Direct counting method and device for detecting number of fluorescence-labeled rare abnormal cells in blood sample

Abstract

The invention discloses a direct counting method and device for detecting the number of fluorescence-labeled rare abnormal cells in a blood sample. Aiming at the problem that direct counting of low-abundance cells is difficult to realize in a conventional blood container by a traditional detection means, the method realizes rapid and direct counting of rare abnormal cells by performing narrow-band filtering, efficient collection and high signal-to-noise ratio detection on a fluorescence signal emitted by a static blood sample. The device comprises a lighting source module, an exciting light and emitted light narrow-band filtering module, a light beam collimation and beam expansion module, a sample bearing module, an emitted light collection and detection module and a data processing and counting module. Accurate counting of the extremely-low-abundance fluorescence labeling abnormal cells can be completed in a conventional transparent container without flow sorting or complex treatment of a blood sample, and the method has the advantages of being easy and convenient to operate, high in detection speed, suitable for clinical screening and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biomedical detection, in particular to a method for directly counting extremely low abundance fluorescently labeled rare abnormal cells in blood by wide-field illumination and high-sensitivity photodetection in a conventional container without relying on microfluidic or sheath flow systems, and a device for implementing the method. BACKGROUND

[0002] With the continuous development of early diagnosis of tumors, immunotherapy, biomarker screening and cell behavior research, the detection of extremely low abundance epithelial-derived cells (such as circulating epithelial cells, circulating tumor cells, etc.) in blood samples has gradually become an important research direction of precision medicine. Such rare abnormal cells are derived from solid tumor tissues or epithelial organs, and their concentration in peripheral blood is extremely low, usually only a few to a few hundred per milliliter, much lower than the number of white blood cells (10 6 -10 7 / mL) and red blood cells (5x10 9 / mL). Therefore, how to efficiently identify and accurately count these rare abnormal cells in the presence of a large number of background blood cells is a key challenge faced by both basic research and clinical detection.

[0003] Traditional detection methods mainly include flow cytometry and microfluidic-based cell sorting technology. Although these systems have multi-parameter analysis capabilities, they still have obvious limitations when faced with extremely low abundance target cells: on the one hand, flow detection relies on sheath flow focusing, high-precision flow control and complex optical paths, which require strict maintenance of liquid flow stability and optical alignment accuracy. The actual passing probability of rare abnormal cells is low, the detection efficiency is limited, and non-specific scattering or weak fluorescence is easily shielded in the presence of a large number of background blood cells. On the other hand, the microfluidic channel structure is delicate and highly sensitive to viscosity, cell concentration and impurity content. For blood samples that have not been highly purified, the channel is prone to adsorption, clogging, shear damage and other phenomena, resulting in a decrease in detection reliability. In addition, microfluidic systems usually require a long detection time, limiting high-throughput applications.

[0004] Due to the above limitations, existing methods are difficult to accurately count and identify fluorescently labeled rare abnormal cells at extremely low abundance levels in a conventional blood cell container (such as a transparent centrifuge tube or a well plate) by a fast and simple method. Therefore, the present application proposes a direct counting method and device for detecting the number of fluorescently labeled rare abnormal cells in a blood sample, which can achieve fast, high-sensitivity, high signal-to-noise ratio direct counting of fluorescently labeled rare abnormal cells in a static blood sample. SUMMARY

[0005] The present application aims to provide a direct counting method and device for detecting the number of fluorescently labeled rare abnormal cells in a blood sample, and to solve the problems of complex equipment, slow detection speed, and high sample processing requirements caused by the dependence on microfluidic channels or sheath flow systems in the prior art. The core is to realize the direct calculation of the total number of fluorescently labeled rare abnormal cells in a static sample without physical separation or individual detection by combining a pre-constructed quantitative model with high-selectivity spectral filtering and high-sensitivity photodetection.

[0006] To achieve the above-mentioned purpose, the present application provides a direct counting method for detecting the number of fluorescently labeled rare abnormal cells in a blood sample, characterized by comprising the following steps:

[0007] S1: providing a laser light source with an emission wavelength corresponding to the excitation spectrum peak wavelength of the target fluorescent label; performing excitation light narrowband filtering on the laser output by the laser light source to obtain an excitation light beam with a center wavelength matching the excitation spectrum peak of the fluorescent label; the fluorescent label has a specific excitation spectrum and emission spectrum, and at least a certain spectral separation between the two;

[0008] S2: preparing a human or animal peripheral blood or umbilical cord blood sample with red blood cells removed, and adding a known number of fluorescently labeled rare abnormal cells to the sample; the rare abnormal cells are cells derived from non-hematopoietic tissues with a concentration of less than 1000 cells per milliliter of blood sample;

[0009] S3: using the excitation light narrowband filtered laser to uniformly cover and irradiate the entire detection area of the blood sample in a wide-field illumination manner, so that all fluorescently labeled rare abnormal cells in the sample are excited synchronously;

[0010] S4: performing emission light narrowband filtering on the fluorescent signal emitted by the sample; the center wavelength of the transmission band of the emission light narrowband filtering corresponds to the emission spectrum peak wavelength of the fluorescent label, and the transmission spectral band does not overlap with the transmission spectral band of the excitation light narrowband filtering to suppress excitation light crosstalk;

[0011] S5: using a photodetector to receive and record the total fluorescent signal after emission light narrowband filtering, the photodetector including a photomultiplier tube, an avalanche photodiode, or a high-sensitivity silicon photodiode, with single-molecule-level fluorescence detection capability;

[0012] S6: repeating steps S1 to S5 at least ten times, each time using a different known number of fluorescently labeled rare abnormal cells as a calibration sample; based on the recorded fluorescent signal intensity, laser power, filter parameters, and rare abnormal cell addition quantity, a quantitative relationship model between the fluorescent signal and the number of rare abnormal cells is established;

[0013] S7: performing steps S1, S3, S4 and S5 on the blood sample to be tested (containing an unknown number of fluorescently labeled rare abnormal cells) under the same optical conditions to obtain a measured fluorescent signal; and inversely deducing the actual number of fluorescently labeled rare abnormal cells in the sample from the measured fluorescent signal according to the quantitative relationship model constructed in step S6.

[0014] Preferably, the fluorescent label comprises any one of green fluorescent protein (GFP), red fluorescent protein (RFP), enhanced yellow fluorescent protein (EYFP), mCherry, tdTomato, Alexa Fluor series dye, Cyanine (Cy) series dye, BODIPY series dye, or ATTO series dye.

[0015] Preferably, the full width at half maximum (FWHM) of the filter used for excitation light narrowband filtering and / or emission light narrowband filtering is less than 20 nm, more preferably 10 nm.

[0016] Preferably, the wide-field illumination is achieved by collimating beam expansion, and the illumination spot uniformly covers the entire blood sample detection area.

[0017] Preferably, the quantitative relationship model comprises a linear regression model, a polynomial fitting model, a multivariate correction model, a machine learning model, or a deep learning model, for accurately calculating the number of fluorescently labeled rare abnormal cells in the blood sample according to the photodetector signal.

[0018] Preferably, the photodetector is configured in an integration mode or a photon counting mode for collecting continuous light intensity signals or discrete photon event streams, and the integration time is controllable.

[0019] Preferably, the blood sample is pretreated by lysing red blood cells, centrifugal washing, or magnetic bead enrichment before detection to reduce non-specific absorption and background fluorescence interference.

[0020] In particular, the method does not rely on microfluidic channels or sheath flow systems, and can complete full-field fluorescence signal acquisition in a static blood sample through a single wide-field illumination, and realize the calculation of the number of fluorescently labeled rare abnormal cells.

[0021] Correspondingly, the present application also provides a direct counting device for detecting the number of fluorescently labeled rare abnormal cells in a blood sample, which is suitable for implementing any of the above methods, and is characterized in that it comprises:

[0022] An illumination light source module is configured with a laser, and the wavelength emitted by the laser matches the peak value of the excitation spectrum of the target fluorescent label;

[0023] An excitation light narrow-band filtering module is arranged behind the illumination light source module, and includes at least one band-pass filter with a center wavelength matching a peak value of an excitation spectrum of the fluorescent marker, for filtering out spectral components of the illumination light source outside the excitation spectrum of the fluorescent marker, so as to avoid interference with detection of subsequent fluorescent emission signals;

[0024] A light beam collimation and expansion module is arranged behind the excitation light narrow-band filtering module, for spatially collimating and expanding the filtered excitation light, so as to form an illumination light beam suitable for wide-field illumination.

[0025] A sample carrying module is arranged behind the light beam collimation and expansion module, for accommodating a blood sample to be detected, and the sample carrying module is made of an optically transparent material selected from quartz, borosilicate glass or optical-grade polymer.

[0026] An emission light collecting module is arranged behind the sample carrying module, for converging fluorescent emission light beams generated by the blood sample, and effectively guiding the light beams into an effective light passing aperture range of an emission light narrow-band filtering module and a signal detection module.

[0027] An emission light narrow-band filtering module is arranged between the emission light collecting module and the signal detection module, and includes at least one band-pass filter with a center wavelength matching a peak value of an emission spectrum of the fluorescent marker, and a transmission spectrum range of the band-pass filter is free of spectral overlap with a transmission spectrum range of the band-pass filter in the excitation light narrow-band filtering module.

[0028] The signal detection module includes one or more high-sensitivity photodetectors, such as photomultiplier tubes, avalanche photodiodes or high-sensitivity silicon photodiodes, which have single-molecule-level fluorescent detection capability, for collecting and quantifying total fluorescent signals after emission light narrow-band filtering.

[0029] The data processing and counting module is in communication connection with the signal detection module, and is internally provided with a memory and a processor, for generating and running the quantitative relationship model as claimed in claim 1, and calculating the number of fluorescently labeled rare abnormal cells in the blood sample according to the input fluorescent signals.

[0030] Preferably, the sample carrying module is a detachable container, including a transparent centrifuge tube, a microcuvette or a multi-well plate (such as a 6-well plate or a 12-well plate), with a volume of 10 μL to 15 mL, and suitable for batch sample detection.

[0031] Preferably, the data processing and counting module is further provided with a user interface unit, for displaying detection results, calibration curves and quality control parameters, and supporting export of a standardized report format.

[0032] Compared with the prior art, the present application has the following advantages:

[0033] a) Flow-free detection: Abandoning traditional microfluidic or sheath flow systems, without the need for cells to pass through the detection point one by one, the whole field excitation and signal acquisition can be completed at one time in a static sample, greatly simplifying the operation process;

[0034] b) High detection throughput: The whole fluorescent response can be obtained at a single measurement, suitable for high-throughput screening;

[0035] c) Strong anti-interference ability: The complete spectral separation of excitation light and emission light is realized through narrow-band filtering, effectively suppressing background noise and crosstalk;

[0036] d) Wide adaptability: Support for a variety of common experimental containers, without the need for special chips or customized consumables;

[0037] e) Easy to automate integration: The device structure is compact, easy to use with automatic sampling, temperature control module, and promote the development of fully automated detection platform. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a flowchart of the direct counting method for detecting the number of fluorescently labeled rare abnormal cells in a blood sample described in the present application, which shows the complete technical path from excitation light preparation, calibration sample preparation, wide-field synchronous excitation, emission light narrow-band filtering, fluorescent signal acquisition, counting model construction to unknown sample calculation.

[0039] Figure 2 is a structural schematic diagram of the direct counting device for detecting the number of fluorescently labeled rare abnormal cells in a blood sample provided in the embodiment of the present application, which mainly includes the following eight functional modules:

[0040] 1: Illumination light source module; 2: Excitation light narrow-band filtering module; 3: Beam collimation and expansion module; 4: Sample bearing module; 5: Emission light collection module; 6: Emission light narrow-band filtering module; 7: Signal detection module; 8: Data processing and counting module. DETAILED DESCRIPTION

[0041] The present application will be further described in detail below in conjunction with the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for the purpose of explaining the present application and do not limit the protection scope of the present application.

[0042] The present embodiment is aimed at the following application scenario: in a human peripheral blood sample treated by red blood cell removal, containing about 5x10 6 to 5x10 7 blood cells (mainly CD45 + white blood cells), the number of rare abnormal cells of epithelial origin expressing green fluorescent protein (GFP) as a fluorescent marker (CD45-GFP +actual number of cells of interest in the sample. The number of such cells in the sample system typically fluctuates between 0-50, belonging to a typical very low abundance rare cell population. The detection process needs to be completed in common laboratory containers, including transparent tubes with a diameter of about 10 mm, 1.5 mL round-bottom centrifuge tubes, flat-bottom 6-well plates, or standard cuvettes, etc.

[0043] The specific implementation process of the present application is as follows:

[0044] (I) Device configuration:

[0045] The device used in this embodiment is shown in Figure 2 and includes the following core modules:

[0046] Illumination light source module 1: a semiconductor laser with a wavelength of 488 nm is used, which is highly matched with the excitation spectrum peak (about 488 nm) of GFP;

[0047] Excitation light narrowband filtering module 2: one or more bandpass filters with a center wavelength of 488 nm and a full width at half maximum (FWHM) of 10 nm are inserted, which effectively removes the spectral components outside the non-GFP excitation spectral peak wavelength range in the laser source;

[0048] Beam collimation and expansion module 3: a beam expansion system composed of a pair of convex lenses is used to expand the incident laser beam to a parallel light beam with a diameter ≥ 15 mm, ensuring coverage of the bottom detection area of all types of containers;

[0049] Sample carrying module 4: a 10 mm optical path cuvette is selected, and the container material is a high-transmittance quartz crystal;

[0050] Emission light collection module 5: a plurality of aspherical lenses with a focal length of 50 mm are used and placed around the sample emission light scattering direction, which are used to efficiently collect fluorescence emission signals from all directions of the entire sample;

[0051] Emission light narrowband filtering module 6: a bandpass filter with a center wavelength of 510 nm (the emission spectrum peak of GFP is about 509 nm) and FWHM = 10 nm is configured, and its transmission spectrum has no overlap with the excitation light filter;

[0052] Signal detection module 7: a high-sensitivity photomultiplier tube (PMT) is selected, which works in photon counting mode, is equipped with a high-voltage power supply and a preamplifier, and has single-photon-level fluorescence signal detection capability;

[0053] Data processing and counting module 8: developed based on an embedded ARM processor, integrated with data analysis programs in Python environment, supports real-time signal reading, background subtraction, normalization processing and model calling, and is equipped with a touch screen for result display and parameter setting.

[0054] (ii) Detection procedure

[0055] As shown in the flowchart in Figure 1 , the implementation procedure of the method of the present application is as follows:

[0056] S1: Excitation light preparation. A laser with a central wavelength matching the excitation peak of the target fluorescent protein (GFP) (a semiconductor laser with a wavelength of 488 nm) is used. The laser first passes through the excitation light narrowband filtering module 2 for light source filtering. Subsequently, the light beam passes through the beam collimation and expansion module 3 to form a uniform light spot with a diameter of about 15 mm.

[0057] S2: Calibration sample preparation. Peripheral blood of healthy people is taken, treated with red blood cell lysis solution, and centrifuged to remove red blood cells to obtain a suspension rich in white blood cells. The in vitro cultured GFP transfected human breast cancer cell line MCF-7 (CD45-GFP + ) is diluted in gradient and added to the above sample to prepare a series of calibration samples containing 0, 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, and 50 target cells (each quantity of sample is measured 5 times), with a volume of 1 mL per sample.

[0058] S3: Wide-field synchronous excitation. Each group of samples is placed in the sample carrying module 4 to receive wide-field illumination excitation. The light spot uniformly covers all the sample areas to be tested.

[0059] S4: Emission light narrowband filtering. The green fluorescence emitted by the sample is collected by the emission light collection module 5. Subsequently, the emission light narrowband filtering module 6 performs narrowband filtering of the green fluorescence. The bandpass filter in the emission light narrowband filtering module 6 ensures that there is no spectral overlap with the bandpass filter in the excitation light narrowband filtering module 2.

[0060] S5: Total fluorescence signal detection. The fluorescence signal after emission light narrowband filtering is introduced into the PMT detector in the signal detection module 7. The PMT accumulates and collects for 1 s in photon counting mode, and records the total photon event number of each sample. At the same time, a blank sample (without GFP cells) is collected as a background control.

[0061] S6: Construction of quantitative counting model. After completing the fluorescence signal collection of multiple calibration samples, the deep learning method is used in this embodiment to establish a non-linear mapping relationship from the total fluorescence signal intensity to the number of rare abnormal cells. The specific procedure is as follows:

[0062] a) Collect independent experimental data (N groups in total) in the calibration process of S2, each group containing a known number of CD45-GFP +Cell and its corresponding multi-dimensional input features, including: fluorescence signal intensity (background deduction), laser power, filter parameters and rare abnormal cell addition quantity, PMT integration time, background noise level (from blank sample measurement) and the like. After standardization, all data constitute a training data set wherein is the input feature vector (d is the feature dimension), is the corresponding true rare abnormal cell quantity.

[0063] b) A deep neural network is constructed, which consists of an encoder and a decoder. Before inputting the encoder, a one-dimensional signal extraction module is used for feature preprocessing. To avoid feature loss during downsampling, a skip connection is introduced between the encoder and the decoder to supplement key information. Finally, the network completes the prediction of the rare abnormal cell quantity through the output layer.

[0064] c) The collected data is divided into training set, test set and validation set according to the ratio of 7:2:1. During model training, the loss function is composed of a normalized loss term and a mean square error (MSE) term to constrain the prediction bias and scale consistency. The Adam optimizer is used for parameter iteration update, and the initial learning rate is set to 1x10-3. To avoid training instability or falling into local optimum caused by high initial learning rate, a progressive learning rate warm-up strategy is introduced in the training stage, and an adaptive learning rate decay mechanism is used in the subsequent stage to improve the convergence performance. The training batch size is set to 8, and the maximum training round is 200. To reduce the risk of overfitting, an early stopping mechanism is enabled during training to monitor the performance of the validation set in real time and terminate training in advance when the performance no longer improves. After training is completed, the counting accuracy of the model is evaluated on the test set, and the internal parameters and inference weights of the model are further updated to obtain the final cell quantity prediction model.

[0065] d) The optimal model parameters after training are fixed to the embedded system in the data processing and counting module 8, which runs in a lightweight format to support real-time calculation.

[0066] S7: Estimation of the number of rare abnormal cells in unknown samples. After the same pretreatment, the blood sample containing unknown number of fluorescently labeled rare abnormal cells is loaded into the sample carrying module 4, and the measured fluorescence signal is obtained according to S1, S3, S4 and S5. The system automatically extracts the multi-dimensional input features under the current measurement conditions, and calls the deep learning model to output the predicted rare cell quantity.

[0067] The key of the method is: 1) the excitation / emission double narrow-band filtering greatly suppresses the interference of excitation light scattering and non-target autofluorescence; 2) the high-sensitivity detector such as PMT is used to realize the integral detection of weak group fluorescence; 3) through the model construction in the early stage, the complex group fluorescence signal intensity is associated with the number of rare abnormal cells, so that the cumbersome process of spatial resolution and identification of each cell is avoided.

[0068] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A direct counting method for detecting the number of fluorescently labeled rare abnormal cells in a blood sample, characterized in that, The method includes the following steps: S1: Provide a laser source whose emission wavelength corresponds to the peak wavelength of the excitation spectrum of the target fluorescent marker; perform narrowband filtering on the laser output from the laser source to obtain an excitation beam whose center wavelength matches the peak wavelength of the excitation spectrum of the fluorescent marker; the fluorescent marker has a specific excitation spectrum and emission spectrum, and there is at least a certain spectral separation between them; S2: Prepare a human or animal peripheral blood or umbilical cord blood sample from which red blood cells have been removed, and add a known number of rare abnormal cells labeled with a fluorescent marker to the sample; the rare abnormal cells are cells derived from non-hematopoietic tissues with a concentration of less than 1,000 per milliliter of blood sample; S3: The laser, after being filtered by a narrow band after excitation, is used to uniformly cover the entire detection area of ​​the blood sample in a wide field illumination mode, so that all fluorescently labeled rare abnormal cells in the sample are excited synchronously. S4: Perform emission light narrowband filtering on the fluorescence signal emitted by the sample; the center wavelength of the transmission band of the emission light narrowband filter corresponds to the peak wavelength of the emission spectrum of the fluorescent marker, and its transmission spectrum has no spectral overlap with the transmission spectrum of the excitation light narrowband filter, so as to suppress excitation light crosstalk. S5: Receive and record the total fluorescence signal after narrowband filtering of the emitted light using a photodetector. The photodetector includes a photomultiplier tube, an avalanche photodiode, or a high-sensitivity silicon photodiode, and has single-molecule-level fluorescence detection capability. S6: Repeat steps S1 to S5 at least ten times, each time using a different known number of fluorescently labeled rare abnormal cells as calibration samples; based on the recorded fluorescence signal intensity, laser power, filter parameters and the number of rare abnormal cells added, establish a quantitative relationship model between fluorescence signal and the number of rare abnormal cells. S7: Perform steps S1, S3, S4 and S5 on the blood sample to be tested (containing an unknown number of fluorescently labeled rare abnormal cells) under the same optical conditions to obtain the measured fluorescence signal; based on the quantitative relationship model constructed in step S6, the actual number of fluorescently labeled rare abnormal cells in the sample is deduced from the measured fluorescence signal.

2. The method of claim 1, wherein the fluorescent marker comprises any one of green fluorescent protein (GFP), red fluorescent protein (RFP), enhanced yellow fluorescent protein (EYFP), mCherry, tdTomato, Alexa Fluor series dyes, Cyanine (Cy) series dyes, BODIPY series dyes, or ATTO series dyes.

3. The method as described in claim 1, wherein the full width at half maximum (FWHM) of the filter used for the excitation light narrowband filter and / or the emission light narrowband filter is less than 20 nm, preferably 10 nm.

4. The method of claim 1, wherein the wide-field illumination is achieved by collimation and beam expansion, and the illumination spot uniformly covers the entire blood sample detection area.

5. The method of claim 1, wherein the quantitative relationship model includes a linear regression model, a polynomial fitting model, a multivariate correction model, a machine learning model, or a deep learning model, for accurately calculating the number of fluorescently labeled rare abnormal cells in a blood sample based on the photodetector signal.

6. The method of claim 1, wherein the photodetector is configured with an integration mode or a photon counting mode for acquiring continuous light intensity signals or discrete photon event streams, and the integration time is controllable.

7. The method of claim 1, wherein the blood sample is pretreated by lysing red blood cells, centrifugation and washing, or magnetic bead enrichment before detection to reduce nonspecific absorption and background fluorescence interference.

8. The method of claim 1, wherein the method does not rely on microfluidic channels or sheath flow systems, and can complete the acquisition of full-field fluorescence signals in static blood samples through a single wide-field illumination, and realize the estimation of the number of fluorescently labeled rare abnormal cells.

9. A direct counting device for detecting the number of fluorescently labeled rare abnormal cells in a blood sample, suitable for implementing the method as described in any one of claims 1-8, characterized in that, include: The illumination source module is equipped with a laser, the emission wavelength of which matches the excitation spectral peak of the target fluorescent marker; An excitation light narrowband filter module, disposed after the illumination source module, includes at least one bandpass filter with a center wavelength matching the peak value of the excitation spectrum of the fluorescent marker, used to filter out spectral components in the illumination source that are outside the excitation spectrum of the fluorescent marker, so as to avoid interference with the detection of subsequent fluorescence emission signals; A beam collimation and beam expansion module is located after the excitation light narrowband filter module. It is used to spatially collimate and expand the filtered excitation light to form an illumination beam suitable for wide-field illumination. The sample carrier module, located after the beam collimation and beam expander module, is used to hold the blood sample to be tested. The sample carrier module is made of an optically transparent material, selected from quartz, borosilicate glass or optical grade polymer. The emission light collection module, located after the sample carrier module, is used to gather the fluorescent emission beam generated by the blood sample and effectively guide the beam into the effective aperture range of the emission light narrowband filter module and the signal detection module. The emitted light narrowband filter module is disposed between the emitted light collection module and the signal detection module. It includes at least one bandpass filter whose center wavelength matches the emission spectrum peak of the fluorescent marker, and its transmission spectrum does not overlap with the transmission spectrum of the bandpass filter used in the excitation light narrowband filter module. The signal detection module includes one or more high-sensitivity photodetectors, which include photomultiplier tubes, avalanche photodiodes or high-sensitivity silicon photodiodes, and have single-molecule-level fluorescence detection capabilities, used to collect and quantify the total fluorescence signal after narrowband filtering of emitted light. The data processing and counting module, which is communicatively connected to the signal detection module, has a built-in memory and processor for generating and running the quantitative relationship model as described in claim 1, and calculating the number of fluorescently labeled rare abnormal cells in the blood sample based on the input fluorescence signal.

10. The apparatus of claim 9, wherein the sample carrying module is a detachable container, including a transparent centrifuge tube, a micro-cube, or a multi-well plate, with a volume of 10 μL to 15 mL, suitable for batch sample detection.

Images