Intelligent tendon-bone healing evaluation system based on artificial intelligence

Abstract

The invention relates to the technical field of noninvasive evaluation of tendon-bone healing states, and discloses an intelligent evaluation system for tendon-bone healing based on artificial intelligence, which comprises the following steps: passively sensing myogenic vibration before detection under standard static pre-pressure, adaptively determining a trigger threshold value by using a cardiogenic pulsation signal, and determining the state of tendon-bone healing according to the trigger threshold value; a non-imaging acoustic detection signal is emitted only when the mechanical environment meets the threshold value, then an acoustic impedance spectrum representing tissue biomechanical attributes is calculated by analyzing echoes, and a quantitative index is generated. Standardization of a measurement environment is internalized into a system closed-loop quality control function from external requirements, physiological noise interference can be penetrated, and an evaluation index capable of stably reproducing and directly reflecting the intrinsic mechanical function of the deep healing tissue is obtained.

Description

Technical Field

[0001] This invention relates to an intelligent assessment system for tendon and bone healing based on artificial intelligence, belonging to the field of non-invasive assessment technology for tendon and bone healing status. Background Technology

[0002] Currently, as clinical practice places higher demands on personalized and accelerated rehabilitation, especially in scenarios where the core goal is the recovery of mechanical function, such as tendon and bone healing, the focus of assessment has shifted from the integrity of tissue morphology to the adequacy of its biomechanical function. Under this requirement, a limitation of existing imaging methods stemming from their measurement principles becomes apparent: a tool designed to observe the static morphology of tissues cannot directly quantify the functional load-bearing capacity of tissues as mechanical components by acquiring physical information. Simply increasing image resolution cannot bridge the gap between morphological information and mechanical function.

[0003] To address this, the industry has adopted techniques such as ultrasonic elastography to measure tissue stiffness, which to some extent reflects the physical properties of the tissue. However, it still falls within the scope of mapping a static physical parameter of the tissue. In monitoring the healing process, which requires multiple longitudinal comparisons, this method also has inherent limitations. For example, in the Chinese invention patent with authorization announcement number CN119235288A, the industry has attempted to apply artificial intelligence to improve the automation of assessment. For instance, it uses convolutional neural networks to automatically segment and analyze MRI images to assess the maturity of ligament grafts and changes in bone tunnels. However, the core of such methods is still the quantitative analysis of static morphological or image signal characteristics of tissues. Although this improves the efficiency and objectivity of morphological assessment, it does not solve the problem of the difference between morphological information and biomechanical function, nor does it handle the impact of physiological interferences such as instantaneous muscle stress on the assessment results during the measurement process.

[0004] Specifically, existing technologies have the following shortcomings: 1. Existing technologies acquire the static physical properties or anatomical morphology of tissues, lacking a technical method that can directly generate comprehensive functional indicators to quantitatively characterize the dynamic mechanical response of tissues; 2. The apparent mechanical properties of tissues are easily affected by the instantaneous stress of the superficial tissues above and the surrounding muscles. Existing technologies lack a mechanism to actively identify and correct such physiological interferences during the measurement process, thus affecting the consistency and comparability of measurement results at different time points. Therefore, how to utilize widely available diagnostic equipment to establish a method that can actively detect and acquire the biomechanical functional response information of healing tissues, and eliminate the influence of physiological interference factors on the results during the measurement process, in order to obtain an objective and longitudinally comparable quantitative evaluation index, is the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides an intelligent assessment system for tendon and bone healing based on artificial intelligence. Its main purpose is to solve the problems that existing technologies cannot directly obtain tissue biomechanical functional information and that measurements are easily affected by physiological interference, thus affecting the consistency of longitudinal comparisons.

[0006] To achieve the above objectives, the present invention provides an intelligent assessment system for tendon and bone healing based on artificial intelligence, the system comprising: A calibration module is configured to control the ultrasound probe to apply and maintain static pre-pressure on the target healing tissue, wherein the static pre-pressure is limited to a preset pressure value range; A state sensing module is configured to, during static pre-stress maintenance, first control the ultrasound probe to operate in passive receiving mode to sense mechanical vibrations generated by muscle activity around the target healing tissue before emitting any active acoustic signals, and extract a periodic signal component synchronized with the patient's cardiac pulsation from the sensed mechanical vibrations. A threshold determination unit is configured to determine a trigger threshold based on an amplitude feature of a periodic signal component, wherein the trigger threshold is proportional to the amplitude feature of the periodic signal component. A detection module whose operation is constrained by gating logic is configured to trigger the detection module to drive the ultrasound probe to emit broadband acoustic detection signals to the tissue and receive its echo signals only when the mechanical vibration sensed by the state sensing module is below the trigger threshold. An analysis module is configured not to generate anatomical images, but to calculate the acoustic impedance spectrum characterizing the biomechanical properties of tissues by jointly analyzing the probe and echo signals. An evaluation module is configured to generate a quantitative index for assessing the state of tissue healing by performing a time-series analysis on one or more spectral features of an acoustic impedance spectrum characterizing the biomechanical properties of tissue obtained from one or more measurements.

[0007] Preferably, the analysis module is configured to obtain the acoustic impedance spectrum by calculating the transfer function between the probe signal and the echo signal, and the spectral features include at least one of the resonant frequency and the peak width.

[0008] Preferably, the analysis module is further configured to: determine the geometric parameters of the superficial tissue above the target healing tissue by identifying characteristic reflection peaks from different tissue interfaces and calculating the time delay between the characteristic reflection peaks based on the time domain characteristics of the echo signal; and to build a superficial tissue acoustic transfer function model, which is configured to receive geometric parameters as input and output a correction factor characterizing the acoustic response of the superficial tissue, and then use the correction factor to perform correction operations on the acoustic impedance spectrum.

[0009] Preferably, the state sensing module is configured to analyze the energy spectrum of the sensed mechanical vibrations with frequencies in the range of 5 Hz to 100 Hz to identify muscle activity.

[0010] Preferably, the threshold determination unit is configured to determine the trigger threshold according to the following rules. : ,in, The average amplitude of the periodic signal component is denoted as , while It is a preset multiplier factor, with a value ranging from 3 to 10.

[0011] Preferably, the calibration module further includes a feedback unit configured to provide the operator with real-time feedback on whether the static prepressure has been reached and maintained within a preset pressure range, based on data from the pressure sensor.

[0012] Preferably, the evaluation module is configured to generate a quantification index by analyzing at least one of the slope of the time series curve composed of spectral features and plateau features.

[0013] Preferably, the broadband acoustic detection signal is a broadband linear frequency modulated acoustic pulse that is not intended for imaging.

[0014] Preferably, the analysis module is configured to perform mathematical deconvolution operations to perform correction operations on the acoustic impedance spectrum.

[0015] Preferably, the state sensing module is further configured to: compare the signal-to-noise ratio of the periodic signal components with a preset coupling quality threshold to generate an indication signal characterizing the acoustic coupling quality between the ultrasound probe and the skin.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. By applying a standardized static pre-pressure to the target tissue using a standard ultrasound probe, and then driving the probe to emit a series of specific broadband acoustic detection signals, a stable acoustic-mechanical coupling measurement environment is established. In this environment, the analysis module does not construct an anatomical image of the tissue, but calculates the acoustic impedance spectrum of the entire coupling system by analyzing the transfer function between the detection signal and the echo signal. The acoustic impedance spectrum obtained in this process directly reflects the macroscopic biomechanical properties of the healed tissue in the calibrated area under the predetermined pressure, providing a physical quantity about the functional state of the tissue that is different from morphological observation for diagnosis.

[0017] 2. While calculating the acoustic impedance spectrum, the analysis module also uses the time-domain characteristics of the same echo signal to determine the geometric parameters of the superficial tissue above the target tissue. These geometric parameters are used as input to a physical model that describes the known influence of different superficial tissues on the acoustic impedance spectrum. Finally, the analysis module uses this model to correct the initially calculated composite impedance spectrum. This mechanism of using the same signal source to analyze different information dimensions and achieve self-consistent correction ensures that the final output impedance spectrum eliminates interference caused by individual differences in superficial tissues or postoperative edema, reflecting the more intrinsic mechanical properties of the target healing tissue.

[0018] 3. To address the disturbances to the instantaneous stiffness of healing tissue caused by involuntary muscle activity of patients in real-world measurement environments, the technical solution of this application adds a state sensing step before the detection module transmits the detection signal. This step controls the ultrasound probe to operate in passive receiving mode to sense the mechanical vibrations generated by muscle activity around the target tissue. Only when the mechanical vibration is below a preset threshold is the detection module triggered to perform subsequent detection actions. This non-silent, non-measuring gating logic transforms the conformity of the measurement environment from an external requirement for the operator into an internal, closed-loop quality control function of the system itself, ensuring the consistency of the initial mechanical environment for each effective measurement.

[0019] 4. Based on the state sensing step, the technical solution of this application is further configured to extract a periodic signal component that is synchronized with the patient's cardiac pulsation from the sensed background mechanical vibration. The system uses the characteristics of this periodic signal component, such as its amplitude, to dynamically adjust the aforementioned preset threshold. This means that the standard used to judge whether muscle activity is excessive is no longer a fixed absolute value, but a relative value associated with the background noise of the current individual's own physiological baseline. This mechanism enables the entire system to adapt to patients with different physiological characteristics and perform measurements in a signal-to-noise ratio environment that matches the individual's physiological baseline, thereby improving the comparability of measurement results between different individuals. Attached Figure Description

[0020] Figure 1 This is a logic flowchart of the intelligent gating evaluation method of the present invention; Figure 2 This is a diagram of the hardware core software and cloud service architecture of the system of this invention; Figure 3 This is a timing diagram for the dynamic judgment of the gate logic triggering conditions of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. It should be noted that the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] This invention provides an artificial intelligence-based intelligent assessment system for tendon-bone healing. Its overall architecture mainly includes a calibration module, a state sensing module, a threshold determination unit, a detection module constrained by gating logic, an analysis module, and an assessment module. These modules work collaboratively to provide a technical method that can eliminate physiological noise interference and obtain an assessment index reflecting the intrinsic biomechanical function of deep healing tissue. In a longitudinal monitoring application scenario targeting the tendon-bone healing interface in patients after anterior cruciate ligament reconstruction, the specific operation flow of this system is configured as a series of interrelated steps; to enable… The operation of each module of the system has a defined engineering benchmark. The calibration procedure for its key operating parameters is configured as follows: First, a static pre-pressure of 0 to 20 Newtons in increments of 0.5 Newtons is applied to a set of tissue simulation phantoms with known Young's moduli, while simultaneously recording the acoustic coupling signal-to-noise ratio and the resonance peak frequency of the acoustic impedance spectrum. By analyzing the two pressure response curves, the pressure point where the signal-to-noise ratio reaches 90% of its saturation plateau region value is defined as the lower limit of the preset pressure range, and the pressure point where the resonance peak frequency value begins to show a nonlinear shift of more than 5% is defined as the upper limit. Second, the trigger threshold in the gating logic is... Its calculation formula is ,in, This is the final trigger threshold. The average amplitude of the cardiac impulse signal component extracted by the state sensing module, and the multiplier factor. The value range is determined by applying a weak mechanical vibration of 30 Hz frequency and 1 micrometer amplitude to the phantom, simulating muscle fiber tremor, through... The system performs step scans from 1 to 15, and records the detection rate of the vibration and the false alarm rate of the background noise when there is no vibration. After plotting the receiver operation characteristic curve, the numerical range of 3 to 10 corresponding to the inflection point of the curve is selected as its working range.

[0023] At the beginning of the diagnostic process, to ensure a consistent initial mechanical environment for each measurement and reduce measurement variability introduced by differences in operator technique, the system first activates the calibration module. This calibration module has a built-in feedback unit that monitors the static pre-pressure applied between the probe and the patient's skin in real time using a pressure sensor integrated with a standard ultrasound probe. It compares this real-time pressure value with a preset pressure range, such as 5N to 10N. Correspondingly, a graphical user interface provides visual feedback to the operator, such as a dynamically changing pressure bar. When the actual pressure enters and stabilizes within the preset range for more than a predetermined time, such as 500ms, the interface provides a clear calibration completion indication. At this point, the calibration module controls the ultrasound probe to... The static pre-pressure is applied to and maintained on the target healing tissue to establish a standardized mechanical reference surface for subsequent acoustic detection. After the static pre-pressure is stably maintained, the system does not immediately emit an active detection signal to address the interference of transient stress generated by involuntary muscle activity around the healing tissue on the tissue's apparent mechanical properties. Instead, the state sensing module first executes an acoustic silent sensing procedure. Under this procedure, the ultrasound probe is controlled to operate in a highly sensitive passive receiving mode to sense and acquire mechanical vibration signals with frequencies within a specific range transmitted through the tissue medium. The state sensing module is configured to focus on analyzing the energy spectrum of the sensed mechanical vibrations with frequencies in the range of 5Hz to 100Hz to identify muscle activity.

[0024] To adapt the muscle activity recognition criteria to the physiological differences of different individuals, the system, while performing acoustic silence sensing, also utilizes a threshold determination unit to extract a periodic signal component synchronized with the patient's cardiac impulses from the sensed broadband mechanical vibrations through a bandpass filter with a center frequency of 1.2 Hz and a bandwidth of 1 Hz. The signal-to-noise ratio of this periodic signal component can be compared with a preset coupling quality threshold to generate an indicator signal characterizing the acoustic coupling quality between the ultrasound probe and the skin, thereby providing the operator with immediate feedback on the probe placement quality. The threshold determination unit is configured to base its judgment on an amplitude characteristic of this periodic signal component, such as its average amplitude over the past 2 seconds. This allows for the dynamic determination of a trigger threshold for gating measurements. The trigger threshold The determination of follows a quantitative procedure, namely: In the formula, The average amplitude of the periodic signal component is expressed in millivolts (mV). This is a preset multiplier factor, ranging from 3 to 10. This range was determined by an offline experiment that balanced the detection rate of effective signals with the suppression rate of noise signals at different muscle tone levels, thus establishing this working window. For example, if the system measures the average amplitude of a patient's cardiac impulses... It is 10mV, and the system is preset. A value of 5 indicates the trigger threshold used for this measurement. That is, it is determined to be 50mV; the operation of the detection module is constrained by the aforementioned gating logic. Specifically, it will only operate when the total vibration energy sensed by the state sensing module in the frequency range of 5Hz to 100Hz is lower than the trigger threshold determined by the threshold determination unit. Only when the vibration energy exceeds a certain threshold is the detection module triggered, driving the ultrasound probe to emit a specific broadband acoustic detection signal towards the tissue and receive its echo signal. If the vibration energy exceeds this threshold, the measurement action is suppressed until the mechanical environment meets the quiescent condition again. This non-quiescent gating logic makes the compliance of the measurement environment an internal quality control step of the system. The broadband acoustic detection signal is a broadband linear frequency modulated sound pulse for non-imaging purposes. Its energy is lower than that required for conventional B-mode ultrasound imaging, but its frequency coverage is wide, for example, from 2MHz linear scan to 8MHz.

[0025] After acquiring the echo signal, the analysis module performs joint analysis of the probe signal and the echo signal to calculate the acoustic impedance spectrum, which characterizes the biomechanical properties of the tissue. The analysis module is configured to obtain this acoustic impedance spectrum by calculating the transfer function between the probe signal and the echo signal. The morphology of this spectrum, such as the resonant frequency and peak width, reflects the stiffness and viscoelasticity of the healing tissue below the probe under standard pre-stress. To reduce interference from shallow tissue in the initially obtained composite spectrum, the analysis module also performs a self-consistent correction process. This process utilizes the time-domain characteristics of the same echo signal, identifies characteristic reflection peaks from different tissue interfaces, and calculates the time delay between these characteristic reflection peaks to determine the geometric parameters of the shallow tissue above the target healing tissue, such as the thickness of each layer. The system has a built-in shallow tissue acoustic transfer function model, which receives the real-time calculated geometric parameters as input and outputs a... A correction factor characterizes the current acoustic response of superficial tissue. Finally, the analysis module performs a mathematical deconvolution operation, using this correction factor to correct the initially calculated composite acoustic impedance spectrum, thus obtaining a purified impedance spectrum reflecting the intrinsic mechanical properties of the target healing tissue. The evaluation module performs time-series analysis on the corrected acoustic impedance spectra obtained from one or more measurements to generate a quantitative index assessing the tissue healing status. The evaluation module extracts one or more key spectral features from the purified impedance spectrum obtained from each measurement, such as resonant frequencies, and constructs a mechanical performance evolution curve according to the measurement time sequence. Then, by analyzing at least one of the slope and plateau characteristics of this time-series curve, the final quantitative index is generated. This index, as a neutral technical indicator, provides objective evidence for clinical decision-making regarding the degree of tissue functional recovery, different from traditional morphological assessments.

[0026] Example 1: In an application for longitudinal monitoring of an athlete's rehabilitation progress after anterior cruciate ligament reconstruction, when the athlete reached the twelfth week post-surgery, their rehabilitation plan faced a decision point: whether to increase training intensity to include high-impact movements. At this time, conventional MRI showed that the graft was morphologically intact and well-positioned, but this morphological information could not quantify the biomechanical strength of its interface with the bone tunnel healing, thus failing to provide direct functional evidence for the next stage of rehabilitation load. To obtain this functional evidence, the system described in the aforementioned specific implementation method was used to monitor the athlete's... The healing site was assessed. During the measurement, due to the athlete's anxiety, the muscle tissue around the healing site exhibited involuntary, persistent tension and intermittent tremors. After the operator applied and maintained a static pre-pressure ranging from 5N to 10N, the system did not immediately transmit a detection signal. Instead, the state sensing module first entered passive receiving mode. This module detected mechanical vibration energy in the 5Hz to 100Hz frequency band. Simultaneously, the threshold determination unit extracted the periodic signal component synchronized with the athlete's heart rate from the same passive sensing signal and determined its average amplitude. Calculate a dynamic trigger threshold specific to the athlete's current physiological state. The system repeatedly compares the sensed muscle activity vibration energy with the dynamic trigger threshold. The comparison was performed, and all results were determined to exceed the threshold. The measurement command of the detection module was continuously withheld until a break after the athlete took a deep breath to relax. At this point, the state sensing module detected a mechanical silence window, that is, the vibration energy of the muscle activity was lower than the trigger threshold marked by its own cardiac pulsation signal. Only then was the gating logic opened.

[0027] At the instant the gating logic is opened, the detection module is triggered, emitting a broadband linear frequency modulated acoustic pulse towards the healing tissue under standardized static pre-stress and in a mechanically silent environment. This coordinated series of operations allows the active acoustic detection step to be performed within a measurement window that meets the preset mechanical silence conditions, ensuring that the mechanical information carried by the echo signal received by the analysis module is not contaminated by instantaneous myogenic pre-stress. The analysis module then calculates the initial composite acoustic impedance spectrum by analyzing the transfer functions of the detection signal and the echo signal. Simultaneously, the analysis module utilizes the time-domain characteristics of the same echo signal to identify the interface reflection between the subcutaneous edema layer and the healing tissue, and uses this to calculate the geometric parameters of the superficial tissue. These parameters are input into a superficial tissue acoustic transfer function model to generate a correction factor, and then deconvolve the model using mathematical methods. The acoustic effects caused by postoperative edema were removed from the composite acoustic impedance spectrum, ultimately outputting an acoustic impedance spectrum reflecting the intrinsic biomechanical properties of the athlete's tendon-bone healing interface. The evaluation module extracted the resonant frequency of this purified acoustic impedance spectrum as a key spectral feature and integrated it with the same spectral feature obtained from measurements taken by the athlete at weeks 6, 8, and 10 postoperatively, constructing a time-series curve characterizing the evolution of the mechanical properties of the healing tissue. This curve shows that the resonant frequency increases linearly from week 6 to week 10, while the rate of increase begins to slow down from week 10 to week 12, showing a trend of entering a plateau phase. This quantitative index generated by the evaluation module and its temporal variation characteristics provide an objective basis for the rehabilitation team to determine that the athlete's tendon-bone healing has entered the biomechanical maturity stage and to decide whether the athlete can enter the next stage of high-intensity training.

[0028] Example 2: To verify the correlation between the quantitative index generated by the system of the present invention and the biomechanical properties of the healing tissue, and to verify the role of the state sensing module in ensuring the consistency of measurement results under external mechanical disturbance, this example designed and executed an in vitro animal model experiment. The experiment used 32 New Zealand white rabbits to establish an animal model of bilateral Achilles tendon repair, and randomly divided them into groups of 8 rabbits each for the second, fourth, and eighth weeks post-surgery, and a healthy control group. The experimental platform consisted of the system of the present invention, a mechanical exciter capable of applying controllable low-frequency vibration, and a material testing machine for final destructive mechanical testing. The mechanical exciter is used to simulate myogenic vibrations around the healing site during measurement. The experimental procedure is as follows: at each predetermined time point, measurements are taken on anesthetized animals. First, the experimental group uses the state sensing module and gating logic enabled, while the control group uses the state sensing module and gating logic disabled. Ten repeated measurements are performed on the same measurement point under a 20Hz external mechanical disturbance. Specifically, the ultrasound probe is placed in the Achilles tendon healing area, and a static pre-pressure of 7N is applied with the assistance of the calibration module, followed by the initiation of a 20Hz mechanical disturbance. During the experimental group measurement, the system only dynamically triggers the threshold. During the window period when the mechanical environment is deemed acceptable, the system performs detection and records the acoustic impedance spectrum. During the control group measurement, the system continuously performs 10 detections and records the acoustic impedance spectrum under a disturbed environment. After completing the non-invasive measurement, an Achilles tendon sample is obtained and subjected to uniaxial tensile testing using a materials testing machine until fracture. Its ultimate tensile strength is recorded as a reference standard for biomechanical properties.

[0029] The experimental data processing and analysis process was as follows: For each animal, the average and standard deviation of the resonant frequencies, which are key spectral features, were calculated in the 10 acoustic impedance spectra measured in the experimental and control groups. Data from the fourth week after surgery were selected as representative for illustration (see Table 1). In the control group, due to continuous mechanical disturbance, the resonant frequency values ​​obtained from 10 repeated measurements fluctuated significantly, with a standard deviation of 1.12 MHz. In the experimental group, however, because the gating logic avoided most of the disturbance window, the resonant frequency values ​​obtained from 10 repeated measurements showed higher consistency, with a standard deviation of 0.21 MHz. This comparison shows that the introduction of the state sensing module reduced the impact of external mechanical disturbance on the consistency of measurement results. The results are shown in Table 1.

[0030] Table 1: Comparison of measurement results between the experimental group and the control group at the fourth week after surgery. Measurement grouping Number of repeated measurements Average frequency of resonance peaks (MHz) Standard deviation of resonant frequency (MHz) control group 10 4.15 1.12 experimental group 10 4.21 0.21

[0031] Furthermore, the average resonant frequency measured at all time points using the experimental group's working method was correlated with the ultimate tensile strength obtained from subsequent destructive testing of the sample. The results showed that from the healthy control group to the eighth week after surgery, there was a positive correlation between the average resonant frequency and the ultimate tensile strength, with a Pearson correlation coefficient of 0.96. These experimental results indicate that the quantitative index represented by the resonant frequency generated by the system and method of this invention can reflect the recovery process of tissue biomechanical function during tendon-bone healing.

[0032] To further verify the core technical solution of the present invention, which is based on adaptive determination of the trigger threshold using cardiac pulsation signals, and to compare with the conventional technical means that those skilled in the art may use when facing the same problem (i.e., using a fixed trigger threshold based on experience), a comparative example 1 is provided as follows.

[0033] Comparative Example 1: This comparative example is intended to be directly compared with the experimental group in this embodiment, using the adaptive threshold of the present invention, in order to verify the progressiveness of the core technical points of the present invention. The experimental subjects, equipment, environmental conditions, and all operating steps except for the gate logic judgment basis of this comparative example are exactly the same as those of the experimental group in Example 2. The only difference is that this comparative example does not perform the step of adaptively determining the trigger threshold based on the cardiac pulsation signal, but adopts a conventional technical path that is usually used by those skilled in the art when designing similar gate logic: that is, a fixed trigger threshold is preset based on large-scale sample statistics or empirical values.

[0034] In this comparative example, the fixed trigger threshold was set to 50mV, a static threshold with universal applicability over a large sample range. The experimental procedure was the same: 10 repeated measurements were performed on the same measurement point of 8 animals at the fourth week after surgery, under the condition of applying a 20Hz external mechanical disturbance. During the measurement, the gating logic only triggered the detection module when the total vibration energy sensed by the state sensing module was lower than the fixed threshold of 50mV. During the experiment, it was observed that for 2 animals with high muscle tone, the background vibration energy under the disturbance environment fluctuated continuously in the range of 55mV to 70mV. The system could not find a measurement window below the fixed threshold of 50mV for a monitoring period of up to 3 minutes, resulting in the measurement not being completed automatically. For the other 6 animals, although the system could complete the measurement, the consistency of the measurement results is shown in Table 2 below.

[0035] Table 2: Comparison of measurement results under different gating methods at the fourth week post-operation. Measurement grouping Number of repeated measurements Average frequency of resonance peaks (MHz) Standard deviation of resonant frequency (MHz) Control group (no gating) 10 4.15 1.12 Comparative Example 1 (Fixed Threshold Gating) 10 4.18 0.67 Experimental group (adaptive threshold gating of this invention) 10 4.21 0.21

[0036] The results of Comparative Example 1 show that, although using a fixed, experience-based trigger threshold for gating measurements can improve the repeatability of the measurement results to some extent (the standard deviation decreased from 1.12MHz to 0.67MHz) compared to the control group without any gating logic, its effect is inferior to the experimental group using the adaptive threshold gating of this invention (the standard deviation is 0.21MHz). In addition, this fixed threshold method also exposes its inherent defect of not being able to adapt to individual physiological differences. There are practical engineering problems such as measurement failure or the need to significantly extend the measurement time on some test subjects with high muscle tone. This comparative experiment confirms that the key to the invention is to dynamically adjust the trigger threshold based on the test subject's own cardiac pulsation signal, rather than any form of simple gating, which is the key to achieving stable reproduction of measurement results and effectively penetrating physiological noise interference.

[0037] Example 3: This example combines Figures 1 to 3 This describes an intelligent assessment system for tendon and bone healing based on artificial intelligence, such as... Figure 1 As shown, the process begins with the ultrasound probe acting on the target healing tissue. First, a calibration module applies and maintains a standardized static pre-pressure to establish a standard mechanical environment. Then, the state sensing module passively senses the myogenic vibration signal generated by myogenic vibration without emitting an active signal, and extracts the cardiac pulsation signal from it. This cardiac pulsation signal is transmitted to a threshold determination unit, which adaptively determines a trigger threshold based on the cardiac pulsation signal and transmits this adaptive trigger threshold to the gating logic judgment stage. Simultaneously, the aforementioned myogenic vibration signal is also input into the gating logic. In the gating logic judgment stage, this stage determines whether the myogenic vibration is below the trigger threshold. If it is above the threshold, the measurement is suppressed and the state sensing module is instructed to continue sensing. If it is below the threshold, the detection module is triggered. The detection module emits a broadband acoustic detection signal and receives its echo. The obtained detection and echo signals are transmitted to the analysis module. The analysis module calculates the acoustic impedance spectrum and corrects for the influence of superficial tissues to generate an acoustic impedance spectrum as time-series data. This spectrum is transmitted to the evaluation module. The evaluation module analyzes the characteristics of the time-series analysis spectrum and finally generates a quantitative evaluation index.

[0038] like Figure 2 As shown, the integrated ultrasound probe is responsible for real-time sensor data transmission, including pressure, vibration, and acoustic echo, and transmits the data to the intelligent assessment host. The host runs core assessment software, which includes calibration and feedback applications, intelligent gating logic, a signal analysis engine, and an assessment model. The intelligent assessment host interacts bidirectionally with a cloud server cluster, uploading anonymous measurement data to the cloud and downloading updates to the assessment model from the cloud. The cloud server cluster includes a model training and data platform, which consists of a historical patient database, model training services, and model distribution services.

[0039] like Figure 3 As shown, the process begins with the ultrasound probe maintaining a static pre-pressure of 7N. At this time, the state sensing module activates the passive receiving mode and continuously senses for 2 seconds, acquiring vibration signals from 5Hz to 100Hz. The heart rate synchronization signal is extracted from this signal and transmitted to the threshold determination unit. This unit processes the signal through a bandpass filter with a center frequency of 1.2Hz, calculates the average amplitude as 10mV, and sets a multiplier factor k=5, thus determining the trigger threshold as 50mV. Subsequently, the gating logic begins to execute the gating judgment loop. When the system senses the current vibration energy as 75mV, because it is greater than the 50mV threshold, the measurement is suppressed and the system waits for silence. When the subsequently sensed vibration energy is 60mV, the detection continues to be suppressed until the current vibration energy is sensed as 35mV. Because it is less than the 50mV threshold, the gating logic is triggered, and the detection module is instructed to emit a frequency-modulated pulse with a linear scanning frequency range of 2MHz to 8MHz. After the ultrasound probe receives the echo signal, the detection is completed, the system is reset, and it waits for the next measurement.

[0040] Example 4: Before the system of this invention is put into application, some core parameters and built-in models need to be calibrated offline and a procedure established to ensure the traceability and consistency of the output quantization index. This procedure is carried out in a standard acoustic laboratory, and its initial state is defined as follows: the object of action is a three-layer structured tissue simulation phantom, which consists of a bottom layer of high acoustic impedance material (simulating bone), a middle layer of polyvinyl alcohol gel (simulating tendon tissue at different healing stages, with Young's moduli of 0.5MPa, 1.0MPa, and 2.0MPa respectively), and a top layer of variable-thickness silicone (simulating shallow tissue with thicknesses of 2mm, 4mm, 6mm, and 8mm respectively). The functional specifications of the enabling environment include the system of this invention, a mechanical vibration table capable of outputting frequencies ranging from 10Hz to 100Hz and amplitudes adjustable from 0.1 micrometers to 10 micrometers, and a standard hydrophone for reference measurement. The procedure first determines the multiple factor in the threshold determination unit. The calibration process is quantified as an optimization process balancing detection sensitivity and specificity. A tissue simulation phantom with a Young's modulus of 1.0 MPa is placed on the test platform. After applying a static preload of 7 N, the system first collects a 1-minute background noise signal under conditions of no external vibration, extracts the periodic signal component synchronized with the cardiac impulse, and calculates its average amplitude. The baseline value was then established; subsequently, the mechanical vibration table was activated, applying a weak vibration with a frequency of 30 Hz and an amplitude of 1 micrometer to simulate early-stage muscle fiber tremors, and the system was allowed to... With the value gradually increasing from 1 to 15, the measurement gating judgment was repeated 100 times, and the rate at which the system could correctly identify vibration and suppress measurement was recorded, i.e., the sensitivity; then the vibration table was turned off, and the same process was repeated... Under conditions ranging from 1 to 15, the system records the rate at which it incorrectly suppresses measurements when there is only background noise; this is the false alarm rate. The sensitivity at a given value is plotted against (1 - false alarm rate) to form a receiver operating characteristic curve. The inflection point of the curve corresponds to... The value range is the working range that balances effective recognition of subtle muscle activity with tolerance to background noise. This calibration determines... The value ranges from 3 to 10.

[0041] Next, the procedure proceeds to the step of establishing the shallow tissue acoustic transfer function model built into the analysis module; this step aims to correlate the shallow tissue geometric parameters determined by the time-domain characteristics of the echo signal with the acoustic response of the shallow tissue to the acoustic impedance spectrum; using the aforementioned calibrated model... The system measures phantoms with shallow tissues of different thicknesses (2mm, 4mm, 6mm, 8mm) using a value (e.g., 5). For each phantom, the analysis module first determines the geometric parameters of the shallow tissue, i.e., its thickness, based on the time-domain characteristics of the echo signal by identifying characteristic reflection peaks from different tissue interfaces and calculating the time delay. Simultaneously, the analysis module calculates the composite acoustic impedance spectrum, which includes the interaction between the shallow tissue and the intermediate gel layer. Since the intrinsic acoustic impedance spectrum of the intermediate gel layer can be obtained by directly measuring a gel phantom without shallow tissue coverage, the acoustic transfer function generated by the shallow tissue of a specific thickness can be derived by performing mathematical deconvolution operations on the composite acoustic impedance spectrum and the intrinsic acoustic impedance spectrum. By fitting the corresponding data of multiple sets of shallow tissue thickness-acoustic transfer functions, a function model is established with the shallow tissue geometric parameters as input and a correction factor as output. By executing the above procedures, the key parameters in the system... The experimental basis was obtained, and the shallow tissue acoustic transfer function model required by the analysis module was also established. Subsequently, when the system performs actual measurements, all its internal judgments and corrections are based on this calibrated and reproducible engineering data, thus ensuring the accuracy and comparability of its final output quantification index.

[0042] Example 5: When the system of the present invention is applied to a test subject with arrhythmia or poor peripheral circulation, its state sensing module may not be able to extract a periodic signal component that is synchronized with the cardiac impulse and whose signal-to-noise ratio and periodicity both meet the preset requirements from the sensed mechanical vibration during the initial passive reception phase. When the system starts the measurement, it will first execute a baseline signal quality assessment procedure for 3 seconds. This procedure requires that the signal-to-noise ratio of the extracted periodic signal component is not less than 3 dB and its main frequency is stable in the range of 0.8 Hz to 2.5 Hz. If the system determines that the quality index of the signal component fails to meet any of the above requirements, the system will not adopt the working mode of dynamically adjusting the trigger threshold based on the signal component.

[0043] Under this operating condition, to maintain the effectiveness of the gating logic, the system is configured to automatically switch to a preset fixed threshold operating mode; in this mode, the threshold determination unit no longer calculates the trigger threshold based on the periodic signal components sensed in real time. Instead, it directly calls a static trigger threshold obtained from large-scale sample statistics in the offline calibration procedure. This static threshold, such as 50 millivolts, is set to be able to trigger suppression under most common myogenic vibrations that can affect the measurement results. The system uses this fixed threshold to complete the subsequent state sensing and gating detection process, thereby ensuring that each acoustic detection is carried out in an environment that meets the basic mechanical silence requirements under boundary conditions where there is a lack of effective intrinsic reference signals.

[0044] Example 6: To ensure the objectivity and traceability of the quantitative index generated by the evaluation module, the system of this invention executes a standardized engineering procedure before deployment, which constructs and periodically updates the evaluation model based on historical data. The initial input of this procedure is an anonymized historical patient dataset containing hundreds of cases that have completed rehabilitation and obtained final clinical results verification. Each data point is a time series containing weekly measurements of formant frequencies as key spectral features. The procedure first quantifies and extracts the slope and plateau features in the time series. For each historical time series data point, the procedure uses a sliding time window with a width of four quadrants. By performing linear regression analysis on the data points within the window, the instantaneous slope at each time point is calculated, and the maximum instantaneous slope occurring throughout the entire healing cycle is defined as the maximum healing rate feature of the sample. The procedure further defines the plateau period as the rate of change of the resonance peak frequency in two or more consecutive measurements being lower than a preset stability threshold. This threshold is set to 5% of the total change in the sample from the initial measurement point to the current measurement point, and the time point at which the plateau period begins is defined as the mechanical maturity initiation point feature. Through this method, each original time series data is processed into a two-dimensional feature vector composed of the maximum healing rate and the mechanical maturity initiation point.

[0045] Furthermore, the procedure constructs a generative model from the two-dimensional feature vector to the final quantitative index through a machine learning process. Using the two-dimensional feature vectors of all historical samples and their corresponding known final clinical rehabilitation results as training data, a logistic regression algorithm is employed to train a classifier model capable of outputting a quantitative index ranging from 0 to 100, representing the maturity of the healing state, based on the input feature vector. Simultaneously, to maintain the long-term effectiveness of this model, an adaptive update mechanism is established: whenever 100 new validated patient data points are added, the system automatically incorporates these new data into the training set and iteratively updates the model parameters. This evaluation model is embedded in the evaluation module and used to process the time-series data of new patients, thereby transforming the evaluation of the healing curve from a qualitative observation method into a data-driven, algorithmic process with quantitative rules.

[0046] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0047] Finally, it should be noted that the above 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent assessment system for tendon and bone healing based on artificial intelligence, characterized in that, The system includes: A calibration module is configured to control the ultrasound probe to apply and maintain static pre-pressure on the target healing tissue, wherein the static pre-pressure is limited to a preset pressure value range; A state sensing module is configured to, during static pre-stress maintenance, first control the ultrasound probe to operate in passive receiving mode to sense mechanical vibrations generated by muscle activity around the target healing tissue before emitting any active acoustic signals, and extract a periodic signal component synchronized with the patient's cardiac pulsation from the sensed mechanical vibrations. A threshold determination unit is configured to determine a trigger threshold based on an amplitude feature of a periodic signal component, wherein the trigger threshold is proportional to the amplitude feature of the periodic signal component. A detection module whose operation is constrained by gating logic is configured to trigger the detection module to drive the ultrasound probe to emit broadband acoustic detection signals to the tissue and receive its echo signals only when the mechanical vibration sensed by the state sensing module is below the trigger threshold. An analysis module is configured not to generate anatomical images, but to calculate the acoustic impedance spectrum characterizing the biomechanical properties of tissues by jointly analyzing the probe and echo signals. An evaluation module is configured to generate a quantitative index for assessing the state of tissue healing by performing a time-series analysis on one or more spectral features of an acoustic impedance spectrum that characterizes the biomechanical properties of tissue, obtained from one or more measurements.

2. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 1, characterized in that, The analysis module is configured to obtain the acoustic impedance spectrum by calculating the transfer function between the probe signal and the echo signal, and the spectral features include at least one of the resonant frequency and the peak width.

3. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 1, characterized in that, The analysis module is also configured to: determine the geometric parameters of the superficial tissue above the target healing tissue by identifying characteristic reflection peaks from different tissue interfaces and calculating the time delay between the characteristic reflection peaks based on the time domain characteristics of the echo signal. It also includes a built-in shallow tissue acoustic transfer function model, which is configured to receive geometric parameters as input and output a correction factor characterizing the acoustic response of shallow tissue.

4. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 1, characterized in that, The state sensing module is configured to analyze the energy spectrum of sensed mechanical vibrations with frequencies in the range of 5 Hz to 100 Hz.

5. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 1, characterized in that, The threshold determination unit is configured to determine the trigger threshold based on the following rules. , ,in, The average amplitude of the periodic signal component is denoted as , while It is a preset multiplier factor, with a value ranging from 3 to 10.

6. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 1, characterized in that, The calibration module also includes a feedback unit configured to provide the operator with real-time feedback on whether the static prepressure has been reached and maintained within a preset pressure range, based on data from the pressure sensor.

7. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 1, characterized in that, The evaluation module is configured to generate a quantification index by analyzing at least one of the slope of the time series curve composed of spectral features and plateau features.

8. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 1, characterized in that, The broadband acoustic detection signal is a broadband linear frequency modulated acoustic pulse that is not intended for imaging.

9. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 3, characterized in that, The analysis module is configured to perform mathematical deconvolution operations.

10. The artificial intelligence-based intelligent assessment system for tendon and bone healing according to claim 1, characterized in that, The state sensing module is also configured to compare the signal-to-noise ratio of the periodic signal components with a preset coupling quality threshold to generate an indication signal characterizing the acoustic coupling quality between the ultrasound probe and the skin.

Images