Non-labeled electrical impedance flow cytometer and detection method for circulating tumor cells

Abstract

The application discloses a kind of non-marked impedance flow cytometry and detection method for circulating tumor cell, including including sampling system, impedance spectrometer, sampling control system and computer, sampling system includes sampling pump, switch valve connected with sampling pump, first centrifugal tube with cleaning solution being connected with the first liquid inlet of switch valve, second centrifugal tube with the sample to be measured being connected with the second liquid inlet of switch valve, third centrifugal tube with standard sample being connected with the third liquid inlet of switch valve, microfluidic chip being connected with the liquid outlet of switch valve and waste liquid collection bottle being connected with microfluidic chip, sampling control system includes stepper motor drive board and control board, control board is connected with switch valve and stepper motor drive board respectively, impedance spectrometer applies multiple frequency excitation signal to microfluidic chip, microfluidic chip exports feedback current signal to impedance spectrometer, and impedance spectrometer exports original signal to computer for signal processing and identifies cell type.

Description

Technical Field

[0001] This invention relates to the field of label-free cell detection, and more particularly to a label-free electrical impedance flow cytometer and detection method for circulating tumor cells. Background Technology

[0002] Circulating tumor cells (CTCs) serve as "seeds" for cancer metastasis and invasion, possessing significant scientific and clinical research value. Currently, widely used methods for sorting, counting, and analyzing CTCs often rely on immunolabeling, which fails to obtain viable CTCs, hindering further analysis of their biophysical characteristics such as drug resistance and environmental adaptation. Therefore, developing label-free detection and species identification techniques for obtaining viable CTCs is crucial. The dielectric properties of cells, as an important biophysical characteristic, can serve as a vital parameter for cell counting and species identification.

[0003] Existing technology CN113866074A discloses a microfluidic chip for white blood cell classification and counting based on position-compensated impedance method. The microfluidic chip includes a fluid channel filled with an electrolyte solution, and seven electrodes symmetrically distributed at equal intervals at the bottom of the fluid channel. The microfluidic chip employs a dual-differential structure. This position-compensated impedance-based white blood cell classification and counting microfluidic chip integrates dual-differential electrode structures and features high throughput, micro-sample handling, label-free operation, and high sensitivity. It can quickly and accurately count and classify white blood cells. Furthermore, its small size and high integration allow for lightweight and portable manufacturing. Operation is simple and time-saving, requiring no professional skills or additional medical equipment. Even in remote and impoverished areas with limited medical resources, timely testing can be achieved in routine diagnoses, eliminating errors from manual counting and reducing the risk of infection. However, the aforementioned existing technology cannot meet diverse testing needs and has poor versatility. In addition, existing cell analyzers are prone to contamination of the sample solution from the previous test with the sample solution from the next test, thus affecting the accuracy and results of the test. Summary of the Invention

[0004] Purpose of the invention: The first objective of this invention is to provide a label-free impedance flow cytometer for circulating tumor cells that uses flow cytometry AC impedance signals at multiple frequencies to identify cell types.

[0005] A second objective of this invention is to provide a label-free electrical impedance flow cytometry method for detecting circulating tumor cells.

[0006] Technical Solution: To achieve the above objectives, this invention discloses a label-free electrical impedance flow cytometer for circulating tumor cells, comprising a sample introduction system, an impedance spectrometer, a sample introduction control system, and a computer. The sample introduction system includes a sample pump, a switching valve connected to the sample pump, a first centrifuge tube containing a washing solution connected to a first inlet of the switching valve, a second centrifuge tube containing a sample to be tested connected to a second inlet of the switching valve, a third centrifuge tube containing a standard sample connected to a third inlet of the switching valve, a microfluidic chip connected to the outlet of the switching valve, and a waste collection bottle connected to the microfluidic chip. The sample introduction control system includes a stepper motor drive board and a control board for driving the sample pump. The control board is connected to the switching valve and the stepper motor drive board, respectively. The impedance spectrometer applies multi-frequency excitation signals to the microfluidic chip, the microfluidic chip outputs feedback current signals to the impedance spectrometer, and the impedance spectrometer outputs raw signals to the computer for signal processing to identify cell types.

[0007] Furthermore, the multi-frequency excitation signal is a hybrid multi-frequency AC signal with a frequency range of 500kHz to 50MHz, and the number of selected frequencies is no more than 8.

[0008] Furthermore, the microfluidic chip includes a glass substrate, and a first electrode, a second electrode, a third electrode, a fourth electrode, and a fifth electrode integrated on the glass substrate and coplanar.

[0009] Preferably, the microfluidic chip has a three-electrode detection mode and a five-electrode detection mode. In the three-electrode detection mode, a multi-frequency excitation signal is applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are not connected to an external signal source. In the five-electrode detection mode, a multi-frequency excitation signal is applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are connected to an external electrical signal with the opposite sign to that on the third electrode.

[0010] Furthermore, the microfluidic chip also includes a microfluidic channel, which is an asymmetric sinusoidal channel. The length L of the detection channel in the asymmetric sinusoidal channel is greater than the electrode distribution width at the bottom of the microfluidic channel. The width W of the detection channel is equal to the width of the channel with small curvature in the asymmetric sinusoidal channel. The height H of the detection channel is greater than the maximum diameter of the cell being measured.

[0011] Furthermore, the control board features cleaning, calibration, and detection modes.

[0012] In cleaning mode, the control board controls the switching valve to switch to the first inlet. The control board drives the stepper motor to control the injection pump to draw the cleaning solution in the first centrifuge tube and pump it to the switching valve. The switching valve switches to the outlet. The cleaning solution flows to the microfluidic chip through the liquid pipeline to clean the microfluidic chip.

[0013] In calibration mode, the control board controls the switching valve to switch to the third injection port. The control board drives the stepper motor to control the injection pump to draw the standard solution in the third centrifuge tube and pump it to the switching valve. The switching valve switches to the outlet. The standard solution flows to the microfluidic chip through the liquid pipeline. A multi-frequency excitation signal is applied to the microfluidic chip. The microfluidic chip outputs a feedback current signal to the impedance spectrometer. The impedance spectrometer outputs the raw signal to the computer as the reference value of the cell electrical impedance signal.

[0014] In detection mode, the control board controls the switching valve to switch to the second sample inlet. The control board drives the stepper motor to control the sample pump to draw the test solution in the second centrifuge tube and pump it to the switching valve. The switching valve switches to the liquid outlet. The test solution flows to the microfluidic chip through the liquid path. A multi-frequency excitation signal is applied to the microfluidic chip. The microfluidic chip outputs a feedback current signal to the impedance spectrometer. The impedance spectrometer outputs the raw signal to the computer for signal processing to identify cell types.

[0015] Furthermore, it also includes a power supply module and a DC-DC converter module for power conversion. The power supply module is connected to the computer, impedance spectrometer and DC-DC converter module respectively, and the DC-DC converter module is connected to the control board, stepper motor drive board, sample pump and switching valve respectively.

[0016] This invention discloses a label-free electrical impedance flow cytometry detection method for circulating tumor cells, comprising the following steps:

[0017] When in three-electrode detection mode

[0018] The control board controls the switching valve to switch to the second sample inlet. The control board controls the sample pump to draw the solution to be tested in the second centrifuge tube and pump it to the switching valve via the stepper motor drive board. The switching valve switches to the liquid outlet and the solution to be tested flows to the microfluidic chip through the liquid pipeline.

[0019] Multi-frequency excitation signals are applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are not connected to external signal sources. The impedance spectrometer outputs the raw signal to the computer. The computer first performs noise reduction processing on the raw signal, flattens the signal baseline, extracts the cell impedance signal, and identifies the cell type based on the convolutional neural network model.

[0020] In the convolutional neural network model, the number of alternating current frequencies applied to the third electrode is k. A single cell undergoes n signal samplings while passing through the impedance detection region at a fixed flow rate. The cell impedance signals intercepted at each frequency are combined into a 2k x n matrix. The first row of the matrix is ​​the real part of the cell AC impedance signal at the first frequency, the second row is the imaginary part of the cell AC impedance signal at the first frequency, and so on, with the 2kth row being the imaginary part of the cell AC impedance signal at the kth frequency. After performing convolution calculation on this matrix, the result is fed into a max pooling layer, then into a fully connected neural network, and finally the cell type information is output.

[0021] This invention discloses a label-free electrical impedance flow cytometry detection method for circulating tumor cells, comprising the following steps:

[0022] When in five-electrode detection mode

[0023] The control board controls the switching valve to switch to the second sample inlet. The control board controls the sample pump to draw the solution to be tested in the second centrifuge tube and pump it to the switching valve via the stepper motor drive board. The switching valve switches to the liquid outlet and the solution to be tested flows to the microfluidic chip through the liquid pipeline.

[0024] A multi-frequency excitation signal is applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are connected to electrical signals with opposite signs to those on the third electrode. The impedance spectrometer outputs the raw signal to the computer, which first performs noise reduction on the raw signal, flattens the signal baseline, extracts the cell impedance signal, and identifies the cell type based on the convolutional neural network model.

[0025] In the convolutional neural network model, the number of alternating current frequencies applied to the third electrode is k. A single cell undergoes n signal samplings while passing through the impedance detection region at a fixed flow rate. The cell impedance signals intercepted at each frequency are combined into a 2k x n matrix. The first row of the matrix is ​​the real part of the cell AC impedance signal at the first frequency, the second row is the imaginary part of the cell AC impedance signal at the first frequency, and so on, with the 2kth row being the imaginary part of the cell AC impedance signal at the kth frequency. After performing convolution calculation on this matrix, the result is fed into a max pooling layer, then into a fully connected neural network, and finally the cell type information is output.

[0026] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0027] (1) The present invention does not require the application of immunofluorescence labeling technology for the identification of cell types, and it is easy to obtain active cells after detection for further scientific research.

[0028] (2) The cell impedance detection microfluidic chip in this invention can realize single-stage differential impedance signal detection of cells based on three electrodes or double-stage differential impedance signal detection based on five electrodes. These two detection modes can be easily switched. Single-stage differential detection based on three electrodes is suitable for high-throughput, high-cell-concentration sample detection. Double-stage differential detection based on five electrodes meets the detection requirements of high sensitivity, high signal-to-noise ratio, and low-concentration cell samples.

[0029] (3) The cell impedance detection microfluidic chip based on asymmetric sinusoidal flow channel in this invention can realize high-throughput single-row passive focusing of cells in the flow channel to improve the detection accuracy of cell impedance signals;

[0030] (4) The cell species identification algorithm based on cell impedance detection signal in this invention can effectively improve the generalization ability of the cell species identification model; it is less affected by the processing error of the impedance detection chip. Attached Figure Description

[0031] Figure 1(a) is a front view of the present invention;

[0032] Figure 1(b) is a schematic diagram of the internal structure of the present invention;

[0033] Figure 1(c) is a schematic diagram of the internal structure of the present invention. Figure 2 ;

[0034] Figure 2 This is a schematic diagram of the present invention;

[0035] Figure 3(a) is a schematic diagram of the structure of the microfluidic chip and fixture in this invention;

[0036] Figure 3(b) is a schematic diagram of the assembly of the microfluidic chip and the fixture in this invention;

[0037] Figure 4 This is a schematic diagram of the electrode arrangement in the impedance detection area of ​​the present invention;

[0038] Figure 5 This is a schematic diagram of the asymmetric sinusoidal flow channel in this invention;

[0039] Figure 6 This is a schematic diagram of the working principle of the liquid circuit of the sample introduction system in this invention;

[0040] Figure 7 This is a comparison diagram of the detection performance of the planar three-electrode and five-electrode structures of this invention;

[0041] Figure 8 This is a diagram illustrating the matrixing process of the electrical impedance signal of a single cell in this invention;

[0042] Figure 9This is a diagram of the artificial neural network structure for cell species identification based on convolution algorithm in this invention.

[0043] Figure 10 This is a schematic diagram illustrating the performance of the cell species identification model based on artificial neural networks in this invention. Detailed Implementation

[0044] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0045] The present invention discloses a label-free electrical impedance flow cytometer for circulating tumor cells, comprising an instrument housing 5, a sample introduction system, an impedance spectrometer 6, a transimpedance amplifier 7, a sample introduction control system, and a computer 8. The sample introduction system includes the housing 5, a sample pump 10, a switching valve 11, a first centrifuge tube 2, a second centrifuge tube 3, a third centrifuge tube 4, microfluidic chips 1-3, and a waste collection bottle 9. The sample introduction control system includes a stepper motor drive board 13, a control board 15, a power module 12, and a DC-DC converter module 14. Switching valve 11 is connected to sample injection pump 10. The first centrifuge tube 2 contains a cleaning solution, which can be a phosphate buffer solution. The first centrifuge tube 2 is connected to the first inlet P1 of switching valve 11. The second centrifuge tube 3 contains the sample to be tested and is connected to the second inlet P3 of switching valve 11. The third centrifuge tube 4 contains a standard sample, which can be a phosphate buffer solution containing dissolved hyaluronic acid and suspended standard micron-sized particles. The third centrifuge tube 4 is connected to the third inlet P2 of the switching valve. The outlet P4 of switching valve 11 is connected to microfluidic chip 1-3, and a waste collection bottle 9 is also connected to microfluidic chip 1-3. Control board 15 is connected to switching valve 11 and stepper motor drive board 13, which is connected to sample injection pump 10 to drive the pump. The power module is connected to the computer 8, impedance spectrometer 6, and DC-DC converter 14. The DC-DC converter 14 is connected to the control board 15, stepper motor drive board 13, sample pump 10, and switching valve 11. The power module serves as the main power interface, and the entire instrument is powered by this interface. The sample pump 10 provides positive or negative pressure to the liquid path of the sample introduction system to push or aspirate the liquid sample. The rotation of the valve body in the switching valve 11 switches the liquid flow path. Because the sample pump and switching valve have narrow gaps that are difficult to clean and dead volumes that are difficult to empty, the sample storage tubing of this invention is a section with an internal volume larger than the maximum sample pumping capacity to avoid contact with samples containing cells. When the liquid sample in the storage tubing is pumped into the microfluidic chip, the switching valve switches to the outlet P4 position, forming a fluid path between the sample pump, switching valve, and microfluidic chip, while other valve ports remain closed, without affecting the ongoing pumping action. Taking the cleaning mode as an example, the operation logic of the injection pump and the switching valve is as follows: (1) The switching valve switches to the first inlet P1 → (2) The injection pump draws a fixed amount of cleaning solution → (3) The switching valve switches to the outlet P4 → (4) The injection pump delivers the cleaning solution at the set speed. Other modes follow similar operation logic, which will not be elaborated here.

[0046] Impedance spectrometer 6 applies multi-frequency excitation signals to microfluidic chip 1-3. Microfluidic chip 1-3 outputs feedback current signals to impedance spectrometer 6. Impedance spectrometer 6 outputs raw signals to computer 8 for signal processing to identify cell types. The multi-frequency excitation signal is a mixed multi-frequency AC signal with a frequency range of 500kHz to 50MHz. The number of selected frequencies is no more than 8. Cell types are identified using the flow cytometry AC impedance signals of cells at multiple frequencies. Microfluidic chip 1-3 is fixed by clamp 1-1 and mounted on the upper cover of instrument housing 5. It is connected to the internal detection circuit of the instrument via spring pins 1-2. The detection circuit includes an impedance spectrometer and a differential amplifier. Microfluidic chip 1-3 and clamp 1-1 at the upper cover of the instrument can be opened without disassembly to replace microfluidic chip 1-3, as shown in Figure 3. The microfluidic chip 1-3 includes a glass substrate, a first electrode 1-3-1, a second electrode 1-3-2, a third electrode 1-3-3, a fourth electrode 1-3-4, and a fifth electrode 1-3-5 integrated on the glass substrate and coplanar.

[0047] A schematic diagram of the electrode arrangement in the impedance detection region near the outlet of a microfluidic chip for microcellular impedance detection is shown below. Figure 4 As shown, five coplanar microelectrodes are integrated on the glass substrate of the microfluidic chip 1-3, enabling two detection modes for cell impedance signals: a three-electrode detection mode and a five-electrode detection mode. In the three-electrode detection mode, a multi-frequency excitation signal is applied to the third electrode 1-3-3. The feedback current signals output from the second electrode 1-3-2 and the fourth electrode 1-3-4 are transmitted through a differential amplifier 7, and then phase-locked by an impedance spectrometer to extract specific frequency signals, which are then amplified and input into the impedance spectrometer 6. The first electrode 1-3-1 and the fifth electrode 1-3-5 are not connected to an external signal source. The three-electrode detection mode is suitable for high-throughput, high-cell-concentration sample detection. High throughput refers to greater than 100 μL / min, and high cell concentration refers to greater than 10 μL / min. 5 cells / mL. Differential amplification of the feedback current signals from the second electrode 1-3-2 and the fourth electrode 1-3-4 allows for a high signal-to-noise ratio in the cell impedance detection signal. Because the cell impedance detection region is small and occupies a short length within the flow channel in this detection mode, it is suitable for high-throughput, high-cell-concentration sample detection.

[0048] In the five-electrode detection mode, a multi-frequency excitation signal is applied to the third electrode 1-3-3. The second electrode 1-3-2 and the fourth electrode 1-3-4 output feedback current signals to the impedance spectrometer 6. The first electrode 1-3-1 and the fifth electrode 1-3-5 are externally connected to electrical signals with opposite signs to those on the third electrode 1-3-3. The five-electrode detection mode has a higher detection signal-to-noise ratio, suitable for the detection needs of high sensitivity, high signal-to-noise ratio, and low-concentration cell samples. High sensitivity refers to a cell volume fraction >320ppm in the detection area, and low-concentration cell samples refer to less than 10... 5 The cells / mL. The AC signals with opposite signs to those on the third electrode 1-3-3, connected externally to the first electrode 1-3-1 and the fifth electrode 1-3-5, can be achieved using a dedicated passive inverter. The main feature of the five-electrode detection mode is the passive second-order differential division of the current signal. This means that the five-electrode detection mode eliminates the need for other electrical components to achieve second-order differential division, resulting in a higher signal-to-noise ratio cell impedance signal, suitable for detecting high-sensitivity and low-concentration cell samples.

[0049] This invention selects the same microfluidic chip with a five-electrode structure and connects it to the detection circuit. For example... Figure 7 As shown in (a), the inverted signal that should have been applied to the two edge electrodes is cut off. At this point, the detection circuit is equivalent to a single-shot differential impedance detection method based on a three-electrode structure. A sinusoidal AC signal with a frequency of 500 kHz and a peak-to-peak voltage of 1 V is applied to the middle electrode. Standard polystyrene particles with a diameter of 4 μm are introduced into the microfluidic chip at a flow rate of 80 μL / min. After recording a certain amount of raw data, the inverted signal on the edge electrodes is turned on under the same conditions, as shown in (a). Figure 7 As shown in (b), the raw data of the impedance signal is saved and recorded as another set of data. Figure 7 The vertical axis of the data graph has been converted to arbitrary units (au) using the same method to compare the differences between the two detection methods.

[0050] Depend on Figure 7 As shown in the signal graph of 4μm particles in (a), the single-shot differential impedance detection method based on the three-electrode structure failed to detect a significant peak. Although this figure only shows the signal within a randomly selected 75ms interval, this invention examined the signal over a time span of approximately 30s under the same detection conditions, and no signal peaks clearly consistent with the characteristics of 4μm particles were found. In this case, the signal noise range is approximately ±300 A.u. Therefore, it can be concluded that the single-shot differential impedance detection method based on the three-electrode structure cannot effectively detect particle signals with a diameter no greater than 4μm. Figure 7 As shown in (b), the signal peak of 4μm particles can be clearly identified in the particle impedance second differential detection mode based on the five-electrode structure.

[0051] In summary, the instrument involved in this invention can freely switch between three-electrode and five-electrode detection modes, and the signal-to-noise ratio of the secondary differential impedance detection based on the five-electrode structure is higher than that of the single differential impedance detection mode based on the three-electrode structure. However, the length of the detection area in both detection modes determines that the five-electrode structure is only suitable for detecting samples with low cell concentrations; otherwise, multiple cells may enter the detection area simultaneously, which is not conducive to data decoupling and analysis.

[0052] The microfluidic chip 1-3 also includes a microfluidic channel, which is an asymmetric sinusoidal channel. The length L of the detection channel in the asymmetric sinusoidal channel is greater than the electrode distribution width at the bottom of the microfluidic channel. The width W of the detection channel is equal to the width of the channel with small curvature in the asymmetric sinusoidal channel. The height H of the detection channel is greater than the maximum diameter of the measured cell. This invention utilizes the coupling effect of inertial and elastic forces in the asymmetric sinusoidal channel to form a single-row focus of cells 1-3-6 within the microfluidic channel, thereby improving the detection accuracy of cell impedance signals. The cell sample suspension is a phosphate buffer solution dissolving hyaluronic acid (HA), a non-Newtonian fluid. Under the coupling effect of inertial and elastic forces in the flow field within the channel, cells can achieve precise single-row focus within the channel, resulting in good uniformity of the impedance signal of the standard polystyrene microspheres, providing the necessary conditions for the accurate acquisition of subsequent cell impedance signals. The dimensional and structural characteristics of the asymmetric sinusoidal channel and the forces exerted by its internal flow field on the cells can be found in the literature. [1,2] The length L of the detection channel should be greater than the width of the electrode distribution at its bottom. The width W should be comparable to the width of the channel in the smaller curvature section of the asymmetric sinusoidal channel. The height H should be slightly larger than the maximum diameter of the cell being measured, typically 30–40 micrometers.

[0053] In this invention, the control board 15 has cleaning mode, calibration mode, and detection mode. The microcontroller-based control board 15 is responsible for receiving instructions from the computer and reporting the status of the operation to the host computer. The computer transmits the mode to be executed (cleaning mode, calibration mode, or detection mode) to the control board 15, which then converts the specific instructions and transmits them to the stepper motor drive board 13 or the switching valve 11. The stepper motor drive board 13 controls the speed and rotation amount of the sample injection pump 10.

[0054] In cleaning mode, the control board controls the switching valve to switch to the first inlet. The control board drives the stepper motor to control the injection pump to draw the cleaning solution in the first centrifuge tube and pump it to the switching valve. The switching valve switches to the outlet. The cleaning solution flows to the microfluidic chip through the liquid pipeline to clean the microfluidic chip.

[0055] In calibration mode, the control board controls the switching valve to switch to the third injection port. The control board, through the stepper motor drive board, controls the injection pump to draw the standard solution in the third centrifuge tube and pump it to the switching valve. The switching valve switches to the outlet, and the standard solution flows to the microfluidic chip through the liquid path. A multi-frequency excitation signal is applied to the microfluidic chip, and the microfluidic chip outputs a feedback current signal to the impedance spectrometer. The impedance spectrometer outputs the raw signal to the computer, which serves as the reference value for the cell impedance signal in calibration mode. In calibration mode, the specific role of the particle impedance data is as follows: Under low-frequency alternating current excitation (frequency not higher than 500kHz), the peak value P of the detection current fluctuation caused by the particle or cell passing over the microelectrode is directly proportional to its volume V, P = k·V. The low-frequency impedance data of the particles obtained in calibration mode can be used to calculate this proportionality coefficient k, which is then used in the subsequent rapid calculation of cell diameter.

[0056] In detection mode, the control board controls the switching valve to switch to the second sample inlet. The control board drives the stepper motor to control the sample pump to draw the test solution in the second centrifuge tube and pump it to the switching valve. The switching valve switches to the liquid outlet. The test solution flows to the microfluidic chip through the liquid path. A multi-frequency excitation signal is applied to the microfluidic chip. The microfluidic chip outputs a feedback current signal to the impedance spectrometer. The impedance spectrometer outputs the raw signal to the computer for signal processing to identify cell types.

[0057] This invention provides a label-free electrical impedance flow cytometry detection method for circulating tumor cells, comprising the following steps:

[0058] When in three-electrode detection mode

[0059] The control board controls the switching valve to switch to the second sample inlet. The control board controls the sample pump to draw the solution to be tested in the second centrifuge tube and pump it to the switching valve via the stepper motor drive board. The switching valve switches to the liquid outlet and the solution to be tested flows to the microfluidic chip through the liquid pipeline.

[0060] Multi-frequency excitation signals are applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are not connected to external signal sources. The impedance spectrometer outputs the raw signal to the computer. The computer first performs noise reduction processing on the raw signal, flattens the signal baseline, extracts the cell impedance signal, and identifies the cell type based on the convolutional neural network model.

[0061] The cell species identification algorithm is a deep learning algorithm based on convolutional neural networks. In the convolutional neural network model, the number of alternating current frequencies applied to the third electrode is k. A single cell undergoes n signal samplings while passing through the impedance detection area at a fixed flow rate. The cell impedance signals intercepted at each frequency are combined into a 2k x n matrix. In the matrix, the first row is the real part of the cell AC impedance signal at the first frequency, the second row is the imaginary part of the cell AC impedance signal at the first frequency, and so on, with the 2kth row being the imaginary part of the cell AC impedance signal at the kth frequency. After performing convolution calculation on this matrix, the result is fed into a max pooling layer, then into a fully connected neural network, and finally the cell species information is output.

[0062] This invention provides a label-free electrical impedance flow cytometry detection method for circulating tumor cells, comprising the following steps:

[0063] When in five-electrode detection mode

[0064] The control board controls the switching valve to switch to the second sample inlet. The control board controls the sample pump to draw the solution to be tested in the second centrifuge tube and pump it to the switching valve via the stepper motor drive board. The switching valve switches to the liquid outlet and the solution to be tested flows to the microfluidic chip through the liquid pipeline.

[0065] A multi-frequency excitation signal is applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are externally connected to electrical signals with opposite signs to those on the third electrode. The impedance spectrometer outputs the raw signal to a computer, which first performs noise reduction, baseline flattening, and cell impedance signal extraction. Noise reduction: Wavelet decomposition algorithm is used to remove high-frequency noise (>100kHz) from the signal. Baseline flattening: To remove the jitter of the signal baseline over time, wavelet decomposition algorithm is used to remove low-frequency components (<1000Hz) from the signal. Cell impedance signal extraction: Peaks are found in the denoised and baseline-flattened signal, and the peak values ​​are recorded as the cell's electrical signal. Cell types are identified based on a convolutional neural network model.

[0066] In the convolutional neural network model, the number of alternating current frequencies applied to the third electrode is k. A single cell undergoes n signal samplings while passing through the impedance detection region at a fixed flow rate. The cell impedance signals intercepted at each frequency are combined into a 2k x n matrix. The first row of the matrix is ​​the real part of the cell AC impedance signal at the first frequency, the second row is the imaginary part of the cell AC impedance signal at the first frequency, and so on, with the 2kth row being the imaginary part of the cell AC impedance signal at the kth frequency. After performing convolution calculation on this matrix, the result is fed into a max pooling layer, then into a fully connected neural network, and finally the cell type information is output.

[0067] The cell species identification algorithm based on cell electrical impedance detection signals involved in this invention can effectively improve the generalization ability of cell species identification models. The cell electrical impedance signals are matrixed. Assume that the AC frequencies used in the multi-frequency cell electrical impedance detection are aMHz, bMHz, cMHz, and dMHz. The signal is a vector with a length of n sampling points. If it precisely contains the signal fluctuations caused by a single cell passing through the impedance detection region, then record them separately. and Let be the amplitude and phase vector of the impedance signal at aMHz. Then, similar to how pixels in an image form a matrix, the impedance signal of a single cell can be represented as follows: Figure 8 The matrix form shown.

[0068] For the convolution operation of matrix C, the embodiments of the present invention select 8 convolution kernels. The initial values ​​of the elements in K are set to random numbers. The convolution operation has a stride of 2 in both directions of matrix C. The element values ​​in each convolution kernel K will be adjusted to suitable values ​​by the optimization algorithm as the number of training iterations increases. After the convolution operation, the result is fed into a 2×2 max-pooling layer. Then, the max-pooling result is reconstructed into a one-dimensional vector l and fed into a fully connected neural network. The number of neurons in each layer of the fully connected network is 120, 512, 512, 120, and 8 respectively, and the cell type is output. The overall neural network uses the cross-entropy classification loss function, and the optimization function is the Root Mean Square Propagation (RMSProp) method with a learning rate lr = 0.0005. The activation function of the hidden layers is the ReLU function, and the activation function of the output layer is the Softmax function. The structure of the entire neural network is as follows: Figure 9 As shown.

[0069] like Figure 10 As shown, this embodiment focuses on training a cell species identification model with strong generalization ability using multi-chip data, and focuses on distinguishing WBCs from tumor cells. Electrical impedance data of 452,634 cells measured from four different batches of microfluidic chips were fed into the neural network for training, and a confusion matrix of cell species identification accuracy on the test set was obtained. The confusion matrix shows that the neural network has good generalization ability.

[0070] This invention provides a label-free electrical impedance flow cytometer and detection method for circulating tumor cells. The above description is only a preferred embodiment of this invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A non-labeled electrical impedance flow cytometer for circulating tumor cells, characterized in that, The system includes a sample introduction system, an impedance spectrometer, a sample introduction control system, and a computer. The sample introduction system comprises a sample pump, a switching valve connected to the sample pump, a first centrifuge tube containing a cleaning solution connected to the first inlet of the switching valve, a second centrifuge tube containing the sample to be tested connected to the second inlet of the switching valve, a third centrifuge tube containing a standard sample connected to the third inlet of the switching valve, a microfluidic chip connected to the outlet of the switching valve, and a waste collection bottle connected to the microfluidic chip. The sample introduction control system includes a stepper motor driver board and a control board for driving the sample pump. The control board is connected to the switching valve and the stepper motor driver board, respectively. The impedance spectrometer applies multi-frequency excitation signals to the microfluidic chip, and the microfluidic chip outputs a feedback current signal to the impedance spectrometer. The instrument, an impedance spectrometer, outputs raw signals to a computer for signal processing to identify cell types. The microfluidic chip includes a glass substrate and five electrodes integrated on the glass substrate and coplanar. The microfluidic chip has a three-electrode detection mode and a five-electrode detection mode. In the three-electrode detection mode, a multi-frequency excitation signal is applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are not connected to an external signal source. In the five-electrode detection mode, a multi-frequency excitation signal is applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are connected to an external electrical signal with the opposite sign to that on the third electrode.

2. The non-labeled electrical impedance flow cytometer for circulating tumor cells according to claim 1, wherein: The multi-frequency excitation signal is a hybrid multi-frequency AC signal with a frequency range of 500kHz to 50MHz, and the number of selected frequencies is no more than 8.

3. The non-labeled electrical impedance flow cytometer for circulating tumor cells according to claim 1, wherein: The microfluidic chip also includes a microfluidic channel, which is an asymmetric sinusoidal channel. The length L of the detection channel in the asymmetric sinusoidal channel is greater than the electrode distribution width at the bottom of the microfluidic channel. The width W of the detection channel is equal to the width of the channel with small curvature in the asymmetric sinusoidal channel. The height H of the detection channel is greater than the maximum diameter of the cell being measured.

4. The label-free electrical impedance flow cytometer for circulating tumor cells according to claim 1, characterized in that: The control board has cleaning mode, calibration mode and detection mode. In cleaning mode, the control board controls the switching valve to switch to the first inlet. The control board drives the stepper motor to control the injection pump to draw the cleaning solution in the first centrifuge tube and pump it to the switching valve. The switching valve switches to the outlet. The cleaning solution flows to the microfluidic chip through the liquid pipeline to clean the microfluidic chip. In calibration mode, the control board controls the switching valve to switch to the third injection port. The control board drives the stepper motor to control the injection pump to draw the standard solution in the third centrifuge tube and pump it to the switching valve. The switching valve switches to the outlet. The standard solution flows to the microfluidic chip through the liquid pipeline. A multi-frequency excitation signal is applied to the microfluidic chip. The microfluidic chip outputs a feedback current signal to the impedance spectrometer. The impedance spectrometer outputs the raw signal to the computer as the reference value of the cell electrical impedance signal. In detection mode, the control board controls the switching valve to switch to the second sample inlet. The control board drives the stepper motor to control the sample pump to draw the test solution in the second centrifuge tube and pump it to the switching valve. The switching valve switches to the liquid outlet. The test solution flows to the microfluidic chip through the liquid path. A multi-frequency excitation signal is applied to the microfluidic chip. The microfluidic chip outputs a feedback current signal to the impedance spectrometer. The impedance spectrometer outputs the raw signal to the computer for signal processing to identify cell types.

5. The label-free electrical impedance flow cytometer for circulating tumor cells according to claim 1, characterized in that: It also includes a power supply module and a DC-DC converter module for power conversion. The power supply module is connected to the computer, impedance spectrometer and DC-DC converter module respectively. The DC-DC converter module is connected to the control board, stepper motor drive board, sample pump and switching valve respectively.

6. A method for detecting circulating tumor cells using a label-free electrical impedance flow cytometer according to any one of claims 1 to 5, characterized in that, The steps include the following: When in three-electrode detection mode The control board controls the switching valve to switch to the second sample inlet. The control board controls the sample pump to draw the solution to be tested in the second centrifuge tube and pump it to the switching valve via the stepper motor drive board. The switching valve switches to the liquid outlet and the solution to be tested flows to the microfluidic chip through the liquid pipeline. Multi-frequency excitation signals are applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are not connected to external signal sources. The impedance spectrometer outputs the raw signal to the computer. The computer first performs noise reduction processing on the raw signal, flattens the signal baseline, extracts the cell impedance signal, and identifies the cell type based on the convolutional neural network model. In the convolutional neural network model, the number of alternating current frequencies applied to the third electrode is k. A single cell undergoes n signal samplings while passing through the impedance detection region at a fixed flow rate. The cell impedance signals intercepted at each frequency are combined into a 2k x n matrix. The first row of the matrix is ​​the real part of the cell AC impedance signal at the first frequency, and the second row is the imaginary part of the cell AC impedance signal at the first frequency. Similarly, the 2kth row is the imaginary part of the cell AC impedance signal at the kth frequency; after performing convolution calculation on this matrix, the result is fed into a max pooling layer, then into a fully connected neural network, and finally the cell type information is output.

7. A method for detecting circulating tumor cells using a label-free electrical impedance flow cytometer according to any one of claims 1 to 5, characterized in that, The steps include the following: When in five-electrode detection mode The control board controls the switching valve to switch to the second sample inlet. The control board controls the sample pump to draw the solution to be tested in the second centrifuge tube and pump it to the switching valve via the stepper motor drive board. The switching valve switches to the liquid outlet and the solution to be tested flows to the microfluidic chip through the liquid pipeline. A multi-frequency excitation signal is applied to the third electrode, and the second and fourth electrodes output feedback current signals to the impedance spectrometer. The first and fifth electrodes are connected to electrical signals with opposite signs to those on the third electrode. The impedance spectrometer outputs the raw signal to the computer, which first performs noise reduction on the raw signal, flattens the signal baseline, extracts the cell impedance signal, and identifies the cell type based on the convolutional neural network model. In the convolutional neural network model, the number of alternating current frequencies applied to the third electrode is k. A single cell undergoes n signal samplings while passing through the impedance detection region at a fixed flow rate. The cell impedance signals intercepted at each frequency are combined into a 2k x n matrix. The first row of the matrix is ​​the real part of the cell AC impedance signal at the first frequency, and the second row is the imaginary part of the cell AC impedance signal at the first frequency. Similarly, the 2kth row is the imaginary part of the cell AC impedance signal at the kth frequency; after performing convolution calculation on this matrix, the result is fed into a max pooling layer, then into a fully connected neural network, and finally the cell type information is output.

Images