Peripheral blood cell morphology analysis system and heart failure degree judgment method

Abstract

The invention provides a peripheral blood cell morphological analysis system and a heart failure degree judgment method, and the method comprises the steps: taking the peripheral blood cell morphological analysis system as the basis, obtaining the fluorescence signal, intensity, phase, bright field and dark field images of red blood cells, building the relation between the heart failure degree and red multi-dimensional features, and determining the heart failure degree of the red blood cells; and constructing, training and optimizing an erythrocyte morphological feature extraction network model based on machine learning. On the basis, a red blood cell characteristic image feature extraction network model with a fluorescence label is adopted to analyze a red blood cell image obtained by detecting a peripheral blood sample by a multi-parameter flow cytometry imaging system, the correlation between the characteristics of red blood cells in the peripheral blood sample and the heart failure degree is analyzed, and the heart failure degree is judged quickly and precisely.

Description

Technical Field

[0001] The present invention relates to the technical field of optical imaging, and particularly to a peripheral blood cell morphology analysis system and a method for judging the degree of heart failure. Background Art

[0002] Heart failure (HF) is a complex clinical syndrome and the late manifestation of various cardiovascular diseases, characterized by the heart's inability to pump blood effectively to meet the body's metabolic needs. Heart failure is a severe manifestation or late stage of various heart diseases, with high mortality and readmission rates. According to statistical data, more than 64 million people worldwide suffer from heart failure. Therefore, joint efforts to reduce its social and economic burden have become a major global public health priority. Despite the latest advances in the treatment of heart failure, the five-year survival rate remains stubbornly below 50%, a figure strikingly similar to the mortality rates of various cancers. The diagnosis and assessment of heart failure are the cornerstones of guiding patient treatment. Effectively predicting the severity of heart failure helps identify patients at different risk levels and guide their medical care accordingly.

[0003] However, as a complex clinical syndrome, the diagnostic methods for heart failure are intricate. In clinical practice, echocardiography, B-type natriuretic peptide (BNP), and cardiac imaging are commonly used as diagnostic criteria. BNP is a bioactive hormone synthesized by cardiomyocytes and increases when the myocardium is damaged. At the same time, echocardiography is a commonly used clinical imaging modality that can evaluate both the heart's morphology and function. However, these two methods are also susceptible to the influence of other diseases. For example, in situations other than heart failure, such as atrial fibrillation, atrial flutter, pulmonary infection, and renal insufficiency, BNP levels can also increase, leading to non-specific elevation. In addition, the diagnostic threshold of BNP is affected by age and body mass index. On the other hand, for patients with severe emphysema, chest deformity, or severe obesity, echocardiography may be affected, resulting in poor imaging and affecting the diagnostic quality. Moreover, these methods require experienced and skilled practitioners, so there is a certain degree of subjectivity.

[0004] Red blood cells, as easily obtainable clinical samples, are also key indicators for the prognosis of heart failure, and their red blood cell distribution width is an independent predictor of heart failure prognosis. However, there is currently no specific explanation for the relationship between red blood cell function and the progression of heart failure. Currently, the detection of red blood cells in clinical practice still relies on blood routine examinations, which have limitations in providing comprehensive information about red blood cells.

[0005] In existing research, the disclosed invention patent (application number: CN202110723669.3) provides a blood cell morphology analyzer based on cloud computing and artificial intelligence. This invention statically scans a glass slide coated with blood cells based on a two-dimensional translation stage, uploads the collected images, and performs cloud computing. The disclosed invention patent (application number: CN202411311051.6) provides a MYH9-RD assisted screening system and method based on blood cell morphology, including an image acquisition module responsible for obtaining Wright-stained peripheral blood images of cases to be screened, an image detection module responsible for obtaining individual images of each blood cell, an image analysis module responsible for obtaining blood cell category information, white blood cell morphology information, and platelet size information, a counting and processing module responsible for generating analysis results of cases to be screened, a related disease judgment module responsible for generating MYH9-RD screening results, and a result output module responsible for outputting screening results. The disclosed invention patent (application number: CN202010624204.8) provides a cell analysis system that, after grading an agglutinated sample, takes images of and analyzes the cells in the test smear. The disclosed invention patent (CN202311174343.5) provides a blood cell image acquisition method based on automatic image recognition, that is, an automatic image recognition algorithm determines an area suitable for clinical blood cell analysis in an image as a suitable recognition area; uses a low-power objective lens to locate a reference point in the suitable recognition area and calculates a key recognition area for scanning under a high-power lens; uses a high-power objective lens to scan the key recognition area starting from the reference point to obtain blood cell images of the key recognition area.

[0006] In the above technologies, using machine learning to analyze cell microscopic images or developing new judgment methods related to heart failure still have problems such as low monitoring throughput and the need for complex staining operations. Therefore, comprehensive and label-free peripheral blood cell detection is urgently needed to achieve accurate assessment of heart failure status. Summary of the Invention

[0007] The present invention provides a peripheral blood cell morphology analysis system and a method for judging the degree of heart failure, aiming to solve the defect of complex implementation of the existing technology for heart failure detection, and realizing the construction of a high-throughput and multi-parameter flow imaging system to detect red blood cells in peripheral blood, and then identifying and analyzing the characteristics of the obtained red blood cell images and fluorescence signals through machine learning algorithms. Making full use of multi-dimensional feature information, combining the analysis ability of machine learning, establishing the connection between red blood cells and the degree of heart failure, improving the accuracy of judging the degree of heart failure, and promoting the development and application of heart failure judgment.

[0008] In the first aspect, the present invention provides a peripheral blood cell morphology analysis system, including: A cell fluorescence excitation and acquisition section, a cell intensity image acquisition section, a cell phase image acquisition section, a cell polarization image acquisition section, a broadband femtosecond laser, a microfluidic chip, and a computer; The cell fluorescence excitation and acquisition section is used to acquire the fluorescence signal of red blood cells; The cell intensity image acquisition section is used to acquire the intensity image of red blood cells; The cell phase image acquisition section is used to acquire the phase image of red blood cells; The cell polarization image acquisition section is used to acquire the bright-field image and dark-field image of red blood cells; The broadband femtosecond laser is used to generate femtosecond pulses with a wide spectrum and high repetition rate; The microfluidic chip is used to make red blood cells flow smoothly through the channel and assist the system in outputting various images; The computer is used to obtain the judgment result of heart failure degree based on the fluorescence signal, intensity image, phase image, bright-field image, and dark-field image.

[0009] According to a peripheral blood cell morphology analysis system provided by the present invention, the cell fluorescence excitation and acquisition section includes a continuous laser, a photomultiplier tube, and a dichroic mirror; The fluorescence marker in the red blood cell sample is excited by the continuous laser, and the generated fluorescence is reflected by the dichroic mirror and propagated to the photomultiplier tube to obtain the fluorescence signal of red blood cells.

[0010] According to a peripheral blood cell morphology analysis system provided by the present invention, the cell intensity image acquisition section includes a single-mode optical fiber, a diffraction grating, a microscope objective, a photodetector, and a high-speed oscilloscope; The cell phase image acquisition section includes a delay module, a beam splitter, a mirror, and a photodetector; By connecting a beam splitter behind the single-mode optical fiber to divide the optical pulse into a probe light and a reference light, a quantitative phase imaging system based on the interference principle is constructed. The optical path of the reference light is adjusted by the delay module so that the probe light and the reference light reach the photodetector simultaneously to achieve spatial interference. After digital signal processing, the intensity image and phase image of red blood cells are restored from the acquired time-domain interference signal.

[0011] According to a peripheral blood cell morphology analysis system provided by the present invention, the cell polarization image acquisition section includes a half-wave plate, a polarization beam splitter, a linear polarizer, and a photodetector; By adding a linear polarizer in front of the diffraction grating, the optical pulse passing through the red blood cell becomes a linearly polarized light. The linearly polarized light passes through the half-wave plate and the polarization beam splitter and reaches different photodetectors respectively, constructing an optical polarization imaging system. By adjusting the half-wave plate and the polarization beam splitter, the bright-field image and dark-field image of red blood cells are obtained respectively.

[0012] In a second aspect, the present invention further provides a method for judging the degree of heart failure, including: Introduce red blood cells and PBS solution into the inlet of the microfluidic chip, and make the red blood cells flow smoothly in the detection channel of the microfluidic chip; Collect the fluorescence signal, intensity image, phase image, bright-field image and dark-field image of red blood cells by the peripheral blood cell morphology analysis system; Send the obtained intensity, phase, bright-field image and dark-field image of red blood cells into the trained deep learning-based red blood cell morphology classification network model to obtain the judgment result of the degree of heart failure.

[0013] According to a method for judging the degree of heart failure provided by the present invention, the deep learning-based red blood cell morphology classification network model is obtained through the following steps: Establish an image database containing all red blood cell categories, and distinguish red blood cells with different degrees of heart failure through fluorescence labeling to obtain a red blood cell image feature database; Construct a red blood cell morphology classification network framework including an image feature extraction module, a feature stacking part and a feature classification module; Introduce an attention mechanism into the red blood cell morphology classification network framework to reduce the dimensions of the image and fluorescence signal, and extract potential features to guide the multi-channel neural network; Introduce an attention module into the red blood cell morphology classification network framework to perform refined attention feature induction and learning on the deep features extracted by the residuals; Use the red blood cell image feature database to train the adjusted red blood cell morphology classification network framework to obtain a red blood cell morphology classification network model.

[0014] According to a method for judging the degree of heart failure provided by the present invention, establish an image database containing all red blood cell categories, and distinguish red blood cells with different degrees of heart failure through fluorescence labeling to obtain a red blood cell image feature database, including: Collect peripheral blood samples of a specified population, and label the red blood cells in the peripheral blood through fluorescence staining; Obtain red blood cell images with different fluorescence labels of different heart failure patients by the peripheral blood cell morphology analysis system; Adopt a multi-channel neural network model guided by an attention mechanism to extract features from the obtained red blood cell images with different fluorescence labels, establish the connection between red blood cell features and the degree of heart failure through fluorescence labeling, and construct a red blood cell image feature database.

[0015] According to a method for judging the degree of heart failure provided by the present invention, send the obtained intensity, phase, bright-field image and dark-field image of red blood cells into the trained deep learning-based red blood cell morphology classification network model to obtain the judgment result of the degree of heart failure, including: Obtain a peripheral blood sample to be detected, preprocess the peripheral blood sample to obtain a preprocessed red blood cell sample; Input the preprocessed red blood cell sample into a peripheral blood cell morphology analysis system to obtain an intensity image, a phase image, a bright-field image, and a dark-field image of the preprocessed red blood cell sample; Input the intensity image, phase image, bright-field image, and dark-field image of the preprocessed red blood cell sample into a red blood cell morphology classification network model, and output a heart failure degree judgment result.

[0016] According to a method for judging the degree of heart failure provided by the present invention, it further includes: Verify the heart failure degree judgment result by using the clinical heart failure degree judgment result, and optimize the red blood cell morphology classification network model.

[0017] In a third aspect, the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the method for judging the degree of heart failure as described in any one of the above.

[0018] The present invention provides a peripheral blood cell morphology analysis system and a method for judging the degree of heart failure. Through the peripheral blood cell morphology analysis system, a large amount of red blood cell image data can be obtained to achieve a comprehensive evaluation of red blood cells in peripheral blood. In addition, due to the high-throughput and efficient imaging method of this method, the operation steps are simplified and the detection efficiency is improved. By analyzing the red blood cell image database through a machine learning algorithm that fuses multiple models, extracting and classifying the red blood cell image features, a fast and accurate judgment of the degree of heart failure can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is a schematic structural diagram of a peripheral blood cell morphology analysis system provided by the present invention; Figure 2 It is a schematic flow diagram of a method for judging the degree of heart failure provided by the present invention; Figure 3 It is a schematic diagram of the principle of red blood cell morphology classification based on machine learning provided by the present invention; Figure 4 It is a schematic flow diagram of heart failure degree classification provided by the present invention; Figure 5 It is a schematic structural diagram of an electronic device provided by the present invention.

[0021] Reference numerals: 101 broadband femtosecond laser; 102 single-mode optical fiber; 103 beam splitter; 104 polarizer; 105 diffraction grating; 106 microscope objective; 107 microfluidic chip; 108 microscope objective; 109 diffraction grating; 110 beam splitter; 111 half-wave plate; 112 mirror; 113 delay module; 114 beam splitter; 115 polarization beam splitter; 116, 117, 118 photodetectors; 119 high-speed oscilloscope; 120 computer; 121 continuous laser, 122 dichroic mirror, 123 dichroic mirror, 124 photomultiplier tube. Detailed implementation manners

[0022] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0023] In view of the limitations existing in the prior art, the present invention proposes a peripheral blood cell morphology analysis system and a heart failure degree judgment method. Among them, the peripheral blood cell morphology analysis system includes: A cell fluorescence excitation and acquisition part, a cell intensity image acquisition part, a cell phase image acquisition part, a cell polarization image acquisition part, a broadband femtosecond laser, a microfluidic chip and a computer; The cell fluorescence excitation and acquisition part is used to acquire the fluorescence signal of red blood cells; The cell intensity image acquisition part is used to acquire the intensity image of red blood cells; The cell phase image acquisition part is used to acquire the phase image of red blood cells; The cell polarization image acquisition part is used to acquire the bright-field image and dark-field image of red blood cells; The broadband femtosecond laser is used to generate femtosecond pulses with a wide spectrum and high repetition frequency; The microfluidic chip is used to make red blood cells flow smoothly through the channel and assist the system to output various images; The computer is used to obtain the heart failure degree judgment result based on the fluorescence signal, intensity image, phase image, bright-field image and dark-field image.

[0024] Specifically, as Figure 1As shown in the figure, the main structure of the peripheral blood cell morphology analysis system in this embodiment includes a cell fluorescence excitation / acquisition part, a cell intensity image acquisition part, a cell phase image acquisition part, a cell polarization image acquisition part, a continuous laser 121, a broadband femtosecond laser 101, a microfluidic chip 107, and a computer 120.

[0025] The present invention will develop a high-throughput and multi-parameter flow imaging system based on an optical flow control time-domain stretching imaging system, and study imaging and fluorescence signal excitation / acquisition technologies suitable for red blood cell morphology. By introducing fluorescence excitation / acquisition technology, quantitative phase detection technology, and optical polarization detection technology into the optical flow control time-domain stretching imaging system, the imaging system can simultaneously measure the fluorescence signal intensity, light pulse intensity, phase, bright field, and dark field information of cells. The size, structure, dry mass, roughness, and other multi-dimensional information of cells are retrieved using the optical pulse signal to provide complete cell parameters. The computer extracts and analyzes the fluorescence signal, intensity, phase, bright field, and dark field images of red blood cells collected, establishes the relationship between red blood cell characteristics and the degree of heart failure, and realizes the judgment of the degree of heart failure of patients.

[0026] Among them, the cell intensity image acquisition part includes a 1.5-kilometer-long single-mode optical fiber 102, diffraction gratings 105, 109, microscope objectives 106 and 108, a photodetector 118, and a high-speed oscilloscope 119; the cell phase image acquisition part includes a delay module 113, beam splitters 103 / 114 / 110, a mirror 112, and a photodetector 118; the cell polarization image acquisition part includes a half-wave plate 111, a polarization beam splitter 115, a linear polarizer 104, and photodetectors 116 and 117; the fluorescence excitation / acquisition part includes a continuous laser 121, a photomultiplier tube 124, dichroic mirrors 122 and 123.

[0027] The peripheral blood cell morphology analysis system in the embodiment of the present invention is used to acquire fluorescence signals, intensity, phase, bright field, and dark field images of red blood cells on the premise of high speed; the broadband femtosecond laser is used to generate femtosecond pulses with a wide spectrum and high repetition frequency; the microfluidic chip is used to make cells flow through the channel stably and quickly, so that the multi-parameter flow cytometry imaging device can stably image; the high-speed oscilloscope is used to quickly collect the electrical signals output by the photodetector; the computer uses machine learning methods to adaptively analyze the massive cell data and multi-dimensional information obtained from high-throughput and multi-parameter flow cytometry detection. By analyzing the fluorescence signal, intensity, phase, bright field, and dark field images, the morphological characteristics of red blood cells with specific markers are obtained, and a red blood cell classification model based on machine learning algorithms is constructed; finally, the trained model is used to predict the degree of heart failure of unknown peripheral blood samples.

[0028] The optical flow control time-domain stretching imaging system involved uses femtosecond pulses as the probe light. First, a dispersion fiber is used to separate different wavelength components of the ultrashort pulse in the time domain to achieve frequency-time mapping. Then, a diffraction grating is used to project different wavelength components of the ultrashort pulse spectrum to different positions in space to achieve frequency-space mapping. Subsequently, a microscope objective lens is used to focus the spatially dispersed optical pulse on the cells in the microfluidic channel, thus realizing the mapping from the spatial information of the cells to the time-domain waveform of the pulse. Finally, a single-pixel photodetector is used to collect the signals. As the cells flow in the microfluidic channel, each optical pulse detects the information of a cross-section of the cells. At the backend, the pulses are stitched together through digital signal processing to obtain the complete two-dimensional image of the flowing cells, which can achieve a detection throughput of more than 1,000,000 cells / second and a spatial resolution better than 780 nm; In the fluorescence excitation / collection part, a continuous laser is used to excite the fluorescent markers in the sample, and the generated fluorescent signal is reflected by a dichroic mirror and propagated to a photomultiplier tube to achieve the collection of the fluorescent signal In the intensity and phase imaging part, a beam splitter is introduced after the dispersion fiber to divide the optical pulse into a probe light and a reference light to construct a quantitative phase imaging system based on the interference principle. The optical path of the reference light is adjusted by a delay module so that the probe light and the reference light reach the photodetector simultaneously to achieve spatial interference. After digital signal processing, the intensity and phase images of the cells are restored from the collected time-domain interference signals; In the polarization imaging part, a linear polarizer is added in front of the diffraction grating so that the optical pulse passing through the cells is a linearly polarized light. Subsequently, the linearly polarized light passing through the cells reaches different photodetectors after passing through a half-wave plate and a polarization beam splitter to construct an optical polarization imaging system. By adjusting the half-wave plate and the polarization beam splitter, the bright-field and dark-field images of the cells can be obtained respectively.

[0029] Optionally, in terms of device selection, in the embodiments of the present invention, the broadband femtosecond laser 101 is a Vitara-P laser from Coherent Corporation of the United States, with a central wavelength / repetition frequency / spectral width / pulse width of 800 nm / 80 MHz / 40 nm / 20 fs; the single-mode optical fiber 102 with a length of 1.5 km is a single-mode optical fiber of YOFC-780-1.5 model from FiberHome Corporation; the beam splitter 103 with a beam splitting ratio of 90:10 is a beam splitter BS041 from Thorlabs; the reflective diffraction gratings 105 and 109 with a grating line density of 1200 lines / mm are GR26-0608 from Thorlabs; the microscopic objectives 106 and 108 with a magnification of 50X and a numerical aperture of 0.65 are LCPLN-IR 50X from Olympus; the beam splitter 110 with a beam splitting ratio of 50:50 is BS005 from Thorlabs; the photodetector is a 1544-B-50 from Newport, with a bandwidth of 12.5 GHz; the high-speed oscilloscope 119 with a sampling rate of 40 GSa / s is a DSA91304A from Keysight Technologies of the United States; the parameters of the microfluidic chip 107 are a channel width of 80 µm and a channel height of 40 µm; the polarization beam splitter is a PBS-612 from Luban Corporation; the linear polarizer 104 is an FLP25-NIR-M from Luban Corporation; the half-wave plate 111 is an AHWP20-SNIR from Luban Corporation; the continuous laser 121 is an OBIS 488 nm LS laser from Coherent Corporation of the United States; the dichroic mirror 122 is a DMLP505 from Thorlabs; the dichroic mirror 123 is a DMLP638 from Thorlabs; the photomultiplier tube 124 is a PMM02 from Thorlabs.

[0030] Figure 2 It is a schematic flow chart of the heart failure degree judgment method provided by the embodiments of the present invention. As Figure 2 shown, it includes: Step 100: Introduce red blood cells and PBS solution into the inlet of the microfluidic chip to make the red blood cells flow smoothly in the detection channel of the microfluidic chip; Step 200: Collect the fluorescence signal, intensity image, phase image, bright-field image and dark-field image of red blood cells by the peripheral blood cell morphology analysis system; Step 300: Send the obtained intensity, phase, bright-field image and dark-field image of red blood cells into the trained deep learning-based red blood cell morphology classification network model to obtain the heart failure degree judgment result.

[0031] Specifically, in the embodiments of the present invention, red blood cells in a peripheral blood sample are detected by a peripheral blood cell morphological analysis system to obtain fluorescence signals, intensities, phases, bright-field and dark-field images of the red blood cells; an image database containing all red blood cell categories is established, and red blood cells with different heart failure degrees are distinguished by fluorescence labeling; a deep learning-based red blood cell image analysis network framework is constructed, multi-dimensional features of red blood cell images of patients with known heart failure degrees are extracted and analyzed, the connection between red blood cell features and heart failure degrees is established, and a high-precision classification network model for patients with heart failure degrees is trained; the trained network model is used to judge the heart failure degree to which the red blood cells in an unknown peripheral blood sample detected by a multi-parameter flow cytometry imaging system belong.

[0032] In one embodiment, as Figure 3 shown, a machine learning-based red blood cell image analysis network framework is constructed, multi-dimensional features of fluorescently labeled red blood cell images are extracted and analyzed, and a high-precision red blood cell morphology classification network model is established and trained, including the following steps: Construct a multi-model fusion neural network including an image feature extraction module and a feature classification module for extracting and analyzing multi-dimensional features of red blood cell fluorescence signals, intensities, phases, bright-field and dark-field images; Introduce an attention mechanism into the network to reduce the dimensions of images and fluorescence signals and extract potential features to guide the multi-channel neural network; Introduce an attention module into the network to perform refined attention feature induction and learning on the deep features extracted by the residuals; Use the established red blood cell image feature database to train the constructed machine learning-based heart failure degree diagnosis network model.

[0033] It should be noted that for the part of classifying the morphology of red blood cells of heart failure patients based on deep learning, a neural network including an image feature extraction module, a feature stacking part and a feature classification module is constructed; among them, the feature extraction module is mainly composed of Yolo-V3 and ShumleNet-V2. Yolo-V3 is used to extract information such as the size, dry mass, texture, and refractive index of red blood cell intensity and phase images. The ShumleNet-V2 network with grouped convolutions efficiently extracts feature information such as the birefringence, structure, and contour of red blood cells; the Concate method is used to stack the features extracted from various images, and the VGG-16 module is used to analyze the image data of each dimension of red blood cells. The global information such as the contour and size of red blood cells and the local information such as cell structure, texture, and birefringence are jointly analyzed through convolution operations to classify the heart failure degree.

[0034] In one embodiment, an image database of red blood cell morphological features for all different heart failure degrees is established, and patients with different heart failure degrees are distinguished by fluorescence labeling, including the following steps: Collect peripheral blood samples of a specific population and label the red blood cells in the peripheral blood by fluorescence staining; After detection by a multi-parameter flow cytometry imaging system, obtain a large number of red blood cell images with different fluorescence labels of different heart failure patients; Adopt the method of a multi-channel neural network model guided by an attention mechanism to initially extract the features of the red blood cell images of heart failure patients with fluorescence signals, and establish the connection between the red blood cell features and the degree of heart failure through fluorescence labeling.

[0035] In one embodiment, as Figure 4 shown, use the trained network model to judge the red blood cell features in an unknown peripheral blood sample detected by a multi-parameter flow cytometry imaging system, including the following steps: Collect an unknown peripheral blood sample, perform simple preprocessing on the peripheral blood sample, and prepare a red blood cell sample suitable for detection by a multi-parameter flow cytometry imaging system; The cell intensity, phase image acquisition part and the cell bright field and dark field image acquisition part of the multi-parameter flow cytometry imaging system acquire the intensity, phase, bright field and dark field images of red blood cells; Send the obtained red blood cell intensity, phase, bright field and dark field images into the trained red blood cell classification network model based on deep learning to judge the degree of heart failure of the patient.

[0036] Finally, verify the classification result of the device through the clinical heart failure degree result, and optimize the machine learning algorithm according to the result.

[0037] The present invention obtains multi-dimensional red blood cell morphology information through a peripheral blood cell morphology analysis system, and uses this information as the training set of a machine learning algorithm to input into a high-accuracy, high-robustness, and lightweight classification model, obtaining a classification model that can achieve high-accuracy classification. When actually detecting the morphology of a patient's red blood cells, collect the intensity, phase, bright field, and dark field morphology images of red blood cells without labels at a flux of 1,000,000 cells per second, and input the obtained multi-dimensional image data into the trained classification model to achieve rapid and high-accuracy classification of the patient's red blood cells and the corresponding proportion. Finally, optimize the result of the system through the clinical heart failure degree diagnosis result to obtain a device for judging the heart failure degree with high accuracy and high speed.

[0038] Figure 5 Illustrates a schematic physical structure diagram of an electronic device, as Figure 5As shown in the figure, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communication bus 540. Among them, the processor 510, the communications interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logical instructions in the memory 530 to execute the heart failure degree judgment method, which includes: introducing red blood cells and PBS solution into the inlet of the microfluidic chip to make the red blood cells flow smoothly in the detection channel of the microfluidic chip; collecting the fluorescence signal, intensity image, phase image, bright-field image, and dark-field image of the red blood cells by the peripheral blood cell morphology analysis system; and sending the obtained intensity, phase, bright-field image, and dark-field image of the red blood cells into the trained deep learning-based red blood cell morphology classification network model to obtain the heart failure degree judgment result.

[0039] In addition, when the logical instructions in the above-mentioned memory 530 are implemented in the form of software functional units and sold or used as independent products, they may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, may be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0040] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0041] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A peripheral blood cell morphology analysis system, characterized in that Comprising: A cell fluorescence excitation and acquisition part, a cell intensity image acquisition part, a cell phase image acquisition part, a cell polarization image acquisition part, a broadband femtosecond laser, a microfluidic chip, and a computer; The cell fluorescence excitation and acquisition part is used to acquire the fluorescence signal of red blood cells; The cell intensity image acquisition part is used to acquire the intensity image of red blood cells; The cell phase image acquisition part is used to acquire the phase image of red blood cells; The cell polarization image acquisition part is used to acquire the bright-field image and dark-field image of red blood cells; The broadband femtosecond laser is used to generate femtosecond pulses with a wide spectrum and high repetition frequency; The microfluidic chip is used to make red blood cells flow smoothly through the channel and assist the system to output various images; The computer is used to obtain the heart failure degree judgment result based on the fluorescence signal, intensity image, phase image, bright-field image, and dark-field image.

2. The peripheral blood cell morphology analysis system according to claim 1, wherein The cell fluorescence excitation and acquisition part includes a continuous laser, a photomultiplier tube, and a dichroic mirror; The fluorescence marker in the red blood cell sample is excited by the continuous laser, and the generated fluorescence is reflected by the dichroic mirror and transmitted to the photomultiplier tube to obtain the fluorescence signal of red blood cells.

3. The peripheral blood cell morphology analysis system according to claim 1, wherein The cell intensity image acquisition part includes a single-mode fiber, a diffraction grating, a microscope objective, a photodetector, and a high-speed oscilloscope; The cell phase image acquisition part includes a delay module, a beam splitter, a mirror, and a photodetector; By connecting a beam splitter behind the single-mode fiber, the optical pulse is divided into a probe light and a reference light to construct a quantitative phase imaging system based on the interference principle. The optical path of the reference light is adjusted by the delay module so that the probe light and the reference light reach the photodetector simultaneously to achieve spatial interference. After digital signal processing, the intensity image and phase image of red blood cells are restored from the collected time-domain interference signal.

4. The peripheral blood cell morphology analysis system according to claim 1, wherein, The cell polarization image acquisition part includes a half-wave plate, a polarization beam splitter, a linear polarizer, and a photodetector; By adding a linear polarizer in front of the diffraction grating, the optical pulse passing through the red blood cells is a linearly polarized light. The linearly polarized light passes through the half-wave plate and the polarization beam splitter and reaches different photodetectors respectively to construct an optical polarization imaging system. By adjusting the half-wave plate and the polarization beam splitter, the bright-field image and dark-field image of red blood cells are obtained respectively.

5. A method for judging the degree of heart failure, based on the peripheral blood cell morphology analysis system according to any one of claims 1 to 4, characterized in that, Comprising: The red blood cells and PBS solution are introduced into the inlet of the microfluidic chip to make the red blood cells flow smoothly in the detection channel of the microfluidic chip; The fluorescence signal, intensity image, phase image, bright-field image, and dark-field image of red blood cells are collected by the peripheral blood cell morphology analysis system; The obtained intensity, phase, bright-field image, and dark-field image of red blood cells are sent into the trained deep learning-based red blood cell morphology classification network model to obtain the heart failure degree judgment result.

6. The method for judging the degree of heart failure according to claim 5, characterized in that, The deep learning-based red blood cell morphology classification network model is obtained through the following steps: An image database containing all red blood cell categories is established, and red blood cells with different heart failure degrees are distinguished by fluorescence labeling to obtain a red blood cell image feature database; A red blood cell morphology classification network framework including an image feature extraction module, a feature stacking part, and a feature classification module is constructed; An attention mechanism is introduced into the red blood cell morphology classification network framework to reduce the dimensions of the image and fluorescence signal, extract potential features to guide the multi-channel neural network; An attention module is introduced into the red blood cell morphology classification network framework to perform refined attention feature induction and learning on the deep features extracted by the residual network; Using the red blood cell image feature database, the adjusted red blood cell morphology classification network framework is trained to obtain a red blood cell morphology classification network model.

7. The method for judging the degree of heart failure according to claim 6, wherein An image database containing all red blood cell categories is established, and red blood cells with different heart failure degrees are distinguished by fluorescence labeling to obtain a red blood cell image feature database, including: Collect peripheral blood samples of a specified population and label the red blood cells in the peripheral blood by fluorescence staining; Obtain red blood cell images with different fluorescence labels of different heart failure patients by a peripheral blood cell morphology analysis system; A multi-channel neural network model guided by an attention mechanism is used to extract features from the obtained red blood cell images with different fluorescence labels, establish the connection between red blood cell features and heart failure degrees through fluorescence labeling, and construct a red blood cell image feature database.

8. The method for judging the degree of heart failure according to claim 6, wherein The intensity, phase, bright-field image, and dark-field image of the obtained red blood cells are input into the trained deep learning-based red blood cell morphology classification network model to obtain the heart failure degree judgment result, including: Obtain a peripheral blood sample to be detected, preprocess the peripheral blood sample to obtain a preprocessed red blood cell sample; Input the preprocessed red blood cell sample into a peripheral blood cell morphology analysis system to obtain the intensity image, phase image, bright-field image, and dark-field image of the preprocessed red blood cell sample; Input the intensity image, phase image, bright-field image, and dark-field image of the preprocessed red blood cell sample into the red blood cell morphology classification network model to output the heart failure degree judgment result.

9. The method for judging the degree of heart failure according to claim 6, wherein, It further includes: Using the clinical heart failure degree judgment result to verify the heart failure degree judgment result and optimize the red blood cell morphology classification network model.

10. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, When the processor executes the program, it implements the heart failure degree judgment method according to any one of claims 5 to 9.

Images