Intelligent recognition method and system for breast core needle biopsy tissue characteristics

Abstract

The application discloses a breast rotary cutting tissue characteristic intelligent identification method and system, which collects tissue characteristic data in real time during the rotary cutting process, and adopts innovative signal processing and pattern recognition algorithms for analysis, and can accurately identify the type and characteristics of the cut tissue. The method comprises the steps of real-time data acquisition, signal preprocessing, feature parameter extraction and intelligent classification and identification, and the identification accuracy is improved through parameter self-adaptive optimization. The system is composed of a sensing module, a data processing module, a pattern recognition module and an output interaction module, can be compatible with various rotary cutting devices, and supports various data formats through a standardized interface. The system can output tissue category information in real time according to the identification result, and automatically adjust the working parameters of the rotary cutting device when necessary, so as to ensure the stable and reliable cutting effect under different tissue conditions. The application realizes real-time intelligent identification of tissue characteristics in breast rotary cutting biopsy surgery, and has significant improvement in accuracy and efficiency.

Description

Technical Field

[0001] This invention relates to the fields of medical devices and signal processing technology, specifically to a method and system for intelligent identification of the characteristics of breast excised tissue. Background Technology

[0002] Vacuum-assisted excision biopsy of the breast has been widely used for minimally invasive diagnosis and removal of suspicious breast lesions. However, existing excision biopsy systems mainly rely on intraoperative image guidance and surgeon experience, lacking accurate technical means for real-time assessment of tissue characteristics. For example, in current technologies, physicians typically determine whether the target lesion tissue has been removed based on the appearance of the extracted tissue or postoperative images. Some systems offer real-time imaging verification functions or presets such as "dense tissue mode" to manually handle different tissue densities. However, these methods either verify after tissue sampling, failing to provide immediate feedback on tissue properties during the cutting process, or require manual judgment and mode switching, making it difficult to identify different types of tissue in a timely and accurate manner. Consequently, the problem of inaccurate tissue characteristic identification in breast excision surgery remains prominent: the inability to intelligently determine whether the removed tissue is lesion tissue or its density during surgery may lead to insufficient removal or misremoval of normal tissue, affecting surgical efficiency and diagnostic accuracy.

[0003] Therefore, there is an urgent need for a technical solution that can identify tissue characteristics in real time and accurately during breast excision to address the shortcomings of existing technologies that cannot intelligently identify tissue types and characteristics. Summary of the Invention

[0004] Technical Objective: To address the shortcomings of inaccurate tissue characteristic identification in existing breast excision surgery techniques, this invention discloses an intelligent identification method and system for breast excision tissue characteristics. This method can collect and analyze tissue characteristic data in real time during the excision surgery, intelligently and accurately identify the properties of the excised tissue, thereby significantly improving the accuracy and efficiency of the excision surgery.

[0005] Technical solution: To achieve the above technical objectives, the present invention adopts the following technical solution:

[0006] A method for intelligent identification of breast excised tissue characteristics specifically includes the following steps:

[0007] During the breast biopsy procedure, signal data characterizing tissue properties are acquired in real time through a sensing module. The signal data includes at least the motor current signal and vacuum negative pressure signal when the biopsy device is running.

[0008] The acquired signal data is filtered, denoised, and calibrated to obtain an effective signal reflecting the tissue cutting status;

[0009] Multiple characteristic parameters that can characterize tissue properties are extracted from the processed effective signal, including motor load characteristics and negative pressure change characteristics during the cutting process;

[0010] The extracted feature parameters are input into a pre-trained pattern recognition model, and the pattern recognition algorithm is used to classify and identify the characteristics of the currently excised tissue to obtain the tissue characteristic identification results.

[0011] Output information is generated based on the identification results and provided to the surgical operator or equipment control unit to indicate tissue characteristics or adjust the operating parameters of the rotary cutting device accordingly.

[0012] Preferably, the sensing module includes multiple sensors for acquiring different types of signal data, and the signal data includes at least one of the following signals: motor drive current signal, rotary cutting needle head vibration signal, vacuum suction pressure signal, acoustic signal, optical spectral signal, and bioelectric impedance signal.

[0013] Preferably, adaptive filtering or wavelet transform is used to denoise the signal data, and baseline calibration is performed on the motor current signal to eliminate the influence of the equipment's static bias on feature extraction.

[0014] Preferably, the integral value of the motor current signal during a single rotary cutting sampling process is calculated to obtain tissue hardness characteristic parameters, the decrease amplitude and recovery time of the vacuum negative pressure signal during the sampling process are calculated to obtain tissue density and patency characteristic parameters, and the main frequency of the vibration or sound signal is extracted to assist in the identification of tissue type.

[0015] Preferably, the pattern recognition model is constructed using a machine learning classification algorithm, including a support vector machine classifier or an artificial neural network model. The model is trained in advance using breast tissue data of known categories, and the feature threshold is adaptively adjusted during operation to improve the recognition accuracy.

[0016] Preferably, when the identification result indicates that the current tissue is dense tissue, a control signal is automatically sent to the rotary cutting device to increase the vacuum negative pressure or reduce the cutting speed of the blade, thereby optimizing the cutting and suction of dense tissue; when the identification result indicates that the target lesion tissue has been removed, a prompt signal is generated to remind the user to stop sampling.

[0017] A smart identification system for breast excision tissue characteristics, used to implement the smart identification method for breast excision tissue characteristics as described above, including:

[0018] The sensing module is used to collect various signal data in real time during the surgical process when installed on the breast biopsy device. The sensing module includes a current sensor, a pressure sensor, as well as a vibration sensor, an acoustic sensor, an optical sensor or a bioimpedance sensor to obtain electrical signals, mechanical signals and optical / electrical characteristic signals related to tissue cutting.

[0019] The data processing module, connected to the sensing module, is used to receive and process signal data and extract characteristic parameters that characterize tissue properties.

[0020] The pattern recognition module, embedded in the data processing module, is used to execute pattern recognition algorithms to classify and analyze feature parameters and generate tissue characteristic identification results.

[0021] The output interaction module, connected to the data processing module, is used to provide information output and user prompts based on the identification results of tissue characteristics, or to use the identification results to control the adjustment of the working parameters of the breast biopsy device.

[0022] Preferably, the data processing module includes a signal conditioning unit and a feature extraction unit. The signal conditioning unit filters, amplifies, and performs analog-to-digital conversion on the raw signals from each sensor. The feature extraction unit calculates multiple feature parameters, including the integral value of the motor current, the rate of change of the vacuum pressure, and the spectral characteristics of the vibration signal.

[0023] Preferably, the pattern recognition module employs a trained artificial intelligence algorithm to classify and judge the feature parameters in real time; wherein the artificial intelligence algorithm is a convolutional neural network or support vector machine model, which can calculate the input feature vector according to the pre-stored model parameters and output a result signal indicating the tissue category.

[0024] Preferably, the output interaction module includes a human-machine interface and a control interface. The human-machine interface is used to display real-time tissue characteristic category information and related indicators to the user. The control interface is used to send control commands to the breast biopsy device when a predetermined type of tissue is detected, so as to automatically adjust the vacuum negative pressure intensity or the speed of the biopsy blade.

[0025] Beneficial Effects: The intelligent identification method and system for breast excision tissue characteristics provided by this invention have the following beneficial effects:

[0026] 1. This invention acquires multi-source data in real time during breast biopsy, including motor current signals, vacuum negative pressure signals, and vibration, acoustic, optical, or bioimpedance signals. After signal processing steps such as filtering and noise reduction and baseline calibration, multiple quantitative feature parameters reflecting tissue characteristics are extracted. These features are then input into a pre-trained and adaptively adjustable pattern recognition model for classification and analysis, outputting tissue characteristic identification results in real time. If necessary, the output interaction module automatically adjusts the operating parameters of the biopsy device. Through multi-source signal fusion and multi-dimensional feature extraction, the pattern recognition model of this invention can more comprehensively reflect the tissue's hardness, patency, and vibration spectrum characteristics, thereby improving the differentiation between different tissue types and reducing misjudgments and omissions caused by human experience.

[0027] 2. The signal processing and pattern recognition of this invention are completed within milliseconds, enabling real-time output of identification results during tissue cutting. Based on the results, operating parameters such as cutting speed and vacuum pressure are dynamically adjusted to achieve adaptive optimization of the surgical process. The modular architecture and standardized data interface design allow the system to adapt to different models of breast biopsy devices and can be expanded with new sensing modules and feature dimensions as needed, improving its applicability under various clinical conditions. By driving the device with the identification results, torque and negative pressure are increased when cutting dense tissue, while efficiency is improved and negative pressure is reduced when cutting soft tissue, thereby reducing the risk of surgical trauma and avoiding unnecessary removal of normal tissue. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0029] Figure 1 This is a flowchart of the method of the present invention;

[0030] Figure 2 This is a system block diagram of the present invention;

[0031] Figure 3 This is a schematic diagram of the motor current variation curve of the present invention;

[0032] Figure 4 This is a schematic diagram of the vacuum pressure recovery curve of the present invention. Detailed Implementation

[0033] The present invention will now be described more clearly and completely by way of a preferred embodiment in conjunction with the accompanying drawings, but this does not limit the invention to the scope of the described embodiment.

[0034] like Figure 1 As shown, a method for intelligent identification of breast tissue characteristics includes the following steps:

[0035] S1. During the breast biopsy procedure, the sensor module acquires signal data characterizing tissue properties in real time.

[0036] The signal data includes various signals from the rotary cutting handle or needle, such as the current and torque signals of the rotary cutting motor, vacuum negative pressure signals, blade vibration signals, and optional acoustic or optical spectral signals. These signals are acquired through a sensing module connected to the rotary cutting device to ensure real-time capture of tissue characteristic information during tissue cutting.

[0037] S2. Filter, reduce noise, and calibrate the acquired signal data to obtain an effective signal reflecting the tissue cutting status.

[0038] Preferably, digital signal processing algorithms are used to denoise and calibrate the signal. For example, bandpass filters are used to remove low-frequency drift and high-frequency noise from the environment, or wavelet transform is used to separate useful signal components. By adaptively adjusting the parameters, the cutoff frequency and gain of the filter are optimized to adapt to different models of rotary cutting devices and individual patient differences, ensuring that the processed data is accurate and reliable.

[0039] S3. Extract multiple characteristic parameters that can characterize tissue properties from the processed effective signal, including motor load characteristics and negative pressure change characteristics during the cutting process.

[0040] Specifically, this includes, but is not limited to: extracting cutting resistance characteristics from motor current signals, extracting tissue extraction ease characteristics from vacuum pressure signals, extracting spectral characteristics from vibration or acoustic signals, and extracting tissue composition characteristics from optical or electrical signals. For example, calculating the integral of the motor current during each rotary cutting process to characterize the tissue density hardness index:

[0041]

[0042] Where I(t) is the instantaneous current of the rotary cutting motor during the cutting process, I0 is the reference current when idling without load, and T is the time length of a single rotary cutting sampling process. This hardness index H reflects the energy consumed by the cutting head in cutting the tissue; a higher value indicates a denser and harder tissue. For example, spectral analysis of the sensed vibration or acoustic signal can determine its dominant frequency f. p ,like:

[0043]

[0044] Where X(f) represents the amplitude spectrum of the signal in the frequency domain f. Different tissue types may correspond to different characteristic frequency distributions: for example, dense fibrous tissue may produce a higher dominant frequency peak, while adipose tissue has a lower dominant frequency. Feature vectors are constructed by extracting the aforementioned feature parameters.

[0045] S4. Input the extracted feature parameters into the pre-trained pattern recognition model, and use the pattern recognition algorithm to classify and identify the characteristics of the currently excised tissue to obtain the tissue characteristic identification results.

[0046] This invention preferably employs machine learning or deep learning algorithms to construct an intelligent recognition model. For example, a support vector machine (SVM) or neural network classifier is used to compare and classify real-time features with features of normal breast tissue and tumor tissue in a training library. The pattern recognition module improves recognition accuracy through prior parameter training and real-time adaptive optimization. The model parameters can be obtained through training with a large amount of ex vivo tissue experimental data and corrected according to specific equipment characteristics or patient conditions during system deployment. Through such parameter optimization, the pattern recognition module can achieve high-precision identification of tissue characteristics. For example, when the hardness index H is higher than a threshold and the dominant frequency f of the vibration signal... p When the elevation is significant, the model determines that the currently cut tissue may be a suspected lesion with dense fibrous tissue; otherwise, it is identified as ordinary glandular or adipose tissue. The entire identification process is completed within milliseconds, achieving real-time intelligent discrimination of tissue characteristics.

[0047] S5. Generate output information based on the identification results and provide the output information to the surgical operator or equipment control unit to indicate tissue characteristics or adjust the operating parameters of the rotary cutting device accordingly.

[0048] The output can be a real-time display of the tissue type determination result on the device's touchscreen interface (e.g., prompting "Current tissue: Benign fibrous tissue" or "Suspected tumor tissue") for the surgeon's reference. Furthermore, the system can automatically adjust the operating parameters of the rotary cutting device based on the identification results, thereby optimizing the surgical procedure. For example, when the current tissue is identified as dense tissue, the system can automatically increase the vacuum negative pressure or slow down the rotary cutting speed to ensure that the dense tissue is smoothly drawn into the blade groove and completely cut; when the target lesion tissue has been almost completely removed and the section is mostly normal tissue, the system can remind the surgeon to end the sampling to avoid excessive removal of normal tissue. Through this closed-loop feedback mechanism, intelligent control is integrated to further improve the safety and efficiency of the surgery.

[0049] A smart identification system for breast excision tissue characteristics, used to implement the smart identification method for breast excision tissue characteristics as described above, including:

[0050] The sensing module is installed on the breast biopsy device to collect various signal data in real time during the surgical procedure.

[0051] The sensing module includes one or more sensors for detecting signals generated by the interaction between the rotary cutting needle and the tissue. For example, it may include current and voltage sensors to monitor motor drive signals, pressure sensors to monitor changes in vacuum pressure, accelerometers / acoustic sensors to detect blade vibration or cutting audio signals, optical sensors (such as fiber optic spectral probes or lasers with photoacoustic transducers) to detect the optical / photoacoustic characteristics of the tissue, and bioimpedance electrodes to measure the electrical properties of the tissue. The sensing module can select at least one of the above-mentioned sensor types in combination as needed to comprehensively acquire tissue characteristic information. Structurally, the sensing module connects to the rotary cutting surgical device through a standardized interface, enabling it to adapt to different models of rotary cutting needles and equipment, ensuring the accuracy and real-time nature of data acquisition.

[0052] The data processing module, connected to the sensing module, is used to receive and process signal data and extract characteristic parameters that characterize tissue properties.

[0053] This module includes sub-modules such as a signal conditioning unit and a feature extraction unit. The signal conditioning unit filters, amplifies, and digitizes signals from different sources, using programmable parameters to adapt to the signal amplitude and frequency range of different devices. The feature extraction unit performs various feature calculations in the above-mentioned method steps, such as calculating the hardness index H and the dominant frequency f. p The system calculates data such as signal energy ratios and constructs feature vectors from these results, which are then provided to the recognition module. The data processing module employs a high-speed digital signal processor or embedded AI chip to achieve parallel processing and real-time computation of multi-channel data.

[0054] The pattern recognition module, embedded in the data processing module, is used to execute pattern recognition algorithms to classify and analyze feature parameters and generate tissue characteristic identification results.

[0055] This module can be implemented using software algorithms (such as machine learning models deployed within the data processing module) or accelerated using dedicated hardware circuitry. The pattern recognition module incorporates a trained tissue characteristic classification model (such as a model based on neural networks or support vector machines). During operation, the module can also adaptively adjust certain parameters based on the actual collected data (e.g., fine-tuning the discrimination threshold based on the tissue characteristics obtained from the initial few cuts) to improve recognition accuracy and robustness. The output of this module is the characteristic determination result for the currently collected tissue.

[0056] The output interaction module, connected to the data processing module, is used to provide information output and user prompts based on the identification results of tissue characteristics, or to use the identification results to control the adjustment of the working parameters of the breast biopsy device.

[0057] The output interaction module includes a human-machine interface (such as a touch screen, audible and visual alarm devices, etc.) that provides real-time feedback of identification information to the surgical operator. For example, when suspected tumor tissue is detected, a warning is highlighted on the interface, and relevant quantitative indicators (such as hardness index values) can be displayed; when the tissue is normal, a normal sampling result is generally indicated. If the system is equipped with automatic control functions, the output interaction module will also transmit the identification results to the control unit of the rotary cutting device to achieve dynamic parameter adjustment, such as automatically controlling the vacuum pump power or motor speed, to cope with different tissue conditions with optimal parameters. The output interaction module uses standardized data interfaces (such as UART, CAN bus, or Ethernet, etc.) to communicate with other hospital systems. The data interface follows a unified protocol and can support the output and recording of identification results and surgical data in multiple formats, facilitating postoperative analysis and cross-device compatibility.

[0058] The system of this invention preferably adopts a modular design, with each module interconnected through standard interfaces. For example, the sensing module and data processing module are connected via high-speed data interfaces (USB 3.0, SPI, etc.), the pattern recognition module is integrated as a software unit into the processor, and the output interaction module is connected to the processing unit via a communication interface. This modular design effectively solves system integration barriers, enabling flexible deployment or upgrades. When it is necessary to adapt to new rotary cutting equipment or add new sensors, only the corresponding module needs to be replaced or the interface protocol modified, without affecting the functionality of other modules. Simultaneously, the system is equipped with necessary auxiliary units such as power management, chassis, and mounting brackets to ensure reliable operation in surgical environments.

[0059] Example 1

[0060] like Figure 2 As shown, the intelligent identification system for breast biopsy tissue characteristics in this embodiment includes a sensing module, a data processing module, a pattern recognition module, and an output interaction module connected to the breast biopsy device. The various parts of the system are connected as a single unit via wired or wireless means. The sensing module is installed on the breast biopsy device to collect various signal data during the surgical procedure in real time. The data processing module is connected to the sensing module to analyze and process the collected data. The pattern recognition module is embedded in the processor of the data processing module to intelligently judge the extracted features. The output interaction module communicates with the data processing module to provide the recognition results to the surgeon or feedback control device.

[0061] Before the minimally invasive breast biopsy begins, all sensors in the sensing module are calibrated to their initial state. For example, the cutting needle is allowed to idle to determine the reference motor current I0 and reference noise level, and the vacuum pressure sensor reading is calibrated at zero point. This ensures the accuracy and reliability of subsequent measurements and reduces the impact of individual equipment differences on the recognition algorithm. Subsequently, during the biopsy procedure, all sensors in the sensing module synchronously begin data acquisition. The current sensor continuously monitors changes in the driving current of the cutting motor, the pressure sensor measures the real-time pressure value of the negative pressure suction, the accelerometer / microphone sensor is attached to the cutting needle handle to capture the blade vibration and cutting sound, and the optical / electrical sensor (if configured) performs rapid spectral or impedance measurements on the extracted tissue at the end of each cutting cycle. All sensor data is sent to the data processing module via a high-speed digital interface.

[0062] like Figure 3 As shown, the curve illustrates the typical variation of the instantaneous current I(t) of the rotary cutting motor during the breast rotary cutting process. The vertical axis represents the current I, and the horizontal axis represents time t; the horizontal dashed line represents the reference current I0 when idling without load. The curve includes rising, slight overshoot, and rippled steady-state segments, reflecting the characteristics of blade engagement and tissue load changes. This curve is used to calculate the tissue hardness characteristic parameter H.

[0063] The data processing module performs real-time preprocessing on data streams from different sources according to their respective characteristics. For example, it applies Fast Fourier Transform (FFT) to the motor current signal to obtain its spectrum and observe for abnormal peaks; or it calculates the current mean and variance using a sliding window to estimate the trend changes in the cutting load. For vacuum negative pressure signals, the data processing module calculates their drop amplitude and recovery time to assess the ease and patency of tissue suction. If the negative pressure fails to recover for an extended period (potentially indicating tissue blockage or incomplete cutting), the system can record this anomaly for the identification module's reference. For vibration and sound signals, time-frequency analysis methods such as wavelet transform are preferred to extract features such as energy proportion and instantaneous frequency changes in specific frequency bands, thereby capturing the differences in vibration modes when cutting different tissue materials. A series of original characteristic parameters are calculated, such as the hardness index H and the dominant frequency f. p Vacuum pressure recovery time constant Vibration signal energy ratio E, etc.

[0064] In a rotary cutting and suction cutting event, the vacuum pressure signal first drops to a local minimum. It then rose back to steady-state pressure. The sampling frequency is The time to the local minimum is denoted as To suppress noise interference, the pressure signal is first filtered and subjected to moving average processing, and then a linearized sequence is constructed within the recovery segment.

[0065]

[0066] in This represents the logarithmic transformation value of the vacuum pressure decay curve. It is a time-sampled sequence. The vacuum pressure at the time of sampling. This is the sampling point number corresponding to the minimum vacuum pressure. To prevent logarithmic operations from diverging to extremely small positive numbers, robust linear regression is used to obtain the slope. The first estimate was obtained. Simultaneously calculate the calibration value of "63.2% arrival time". ,in To restore stress The earliest moment, This is the difference between the steady-state pressure and the minimum vacuum pressure. The final weighted fusion result is taken.

[0067]

[0068] in It adaptively adjusts according to the fitting residuals and fluctuations. , , The unit is kPa. The unit is seconds; the larger the value, the denser the organization and the less unobstructed the channels.

[0069] like Figure 4 The figure shows the vacuum pressure recovery process during a single rotary cutting and suction cutting event. The curve at t min The minimum value p is reached at point min It then tends towards a steady state p ss When the pressure first reaches p ss The time when −0.368Δp is denoted as t. 63 ,definition .

[0070] Within a single rotary cutting and suction cutting event window W, the acceleration signal at the tool holder is acquired. Or microphone wind sound pressure signal The sampling frequency is The amplitude spectrum of the signal at time slice k is obtained by calculating the short-time Fourier transform. In the available frequency domain Within, calculate the average power spectrum of the event. Where K is the number of time window slices, adaptively finding the low-frequency main peak. With high frequency main peak Construct low-frequency bands respectively and high frequency band .calculate

[0071]

[0072] Again

[0073]

[0074] in This represents the total energy in the low-frequency band. This represents the total energy in the high-frequency band. For frequency resolution, To prevent extremely small positive numbers with a denominator of zero, As a dimensionless quantity, a high proportion of high-frequency energy usually corresponds to fibrous dense tissue or hard lesion tissue.

[0075] After feature extraction is complete, the pattern recognition module performs fusion analysis on these features. In this embodiment, the pattern recognition model is a trained three-layer artificial neural network classifier, whose input is the extracted feature vector (H, f). p The system outputs corresponding classification results, including categories such as "normal tissue," "fibrous dense tissue," or "suspected tumor tissue." To ensure the model's adaptability to different patients and devices, several adjustable parameters are pre-introduced into the model. For example, the final classification threshold of the neural network can be fine-tuned based on samples obtained from the first few rotary cuts during surgery—if the initial tissue samples are quickly pathologically confirmed to be benign, the model can appropriately increase the feature weight threshold required to determine malignancy, reducing the false alarm rate; conversely, the opposite is also true. The pattern recognition module executes the following recognition logic: first, the input features are standardized (subtracting the mean, dividing by the standard deviation, etc.), then a nonlinear combination is calculated through the hidden layer of the network, and finally, the confidence scores of each category are obtained in the output layer. If the confidence score of a certain category (such as "suspected tumor") exceeds the preset threshold, the system determines that the currently cut tissue belongs to that category. The entire judgment process takes a very short time (typically less than 100 milliseconds), which can be considered as real-time relative to the mechanical movement of the rotary cutter.

[0076] Once the pattern recognition module arrives at its decision, the data processing module immediately sends the result to the output interaction module. The output interaction module displays a prompt message on the surgical device's screen, such as "Tissue recognition result: Suspected tumor tissue, please pay attention to the resection margins," accompanied by a buzzer sound to draw the surgeon's attention. Simultaneously, the system records and stores the result, marking the corresponding sample number for postoperative pathological comparison. If the recognition results show that the current tissue is entirely normal and no abnormalities are detected consecutively, the system interface may indicate "The target lesion may have been completely removed," assisting the surgeon in deciding whether to terminate the surgery.

[0077] In this embodiment, the system also uses the recognition results to automatically adjust surgical parameters. When pattern recognition determines that the tissue density is high (e.g., two consecutive samples are both identified as dense fibrous tissue), the output module instructs the breast biopsy device to enter the preset "dense tissue mode" via the control interface. In this mode, the vacuum control system automatically increases the negative pressure suction force and extends the suction time to ensure that tough tissue fragments are smoothly extracted from the blade groove; at the same time, the motor control reduces the cutting speed to provide greater torque and avoid needle jamming. Conversely, when the detected tissue is soft (fatty tissue) or normal, the system can restore or switch to the "normal tissue mode," that is, reduce the negative pressure and increase the cutting speed to improve sampling efficiency and reduce damage to surrounding normal tissue. The entire process is completed automatically by the system without human intervention, achieving true intelligent assistance and adaptive control.

[0078] Example 2

[0079] The system of the present invention has high modularity and scalability. This embodiment introduces an improved scheme that integrates photoacoustic sensing technology for identifying pathological features of tissues.

[0080] Building upon the standard configuration, the sensing module incorporates a photoacoustic sensing submodule, including a pulsed laser diode and a high-sensitivity piezoelectric ultrasonic sensor. This submodule operates during the interval between each rotary cutting sampling. A short-pulse laser (e.g., wavelength 808 nm, pulse width 25 ns) is irradiated into the tissue within the rotary cutting groove by the laser diode. The laser energy is absorbed by the tissue, causing instantaneous thermoelastic expansion, thereby exciting an ultrawideband ultrasonic signal (i.e., a photoacoustic signal). After the piezoelectric sensor acquires the photoacoustic signal, the data processing module performs frequency domain analysis to calculate the power spectrum of the photoacoustic signal. Studies have shown significant differences in the photoacoustic spectra of different breast tissues: for example, breast tissue with fibrocystic changes produces a main spectral peak of approximately 1.60 MHz, while the peak frequency of normal breast tissue is approximately 0.26 MHz. This invention utilizes these interdisciplinary findings, incorporating photoacoustic spectral characteristics (such as main peak frequency, average frequency, and photoacoustic energy) into the feature extraction and pattern recognition process. When the photoacoustic sensing submodule detects a main frequency significantly higher than normal and high photoacoustic energy, the pattern recognition module uses this as an auxiliary criterion for identifying malignant or dense lesions, thereby further improving the accuracy of recognition. The addition of the photoacoustic sensing submodule fully demonstrates the compatibility and scalability of this system: by introducing advanced technologies from the fields of acoustics and optical imaging, this system can obtain tissue molecular composition and density information that traditional motors and pressure signals cannot provide, achieving a clever fusion of different technologies in the scenario of this invention.

[0081] It should be noted that, when applying the aforementioned photoacoustic technology, to ensure safety, both the laser output power and irradiation time are strictly controlled within medical safety standards, and the laser is only triggered when there is no remaining uncut tissue in the scalpel or when there is no direct irradiation to the patient. This design ensures that additional tissue information is obtained without increasing the risk to the patient.

[0082] Example 3

[0083] This embodiment provides test results of the system of the present invention under simulated extreme tissue conditions to demonstrate the stability and reliability of the system. The test is divided into two extreme scenarios: (1) extreme hardness: using materials with physical properties close to highly fibrotic lesions (e.g., high-strength rubber blocks) as prosthetic tissue for rotary cutting; (2) extreme softness: using soft materials with properties close to pure adipose tissue (e.g., low-density sponge) as prosthetic tissue. The system of the present invention is connected to a commercial breast rotary biopsy device, and continuous rotary cutting sampling is performed on these two extreme prostheses, each for 50 cycles. The results show that for the high-hardness prosthesis, the system sensing module successfully recorded a significantly increased motor load signal and a delayed negative pressure recovery time. The pattern recognition module accurately identified it as "high-density tissue" and automatically triggered the densification mode to increase negative pressure suction. For the extremely soft prosthesis, the system signal characteristics show that the motor load is very low and the vacuum pressure recovers rapidly. The recognition module identifies it as "soft tissue". The system can be smoothly cut in normal working mode without special adjustment. Throughout the testing process, the system achieved an accuracy rate exceeding 98%, with no misclassifications or system instability, demonstrating the algorithm's high robustness and stability even under extreme conditions. Simultaneously, the system hardware operated smoothly, with excellent inter-module interface coordination and no data loss or delays. Therefore, this invention's system has undergone rigorous testing and can ensure reliable tissue characteristic identification results even in exceptionally difficult or unique tissue conditions during actual surgery, without affecting surgical decisions due to isolated extreme situations.

[0084] Without departing from the inventive concept, those skilled in the art can make various modifications, combinations, or substitutions to the specific technical features in the embodiments, all of which fall within the scope of protection of this invention. For example, the specific implementation of the pattern recognition algorithm can use a convolutional neural network instead of the fully connected network in the embodiments herein to utilize the local features of the time-series signal; furthermore, the system's output module can interface with the hospital's information system to record the real-time recognition results in the patient's file. These improvements do not affect the realization of the function of this invention. In summary, this invention provides a novel technical tool for breast excision surgery with its unique multi-sensor fusion and intelligent algorithm. Its core idea and advantage lie in the real-time, intelligent, and high-precision identification of tissue characteristics, thus significantly distinguishing it from existing technologies and improving clinical outcomes. All combinations and additional variations between the various embodiments that do not depart from the principles of this invention should be considered within the scope of protection of this invention.

Claims

1. A smart identification system for the characteristics of breast excised tissue, characterized in that, include: A sensing module is used to collect various signal data in real time during the surgical process when installed on the breast biopsy device. The signal data includes at least the motor current signal and vacuum negative pressure signal when the breast biopsy device is running. The sensing module includes a current sensor, a pressure sensor, as well as a vibration sensor, an acoustic sensor, an optical sensor or a bioimpedance sensor to obtain electrical signals, mechanical signals and optical signals related to tissue cutting. The data processing module, connected to the sensing module, includes a signal conditioning unit and a feature extraction unit. The signal conditioning unit filters, amplifies, and performs analog-to-digital conversion on the raw signals from each sensor, and performs baseline calibration on the motor current signal to eliminate the influence of the device's static bias on feature extraction. The feature extraction unit calculates the motor current signal I(t) within the duration T of a single rotary cutting sampling process, minus the idling reference current. The integral value is then used to obtain the characteristic parameters characterizing tissue stiffness. The decrease amplitude and recovery time constant τ of the vacuum negative pressure signal during the sampling process are calculated to obtain characteristic parameters characterizing tissue density and patency. ,in , , This indicates the first estimate. This represents the slope obtained using robust linear regression. This represents the calibration value indicating "63.2% arrival time". This indicates that the pressure has reached its first peak. At that moment, Represents the time at a local minimum. For steady-state pressure, The difference between the steady-state pressure and the minimum vacuum pressure is... , This is the minimum vacuum pressure. The system adaptively adjusts to the fitting residuals and fluctuations, extracting the ratio E of the energy in the low-frequency band to the energy in the high-frequency band of the vibration or sound signal to obtain spectral feature parameters characterizing tissue type. These spectral feature parameters are then combined into a multi-dimensional feature vector. in This represents the total energy in the low-frequency band. This represents the total energy in the high-frequency band. To prevent extremely small positive numbers with a denominator of zero; The pattern recognition module uses a pre-trained machine learning classification model to perform real-time multi-class classification on multi-dimensional feature vectors and output tissue category identification results indicating normal tissue, fibrous dense tissue or suspected tumor tissue. The output interaction module, connected to the data processing module, includes a human-machine interface and a control interface. The human-machine interface displays tissue characteristic category information and related quantitative indicators to the user in real time. When the identification result indicates that the current tissue is dense tissue, the control interface automatically sends a control signal to the breast biopsy device to increase the vacuum negative pressure suction force and reduce the cutting speed, so as to provide greater torque to ensure that the tissue is smoothly cut and aspirated. When the identification result indicates that the current tissue is soft tissue or normal tissue, the vacuum negative pressure is reduced and the cutting speed is increased to improve sampling efficiency and reduce damage to surrounding normal tissue. When the identification result indicates that the target lesion tissue has been removed, a prompt signal is generated to remind the user to stop sampling.

2. The intelligent identification system for breast excision tissue characteristics according to claim 1, characterized in that, The signal conditioning unit of the data processing module performs adaptive filtering or wavelet transform noise reduction on the raw signals from each sensor.

3. The intelligent identification system for breast excision tissue characteristics according to claim 1, characterized in that, The pattern recognition module employs a convolutional neural network or support vector machine model, which can calculate the input multidimensional feature vector based on pre-stored model parameters, output a result signal indicating the tissue category, and adaptively adjust the classification threshold during operation to improve recognition accuracy.

Images