A multi-modal based lumbar spine holistic monitoring system

Abstract

The application provides a lumbar vertebra overall monitoring system based on multi-modal, and relates to the technical fields of biomedical engineering and motion monitoring, and comprises the following steps: an HD-sEMG array is arranged on both sides of T12-S1 vertebrae, which is used for monitoring the electromyographic signals of erector spinae and quadratus lumborum; a piezoelectric sensor is arranged on L3-L4 spinous process, which is used for detecting the dynamic tension of muscle fascia; three IMU sensor groups are distributed on T12-L1 interspinous process, L3-L4 interspinous process and S1 spinous process, which are respectively used for detecting the thoracolumbar flexion / tilt angle, lumbar vertebra rotation torque and pelvic tilt angle reference; a data fusion module is used for processing multi-source sensing data and driving the OpenSim lumbar vertebra model to reversely calculate the deep muscle activity, and the deep core muscle group is quantified; the spatiotemporal characteristics of surface electromyography, fascia tension gradient and joint kinematic parameters are fused, the activation intensity and biomechanical load rate of deep stabilizing muscles such as psoas and multifidus are reversely deduced, and the calculation error is reduced by more than 60% compared with the traditional indirect estimation method.

Description

Technical Field

[0001] This invention belongs to the field of biomedical engineering and motion monitoring technology, and more specifically, it relates to a multimodal lumbar spine monitoring system. Background Technology

[0002] Analysis of existing technical defects and their root causes:

[0003] Physiological limitations of a single sensing modality:

[0004] Surface electromyography (sEMG): It can only capture electrical signals of superficial muscles (such as erector spinae), has a significant attenuation effect on the electrical activity of deep core stabilizing muscles (multifidus, iliopsoas), and cannot sense changes in the mechanical tension of fascial tissue.

[0005] Inertial Measurement Unit (IMU): Although it can provide joint angle data, the physiological correlation between kinematic parameters and muscle activation state and intervertebral disc load is missing.

[0006] Biomechanical model disconnect: Traditional OpenSim simulations rely on laboratory optical motion capture systems for input, which cannot acquire individualized data in real time in natural motion scenarios, causing the model output to deviate from the actual physiological state.

[0007] Anatomical misalignment of sensor deployment: Current wearable solutions neglect key biomechanical nodes of the lumbar spine.

[0008] Thoracolumbar region (T12-L1): The transition zone from thoracic kyphosis to lumbar lordosis, accounting for 60% of the total lumbar spine range of motion;

[0009] Core load zone (L3-L4): Bears the peak axial pressure and is the stress concentration area of ​​the intervertebral disc;

[0010] Sacral reference plane (S1): Pelvic tilt directly affects lumbar curvature. Failure to optimize sensor layout for the above nodes resulted in a break in the spinal dynamics chain data.

[0011] Multi-source data fusion is hampered by time-domain / frequency-domain mismatches in electromyography (millisecond transients), mechanical signals (millisecond-second gradual changes), and kinematic signals (second-level trends). Existing systems lack hardware-level synchronization mechanisms, which reduces the reliability of model input data. Summary of the Invention

[0012] To address the aforementioned technical problems, this invention provides a multimodal lumbar spine monitoring system.

[0013] A lumbar spine multimodal monitoring system includes: a high-density surface electromyography (HD-sEMG) electrode array deployed on both sides of the T12 to S1 vertebral bodies; a piezoelectric fascial tension sensor located at the L3-L4 spinous processes; three inertial measurement units (IMUs) fixed between the T12-L1 spinous processes, between the L3-L4 spinous processes, and at the S1 spinous process, respectively; and a personalized lumbar spine biomechanical model constructed based on OpenSim, which receives the three-modal sensing data and inversely calculates the activation degree of deep muscles.

[0014] Preferably, the HD-sEMG array electrode spacing is 5-10mm, covering the erector spinae and quadratus lumborum muscle belly regions. The piezoelectric sensor is a polyvinylidene fluoride (PVDF) film with a thickness ≤0.1mm and a dynamic response frequency band of 0.1-50Hz. The IMU measures the sagittal flexion-extension angle at nodes T12-L1, measures the horizontal rotational torque at nodes L3-L4, and establishes a pelvic motion reference coordinate system at node S1. The OpenSim model solves the joint torque through inverse dynamics and optimizes muscle force distribution by integrating electromyographic activation and fascial stiffness constraints. It includes an FPGA hardware synchronization module to achieve a time alignment error of <1ms for the three-modal signals. The deep muscle activation includes the real-time activation intensity and biomechanical load percentage of the iliopsoas and multifidus muscles.

[0015] A method for assessing lumbar spine function includes the following steps: simultaneously acquiring electromyography (EMG), fascial tension, and joint kinematics data; extracting EMG time-frequency domain features, fascial tension change rate, and joint angle parameters; driving a personalized OpenSim model to inversely calculate deep muscle activation; generating a biomechanical assessment report containing muscle dysfunction warnings; utilizing sacral IMU data to eliminate interference from pelvic movement on lumbar spine dynamics calculations; and using L3-L4 piezoelectric signal peaks to trigger iterative convergence of a muscle strength optimization algorithm during bending movements.

[0016] Constructing a closed-loop system integrating a multimodal sensor network and a biomechanical model includes:

[0017] High-density surface electromyography (HD-sEMG) array:

[0018] The electrodes were placed in the bilateral erector spinae and quadratus lumborum regions of the T12-S1 vertebral bodies to capture the spatiotemporal changes in electromyography using a two-dimensional electrode grid.

[0019] piezoelectric sensing layer:

[0020] It is attached to the spinous processes of L3-L4 and the myofascial tension is quantified in real time through dynamic pressure resistance response.

[0021] Multi-node IMU attitude sensing unit:

[0022] First IMU: Between the spinous processes of T12 and L1, to detect flexion / extension / lateral tilt angles of the thoracolumbar region;

[0023] Second IMU: Between the spinous processes of L3 and L4, measuring the rotational torque of the lumbar spine;

[0024] Third IMU: S1 spinous process, used as a reference for pelvic tilt angle.

[0025] OpenSim model reverse computation engine:

[0026] By integrating data from three types of sensors, a personalized lumbar spine model is driven to reverse-engineer the activation level and biomechanical load of deep muscles (iliopsoas and multifidus).

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] In this invention, a multimodal collaborative sensing mechanism is proposed: a pioneering three-dimensional data coupling architecture of HD-sEMG (electrophysiology) + piezoelectric sensing (mechanics) + IMU (kinematics) covers the entire "muscle-fascia-skeleton" link.

[0029] In this invention, the key biomechanical node layout is as follows: the sensor positions are optimized based on the spinal load transmission path (e.g., L3-L4 is the stress concentration area of ​​the lumbar spine) to ensure that the data has clear physiological significance.

[0030] In this invention, a dynamic model closed-loop drive is used: OpenSim model parameters are iteratively corrected through real-time sensor data to achieve non-invasive reverse calculation of deep muscle activity.

[0031] Achieving comprehensive dynamic analysis of the lumbar spine's biomechanical state:

[0032] This invention, through a multimodal sensing collaborative mechanism, is the first to simultaneously capture three core physiological parameters of the lumbar spine system under natural activity conditions:

[0033] Muscle electrophysiological activity: HD-sEMG array monitors the excitation conduction pathways of erector spinae and quadratus lumborum muscles with millimeter-level spatial resolution, accurately identifying abnormal muscle activation timing (such as bilateral muscle strength asymmetry).

[0034] Soft tissue mechanical load: Piezoelectric sensors on the L3-L4 spinous processes directly quantify the dynamic tension of the fascia and provide real-time feedback on the compressive stress on the intervertebral disc and the ligament stretching status.

[0035] Spinal kinematic characteristics: A three-node IMU group constructs a "thoracolumbar hub-lumbar core-pelvic reference" kinematic chain, accurately calculating the six-degree-of-freedom motion trajectory of lumbar flexion, extension, lateral bending, and rotation. Breakthrough significance: It completely overcomes the one-sided limitations of traditional single-modal devices (such as simple sEMG or IMU), providing a complete dynamic profile of the "muscle-fascia-skeleton" coupled system.

[0036] A revolutionary non-invasive assessment paradigm for deep muscle function:

[0037] Based on the OpenSim model, a multi-source data reverse calculation engine has achieved breakthroughs in two major clinical bottlenecks: quantification of deep core muscle groups: by integrating the spatiotemporal characteristics of surface electromyography, fascial tension gradient and joint kinematic parameters, the activation intensity and biomechanical load rate of deep stabilizing muscles such as iliopsoas and multifidus can be reversed, and the calculation error is reduced by more than 60% compared with the traditional indirect estimation method.

[0038] Muscle Synergy Pattern Analysis: Through a dynamic optimization algorithm, this study visualizes the synergistic relationship of force transmission between superficial motor muscles (erector spinae) and deep stabilizing muscles (multifidus) during lumbar flexion-extension, accurately identifying pathological patterns such as "muscle compensation" or "disability inhibition." Clinical Value: This provides quantitative diagnostic evidence for early muscle function decline in diseases such as lumbar disc degeneration and chronic low back pain.

[0039] Constructing a precise, individualized biomechanical early warning system:

[0040] Anatomy-driven risk warning: When the L3-L4 piezoelectric sensor detects fascial tension exceeding the threshold (>150N) and HD-sEMG shows delayed erector spinae muscle activation, the model automatically marks "intervertebral disc overload risk"; when the S1 reference plane IMU monitors pelvic tilt angle >15° and multifidus muscle load rate <50%, a "core stability failure" alarm is triggered.

[0041] Dynamic adaptive optimization: For subjects with different body types, the OpenSim model's muscle lever arm parameters and tendon stiffness coefficients are automatically corrected through standard movement calibration (e.g., 60° forward flexion), outputting a personalized biomechanical baseline. Application advantages: Significantly improves the scientific rigor of rehabilitation training guidance and occupational safety, and reduces the recurrence rate of lumbar spine injuries.

[0042] Driving technological change in wearable medical devices:

[0043] Adaptation to natural movement scenarios: Flexible piezoelectric films and micro IMU modules can be directly embedded in rehabilitation gears to support continuous monitoring of daily movements such as walking, bending over, and carrying, breaking through the spatial limitations of fixed laboratory equipment;

[0044] Real-time closed-loop feedback: The system can complete the entire process of "data acquisition → model calculation → risk classification" within 200ms, providing wearers with vibration and tactile or sound and light warnings, achieving true dynamic protection. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the data processing algorithm flow of the present invention;

[0046] Figure 2 This is a schematic diagram of the motion joint sensing process of the present invention. Detailed Implementation

[0047] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0048] Please see Figures 1-2 This invention provides a lumbar spine monitoring system based on multimodal sensing, comprising: an HD-sEMG array deployed on both sides of the T12-S1 vertebral bodies for monitoring electromyographic signals of the erector spinae and quadratus lumborum muscles; a piezoelectric sensor placed on the L3-L4 spinous processes for detecting dynamic tension of the myofascia; three IMU sensor groups distributed between the T12-L1 spinous processes, between the L3-L4 spinous processes, and the S1 spinous process to detect flexion / extension / lateral tilt angles of the thoracolumbar spine, lumbar rotational torque, and pelvic tilt reference angles, respectively; and a data fusion module for processing multi-source sensor data and driving the OpenSim lumbar spine model to inversely calculate deep muscle activity.

[0049] The HD-sEMG array covers the target muscle group with a grid topology, with an electrode spacing of ≤10mm. The piezoelectric sensor is a flexible PVDF film with a thickness of ≤0.1mm, which is attached to the surface of the L3-L4 spinous process. The three-node layout of the IMU sensor forms a spinal dynamics chain reference system. The deep muscle activity includes the real-time activation intensity and biomechanical load of the iliopsoas and multifidus muscles. The data fusion module includes a spatiotemporal synchronization unit with a synchronization accuracy of <1ms. The OpenSim model is personalized based on individual anatomical parameters.

[0050] A method for assessing lumbar spine function includes: simultaneously acquiring electromyographic, fascial tension, and kinematic data; calculating and outputting deep muscle activity through inverse dynamics; generating a lumbar spine biomechanical state assessment report; the data processing includes wavelet denoising of electromyographic signals, dynamic calibration of piezoelectric signals, and IMU posture quaternion fusion; using S1 spinous process IMU data as a pelvic reference to correct lumbar spine-pelvic coupling motion errors.

[0051] Multimodal physiological signal coupling mechanism

[0052] Electro-mechanical-motion joint sensing:

[0053] Key node collaborative feedback: L3-L4 piezoelectric sensors detect abnormal increases in fascial tension in real time, triggering the model to prioritize the optimization of the muscle force distribution algorithm for that segment.

[0054] Dynamic calibration based on anatomical features:

[0055] When subjects performed standard forward flexion movements, the ratio between L3-L4 rotational torque and erector spinae sEMG amplitude was calibrated using S1-IMU data as a baseline.

[0056] During the walking cycle, the three-modal data stream is synchronized using heel impact events to eliminate phase delay between sensors.

[0057] Non-invasive reverse calculation of muscle activation

[0058] Define the objective function:

[0059] in:

[0060] τ IMU Measured lumbar torque using an IMU;

[0061] The force is a complex force between muscles and tendons, constrained by a combination of HD-sEMG activation and piezoelectric stiffness.

[0062] r i The lever arm represents the muscle. The activation time series of the iliopsoas and multifidus muscles are solved using the gradient descent method.

[0063] Sensor deployment

[0064] Step 1: HD-sEMG electrodes were arranged in a 4×6 array to cover the bilateral erector spinae muscles from T12 to L1 to S1, with a sampling frequency of 2kHz.

[0065] Step 2: Attach a flexible piezoelectric film to the L3-L4 spikes, with a sensitivity of 0.5V / N and a response bandwidth of 0-50Hz.

[0066] Step 3: Fix the three-axis IMU modules to:

[0067] T12-L1: Measures sagittal flexion-extension angles (±90°, accuracy 0.1°)

[0068] L3-L4: Detects horizontal rotational torque (range ±10Nm)

[0069] S1: Output pelvic tilt reference (error < 0.5°).

[0070] Data processing flow

[0071] Spatiotemporal synchronization: Multi-source data timestamps are aligned via FPGA hardware (synchronization error < 1ms).

[0072] Feature fusion:

[0073] Electromyographic characteristics: extraction of time-frequency parameters such as RMS and MF;

[0074] Piezoelectric signal: Calculates the gradient of tension change;

[0075] IMU data: quaternion-based calculation of joint angles.

[0076] Model-driven: Input the feature vector into the OpenSim customized lumbar spine model and use the inverse dynamics algorithm to solve the iliopsoas / multifidus muscle activation.

[0077] Test scenario: Subjects pick up a 5kg weight from a standing position with a forward bend of 60°.

[0078] Raw data:

[0079] HD-sEMG: The RMS value of the erector spinae muscles increased to 320% of the resting state;

[0080] Piezoelectric sensor: L3-L4 fascia tension reaches 142N;

[0081] IMU: L3-L4 segment buckling angle 58.7°, rotational torque 4.2 Nm;

[0082] Model output:

[0083] Iliopsoas muscle activation: 76.4% (traditional models can only estimate superficial muscles);

[0084] Multifidus muscle loading rate: 89.1%, significantly higher than the healthy threshold (<70%);

[0085] Clinical significance: High load rate indicates a risk of multifidus muscle dysfunction, requiring core stability training intervention.

[0086] Hardware deployment process:

[0087] Skin pretreatment: Wipe the T12-S1 spinous process area with 75% alcohol to reduce skin resistance;

[0088] HD-sEMG array positioning:

[0089] Longitudinal: Using the spinous processes of T12, L3, and S1 as landmarks, the center line of the array is parallel to the direction of the erector spinae muscles;

[0090] Lateral: The bilateral arrays are symmetrically distributed 3 cm lateral to the spinous process;

[0091] Piezoelectric film mounting:

[0092] Align the center of the PVDF sensor with the interspinous space between L3 and L4, and secure it with the edge of medical tape;

[0093] Apply a preload of 0.5N to ensure mechanical coupling;

[0094] IMU module fixed:

[0095] T12-L1 unit: Rigid bracket mounted across the spinous process;

[0096] L3-L4 units: embedded in a flexible waistband and pressurized for attachment;

[0097] S1 unit: Adhere to the median crest of the sacrum.

[0098] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A lumbar spine multimodal monitoring system, characterized in that, include: High-density surface electromyography (HD-sEMG) electrode arrays deployed on both sides of the T12 to S1 vertebral bodies; Piezoelectric fascial tension sensor located at the L3-L4 spinous process; Three inertial measurement units (IMUs) were fixed between the spinous processes of T12-L1, between the spinous processes of L3-L4, and at the spinous process of S1, respectively. A personalized lumbar spine biomechanical model built on OpenSim receives trimodal sensor data and inversely calculates the activation level of deep muscles.

2. The system as described in claim 1, characterized in that: The HD-sEMG array electrodes are spaced 5-10 mm apart, covering the erector spinae and quadratus lumborum muscle belly regions.

3. The system as described in claim 1, characterized in that: The piezoelectric sensor is a polyvinylidene fluoride (PVDF) film with a thickness of ≤0.1mm and a dynamic response frequency band of 0.1-50Hz.

4. The system as described in claim 1, characterized in that: The IMU measures the sagittal flexion-extension angle at nodes T12-L1, measures the horizontal rotational torque at nodes L3-L4, and establishes a pelvic motion reference coordinate system at node S1.

5. The system as described in claim 1, characterized in that: The OpenSim model solves for joint torques through inverse dynamics and optimizes muscle force distribution by incorporating electromyographic activation and fascial stiffness constraints.

6. The system as described in claim 1, characterized in that: It includes an FPGA hardware synchronization module to achieve a time alignment error of <1ms for three-mode signals.

7. The system as described in claim 1, characterized in that: Deep muscle activation includes the real-time activation intensity and biomechanical load percentage of the iliopsoas and multifidus muscles.

8. A method for assessing lumbar spine function based on claims 1-7, comprising the following steps: Simultaneous acquisition of electromyography, fascial tension, and joint kinematics data; Extract time-frequency domain features of electromyography, rate of change of fascial tension, and joint angle parameters; Drive personalized OpenSim models to inversely calculate deep muscle activation; Generate a biomechanical assessment report that includes early warnings of muscle dysfunction.

9. The method as described in claim 8, characterized in that: Sacral IMU data was used to eliminate the interference of pelvic motion on lumbar spine dynamics calculations.

10. The method as described in claim 8, characterized in that: During the bending motion, the peak value of the L3-L4 piezoelectric signal triggers the iterative convergence of the muscle force optimization algorithm.

Images