A method for generating a bone health index

Abstract

The present invention discloses a measurement and judgment method for orthopedic diagnosis, which belongs to the technical field of orthopedic diagnosis and treatment, comprising: collecting magnetic field signals of bone metabolism through a magnetic field sensor array, collecting electrical impedance spectroscopy data through a four-electrode method to form an original data set; filtering and denoising the magnetic field signals and correcting motion artifacts, and then fitting the electrical impedance spectroscopy data with a Cole-Cole model to extract bone density, porosity, and moisture content; calculating bone metabolic current density through a magnetic field-current inversion algorithm, combining electrical impedance characteristics and current characteristics, and generating a bone health index through a dynamic weight distribution mechanism; by combining the magnetic field-current inversion algorithm with Cole-Cole model analysis, multi-dimensional information of bone metabolic activity and bone tissue structure can be simultaneously obtained, thereby effectively overcoming the problem of insufficient information from a single physical field and significantly improving the accuracy of diagnosis.

Description

Technical Field

[0001] The present invention relates to a method for generating a medical bone health index, and in particular to a method for generating a bone health index. Background Art

[0002] In the field of orthopedics, generating accurate bone health indices and effective treatments are key to improving patients' quality of life and preventing the progression of bone diseases. Traditional bone metabolic activity assays, such as biochemical markers or imaging studies (e.g., X-rays, CT scans, and MRIs), while able to reflect bone health to a certain extent, have numerous limitations. Biochemical markers typically provide only indirect information and exhibit a lag, failing to reflect dynamic changes in bone metabolism in real time. Imaging studies may involve radiation exposure and lack sensitivity for identifying early-stage bone lesions.

[0003] In recent years, electrical impedance technology has gained increasing attention in the medical field as a non-invasive, radiation-free testing method. By measuring the electrical properties of biological tissue, this technology can reveal changes in its structure, composition, and physiological state. However, single-modality electrical impedance data is susceptible to interference from various factors, such as individual differences and measurement conditions, making it difficult to comprehensively and accurately assess bone health. Summary of the Invention

[0004] The present invention overcomes the deficiencies of the prior art and provides a method for generating a bone health index.

[0005] To achieve the above object, the present invention adopts the following technical solution: a method for generating a bone health index, comprising the following steps:

[0006] The magnetic field sensor array collects magnetic field signals of bone metabolism, performs filtering and noise removal, and corrects motion artifacts. Then, the electrical impedance spectroscopy data collected by the four-electrode method is fitted with a Cole-Cole model to extract bone density, porosity, and water content.

[0007] The bone metabolic current density is calculated using the magnetic field-current inversion algorithm, and the electrical impedance characteristics and current characteristics are combined to generate the bone health index through a dynamic weight distribution mechanism.

[0008] In a preferred embodiment of the present invention, personalized ultrasound intervention parameters are output through a reinforcement learning algorithm based on the temporal changes of the bone health index.

[0009] In a preferred embodiment of the present invention, the magnetic field sensor array adopts a three-dimensional orthogonal layout, and the sampling frequency 1kHz, spatial resolution 0.1mm3, the acquisition frequency range of electrical impedance spectrum is 10Hz~1MHz.

[0010] In a preferred embodiment of the present invention, the Cole-Cole model fitting adopts nonlinear least squares method, and the optimization objective function is:

[0011] ;

[0012] Where, are the parameters to be optimized for the Cole-Cole model; is the measured impedance; N is the total number of frequency points; i is the time sequence frequency point; is the angular frequency of the i-th frequency point.

[0013] In a preferred embodiment of the present invention, the magnetic field-current inversion algorithm converts the data using the following formula:

[0014] ;

[0015] Where, The current density generated by bone cell metabolism; The Hamiltonian operator is used in electromagnetism to describe the spatial variation characteristics of vector fields; is the vacuum permeability; It is the real-time magnetic field strength, collected by the magnetic field sensor array; The electrical conductivity of bone tissue is calibrated in real time through electrical impedance data;

[0016] Among them, the conductivity is corrected in real time through the electrical impedance data, and the correction formula is:

[0017] ;

[0018] Where, is bone density; is bone porosity; is the real-time moisture content; It is the baseline value of water content in healthy bone tissue; is the pathological threshold.

[0019] In a preferred embodiment of the present invention, the dynamic weight allocation mechanism is implemented by an attention network, and the weight update formula is:

[0020] ;

[0021] Where, is the updated i-th modal weight, which is dynamically adjusted through the attention mechanism; is the state-action value function; is the feature dimension; the Softmax function ensures weight normalization.

[0022] In a preferred embodiment of the present invention, the reward function of the reinforcement learning algorithm is:

[0023] ;

[0024] Where r is the timely reward; is the improvement degree of bone health index; is the change value of ultrasound intensity; is the change in treatment time; 、 、 All are adjustable weight coefficients to balance the effects of bone health index improvement, ultrasound intensity changes, and treatment time changes.

[0025] In a preferred embodiment of the present invention, the dynamic mapping relationship between the bone health index and the ultrasound parameters is described by the following formula:

[0026] ;

[0027] ;

[0028] ;

[0029] Where, f is the actual ultrasonic frequency and the reference frequency; is the default treatment frequency; is the time derivative of the bone health index, reflecting the rate of change of bone metabolic activity; I is the actual ultrasound intensity and the baseline intensity; is the safety threshold; T is the actual treatment time and the benchmark time; The duration of a single treatment; is the frequency adjustment coefficient; Intensity adjustment coefficient; is the time adjustment coefficient.

[0030] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0031] (1) By combining the magnetic field-current inversion algorithm with the Cole-Cole model analysis, multi-dimensional information on bone metabolic activity and bone tissue structure can be obtained simultaneously. This multimodal fusion strategy effectively overcomes the problem of insufficient information from a single physical field and significantly improves the accuracy of bone health index generation.

[0032] (2) By modeling the temporal correlation of data through long short-term memory networks, it is possible to capture subtle changes in bone metabolic activity and achieve early identification of micro-injuries, which in turn helps to intervene in the early stages of the disease, prevent the disease from worsening, and improve the treatment effect.

[0033] (3) Based on the temporal changes of the bone health index, a personalized ultrasound intervention plan is generated through a reinforcement learning algorithm, so that the adaptive treatment algorithm can dynamically adjust the ultrasound parameters according to the patient's real-time status to ensure the personalization and accuracy of the treatment plan.

[0034] (4) Personalized ultrasound intervention plans can more accurately target the lesion site and improve the treatment effect. At the same time, through continuous monitoring and dynamic adjustment, it can help reduce unnecessary treatment side effects and improve patients' treatment compliance and quality of life. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive efforts.

[0036] Figure 1 4 is a flow chart of a method for generating a bone health index according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0038] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0039] The present invention provides a method for generating a bone health index. By utilizing a three-dimensional orthogonal array of magnetic field sensors, the method collects magnetic field signals generated by bone metabolism in real time. This ensures precise spatial resolution and is capable of capturing magnetic field changes caused by the weak currents generated by bone cell metabolism.

[0040] At the same time, the four-electrode method collects electrical impedance spectroscopy data over a wide frequency band (10 Hz to 1 MHz). This data contains key information about the electrical properties of bone tissue, such as bone density and porosity.

[0041] The collected magnetic field signals are filtered, de-noised and corrected for motion artifacts to eliminate the impact of external interference and patient movement on the data and improve data accuracy.

[0042] The Cole-Cole model was used to fit the electrical impedance spectroscopy data and extract biomechanical parameters such as bone density, porosity, and water content. These parameters can reflect changes in the structure and composition of bone tissue.

[0043] Bone metabolic current density is calculated based on a magnetic field-current inversion algorithm combined with real-time calibrated conductivity. This step converts magnetic field signals into current density information reflecting bone metabolic activity.

[0044] Using an attention network, the system dynamically assigns weights based on the reliability and relevance of each modality to generate a bone health index (BHI). The BHI integrates information from magnetic field, electrical impedance, and current modalities to comprehensively reflect the overall health of the skeleton.

[0045] Based on the temporal changes in BHI, a reinforcement learning algorithm is used to output personalized ultrasound intervention parameters. These parameters, including ultrasound frequency, intensity, and duration, can be adjusted according to the patient's real-time condition.

[0046] Through a pre-set mapping relationship, changes in BHI are converted into specific ultrasound treatment parameters. This step ensures the personalization and precision of ultrasound treatment, and can provide an optimized treatment plan based on the patient's bone health status.

[0047] Specifically, if Figure 1 As shown, a method for generating a bone health index comprises the following steps:

[0048] S1. The magnetic field signal of bone metabolism is collected through a magnetic field sensor array, and the electrical impedance spectroscopy data is collected through a four-electrode method to form an original data set.

[0049] The magnetic field sensor array adopts a three-dimensional orthogonal layout, and the sampling frequency 1kHz, spatial resolution 0.1mm3, the acquisition frequency range of electrical impedance spectroscopy is 10Hz~1MHz. A three-dimensional orthogonal magnetic field sensor array with a sampling frequency of 1kHz and a spatial resolution of 0.1mm³ is attached to the surface of the bone to be tested (such as the lumbar spine and femoral neck) to continuously acquire time-varying magnetic field signals related to bone metabolism. Using the four-electrode method, an impedance analyzer with a frequency range of 10Hz~1MHz and an excitation current of 10 , collect electrical impedance spectrum data of bone tissue, and synchronously record temperature and contact pressure to correct measurement deviation.

[0050] The magnetic field sensor array collects magnetic field signals of bone metabolism, while the four-electrode method simultaneously collects electrical impedance spectroscopy data to form a raw data set. This step aims to obtain multi-physics information about bone tissue metabolism, providing basic data for subsequent analysis. The magnetic field signal reflects the weak current generated by bone cell metabolism, while the electrical impedance spectrum contains structural information such as bone density and porosity.

[0051] A three-dimensional orthogonal magnetic field sensor array ensures a spatial resolution better than 0.1 mm³, capturing the bone metabolism-related magnetic field signal B(t) in real time. Simultaneously, a four-electrode method acquires electrical impedance spectroscopy data in the 10 Hz to 1 MHz frequency range, generating a raw data set containing both time and frequency domain information.

[0052] This invention uses a magnetic field sensor array and a four-electrode method to simultaneously collect bone metabolism magnetic field signals and electrical impedance spectroscopy data. This dual-modal data acquisition method can simultaneously capture information about bone cell metabolic activity and the electrical properties of bone tissue, providing a comprehensive data foundation for subsequent multimodal fusion analysis. This method generates high-temporal and spatial resolution magnetic field data and broadband electrical impedance spectroscopy, laying the foundation for subsequent feature extraction and multimodal fusion.

[0053] S2. Filter and denoise the magnetic field signal and correct for motion artifacts, then fit the electrical impedance spectroscopy data with the Cole-Cole model to extract bone density, porosity, and moisture content.

[0054] The Cole-Cole model is fitted using the nonlinear least squares method, and the optimization objective function is:

[0055] ;

[0056] Where, are the parameters to be optimized for the Cole-Cole model; is the measured impedance; N is the total number of frequency points; i is the time sequence frequency point; is the angular frequency of the i-th frequency point.

[0057] The magnetic field signal was filtered, de-noised, and motion artifact corrected. The electrical impedance spectroscopy data were then fitted with the Cole-Cole model to extract bone density, porosity, and water content.

[0058] Specifically, after fitting the Cole-Cole model to the electrical impedance spectroscopy data, the model parameters are converted into actual physiological indicators using pre-established biocalibration relationships. Bone density is modeled by linearly regressing the low-frequency electrical impedance term with bone density values ​​measured by dual-energy X-ray absorptiometry to form a quantitative conversion formula. Porosity is nonlinearly fitted with distribution parameters and porosity data obtained by micro-CT scanning to establish an exponential correlation model with porosity. Moisture content is logarithmically calibrated with moisture content measured by nuclear magnetic resonance (NMR) using the imaginary eigenfrequency to form a piecewise linear function with moisture content. This conversion process requires parameter training using standard samples upon initial use of the device and regular dynamic calibration using phantoms with known physiological indicators during clinical application to ensure the timeliness of the conversion relationship and adaptability to individual variability. This multimodal parameter fusion approach resolves the nonlinear mapping problem between Cole-Cole model parameters and physiological indicators, enabling quantitative conversion from electrical properties to histological features, improving data quality, eliminating noise interference, and resolving key biomechanical parameters from electrical impedance spectroscopy.

[0059] A wavelet threshold denoising algorithm is then used to effectively suppress high-frequency noise; the patient's motion status is monitored in real time through an inertial measurement unit to correct motion artifacts.

[0060] The nonlinear least squares method was then used to fit the Cole-Cole model, optimize the objective function, and extract bone density, porosity, and moisture content.

[0061] By converting magnetic field signals into current density information reflecting bone metabolic activity using a magnetic field-current inversion algorithm, and simultaneously fitting the electrical impedance spectroscopy data using the Cole-Cole model, key biomechanical parameters such as bone density, porosity, and moisture content are extracted. This multimodal data fusion strategy enables a more comprehensive assessment of bone health, improving the magnetic field signal-to-noise ratio by over 20dB, achieving a Cole-Cole model fitting error of less than 2%, and achieving a bone density measurement accuracy of 0.01g / cm³.

[0062] S3. Calculate the bone metabolic current density through the magnetic field-current inversion algorithm, combine the electrical impedance characteristics with the current characteristics, and generate the bone health index through a dynamic weight distribution mechanism.

[0063] The magnetic field-current inversion algorithm converts the data using the following formula:

[0064] ;

[0065] Where, The current density generated by bone cell metabolism; The Hamiltonian operator is used in electromagnetism to describe the spatial variation characteristics of vector fields; is the vacuum permeability; It is the real-time magnetic field strength, collected by the magnetic field sensor array; The electrical conductivity of bone tissue is calibrated in real time through electrical impedance data;

[0066] Among them, the conductivity is corrected in real time through the electrical impedance data, and the correction formula is:

[0067] ;

[0068] Where, is bone density; is bone porosity; is the real-time moisture content; It is the baseline value of water content in healthy bone tissue; is the pathological threshold.

[0069] The dynamic weight allocation mechanism is implemented through the attention network, and the weight update formula is:

[0070] ;

[0071] Where, is the updated i-th modal weight, which is dynamically adjusted through the attention mechanism; It is the state-action value function, specifically: Next action The transition-action value function, that is, the long-term expected cumulative reward starting from the current state-action pair, is used in the bone health index generation system. May contain Bone Health Index (BHI), electrical impedance spectroscopy data, or magnetic field signals, while It may be the adjustment of parameters such as the frequency and intensity of ultrasound intervention; is the feature dimension; the Softmax function ensures weight normalization.

[0072] It is further explained that the state-action value function The derivation process includes:

[0073] ;

[0074] Where, It is a conditional expectation operator, which means that given the current state and actions The statistical average of future cumulative rewards under the condition of . It reflects the system's ability to model uncertainty, such as individual differences in patients or measurement noise; To perform an action The immediate reward obtained after is defined by the reward function; is a discount factor used to balance the importance of immediate rewards and future rewards. When it is close to 1, the system focuses more on long-term benefits; when it is close to 0, it focuses on short-term returns. The value of needs to be adjusted in accordance with clinical timeliness requirements; Indicates the next state All possible actions The maximum Q value of is the optimal future benefit. It reflects the balance of reinforcement learning. The system maximizes the global benefit by selecting the current action and anticipating the optimal action in the future state.

[0075] Bone metabolic current density is calculated using a magnetic field-current inversion algorithm. Combining electrical impedance and current characteristics, a dynamic weighting mechanism is used to generate a Bone Health Index (BHI). This quantifies bone metabolic activity and integrates multimodal information to generate an index that reflects the overall state of the skeleton.

[0076] Based on a magnetic field-current inversion algorithm and real-time calibrated conductivity, bone metabolic current density is calculated. Using an attention network, the weights of magnetic field, electrical impedance, and current modalities are dynamically adjusted through a softmax function to generate a bone health index.

[0077] An attention network is used to implement a dynamic weight allocation mechanism, dynamically adjusting the weights of different modal data in the Bone Health Index (BHI) calculation based on the reliability and relevance of each modality. This dynamic weight allocation mechanism ensures that the BHI can more accurately and comprehensively reflect the overall health of the skeleton. The correlation between the BHI and bone metabolic activity is over 0.85. Dynamic weight allocation improves the accuracy of multimodal fusion by 15%.

[0078] Specifically, the structural design of the attention network adopts a multi-head attention mechanism, taking three types of features: magnetic field signals, electrical impedance spectrum data, and bone metabolic current density as input, generating query vectors, key vectors, and value vectors through linear transformation, and using scaled dot product attention to calculate the correlation between each modality. Finally, the weights are normalized through the Softmax function to ensure that the sum of the weights of different modal data is 1.

[0079] The weight calculation is based on the ratio of the state-action value function to the feature dimension. The state-action value function reflects the long-term expected reward of the combination of the current bone state and the ultrasound intervention action, while the feature dimension controls the smoothness of the weight distribution. This calculation method gives higher weights to high-reward modalities and automatically reduces the weights of low-quality or low-relevance modalities. The dynamic adjustment strategy is implemented using two dimensions: the reliability dimension assesses data quality using the signal-to-noise ratio of the magnetic field signal and the contact pressure stability of the electrical impedance spectroscopy data; and the relevance dimension uses the Pearson correlation coefficient to measure the strength of the association between each modal feature and the bone health index. The conditional expectation operator is introduced during weight update to account for individual patient differences and measurement noise, ensuring robust weight adjustment. The interaction with the reinforcement learning algorithm is manifested at three levels: the state space contains the temporal variation characteristics of the bone health index, the action space is defined as a combination of ultrasound frequency, intensity, and duration parameters, and the reward function balances the BHI improvement magnitude, ultrasound intensity change, and treatment time cost through a weighted sum. The dynamic weights generated by the attention network directly participate in the update of the Q function, forming a closed-loop optimization of the bone health index generation, intervention, and feedback loop.

[0080] Establishing this mechanism requires the following process: In the data preprocessing stage, a three-dimensional orthogonal magnetic field sensor array is first deployed to collect raw magnetic field signals at a sampling frequency of 1kHz and a spatial resolution of 0.1mm³. Simultaneously, electrical impedance spectroscopy data are acquired in the frequency range of 10Hz to 1MHz using a four-electrode method. During the acquisition process, an inertial measurement unit is used to monitor the patient's motion status in real time. The magnetic field signal is subjected to wavelet threshold denoising and motion artifact correction. After temperature and contact pressure compensation are performed on the electrical impedance spectroscopy data, the Cole-Cole model is fitted using the nonlinear least squares method to extract biomechanical parameters such as bone density, porosity, and moisture content.

[0081] During the model training phase, a biological calibration relationship library is first established to map the Cole-Cole model parameters to actual physiological indicators. Subsequently, the attention network weights are initialized, and backpropagation training is performed using a standard sample dataset. During clinical application, dynamic calibration is performed weekly using a phantom containing known physiological indicators to ensure the individual adaptability of the conversion relationship.

[0082] The weight update rule uses a three-stage mechanism: the initialization phase sets prior weights based on historical diagnosis and treatment data; the runtime phase calculates the ratio of the current state-action value function to the feature dimension at each time step, generating real-time weights through softmax; the feedback phase adjusts the attention network parameters based on the reinforcement learning reward function, and automatically triggers weight redistribution when the BHI improvement falls below a threshold, ensuring that the intervention strategy always focuses on the optimal feature combination. This mechanism, through the continuous integration of multi-source physiological signals and treatment feedback data, enables bone health assessment to transition from a static indicator to a dynamic predictive model.

[0083] Based on the temporal changes of bone health index, personalized ultrasound intervention parameters are output through reinforcement learning algorithm.

[0084] The reward function of the reinforcement learning algorithm is:

[0085] ;

[0086] Where r is the timely reward; is the improvement degree of bone health index; is the change value of ultrasound intensity; is the change in treatment time; 、 、 All are adjustable weight coefficients to balance the effects of bone health index improvement, ultrasound intensity changes, and treatment time changes.

[0087] The dynamic mapping relationship between bone health index and ultrasound parameters is described by the following formula:

[0088] ;

[0089] ;

[0090] ;

[0091] Where, f is the actual ultrasonic frequency and the reference frequency; is the default treatment frequency; is the time derivative of the bone health index, reflecting the rate of change of bone metabolic activity; I is the actual ultrasound intensity and the baseline intensity; is the safety threshold; T is the actual treatment time and the benchmark time; The duration of a single treatment; is the frequency adjustment coefficient; Intensity adjustment coefficient; is the time adjustment coefficient.

[0092] Based on the temporal changes in the BHI, a reinforcement learning algorithm is used to output personalized ultrasound intervention parameters. Based on the BHI change trend, ultrasound treatment parameters are dynamically adjusted to achieve personalized intervention.

[0093] The Q-learning algorithm is used, the state space is BHI and its temporal changes, and the action space is ultrasonic frequency f, intensity I, and action time T.

[0094] pass ; ; Realize dynamic mapping of BHI to treatment parameters.

[0095] As a result, the response time for adjusting ultrasound treatment parameters was less than 5 seconds, the BHI improvement rate increased by 30%, and the patient's pain score decreased by 40%.

[0096] Through these steps, this technology achieves a complete closed loop from quantifying bone metabolic activity to personalized ultrasound intervention. It not only accurately assesses bone health but also provides personalized treatment plans based on the patient's real-time condition, significantly improving treatment outcomes and quality of life.

[0097] Example 1

[0098] (1) Patient information

[0099] Age: 50; Gender: Male; Symptoms: Lumbar pain, suspected osteoporosis

[0100] Medical history: No history of major trauma, but the patient has been engaged in physical labor for a long time, and the pain has worsened in the past six months.

[0101] (2) Data collection

[0102] A three-dimensional orthogonal magnetic field sensor array with a sampling frequency of 1kHz and a spatial resolution of 0.1mm³.

[0103] The sensor array was attached to the surface of the patient's lumbar spine to continuously collect magnetic field signals of bone metabolism for 10 minutes. The acquired magnetic field signal waveform showed a peak intensity of 0.5 nT and a spectral distribution concentrated between 10 and 100 Hz. The four-electrode electrical impedance spectroscopy acquisition system had a frequency range of 10 Hz to 1 MHz and an excitation current of 10 mA.

[0104] Electrodes were placed symmetrically on both sides of the lumbar spine, and electrical impedance spectroscopy data were collected simultaneously, recording an ambient temperature of 25°C and a contact pressure of 0.5 N. The impedance spectrum showed an impedance modulus of 150Ω and a phase angle of -15° at a frequency of 1 kHz.

[0105] (3) Data processing

[0106] Using a wavelet threshold denoising algorithm with a threshold of 3, the signal-to-noise ratio increased by 20dB. The inertial measurement unit (IMU) monitored the patient's breathing and micro-displacement, and the signal was corrected in real time. After correction, the residual error was less than 2%.

[0107] Using a nonlinear least squares method, the bone density was fitted to 0.9 g / cm³, the porosity to 15%, and the water content to 30%. The model fitting error was 1.8%. A magnetic field-current inversion algorithm, combining magnetic field signals with electrical impedance characteristics, calculated a bone metabolic current density of 0.1 μA / cm², which is positively correlated with bone metabolic activity.

[0108] Through the attention network, features such as bone density, porosity, and current density are weightedly fused.

[0109] BHI value: The initial BHI is 65, ranging from 0 to 100, reflecting a lower-than-average bone health status.

[0110] (4) Personalized ultrasound intervention parameter output

[0111] Q-learning algorithm, the state space is BHI value, the action space is ultrasound frequency 1-3MHz, intensity 0.5-2W / cm², and action time 10-30 minutes.

[0112] The optimization goal was to maximize the improvement in BHI while limiting the ultrasound intensity change rate to <0.5 W / cm² / min and the treatment time to <30 minutes.

[0113] (5) Parameter output

[0114] Ultrasonic frequency: 1.5 MHz, intensity: 1.0 W / cm², duration: 20 minutes. Based on the BHI timing, the parameters were updated every 5 minutes and adjusted to frequency: 2.0 MHz, intensity: 1.2 W / cm², duration: 25 minutes.

[0115] The signal-to-noise ratio of the magnetic field signal is improved by 20dB, the Cole-Cole model fitting error is 1.5%, and the bone density measurement accuracy is 0.01g / cm³.

[0116] The correlation between BHI and bone metabolic activity was 0.88, and multimodal fusion accuracy improved by 18%. The response time for ultrasound therapy parameter adjustment was 4 seconds, and the BHI improvement rate was 35%.

[0117] The method for generating a bone health index proposed in this paper demonstrates significant innovation in both multimodal fusion and personalized treatment. Through the application of bimodal data acquisition, multimodal data fusion analysis, a dynamic weight allocation mechanism, and a reinforcement learning algorithm, the present invention can more comprehensively and accurately assess bone health and provide personalized treatment plans. Compared with traditional methods, the present invention has higher accuracy and therapeutic efficacy in generating a bone health index, providing new ideas and methods for the prevention and treatment of orthopedic diseases.

[0118] The above description is based on the ideal embodiment of the present invention. Based on the above description, relevant personnel can make various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the content of the specification and must be determined according to the scope of the claims.

Claims

1. A method for generating a bone health index, characterized in that: The following steps are involved: The magnetic field sensor array collects magnetic field signals of bone metabolism, performs filtering and noise removal, and corrects motion artifacts. Then, the electrical impedance spectroscopy data collected by the four-electrode method is fitted with a Cole-Cole model to extract bone density, porosity, and water content. The bone metabolic current density is calculated using a magnetic field-current inversion algorithm. Combining electrical impedance characteristics with current characteristics, a dynamic weight distribution mechanism is used to generate a bone health index. The magnetic field-current inversion algorithm converts the data using the following formula: ； Where, The current density generated by bone cell metabolism; The Hamiltonian operator is used in electromagnetism to describe the spatial variation characteristics of vector fields; is the vacuum permeability; It is the real-time magnetic field strength, collected by the magnetic field sensor array; The electrical conductivity of bone tissue is calibrated in real time through electrical impedance data; Among them, the conductivity is corrected in real time through the electrical impedance data, and the correction formula is: ； Where, is bone density; is bone porosity; is the real-time moisture content; It is the baseline value of water content in healthy bone tissue; is the pathological threshold.

2. The method for generating a bone health index according to claim 1, wherein: The magnetic field sensor array adopts a three-dimensional orthogonal layout, and the sampling frequency 1kHz, spatial resolution 0.1mm3, the acquisition frequency range of electrical impedance spectrum is 10Hz~1MHz.

3. The method for generating a bone health index according to claim 1, wherein: The Cole-Cole model fitting adopts nonlinear least squares method, and the optimization objective function is: ； Where, are the parameters to be optimized for the Cole-Cole model; is the measured impedance; N is the total number of frequency points; i is the time sequence frequency point; is the angular frequency of the i-th frequency point.

4. The method for generating a bone health index according to claim 1, wherein: The dynamic weight allocation mechanism is implemented through the attention network, and the weight update formula is: ； Where, is the updated i-th modal weight, which is dynamically adjusted through the attention mechanism; is the state-action value function; is the feature dimension; The Softmax function ensures weight normalization.

5. The method for generating a bone health index according to claim 1, wherein: The dynamic mapping relationship between the bone health index and ultrasound parameters is described by the following formula: ； ； ； Where, f is the actual ultrasonic frequency and the reference frequency; is the default treatment frequency; It is the time derivative of the bone health index, reflecting the rate of change of bone metabolic activity; I is the actual ultrasonic intensity and the reference intensity; is the safety threshold; T is the actual treatment time and the benchmark time; The duration of a single treatment; is the frequency adjustment coefficient; Intensity adjustment coefficient; is the time adjustment coefficient.

Images