Multi-modal image joint screening method and system for preventive physical examination

Abstract

The application discloses a multi-modal image joint screening method and system for preventive physical examination, relates to the technical field of multi-modal image joint screening, and comprises the following steps: synchronously acquiring a laryngeal surface electromyogram and a vocal cord vibration audio signal of a subject during self-sound emission; calculating energy features and audio instantaneous frequency fluctuation rates in real time; generating a trigger pulse when the energy features both satisfy a threshold value and the frequency fluctuation rate is lower than a tolerance and lasts for a preset time; controlling an ultrasonic probe to emit a plane wave at a target pulse repetition frequency based on the pulse, synchronously acquiring an ultrasonic radio frequency echo signal and a structure image; generating a tissue mask to remove a liquid area based on the structure image; extracting forward propagation wave field displacement information and calculating a wave peak arrival time difference based on the radio frequency echo signal; and combining the mask, the time difference and a preset tissue density to calculate a local shear modulus of the tissue; the application precisely locks a sound emission steady state window through multi-modal signals, and improves the accuracy and repeatability of shear modulus measurement.

Description

Technical Field

[0001] This invention relates to the field of multimodal image combined screening technology, specifically to a multimodal image combined screening method and system for preventive physical examinations. Background Technology

[0002] This involves the field of thyroid ultrasound elastography technology, specifically a screening technique that assesses tissue shear modulus by capturing endogenous vibration signals caused by the subject's spontaneous vocalizations. It is widely used in preventive physical examinations for early risk assessment and quantitative analysis of thyroid nodules.

[0003] Existing technologies typically employ continuous ultrasound acquisition or manual triggering to capture minute tissue displacements during the vocalization process. However, due to the complex neuromuscular coupling involved in human vocalization, the resulting mechanical waves exhibit high transientity and randomness in terms of energy intensity and frequency stability. Existing technologies lack real-time monitoring and feedback control logic for the physiological signal state of vocalization, resulting in ultrasound sampling being unable to accurately lock onto the steady-state window of mechanical energy output. This easily introduces significant measurement noise due to sampling timing deviations, severely impacting the accuracy of risk assessment and the reliability of clinical diagnosis.

[0004] Therefore, a multimodal imaging joint screening method for preventive physical examinations is proposed. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multimodal image-based joint screening method and system for preventive physical examinations.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] In a first aspect, the present invention discloses a multimodal image-based joint screening method for preventive physical examinations, comprising the following steps:

[0008] Simultaneously acquire electromyographic signals of the laryngeal surface and audio signals of vocal cord vibration during the subject's spontaneous vocalization process;

[0009] Real-time calculation of the energy characteristics of electromyographic signals on the laryngeal surface and the energy characteristics and instantaneous frequency fluctuation rate of vocal cord vibration audio signals;

[0010] A trigger pulse signal is generated when the energy characteristics of the electromyography signal on the laryngeal surface and the energy characteristics of the vocal cord vibration audio signal both meet the preset energy threshold, and the instantaneous frequency fluctuation rate of the vocal cord vibration audio signal is lower than the preset fluctuation tolerance and lasts for a preset time.

[0011] Based on the trigger pulse signal, the ultrasonic probe is controlled to emit a plane wave at the target pulse repetition frequency, and ultrasonic radio frequency echo signal and structural image are acquired simultaneously.

[0012] Tissue masks for target regions are generated based on structural images to eliminate fluid-filled areas;

[0013] Based on the ultrasonic radio frequency echo signal, the wave field displacement information of the forward propagation is extracted, and the wave peak arrival time difference is calculated.

[0014] By combining the tissue mask, the time difference of arrival of the wave peak, and the preset tissue density, the local shear modulus of the tissue in the target area is calculated.

[0015] Secondly, the present invention discloses a multimodal imaging joint screening system for preventive physical examinations, including: a signal acquisition module, used to simultaneously acquire the electromyographic signals of the laryngeal surface and the audio signals of vocal cord vibration during the subject's spontaneous vocalization process;

[0016] The signal triggering module is used to calculate the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics and instantaneous frequency fluctuation rate of the vocal cord vibration audio signal in real time. When the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics of the vocal cord vibration audio signal both meet the preset energy threshold, and the instantaneous frequency fluctuation rate of the vocal cord vibration audio signal is lower than the preset fluctuation tolerance and lasts for a preset time, a trigger pulse signal is generated.

[0017] The signal acquisition module is used to control the ultrasound probe to emit plane waves at the target pulse repetition frequency based on the trigger pulse signal, and simultaneously acquire ultrasound radio frequency echo signals and structural images.

[0018] The tissue mask generation module is used to generate tissue masks for target regions based on structural images to eliminate fluid-containing areas;

[0019] The wave field displacement processing module is used to extract the forward propagation wave field displacement information based on the ultrasonic radio frequency echo signal and calculate the wave peak arrival time difference.

[0020] The shear modulus calculation module is used to calculate the local shear modulus of the target area by combining the tissue mask, the time difference of arrival of the peak, and the preset tissue density.

[0021] Compared with existing technologies, the beneficial effects of this invention are as follows: by jointly judging dual-modal signals, the stability and repeatability of the excitation source are ensured; by plane wave high-speed imaging and directional filtering, the propagation process of endogenous shear waves is accurately captured; by tissue masking and time-of-flight method, the local shear modulus of the target area is reliably calculated; by pseudo-color mapping and fusion display, elastic information is transformed into intuitive clinical visualization images, providing a technical means for early screening and benign / malignant assessment of thyroid nodules. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is an overall block diagram of the method in Embodiment 1 of the present invention;

[0024] Figure 2 This is a schematic diagram of the dual-modal triggering timing in Embodiment 1 of the present invention;

[0025] Figure 3 This is an overall block diagram of the system in Embodiment 2 of the present invention. Detailed Implementation

[0026] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Example 1:

[0028] like Figures 1-2 As shown, the multimodal imaging combined screening method for preventive physical examinations includes the following steps:

[0029] To facilitate understanding, this example illustrates a thyroid nodule screening procedure performed on a 45-year-old patient in the endocrinology department of a hospital. Suppose that during a routine physical examination, a hypoechoic nodule approximately 8 mm in diameter was found in the right lobe of the thyroid gland, requiring further evaluation for its benign or malignant nature. Traditional ultrasound elastography requires the doctor to manually press the probe to apply external force, but inconsistencies in the pressure, frequency, and direction of this pressure can lead to poor repeatability of the measurement results. This application utilizes the intrinsic acoustic vibrations generated by the subject's spontaneous vocalization as the excitation source. By simultaneously acquiring electromyographic signals from the laryngeal surface and vocal cord vibration audio signals, a non-invasive quantitative assessment of the shear modulus of thyroid tissue is achieved.

[0030] In this step, synchronous acquisition refers to the acquisition of surface electromyographic signals of the larynx under the control of a unified hardware clock. Harmony vocal cord vibration audio signal Time-aligned acquisition was performed. The electromyographic signal on the surface of the larynx reflects the neuroelectrical activity of the laryngeal muscle group, with amplitude in mV; the vocal cord vibration audio signal reflects the sound pressure fluctuations generated by vocal cord vibration, with amplitude in Pa or normalized amplitude.

[0031] It should be noted that voluntary vocalization is a neuromuscular physiological process. When the brain issues a vocalization command, the neural electrical signals are first transmitted to the laryngeal muscles, causing depolarization and generating electromyographic signals that can be detected on the body surface. Subsequently, muscle contraction generates tension that overcomes airflow resistance, causing the vocal cords to vibrate and producing an audio signal that can be captured by a microphone. These two signals have an inherent causal relationship in time: the electromyographic signal appears before the audio signal, and their energy envelopes are highly correlated during the steady-state vocalization phase. This application utilizes this physiological causal relationship to accurately identify the steady-state window during vocalization through joint analysis of the two modal signals.

[0032] Specifically, the acquisition of laryngeal surface electromyography (EMG) signals is achieved by attaching a pair of surface electrodes symmetrically to each side of the Adam's apple on the front of the subject's neck. The electrodes are made of medical-grade Ag / AgCl material, with an electrode spacing of approximately 20 mm and a sampling rate of no less than 1000 Hz. To reduce interference from the power frequency trunk line and motion artifacts, a 50 Hz notch filter and a 0.5-500 Hz bandpass filter are integrated into the acquisition circuit.

[0033] Following the previous patient case, before the examination began, medical staff attached surface electrodes approximately 15mm to either side of the patient's Adam's apple on the front of the neck. Conductive gel was applied between the electrodes and the skin to reduce contact resistance. The patient was instructed to continuously pronounce the "a" sound at a comfortable volume, at which point the acquisition system began recording the surface electromyographic signals of the larynx. Suppose at a certain moment... The amplitude of the raw electromyographic signal was approximately 0.8 mV, and the effective signal amplitude after bandpass filtering was approximately 0.6 mV.

[0034] The vocal cord vibration audio signal was acquired using a high-sensitivity microphone placed approximately 10 cm in front of the subject's mouth. The microphone had a sampling rate of at least 10 kHz and a frequency response range covering 80-8000 Hz, effectively capturing the fundamental frequency of the human voice and its harmonic components. The acquired audio signal was pre-amplified and converted from analog to digital before being synchronously recorded with the electromyographic signal under the same hardware clock.

[0035] Using the aforementioned patient case, when the patient continuously pronounced the "a" sound, the microphone collected the audio signal. It exhibits periodic oscillation characteristics. Assuming the patient's fundamental frequency is approximately 220Hz (corresponding to pitch A3), the main period of the audio signal is approximately 4.5ms. During the steady-state vocalization phase, the normalized amplitude of the audio signal stabilizes at around 0.7, while at the beginning and end of vocalization, the amplitude shows distinct rising and falling edges.

[0036] To ensure the time alignment accuracy of the two signals, this application adopts a hardware synchronous acquisition scheme based on FPGA. The FPGA integrates a unified master clock (frequency not less than 100MHz), and generates sampling trigger pulses for the electromyography (EMG) and audio signals through clock frequency division, ensuring strict synchronization of the sampling times of the two signals with a time alignment error of less than 10μs. This hardware-level synchronization mechanism avoids the time jitter that may be introduced by multi-threaded acquisition at the software level, providing a reliable time reference for subsequent steady-state window identification.

[0037] Using the above method, the system obtained time-aligned laryngeal surface electromyography signals. Harmony vocal cord vibration audio signal This lays the data foundation for subsequent energy envelope calculation and steady-state window selection. This dual-modal synchronous acquisition design fully utilizes the physiological characteristics of the human vocalization process, enabling the system to comprehensively monitor the vocalization state from multiple links of neuromuscular-mechanical coupling. Compared with solutions that rely solely on audio signals, it has higher robustness and anti-interference ability.

[0038] Continuing with the aforementioned patient case, after completing the laryngeal surface electromyography signal... Harmony vocal cord vibration audio signal After synchronous acquisition, the system enters the real-time identification phase of the steady-state window. The core objective of this phase is to accurately identify the steady-state window during the sound generation process, providing a precise time reference for subsequent ultrasonic triggering.

[0039] Specifically, the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics and instantaneous frequency fluctuation rate of the vocal cord vibration audio signal are calculated in real time, including: based on a preset sliding time window, calculating the root mean square value of the laryngeal surface electromyography signal and the vocal cord vibration audio signal within the time window as the corresponding energy characteristics.

[0040] In this step, the energy characteristics of the laryngeal surface electromyography (EMG) signal are characterized by the root mean square (RMS) value within a sliding time window. For the laryngeal surface EMG signal... The formula for calculating its root mean square value is:

[0041] ;in, The sliding time window is set to a value ranging from 10ms to 100ms, with 50ms being the preferred value. The selection of the sliding time window requires balancing the temporal resolution and stability of the energy estimation: a window that is too short will cause large fluctuations in the energy estimate, easily leading to false triggers; a window that is too long will cause response delays, missing the optimal ultrasonic sampling opportunity. The unit is the same as the original electromyographic signal, still in mV, and this value reflects the time window. Effective energy level of internal electromyographic signals.

[0042] Similarly, the energy characteristics of the vocal cord vibration audio signal are also characterized by the root mean square value within a sliding time window, calculated using the following formula: ;

[0043] This reflects the effective energy level of the audio signal within the time window. When the audio signal is represented using normalized amplitude... The value range is from 0 to 1.

[0044] Using the aforementioned patient case, assuming a long sliding time window With a value of 50ms and a system sampling rate of 10kHz, each time window contains 500 sampling points. After the patient begins uttering the "a" sound, the energy characteristics of both the electromyographic and audio signals begin to rise. Assuming at a certain moment... Root mean square value of electromyography signal The calculated root mean square value of the audio signal is 0.55mV. The calculated value is 0.72. These energy characteristic values ​​will be used for subsequent threshold determination.

[0045] Furthermore, the instantaneous frequency of the vocal cord vibration audio signal is extracted, and the rate of change of the instantaneous frequency over time is calculated as the instantaneous frequency fluctuation rate.

[0046] Instantaneous frequency is a key indicator characterizing the stability of vocal cord vibration. During steady-state phonation, the vocal cords vibrate at a relatively stable fundamental frequency, with minimal fluctuations in instantaneous frequency. However, at the beginning, end, or unstable stages of phonation, significant fluctuations in instantaneous frequency occur. This application employs short-time Fourier transform or Hilbert transform methods to extract the instantaneous frequency of the audio signal. Taking the Hilbert transform as an example, firstly, the audio signal... The Hilbert transform is performed to obtain the analytic signal, and then the instantaneous frequency is extracted from the phase information of the analytic signal.

[0047] The instantaneous frequency fluctuation rate is defined as the absolute value of the rate of change of the instantaneous frequency over time, that is: ;

[0048] This indicator reflects the frequency stability of vocal cord vibration. During steady-state vocalization, the instantaneous frequency fluctuation rate is typically less than 5% / s, meaning that the frequency change per second does not exceed 5% of the current frequency.

[0049] Using the aforementioned patient case, assuming the patient's fundamental frequency is approximately 220Hz, during the steady-state vocalization phase, the instantaneous frequency fluctuates stably between 218Hz and 222Hz, with an instantaneous frequency fluctuation rate of approximately 2Hz / s, or about 0.9% / s of the fundamental frequency, far below the threshold of 5% / s. However, at the onset of vocalization, the instantaneous frequency may rapidly rise from 0Hz to 220Hz, with the instantaneous frequency fluctuation rate potentially reaching hundreds of Hz / s, far exceeding the threshold.

[0050] Through the above calculations, the system obtained the energy characteristics of the surface electromyography signal of the larynx. Energy characteristics of vocal cord vibration audio signals and instantaneous frequency fluctuation rate These three key indicators provide a quantitative basis for subsequent judgment of the steady-state window period.

[0051] Next, under the condition that the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics of the vocal cord vibration audio signal both meet the preset energy threshold, and the instantaneous frequency fluctuation rate of the vocal cord vibration audio signal is lower than the preset fluctuation tolerance and lasts for a preset time, a trigger pulse signal is generated, including: synchronously acquiring the laryngeal surface electromyography signal and the vocal cord vibration audio signal through a unified hardware clock.

[0052] Traditional elastography relies on external mechanical excitation, with the timing and intensity of which are manually controlled by the operator, resulting in subjectivity and non-repeatability. This application, by detecting the steady-state window during the vocalization process and utilizing human spontaneous vocalization as an endogenous excitation source, achieves automated identification and standardized control of the excitation timing.

[0053] Determining the steady-state window requires the simultaneous fulfillment of four conditions: First, the energy characteristics of the electromyographic signal on the laryngeal surface. Greater than the preset electromyographic energy threshold Second, the energy characteristics of the vocal cord vibration audio signal. Greater than the preset audio energy threshold Third, instantaneous frequency fluctuation rate Less than the preset fluctuation tolerance Fourth, the duration for which all three conditions above are met exceeds the steady-state maintenance time threshold. .

[0054] Trigger pulse signal The generation logic can be expressed by the following formula:

[0055] ;

[0056] in, The electromyographic energy threshold ranges from 0.3mV to 0.8mV and can be adaptively adjusted according to individual differences of the subject and electrode position. The audio energy threshold is used when the audio signal uses a normalized amplitude. The value range is from 0.3 to 0.6; The frequency fluctuation tolerance ranges from 3% / s to 10% / s, with a preferred value of 5% / s. The steady-state hold-up time threshold is set between 30ms and 100ms, with 50ms being the preferred value.

[0057] It should be noted that a steady-state hold-up time threshold is introduced. The goal is to filter out transient pseudo-steady states. At the onset of vocalization, the energy of electromyographic and audio signals may briefly exceed a threshold, and the instantaneous frequency fluctuation rate may briefly decrease; however, this state is unstable and typically lasts less than 30ms. By setting... The constraints ensure that the system only triggers ultrasonic sampling after it has truly entered the steady-state sound generation stage, thereby guaranteeing the energy stability and frequency consistency of the shear wave excitation source.

[0058] Furthermore, when the root mean square value is detected to be greater than the corresponding energy threshold, and the absolute value of the instantaneous frequency change rate is less than the preset fluctuation tolerance for a duration exceeding the steady-state holding time threshold, a hardware trigger pulse signal is sent to the ultrasound front end.

[0059] In this step, the hardware trigger pulse signal is generated by the FPGA and sent to the ultrasound front end. The FPGA internally implements a three-input AND gate logic: the first input is the electromyographic energy comparison result (…). The second input is the audio energy comparison result. The third input is the frequency stability comparison result. When all three inputs are high, a counter is started to count. When the count value exceeds a preset threshold (corresponding to...), the counter continues to count. When the time is specified, a high-level trigger pulse is output. .

[0060] Using the aforementioned patient case, let's assume the electromyographic energy threshold. Set to 0.4mV, audio energy threshold Set to 0.5, frequency fluctuation tolerance Set to 5% / s, steady-state hold time threshold Set to 50ms. At a certain moment after the patient begins to pronounce the "a" sound. electromyography energy Audio energy Instantaneous frequency fluctuation At this point, all three conditions are met simultaneously, and the counter inside the FPGA starts counting. Assuming the system clock frequency is 100MHz, the count value will reach... At 50ms, the FPGA sends a positive trigger pulse signal with a width of 10μs to the ultrasound front end.

[0061] Upon receiving the trigger pulse, the ultrasound front end immediately initiates a high-speed plane wave imaging sequence to begin acquiring ultrasound radiofrequency data of the thyroid tissue. Because the trigger moment is precisely locked within the steady-state vocalization window, the mechanical waves generated by vocal cord vibration are propagating in the neck tissue at a stable frequency and energy, providing high-quality excitation conditions for subsequent shear wave velocity calculation and shear modulus inversion.

[0062] It should be noted that the triggering mechanism of this application is based on the physical constraints of causality and energy conservation. The neural electrical signal first reaches the laryngeal muscles, causing depolarization (generating...). Muscle contraction generates tension to overcome resistance, causing the vocal cords to vibrate (producing...). The mechanical waves generated by vocal cord vibration propagate through the neck tissue to the thyroid region. This causal chain dictates that the thyroid region must receive sufficiently strong endogenous shear wave excitation only if both electromyographic energy and audio energy are adequate. Relying solely on audio signals may miss cases of insufficient muscle activation, leading to unstable excitation source energy; while relying solely on electromyographic signals cannot confirm whether the vocal cords have truly entered a steady-state vibration. This application comprehensively monitors the vocal state from multiple aspects of the neuromuscular-mechanical coupling through joint judgment of dual-modal signals, ensuring the accuracy and reliability of the triggering moment.

[0063] Furthermore, to accommodate individual differences among examinees, this application supports adaptive adjustment of the energy threshold. Before the examination begins, the examinee can be instructed to perform a trial vocalization, and the system records the peak values ​​of electromyographic energy and audio energy during this process. The electromyographic energy threshold is then calculated. Set the audio energy threshold to 60% to 80% of the peak value. The peak value is set to 70% to 90%. This adaptive mechanism can effectively cope with the differences in signal amplitude caused by factors such as age, gender, and vocal habits of different subjects, thus improving the applicability and robustness of the system.

[0064] Using the aforementioned patient case, let's assume that during the vocalization trial phase, the patient's peak electromyographic energy was 0.7mV and the peak audio energy was 0.85mV. The system automatically sets the electromyographic energy threshold to... Set the audio energy threshold to During the formal examination, when the electromyographic energy and audio energy reached 0.55mV and 0.72mV respectively, they both exceeded the adaptively adjusted thresholds and met the triggering conditions.

[0065] Through the above-described scheme, this application achieves automatic identification of the steady-state window period of vocalization and precise control of ultrasound triggering. Compared with the traditional manual button triggering scheme, the triggering time of this application is entirely determined by the real-time calculation results of physiological signals, eliminating the influence of operator subjectivity, ensuring consistent excitation conditions for each measurement, and significantly improving the repeatability and standardization of shear modulus measurement. Simultaneously, the joint judgment mechanism of dual-modal signals (electromyography and audio) has higher robustness than the single audio signal judgment scheme, effectively eliminating interference from environmental noise and muscle artifacts, and avoiding false triggering.

[0066] Continuing with the aforementioned patient case, after the FPGA sends a trigger pulse signal to the ultrasound front end, the system enters the ultrasound data acquisition phase. The core objective of this phase is to capture the minute displacement field caused by endogenous acoustic vibration within the thyroid tissue with sufficiently high temporal resolution during the acoustic steady-state window, providing the raw data foundation for subsequent shear wave velocity calculation and shear modulus inversion.

[0067] Specifically, initiating a rapid ultrasound imaging sequence based on a trigger pulse signal refers to the ultrasound front-end receiving a trigger pulse. Immediately afterwards, the pre-configured plane wave transmission sequence is initiated. Plane wave imaging is a high-speed ultrasound imaging technique. Its core principle is to use all array elements of the ultrasound probe to simultaneously emit phase-aligned ultrasound waves, forming an approximate plane wavefront in the tissue. Compared with the traditional focused beam line-by-line scanning method, plane wave imaging can obtain echo data of the entire imaging area after a single transmission, with an imaging frame rate of up to thousands of frames per second, meeting the time resolution requirements for tracking the propagation process of high-frequency shear waves.

[0068] Using the aforementioned patient case, assume the ultrasound probe is a linear array probe with 128 elements, an element spacing of 0.3 mm, and a center frequency of 7.5 MHz. Upon receiving the trigger pulse, the ultrasound front-end controls all 128 elements to simultaneously emit short pulses with a width of approximately 0.2 μs and a pulse center frequency of 7.5 MHz. Because the emission times of each element are strictly synchronized and phase-aligned, within a depth range of approximately 20 mm to 60 mm from the probe surface (covering the thyroid region), the ultrasound wavefront is approximately a plane wave, parallel to the probe surface. This plane wave propagates through the tissue, and reflections and scattering occur when it encounters interfaces with different acoustic impedances, generating echo signals. All elements simultaneously receive the echo signals, which are then amplified, filtered, and converted from analog to digital by the analog front-end to obtain 128 channels of ultrasound radio frequency echo signals. , where x is the horizontal coordinate (parallel to the probe surface), z is the axial coordinate (perpendicular to the probe surface), and t is time.

[0069] It should be noted that the ultrasonic radio frequency echo signal It is the original high-frequency oscillation signal, with the same frequency as the center frequency of the transmitted pulse, approximately 7.5 MHz, and the amplitude unit is V (voltage). This signal contains information about the position and motion of scatterers within the tissue. In static tissue, radio frequency signals acquired at adjacent moments are highly similar; however, when shear waves propagate within the tissue, the scatterers undergo minute displacements along with the tissue particles, resulting in a phase difference between radio frequency signals at adjacent moments. By calculating this phase difference, the tissue's displacement field can be inferred.

[0070] Furthermore, the highest frequency of the intrinsic shear wave is set to be greater than twice the target pulse repetition frequency to satisfy the sampling theorem requirements. This is one of the key technical constraints of this application, and its theoretical basis is the Nyquist-Shannon sampling theorem.

[0071] According to the sampling theorem, in order to reconstruct a band-limited signal without distortion, the sampling frequency must be greater than twice the highest frequency of the signal. In this application, the endogenous shear wave is formed by the propagation of mechanical waves generated by vocal cord vibration in the neck tissue, and its frequency characteristics are closely related to the fundamental frequency of vocalization. The fundamental frequency range of human vocalization is typically 80Hz to 300Hz, with an average of approximately 120Hz for men and approximately 220Hz for women. The mechanical waves generated by vocal cord vibration contain not only the fundamental frequency component but also several harmonic components. However, due to the attenuation effect of biological tissue on high-frequency mechanical waves, the shear wave that can actually be detected in the thyroid region is mainly concentrated at the fundamental frequency and its first few harmonics, with the highest frequency typically not exceeding 1000Hz.

[0072] To ensure accurate capture of the time-domain waveform of the shear wave and avoid spectral aliasing, this application requires a specific ultrasonic pulse repetition frequency. satisfy: ;in, This is the highest frequency of the intrinsic shear wave. Considering the safety margin in practical applications, the preferred frequency is... That is, 2000 plane waves are emitted per second, corresponding to an imaging frame rate of 2000 frames per second, with an inter-frame time interval of 0.5ms.

[0073] Using the aforementioned patient case, assuming the patient's fundamental frequency is 220Hz, and considering the first three harmonics (440Hz, 660Hz, 880Hz), the effective frequency components of the intrinsic shear wave are mainly distributed in the range of 220Hz to 880Hz, with the highest frequency being approximately 880Hz. According to the sampling theorem, theoretically... This will satisfy the requirements. This application sets... It provides a safety margin of approximately 14%, ensuring that the time-domain waveform of the shear wave can still be accurately reconstructed even if the fundamental frequency of the sound fluctuates slightly or higher-order harmonics are present.

[0074] It should be noted that the selection of the pulse repetition frequency requires a trade-off between temporal resolution and imaging depth. A higher pulse repetition frequency results in better temporal resolution, but also a shorter time interval between adjacent transmissions. If the interval is too short, the echo from the previous transmission may not have fully returned to the probe before the next transmission begins, leading to temporal aliasing of the echo signal and affecting image quality. For thyroid imaging, the typical imaging depth is 20mm to 60mm, the propagation speed of ultrasound in soft tissue is approximately 1540m / s, and the round-trip time is approximately 26μs to 78μs. At that time, the transmission interval is 500μs, which is much longer than the round-trip time of the echo, so there will be no time aliasing.

[0075] Based on the above scheme, simultaneously acquiring ultrasonic radio frequency echo signals and structural images means that after each plane wave emission, not only are the radio frequency echo signals used for shear wave tracing recorded. Furthermore, beamforming algorithms were used to generate B-mode ultrasound structural images. Structural images reflect the acoustic impedance distribution of the tissue and are used to identify the anatomical boundaries of the thyroid gland, nodule locations, and internal fluid-filled and calcified areas.

[0076] Specifically, the generation process of B-mode ultrasound structural images is as follows: First, delayed superposition beamforming is performed on the 128-channel radio frequency echo signals to compensate for the propagation time difference from different array elements to the imaging point, achieving dynamic focusing; then, envelope detection is performed on the synthesized radio frequency signals to extract amplitude information; finally, logarithmic compression and grayscale mapping are performed to obtain a grayscale image. The grayscale value reflects the echo intensity of the tissue. In this image, necrotic areas (such as cysts) appear as hypoechoic or anechoic, with a grayscale value close to 0; normal thyroid parenchyma appears as moderately echogenic, with a grayscale value of approximately 128 (8-bit grayscale); calcified areas appear as hyperechoic, with a grayscale value close to 255.

[0077] Using the aforementioned patient case, assume that after the trigger pulse is emitted, the system continuously emits 100 plane waves, with a duration of 50ms (corresponding to...). After each transmission, the system simultaneously records the radio frequency echo signal and generates a structural image. The first frame of the structural image shows a hypoechoic nodule approximately 8 mm in diameter in the right lobe of the patient's thyroid gland. The nodule has clear boundaries, uniform internal echoes, and a grayscale value of approximately 80, while the surrounding normal thyroid parenchyma has a grayscale value of approximately 120. This structural image provides anatomical reference for subsequent region of interest selection and mask generation.

[0078] Furthermore, structural images are also used to generate effective tissue masks. The mask is a binary image, where... This indicates that the location is a valid solid tissue, and shear modulus calculation can be performed. This indicates that the location is a liquid dark area, a calcified area, or an area outside the probe, and is not suitable for shear modulus calculation. The mask is generated based on grayscale threshold segmentation: for areas with grayscale values ​​below a preset lower threshold (e.g., 30), it is determined to be a liquid dark area, and M=0 is set; for areas with grayscale values ​​above a preset upper threshold (e.g., 230), it is determined to be a calcified area, and M=0 is set; for areas with grayscale values ​​within the threshold range, it is determined to be valid solid tissue, and M=1 is set.

[0079] Using the aforementioned patient case, we assume that threshold segmentation is performed on the first frame of the structural image, with a lower threshold of 30 and an upper threshold of 230. Since the grayscale value inside the nodule in this patient is approximately 80, and the grayscale value of the surrounding normal thyroid parenchyma is approximately 120, both are within the threshold range. Therefore, the mask value for both the nodule region and the surrounding parenchyma region is M=1, indicating that these regions can be used for shear modulus calculation. If the patient's thyroid gland contains cystic dark areas (grayscale value close to 0) or calcifications (grayscale value close to 255), the mask value for these regions will be set to M=0, and they will be removed in subsequent shear modulus calculations.

[0080] It should be noted that the purpose of introducing the masking mechanism is to avoid artifacts in areas unsuitable for elasticity measurements. In liquid dark regions, since shear waves cannot propagate in liquids, forcibly calculating the shear modulus of this region will yield non-physical results close to zero or negative. In calcified regions, the abrupt change in acoustic impedance leads to strong acoustic reflection and mode conversion, significantly altering the propagation direction and energy distribution of shear waves. The time-of-flight method based on the unidirectional propagation assumption will introduce systematic errors. By using the masking mechanism, these regions are excluded from the calculation range, ensuring that the final output shear modulus map only includes the effective solid tissue region, thus improving the reliability of the results.

[0081] Building upon the aforementioned scheme, to further improve the accuracy of shear wave tracking, this application employs a strategy of continuous multiple plane wave transmissions. After a single trigger pulse, the system does not transmit a plane wave only once, but rather transmits multiple plane waves continuously at the target pulse repetition frequency, forming a time-series radio frequency dataset. Where N is the number of launches, This refers to the time corresponding to the nth launch.

[0082] Using the aforementioned patient case, let's assume that after a single trigger pulse, the system... 100 consecutive plane wave transmissions yield 100 frames of radio frequency data, spanning 50 ms. These 100 frames fully record the dynamic process of endogenous shear waves propagating in thyroid tissue. Since the propagation speed of shear waves is typically 1 m / s to 5 m / s, they can propagate 50 mm to 250 mm within 50 ms, which is much larger than the lateral dimension of the thyroid gland (approximately 40 mm). Therefore, this time window is sufficient to capture the complete process of shear waves propagating from the near field to the far field.

[0083] Through the above-described scheme, this application achieves simultaneous acquisition of ultrasound radiofrequency echo signals and structural images with high temporal resolution within the steady-state acoustic window. Compared to traditional low-frame-rate ultrasound imaging (typically 30 to 100 frames per second), the plane wave high-speed imaging technology employed in this application increases the frame rate to 2000 frames per second, improving the temporal resolution by 20 to 60 times. This accurately captures the high-frequency oscillation characteristics of intrinsic shear waves, providing high-quality raw data for subsequent wave velocity estimation and modulus inversion. Simultaneously, by simultaneously acquiring structural images and generating masks, it ensures that shear modulus calculations are performed only within effective solid tissue regions, avoiding artifacts caused by fluid-filled dark areas and calcified regions, thus improving the reliability and clinical applicability of the measurement results.

[0084] Furthermore, to verify the rationality of the target pulse repetition frequency setting, a combination of theoretical derivation and experimental verification can be used. From a theoretical perspective, based on the propagation characteristics of shear waves in biological soft tissues, the shear wave velocity... With the shear modulus of the tissue and density The relationship between them is satisfied For thyroid tissue, the typical range of shear modulus is 1 kPa to 100 kPa, and the density is approximately 1000 kg / m³, therefore the shear wave velocity ranges from approximately 1 m / s to 10 m / s. Assume the wavelength of the shear wave is... The frequency is Then it satisfies When the shear wave frequency is 220 Hz (corresponding to the fundamental frequency), the wavelength is approximately 4.5 mm to 45 mm; when the frequency is 880 Hz (corresponding to the third harmonic), the wavelength is approximately 1.1 mm to 11 mm. To spatially resolve these wavelengths, the lateral resolution of ultrasound imaging needs to be better than half the minimum wavelength, approximately 0.5 mm, which can be achieved using a linear array probe with a center frequency of 7.5 MHz. Simultaneously, to accurately reconstruct these frequency components in time, the pulse repetition frequency needs to satisfy the sampling theorem, i.e., This application sets This satisfies the requirements of the sampling theorem while also taking into account the safety margin in practical applications.

[0085] From an experimental verification perspective, the impact of different pulse repetition frequencies on shear wave tracking accuracy can be evaluated through phantom experiments. A series of agar phantoms with known shear moduli (ranging from 5 kPa to 50 kPa) were prepared. Sinusoidal excitations of known frequencies (e.g., 200 Hz, 400 Hz, 800 Hz) were applied to the phantom surface using a mechanical vibrator. , and Ultrasonic tracking was performed, and the calculated shear modulus was compared with the actual value of the phantom. Experimental results show that when... At an excitation frequency of 800Hz, the sampling theorem is not satisfied ( The calculated shear modulus error exceeds 30%; when At all test frequencies, the calculated shear modulus error was less than 10%; While accuracy is slightly improved, the error improvement is not significant (less than 5%), and the higher pulse repetition frequency increases the data transmission and processing burden on the system. Therefore, considering both accuracy and efficiency, It is a reasonable compromise.

[0086] Using the aforementioned patient case, after acquiring 100 frames of radiofrequency data and structural images, the system obtained a complete spatiotemporal dataset, laying the data foundation for subsequent shear wave velocity calculation and shear modulus inversion. This dataset not only contains dynamic information on shear wave propagation but also provides anatomical references through structural images and masks, enabling subsequent calculations to focus on effective solid tissue areas and avoiding interference from fluid-filled dark areas and calcified regions.

[0087] Continuing with the aforementioned patient case, after simultaneously acquiring ultrasound radiofrequency echo signals and structural images, the system enters the tissue mask generation stage. The core objective of this stage is to identify and eliminate fluid regions unsuitable for elasticity measurement based on the structural images, ensuring that subsequent shear modulus calculations are performed only within valid solid tissue regions, thus avoiding non-physical artifacts.

[0088] Specifically, generating a tissue mask for the target region based on the structural image to remove fluid-containing regions includes: extracting grayscale information from the structural image.

[0089] In this step, the structural image A grayscale image is obtained by beamforming, envelope detection, logarithmic compression, and grayscale mapping of ultrasound radio frequency echo signals. The grayscale value of each pixel in the image reflects the ultrasound echo intensity of the tissue at that location, and its value range is usually from 0 to 255 (8-bit grayscale). Different types of tissue structures exhibit different grayscale characteristics in ultrasound images: fluid-filled dark areas (such as cysts and vascular cavities) exhibit low echo or no echo characteristics due to the large difference in acoustic impedance compared with surrounding tissues and the absence of internal scatterers, with grayscale values ​​usually below 30; normal thyroid parenchyma exhibits medium echo characteristics due to the presence of uniformly distributed scatterers within it, with grayscale values ​​usually between 80 and 180; calcified areas exhibit strong echo characteristics due to extremely high acoustic impedance leading to strong reflection, with grayscale values ​​usually above 230.

[0090] Using the aforementioned patient case, let's assume that in the first frame of the structural image, the grayscale value of the hypoechoic nodule region in the right lobe of the thyroid gland is approximately 80, while the grayscale value of the surrounding normal thyroid parenchyma is approximately 120. If the patient's thyroid gland contains a cystic hypoechoic area with a diameter of approximately 3 mm, the grayscale value of this area will be close to 5; if a calcification lesion with a diameter of approximately 1 mm is present, the grayscale value of this area will be close to 250. The system first performs a pixel-by-pixel scan of the entire structural image, extracting the location of each pixel. Corresponding grayscale value This forms a grayscale information matrix.

[0091] It should be noted that grayscale information extraction is the foundation for subsequent thresholding. Since the grayscale distribution of ultrasound images is influenced by various factors, including the acoustic properties of the tissue, imaging depth, and gain settings, the selection of the grayscale threshold needs to be adjusted according to specific imaging parameters and tissue type in practical applications. To improve the robustness of thresholding, preprocessing can be performed after grayscale information extraction, such as applying mean-mode filtering to remove salt-and-pepper noise, or applying morphological opening operations to remove isolated small bright spots.

[0092] Furthermore, the structural image is binarized based on a preset grayscale threshold, and pure liquid dark areas below the grayscale threshold are identified as invalid areas, and a tissue mask is generated to characterize the effective physical tissue.

[0093] In this step, binarization refers to processing based on a preset grayscale threshold. This classifies each pixel in the structured image into a valid or invalid region. The specific determination rule is: if the gray value of a pixel... If the pixel is identified as a liquid dark area and thus an invalid region, it will be set in the tissue mask. ;like If the pixel is determined to be substantial tissue and belongs to the valid area, then a tissue mask is set. .

[0094] Gray threshold The selection of the grayscale threshold is crucial in this step. According to the physical principles of ultrasound imaging, the grayscale value of a fluid-filled hypoechoic area is typically significantly lower than that of solid tissue. In clinical practice, the grayscale value of a fluid-filled hypoechoic area is usually below 30, while the grayscale value of normal thyroid parenchyma is usually above 60. To ensure complete removal of the fluid-filled hypoechoic area while avoiding misidentification of hypoechoic solid tissue as a fluid-filled hypoechoic area, this application uses a specific grayscale threshold. The value is set between 30 and 50, with 40 being the preferred value. This threshold setting is based on statistical analysis of a large number of clinical ultrasound images and can accurately distinguish between hypoechoic areas and solid tissue in most cases.

[0095] Using the aforementioned patient case, let's assume a grayscale threshold. The value is set to 40. For the hypoechoic nodule region in the right lobe of the thyroid gland, the gray value is approximately 80, satisfying 80 > 40, therefore this region is identified as effective solid tissue, and the mask value M = 1; for the surrounding normal thyroid parenchyma, the gray value is approximately 120, also satisfying 120 > 40, and the mask value M = 1; for the cystic dark area with a diameter of approximately 3 mm, the gray value is approximately 5, which does not satisfy 5 > 40, therefore this region is identified as a fluid-filled dark area, and the mask value M = 0. After binarization processing, the generated tissue mask... It is a binary matrix with the same size as the structure image, where the region M=1 corresponds to the effective solid tissue and the region M=0 corresponds to the liquid dark area that needs to be removed.

[0096] It is important to note that the physical basis for introducing the tissue masking mechanism lies in the propagation characteristics of shear waves. A shear wave is a transverse wave whose propagation depends on the shear stiffness of the medium. In solid or viscoelastic media (such as biological soft tissue), shear waves can propagate effectively, and their propagation speed is positively correlated with the shear modulus of the medium. However, in pure liquids, because the liquid cannot withstand shear stress, shear waves cannot propagate, or in other words, the shear modulus is zero. Therefore, forcibly calculating the shear modulus in a liquid-like dark region will yield non-physical results close to zero or negative. This not only has no clinical significance but also produces significant artifacts in the shear modulus mapping, interfering with the doctor's interpretation of the elastic properties of the surrounding solid tissue.

[0097] Furthermore, to improve the accuracy of mask generation, morphological post-processing can be performed after binarization. Morphological post-processing includes two steps: First, morphological closing operations (dilation followed by erosion) are applied to fill small holes inside the mask, which are often misjudged due to image noise or local grayscale fluctuations. Second, morphological opening operations (erosion followed by dilation) are applied to remove small protrusions and isolated small regions at the mask edges; these regions are usually too small to be statistically significant. The structuring elements for morphological operations are typically chosen to be circular or square, with dimensions ranging from 3×3 pixels to 5×5 pixels.

[0098] Using the aforementioned patient case, suppose that after initial binarization, due to local image noise, several small holes with an area of ​​approximately 1 pixel appear inside the nodule region (mask value M=0). After applying morphological closing operations, these small holes are filled, and the mask value is corrected to M=1, making the mask of the nodule region more continuous and complete. Simultaneously, if there are several isolated small regions with an area of ​​less than 5 pixels at the edge of the mask, these small regions are removed after applying morphological opening operations, avoiding the introduction of unstable edge effects in subsequent calculations.

[0099] Building upon the aforementioned approach, to further improve mask accuracy, ultrasound Doppler blood flow information can be incorporated for auxiliary judgment. Due to blood flow within the vascular lumen, a clear blood flow signal is displayed in color Doppler imaging, while cystic dark areas lack this signal. By overlaying and analyzing color Doppler images with grayscale structural images, the distinction between vascular lumens and cystic dark areas can be made more accurate: areas with grayscale values ​​below a threshold and exhibiting blood flow signals are identified as vascular lumens, and the mask value is set to M=0; areas with grayscale values ​​below the threshold and lacking blood flow signals are identified as cystic dark areas, and the mask value is also set to M=0. This multimodal information fusion strategy further enhances the reliability of mask generation, preventing misidentification of blood vessels as solid tissue.

[0100] Using the aforementioned patient case, suppose there is a thyroid artery branch with a diameter of approximately 2 mm near a nodule in the right lobe of the thyroid gland. This vessel appears as a hypoechoic tubular structure in the grayscale image, with a grayscale value of approximately 15, below the threshold of 40. If judged solely based on the grayscale threshold, this vessel would be misidentified as a cystic dark area, and the mask value would be set to M=0. However, by overlaying a color Doppler image, a clear red blood flow signal was found in this area (indicating blood flowing towards the probe), thus confirming that this area is a vascular lumen, and the mask value was correctly set to M=0. This multimodal fusion strategy ensures that the mask can remove both cystic dark areas and vascular lumens, allowing subsequent shear modulus calculations to be performed only in pure solid tissue regions.

[0101] It should be noted that tissue mask generation is not only used to eliminate liquid dark areas, but also to eliminate other areas unsuitable for elasticity measurements. For example, strong echo regions with gray values ​​exceeding the upper threshold (e.g., 230) typically correspond to calcifications or gas artifacts. These regions experience strong acoustic reflections and mode transitions due to abrupt changes in acoustic impedance, significantly altering the propagation direction and energy distribution of shear waves. Time-of-flight methods based on the unidirectional propagation assumption will introduce systematic errors. Therefore, an upper threshold judgment can be added during mask generation: if... (For example If so, then set it the same way. This area is excluded from the calculation.

[0102] Using the previous patient case, suppose the patient has a calcification lesion approximately 1 mm in diameter within their thyroid gland. This area appears as a strongly echogenic point in the structural image, with a grayscale value of approximately 250, higher than the upper threshold of 230. Through dual thresholding (lower threshold 40, upper threshold 230), the calcification is correctly identified and set to M=0 in the mask, avoiding artifacts generated during shear modulus calculation in this area. The final generated tissue mask... It only includes solid tissue areas with gray values ​​between 40 and 230, which are neither liquid dark areas nor calcified areas, and are suitable for reliable elasticity measurements.

[0103] Furthermore, to verify the accuracy of the generated mask, it can be compared with a region of interest (ROI) manually drawn by the physician. In clinical practice, physicians typically manually delineate the solid tissue areas requiring elasticity measurements on ultrasound images, excluding cystic areas, blood vessels, and calcifications. The overlap between the automatically generated tissue mask and the physician-delineated ROI is analyzed, and the Dice similarity coefficient is calculated.

[0104] ;

[0105] in, For automatically generated masks, The mask manually drawn by the doctor. Indicates the area of ​​the region. The intersection is indicated. The Dice coefficient ranges from 0 to 1, with a value closer to 1 indicating a higher degree of overlap between the automatic and manual masks. In validation with a large amount of clinical data, the Dice coefficient of the automatic mask generation method in this application is typically higher than 0.85, indicating a high degree of consistency between the automatic mask and the ROI drawn manually by the doctor, which can meet the accuracy requirements of clinical applications.

[0106] Through the above-described scheme, this application achieves automatic tissue mask generation based on structural images, accurately identifying and eliminating fluid-filled dark areas, vascular cavities, and calcified regions, ensuring that subsequent shear modulus calculations are performed only within the effective solid tissue area. Compared to traditional full-area calculation methods, the masking mechanism of this application significantly reduces artifacts caused by fluid-filled dark areas and calcified regions, improving the accuracy and clinical interpretability of shear modulus measurement. Simultaneously, the automatic mask generation method avoids the subjectivity and time-consuming nature of doctors manually delineating ROIs, improving examination efficiency and standardization, making this technology more suitable for large-scale clinical application.

[0107] It should be noted that tissue mask generation is based on static segmentation using grayscale information from structural images. Its accuracy depends on the quality of the structural image and the appropriate setting of the grayscale threshold. In practical applications, if the structural image contains severe noise or artifacts, or if the grayscale distribution of the tissue differs significantly from typical cases, misjudgments may occur during mask generation. To improve robustness, a manual review step can be added after mask generation: the system overlays the automatically generated mask onto the structural image and displays it to the doctor, who can then fine-tune the mask based on clinical experience. For example, the doctor can manually add hypoechoic solid tissue that was mistakenly identified as a fluid-filled dark area, or manually delete small cysts that were mistakenly identified as solid tissue. This hybrid mode of automatic generation + manual review ensures both efficiency and accuracy, representing best practice in clinical applications.

[0108] Using the aforementioned patient case, after automatically generating the tissue mask, the system displays the mask on the structural image as a semi-transparent color overlay: the effective solid tissue area (mask value M=1) is displayed as a green semi-transparent cover, while the invalid area (mask value M=0) is not covered. During review, the doctor discovered a small hypoechoic area (approximately 2 square millimeters in area, with a grayscale value of approximately 35) at the edge of the nodule, which was misidentified by the automatic mask as a liquid dark area (mask value M=0). However, based on the doctor's clinical experience, this area was actually the hypoechoic portion of the nodule and should be included in the elasticity measurement range. The doctor clicked on this area with the mouse, and the system corrected the mask value of this area to M=1. After manual review and fine-tuning, the final tissue mask more accurately reflected the actual distribution of solid tissue, providing a reliable spatial constraint for subsequent shear modulus calculations.

[0109] Continuing with the aforementioned patient cases, after completing the tissue masking... After the generation of the signal, the system enters the shear wave propagation information extraction and shear modulus inversion stage. The core objective of this stage is to recover the minute displacement field generated by acoustic excitation within the thyroid tissue from the continuously acquired ultrasound radio frequency echo signals, and after eliminating reflected wave interference, accurately estimate the shear wave propagation velocity, thereby obtaining the local shear modulus of the target area.

[0110] Based on the ultrasonic radio frequency echo signal, the wave field displacement information of the forward propagation is extracted, and the wave peak arrival time difference is calculated.

[0111] The reason for first extracting the forward-propagating wavefield displacement information is that thyroid tissue is not an ideal, infinitely homogeneous medium. In actual human tissue, when shear waves propagate to the boundaries of nodules, capsules, blood vessels, or near local calcifications, reflection, scattering, and local mode transitions occur. If the wave velocity is directly calculated using the simple time-of-flight method for the mixed wavefield, it is easy to misjudge the reflected waves as forward-propagating waves, leading to an excessively long transit time. The estimation error is amplified nonlinearly by the square in the calculation of shear modulus. Therefore, this application does not directly estimate the wave velocity on the original displacement field, but first recovers the displacement from the ultrasonic radio frequency echo signal, and then further separates the forward propagating wave field component, thereby improving the stability and physical consistency of the subsequent modulus inversion.

[0112] Extracting forward-propagating wavefield displacement information based on ultrasonic radio frequency echo signals includes: performing one-dimensional cross-correlation calculation on continuously acquired ultrasonic radio frequency echo signals to obtain the axial displacement information of tissue particles in the target area; using a spatiotemporal direction filter to filter the axial displacement information, separating and extracting the forward wavefield displacement information propagating in a single direction to eliminate interference from boundary reflected waves.

[0113] In this step, the continuously acquired ultrasonic radio frequency echo signals can be represented as a time series. Where x represents the horizontal coordinate and z represents the axial coordinate. This indicates the acquisition time corresponding to the nth frame echo. In the aforementioned patient case, the system uses... 100 frames of radio frequency data are acquired continuously, so the time interval between two adjacent frames is 0.5ms, and the total time span is 50ms.

[0114] One-dimensional cross-correlation is a process of similarity matching between axial radio frequency signals at adjacent time points from the same spatial location to estimate the minute displacement of the scatterer along the ultrasonic propagation direction. Since ultrasonic radio frequency signals are most sensitive to axial displacement, displacement estimation along the axial direction is preferred. For position... Location, adjacent frames and The local radio frequency window function between them can be used to determine the optimal time delay through the cross-correlation function. Then convert the time delay into axial displacement. Or more generally written as Its basic relationship can be expressed as: ;

[0115] in, The speed at which ultrasound travels through soft tissue is typically taken as 1540 m / s; the coefficient 1 / 2 indicates that the ultrasound echo has traveled back and forth. The time delay is given by the peak value of the local cross-correlation between adjacent frame signals, and its unit is seconds. Therefore, the displacement... The unit is m, which is usually in the micrometer range in actual thyroid tissue.

[0116] In engineering implementation, cross-correlation calculations can be performed using axial windows of several ultrasonic wave lengths near each pixel to balance displacement estimation accuracy and spatial resolution. If the window is too short, the cross-correlation peaks are not obvious and are easily affected by noise; if the window is too long, it smooths out local displacement gradients. For a probe with a center frequency of 7.5 MHz, the wavelength in soft tissue is approximately 0.205 mm, so an axial correlation window of approximately 0.5 mm to 2 mm can be selected. Since the focus of this application is not on the specific form of the cross-correlation window, its details will not be elaborated here.

[0117] Using the aforementioned patient case, let's assume a nodule at a certain depth in the right lobe of the thyroid gland. At this point, the system performs one-dimensional cross-correlation analysis on two consecutive frames of radio frequency signals to obtain the optimal time delay. The corresponding axial displacement is approximately: ;

[0118] This order of magnitude is consistent with the minute vibrational displacement of biological soft tissue under acoustic excitation, which also indicates that it is reasonable to use the displacement tracking method for shear wave analysis under the previously established linear elastic small strain assumption.

[0119] It should be noted that the original axial displacement information obtained from one-dimensional cross-correlation includes a mixture of forward propagating waves, boundary reflected waves, and local scattering. Especially near the nodule edge or thyroid capsule, if a clear echo interface exists, the reflected waves will superimpose with the main propagating wave in the spatiotemporal domain, resulting in phenomena such as double peaks, tails, or local inversion of the wave crest trajectory. Based on this, this application further employs a spatiotemporal direction filter for separation.

[0120] The basic principle of the spatiotemporal directional filter is that: or In the joint spacetime domain, waves propagating in different directions have different slopes and spectral orientations. Taking a shear wave propagating in the transverse x-direction as an example, if the wave field can be approximated as... In the spatial and temporal domain, the forward propagation wave corresponds to a set of energy trajectories arranged along the positive slope direction; the reverse propagation wave corresponds to energy trajectories along the opposite slope direction. Therefore, a directional filter that retains only the forward propagation slope component can be constructed to suppress the reverse propagation energy, thereby obtaining the forward wavefield displacement information. .

[0121] In this application, directional filtering does not alter the physical source of the displacement field; instead, it extracts the component consistent with the target propagation direction from the mixed displacement field. In other words, what is obtained after directional filtering is not a new displacement, but a forward propagation wave field separated from the original axial displacement information, which is more suitable for analysis using the time-of-flight method. The technical advantages of this approach are twofold: firstly, it reduces the interference from reflected waves caused by nodule edges and envelope boundaries; secondly, it makes the wave crest trajectory more continuous and monotonic on the spatiotemporal map, facilitating subsequent stable extraction of arrival times.

[0122] Using the previous patient case, let's assume an analysis line, approximately 12mm long, is selected across the center of the right lobe nodule. After applying spatiotemporal direction filtering to the axial displacement information along this analysis line, we can observe that... to The bidirectional intersecting fringes within the interval are significantly reduced, retaining only the dominant wavefield component propagating from left to right. At this point, the wave crest trajectory in the forward wavefield displacement information is clearer, indicating that directional filtering effectively suppresses aliasing caused by reflected waves.

[0123] The calculation of the wave crest arrival time difference includes: tracking the shear wave crest in the forward wavefield displacement information based on the time-of-flight method; and calculating the transit time difference when the shear wave crest passes through two adjacent detection points within a preset lateral detection interval, which is used as the wave crest arrival time difference.

[0124] In this step, the essence of the time-of-flight method is to estimate the propagation velocity using the time difference of arrival of the same wave crest at different lateral positions. Since the forward wavefield displacement information has already been extracted in the previous steps, it can be assumed that the wave crest to be tracked mainly propagates along a single lateral direction. For two adjacent lateral detection points at the same depth z or within the same small region... and Let its lateral detection spacing be... If the corresponding peak arrival times are respectively and Then the time difference of the wave crest arrival is: .

[0125] Correspondingly, shear wave velocity It can be represented as: .

[0126] in, The unit is m. The unit is s, therefore The unit is m / s. In thyroid tissue, the shear wave velocity is typically in the range of 1 m / s to 10 m / s. It should not be set too small, otherwise it will be limited. If the time difference between the wave crest crossings is too short, the time quantization error will increase significantly. It should not be set too large, otherwise it will reduce spatial resolution and violate the assumption of local uniform propagation when crossing excessively large structural boundaries. For thyroid ultrasound images, the lateral detection spacing can be selected according to the probe element spacing, reconstructed pixel size, and target tissue size, and is usually 0.5 mm to 3 mm, preferably about 1 mm to 2 mm.

[0127] Furthermore, the shear modulus is highly sensitive to time difference measurement errors. If the shear modulus is defined as... Then its partial derivative with respect to the time difference is: ;

[0128] This formula shows that when When the error is small, the time measurement error will be as follows: The magnitude is amplified. Therefore, this application employs the following in the preliminary steps: The goal of this rapid imaging process, which involves first extracting the forward wavefield and then tracing the peaks, is to minimize the impact of wave size. The adverse effects of extraction errors on the final modulus result.

[0129] Using the aforementioned patient case, let's assume that two adjacent lateral detection points are selected near the depth of the nodule's center, with a lateral detection distance of denoted as [missing information]. After directional filtering, at the detection point on the left... The arrival time of a certain shear wave peak was detected at [location]. At the right detection point The same wave peak was detected at the time of arrival. Then the time difference of the wave crest arrival is: Therefore, the shear wave velocity in this local region is: This velocity is within the typical range acceptable for thyroid soft tissue, indicating that the current peak tracking results have reasonable physical significance.

[0130] It should be noted that the so-called wave crest is not limited to the global maximum displacement peak; it can also be a local main peak with a clear and continuous propagation trajectory within a preset time window. The key is that what is being tracked is the corresponding characteristics of the same propagating wavefront at different detection points. To improve stability, cross-correlation, peak search, or local time delay matching methods can be combined between adjacent detection points to assist in confirming the correspondence, but the goal is always to obtain reliable... .

[0131] Furthermore, in two-dimensional space, if certain local areas have complex boundaries that cause slight distortions in the peak shape, this can be addressed using tissue masks. Within the effective region, estimation is performed point-by-point or block-by-block according to local neighborhoods. That is, it is not required that a single wave velocity be used for the entire target region, but rather different locations are allowed to use the same wave velocity. Local wave velocities are calculated based on the arrival time differences of local wave peaks within their neighborhoods, thus yielding a spatially distributed shear modulus map. This is particularly important in clinical scenarios where there are elastic differences between nodules and surrounding parenchyma.

[0132] Combining the tissue mask, the time difference of arrival of the wave crest, and the preset tissue density, the local shear modulus of the target area is calculated, including: taking the ratio of the lateral detection spacing to the time difference of arrival of the wave crest as the shear wave velocity; multiplying the square of the shear wave velocity by the preset tissue density to obtain the initial shear modulus; mapping the initial shear modulus to the tissue mask point by point, setting the shear modulus of the invalid area to zero, and obtaining the final local shear modulus of the tissue.

[0133] In this step, the calculation of the local shear modulus is based on the propagation relationship of shear waves in an incompressible, isotropic, linear elastic medium. According to the equation of motion: ;

[0134] The following relationship can be obtained between shear wave velocity and shear modulus: From this, we can deduce that: ;

[0135] in, The local shear modulus of the tissue is expressed in Pa. Tissue density, unit: ; The shear wave velocity is expressed in m / s. For biological soft tissues such as the thyroid gland, the density is typically close to that of water; therefore, this application presupposes a tissue density. Desirable This value was chosen because thyroid tissue has a high water content, and it is approximated as such in engineering implementation. It conforms to physiological realities and facilitates standardized calculations. If higher precision is required in specific applications, other density values ​​that are close to reality can be preset according to tissue type, without changing the basic calculation principle of this application.

[0136] Combining the aforementioned time-of-flight method, it can be further written as: ;

[0137] in, For tissue masking; when When, it indicates that the point belongs to the effective substantive organizational area; when When the value is displayed, it indicates that the point is a liquid dark area, a calcified area, a vascular cavity, or other invalid area. This represents the local peak transit time difference extracted based on the forward wave field. After multiplying the mask by the initial shear modulus point by point, the modulus value of the invalid region is automatically set to zero, thus ensuring that the final output modulus map only reflects the elastic information of the effective substantial tissue region.

[0138] Using the aforementioned patient case, based on the local measurement results described above, if there are [missing information] near the center of the nodule The initial shear modulus is: If the tissue mask value corresponding to this location is The final local shear modulus of the tissue remains 4.0 kPa. Conversely, if a location is in a cystic dark area or calcification, the corresponding mask value is... Therefore, even if a certain value is obtained from the original calculation, it will be set to zero after mapping, that is: The technical advantage of this approach is that it avoids generating spurious elasticity results in regions where the shear wave propagation assumption is not physically satisfied, thereby enhancing the clinical reliability of the results.

[0139] To illustrate the necessity of this mask mapping step, it can be understood in conjunction with boundary conditions. When When this occurs, it means that the wave can hardly propagate in that region, then there is... This aligns with the physical intuition that a pure liquid region has virtually no shear resistance; however, when the local tissue is extremely stiff, Decrease Increase, corresponding Increase. Therefore, it can be seen that adopting... It has a clear physical basis, and combining it with a mask can further ensure that the formula is used only in areas that meet the applicable conditions.

[0140] Using the aforementioned patient case, if point-by-point calculations are performed on the central region of the nodule and the surrounding solid tissue, a two-dimensional shear modulus distribution map can be obtained. For example, the arrival time difference of the peak in the normal solid region surrounding the nodule is approximately 1.06 ms. Under these conditions, the corresponding shear wave velocity is approximately 1.42 m / s, and the shear modulus is approximately: .

[0141] The calculated value for the central region of the nodule is approximately 4.0 kPa. This indicates that the local shear modulus of the nodule region is higher than that of the surrounding normal tissue, suggesting increased local tissue stiffness. For clinicians, this modulus difference can serve as important supplementary information for screening and assessment in conjunction with ultrasound morphological features.

[0142] It should be noted that the focus of this application in this step is not simply on using formulas. The key lies in the combined process of forward wavefield extraction, peak arrival time difference calculation, and tissue mask mapping, which allows the formula to adapt to complex biological tissue environments. Specifically: directional filtering solves the problem of the unidirectional propagation assumption failing due to boundary reflection; tissue masking eliminates liquid dark areas and calcified regions, avoiding erroneous inversion in areas that do not meet the shear wave propagation conditions; and high... Peak tracking under certain conditions reduces Measurement errors amplify the modulus results. Therefore, this application not only theoretically satisfies the elastic mechanical relationship, but also possesses engineering feasibility and clinical application value.

[0143] Using the above scheme, the system can reliably extract the forward-propagating shear wave displacement information from the ultrasound radiofrequency echo signal acquired under acoustic excitation, and calculate the spatially distributed local shear modulus of the tissue based on the tissue mask. Compared with the scheme of directly estimating elastic parameters on the original mixed wave field, this application can effectively suppress artifacts caused by reflected waves and invalid regions, improve the accuracy, stability and interpretability of modulus mapping results, and provide a reliable quantitative basis for non-invasive combined screening of thyroid nodules.

[0144] Continuing with the aforementioned patient case, the local shear modulus of the target tissue area was reduced. After calculation, the system enters the elasticity distribution visualization and clinical interpretation stage. The core objective of this stage is to transform the spatially distributed shear modulus data into an intuitive two-dimensional elasticity distribution map, and to display it through pseudo-color mapping and structural image fusion, helping clinicians quickly identify lesion areas with abnormal hardness and providing quantitative reference for the benign and malignant assessment of thyroid nodules.

[0145] Specifically, generating a two-dimensional elasticity distribution map based on the local shear modulus of tissue at various points within the target region refers to using the shear modulus calculated in the aforementioned steps. According to its spatial coordinates The elements are arranged to form a two-dimensional matrix that matches the size and resolution of the ultrasound structural image. The value of each pixel in this matrix corresponds to the shear modulus at that location, expressed in Pa or kPa. Since the preceding steps have already been performed using a tissue mask... The modulus of the invalid regions is set to zero. Therefore, only the effective solid tissue regions have non-zero modulus values ​​in the two-dimensional elasticity distribution map, while the modulus values ​​of invalid regions such as fluid dark areas, calcified areas, and vascular cavities are zero.

[0146] Using the aforementioned patient case, assuming the imaging area in the right lobe of the thyroid gland has a lateral range of approximately 40 mm and an axial range of approximately 30 mm, and the pixel resolution of the ultrasound image is 0.1 mm, then the size of the two-dimensional elastography map is: Pixels. In this matrix, the shear modulus of the central region of the nodule (approximately 8 mm in diameter) is approximately 4.0 kPa, the shear modulus of the surrounding normal thyroid parenchyma is approximately 2.0 kPa, and the shear modulus of the cystic dark area (approximately 3 mm in diameter) and calcifications (approximately 1 mm in diameter) is 0 kPa. This spatial distribution clearly shows the elastic difference between the nodule region and the surrounding parenchyma, as well as the spatial location of the ineffective region.

[0147] It should be noted that generating a two-dimensional elasticity distribution map does not change the physical meaning of the shear modulus. Instead, it organizes the modulus values ​​at discrete spatial sampling points into a continuous two-dimensional image, facilitating subsequent visualization and clinical interpretation. In practical implementation, if some pixel locations are not directly involved in wave velocity calculation (e.g., located between adjacent detection points), spatial interpolation methods (such as bilinear interpolation or cubic spline interpolation) can be used to fill in the modulus values ​​at these locations to obtain a smooth and continuous elasticity distribution map. The rationale for interpolation lies in the fact that within locally homogeneous tissue regions, the elastic properties of adjacent pixels are usually continuous; therefore, the interpolation result can reasonably reflect the spatial distribution trend of the local elastic field.

[0148] Furthermore, the two-dimensional elastic distribution map is subjected to pseudo-color mapping and fused with the structural image to assist in the identification of lesion areas with abnormal hardness. This includes: mapping the shear modulus values ​​in the two-dimensional elastic distribution map to the corresponding colors according to a preset color mapping table to generate a pseudo-color elastic image; spatially registering and overlaying the pseudo-color elastic image with the structural image to generate a fused display image, where the structural image provides anatomical reference and the pseudo-color elastic image provides elastic information. The superposition of the two can simultaneously display the morphological characteristics and hardness distribution of the tissue.

[0149] In this step, pseudo-color mapping is the process of mapping the numerical range of shear modulus to a color space. Since the human eye's perception of color is far stronger than its perception of grayscale, pseudo-color mapping can significantly improve the readability and contrast of elasticity information. Commonly used color mapping tables include heatmaps (from blue to red), rainbow maps (from purple to red), or custom clinical color schemes. In this application, a heatmap mapping scheme from blue (low modulus) to red (high modulus) is preferred, as this scheme aligns with clinicians' intuitive understanding: blue indicates softer tissue, and red indicates firmer tissue.

[0150] Specifically, a color mapping table can be defined as a piecewise linear function or lookup table that uniformly divides the numerical range of shear modulus (e.g., 0 kPa to 10 kPa) into several color intervals, each interval corresponding to a different color. For example, 0 kPa to 1 kPa can be mapped to dark blue, 1 kPa to 2 kPa to light blue, 2 kPa to 3 kPa to green, 3 kPa to 4 kPa to yellow, 4 kPa to 5 kPa to orange, and above 5 kPa to red. This piecewise mapping scheme can highlight the boundaries of different hardness intervals while maintaining color continuity, making it easier for doctors to quickly identify areas of abnormal hardness.

[0151] It should be noted that the dynamic range (i.e., minimum and maximum values) of the color map needs to be adaptively adjusted based on the actual measured shear modulus distribution. If the dynamic range is set too narrow, it will cause saturation in high-modulus areas (displayed entirely in red), making it impossible to distinguish different degrees of hardening; if the dynamic range is set too wide, the color differences in low-modulus areas will be indistinct, reducing contrast. In this application, the dynamic range of the color map can be automatically determined based on the statistical distribution of the shear modulus within the target area (e.g., from the 5th percentile to the 95th percentile), which can both cover the vast majority of valid data and avoid the impact of extreme values ​​on the overall display effect.

[0152] After generating the pseudocolor elastography image, it is fused with the structural image for display. The purpose of fusion display is to overlay elastic and anatomical information in the same image, allowing doctors to simultaneously observe the morphological features of the tissue (such as nodule boundaries, internal echoes, and the location of calcifications) and hardness distribution (such as the hardness of the nodule center, the hardness of the surrounding parenchyma, and the hardness gradient). The fusion display is achieved through two key steps: first, spatial registration, ensuring that the pseudocolor elastography image and the structural image are strictly aligned in spatial coordinates; second, transparency overlay, where the pseudocolor elastography image is semi-transparently overlaid on the structural image, allowing both information to be visible simultaneously.

[0153] Spatial registration is a prerequisite for fused display. Since both the pseudo-color elastic image and the structural image originate from the same ultrasound acquisition process and use the same imaging coordinate system, they are naturally aligned in space, eliminating the need for complex registration algorithms. However, in practical implementations, if the resolution of the elastic image differs from that of the structural image (for example, if the elastic image uses coarser spatial sampling to improve computational efficiency), interpolation or resampling is required to adjust the elastic image to the same resolution as the structural image to ensure pixel-by-pixel correspondence.

[0154] Transparency overlay is the core of the blended display. In this step, each pixel of the pseudo-color elastic image is weighted and blended with the corresponding pixel of the structured image to generate the blended display image. The blending formula can be expressed as:

[0155] ;

[0156] in, To blend and display the pixel values ​​of the image, These are the pixel values ​​of the pseudo-color elastic image. The pixel values ​​of the structured image. This is the transparency coefficient, with a value ranging from 0 to 1. When... When the image is fully fused, the structural image is displayed; when... When the image is fully displayed, the fused image completely shows the pseudo-color elastic image; when When the median value (e.g., 0.5) is taken, the fused image displays information from both images simultaneously, with equal weights.

[0157] In this application, the transparency factor This can be adjusted according to clinical needs. In the initial screening phase, doctors may focus more on flexibility information; in this case, [the following can be added]: Setting it to a higher value (e.g., 0.7) makes the pseudocolor elasticity image more prominent; for detailed diagnostic stages, doctors may need to observe both morphology and elasticity simultaneously, in which case... Set it to a middle value (e.g., 0.5) to balance the information displayed from both sources. The system offers an interactive slider that allows doctors to adjust it in real time. The display effect is optimized based on the characteristics of specific cases.

[0158] It's important to note that the fused image display is not only for static observation but also for dynamic interactive analysis. For example, doctors can click or select a region of interest on the fused image, and the system will display statistical information such as the mean shear modulus, standard deviation, maximum and minimum values ​​of that region in real time. This quantitative analysis function helps doctors more accurately assess the hardness characteristics of nodules, such as determining whether there is uneven hardness (indicating complex internal structure) or the hardness ratio of the nodule to the surrounding parenchyma (indicating malignancy risk).

[0159] Furthermore, to aid in identifying lesions with abnormal hardness, the system can automatically mark the boundaries of these areas on the fused image. The definition of abnormal hardness areas can be based on statistical or clinically experienced thresholds. For example, an area with a shear modulus higher than the average of the surrounding normal parenchyma plus two standard deviations can be defined as an area with abnormal hardness, i.e.: ;

[0160] in, The average shear modulus of the surrounding normal material. The standard deviation is denoted as . Pixels that meet this condition are marked as having abnormal hardness. The system outlines the boundaries of these areas with red contour lines on the fused image and labels them with the words "abnormal hardness" to remind doctors to pay close attention.

[0161] It should be noted that the criteria for determining abnormal hardness can be adjusted according to clinical needs. For the screening stage, a lower threshold (e.g., the mean plus one standard deviation) can be used to improve sensitivity and avoid missed diagnoses; for the diagnostic stage, a higher threshold (e.g., the mean plus three standard deviations) can be used to improve specificity and reduce false positives. The system provides a variety of preset criteria for doctors to choose from, or allows doctors to customize the threshold according to the characteristics of specific cases.

[0162] Building upon the aforementioned approach, to further enhance the clinical applicability of fused image display, the system can also provide multiple display modes for physicians to switch between. For example, it can offer a structural image mode (displaying only grayscale structural images), an elasticity image mode (displaying only pseudo-color elasticity images), a fused display mode (displaying both images overlaid), and a contrast display mode (displaying both images side-by-side). Physicians can select the appropriate display mode according to different stages of the diagnostic process: in the initial observation stage, the structural image mode is used to understand tissue morphology; in the elasticity assessment stage, the elasticity image mode is used to focus on stiffness distribution; and in the comprehensive interpretation stage, the fused display mode or contrast display mode is used to simultaneously analyze morphological and elasticity information.

[0163] Through the above-described scheme, this application achieves a complete conversion process from shear modulus values ​​to clinically visualized images. By using pseudo-color mapping and fusion display, abstract elastic mechanical parameters are transformed into intuitive color distribution maps, which are then overlaid with anatomical structure images, enabling physicians to simultaneously obtain comprehensive morphological and elastic information from a single image.

[0164] Furthermore, to validate the clinical value of fused imaging, it can be compared with the gold standard of pathology. In clinical practice, for highly suspicious nodules, fine-needle aspiration biopsy or surgical resection is usually performed to obtain a pathological diagnosis. Spatial correspondence analysis between the areas of abnormal hardness marked in the fused imaging and the lesion areas in the pathological sections can assess the accuracy of elastography in lesion localization.

[0165] In summary, the thyroid ultrasound screening method based on endogenous acoustic vibration proposed in this application realizes a complete technical chain from acoustic excitation, steady-state window identification, high-speed ultrasound imaging, shear wave tracking, shear modulus inversion, to fusion display. By jointly judging dual-modal signals, the stability and repeatability of the excitation source are ensured; by using plane wave high-speed imaging and directional filtering, the propagation process of endogenous shear waves is accurately captured; by using tissue masking and time-of-flight method, the local shear modulus of the target area is reliably calculated; and by using pseudo-color mapping and fusion display, elastic information is transformed into intuitive clinical visualization images, providing a novel technical means for the early screening and benign / malignant assessment of thyroid nodules.

[0166] Example 2:

[0167] like Figure 3 As shown, the multimodal imaging combined screening system for preventive physical examinations includes:

[0168] The signal acquisition module is used to simultaneously acquire the electromyographic signals on the surface of the larynx and the audio signals of vocal cord vibration during the subject's spontaneous vocalization process;

[0169] The signal triggering module is used to calculate the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics and instantaneous frequency fluctuation rate of the vocal cord vibration audio signal in real time. When the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics of the vocal cord vibration audio signal both meet the preset energy threshold, and the instantaneous frequency fluctuation rate of the vocal cord vibration audio signal is lower than the preset fluctuation tolerance and lasts for a preset time, a trigger pulse signal is generated.

[0170] The signal acquisition module is used to control the ultrasound probe to emit plane waves at the target pulse repetition frequency based on the trigger pulse signal, and simultaneously acquire ultrasound radio frequency echo signals and structural images.

[0171] The tissue mask generation module is used to generate tissue masks for target regions based on structural images to eliminate fluid-containing areas;

[0172] The wave field displacement processing module is used to extract the forward propagation wave field displacement information based on the ultrasonic radio frequency echo signal and calculate the wave peak arrival time difference.

[0173] The shear modulus calculation module is used to calculate the local shear modulus of the target area by combining the tissue mask, the time difference of arrival of the peak, and the preset tissue density.

[0174] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A multimodal imaging combined screening method for preventive physical examinations, characterized in that, Includes the following steps: Simultaneously acquire surface electromyographic signals of the larynx and vocal cord vibration audio signals during the subject's spontaneous vocalization process; Real-time calculation of the energy characteristics of the electromyographic signal on the surface of the larynx and the energy characteristics and instantaneous frequency fluctuation rate of the vocal cord vibration audio signal; When the energy characteristics of the electromyography signal on the surface of the larynx and the energy characteristics of the vocal cord vibration audio signal both meet the preset energy threshold, and the instantaneous frequency fluctuation rate of the vocal cord vibration audio signal is lower than the preset fluctuation tolerance and lasts for a preset time, a trigger pulse signal is generated. Based on the trigger pulse signal, the ultrasonic probe is controlled to emit a plane wave at the target pulse repetition frequency, and ultrasonic radio frequency echo signal and structural image are acquired simultaneously. A tissue mask for the target region is generated based on the structural image to eliminate fluid-filled areas; Based on the ultrasonic radio frequency echo signal, the wave field displacement information of the forward propagation is extracted, and the wave peak arrival time difference is calculated. The local shear modulus of the target region is calculated by combining the tissue mask, the peak arrival time difference, and the preset tissue density.

2. The multimodal image-based joint screening method for preventive physical examinations according to claim 1, characterized in that, Real-time calculation of the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics and instantaneous frequency fluctuation rate of the vocal cord vibration audio signal, including: Based on a preset sliding time window, the root mean square values ​​of the electromyographic signal on the laryngeal surface and the audio signal of vocal cord vibration within the time window are calculated respectively, and used as the corresponding energy features; The instantaneous frequency of the vocal cord vibration audio signal is extracted, and the rate of change of the instantaneous frequency over time is calculated as the instantaneous frequency fluctuation rate.

3. The multimodal image joint screening method for preventive physical examinations according to claim 2, characterized in that, When the energy characteristics of the electromyography signal on the laryngeal surface and the energy characteristics of the vocal cord vibration audio signal both meet a preset energy threshold, and the instantaneous frequency fluctuation rate of the vocal cord vibration audio signal is lower than a preset fluctuation tolerance for a preset time, a trigger pulse signal is generated, including: The electromyographic signals on the surface of the larynx and the audio signals of vocal cord vibration are acquired synchronously using a unified hardware clock. When the root mean square value is detected to be greater than the corresponding energy threshold, and the absolute value of the instantaneous frequency change rate is less than the duration of the preset fluctuation tolerance exceeding the steady-state holding time threshold, a hardware trigger pulse signal is sent to the ultrasound front end.

4. The multimodal image joint screening method for preventive physical examinations according to claim 1, characterized in that, Based on the trigger pulse signal, controlling the ultrasound probe to emit a plane wave at the target pulse repetition frequency includes: The ultrasonic high-speed imaging sequence is initiated based on the trigger pulse signal; The target pulse repetition frequency is set to be greater than twice the highest frequency of the intrinsic shear wave to satisfy the sampling theorem requirements.

5. The multimodal image-based joint screening method for preventive physical examinations according to claim 1, characterized in that, Generating a tissue mask for the target region based on the structural image to remove fluid-filled areas includes: Extract the grayscale information of the structure image; The structure image is binarized based on a preset grayscale threshold. Pure liquid dark areas below the grayscale threshold are identified as invalid areas, and a tissue mask is generated to characterize the effective physical tissue.

6. The multimodal image-based joint screening method for preventive physical examinations according to claim 1, characterized in that, Based on the ultrasonic radio frequency echo signal, the forward propagation wavefield displacement information is extracted, including: One-dimensional cross-correlation calculation is performed on the continuously acquired ultrasonic radio frequency echo signals to obtain the axial displacement information of tissue particles in the target area; The axial displacement information is filtered by a spatiotemporal direction filter to separate and extract the positive wave field displacement information propagating in a single direction, so as to eliminate the interference of boundary reflected waves.

7. The multimodal image-based joint screening method for preventive physical examinations according to claim 6, characterized in that, Calculating the time difference of arrival of the wave crest includes: Based on the time-of-flight method, the shear wave peak is tracked in the positive wave field displacement information; The transit time difference between two adjacent detection points within a preset lateral detection distance of the shear wave peak is calculated and used as the arrival time difference of the wave peak.

8. The multimodal image-based joint screening method for preventive physical examinations according to claim 7, characterized in that, Combining the tissue mask, the peak arrival time difference, and the preset tissue density, the local shear modulus of the target region is calculated, including: The ratio of the lateral detection spacing to the time difference of the wave peak arrival is taken as the shear wave velocity; The initial shear modulus is obtained by multiplying the square of the shear wave velocity by the preset tissue density. The initial shear modulus is mapped by multiplying the tissue mask point by point, and the shear modulus of the invalid region is set to zero to obtain the final local shear modulus of the tissue.

9. The multimodal image-based joint screening method for preventive physical examinations according to claim 1, characterized in that, After calculating the local shear modulus of the target tissue region, the method further includes: A two-dimensional elasticity distribution map is generated based on the local shear modulus of the tissue at each point within the target area. The two-dimensional elastic distribution map is subjected to pseudo-color mapping and then fused with the structural image for display, in order to help identify lesion areas with abnormal hardness.

10. A multimodal image-based joint screening system for preventive physical examinations, characterized in that: Using a multimodal imaging combined screening method for preventive physical examinations as described in any one of claims 1-9, comprising: The signal acquisition module is used to simultaneously acquire the electromyographic signals on the surface of the larynx and the audio signals of vocal cord vibration during the subject's spontaneous vocalization process; The signal triggering module is used to calculate in real time the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics and instantaneous frequency fluctuation rate of the vocal cord vibration audio signal; when the energy characteristics of the laryngeal surface electromyography signal and the energy characteristics of the vocal cord vibration audio signal both meet the preset energy threshold, and the instantaneous frequency fluctuation rate of the vocal cord vibration audio signal is lower than the preset fluctuation tolerance and lasts for a preset time, a trigger pulse signal is generated. The signal acquisition module is used to control the ultrasound probe to emit a plane wave at a target pulse repetition frequency based on the trigger pulse signal, and simultaneously acquire ultrasound radio frequency echo signals and structural images. The tissue mask generation module is used to generate a tissue mask for the target region based on the structural image to eliminate fluid-containing areas; The wave field displacement processing module is used to extract the forward propagation wave field displacement information based on the ultrasonic radio frequency echo signal and calculate the wave peak arrival time difference. The shear modulus calculation module is used to calculate the local shear modulus of the target region by combining the tissue mask, the peak arrival time difference, and the preset tissue density.

Images