Multimodal flow-based particle detection method based on PT-symmetric circuit
By combining PT symmetric circuits with microfluidic chips, the time-domain and frequency-domain feature information of particles is obtained. Combined with machine learning models, the problems of high throughput, low sensitivity, and label-free detection of particles in existing technologies are solved, and accurate classification and state prediction of particles are achieved.
Patent Information
- Application Number
- PCT/CN2025/107488
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-15
- Filing Date
- 2025-07-08
- Publication Date
- 2026-11-19
AI Technical Summary
Existing particulate detection technologies in the biomedical field suffer from problems such as high throughput, low sensitivity, low precision, and inability to perform label-free detection. In particular, optical flow cytometry and impedance flow cytometry are insufficient in terms of detection complexity and accuracy.
By combining PT symmetric circuits with microfluidic chips, and adjusting circuit parameters to singular points, the extremely high sensitivity of voltage waveforms is utilized to acquire time-domain and frequency-domain feature information. This information is then combined with machine learning models for label-free detection and accurate classification of particles.
It achieves high-throughput, high-sensitivity label-free detection and accurate classification of microparticles, and can identify the quantity, size and type of various microparticles, and predict the clinical cycle and status of biological microparticles.
Smart Images

Figure CN2025107488_19112026_PF_FP_ABST
Abstract
Description
A multimodal flow particle detection method based on PT symmetric circuit TECHNICAL FIELD
[0001] The present application belongs to the technical field of biomedical detection, and specifically relates to a multimodal flow particle detection method based on a PT symmetric circuit. BACKGROUND
[0002] Traditional particle detection technologies mainly include optical flow cytometry and impedance flow cytometry. An invention patent application with publication number CN104503101A discloses a flow cytometer light beam forming system based on a diffractive optical shaping device, which comprises a laser light source, a diffractive optical shaping device, a reflector, a beam splitter, and a lens. The light emitted by each laser light source forms a light spot with uniform intensity distribution, including but not limited to a rectangular light spot, after passing through the diffractive optical shaping device, and finally irradiates on the cell liquid flow of the flow cytometer system. Compared with the results of the traditional flow cytometer light beam forming system, the patent application applies the diffractive optical shaping device to the light beam forming system of the flow cytometer, forms an illumination light spot with uniform intensity distribution, and improves the energy utilization rate and measurement and analysis accuracy of the flow cytometer system. However, the optical flow cytometry disclosed in the invention patent application relies on cell biomarkers, and has problems such as complex operation, difficulty in detecting cells lacking markers, and possible alteration of the intrinsic characteristics of cells.
[0003] An invention patent application with publication number CN116371494A discloses a microfluidic chip for identifying and separating abnormal cells in CAR-T cells and a preparation method thereof. The microfluidic chip comprises a microchannel, the microchannel comprises an inlet for cells to be detected to enter and at least two outlets for the cells to be detected to leave, a microchannel separation channel, and a detection channel. A pair of detection electrodes distinguishes the cells as abnormal cells or normal cells by detecting the impedance. A pair of positioning electrodes confirms which outlet the cells pass through according to the detected impedance. However, the impedance flow cytometry disclosed in the invention patent application relies on external platform parameters, and has problems such as low detection throughput, insufficient precision, poor data consistency, and inability to identify cells at different clinical stages.
[0004] Therefore, there is still a need to develop new technologies to meet the demand for high-throughput, high-sensitivity, and low-cost label-free cell detection in the field of biomedicine.
[0005] The singular point effect of the PT (PT-symmetric system) symmetric system has extremely high sensitivity and detection stability, and can produce a great response to the change of a small signal under the condition of balanced gain and loss.
[0006] The patent application with the publication number CN114812371A discloses a metal thin film thickness detection system based on a PT symmetric circuit. It includes a first resonant circuit and a second resonant circuit, which are RLC series resonant circuits with the same inherent resonant frequency and are connected in series at both ends of a coupled inductor. By adjusting the capacitance value in the first resonant circuit and the second resonant circuit, the gain loss coefficient is changed, and the gain loss coefficients of the two are opposite to each other, so that a PT symmetric resonant system is obtained. The coupled inductor of the system and the metal thin film to be measured change in self-induction under the influence of the eddy current effect, the coupling coefficient between the first and second resonant circuits changes, and the resonant frequency of the PT symmetric resonant system is changed.
[0007] However, the biological application of the PT symmetric system needs to be further studied and discussed, and the rapid detection method of the PT symmetric system needs to be developed. SUMMARY
[0008] The present application provides a multi-modal flow micro-particle detection method based on a PT symmetric circuit, which aims to realize label-free detection and accurate classification of micro-particles by utilizing its extremely high sensitivity to small perturbations.
[0009] The present application provides a multi-modal flow micro-particle detection method based on a PT symmetric system, which comprises:
[0010] The conductive electrodes of the microfluidic chip are connected to the PT symmetric circuit, and the parameters of the PT symmetric circuit are adjusted to the singular point. The output voltage waveform of the circuit is segmented and processed to obtain time domain feature information and frequency domain feature information. When the micro-particles pass between the electrodes, a wave peak or a wave trough appears in the feature information. The extreme points corresponding to the wave peak or the wave trough are taken as the characteristic values of the micro-particles, and the number of extreme points corresponds to the number of micro-particles, which can be used for counting.
[0011] The characteristic values of multiple pairs of micro-particles and the corresponding micro-particle labels are divided into a training sample set and a verification sample set. A machine learning model is trained by a training data set to obtain a micro-particle prediction model.
[0012] In application, the extreme points of the time domain feature information and / or the frequency domain feature information of the micro-particles are input into the micro-particle detection model to obtain the category prediction value, the clinical cycle prediction value, the variant prediction value, or the different state prediction value of the micro-particles. When the micro-particles are biological micro-particles, the different states are the initial stage, the growth stage, or the apoptosis stage of the biological micro-particles.
[0013] Preferably, the method for counting and flux calculation of m kinds of micro-particles respectively comprises:
[0014] The PT-symmetry circuit-based multi-modal flow particle detection method is used to test m kinds of particles respectively, corresponding training sample sets are obtained, machine learning training is performed, a particle prediction model is obtained, and then the obtained particle prediction model is applied to sample detection containing m kinds of particles at the same time, the feature change caused by each particle passing through the electrode is input into the particle prediction model, the type is predicted, and then the number n of the ith kind of particle in the time Δt is counted i ;
[0015] The detection flux of the ith kind of particle is calculated as K i = n i / Δt, wherein n i (i=1-m) is the number of peaks or troughs of the ith kind of particle in Δt time;
[0016] The total flux of the particles is K, K=Σn i / Δt.
[0017] Preferably, the voltage waveform is segmented and processed to obtain time domain feature information, including:
[0018] The segmented voltage waveform is analyzed by using data processing software such as matlab or python, and one or more of the standard deviation, kurtosis, envelope frequency, envelope peak value, envelope valley value, and envelope amplitude information of the voltage waveform at different times are constructed to obtain time domain feature information.
[0019] Preferably, the voltage waveform is segmented and processed to obtain frequency domain feature information, including:
[0020] The segmented voltage waveform is subjected to Fourier transform, and one or more of the frequency shift and amplitude change information of the particle passing through the microfluidic chip compared with the eigenmode are constructed to obtain frequency domain feature information.
[0021] Preferably, the time for a single particle to pass through the electrodes is 0.1-10 ms.
[0022] Preferably, the PT-symmetry circuit is composed of LC circuits with gain and loss, two LC circuits are connected by capacitive coupling, inductive coupling or equivalent transmission line, and the parameters of the two LC circuits satisfy the PT-symmetry condition.
[0023] Preferably, the voltage waveform information is obtained by reading the voltage signal of the gain or loss LC circuit of the PT-symmetry circuit through one of an oscilloscope, an FPGA circuit, and a single-chip microcomputer circuit.
[0024] Preferably, the microfluidic chip comprises:
[0025] a substrate;
[0026] A micro-channel is arranged on the substrate, and a groove is arranged on the micro-channel, and a measuring position is arranged between the front and back ends of the groove, and the micro-particles flow from one end of the groove to the other end and pass through the measuring position.
[0027] and a conductive electrode arranged on the substrate and connected with the PT symmetric circuit, the conductive electrode is one or more pairs of electrodes, and the ends of the two electrodes in each pair of electrodes are arranged left and right or on the same side at the measuring position, for measuring the impedance when the micro-particles pass.
[0028] Preferably, the method for obtaining the size of the micro-particles comprises:
[0029] When the ratio of the distance between the electrodes, the cross-sectional size of the pipeline and the size of the micro-particles is less than 2, the injection flow rate Q pushed by the injection pump and the cross-sectional area S near the electrodes are used to calculate the flow rate V of the micro-particles passing through the electrodes, and the time t of the micro-particles passing through the electrodes is further combined to calculate the diameter X of the micro-particles, that is, X = V·t = Q·t / S, and a coefficient a is introduced for correction, so that the final calculation method of the diameter of the micro-particles is X = V·a·t = Q·a·t / S.
[0030] Preferably, the machine learning model is obtained by training a machine learning model through a training sample set, and the convolutional neural network, the machine learning model comprises a convolutional neural network, a decision tree, an xgboost or a KNN algorithm.
[0031] Preferably, the multi-modal flow micro-particle detection method distinguishes the number (count), size and type of the micro-particles including cells, bacteria, proteins, exosomes or micro-plastic particles, when the micro-particles refer to cells, the method can further distinguish the varieties, clinical cycles and cell cycles of the cells; when the micro-particles refer to bacteria, the method can further distinguish the growth states of the bacteria.
[0032] Compared with the prior art, the present application has the following beneficial effects:
[0033] The present application can make the circuit output voltage waveform extremely sensitive when the micro-particles pass through the conductive electrode by regulating the PT symmetric circuit to a singular point, and the time domain characteristic information and the frequency domain characteristic information can be obtained by processing the voltage waveform, the extreme points of the wave peaks or wave troughs appearing in the characteristic information are taken as the characteristic values of the micro-particles, so that the characteristic value information of the micro-particles is more abundant, and the machine learning model trained by the characteristic values of the micro-particles can perform no-label detection and accurate classification of the micro-particles. BRIEF DESCRIPTION OF DRAWINGS
[0034] FIG. 1 is a circuit schematic diagram of a multi-modal flow micro-particle detection system based on PT symmetry provided by an embodiment of the present application;
[0035] FIG. 2 is a schematic diagram of a negative resistance circuit based on an operational amplifier;
[0036] Figure 3 is a schematic diagram of a microfluidic chip provided in a specific embodiment of the present invention;
[0037] Figure 4 is a physical simulation diagram of the multimodal flow cytometry particle detection system based on PT symmetry provided in a specific embodiment of the present invention;
[0038] Figure 5 is a flowchart of a multimodal flow cytometry particle detection method based on PT symmetric circuits provided in a specific embodiment of the present invention;
[0039] Figure 6 is a diagram showing the system voltage waveform changes caused by the cells passing through the electrodes according to a specific embodiment of the present invention;
[0040] Figure 7 shows the changes in the time-domain characteristics of the system caused by the cells passing through the electrodes according to a specific embodiment of the present invention (1);
[0041] Figure 8 shows the change in the system time-domain characteristics caused by the cells passing through the electrodes according to a specific embodiment of the present invention (2);
[0042] Figure 9 is an FFT spectrum diagram corresponding to the system voltage waveform change caused by the cell passing through the electrode in a specific embodiment of the present invention.
[0043] Figure 10 is a diagram showing the change in system frequency domain characteristics caused by cells passing through electrodes according to a specific embodiment of the present invention;
[0044] Figure 11 is a system frequency offset diagram in a high-throughput cell assay provided in a specific embodiment of the present invention;
[0045] Figure 12 is a system frequency offset diagram in the cell survival ratio test provided in a specific embodiment of the present invention;
[0046] Figure 13 shows the prediction results of the time-frequency domain full-feature convolutional neural network provided in a specific embodiment of the present invention;
[0047] Figure 14 shows the prediction results of the temporal feature convolutional neural network provided in a specific embodiment of the present invention. Detailed Implementation
[0048] To enhance understanding of the present invention, the embodiments will be described in detail below with reference to the accompanying drawings.
[0049] The voltage waveform of the micro-particle passing through the micro-fluid chip can be obtained by using the PT symmetric circuit in the embodiment of the present application, and the time domain characteristic information and the frequency domain characteristic information are obtained by processing the voltage waveform, when the micro-particle passes through the conductive electrode of the micro-fluid chip, the wave peak or the wave trough appears in the time domain characteristic information and the frequency domain characteristic information, and the extreme point of the wave peak or the wave trough is taken as the characteristic value of the micro-particle, so that the characteristic value of the micro-particle is extremely rich, and the model trained by using the characteristic value of the micro-particle can quickly and accurately detect the category of the micro-particle in the clinical period.
[0050] The embodiment of the present application provides a multi-modal flow particle detection method based on a PT symmetric circuit, which comprises the following steps:
[0051] S1, the conductive electrode of the micro-fluid chip is connected with the PT symmetric circuit, the parameters of the PT symmetric circuit are adjusted to the singular point, and the voltage waveform information of the micro-particle passing through the micro-fluid chip is obtained. When the parameters of the PT symmetric circuit are to the singular point, when the micro-particle passes through the conductive electrode, it can be extremely sensitive to detect that the voltage waveform information changes obviously.
[0052] In an embodiment, the PT symmetric circuit provided by the embodiment comprises LC circuits with gain and loss, two LC circuits are connected through capacitive coupling, inductive coupling or equivalent transmission line, the parameters of the two LC circuits satisfy the PT symmetric condition, and the two LC circuits are series LC circuits or parallel LC circuits.
[0053] As shown in FIG. 1, the PT symmetric circuit provided by the embodiment of the present application comprises a detection end and a reading end, wherein the detection end comprises a capacitive element (C s ), a resistive element (R s ) and an inductive element (L s ), which is an LC parallel resonance circuit with loss (R s ); the reading end comprises a capacitive element (C r ), a negative resistive element (R r ) and an inductive element (L r ), which is an LC parallel resonance circuit with gain (R r ), the detection end and the reading end are connected through inductive coupling, and can also be connected through capacitive coupling or equivalent transmission line coupling, forming a PT symmetric structure, and the parameters satisfy the PT symmetric condition (i.e. L r =xL s , xC r =C s and R r =-xR s, x is a scaling factor.); a microfluidic chip is connected to the detection end, and the chip is provided with a conductive interdigital electrode for detecting the change of the electrical signal caused by the passing of the microparticles (here, the impedance of the microfluidic chip is equivalent to a parallel RC circuit, and the R s , C s ) has been considered.
[0054] The part with loss characteristics (R s ) can be composed of one of ordinary resistance or two-dimensional material, metasurface with equivalent impedance value; the part with gain characteristics (R r ) can be composed of one of amplification circuit composed of operational amplifier, mos tube, geng type diode, tunnel diode, two-dimensional material, metasurface with the same impedance value, as shown in figure 2 is based on the schematic diagram of negative resistance circuit of operational amplifier.
[0055] As shown in figure 3 and figure 4, the microfluidic chip provided by the embodiment comprises a substrate 201, a microchannel 202 and a conductive electrode 203.
[0056] The microchannel 202 provided by the embodiment is located on the substrate 201, and the microchannel 202 has a groove, and the groove is provided with a measurement position between the front and back ends, and the microparticles flow from one end of the groove to the other end and pass through the measurement position.
[0057] The microfluidic chip provided by the embodiment can also be composed of other materials, such as a glass layer fixed on a silicon substrate by thermal bonding.
[0058] The conductive electrode 203 provided by the embodiment is located on the substrate 201 and connected with the PT symmetric circuit, and the conductive electrode 203 is one or more pairs of electrodes, and the end portions of the two electrodes in each pair of electrodes are arranged oppositely at the measurement position, and in an embodiment, the distance between the end portions of the oppositely arranged two electrodes is 10nm-100μm, which is used for measuring the impedance when the microparticles pass through.
[0059] The conductive electrode provided by the embodiment is one or more pairs of interdigital electrodes and one or more pairs of positive electrodes.
[0060] The material of the conductive electrode provided by the embodiment is one of conductive nanomaterial, metal and compound, and the substrate and microchannel can be composed of one of glass and silicon.
[0061] In a specific embodiment, as shown in figure 4, the inductance value of the coil (L s , L r ) is 10μH, the capacitance (C s , C r ) is adjustable capacitor (TZB4R500AB10R00, Murata), and the resistance (R s) is an adjustable resistor (3314J-1-502E, Bourns), wherein the gain unit (negative resistance) is composed of a positive feedback circuit of an operational amplifier chip (OPA695, Texas instruments) (as shown in FIG. 2), R G = R L = 511Ω, according to the virtual short and virtual open properties of an ideal operational amplifier, it is easy to deduce that the size of the negative resistance is controlled by R F (3314J-1-502E, Bourns). Therefore, when the system parameters satisfy C s = C r , R s = R F (Here, the impedance of the microfluidic chip is equivalent to a parallel RC circuit, and R s , C s have been considered) , The system works in the PT symmetric state.
[0062] The voltage waveform information provided by the embodiment is obtained by reading the voltage signal of the LC circuit of the PT symmetric circuit through one of an oscilloscope, an FPGA circuit and a single-chip microcomputer circuit.
[0063] The microparticles provided by the embodiment include cells, bacteria, proteins, exosomes or microplastics.
[0064] S2, the voltage waveform is segmented, processed to extract the peak value, and then the baseline is removed to obtain time domain waveform information and frequency domain waveform information, when the microparticles pass between the electrodes, the time domain characteristic information and the frequency domain characteristic information appear wave peaks or wave troughs, and the extreme points corresponding to the wave peaks or wave troughs are taken as the characteristic values of the microparticles, such as characteristics 1-10 of cell 1 in FIG. 5.
[0065] In an embodiment, the embodiment provides specific steps of segmenting and processing the voltage waveform to obtain time domain waveform characteristics, including:
[0066] The segmented voltage waveform is analyzed and the baseline is removed by using data processing software such as matlab or python, and one or more of the standard deviation, kurtosis, envelope frequency, envelope peak value, envelope valley value and envelope amplitude of the voltage waveform at different times obtained after removing the baseline are taken as the time domain characteristic information. Since the time domain characteristic information is more diverse, it is beneficial to increase the accuracy of the training model. At the same time, the time domain information processing speed is faster, which is beneficial to further improve the detection speed in the later stage.
[0067] The embodiment of the application realizes the counting of microparticles by counting the number of wave peaks or wave troughs in the standard deviation, kurtosis, envelope frequency, envelope peak value, envelope valley value or envelope amplitude characteristic information of the voltage waveform.
[0068] In an embodiment, the embodiment provides segmenting and processing voltage waveforms to obtain frequency domain information, including:
[0069] The system operating frequency interval obtained by Fourier transforming the segmented voltage waveform information is between 1-100MHz, and one or more of the frequency offset and amplitude change characteristic information of the micro-particles compared with the eigenmode when passing through the microfluidic chip is taken as the frequency domain characteristic information.
[0070] The embodiment of the application realizes the counting of a single micro-particle by counting the number of wave peaks or wave troughs in the characteristic information of the frequency offset or amplitude change.
[0071] S3, the characteristic values (characteristics 1-10) of the multiple pairs of micro-particles and the corresponding micro-particle labels are divided into a training set and a validation set, a machine learning model is trained through the training set to obtain a micro-particle detection model, and since the characteristic values of the micro-particles provided by the embodiment of the application are extremely rich, a model capable of accurately identifying the category and clinical period of the micro-particles can be trained, and the machine learning model includes a convolutional neural network, a decision tree, an xgboost or a KNN algorithm.
[0072] When applied, the extreme points of the time domain characteristics and / or frequency domain characteristics of the micro-particles are input into the micro-particle detection model to obtain the category prediction value, the clinical period prediction value, the variant prediction value or the different state prediction value of the micro-particles, and the different state is the initial stage, the growth period or the apoptosis period or the clinical period of the biological micro-particles when the micro-particles are biological micro-particles. It can be understood that as long as the application can provide different categories of micro-particles, and the clinical period, variant or different state of biological micro-particles, the micro-particle detection model capable of accurately predicting the category prediction value, the clinical period prediction value, the variant prediction value or the different state of the micro-particles can be obtained by training in the system provided by the embodiment of the application.
[0073] In a specific embodiment, the convolutional neural network provided by the embodiment includes 10 input neurons, 100, 50 and 30 neurons in the three hidden layers respectively, and 5 output neurons, which are used for cell identification to obtain the predicted cell category or clinical period, i.e., the predicted label, and a confusion matrix is constructed by the true label and the predicted label.
[0074] The embodiment of the application further provides a method for obtaining the size of the micro-particles, including:
[0075] When the ratio of the distance between the electrodes, the pipe cross-sectional size and the size of the microparticles is less than 2, the bolus flow rate Q pushed by the injection pump, and the cross-sectional area S near the electrodes, we can calculate the flow rate V of the microparticles passing through the electrodes = Q / S, and further combine the time t of the microparticles passing through the electrodes, to calculate the size X of the microparticles = V·t = Q·t / S. Considering the influence of the actual width of the electrode on the calculation time t, a coefficient a is introduced for correction, so the final microparticle diameter calculation method is X = V·a·t = Q·a·t / S.
[0076] The time for a single microparticle to pass between the electrodes using the above-mentioned multi-modal streaming microparticle detection method based on PT symmetric circuit is 0.1-10 ms.
[0077] The specific embodiments of the present application also provide a method for counting different microparticles respectively, comprising:
[0078] By separately testing m kinds of microparticles, corresponding training sample sets are obtained, and machine learning training is performed to obtain a microparticle prediction model, and then the obtained microparticle prediction model is applied to sample detection containing m kinds of microparticles at the same time, the feature change caused by each microparticle passing through the electrodes is input into the microparticle prediction model, the type is predicted, and then the number n of the i-th kind of microparticle in the time Δt is counted. i The detection flux of the i-th kind of microparticle is calculated as K i = n i / Δt, where n i (i = 1-m) is the number of peaks or troughs corresponding to the i-th kind of microparticle in the time Δt. The total flux of the microparticles is K, K = Σn i / Δt.
[0079] In an embodiment, the detection sample is breast cancer cells (MCF-7), and in this embodiment, the injection pump is used to push the cell solution, the oscilloscope is used to record and read the voltage waveform and save it, and then offline data processing is performed. According to the size of the system LC element value, the system operating frequency is estimated to be less than 10 MHz, so the sampling rate of the oscilloscope is set to 40M / s, and the sampling point number is set to 100Mpts. After saving the data, signal analysis is performed using matlab, or real-time data processing is performed directly using FPGA+ADC. As shown in FIG. 6, the waveform signal change during the process of the cell passing through the electrode is first shown (to more obviously show the signal change caused by the cell flowing through the electrode, only the voltage waveform information with a time length of 8μs (320pts) is shown here), and in this embodiment, the pattern between the adjacent two troughs of the outer envelope is regarded as a small envelope, and the time length is T, so the envelope frequency of the outer envelope is 1 / T. The waveforms of (a) and (c) of FIG. 6 respectively represent before and after the cell flows through the electrode, and the waveform of (b) of FIG. 6 represents that the cell is between the electrodes. It can be seen that when the cell flows through the electrode, the inner envelope voltage amplitude Va The decrease and the increase of the number of envelopes in the same time (i.e. the increase of the envelope frequency) indicate that the cell passing the electrode will cause a significant change in the waveform.
[0080] In order to quantitatively analyze the signal, the waveform signal of 100Mpts length is segmented by 20kpts (corresponding to 0.5ms) in this embodiment, and finally divided into 5000 segments. And for each segment of 20kpts length signal, the mean square error, kurtosis, envelope frequency is calculated, so as to obtain the change of these parameters over time. As shown in Fig. 7(a), the standard deviation analysis result of 11 continuous 100Mpts data files, each downward peak represents a change caused by a cell passing the electrode. According to the size difference between different downward peaks, it reflects the response difference between different cells. Fig. 7(b) shows the standard deviation change of a single cell passing the electrode after amplification. Combined with the waveform change of the three time points of the cell passing the electrode in Fig. 6, it can be seen that when the cell passes the electrode, the number of envelopes per unit time increases, the waveform density increases, and the standard deviation decreases to 0.352V. When the cell is far away from the electrode, the standard deviation value gradually returns to the baseline, i.e. 0.372V. It can be known from Fig. 7(b) that the time of a cell passing the electrode is about 10ms. Correspondingly, there are also kurtosis, envelope frequency (1 / T), envelope valley (V a / 2), envelope peak (V b ), envelope amplitude (V b -V a / 2) and other time domain characteristics, as shown in Fig. 7(c)-(f) and Fig. 8(a)-(f). These characteristics all reflect the influence of the impedance change of the cell passing the electrode on the system.
[0081] Then the Fourier transform of each segment of the waveform is performed in this embodiment to obtain the frequency domain information. Fig. 9 is the frequency domain change during the process of the cell passing the electrode, which corresponds to the three time points in Fig. 7(b) respectively. It can be found that the frequency domain information of the waveform shows two eigenmodes of the PT system, and the frequency and amplitude of the two eigenmodes will obviously shift when the cell passes the electrode. If the original frequency and amplitude of the left peak are denoted as (f1, A1), and the original frequency and amplitude of the right peak are denoted as (f2, A2), and the frequency and amplitude of the left peak when the cell passes the electrode are denoted as (f1 ’ , A1 ’ ), and the original frequency and amplitude of the right peak are denoted as (f2 ’ , A2 ’ ), then the signal change of the left peak is Δf1 = f1-f1 ’ , ΔA1 = A1-A1 ’ , and the signal change of the right peak is Δf2 = f2-f2 ’, ΔA2 = A2 - A2 ’ Fig. 10 (a)-(d) respectively show the corresponding frequency domain signal changes, while showing the high signal-to-noise ratio of the system. According to each peak, the cells can be counted and analyzed.
[0082] Further, the embodiment accelerates the injection speed of the solution of the injection pump for high-throughput cell analysis. As shown in Fig. 11 (a) and (b), the system frequency domain characteristics after the flow rate is accelerated are shown, and from the enlarged view on the right, it can be seen that the time of a single cell passing through the electrode is 0.1 ms, which shows a test throughput of 10,000 cells per second.
[0083] Then the embodiment uses formaldehyde solution to inactivate MCF-7 cells, and mixes them with living cells to obtain cell solutions with different survival rates, and uses them for testing. We extract the extreme points (i.e. the time when the cells are at the electrode) in Fig. 10 (b) and (d) as the final response of the system, so as to obtain the response data set of different cells, and the frequency domain test results are shown in Fig. 12 a and b, the box plot shows the 95% confidence ellipse distribution interval of three kinds of cell solutions, and the pentagram in the distribution interval represents the centroid point of the ellipse. The box on the opposite side of the xy axis corresponds to the minimum value, lower quartile, median, upper quartile and maximum value of the frequency domain response data set of each cell solution.
[0084] As can be seen from Fig. 12a, as the survival rate of cells decreases, Δf1 at the centroid point deviates from negative to positive, showing a clear trend. As can be seen from Fig. 12b, as the survival rate of cells decreases, ΔA1 and ΔA2 at the centroid point gradually increase, also showing a clear trend, which is mainly because as the cells die, the permeability of the cell membrane changes, and the resistance characteristics of the cells at the test frequency of 7MHz gradually weaken. The above results show the potential of the PT system in cell activity analysis.
[0085] The multi-modal flow micro-particle detection method provided by the embodiment distinguishes the number (count), size and type of the micro-particles, including cells, bacteria, proteins, exosomes or micro-plastic particles, when the micro-particles refer to cells, it can further distinguish their varieties, clinical periods and cell cycles; when the micro-particles refer to bacteria, it can further distinguish their growth states.
[0086] In a specific embodiment, different types of cancer cells, cervical cancer cells (Hela), breast cancer cells (MCF-7), and small cell lung cancer cells of different clinical periods, early (PC-9), medium (NCI-1373), and late (NCI-1792), are used as test samples, voltage waveforms during the test process are obtained, and the extreme value points of the time-frequency domain feature changes of the cells passing through the electrodes are extracted as feature values, including 6 time domain features such as standard deviation, kurtosis, envelope frequency, envelope peak value, envelope valley value, and envelope amplitude, and 4 frequency features such as frequency shift and amplitude change of two eigenmodes, and each type of cell data is 1700. We use matlab to add labels to them, and perform random sampling according to the ratio of training set: test set = 7:3 to obtain the training set and the test set, which are used for convolutional neural network training (where the number of input neurons is 10, the number of hidden layer neurons is 100, 50, and 30 respectively, and the number of output neurons is 5), for cell recognition. The results are shown in FIG. 13, and the classification accuracy can reach more than 97.13%.
[0087] In order to show the importance of the newly introduced frequency domain features of the present application, the frequency domain features are further excluded, and only the time domain features are used for training, and the results are shown in FIG. 14. The accuracy of each type of cell is significantly decreased, and the accuracy of 1792 cells is directly reduced to 81.12%, causing a great recognition error.
[0088] The method provided by the specific embodiment of the present application has application potential in label-free, high-throughput, and high-sensitivity single cell detection and classification, and is suitable for fields such as cancer diagnosis, drug screening, and immune detection.
[0089] It should be noted that the above embodiments are not intended to limit the scope of protection of the present application, and equivalent transformations or substitutions made on the basis of the above technical solutions all fall within the scope of protection of the claims of the present application.
Claims
1. A method for multi-modal flow particle detection based on PT-symmetric circuit, characterized in that, The method comprises the following steps: The conductive electrodes of the microfluidic chip are connected with the PT symmetric circuit, the parameters of the PT symmetric circuit are adjusted to the singular point, the time domain characteristic information and the frequency domain characteristic information are obtained by segmenting and processing the output voltage waveform of the circuit, when the microparticles pass between the electrodes, a wave peak or a wave trough appears in the characteristic information, the extreme points corresponding to the wave peak or the wave trough are taken as the characteristic values of the microparticles, and the number of the extreme points corresponds to the number of the microparticles, which can be used for counting; The characteristic values of the microparticles and the corresponding microparticle labels are divided into a training sample set and a verification sample set, and a microparticle prediction model is obtained by training a machine learning model based on the training data set; In application, the extreme points of the time domain characteristic information and / or the frequency domain characteristic information of the microparticles are input into the microparticle detection model to obtain the category prediction value, the clinical cycle prediction value, the variant prediction value or the different state prediction value of the microparticles, wherein the different state is the nascent stage, the growth stage or the apoptosis stage of the biological microparticles when the microparticles are biological microparticles.
2. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 1, wherein, The method for counting and flux calculation of m kinds of microparticles comprises the following steps: The m kinds of particles are respectively tested by the multi-modal flow particle detection method based on the PT symmetric circuit, corresponding training sample sets are obtained, machine learning training is performed, a particle prediction model is obtained, and then the obtained particle prediction model is applied to sample detection containing m kinds of particles at the same time, the feature change caused by each particle when passing through the electrode is input into the particle prediction model, the type is predicted, and then the number n of the i-th kind of particle in the time Δt is counted i ; The detection flux corresponding to the ith kind of particle is calculated as K i = n i / Δt, wherein n i (i=1-m) is the number of peaks or troughs corresponding to the ith kind of particle in the time Δt. The total flux of the particles is K, K =∑n i / Δt.
3. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 1, wherein, The time domain characteristic information is obtained by segmenting and processing the voltage waveform, which comprises the following steps: The segmented voltage waveform is analyzed by using a data processing software such as matlab or python, and one or more of the standard deviation, the kurtosis, the envelope frequency, the envelope peak value, the envelope valley value and the envelope amplitude information of the voltage waveform at different times are used to construct the time domain characteristic information.
4. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 1, wherein, The frequency domain characteristic information is obtained by segmenting and processing the voltage waveform, which comprises the following steps: The segmented voltage waveform is subjected to Fourier transform, and one or more of the frequency shift and the amplitude change information of the microparticles compared with the eigenmode are used to construct the frequency domain characteristic information.
5. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 1, wherein, The time for a single microparticle to pass between the electrodes is 0.1-10 ms.
6. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 1, wherein, The PT symmetric circuit is composed of LC circuits with gain and loss, two LC circuits are connected through capacitive coupling, inductive coupling or equivalent transmission line, and the parameters of the two LC circuits satisfy the PT symmetric condition.
7. The multi-modal flow microparticle detection method based on the PT symmetric circuit according to claim 1 or 6, wherein the voltage waveform information is obtained by reading the voltage signal of the LC circuit with gain or loss of the PT symmetric circuit through one of an oscilloscope, an FPGA circuit and a single-chip microcomputer circuit.
8. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 1, wherein, The microfluidic chip comprises: a substrate; a microchannel located on the substrate, the microchannel has a groove, and a measurement position is arranged between the front and back ends of the groove, microparticles flow from one end of the groove to the other end and pass through the measurement position; and a conductive electrode located on the substrate and connected with the PT symmetric circuit, the conductive electrode is one or more pairs of electrodes, the end portions of the two electrodes in each pair of electrodes are arranged left and right or on the same side at the measurement position, and are used for measuring the impedance when the microparticles pass.
9. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 2, wherein, The method for obtaining the size of the microparticles comprises the following steps: When the ratio of the distance between the electrodes, the pipe cross-sectional size and the size of the particles is less than 2, the bolus flow rate Q pushed by the injection pump, and the cross-sectional area S near the electrodes, the flow rate V of the particles passing through the electrodes is calculated = Q / S, and further combined with the time t of the particles passing through the electrodes, the particle diameter size X = V·t = Q·t / S is calculated, and a coefficient a is introduced for correction, so the final particle diameter calculation method is X = V·a·t = Q·a·t / S.
10. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 1, wherein, The machine learning model is trained by a training sample set to obtain a particle detection model, the convolutional neural network, the machine learning model includes a convolutional neural network, a decision tree, an xgboost or a KNN algorithm.
11. The PT-symmetric circuit based multi-modal streaming microparticle detection method of claim 1, wherein, The multi-modal flow particle detection method distinguishes the number (count), size and type of the particles including cells, bacteria, proteins, exosomes or micro-plastic particles, when the particles refer to cells, the method can further distinguish the varieties, clinical cycles and cell cycles of the cells; when the particles refer to bacteria, the method can further distinguish the growth states of the bacteria.
