Intelligent kit for early screening of urinary system cancer and liver cancer

Abstract

The invention discloses an intelligent kit for early screening of urinary system cancer and liver cancer, and belongs to the technical field of embedded medical supplies for early detection of cancers. The intelligent kit mainly comprises biological rapid detection test paper, a sensor, a neural network embedded chip, a microprocessor, a display screen and the like, a blood or urine sample of a patient is dropped on the detection test paper, and after calculation of the embedded chip, a detection result is reflected on external equipment in real time, so that the detection accuracy is improved. And whether the patient suffers from urinary system cancer, liver cancer or other diseases is further distinguished. In this way, early detection of the urinary system cancer and the liver cancer is achieved, the detection efficiency of the urinary system cancer and the liver cancer is improved, comprehensive improvement of functions and performance of portable instruments and large-scale detection equipment is achieved, and great significance is achieved for improving medical conditions of people and improving the living standard of people.

Description

Technical Field

[0001] The present invention belongs to the technical field of embedded medical products for early detection of cancer, and specifically relates to an intelligent test kit for early screening of urinary system cancer and liver cancer based on a neural network embedded chip and a biological rapid detection test strip. Background Art

[0002] As the average human life span increases, cancer poses an increasing threat to humans. The clinical manifestations of malignant tumors vary depending on the organ, location, and degree of development. However, there are usually no obvious symptoms in the early stages of malignant tumors, and even if there are symptoms, they are often non-specific. By the time patients develop specific symptoms, the tumor is usually already in the middle or late stages, causing many cancer patients to die due to missing the best treatment opportunity.

[0003] Urinary system cancer is a tumor that occurs in any part of the urinary system. It includes tumors of the kidney, renal pelvis, ureter, bladder, and urethra. Urinary system cancer is divided into two major categories: one is the tumor of the hollow organs of the renal pelvis, ureter, bladder, and urethra, whose cavities are covered with urothelium, so what occurs is a transitional epithelial tumor. It is the most common tumor of the urinary system. The other type is renal cancer, Wilms tumor, and non-urothelial tumors of various organs that occur in the renal parenchyma, such as bladder rhabdomyosarcoma. Urinary system cancer is a common malignant tumor worldwide. More than 90% of urinary system tumors originate from the bladder and are bladder cancer, while the incidence of renal pelvic tumors accounts for only 5% of urothelial tumors and 5%-13.6% of kidney tumors. Bladder cancer ranks first in the incidence and mortality of reproductive system tumors worldwide, and increases year by year with age. The disease is more common in middle-aged and elderly people aged 40 to 60, accounting for 3.2% of all malignant tumors. Urinary system tumors often occur after the age of 40. The incidence rate in men is about twice that in women. The age-standardized incidence rate for men is 7.37 / 100,000, and for women it is 1.98 / 100,000. About 100,000 patients die from the disease each year, and it has been gradually increasing in recent years.

[0004] Liver cancer is the sixth most commonly diagnosed cancer and the fourth leading cause of cancer death worldwide, after lung cancer, colorectal cancer, and gastric cancer. Liver cancer is a highly lethal tumor, with most cases detected at an advanced stage and an incidence-to-mortality ratio close to 1. Hepatocellular carcinoma (HCC) accounts for approximately 75-85% of primary liver cancers and is a major health problem worldwide. 72% of cases occur in Asia (more than 50% in China), and most HCC cases occur in the context of chronic liver disease. Cirrhosis is the main risk factor for HCC, independent of liver disease etiology. Between 1990 and 2015, the global incidence of liver cancer increased by 75%. These data reflect changes in etiology, population age distribution, population growth, and age-specific incidence. Cancer surveillance aims to detect tumors early, with a recommended interval of 6 months in the West to minimize the risk of detecting late-stage HCC. The development of new biomarkers for early diagnosis of liver cancer is an urgent issue to be addressed. Although the current diagnostic marker, alpha-fetoprotein, has improved accuracy compared with ultrasound alone, it is still far from meeting the requirements for large-scale accurate screening.

[0005] With the in-depth development of computer technology, Internet technology and single-chip microcomputer technology, embedded systems have entered a new historical development stage after decades of development, that is, from general low-end applications to an era of high-end and low-end parallel development, and constantly improving the level of low-end application technology, which is marked by the development of 32-bit embedded systems. At present, the development of embedded systems in my country is very rapid, and its development is moving from embedded systems to embedded industries. The huge market demand continues to accelerate the industrialization process of embedded systems. In the field of technology, medical instruments and equipment are beginning to show a trend towards portability and networking. There is a strong market demand for blood pressure monitors and blood glucose meters that can be carried with you, and ventilators and electrocardiogram monitors that can be used at home or in small community health hospitals. Based on the inherent characteristics of the above-mentioned medical instruments and equipment and the requirements of the latest development trends, the application of embedded systems in medical instruments and equipment meets the requirements and changes brought about by the development trend. Summary of the invention

[0006] In view of this, the present invention aims to provide an intelligent test kit for early screening of urinary system cancer and liver cancer based on a neural network embedded chip and a biological rapid test strip. The doctor drops the patient's blood or urine sample on the test strip. After calculation by the embedded chip, the test result is reflected in real time on the external device, and further distinguishes whether the patient has urinary system cancer, liver cancer or other diseases. In this way, early detection of urinary system cancer and liver cancer can be achieved, and the detection efficiency of urinary system cancer and liver cancer can be accelerated. The comprehensive improvement of the functions and performance of portable instruments and large-scale detection equipment has been achieved, which is of great significance for improving people's medical conditions and improving people's living standards.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A smart test kit for early screening of urinary system cancer and liver cancer, the smart test kit mainly comprising a biological rapid test strip, a sensor, a neural network embedded chip, a microprocessor and a display screen; the biological rapid test strip can specifically detect Neu5Gc glycoform structures that modify UMOD in human urine and Neu5Gc-modified IgG glycoprotein in blood; the sensor is used to receive signals of T lines and C lines of the biological rapid test strip, and transmit the detection data to the neural network embedded chip inside the test kit; the neural network embedded chip is used to obtain the ratio of T and C from the received fluorescence signals of the T lines and C lines through an algorithm, convert it into a data signal and transmit it to the microprocessor, the microprocessor is used to process the received information through a neural network algorithm, classify the sample data and display it on the display screen of the test kit: the detection of urine shows whether the test subject will suffer from urinary system cancer, and the detection of blood shows whether the test subject will suffer from liver cancer.

[0009] Based on the above technical scheme, further, the biological rapid test paper is pasted on a rubber plate by a filter pad, an immunofluorescent microsphere release pad, an immunonitrocellulose membrane (detection line-T line and quality control line-C line) and a water absorbent pad (tightly connected from left to right in sequence); the preparation process of the immunofluorescent microsphere release pad is: fluorescent microsphere-labeled streptavidin and fluorescent microsphere-labeled mouse anti-IgG are sprayed on the release pad, and the preparation is prepared after drying; the preparation process of the immunonitrocellulose membrane is: rabbit anti-UMOD antibody and sheep anti-mouse IgG solution are respectively streaked on the T and C lines of the nitrocellulose membrane, and the preparation is prepared after drying.

[0010] Based on the above technical solution, further, the neural network algorithm specifically includes:

[0011] (1) In order to balance the number of cancer data and healthy data in urinary system cancer and liver cancer data and improve the efficiency of subsequent network learning and classification accuracy, a simple interpolation oversampling method, OREM, was used;

[0012] (2) LASSO regression model was used to extract key predictive variables and perform feature selection on disease characteristics in urinary system cancer and liver cancer;

[0013] (3) Design an optimized parameter model for urinary system cancer and liver cancer data classification based on LSTM-SENet-DBN to solve the problem of heavy computational load in traditional deep models and improve the learning rate of the network.

[0014] Based on the above technical solution, further, in step (1), OREM discovers potential minority sample regions through two steps, namely, exploring candidate minority sample regions (CMRs) and further identifying clean sub-regions in CMRs that do not contain any majority samples. The clean sub-regions are the desired potential minority sample regions. By filling the synthetic samples into the clean sub-regions, OREM can broaden the minority class region and enhance the conceptual expression of the minority class.

[0015] Based on the above technical solution, further, the process of the OREM algorithm in step (1) is as follows: First, the OREM of each minority sample finds its CMR and defines A(x i )={x i1 ,...xi(|s|-1)} is S(x i ) According to the sample to x i The distance of the rearranged sample set, where x i1 is the distance x i The most recent sample, x i2 is the distance x i The second closest sample, and so on, if there are q consecutive samples x i(k-q+1) ,x i(k-q+2) ,...x ik In A(x i ) comes from the majority sample class, with x i Centered on x i and x i(k-q) The distance between them is the radius of the circular area S(x i x i(k-q) ) is x i Then, OREM further utilizes the CMR of each original minority sample to identify the clean sub-regions; for C(x i ) for each sample x ip , if x i and x ip The midpoint of i and x ip A hypersphere region with a radius of half the distance between them does not contain any majority samples, and the region is a clean subregion; finally, OREM generates the same number of synthetic samples for each original minority sample.

[0016] Based on the above technical solution, further, in step (2), LASSO is used to select features from the balanced urinary system cancer and liver cancer data. LASSO replaces the L2 regularization term with L1, and the penalty function is expressed as:

[0017]

[0018] where y i , xi β i They are the response variable, explanatory variable and corresponding coefficient respectively, through the formula:

[0019]

[0020] Get the coefficients in LASSO The estimator is used for subsequent feature selection. After LASSO, the coefficients of the features in urinary system cancer and liver cancer are compressed and some regression coefficients are turned to 0, thereby realizing the feature extraction and selection of disease features.

[0021] Based on the above technical solution, further, LSTM-SENet-DBN is used in step (3). The key of LSTM (Long Short Tem Memory) is that the cell state horizontal line runs through the top of the graph. LSTM adds or deletes information to the cell state C through the "gate", and then completes the information transmission through the cell state C. This process consists of σ function and point-by-point multiplication operation.

[0022] Based on the above technical solution, further, the process of completing the forward calculation of the three gates of the LSTM layer includes three stages, including: "Forget gate" is used to discard information from the cell state, then the information value is updated to update the "cell" state, and finally the output value ht is determined; SENet (Squeeze-and-Excitation networks) is used to enhance the influence of effective features and weaken the influence of irrelevant features, and each element is weighted and summed with the corresponding attention weight to obtain the output of the self-attention mechanism; the output of SENet is input into the DBN (Deep Belief Network, DBN) model, DBN contains a restricted Boltzmann machine (restricted Boltzmann machine, RBM), which has the characteristics of extracting complex data features and reducing dimensionality when obtaining the probability distribution of input data, and ends with a BP layer.

[0023] Based on the above technical solution, the hardware further selects the NVIDIA JetsonNano core board with relatively large storage resources as the internal integrated core of the chip development board. The core board only has NVIDIA's CUDA computing unit. The base board is surrounded by a self-recovery fuse, a power indicator light, a voltage regulator chip, a heat sink, I / O pins and pins, a Type-C interface, a USB interface, a restart button, a Flash chip and an SD card slot bridged together through a wiring board.

[0024] Based on the above technical solution, the development board further adopts the Cortex-M7ARM processor as the SOC of the substrate. The development board generates a link library through the python algorithm, and uses the C / C++ language to directly call the functions in the library to configure the hardware and select the interface according to the project requirements; the LCD1602 display is used to realize the display function. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention, the drawings related to the embodiments are briefly introduced below.

[0026] Figure 1 It is the overall logic block diagram of the present invention;

[0027] Figure 2 This is a flow chart of preprocessing and feature selection of the present invention;

[0028] Figure 3 Reagent card composition diagram;

[0029] Figure 4 This is a neural network architecture diagram of the present invention;

[0030] Figure 5 This is the architecture diagram of the embedded chip of the present invention;

[0031] Figure 6 This is a schematic diagram of the test paper of the present invention;

[0032] Figure 7 This is a graph showing the detection results of the algorithm of the present invention on bladder cancer data;

[0033] Figure 8 This is the waveform algorithm data diagram inside the test kit;

[0034] Fig. 9 Kernel hardware parameters for embedded devices. DETAILED DESCRIPTION

[0035] The present invention is described in detail below with reference to the embodiments, but the embodiments of the present invention are not limited thereto. Obviously, the embodiments described below are only partial embodiments of the present invention. For those skilled in the art, other similar embodiments obtained without creative labor all fall within the protection scope of the present invention.

[0036] Example 1

[0037] The present invention provides an intelligent kit for early screening of urinary system cancer and liver cancer based on a neural network embedded chip and a biological rapid test strip, which is carried out according to the following steps:

[0038] Step 1: Based on the specific recognition of the UMOD glycoform structure in human urine by the lamprey immune protein LIP protein (see Chinese invention patent CN112763713B) and the glycoform structure characteristics of Neu5Gc-modified IgG glycoprotein in blood (see Chinese invention patent CN114167060B), an early diagnosis kit for urinary system cancer and liver cancer was established;

[0039] A new type of lamprey immune-related protein LIP (lamprey immune correlation protein, LIP) was isolated and identified from lampreys. The specific preparation process refers to the Chinese invention patent with patent application number 201310501366.2, entitled "lamprey LIP protein, preparation method and application in the preparation of drugs for the prevention and treatment of tumor diseases". The lamprey immune protein LIP can be obtained by isolating it from the culture medium of lamprey Leptinotarsa ​​bodies, or by a heterologous expression system, such as an Escherichia coli expression system, a yeast expression system, a plant expression system, an insect expression system and a mammalian cell expression system.

[0040] The protein can selectively kill tumor cells. Through ultra-high resolution microscopic imaging system and laser confocal microscopy, it is inferred that the recognition site of LIP for tumor cells is on the cell lipid raft. The results of sugar chip experiments show that LIP interacts with two antenna-type N-sugar chains containing N-glycolylneuraminic acid (Neu5Gc). Biacore data proves that LIP can bind to N-sugar chains containing Neu5Gc on the tumor surface. Based on the principle that LIP protein can specifically identify specific glycoform modifications in the blood and urine of tumor patients as a lectin, the immunochromatographic quantitative technology based on europium chelate fluorescent microspheres and the immunoassay platform of time-resolved immunofluorescent microspheres are combined to establish urine or blood detection technology for early diagnosis of urinary system cancer or liver cancer (see Chinese invention patent CN112763713B).

[0041] The reagent system consists of rabbit anti-UMOD antibody and goat anti-mouse IgG solution streaked on T and C lines of nitrocellulose membrane respectively, and a release pad sprayed with streptavidin labeled with fluorescent microspheres and mouse anti-IgG solution labeled with fluorescent microspheres, a filter pad, an absorbent pad and other components.

[0042] Reagent preparation process

[0043] Preparation of immunofluorescent microsphere release pad: prepare fluorescent microspheres labeled with streptavidin and fluorescent microspheres labeled with mouse anti-IgG, adjust to appropriate concentrations, spray on the release pad, and dry;

[0044] Preparation of immune nitrocellulose membrane: dilute UMOD rabbit polyclonal antibody solution and goat anti-mouse IgG solution to appropriate concentrations, draw the membrane on the nitrocellulose membrane to prepare T and C lines, and dry;

[0045] Paste the filter pad, immunofluorescence microsphere release pad, immunonitrocellulose membrane (test line - T line and quality control line - C line) and absorbent pad (tightly connected from left to right) on the plastic board, and cut the card into reagent strips of uniform width as required; put the reagent strips into a plastic box, and put each test kit and desiccant into an aluminum bag, seal it, and package it.

[0046] Detection principle: The quality control line (C line) on the nitrocellulose membrane in the kit is coated with goat anti-mouse IgG, and the detection line (T line) is coated with UMOD rabbit polyclonal antibody solution. The urine protein of the subject is mixed with the LIP-biotinylated solution and added to the filter pad. The LIP-biotinylated solution will bind to the streptavidin in the immunofluorescent microsphere release pad. When the chromatography (mouse IgG labeled with immunofluorescent microspheres will diffuse to the nitrocellulose membrane due to chromatography) reaches the T line of the immunofluorescent microcellulose membrane, if there is UMOD glycoprotein modified with terminal Neu5Gc bi-antenna in the sample, the Neu5Gc-modified UMOD glycoprotein can bind to the UMOD rabbit polyclonal antibody (mouse IgG labeled with immunofluorescent microspheres will bind to the goat anti-mouse IgG on the C line). According to the proportionality between the fluorescence signal intensity and the concentration of Neu5Gc-modified UMOD glycoprotein in the sample, the sample test results are interpreted and quantitatively analyzed using an appropriate instrument.

[0047] Step 2: Classification of liver cancer data and bladder cancer data based on LSTM-SENet-DBN neural network.

[0048] (1) In order to balance the number of various cancer data and healthy data in the urinary system cancer and liver cancer data extracted from the kit and improve the efficiency of subsequent network learning and the accuracy of classification, a simple interpolation oversampling method, OREM, is used, which is an oversampling method that reliably expands the minority class.

[0049] The key to interpolation sampling is to find potential minority sample regions. OREM uses two steps to find potential minority sample regions, namely exploring candidate minority sample regions (CMRs) and further identifying clean subregions in CMRs that do not contain any majority samples. Clean subregions are the desired potential minority sample regions. By filling the synthetic samples into the clean subregions, OREM can broaden the minority class region and enhance the conceptual expression of the minority class. The process of the OREM algorithm is as follows: First, OREM of each minority sample finds its CMR and defines A(x i )={x i1 ,...xi(|s|-1)} is S(x i ) According to the sample to x i The distance of the rearranged sample set, where x i1 is the distance x iThe most recent sample, x i2 is the distance x i The second closest sample, and so on. If there are q consecutive samples x i(k-q+1) ,x i(k-q+2) ,...x ik In A(x i ) comes from the majority sample class, with x i Centered on x i and x i(k-q) The distance between them is the radius of the circular area S(x i x i(k-q) ) is x i Then, OREM further utilizes the CMR of each original minority sample to identify clean subregions. i ) for each sample x ip , if x i and x ip The midpoint of i and x ip A hypersphere region with a radius of half the distance between them does not contain any majority samples, and the region is a clean subregion. Finally, OREM generates the same number of synthetic samples for each original minority sample.

[0050] (2) The LASSO regression model was used to extract key predictive variables and perform feature selection on the disease characteristics of urinary system cancer and liver cancer.

[0051] The data of urinary system cancer and liver cancer after data balancing are complex and difficult to analyze and use. Therefore, LASSO is used to select features from the balanced urinary system cancer and liver cancer data. LASSO is a special form of ridge regression that can avoid overfitting, ensure prediction accuracy, compress data dimensions, and reduce time costs. LASSO replaces the L2 regularization term with L1, and the penalty function is expressed as:

[0052]

[0053] where y i , x i β i They are the response variable, explanatory variable and corresponding coefficient respectively. Then the formula is:

[0054]

[0055] Get the coefficients in LASSO After LASSO, the coefficients of the features in urinary system cancer and liver cancer can be compressed and some regression coefficients can be turned into 0, thereby realizing the feature extraction and selection of disease features.

[0056] (3) Design an optimized parameter model for urinary system cancer and liver cancer data classification based on LSTM-SENet-DBN to solve the problem of heavy computational load in traditional deep models and improve the learning rate of the network.

[0057] The algorithm uses LSTM-SENet-DBN. The key of LSTM (Long Short Tem Memory) is that the horizontal line of the cell state runs through the top of the graph. LSTM adds or deletes information to the cell state C through the "gate", and then completes the information transmission through the cell state C. This process consists of σ function and point-by-point multiplication operation. The process of completing the forward calculation by the three gates of the LSTM layer includes three stages, including: "Forget gate" to discard information from the cell state, then update the information value, update the "cell" state, and finally determine the output value ht. SENet (Squeeze-and-Excitation networks) is used to enhance the influence of effective features and weaken the influence of irrelevant features. Each element is weighted and summed with the corresponding attention weight to obtain the output of the self-attention mechanism. The output of SENet is input into the DBN (Deep Belief Network, DBN) model. DBN contains restricted Boltzmann machine (RBM), which has the characteristics of extracting complex data features and reducing dimensions when obtaining the probability distribution of input data, and ends with BP layer.

[0058] Step 3: In order to get rid of the need for machines to complete related cancer detection and make cancer detection faster and more immediate, an embedded microfluidic neural network multi-classification chip is designed to complete offline cancer detection.

[0059] (1) Specific steps for combining the hardware part of the kit with the bioluminescence analysis process.

[0060] The internal chip of the test kit receives biological sample data (blood or urine) through photosensitivity and fluorescence sensors. The sample information is transmitted to the chip through the interface for interactive processing. The immune reagent is used as a molecular recognition unit and the fluorescent protein LIP is used as a marker. The T line value and C line value of the test paper are determined through the specific reaction between the antibody and the antigen. The T value and C value will form a waveform graph inside the hardware of the test kit. ( Figure 8), before being transmitted to the microprocessor, the peak area ratio algorithm is used to calculate the ratio of T, C, and transmit it to the microprocessor as a digital signal; the vertical axis of the waveform represents the concentration of the immune protein LIP, and the strength of the concentration will reflect the amount of protein carried by the sample, that is, the number of valid data, and the horizontal axis represents the time the protein exists on this test paper, because only after a period of time will the protein show up on the sample and form a peak; the microprocessor then processes the received information using the neural network algorithm of the above steps, and the neural network algorithm will classify the sample data and display it on the screen of the test kit: the test of urine will show whether the test subject is likely to suffer from urinary system cancer and the result will be displayed on the screen; the test of blood will show whether the test subject is likely to suffer from liver cancer and the result will be displayed on the screen.

[0061] (2) Based on the functions to be achieved, the chip is selected to perform simulation development experiments on the corresponding hardware required for embedded development.

[0062] The hardware selected the NVIDIA Jetson Nano core board with relatively large storage resources as the internal integrated core of the chip development board. The NVIDIA Jetson Nano development kit is a small, powerful computer that can run multiple neural networks in parallel at the same time. It can be used for applications such as image classification, object detection, segmentation, and speech processing, and the operating power consumption is as low as 5 watts; the computing unit of the development board is based on NVIDIA Jetson Nano core board, the core board only has NVIDIA's CUDA computing unit, the base board is jointly completed by the project team through PCB design, and the base board is surrounded by a self-recovery fuse, a power indicator light, a voltage regulator chip, a heat sink, I / O pins and pins, a Type-C interface, a USB interface, a restart button, a Flash chip and an SD card slot bridged together through a wiring board: Among them, the self-recovery fuse will provide overheating protection; the power indicator light prompts whether the development board is working; the voltage regulator chip can stabilize the voltage of the entire electronic device; the heat sink is used for heat dissipation of the development board; the I / O pins and pins provide a place for input devices to interact with output devices, and play a role in connecting signal lines and circuit boards; by connecting the Type-C or USB interface, the development board can be burned with algorithms and powered, and connected to peripherals; the restart button can make the development board work again in a dead state; the Flash chip is mainly used as the memory of this embedded system, storing the system program code, boot code and configuration data, and driving it; the SD card can receive sensor data, store the file system, and detect the security of the development board to make it more reliable. The development board uses the Cortex-M7ARM processor as the SOC of the base board. The Cortex-M7ARM processor is a leading 32-bit processor with a dual-core CPU inside. It has high performance and low dynamic power consumption, and can perform convolutional neural network operations with high performance. The development board generates a link library through the python algorithm and directly calls the functions in the library using the C / C++ language, so that it can configure the hardware and select the interface according to the project requirements; the LCD1602 display is used to realize the display function.

[0063] (3) Configure the corresponding software according to the above embedded hardware requirements to conduct experiments.

[0064] The overall chip development environment is the Keil5 development environment. According to the data of biological samples, the Linux system is selected as the operating system kernel, which also includes device drivers, file systems, network protocols, etc., and the relevant code writing, compilation, burning, debugging and other steps are carried out under the Linux system. The source code includes the main function, keyboard input code, LCD1602 display code, buzzer, delay function code, etc.

[0065] First, write the corresponding program engineering and compile it in the Keil5 development environment under the Linux system. The executable file generated by the compilation is a .hex or .bin format file, which is the internal file system of the chip, including the root file system, library file system, network protocol, etc. After that, Keil5 detects whether the compilation is wrong, and burns it after it is correct; use the JATG debugging interface to connect the chip to the computer for configuration, configure the downloader J-Link, import the .hex or .bin file and the download address (Flash, RAM) and other information, and finally download the burning program, import the above neural network algorithm into the burning program, and embed it into the chip; connect the chip to the computer again to test whether all the functions of the chip can be used normally. Machine language is to use C / C++ language to call the library function of the relevant neural network algorithm and call each interface of the chip to realize the chip-related functions, and use the static memory allocation method to avoid the use of malloc to dynamically allocate memory and generate memory fragmentation. Use GGC, GDB compiler and debugger to compile and debug the application embedded in the Linux system. After the above steps are fully implemented, the system performance and response speed are improved by optimizing the algorithm and code, and the overall system resource consumption is reduced. The test kit will have a chip integrated with the above functions installed inside. The tester will use a test strip to collect his or her own sample data. As long as the reagent strip carrying the sample data passes through the test kit, the tester can observe the result of whether he or she is sick on the display screen installed in the test kit.

[0066] 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 it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent kit for early screening of urinary system cancer and liver cancer, characterized in that: The intelligent test kit mainly includes a biological rapid test strip, a sensor, a neural network embedded chip, a microprocessor and a display screen; the biological rapid test strip can specifically detect the Neu5Gc glycoform structure of UMOD modified in human urine and the Neu5Gc-modified IgG glycoprotein in blood; the sensor is used to receive the signals of the T line and C line of the biological rapid test strip, and transmit the detection data to the neural network embedded chip inside the test kit; the neural network embedded chip is used to obtain the ratio of T and C from the received fluorescence signals of the T line and C line through an algorithm, convert it into a data signal and transmit it to the microprocessor, and the microprocessor is used to process the received information through a neural network algorithm, classify the sample data and display it on the display screen of the test kit: the detection of urine shows whether the test subject will suffer from urinary system cancer, and the detection of blood shows whether the test subject will suffer from liver cancer.

2. The intelligent reagent kit according to claim 1, characterized in that: The biological rapid detection test paper is composed of a filter pad, an immunofluorescent microsphere release pad, an immunonitrocellulose membrane (a detection line-T line and a quality control line-C line) and a water absorbent pad (closely connected from left to right) pasted on a rubber plate; the preparation process of the immunofluorescent microsphere release pad is: streptavidin labeled with fluorescent microspheres and mouse anti-IgG labeled with fluorescent microspheres are sprayed on the release pad, and then dried to prepare; the preparation process of the immunonitrocellulose membrane is: rabbit anti-UMOD antibody and sheep anti-mouse IgG solution are respectively streaked on the T and C lines of the nitrocellulose membrane, and then dried to prepare.

3. The intelligent reagent kit according to claim 1, characterized in that: The neural network algorithm specifically includes: (1) In order to balance the number of cancer data and healthy data in urinary system cancer and liver cancer data and improve the efficiency of subsequent network learning and classification accuracy, a simple interpolation oversampling method, OREM, was used; (2) LASSO regression model was used to extract key predictive variables and perform feature selection on disease characteristics in urinary system cancer and liver cancer; (3) Design an optimized parameter model for urinary system cancer and liver cancer data classification based on LSTM-SENet-DBN to solve the problem of heavy computational load in traditional deep models and improve the learning rate of the network.

4. The intelligent reagent kit according to claim 3, characterized in that: In step (1), OREM discovers potential minority sample regions through two steps, namely exploring candidate minority sample regions (CMRs) and further identifying clean subregions in CMRs that do not contain any majority samples. The clean subregions are the desired potential minority sample regions. By filling the synthetic samples into the clean subregions, OREM can broaden the minority class region and enhance the conceptual expression of the minority class.

5. The intelligent reagent kit according to claim 4, characterized in that: The process of OREM algorithm in step (1) is as follows: First, find the CMR of each minority sample OREM and define A(x i )={x i1 ,...xi(|s|-1)} is S(x i ) According to the sample to x i The distance of the rearranged sample set, where x i1 is the distance x i The most recent sample, x i2 is the distance x i The second closest sample, and so on, if there are q consecutive samples x i(k -q+ 1) ,x i(k -q+2),...x ik In A(x i ) comes from the majority sample class, with x i Centered on x i and x i(k-q) The distance between them is the radius of the circular area S(x i x i(k-q) ) is x i Then, OREM further utilizes the CMR of each original minority sample to identify the clean sub-regions; for C(x i ) for each sample x ip , if x i and x ip The midpoint of i and x ip A hypersphere region with a radius of half the distance between them does not contain any majority samples, and the region is a clean subregion; finally, OREM generates the same number of synthetic samples for each original minority sample.

6. The intelligent reagent kit according to claim 3, characterized in that: In step (2), LASSO is used to select features from the balanced urinary system cancer and liver cancer data. LASSO replaces the L2 regularization term with L1, and the penalty function is expressed as: where y i , x i β i They are the response variable, explanatory variable and corresponding coefficient respectively, through the formula: Get the coefficients in LASSO The estimator is used for subsequent feature selection. After LASSO, the coefficients of the features in urinary system cancer and liver cancer are compressed and some regression coefficients are turned to 0, thereby realizing the feature extraction and selection of disease features.

7. The intelligent reagent kit according to claim 3, characterized in that: In step (3), LSTM-SENet-DBN is used. The key to LSTM (Long Short Tem Memory) is that the cell state horizontal line runs through the top of the graph. LSTM adds or deletes information to the cell state C through the "gate", and then completes the information transmission through the cell state C. This process consists of σ function and point-by-point multiplication operations.

8. The intelligent reagent kit according to claim 7, characterized in that: The process of completing the forward calculation by the three gates of the LSTM layer includes three stages, including: the "forget gate" discards information from the cell state, then updates the information value, updates the "cell" state, and finally determines the output value ht; SENet (Squeeze-and-Excitation networks) is used to enhance the influence of effective features and weaken the influence of irrelevant features, and each element is weighted and summed with the corresponding attention weight to obtain the output of the self-attention mechanism; the output of SENet is input into the DBN (Deep Belief Network, DBN) model, which contains a restricted Boltzmann machine (RBM), which has the characteristics of extracting complex data features and reducing dimensionality when obtaining the probability distribution of the input data, and ends with a BP layer.

9. The intelligent reagent kit according to claim 1, characterized in that: The hardware selects the NVIDIA Jetson Nano core board with relatively large storage resources as the internal integrated core of the chip development board. The core board only has NVIDIA's CUDA computing unit. The baseboard is surrounded by a self-recovery fuse, a power indicator light, a voltage regulator chip, a heat sink, I / O pins and pins, a Type-C interface, a USB interface, a restart button, a Flash chip and an SD card slot bridged together through a wiring board.

10. The intelligent reagent kit according to claim 9, characterized in that: The development board uses the Cortex-M7 ARM processor as the SOC of the base board. The development board generates a link library through the python algorithm and uses the C / C++ language to directly call the functions in the library to configure the hardware and select the interface according to the project requirements; the LCD1602 display is used to realize the display function.

Images