阅读说明：本技术 机器学习辅助的微滴数字核酸检测装置及检测方法 (Machine learning-assisted microdroplet digital nucleic acid detection device and detection method ) 是由 李彪 李雯曦 王宇航 陈以铖 张馨予 帅永 于 2021-09-09 设计创作，主要内容包括：本发明提供一种机器学习辅助的微滴数字核酸检测装置及检测方法。所述微滴数字核酸检测装置包括液滴生成芯片、核酸扩增装置、荧光检测芯片三个部分。本发明利用微流控技术和机器学习辅助的图形处理,针对现有的恒温扩增或非恒温扩增核酸检测操作复杂、难以实现POCT、检测难以绝对定量的核酸检测问题,提出一种新的基于微流控的高通量微液滴数字核酸检测低成本系统,可以满足高生物安全性、便携性、高效性的检测设备需求。(The invention provides a microdroplet digital nucleic acid detection device assisted by machine learning and a detection method. The microdroplet digital nucleic acid detection device comprises a droplet generation chip, a nucleic acid amplification device and a fluorescence detection chip. The invention provides a novel microfluidic-based high-flux micro-droplet digital nucleic acid detection low-cost system by utilizing a microfluidic technology and machine learning-assisted graphic processing, aiming at the problems that the existing constant-temperature amplification or non-constant-temperature amplification nucleic acid detection is complex in operation, POCT is difficult to realize and absolute quantification of detection is difficult, and can meet the requirements of detection equipment with high biological safety, portability and high efficiency.)

1. A microdroplet digital nucleic acid detection device assisted by machine learning is characterized by comprising a droplet generation chip, a nucleic acid amplification device and a fluorescence detection chip; the droplet generation chip is connected with the fluorescence detection chip through a nucleic acid amplification device, and the droplet digital nucleic acid detection device is used under the pressure not lower than the atmospheric pressure;

the liquid drop generates the chip and includes meter font, cross or T shape liquid drop generation passageway (1), import (2), fillter (3), stationary flow section (4) and export (5), every import (2) all communicate fillter (3), every fillter (3) all communicate one stationary flow section (4), a plurality of stationary flow section (4) all communicate the one end of a meter font, cross or T shape liquid drop generation passageway (1), meter font, cross or T shape liquid drop generation passageway (1) intercommunication export (5).

2. The machine learning-assisted droplet digital nucleic acid detection device according to claim 1, wherein each filter section (3) comprises a plurality of sets of filter columns (3-1), and each set of filter columns (3-1) is a plurality of filter columns arranged in a staggered manner.

3. The machine learning-assisted droplet digital nucleic acid detection device according to claim 1, wherein the channel (1) for generating droplets in a shape like a Chinese character 'mi' comprises an inlet for a dispersed phase and an inlet for a continuous phase, the continuous phase of the channel (1) for generating droplets in a shape like a Chinese character 'mi' is located outside the dispersed phase, and the fluid in the shape like a Chinese character 'mi' applies shear force to the fluid in the dispersed phase symmetrically from two sides;

the cross-shaped liquid drop generating channel (1) comprises a dispersed phase inlet and a continuous phase inlet, the continuous phase fluid of the cross-shaped liquid drop generating channel (1) flows in two paths, and the dispersed phase is symmetrically extruded from two ends;

the T-shaped droplet generation channel (1) comprises a dispersed phase inlet and a continuous phase inlet, and the continuous phase fluid of the T-shaped droplet generation channel (1) extrudes the dispersed phase from a single side.

4. The machine learning-assisted droplet digital nucleic acid detection device according to claim 1, wherein the nucleic acid amplification device comprises a cylindrical nucleic acid amplification device and a planar nucleic acid amplification device; cylindrical nucleic acid amplification device includes heat conduction piece I (11) and heat conduction piece II (12) and winding pipeline on it, set up hot-hole I (13) on heat conduction piece I (11), set up a plurality of control by temperature change I (14) in the radiation range of hot-hole I (13), the one end of heat conduction piece II (12) sets up hot-hole II (15), set up control by temperature change hole II (16) in the radiation range of hot-hole II (15), be connected through heat-insulating material between heat conduction piece I (11) and the heat conduction piece II (12), heat conduction piece I (11), heat conduction piece II (12) form similar cylinder structure with heat-insulating material.

5. The machine learning-assisted droplet digital nucleic acid detecting apparatus according to claim 4, wherein the planar nucleic acid amplifying apparatus comprises a heat conducting plate (25) and micro-channels (28) etched thereon, and the heat conducting plate (25) is heated at different positions, and the temperatures of different regions are made different by temperature control.

6. The device for detecting microdroplet digital nucleic acid with machine learning assistance as claimed in claim 4 or 5, wherein the heat conducting block I (11) and the heat conducting block II (12) or the heat conducting plate (25) are divided into several regions, each region corresponds to different temperature requirements of the nucleic acid amplification process, and the heating time of the reaction reagents in the droplets is controlled by the liquid flow speed and the distance, so as to realize amplification in the heated micro-channels.

7. The device for droplet digital nucleic acid detection assisted by machine learning according to claim 1, wherein the fluorescence detection chip comprises an inlet (21), an outlet (22) and a branch (23), the inlet (21) and the outlet (22) are connected by the branch (23), the branch (23) is a straight path or a serpentine-shaped bent section is arranged on the straight path, and the branch (23) allows a plurality of droplets to flow side by side.

8. The device for detecting microdroplet digital nucleic acid with machine learning assistance according to claim 1, wherein the fluorescence detection chip is a micro-channel with a light-transmitting structure and can be transmitted by laser with a specific wavelength.

9. The method for testing a microdroplet digital nucleic acid detection device assisted by machine learning according to any one of claims 1-7, wherein the method for testing an absolute quantitative nucleic acid comprises the following steps:

step 1: liquid to be detected enters the filtering section (3) through the inlet (2), and large-sized foreign matters are filtered through the filtering section (3);

step 2: the liquid to be detected filtered in the step 1 enters the liquid drop generation section (1) through the steady flow section (4) and is prepared into liquid drops at the channel intersection of the liquid drop generation channel (1);

and step 3: flowing the liquid to be detected passing through the droplet generation channel (1) in the step 2 to a nucleic acid amplification device through an outlet (5),

and 4, step 4: the liquid to be detected which passes through the nucleic acid amplifier enters a fluorescence detection chip, a fluorescence image is obtained by imaging equipment and is processed to obtain a detection result of whether the image has fluorescence running nose or not, and the detection result is interpreted by a negative and positive result through a trained machine learning algorithm;

and 5: and discharging the waste liquid after the fluorescence detection through a drainage pipe.

10. The method for testing a droplet digital nucleic acid detecting device assisted by machine learning according to claim 8, wherein the machine learning of step 4 is specifically to train a classifier with limited manual annotations and automatically segment the remaining data, select image features and classifier of applicable droplets; the method comprises the steps of taking the liquid drops showing fluorescence as a foreground, taking the rest parts including liquid drops not emitting fluorescence, continuous phase oil among the liquid drops and PDMS channel materials as a background, judging pixel points in an image through a training classifier, and predicting the probability that the pixel points belong to the fluorescent liquid drops.

Technical Field

The invention belongs to the field of field qualitative detection, and particularly relates to a microdroplet digital nucleic acid detection device assisted by machine learning and a detection method.

Background

The infectious virus has strong transmission, easy variation and great difficulty in prevention and control, and has serious influence on the health and life of people. In the case of infectious virus transmission, the best solution is to quickly and accurately detect the infectious agent and isolate the vector to prevent further transmission. The existing nucleic acid detection operation is complex and long in time consumption, needs to be implemented in a biosafety laboratory, and is difficult to realize field application.

At present, nucleic acid detection technology has been widely applied in the field of clinical medicine, but in public places at the front of epidemic prevention, such as railway stations, airports, clinical sites and the like, people flow is huge, the existing nucleic acid detection means has high requirements on experimental environment and consumes long time, and can not be used for point-of-care testing (POCT), and POCT still faces a series of problems of how to effectively overcome pollution to the surrounding environment, how to reduce cost required by experiments, how to reduce complexity of instruments, how to simplify detection operation to the greatest extent, how to realize absolute quantification and the like.

Disclosure of Invention

The invention provides a machine learning-assisted microdroplet digital nucleic acid detection device and a detection method, which aim at solving the problems that the existing constant-temperature amplification or non-constant-temperature amplification nucleic acid detection operation is complex, POCT is difficult to realize, and absolute quantification is difficult to detect.

The invention is realized by the following technical scheme:

a machine learning assisted microdroplet digital nucleic acid detection device comprises a droplet generation chip, a nucleic acid amplification device and a fluorescence detection chip; the droplet generation chip is connected with the fluorescence detection chip through a nucleic acid amplification device, and the droplet digital nucleic acid detection device is used under the pressure not lower than the atmospheric pressure.

The liquid drop generates the chip and includes rice font, cross or T shape liquid drop generation passageway 1, import 2, fillter section 3, stationary flow section 4 and an export 5, every import 2 all communicates a fillter section 3, every fillter section 3 all communicates a stationary flow section 4, and is a plurality of stationary flow section 4 all communicates the one end of a rice font, cross or T shape liquid drop generation passageway 1, rice font, cross or T shape liquid drop generation passageway 1 intercommunication export 5.

Furthermore, each filtering section 3 comprises a plurality of groups of filtering columns 3-1, and each group of filtering columns 3-1 is formed by staggering a plurality of filtering columns.

Further, the inlet 2 comprises a dispersed phase inlet and a continuous phase inlet, the zigzag-shaped droplet generation channel 1 comprises a dispersed phase inlet and a continuous phase inlet, the continuous phase of the zigzag-shaped droplet generation channel 1 is arranged outside the dispersed phase, and the continuous phase fluid symmetrically applies shearing force to the dispersed phase fluid from two sides;

the cross-shaped liquid drop generating channel 1 comprises a dispersed phase inlet and a continuous phase inlet, the continuous phase fluid of the cross-shaped liquid drop generating channel 1 flows in two paths, and the dispersed phase is symmetrically extruded from two ends;

the T-shaped droplet generation channel 1 comprises a dispersed phase inlet and a continuous phase inlet, and the continuous phase fluid of the T-shaped droplet generation channel 1 extrudes a dispersed phase from a single side;

further, the nucleic acid amplification apparatus comprises a cylindrical nucleic acid amplification apparatus and a planar nucleic acid amplification apparatus; cylindrical nucleic acid amplification device includes heat conduction piece I11 and heat conduction piece II 12 and winding pipeline on it, set up heated hole I13 on the heat conduction piece I11, set up a plurality of temperature control holes I14 in the radiation range of heated hole I13, the one end of heat conduction piece II 12 sets up heated hole II 15, set up temperature control hole II 16 in the radiation range of heated hole II 15, be connected through the heat-insulating material between heat conduction piece I11 and the heat conduction piece II 12, heat conduction piece I11, heat conduction piece II 12 and heat-insulating material form similar columniform structure.

Further, the planar nucleic acid amplification apparatus includes a heat conducting plate 25 and micro-channels 28 etched thereon, and heats different positions on the heat conducting plate 25 and makes different temperatures in different areas by temperature control.

Furthermore, the heat conducting block I11 and the heat conducting block II 12 or the heat conducting plate 25 are divided into a plurality of areas, each area corresponds to different temperature requirements of the nucleic acid amplification process, and the heating time of the reaction reagent in the liquid drop is controlled by the liquid flow speed and the distance, so that the amplification is realized in the heated micro pipeline.

Further, the fluorescence detection chip comprises an inlet 21, an outlet 22 and a branch 23, wherein the inlet 21 and the outlet 22 are connected by the branch 23, a straight passage or a serpentine bent section is arranged on the branch 23, and the branch 23 allows a plurality of droplets to flow side by side.

Furthermore, the fluorescence detection chip is a micro-channel of a light-transmitting structure and can be transmitted by laser with specific wavelength.

A method of testing a machine learning assisted microdroplet digital nucleic acid detection device, the method of absolute quantitative nucleic acid testing comprising the steps of:

step 1: liquid to be detected enters the filtering section 3 through the inlet 2, and large-sized foreign matters are filtered through the filtering section 3;

step 2: the liquid to be detected filtered in the step 1 enters the liquid drop generation section 1 through the steady flow section 4, and the diameters of the liquid drops generated by the liquid to be detected are visually compared at the channel intersection of the liquid drop generation channel 1;

and step 3: the liquid to be detected passing through the droplet generating channel 1 in the step 2 is made to flow toward the nucleic acid amplification apparatus through the outlet 5,

and 4, step 4: the liquid to be detected which passes through the nucleic acid amplifier enters a fluorescence detection chip, a fluorescence image is obtained by imaging equipment and is processed to obtain a detection result of whether the image has fluorescence running nose or not, and the detection result is interpreted by a negative and positive result through a trained machine learning algorithm;

and 5: and discharging the waste liquid after the fluorescence detection through a drainage pipe.

Further, the machine learning of step 4 is specifically to train a classifier with limited manual annotations and automatically segment the remaining data, and customize the image features and the classifier of the applicable droplets; the method comprises the steps of taking the liquid drops showing fluorescence as a foreground, taking the rest parts including liquid drops not emitting fluorescence, continuous phase oil among the liquid drops and PDMS channel materials as a background, judging pixel points in an image through a training classifier, and predicting the probability that the pixel points belong to the fluorescent liquid drops.

The invention has the beneficial effects that:

1. the invention can realize absolute quantitative detection of liquid drops; and identifying the number of the fluorescent liquid drops to obtain the accurate concentration of the virus nucleic acid.

2. Compared with the PCR instrument which meets the requirement of amplification temperature through temperature rise and drop of the instrument, the temperature requirement is met by flowing liquid in a plurality of temperature areas, and the temperature rise and drop process can be omitted to shorten the amplification time.

3. The invention can simultaneously carry out high-flux processing and identification on hundreds of liquid drops on a single image by using machine learning or other algorithms, can effectively improve the detection speed and reduce the time required by a detection section experiment.

4. The invention combines the micro-fluidic technology, achieves the miniaturization and the portability of equipment and has higher biological safety.

Drawings

Fig. 1 is a schematic diagram of a channel of a droplet generation chip according to the present invention, wherein (a) the channel is a rectangular channel, (b) the channel is a cross channel, (c) the channel is a T channel, (d) the filter section is a flow stabilizer, (e) the channel is a rectangular channel, and (f) the channel is a rectangular channel, and (g) the channel is a cross channel, and (h) the channel is a T channel.

FIG. 2 is a schematic diagram of a droplet generation channel chip product of the present invention.

FIG. 3 is a schematic structural view of a cylindrical heater according to the present invention, wherein (a) the copper block is schematically shown in cross section, and (b) the copper block is schematically shown in three dimensions.

FIG. 4 is a schematic representation of a cylindrical heater embodiment of the present invention.

FIG. 5 is a schematic diagram of a detection chip channel according to the present invention.

FIG. 6 is a schematic diagram of the arrangement of droplets in the detection chip of the present invention.

FIG. 7 is a schematic diagram of the results of the classifier of the present invention processing tight fluorescent droplets.

FIG. 8 is a schematic flow chart of an embodiment of the present invention.

FIG. 9 is a schematic view of a planar heater according to the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Taking a common non-isothermal amplification PCR (polymerase Chain reaction) polymerase Chain reaction process as an example, a microdroplet digital nucleic acid detection device assisted by machine learning, wherein the detection process is realized by taking a fluorescent liquid drop image by imaging equipment or photoelectric conversion equipment and processing the image, the method is realized by the assistance of machine learning or other algorithms, and the microdroplet digital nucleic acid detection device comprises a liquid drop generation chip, a nucleic acid amplification device and a fluorescent detection chip; the droplet generation chip is connected with the fluorescence detection chip through a nucleic acid amplification device, and the droplet digital nucleic acid detection device is used under the pressure not lower than the atmospheric pressure;

for a common PCR amplification process, a cylindrical heating device is taken as an example, and fig. 8 is a schematic flow chart thereof. Wherein the image processing method is implemented with the aid of machine learning or other algorithms.

The liquid drop generates the chip and includes rice font, cross or T shape liquid drop generation passageway 1, import 2, fillter section 3, stationary flow section 4 and export 5, every import 2 all communicates a fillter section 3, every fillter section 3 all communicates a stationary flow section 4, and is a plurality of stationary flow section 4 all communicates a rice font, cross or T shape liquid drop generation passageway 1's one end, rice font, cross or T shape liquid drop generation passageway 1 intercommunication export 5.

Because the nucleic acid detection process is easy to generate cross contamination, the micro-fluidic chip has small volume and the whole experiment process is in a closed state, the problems of reliability, false positive and the like of detection can be effectively solved, the chip is only a few square centimeters and is easy to carry, and the miniaturization also makes the detection more convenient.

The invention designs and manufactures a micro-fluidic chip integrating sample processing and detection by utilizing a micro-fluidic technology, processes a detection image by combining a machine learning technology to obtain an absolute quantitative detection result, and provides a new solution for in vitro quantitative detection, particularly for the instant on-site detection of nucleic acid.

Furthermore, a plurality of the filtering sections (3) comprise a plurality of groups of filtering columns, each filtering section 3 comprises a plurality of groups of filtering columns 3-1, and each group of filtering columns 3-1 is formed by staggering a plurality of filtering columns.

The filter structure was formed by arranging round columns of about 50 μm diameter in a staggered manner, as shown in FIG. 1 (d). Due to the complexity of the reaction system and the possible micro-tissue generated in the chip channel manufacturing process and the punching process, in order to prevent the micro-tissue from forming accumulation and blockage at the fine structure of the channel and influencing the flow state of fluid in the channel, and causing the conditions of uneven liquid distribution, different flow states and the like of the symmetrical channels at two sides, a filter structure is designed at the inlet of each channel. The processing precision of PDMS and some micron structure particles such as viruses, proteins and the like which need to be wrapped in the dispersed phase in the experiment are comprehensively considered, and the micro-cylinder distance in the inlet channel is reasonably set so as to block foreign matters with larger sizes and improve the durability of the chip.

Further, the steady flow section 4 is a repeated zigzag pipeline. As shown in fig. 1 (e). Microchips and their experimental devices often have mechanical structural instability: when the micro-injection pump vibrates and the density of the fluid is uneven, the flow can oscillate; meanwhile, the PDMS material has a large elasticity, and thus may deform under a high fluid pressure condition, which finally causes the fluctuation of the fluid flow rate. To reduce the oscillation of the fluid in the microchannels, it is desirable to add a fluid resistance structure to the chip, which is designed for both continuous and dispersed phase microchannels.

Further, the inlet 2 comprises a dispersed phase inlet and a continuous phase inlet, the zigzag-shaped droplet generation channel 1 comprises a dispersed phase inlet and a continuous phase inlet, the continuous phase of the zigzag-shaped droplet generation channel 1 is arranged outside the dispersed phase, and the continuous phase fluid symmetrically applies shearing force to the dispersed phase fluid from two sides;

the cross-shaped liquid drop generating channel 1 comprises a dispersed phase inlet and a continuous phase inlet, the continuous phase fluid of the cross-shaped liquid drop generating channel 1 flows in two paths, and the dispersed phase is symmetrically extruded from two ends;

the T-shaped droplet generation channel 1 comprises a dispersed phase inlet and a continuous phase inlet, and the continuous phase fluid of the T-shaped droplet generation channel 1 extrudes a dispersed phase from a single side;

further, fig. 1(f), (g) and (h) are detail views of channels where two-phase fluids meet to generate liquid drops, four inlets are all tapered and provided with buffer zones, and outlets are tapered. A scale ruler with the size of 20-60 mu m is arranged at the intersection of the channels of the liquid drop generating channel 1 in the shape of a Chinese character 'mi', a cross or a T, so that the diameters of generated liquid drops can be conveniently and visually compared.

Further, the nucleic acid amplification apparatus comprises a cylindrical nucleic acid amplification apparatus and a planar nucleic acid amplification apparatus; cylindrical nucleic acid amplification device includes heat conduction piece I11 and heat conduction piece II 12 and winding pipeline on it, set up heated hole I13 on the heat conduction piece I11, set up a plurality of temperature control holes I14 in the radiation range of heated hole I13, the one end of heat conduction piece II 12 sets up heated hole II 15, set up temperature control hole II 16 in the radiation range of heated hole II 15, be connected through the heat-insulating material between heat conduction piece I11 and the heat conduction piece II 12, heat conduction piece I11, heat conduction piece II 12 and heat-insulating material form similar columniform structure.

The heat conduction block I11 is divided into a plurality of areas, each area corresponds to the temperature of PCR denaturation, annealing and extension, the heating time is controlled by the flow speed and the path, and the liquid passes through the micro-channel wrapped on the surface of the cylinder to realize amplification.

The capillary is wound around the cylindrical heater, so that the heating is more uniform, the cylinder is made of a red copper material with the highest heat conductivity coefficient, and the copper block is provided with a heating hole and a temperature control hole. The middle through hole is a heating hole for placing a heating rod; the through hole close to the edge is a temperature control hole, a thermocouple is partially arranged, the other part of the thermocouple is used for air cooling or water cooling, and the measured value can represent the surface temperature of the copper block and is used for temperature feedback. The thickness and the arc length of the copper block are designed according to the flow and the time required by the use of the reagent. The middle part of the copper block is combined into a complete circle by using a 3D printed heat insulation material model as a support, so that an obvious crease of the capillary can be avoided due to a sharp bend angle, and the 3D printed support is slightly shorter than the length of the copper block to reserve an independent heating area.

Furthermore, each part of the nucleic acid amplification device keeps constant temperature, and the controller realizes smaller temperature fluctuation of different temperature modules.

Furthermore, the two heat-conducting blocks of the nucleic acid amplification device have different sizes, the small size is high-temperature heating, and the large size is low-temperature heating, and the two heat-conducting blocks are respectively designed according to the amplification standard of the used nucleic acid detection reagent. The two heat conducting blocks need to be equipped with corresponding heating and temperature control equipment, and the two heat conducting blocks need to be controlled to reach the target amplification temperature by methods of ventilation, water addition and the like under partial conditions.

Further, the planar nucleic acid amplification apparatus includes a heat conducting plate 25 and micro-channels 28 etched thereon, and heats different positions on the heat conducting plate 25 and makes different temperatures in different areas by temperature control.

Furthermore, the heat conducting block I11 and the heat conducting block II 12 or the heat conducting plate 25 are divided into a plurality of areas, each area corresponds to different temperature requirements of the nucleic acid amplification process, and the heating time of the reaction reagent in the liquid drop is controlled by the liquid flow speed and the distance, so that the amplification is realized in the heated micro pipeline.

Further, the fluorescence detection chip satisfies high throughput detection, and the fluorescence detection chip includes an inlet 21, an outlet 22 and a branch 23, wherein the inlet 21 and the outlet 22 are connected by the branch 23, the branch 23 is a straight path or a serpentine curved section may be disposed on the straight path, and the branch 23 allows a plurality of droplets to flow side by side.

Furthermore, the fluorescence detection chip is a micro-channel of a light-transmitting structure and can be transmitted by laser with specific wavelength.

The goal of this portion of the detection channel is to have the droplets closely spaced, in which case the relative positions of the droplets are more fixed and more droplets can be accommodated in the same picture. As shown in FIG. 5, for a two-channel photo mask design, the straight channel length ensures that the droplets entering the capture frame have maintained a relatively steady flow state. The serpentine segments help to help stabilize the droplets, and the width of each straight channel leg allows multiple droplets to flow side-by-side. The designed channel has small overall size, and the overall length is 15mm and 19mm respectively. As shown in fig. 6, in this channel, the droplets can be arranged in a close and stable manner.

Furthermore, in the fluorescence detection chip, the fluorescence of the amplified reagent is excited by using the irradiation of exciting light with corresponding wavelength, a two-dimensional image of the fluorescent liquid drop is obtained by shooting by an imaging device or a photoelectric conversion device in the detection process, the two-dimensional image is subjected to image processing, and the method is realized by machine learning or assistance of other algorithms.

A method of testing a machine learning assisted microdroplet digital nucleic acid detection device, the method of absolute quantitative nucleic acid testing comprising the steps of:

step 1: step 1: liquid to be detected enters the filtering section 3 through the inlet 2, and large-size foreign matters are filtered through the filtering section 3, wherein the large size is based on the condition that the filtering cylinders which are arranged in a staggered mode cannot pass through;

step 2: the liquid to be detected filtered in the step 1 enters the liquid drop generation section 1 through the steady flow section 4, and the diameters of the liquid drops generated by the liquid to be detected are visually compared at the channel intersection of the liquid drop generation channel 1;

and step 3: flowing the liquid to be detected passing through the droplet generation channel 1 in the step 2 to a nucleic acid amplification device through an outlet 5;

and 4, step 4: the liquid to be detected which passes through the nucleic acid amplifier enters a fluorescence detection chip, a fluorescence image is obtained by imaging equipment and is processed to obtain a detection result of whether the image has fluorescence running nose or not, and the detection result is interpreted by a negative and positive result through a trained machine learning algorithm;

and 5: and discharging the waste liquid after the fluorescence detection through a drainage pipe.

The machine learning of the step 4 is specifically to train a classifier by limited manual annotation and automatically segment the rest data, and select image features and the classifier suitable for the liquid drops; the method comprises the steps of taking liquid drops showing fluorescence as a foreground, taking the rest parts including liquid drops not emitting fluorescence, continuous phase oil among the liquid drops and PDMS channel materials as a background, judging pixel points in an image through a training classifier, and predicting the probability that the pixel points belong to the foreground needing to be identified, namely the fluorescent liquid drops.