concentration areas. An isolated point is defined as a high stress concentration point that is far from all other high stress concentration points, such as a distance greater than a distance threshold. This distance threshold can be set by the implementer based on specific circumstances; for example, an empirical value could be twice the shortest distance between monitoring points. The implementation process of the Otsu's method is existing technology and will not be described further here.

[0134] The third step is to determine the potential risk level of the relatively high stress concentration area for each patient sample based on the sustained relatively high stress performance of each monitoring point within the relatively high stress concentration area.

[0135] In this embodiment, for each patient sample, the average value of the sustained relative high stress performance of all monitoring points in the relative high stress concentration area corresponding to the same patient sample is calculated as the potential risk level of the relative high stress concentration area of ​​the corresponding patient sample.

[0136] S33, determine the attention weight of the relatively high stress concentration area of ​​each patient sample based on the distance between each monitoring point in the relatively high stress concentration area and each marker position in the target location marker sequence.

[0137] The presence of relatively high stress concentration areas is an important factor leading to the potential risk of collapse in patients. In addition, it is necessary to fully consider the distance between the relatively high stress concentration areas and the subchondral bone area. If the relatively high stress concentration areas are close to the subchondral bone area, it means that there is no need to pay too much attention to the potential risk of collapse. The daily wear performance calculated above has already covered the analytical characteristics of high stress concentration, and the weight of attention to relatively high stress concentration areas should be smaller.

[0138] In this embodiment, for each patient sample, the shortest distance between each monitoring point in the relatively high stress concentration area of ​​the patient sample and each marker position in the target position marker sequence is calculated using the Euclidean distance algorithm. The shortest distance value is normalized using the minimum-maximum normalization algorithm, and the normalized value is used as the attention weight of the relatively high stress concentration area of ​​the patient sample.

[0139] S34, the potential risk level and attention weight of the relatively high stress concentration area of ​​the same patient sample are weighted and fused to obtain the local potential risk level of each patient sample.

[0140] In this embodiment, for each patient sample, the product of the potential risk level of the relatively high stress concentration area of ​​the patient sample and the attention weight is calculated as the local potential risk level of the corresponding patient sample.

[0141] Thus, this embodiment has obtained the local potential risk level for each patient sample.

[0142] It should be noted that step S3 above transforms the raw data into a quantitative indicator reflecting the risk of acute, localized failure, solving the problem that traditional methods cannot accurately locate mechanically weak points. Specifically, by analyzing high stress concentration, it can directly identify relatively high stress concentration areas in the bone where microfractures and structural collapse are most likely to occur, which directly corresponds to the pathophysiological mechanism of collapse. Even if the total volume of the necrotic area is not large, the existence of an extremely high local potential risk indicates an extremely high risk of collapse, thus solving the problem of missed diagnoses caused by traditional methods based on overall volume assessment.

[0143] S4. Based on the daily wear and tear performance and local potential risk level of each patient sample, determine the multidimensional comprehensive risk assessment value of each patient sample.

[0144] Here, the multidimensional comprehensive risk assessment value is one of the input variables for training the prediction model, and the prediction model can predict the patient's preoperative risk of femoral head collapse. The prediction model can use a neural network as the model structure.

[0145] As an exemplary implementation, determining the multidimensional comprehensive risk assessment value for each patient sample includes:

[0146] The first step is to count the number of femoral head collapses in the medical records of each patient sample group, and then determine the risk weight of each patient sample in each patient sample group based on the number of femoral head collapses.

[0147] Each patient sample group possesses corresponding group-specific characteristics, meaning each group also exhibits its own risk profile. For high-risk groups, the focus can be on the patients' apparent risks, namely the degree of stress accumulation from daily wear and tear. For low-risk groups, the focus can be on the patients' potential risks, namely the degree of localized potential risk. Therefore, it is necessary to measure the risk profile of each patient sample group by analyzing the number of femoral head collapses recorded in their medical records, and to determine the risk weight for each patient sample within each group.

[0148] In this embodiment, for each patient sample group, the number of femoral head collapses corresponding to that patient sample group is counted. The ratio of the number of femoral head collapses corresponding to that patient sample group to the total number of femoral head collapses is used as the risk weight for each patient sample in the patient sample group. The risk weight is the same for each patient sample within the same patient sample group, and the total number of femoral head collapses refers to the sum of the number of femoral head collapses corresponding to all patient sample groups.

[0149] The second step involves using the risk weights of each patient sample to perform weighted fusion processing on the daily wear and tear performance and local potential risk of each patient sample, thereby obtaining a multidimensional comprehensive risk assessment value for each patient sample.

[0150] As an example, the formula for calculating the multidimensional comprehensive risk assessment value of the nth patient sample can be:

[0151] In the formula, This represents the multidimensional comprehensive risk assessment value of the nth patient sample. This represents the risk weight of the patient sample group to which the nth patient sample belongs. This represents the daily wear and tear performance of the nth patient sample. This represents the local potential risk level of the nth patient sample.

[0152] In the formula for calculating the multidimensional comprehensive risk assessment value, the larger the risk weight of the nth patient sample, the greater the likelihood that the nth patient sample belongs to a high-risk group. Therefore, more attention should be paid to daily wear and tear performance when calculating the multidimensional comprehensive risk assessment value. Conversely, the smaller the risk weight of the nth patient sample... The larger the value, the greater the likelihood that the nth patient sample belongs to a low-risk group. Therefore, more attention should be paid to the local potential risk level when calculating the multidimensional comprehensive risk assessment value.

[0153] Therefore, by referring to the multidimensional comprehensive risk assessment value of the nth patient sample, the multidimensional comprehensive risk assessment value of each patient sample can be obtained.

[0154] It should be noted that step S4 above creates a completely new and comprehensive risk assessment standard through information fusion. This standard combines indicators reflecting long-term health and those reflecting short-term acute risk, and uses the combined result as input to the predictive model. Because the multidimensional comprehensive risk assessment value contains far more information than traditional imaging features such as biomechanical variation, spatial distribution, and temporal cumulative effects, it directly and effectively solves the core deficiency of insufficient representativeness of training data, and is the most crucial link in ultimately improving the model's predictive accuracy.

[0155] After obtaining the multidimensional comprehensive risk assessment value for each patient sample, this value is used as one of the input data for a model predicting preoperative femoral head collapse risk; that is, it serves as training data for the neural network, thus training a well-trained predictive model. The input data for the neural network consists of the multidimensional comprehensive risk assessment value and traditional imaging data. The training process for the neural network is based on existing technology and will not be detailed here.

[0156] It should be noted that if some indicators cannot be calculated when calculating the multidimensional comprehensive risk assessment value of the patient to be predicted, it means that the patient to be predicted lacks the original data for calculating the corresponding indicators. In this case, the corresponding indicators can be directly defaulted to zero.

[0157] In summary, the model training process for predicting preoperative femoral head collapse risk in this invention considers the influence of multiple feature factors on top of traditional single-factor assessment, using the determined multi-dimensional comprehensive risk assessment value as the core input feature of the prediction model. This assessment value, based on traditional imaging assessment, directly incorporates biomechanical factors that determine collapse occurrence, thereby significantly improving the accuracy of preoperative femoral head collapse risk prediction.

[0158] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A preoperative femoral head collapse risk prediction system based on biomechanical analysis, characterized by, The method comprises the following steps: A data acquisition module is configured to acquire pressure monitoring sets and marker position coordinates of each monitoring point in the femoral head necrosis area of a plurality of patient samples, wherein the pressure monitoring set comprises pressure monitoring data under a plurality of load pressures; A first determination module is configured to analyze the relative abnormality of the patient sample according to the pressure monitoring set and the marker position coordinates of each monitoring point corresponding to each patient sample, and determine the daily wear performance degree of each patient sample; A second determination module is configured to analyze the high stress concentration of the patient sample according to the pressure monitoring set and the marker position coordinates of each monitoring point corresponding to each patient sample, and determine the local potential risk degree of each patient sample; A third determination module is configured to determine the multi-dimensional comprehensive risk assessment value of each patient sample according to the daily wear performance degree and the local potential risk degree of each patient sample, wherein the multi-dimensional comprehensive risk assessment value is one of the input variables of the training prediction model; The first determination module comprises: A data determination unit is configured to determine the pressure reference value of each monitoring point of the current patient sample under each load pressure, and acquire the target position marker sequence of the subchondral bone area of the current patient sample, wherein the current patient sample is any one of the patient samples; A first determination unit is configured to determine the pressure abnormal performance degree of each monitoring point of the current patient sample under each load pressure according to the difference between the pressure monitoring data and the pressure reference value of each monitoring point of the current patient sample under each load pressure; A second determination unit is configured to determine the performance weight of each monitoring point of the current patient sample according to the distance between the marker position coordinates of each monitoring point corresponding to the current patient sample and each position marker in the target position marker sequence; A third determination unit is configured to determine the daily wear performance degree of the current patient sample by performing weighted fusion processing on the pressure abnormal performance degree of each monitoring point of the current patient sample under each load pressure and the performance weight of each monitoring point; The second determination module comprises: A fourth determination unit is configured to determine the relative stress stability performance degree and the high stress performance coefficient of each monitoring point according to the pressure monitoring data of each monitoring point corresponding to each patient sample under each load pressure; A fifth determination unit is configured to determine the potential risk degree of the relative high stress concentration area of each patient sample according to the relative stress stability performance degree and the high stress performance coefficient of each monitoring point corresponding to each patient sample; A sixth determination unit is configured to determine the attention weight of the relative high stress concentration area of each patient sample according to the distance between each monitoring point in the relative high stress concentration area of each patient sample and each marker position in the target position marker sequence; A data fusion unit is configured to perform weighted fusion processing on the potential risk degree and the attention weight of the relative high stress concentration area of the same patient sample to obtain the local potential risk degree of each patient sample.

2. A pre-operative femoral head collapse risk prediction system based on biomechanical analysis as claimed in claim 1, wherein, The data acquisition module is further configured to: Acquire a physiological feature vector of the plurality of patient samples, wherein the physiological feature vector at least comprises age, weight, and bone density; According to the difference between the physiological feature vectors of every two patient samples, all the patient samples are divided into groups to obtain a plurality of patient sample groups.

3. A pre-operative femoral head collapse risk prediction system based on biomechanical analysis as claimed in claim 1, wherein, The determination of the pressure reference value of each monitoring point under each load pressure in the group where the current patient sample is located comprises: In the group where the current patient sample is located, the average value of the pressure monitoring data of all the patient samples under the same load pressure at the same monitoring point is calculated as the pressure reference value of the corresponding monitoring point under the corresponding load pressure in the group where the current patient sample is located.

4. A pre-operative femoral head collapse risk prediction system based on biomechanical analysis as claimed in claim 1, wherein, The execution steps of the second determination unit comprise: The distance value between the marker position of each monitoring point and each marker position in the target position marker sequence is calculated, and then all the distance values of the same monitoring point are subjected to data fusion processing to obtain the fusion value of each monitoring point of the current patient sample; The fusion value of each monitoring point of the current patient sample is subjected to inverse proportional normalization processing, and the obtained normalized value is taken as the performance weight of the corresponding monitoring point of the current patient sample.

5. A pre-operative femoral head collapse risk prediction system based on biomechanical analysis as claimed in claim 1, wherein, The execution steps of the fourth determination unit comprise: According to the pressure monitoring data of each monitoring point under each load pressure corresponding to each patient sample, the relative pressure performance value of each monitoring point under each load pressure is determined; According to the relative pressure performance value of each monitoring point under each load pressure, the pressure change fluctuation is analyzed to determine the corresponding stress stability performance degree of each monitoring point; According to the pressure monitoring data of each monitoring point under each load pressure corresponding to each patient sample, the high stress performance coefficient of each monitoring point is determined.

6. A pre-operative femoral head collapse risk prediction system based on biomechanical analysis as claimed in claim 5 wherein, The determination of the high stress performance coefficient of each monitoring point according to the pressure monitoring data of each monitoring point under each load pressure corresponding to each patient sample comprises: For each patient sample, the average value of the pressure monitoring data of the same monitoring point under each load pressure corresponding to the patient sample is calculated, which is denoted as a high stress performance factor; The high stress performance factor of each monitoring point corresponding to the patient sample is subjected to normalization processing, and the obtained normalized value is taken as the high stress performance coefficient of the corresponding monitoring point.

7. A pre-operative femoral head collapse risk prediction system based on biomechanical analysis as claimed in claim 1, wherein, The execution steps of the fifth determination unit comprise: According to the corresponding stress stability performance degree and the high stress performance coefficient of each monitoring point corresponding to each patient sample, the sustained relative high stress performance degree of each monitoring point corresponding to each patient sample is determined; Using the sustained relative high stress performance degree of each monitoring point corresponding to each patient sample, the region of all the monitoring points corresponding to the patient sample is divided to obtain a relative high stress concentration region; According to the sustained relative high stress performance degree of each monitoring point in the relative high stress concentration region, the potential risk degree of the relative high stress concentration region of each patient sample is determined.

8. A pre-operative femoral head collapse risk prediction system based on biomechanical analysis as claimed in claim 2, wherein, The execution steps of the third determination module comprise: The number of femoral head collapses in the medical records of each patient sample group is counted, and the risk weight of each patient sample in each patient sample group is determined according to the number of femoral head collapses; The daily wear performance degree and the local potential risk degree of each patient sample are weighted and fused by using the risk weight of each patient sample to obtain a multi-dimensional comprehensive risk assessment value of each patient sample.

Images