Multi-modal fusion ACLR post-rehabilitation intelligent evaluation method and system

Abstract

The invention relates to the technical field of medical data analysis. By providing the multi-modal fusion ACLR post-rehabilitation intelligent evaluation method and system, the method comprises the steps that peripheral modal data, central modal data and rehabilitation intervention data of an ACLR patient are synchronously collected and processed, and a multi-source heterogeneous data set is obtained; performing space-time alignment processing on the multi-source heterogeneous data set through a preset unified timestamp to generate a time-synchronized cross-modal data set; the cross-modal data set is processed through the trained brain-muscle-bone multi-modal fusion model, and three-dimensional comprehensive evaluation indexes including the structure safety degree, the function recovery degree and the nerve recombination degree are generated; and performing rehabilitation parameter dynamic optimization processing according to the three-dimensional comprehensive evaluation index, and generating a personalized blood flow limitation training pressure value and proprioceptive training intensity, so as to achieve the technical effects of multi-dimensional quantification of a rehabilitation state, dynamic adaptation of training parameters, nerve-muscle-bone collaborative intervention, improvement of a rehabilitation curative effect and reduction of a re-injury risk.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical data analysis, in particular to an ACLR post-rehabilitation intelligent evaluation method and system based on multi-modal fusion. BACKGROUND

[0002] With the deep cross-fusion of sports medicine and intelligent rehabilitation technology, the scientificity and individuality of the post-rehabilitation scheme after anterior cruciate ligament reconstruction (ACLR) have become the core direction to improve the rehabilitation effect and reduce the risk of re-injury.

[0003] In the traditional technology, the evaluation of the post-rehabilitation state of ACLR patients mainly relies on single modal data, lacks the fusion analysis ability of multi-modal data, and is difficult to quantitatively evaluate the multidimensional characteristics of the rehabilitation state; the setting of rehabilitation parameters depends on empirical decision-making, and cannot realize real-time adaptation driven by data in combination with the physiological dynamics of individual patients, which easily leads to the risk of mismatch between training intensity and physiological bearing capacity; in addition, the traditional technology separates the correlation analysis of central nervous remodeling and peripheral musculoskeletal function recovery, making it difficult to construct a collaborative rehabilitation path, resulting in the limitation of intervention means such as proprioceptive training to a programmed process, which cannot dynamically optimize the training intensity and mode according to individual brain activation maps, neural plasticity entropy and other central indicators, thereby limiting the improvement of neuromuscular control ability. SUMMARY

[0004] Therefore, it is necessary to provide an ACLR post-rehabilitation intelligent evaluation method and system based on multi-modal fusion to achieve multidimensional quantification of rehabilitation state, dynamic adaptation of training parameters and neural-musculoskeletal collaborative intervention, and to improve the technical effects of rehabilitation effect and reduce the risk of re-injury.

[0005] In a first aspect, the present application provides an ACLR post-rehabilitation intelligent evaluation method based on multi-modal fusion, which comprises:

[0006] synchronously collecting and processing peripheral modal data, central modal data and rehabilitation intervention data of ACLR patients to obtain a multi-source heterogeneous data set;

[0007] performing spatio-temporal alignment processing on the multi-source heterogeneous data set through a preset unified timestamp to generate a time-synchronized cross-modal data set;

[0008] processing the cross-modal data set through a trained brain-muscle-bone multi-modal fusion model to generate three-dimensional comprehensive evaluation indexes including structural safety, functional recovery and neural reorganization;

[0009] According to the three-dimensional comprehensive evaluation index, the rehabilitation parameter dynamic optimization processing is performed, and a personalized blood flow restriction training pressure value and proprioceptive training intensity are generated.

[0010] In an embodiment, the cross-modal data set is processed by the trained brain-muscle-bone multi-modal fusion model to generate a three-dimensional comprehensive evaluation index including structural safety, functional recovery and neural reorganization, including:

[0011] The biological structure degradation characteristics in the cross-modal data set are quantitatively fused to generate a structural safety;

[0012] The biomechanical function parameters in the cross-modal data set are dynamically coupled to generate a functional recovery;

[0013] The neural plasticity parameters in the cross-modal data set are modeled using the following formula to generate a neural reorganization degree:

[0014]

[0015] wherein R neuro represents the neural reorganization degree, M represents the total number of modes in the cross-modal data set, K represents the number of neural plasticity parameters in each mode, ω ik represents the weight coefficient of the kth parameter in the ith mode, ΔP ik (t) represents the change amount of the kth neural plasticity parameter in the ith mode at time t, Δt represents the time interval, λ ik represents the attenuation coefficient, d ik represents the compensation path distance;

[0016] The structural safety, functional recovery and neural reorganization are adaptively weighted in the rehabilitation stage to generate a three-dimensional comprehensive evaluation index.

[0017] In an embodiment, the biomechanical function parameters in the cross-modal data set are dynamically coupled to generate a functional recovery, including:

[0018] The muscle force dynamic parameters in the cross-modal data set are processed by power spectrum conversion to generate a muscle force synergistic activation vector;

[0019] The proprioceptive function parameters in the cross-modal data set are processed by four-quadrant stability analysis using the following formula to generate a dynamic balance control coefficient:

[0020]

[0021] wherein K dc represents the dynamic balance control coefficient, ω q represents the weight factor of the qth quadrant, P qrepresents a proprioceptive function parameter value in the qth quadrant, S q represents a stability index of the qth quadrant, P ref represents a reference equilibrium point parameter value;

[0022] The muscle strength synergistic activation vector and the dynamic balance control coefficient are subjected to neuromuscular coupling processing to generate a functional recovery degree.

[0023] In an embodiment, the neuroplasticity parameters in the cross-modal data set are subjected to compensatory path modeling processing to generate a neural reorganization degree, including:

[0024] The task-state brain function data in the cross-modal data set is subjected to sensory-motor cortex mapping processing to generate a brain region compensatory activation atlas;

[0025] The resting-state brain function data in the cross-modal data set is subjected to whole brain network entropy calculation processing to generate a neuroplasticity entropy value;

[0026] The brain region compensatory activation atlas and the neuroplasticity entropy value are subjected to reorganization path mining processing to generate a neural reorganization degree.

[0027] In an embodiment, the biological structure degradation characteristics in the cross-modal data set are subjected to quantitative fusion processing to generate a structure safety degree, including:

[0028] The cartilage degradation characteristics in the cross-modal data set are subjected to T2 relaxation mapping processing to generate a cartilage metabolic activity index;

[0029] The muscle degradation characteristics in the cross-modal data set are subjected to fat infiltration quantification processing to generate a muscle fat index;

[0030] The cartilage metabolic activity index and the muscle fat index are subjected to graft risk fusion processing to generate a structure safety degree.

[0031] In an embodiment, peripheral modal data is acquired, including:

[0032] The knee joint of the ACLR patient is subjected to MRI scanning processing to obtain knee joint structure data;

[0033] The quadriceps femoris of the ACLR patient is subjected to isokinetic muscle strength testing processing to obtain muscle strength function data;

[0034] The balance ability of the ACLR patient is subjected to Pro-kin balance testing processing to obtain proprioceptive data;

[0035] The knee joint structure data, muscle strength function data, and proprioceptive data are subjected to peripheral modal integration processing to generate peripheral modal data.

[0036] In an embodiment, the rehabilitation parameter dynamic optimization processing is performed according to the three-dimensional comprehensive evaluation index, and a personalized blood flow restriction training pressure value is generated, including:

[0037] The structural safety degree is graded and processed to generate a structural safety risk level;

[0038] The functional recovery degree is evaluated and processed to generate a functional compensation state index;

[0039] The neural reorganization degree is quantitatively processed to generate a neural remodeling activation intensity;

[0040] Based on the structural safety risk level, an initial blood flow restriction pressure reference value is matched from a preset rehabilitation safety pressure map;

[0041] Using the following formula, the initial blood flow restriction pressure reference value is corrected in the first stage by combining the functional compensation state index and a preset dynamic pressure adjustment algorithm to generate an adaptive blood flow restriction pressure estimate value:

[0042] P adaptive =P initial ·(1+α·C compensation ·Δt1)

[0043] Wherein, P adaptive represents the adaptive blood flow restriction pressure estimate value, P initial represents the initial blood flow restriction pressure reference value, α represents the dynamic adjustment coefficient, C compensation represents the functional compensation state index, and Δt1 represents the time adjustment factor;

[0044] The neural remodeling activation intensity is fused, and the adaptive blood flow restriction pressure estimate value is corrected in the second stage by a preset neural-muscle coordination optimization algorithm to generate an optimized blood flow restriction pressure value;

[0045] The optimized blood flow restriction pressure value is processed for real-time physiological feedback adaptation to generate a personalized blood flow restriction training pressure value.

[0046] In a second aspect, the present application also provides an ACLR post-rehabilitation intelligent evaluation system based on multi-modal fusion, which comprises:

[0047] A multi-source heterogeneous data acquisition module is used to acquire and process peripheral modal data, central modal data and rehabilitation intervention data of ACLR patients simultaneously to obtain a multi-source heterogeneous data set;

[0048] A cross-modal spatiotemporal alignment module is used to perform spatiotemporal alignment processing on the multi-source heterogeneous data set through a preset unified timestamp to generate a time-synchronized cross-modal data set;

[0049] The multi-modal fusion evaluation module is configured to process the cross-modal data set by using the trained brain-muscle-bone multi-modal fusion model, and generate three-dimensional comprehensive evaluation indexes including structural safety, functional recovery and neural reorganization.

[0050] The rehabilitation parameter dynamic optimization module is configured to perform dynamic optimization processing on rehabilitation parameters according to the three-dimensional comprehensive evaluation indexes, and generate individualized blood flow restriction training pressure values and proprioceptive training intensity.

[0051] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of any method of the first aspect of the present application when executing the computer program.

[0052] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of any method of the first aspect of the present application.

[0053] The multi-modal fusion ACLR post-rehabilitation intelligent evaluation method and system provided by the present application directly improves the multi-dimensional information capturing capability of the rehabilitation state by synchronously collecting peripheral modalities, central modalities and rehabilitation intervention data and processing to generate a multi-source heterogeneous data set; reduces the cross-modal data correlation deviation from the source by performing spatio-temporal alignment on the multi-source heterogeneous data set based on a preset unified timestamp; significantly enhances the quantitative accuracy of the rehabilitation state by using a brain-muscle-bone multi-modal fusion model to analyze the cross-modal data to generate three-dimensional comprehensive evaluation indexes; and provides an adaptable and reliable solution for improving the rehabilitation efficacy after ACLR and reducing the risk of re-injury by generating individualized training pressure values and intensity through dynamic optimization of rehabilitation parameters. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed in the embodiment or related art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0055] Figure 1 The flowchart of the multi-modal fusion ACLR post-rehabilitation intelligent evaluation method in an embodiment of the present application;

[0056] Figure 2 The flowchart of the compensation path modeling processing on the neural plasticity parameters in the cross-modal data set in an embodiment of the present application to generate the neural reorganization degree;

[0057] Figure 3A structural diagram of the multi-modal fusion ACLR post-rehabilitation intelligent evaluation system in an embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below. In the following description, a lot of specific details are set forth in order to fully understand the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the scope of the application, so the present application is not limited to the specific embodiments disclosed below.

[0059] Firstly, the application scenario of the embodiments of the present application is described. In the embodiments of the present application, a multi-modal fusion ACLR post-rehabilitation intelligent evaluation method and system suitable for but not limited to the ACLR post-rehabilitation evaluation and training parameter optimization scenario are provided.

[0060] Illustratively, the multi-modal fusion ACLR post-rehabilitation intelligent evaluation method and system provided by the embodiments of the present application can also be applied to other scenarios involving multi-modal physiological data acquisition, cross-modal information integration and individualized rehabilitation program development, which are only exemplified herein and are not limited to specific application scenarios.

[0061] In the present application, the anterior cruciate ligament is an important stabilizing structure in the knee joint, which plays a key role in maintaining the normal movement of the knee joint and preventing the tibia from moving forward excessively. In sports such as football and basketball, which require frequent sudden stops, turns and jumping actions, the anterior cruciate ligament is prone to injury. ACLR is a common treatment for anterior cruciate ligament injury, aiming to restore the stability of the knee joint by reconstructing the damaged ligament, and help patients return to normal life and sports state.

[0062] Magnetic resonance imaging (MRI) is an imaging technology that uses magnetic fields and radio frequency pulses to generate detailed images of internal structures of the human body, and is widely used in the medical field for non-invasive examination of the structure and function of various parts of the human body.

[0063] Functional magnetic resonance imaging (fMRI) is an imaging technology that uses magnetic fields and radio frequency pulses to reflect the functional activity of brain regions by detecting changes in blood oxygenation levels during brain activity.

[0064] The automated anatomical labeling (AAL) 116 brain atlas is a widely used functional brain region template in the field of neuroimaging.

[0065] As Figure 1 shown, the present application provides a multi-modal fusion ACLR post-rehabilitation intelligent evaluation method, which comprises:

[0066] S101: Synchronously collecting and processing peripheral modality data, central modality data and rehabilitation intervention data of ACLR patients to obtain a multi-source heterogeneous data set.

[0067] Exemplarily, when synchronously collecting peripheral modality data, central modality data and rehabilitation intervention data of ACLR patients, corresponding collection methods are adopted for different modalities of data, such as obtaining structural data through knee MRI scanning, obtaining muscle strength function data by means of isokinetic muscle strength test, obtaining multi-axis proprioceptive function data by using Pro-kin balance test, and integrating the above data to generate a peripheral modality data set.

[0068] Through task-state fMRI (ts-fMRI) and resting-state fMRI (rs-fMRI) examination, multi-sensory motor brain activation and whole brain functional connectivity related data are collected to generate a central modality data set; training parameters, execution conditions and other information in routine rehabilitation and additional blood flow restriction training, quantitative proprioceptive training are recorded to generate a rehabilitation intervention data set. Then, the peripheral, central and rehabilitation intervention data collected above are processed to integrate and generate a multi-source heterogeneous data set.

[0069] S102: Temporally and spatially aligning the multi-source heterogeneous data set by a preset unified timestamp to generate a time-synchronized cross-modality data set.

[0070] Exemplarily, after obtaining the multi-source heterogeneous data set, time markers are added to the peripheral modality data, central modality data and rehabilitation intervention data respectively to ensure that each modality of data is associated with the corresponding collection or recording time point. A unified timestamp reference is set to calibrate the time markers of different modalities of data, eliminating the deviation of each modality of data in time recording. The calibrated multi-source heterogeneous data is matched and integrated according to the unified timestamp, so that different modalities of data are correspondingly associated at the same time node, and then a time-synchronized cross-modality data set is generated.

[0071] S103: Processing the cross-modality data set by a trained brain-muscle-bone multi-modal fusion model to generate three-dimensional comprehensive evaluation indexes including structural safety degree, functional recovery degree and neural reorganization degree.

[0072] Exemplarily, after the time-synchronized cross-modal data set is input into the trained brain-muscle-bone multi-modal fusion model, the structural features related to the knee joint cartilage and muscle in the data are quantitatively integrated, the degradation or recovery state thereof is analyzed, and a structure safety degree is generated. The muscle strength parameters and the proprioceptive function parameters in the data are dynamically coupled and processed, the neuromuscular coordination function is evaluated, and a function recovery degree is generated.

[0073] Meanwhile, the neural plasticity parameters reflecting the brain region activation and whole brain functional connectivity in the data are modeled by a compensation path, the central nervous system reorganization is analyzed, and a neural reorganization degree is generated. The structure safety degree, the function recovery degree and the neural reorganization degree are adaptively weighted and fused in combination with the rehabilitation stage of the patient, and a three-dimensional comprehensive evaluation index including the structure safety degree, the function recovery degree and the neural reorganization degree is generated.

[0074] S104: dynamically optimizing the rehabilitation parameters according to the three-dimensional comprehensive evaluation index, and generating a personalized blood flow restriction training pressure value and a proprioceptive training intensity.

[0075] Exemplarily, based on the structure safety degree in the three-dimensional comprehensive evaluation index, the tolerance state of the graft and the surrounding tissue is evaluated, and the initial pressure benchmark of the blood flow restriction training is determined. In combination with the coordination level of the muscle strength and the proprioceptive function in the function recovery degree, the initial pressure benchmark is dynamically adjusted to match the compensation demand of the neuromuscular function. Referring to the brain region activation and the functional connectivity state reflected by the neural reorganization degree, the pressure parameter is further optimized, and a personalized blood flow restriction training pressure value is generated.

[0076] Meanwhile, according to the proprioceptive function recovery degree in the three-dimensional comprehensive evaluation index and the response characteristics of the neural reorganization to the sensory input, the corresponding training module parameters and the difficulty setting are adjusted, and an adapted proprioceptive training intensity is generated.

[0077] The multi-modal fusion ACLR post-rehabilitation intelligent evaluation method provided by an embodiment of the present application directly improves the multi-dimensional information capture capability of the rehabilitation state by synchronously collecting peripheral modalities, central modalities and rehabilitation intervention data and processing to generate a multi-source heterogeneous data set; reduces the cross-modal data correlation deviation from the source by spatiotemporal alignment of the multi-source heterogeneous data set based on a preset unified timestamp; significantly enhances the quantitative precision of the rehabilitation state by analyzing the cross-modal data using a brain-muscle-bone multi-modal fusion model to generate a three-dimensional comprehensive evaluation index; and generates personalized training pressure values and intensities by dynamically optimizing the rehabilitation parameters, thereby providing an adaptable and reliable solution for improving the rehabilitation efficacy after ACLR and reducing the risk of re-injury.

[0078] In an embodiment, the cross-modal data set is processed by a trained brain-muscle-bone multi-modal fusion model to generate a three-dimensional comprehensive evaluation index including a structure safety degree, a function recovery degree and a neural reorganization degree, which includes:

[0079] Quantitative fusion processing of biological structure degradation characteristics in the cross-modal data set generates structure safety degree;

[0080] Dynamic coupling processing of biomechanical function parameters in the cross-modal data set generates function recovery degree;

[0081] Using the following formula, the compensatory path modeling processing of the neuroplasticity parameters in the cross-modal data set generates the neural reorganization degree:

[0082]

[0083] Wherein, R neuro represents the neural reorganization degree, M represents the total number of modes in the cross-modal data set, K represents the number of neuroplasticity parameters in each mode, ω ik represents the weight coefficient of the kth parameter in the ith mode, ΔP ik (t) represents the change amount of the kth neuroplasticity parameter in the ith mode at time t, Δt represents the time interval, λ ik represents the attenuation coefficient, d ik represents the compensatory path distance;

[0084] The structure safety degree, function recovery degree and neural reorganization degree are adaptively weighted in the rehabilitation stage to generate a three-dimensional comprehensive evaluation index.

[0085] Exemplarily, after inputting the time-synchronized cross-modal data set into the trained brain-muscle-bone multi-modal fusion model, the metabolic activity index of the knee joint cartilage, the fat index of the muscle and other parameters reflecting the structure state are extracted, and the above parameters are quantitatively integrated and analyzed to comprehensively evaluate the structure safety of the graft and the surrounding tissue, and generate the structure safety degree.

[0086] The muscle strength dynamic parameters are extracted and converted into muscle strength synergistic activation vectors, and the proprioceptive function parameters are analyzed for stability to generate a dynamic balance control coefficient. The above two types of parameters are dynamically coupled at the neuromuscular level to evaluate the knee function recovery condition to generate the function recovery degree.

[0087] Then, the neuroplasticity parameters in the cross-modal data set are processed, i.e. the brain region compensatory activation atlas generated by the task-state brain function data and the neuroplasticity entropy value generated by the resting-state brain function data are extracted, and the above parameters are modeled for compensatory path to analyze the reorganization of the central nervous system to generate the neural reorganization degree.

[0088] According to the rehabilitation stage of the patient, the structure safety degree, function recovery degree and neural reorganization degree are given corresponding weights for adaptive weighted fusion processing to generate a three-dimensional comprehensive evaluation index including the structure safety degree, function recovery degree and neural reorganization degree.

[0089] In an embodiment, the biomechanical function parameters in the cross-modal data set are dynamically coupled to generate a function recovery degree, including:

[0090] The muscle force dynamic parameters in the cross-modal data set are processed by power spectrum conversion to generate a muscle force synergistic activation vector;

[0091] The proprioceptive function parameters in the cross-modal data set are processed by four-quadrant stability analysis to generate a dynamic balance control coefficient using the following formula:

[0092]

[0093] wherein K dc represents the dynamic balance control coefficient, ω q represents the weight factor of the qth quadrant, P q represents the proprioceptive function parameter value in the qth quadrant, S q represents the stability index of the qth quadrant, P ref represents the reference balance point parameter value;

[0094] The muscle force synergistic activation vector and the dynamic balance control coefficient are processed by neuromuscular coupling to generate a function recovery degree.

[0095] Exemplarily, the muscle force dynamic parameters obtained from the isokinetic muscle force test are extracted from the cross-modal data, which reflect the force changes of the quadriceps muscle and other muscles during movement. The time-domain dynamic characteristics of muscle force are converted into frequency-domain characteristics by power spectrum conversion processing, and then a muscle force synergistic activation vector reflecting the synergistic activation state of different muscle groups during contraction is generated.

[0096] The multi-axis proprioceptive function parameters obtained from the test are extracted from the cross-modal data, which include the kinesthetic, position sense, and weight sense information of the knee joint. They are included in the preset four-quadrant analysis framework, each quadrant corresponds to a specific stability evaluation dimension, and combined with the weight factor of each quadrant, the proprioceptive function parameter value in the quadrant, the stability index of the quadrant, and the reference balance point parameter value, the dynamic balance control coefficient reflecting the balance control ability of the knee joint during dynamic movement is generated by four-quadrant stability analysis processing.

[0097] The correlation between muscle contraction and proprioceptive feedback under the regulation of neural signals is analyzed based on the muscle force synergistic activation vector and the dynamic balance control coefficient. The muscle synergistic working state and the dynamic balance control ability are integrated to comprehensively evaluate the function recovery status of the knee joint at the biomechanical level to generate a function recovery degree.

[0098] As Figure 2The compensation path modeling process is performed on the neural plasticity parameters in the cross-modal data set to generate a neural reorganization degree, including:

[0099] S201: Task-state brain function data in the cross-modal data set is processed through sensory-motor cortex mapping to generate a brain region compensation activation map;

[0100] S202: Resting-state brain function data in the cross-modal data set is processed through whole brain network entropy calculation to generate a neural plasticity entropy value;

[0101] S203: The brain region compensation activation map and the neural plasticity entropy value are subjected to reorganization path mining processing to generate a neural reorganization degree.

[0102] Exemplarily, task-state brain function data is extracted from cross-modal data, which comes from whole brain imaging sequences collected during joint position sense tests. Through sensory-motor cortex mapping processing, the activated brain regions during task execution are located and marked, especially the multisensory motor brain regions related to motor control and sensory feedback, and then a brain region compensation activation map is generated, which can intuitively present the distribution of brain region compensation activation.

[0103] Resting-state brain function data is extracted from cross-modal data, which comes from whole brain imaging sequences under resting state. With the help of whole brain network entropy calculation processing, the brain regions divided by AAL116 brain atlas are taken as network nodes, the complexity and dynamic changes of functional connections between nodes are analyzed, and a neural plasticity entropy value reflecting the level of neural network plasticity of the whole brain is generated.

[0104] The generated brain region compensation activation map and neural plasticity entropy value are subjected to reorganization path mining processing, the compensation path formed by the central nervous system in the functional remodeling process and its characteristics are identified, the degree and effectiveness of neural functional reorganization are comprehensively evaluated, and a neural reorganization degree is generated.

[0105] In an embodiment, the biological structure degradation characteristics in the cross-modal data set are subjected to quantitative fusion processing to generate a structure safety degree, including:

[0106] The cartilage degradation characteristics in the cross-modal data set are processed through T2 relaxation mapping to generate a cartilage metabolic activity index;

[0107] The muscle degradation characteristics in the cross-modal data set are processed through fat infiltration quantification to generate a muscle fat index;

[0108] The cartilage metabolic activity index and the muscle fat index are subjected to graft risk fusion processing to generate a structure safety degree.

[0109] Exemplarily, cartilage structure data obtained from a knee MRI examination of the cross-modal data is extracted, which includes image features of different regions of cartilage, and the MRI signal features of the cartilage are converted into quantitative indicators reflecting the metabolic state of the cartilage tissue through T2 relaxation mapping processing, to generate a cartilage metabolic activity index, thereby reflecting the degree of cartilage degeneration.

[0110] Muscle structure data obtained from a thigh fat quantitative sequence scan of the cross-modal data is extracted, which contains image information of muscles such as quadriceps femoris, and the infiltration degree of fat components in muscle tissue is analyzed and converted into a quantitative parameter through fat infiltration quantification processing, to generate a muscle fat index, thereby reflecting the degeneration state of the muscle.

[0111] The generated cartilage metabolic activity index and muscle fat index are subjected to graft risk fusion processing: the correlation between the two and the safety of the anterior cruciate ligament graft is combined, the risk that cartilage metabolic abnormalities and muscle fatification may bring to the stability and healing process of the graft is analyzed, and the structural safety state in which the graft is located is comprehensively evaluated by integrating the quantitative results of the two indicators, to generate a structural safety degree.

[0112] In an embodiment, peripheral modal data is obtained, including:

[0113] An MRI scan is performed on the knee joint of the ACLR patient to obtain knee joint structure data;

[0114] An isokinetic muscle strength test is performed on the quadriceps femoris of the ACLR patient to obtain muscle strength function data;

[0115] A Pro-kin balance test is performed on the balance ability of the ACLR patient to obtain proprioceptive data;

[0116] Peripheral modal integration processing is performed on the knee joint structure data, muscle strength function data, and proprioceptive data to generate peripheral modal data.

[0117] Exemplarily, an MRI scan is performed on the knee joint of the ACLR patient: multiple sequences are used for scanning to obtain image data including structures such as knee cartilage and quadriceps femoris, from which information such as cartilage degeneration features and muscle fat infiltration is extracted to obtain knee joint structure data.

[0118] An isokinetic muscle strength test is performed on the quadriceps femoris of the ACLR patient: the patient is fixed on a force chair, the force meter axis is adjusted to be aligned with the center of the knee, the knee joint is flexed and extended according to specific requirements, and parameters reflecting muscle contraction strength and function such as peak torque, peak torque to body weight ratio, and power are recorded during the movement to obtain muscle strength function data.

[0119] Balance ability of the ACLR patient is tested and processed: through dynamic multi-axis somatosensory evaluation, the patient controls the balance board to tilt according to the system trajectory diagram to complete the kinesthetic, weight sense and position sense tests, and the average track error (ATE) and average freeweight variance (AFV) are recorded to obtain somatosensory data.

[0120] Peripheral modalities are integrated for the knee joint structure data, muscle strength function data and somatosensory data: the information reflecting the knee joint structure state, muscle function level and somatosensory function in the three types of data is summarized and associated, data redundancy is eliminated and data format is unified to form a comprehensive data set that can more comprehensively reflect the peripheral physiological function state of the patient, and peripheral modality data is generated.

[0121] In an embodiment, according to the three-dimensional comprehensive evaluation index, the rehabilitation parameter dynamic optimization processing is performed to generate a personalized blood flow restriction training pressure value, including:

[0122] The structural safety degree is graded for the graft microenvironment stability to generate a structural safety risk level.

[0123] Illustratively, according to the cartilage metabolic activity, muscle fat degree and overall structure state of the graft reflected by the structural safety degree, the metabolic balance and mechanical support stability of the tissue around the graft are graded by the graft microenvironment stability grading standard to generate a structural safety risk level to clarify the tolerance boundary of the graft to the blood flow restriction pressure.

[0124] The functional recovery degree is evaluated for the neuromuscular function compensation state to generate a functional compensation state index.

[0125] Illustratively, based on the muscle strength synergistic activation state and dynamic balance control ability embodied by the functional recovery degree, the synergistic efficiency of muscle strength output and somatosensory feedback and the improvement degree of compensation mechanism are analyzed by the neuromuscular function compensation state evaluation system to generate a functional compensation state index to reflect the adaptive potential of the current neuromuscular system to low-load resistance training combined with blood flow restriction.

[0126] The neural reorganization degree is quantified for the central remodeling efficiency to generate a neural remodeling activation intensity.

[0127] Illustratively, combined with the multisensory motor brain area activation map and whole brain functional connectivity features reflected by the neural reorganization degree, the brain area compensation activation intensity and the rate of change of neural plasticity entropy value are evaluated by the central remodeling efficiency quantification method to generate a neural remodeling activation intensity to reflect the efficiency of central control of movement instructions and integration of sensory input.

[0128] Based on the structural safety risk level, an initial blood flow restriction pressure reference value is matched from a preset rehabilitation safety pressure map.

[0129] Illustratively, based on the structural safety risk level, a lower limb blood flow restriction basic pressure parameter corresponding to the level is matched from a preset rehabilitation safety pressure map as an initial blood flow restriction pressure reference value, ensuring that the pressure setting does not exceed the graft safety range.

[0130] Using the following formula, combined with the functional compensation state index, the initial blood flow restriction pressure reference value is corrected in the first stage by a preset dynamic pressure adjustment algorithm to generate an adaptive blood flow restriction pressure estimate value:

[0131] P adaptive = P initial ·(1+α·C compensation ·Δt1)

[0132] Wherein, P adaptive represents the adaptive blood flow restriction pressure estimate value, P initial represents the initial blood flow restriction pressure reference value, α represents the dynamic adjustment coefficient, C compensation represents the functional compensation state index, and Δt1 represents the time adjustment factor.

[0133] Illustratively, according to the muscle strength recovery level and balance control ability reflected by the functional compensation state index, the pressure reference value is adjusted to adapt to the current contraction efficiency and load bearing capacity of the muscle, generating an adaptive blood flow restriction pressure estimate value.

[0134] Fusion of neural remodeling activation intensity, the adaptive blood flow restriction pressure estimate value is corrected in the second stage by a preset neural-muscular coordination optimization algorithm to generate an optimized blood flow restriction pressure value.

[0135] Illustratively, the neural remodeling activation intensity is associated with the neural signal conduction efficiency during muscle movement, and the pressure estimate value is optimized to promote more accurate regulation of the muscle by the nerve, generating an optimized blood flow restriction pressure value.

[0136] The optimized blood flow restriction pressure value is processed for real-time physiological feedback adaptation to generate a personalized blood flow restriction training pressure value.

[0137] Illustratively, combined with the balance state monitored during training, the muscle strength change of isokinetic test and the somatosensory feedback, the pressure value is fine-tuned in real time to ensure that the limb circulation state and muscle contraction response when the thigh proximal cuff applies pressure are adapted, generating a personalized blood flow restriction training pressure value.

[0138] It should be understood that although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise explicitly stated herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps.

[0139] In an embodiment, as shown in FIG. 1, the present application also provides a multimodal fusion ACLR post-rehabilitation intelligent evaluation system 300, which comprises: Figure 3

[0140] A multi-source heterogeneous data acquisition module 301 is configured to synchronously acquire and process peripheral modality data, central modality data and rehabilitation intervention data of an ACLR patient to obtain a multi-source heterogeneous data set.

[0141] A cross-modality spatiotemporal alignment module 302 is configured to perform spatiotemporal alignment processing on the multi-source heterogeneous data set by using a preset unified timestamp to generate a time-synchronized cross-modality data set.

[0142] A multi-modal fusion evaluation module 303 is configured to process the cross-modality data set by using a trained brain-muscle-bone multi-modal fusion model to generate three-dimensional comprehensive evaluation indexes including structural safety, functional recovery and neural reorganization.

[0143] A rehabilitation parameter dynamic optimization module 304 is configured to perform rehabilitation parameter dynamic optimization processing according to the three-dimensional comprehensive evaluation indexes to generate individualized blood flow restriction training pressure values and proprioceptive training intensities.

[0144] Specifically, the multi-source heterogeneous data acquisition module 301 acquires peripheral modality data related to structures such as cartilage and muscle by knee joint MRI scanning, acquires peripheral modality data related to muscle strength function and proprioception by isokinetic muscle strength testing, acquires central modality data such as activation of multisensory motor brain areas and whole brain functional connectivity state by ts-fMRI and rs-fMRI, and records training parameters, execution conditions and other rehabilitation intervention data in conventional rehabilitation and additional training. After preprocessing the above data, the data is integrated into a multi-source heterogeneous data set.

[0145] ​The cross-modal spatio-temporal alignment module 302 processes the multi-source heterogeneous data set, adds corresponding acquisition time markers to the peripheral modal data, central modal data and rehabilitation intervention data respectively, calibrates the time record deviation of different modal data based on a preset unified timestamp reference, matches and associates each modal data after time synchronization according to the time node, and generates a time-synchronized cross-modal data set.

[0146] The multi-modal fusion evaluation module 303 inputs the time-synchronized cross-modal data set into the trained brain-muscle-bone multi-modal fusion model, quantitatively integrates the structural features related to the knee cartilage and muscle in the data, analyzes the degradation or recovery state, generates a structure safety degree, and dynamically couples the muscle strength parameters and the proprioceptive function parameters in the data to evaluate the neuromuscular coordination function and generate a function recovery degree.

[0147] Meanwhile, the neuroplasticity parameters reflecting the brain region activation and whole brain functional connectivity in the data are modeled for compensation path, the central nervous system reorganization is analyzed, and a neural reorganization degree is generated. Combined with the rehabilitation stage of the patient, the structure safety degree, the function recovery degree and the neural reorganization degree are adaptively weighted and fused to generate a three-dimensional comprehensive evaluation index including the structure safety degree, the function recovery degree and the neural reorganization degree.

[0148] The rehabilitation parameter dynamic optimization module 304 evaluates the tolerance state of the graft and the surrounding tissue based on the structure safety degree in the three-dimensional comprehensive evaluation index, determines the initial pressure reference of blood flow restriction training, dynamically adjusts the initial pressure reference according to the coordination level of muscle strength and balance function in the function recovery degree to match the compensation demand of neuromuscular function, and further optimizes the pressure parameter according to the brain region activation and functional connectivity state reflected by the neural reorganization degree to generate a personalized blood flow restriction training pressure value.

[0149] Meanwhile, according to the proprioceptive function recovery degree and the response characteristics of the neural reorganization to sensory input in the three-dimensional comprehensive evaluation index, the corresponding training module parameters and difficulty settings are adjusted to generate an adaptive proprioceptive training intensity.

[0150] The multi-modal fusion evaluation module 303 is also used for:

[0151] quantitative fusion processing of biological structure degradation characteristics in the cross-modal data set to generate a structure safety degree;

[0152] dynamic coupling processing of biomechanical function parameters in the cross-modal data set to generate a function recovery degree;

[0153] The following formula is used to model the neuroplasticity parameters in the cross-modal data set for compensation path processing to generate a neural reorganization degree:

[0154]

[0155] wherein R neuro denotes the neural reorganization degree, M denotes the total number of modalities in the cross-modality dataset, K denotes the number of neural plasticity parameters in each modality, ω ik denotes the weight coefficient of the kth parameter in the ith modality, ΔP ik (t) denotes the change amount of the kth neural plasticity parameter in the ith modality at time t, Δt denotes the time interval, λ ik denotes the attenuation coefficient, d ik denotes the compensatory path distance.

[0156] The structural safety degree, the functional recovery degree, and the neural reorganization degree are adaptively weighted in the rehabilitation stage to generate a three-dimensional comprehensive evaluation index.

[0157] The multi-modal fusion evaluation module 303 is further configured to:

[0158] The muscle force dynamic parameters in the cross-modality dataset are processed through power spectrum conversion to generate a muscle force synergistic activation vector.

[0159] The proprioceptive function parameters in the cross-modality dataset are processed through four-quadrant stability analysis to generate a dynamic balance control coefficient using the following formula:

[0160]

[0161] wherein K dc denotes the dynamic balance control coefficient, ω q denotes the weight factor of the qth quadrant, P q denotes the proprioceptive function parameter value in the qth quadrant, S q denotes the stability index of the qth quadrant, P ref denotes the reference balance point parameter value.

[0162] The muscle force synergistic activation vector and the dynamic balance control coefficient are processed through neuromuscular coupling to generate a functional recovery degree.

[0163] The multi-modal fusion evaluation module 303 is further configured to:

[0164] The task-state brain function data in the cross-modality dataset are processed through sensory-motor cortex mapping to generate a brain region compensatory activation map.

[0165] The resting-state brain function data in the cross-modality dataset are processed through whole-brain network entropy calculation to generate a neural plasticity entropy value.

[0166] The brain region compensatory activation map and the neural plasticity entropy value are processed through reorganization path mining to generate a neural reorganization degree.

[0167] The multi-modal fusion evaluation module 303 is further used for:

[0168] The cartilage degradation features in the cross-modal data set are processed through T2 relaxation mapping to generate a cartilage metabolic activity index;

[0169] The muscle degradation features in the cross-modal data set are processed through fat infiltration quantification to generate a muscle fatification index;

[0170] The cartilage metabolic activity index and the muscle fatification index are subjected to graft risk fusion processing to generate a structural safety degree.

[0171] The multi-source heterogeneous data acquisition module 301 is further used for:

[0172] The knee joint of the ACLR patient is subjected to MRI scanning processing to obtain knee joint structure data;

[0173] The quadriceps femoris of the ACLR patient is subjected to isokinetic muscle strength testing processing to obtain muscle strength function data;

[0174] The balance ability of the ACLR patient is subjected to Pro-kin balance testing processing to obtain proprioceptive data;

[0175] The knee joint structure data, muscle strength function data, and proprioceptive data are subjected to peripheral modal integration processing to generate peripheral modal data.

[0176] The rehabilitation parameter dynamic optimization module 304 is further used for:

[0177] The structural safety degree is subjected to graft microenvironment stability classification processing to generate a structural safety risk level;

[0178] The functional recovery degree is subjected to neuromuscular function compensation state evaluation processing to generate a function compensation state index;

[0179] The neural reorganization degree is subjected to central remodeling efficiency quantification processing to generate a neural remodeling activation intensity;

[0180] Based on the structural safety risk level, an initial blood flow restriction pressure reference value is matched from a preset rehabilitation safety pressure map;

[0181] Using the following formula, the initial blood flow restriction pressure reference value is subjected to first-stage correction through a preset dynamic pressure adjustment algorithm in combination with the function compensation state index to generate an adaptive blood flow restriction pressure estimate value:

[0182] P adaptive =P initial ·(1+α·C compensation ·Δt1)

[0183] Wherein, P adaptiverepresents an adaptive blood flow restriction pressure prediction value, P initial represents an initial blood flow restriction pressure reference value, a represents a dynamic adjustment coefficient, C compensation represents a functional compensation state index, Δt1 represents a time adjustment factor;

[0184] fused neural remodeling activation intensity, the adaptive blood flow restriction pressure prediction value is corrected in a second stage through a preset neural-muscular synergy optimization algorithm, and an optimized blood flow restriction pressure value is generated;

[0185] The optimized blood flow restriction pressure value is subjected to real-time physiological feedback adaptation processing, and a personalized blood flow restriction training pressure value is generated.

[0186] In one embodiment, the present application also provides a computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the steps in the above method embodiments when executing the computer program.

[0187] In one embodiment, the present application also provides a computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps in the above method embodiments.

[0188] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts are described with reference to the parts of the method embodiments. The device embodiments described above are only schematic, and the components described as separate components can or can not be physically separate, and the components displayed as a unit can or can not be a physical unit, i.e. they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present disclosure according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0189] The above embodiments only express several implementation manners of the embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the patent scope of the embodiments of the present application. It should be pointed out that, for those skilled in the art, without departing from the concept of the embodiments of the present application, a number of modifications and improvements can be made, which are all within the protection scope of the embodiments of the present application.

Claims

1. A multimodal fusion-based intelligent assessment method for post-ACLR rehabilitation, characterized in that, The method includes: Peripheral modal data, central modal data, and rehabilitation intervention data of ACLR patients were collected and processed simultaneously to obtain a multi-source heterogeneous dataset; The multi-source heterogeneous dataset is spatiotemporally aligned using a preset unified timestamp to generate a time-synchronized cross-modal dataset. The cross-modal dataset is processed by a trained brain-muscle-bone multimodal fusion model to generate a three-dimensional comprehensive evaluation index including structural safety, functional recovery and neural reorganization. Based on the aforementioned three-dimensional comprehensive evaluation indicators, rehabilitation parameters are dynamically optimized to generate personalized blood flow restriction training pressure values ​​and proprioceptive training intensity.

2. The multimodal fusion-based intelligent assessment method for ACLR post-rehabilitation according to claim 1, characterized in that, The cross-modal dataset is processed using a trained brain-muscle-bone multimodal fusion model to generate a three-dimensional comprehensive evaluation index including structural safety, functional recovery, and neural reorganization. The biological structural degradation features in the cross-modal dataset are quantified and fused to generate structural safety. The biomechanical functional parameters in the cross-modal dataset are dynamically coupled to generate functional recoveries. The following formula is used to model compensatory paths for the neural plasticity parameters in the cross-modal dataset, generating neural reorganization degree: Among them, R neuro ω represents the neural reorganization degree, M represents the total number of modalities in the cross-modal dataset, K represents the number of neural plasticity parameters in each modality, and ω represents the neural plasticity degree. ik ΔP represents the weighting coefficient of the k-th parameter in the i-th mode. ik (t) represents the change of the k-th neuroplasticity parameter in the i-th modality at time t, where Δt represents the time interval, and λ ik d represents the attenuation coefficient. ik Indicates the distance of the compensation path; The structural safety, functional recovery, and neural reorganization are subjected to adaptive weighting during the rehabilitation phase to generate the three-dimensional comprehensive evaluation index.

3. The multimodal fusion-based intelligent assessment method for ACLR post-rehabilitation according to claim 2, characterized in that, The dynamic coupling processing of biomechanical functional parameters in the cross-modal dataset to generate functional recoveries includes: The muscle dynamic parameters in the cross-modal dataset are processed by power spectrum conversion to generate a muscle coactivation vector; The following formula is used to generate dynamic equilibrium control coefficients for the proprioceptive function parameters in the cross-modal dataset through four-quadrant stability analysis: Among them, K dc ω represents the dynamic balance control coefficient. q P represents the weighting factor for the q-th quadrant. q S represents the proprioceptive function parameter value in the q-th quadrant. q P represents the stability index in the q-th quadrant. ref Indicates the reference equilibrium point parameter value; The neuromuscular coupling processing is performed on the muscle strength co-activation vector and the dynamic balance control coefficient to generate the functional recovery degree.

4. The multimodal fusion-based intelligent assessment method for ACLR post-rehabilitation according to claim 2, characterized in that, The step of performing compensatory path modeling on the neural plasticity parameters in the cross-modal dataset to generate neural reorganization degree includes: The task-state brain function data in the cross-modal dataset are processed by sensorimotor cortex mapping to generate brain region compensatory activation maps; The resting-state brain function data in the cross-modal dataset are processed by whole-brain network entropy calculation to generate neural plasticity entropy values; The brain region compensatory activation map and the neural plasticity entropy value are subjected to reorganization path mining processing to generate the neural reorganization degree.

5. The multimodal fusion-based intelligent assessment method for ACLR post-rehabilitation according to claim 2, characterized in that, The step of quantifying and fusing biological structural degradation features in the cross-modal dataset to generate structural safety includes: The cartilage degeneration features in the cross-modal dataset are processed by T2 relaxation mapping to generate cartilage metabolic activity indicators; The muscle degeneration features in the cross-modal dataset are quantified by fat infiltration to generate a muscle fattening index; The cartilage metabolic activity index and the muscle fattening index are subjected to graft risk fusion processing to generate the structural safety score.

6. The multimodal fusion-based intelligent assessment method for ACLR post-rehabilitation according to claim 1, characterized in that, Acquiring the peripheral modal data includes: MRI scans of the knee joints of ACLR patients were performed to obtain knee joint structural data. Isokinetic muscle strength testing was performed on the quadriceps femoris muscles of ACLR patients to obtain muscle strength and function data. Prokin balance test was performed on ACLR patients to obtain proprioceptive data; The knee joint structural data, muscle strength and function data, and proprioceptive data are subjected to peripheral modal integration processing to generate the peripheral modal data.

7. The multimodal fusion-based intelligent assessment method for ACLR post-rehabilitation according to claim 1, characterized in that, The step of dynamically optimizing rehabilitation parameters based on the three-dimensional comprehensive evaluation indicators to generate personalized blood flow restriction training stress values ​​includes: The structural safety is graded by the stability of the graft microenvironment to generate a structural safety risk level. The functional recovery level is evaluated by neuromuscular functional compensation status to generate a functional compensation status index. The central remodeling efficiency of the neural remodeling degree is quantified to generate the neural remodeling activation intensity; Based on the structural safety risk level, an initial blood flow restriction pressure benchmark value is obtained by matching from a preset rehabilitation safety pressure spectrum; Using the following formula, combined with the functional compensation state index, a preset dynamic pressure regulation algorithm is used to perform a first-stage correction on the initial blood flow restriction pressure benchmark value, generating an adaptive blood flow restriction pressure estimate: P adaptive =P initial ·(1+α·C compensation ·Δt1) Among them, P adaptive P represents the adaptive blood flow limiting pressure prediction. initial The initial blood flow limiting pressure is the baseline value, α represents the dynamic adjustment coefficient, and C represents the initial blood flow limiting pressure. compensation Δt1 represents the functional compensation state index, and Δt1 represents the time adjustment factor. By integrating the neural remodeling activation intensity, the adaptive blood flow restriction pressure estimate is corrected in the second stage using a preset neuro-muscle synergistic optimization algorithm to generate an optimized blood flow restriction pressure value. The optimized blood flow restriction pressure value is subjected to real-time physiological feedback adaptation processing to generate the personalized blood flow restriction training pressure value.

8. A multimodal fusion-based intelligent assessment system for post-ACLR rehabilitation, characterized in that, The system includes: The multi-source heterogeneous data acquisition module is used to simultaneously collect and process peripheral modal data, central modal data and rehabilitation intervention data of ACLR patients to obtain multi-source heterogeneous datasets; The cross-modal spatiotemporal alignment module is used to perform spatiotemporal alignment processing on the multi-source heterogeneous dataset using a preset unified timestamp to generate a time-synchronized cross-modal dataset. The multimodal fusion assessment module is used to process the cross-modal dataset through a trained brain-muscle-bone multimodal fusion model to generate a three-dimensional comprehensive assessment index including structural safety, functional recovery and neural reorganization. The rehabilitation parameter dynamic optimization module is used to dynamically optimize rehabilitation parameters based on the three-dimensional comprehensive evaluation index, and generate personalized blood flow restriction training pressure values ​​and proprioceptive training intensity.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the multimodal fusion-based intelligent assessment method for ACLR post-rehabilitation according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the multimodal fusion intelligent assessment method for ACLR post-rehabilitation according to any one of claims 1 to 7.

Images