Device and method for research, development and detection of cardiovascular and cerebrovascular drugs

By combining a sample loading unit, a robotic arm, and a temporal convolutional network model, automated detection of cardiovascular and cerebrovascular drugs was achieved, solving the problem of missing specific parameters, improving detection accuracy and efficiency, and reducing costs.

CN118980829APending Publication Date: 2024-11-19佳木斯大学宏大医院有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410979219.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

The lack of effective specific detection equipment and solutions in current technologies leads to a lack of specific parameters in the development of cardiovascular and cerebrovascular drugs, which affects the formulation of clinical treatment plans.

Method used

By combining a sample loading unit, a robotic arm, optical components, and a temporal convolutional network model, automated sample loading, mixing, detection, and analysis are achieved. The drug's specific type is obtained by detecting spectral signals and utilizing the temporal convolutional network model.

Benefits of technology

It improves the accuracy and efficiency of cardiovascular and cerebrovascular drug testing, reduces manual operation, shortens research and development time, reduces costs, and has scalability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118980829A_ABST
    Figure CN118980829A_ABST
Patent Text Reader

Abstract

The invention relates to cardiovascular and cerebrovascular drug research and development detection equipment and method, and relates to the technical field of intelligent equipment, the cardiovascular and cerebrovascular drug research and development detection equipment comprises: a sample loading unit for ensuring enough to-be-detected samples, sampling through a manipulator and a sampling needle, and transferring the samples to an incubation plate for full reaction; after full reaction, transferring the medicine in the incubation tray into a magnetic separation recovery tray for detection; in the process, on one hand, a to-be-detected sample is continuously added into the incubation disc, the reacted sample is transferred out, continuous circulation is realized, meanwhile, detection is performed through an optical assembly in the magnetic separation recovery disc, and a detected reagent is recovered to form cyclic detection; after detection is carried out through the optical assembly every time, the signal is output to the analysis unit to serve as an input signal of the analysis unit, the analysis unit carries out analysis through the network model, manual intervention is not needed in the detection process, and detection is carried out circularly.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent devices, and in particular to a cardiovascular and cerebrovascular drug development and detection device and method. Background Art

[0002] In the early stages of drug development and production, large-scale drug property screening is often required to determine the pharmacological activity of candidate compounds. This screening usually involves testing a large number of samples to evaluate their properties in terms of activity, toxicity, and stability. In the subsequent stages of drug production, given that the drug's properties have been determined during the development phase, only limited drug property testing is often required for each specific drug batch. The purpose of these tests is to ensure that the quality and specific properties of the drug meet expectations, such as checking purity, chemical composition, physical properties, etc.

[0003] For cardiovascular and cerebrovascular drug development, the trend is to pursue higher specificity, that is, to allow drugs to more accurately target specific receptors or pathways to improve therapeutic effects and reduce adverse reactions. The goal of these specific drug developments is to minimize the impact on non-target tissues and organs, thereby providing safer and more effective treatment options.

[0004] However, at present, the development of cardiovascular and cerebrovascular drugs generally relies on clinical experience data for specific detection. In the actual production and development process, there is a lack of specific detection equipment and programs, which leads to a large lack of specific parameters for this type of drug, which is not conducive to the formulation of clinical treatment plans. Summary of the invention

[0005] To solve the above problems, the embodiment of the present invention discloses a cardiovascular and cerebrovascular drug development and testing device, comprising: A sample loading unit, which is loaded with a medicine test tube rack, on which a test tube for loading a medicine to be tested is provided; A first manipulator connected to a sampling needle, the first manipulator is used to transfer the detection reagent from the test kit to the incubation tray, and to transfer the drug to be detected from the test tube to the incubation tray to obtain a first mixed drug; A second manipulator is used to transfer the first mixed drug in the incubation tray to the magnetic separation recovery tray; An optical component, used for detecting a second mixed reagent in the magnetic separation recovery disk, wherein the second mixed reagent is the first mixed reagent to which an excitation light liquid is added; An analysis unit, used for receiving a detection signal of the optical component, and inputting the detection signal into a preset temporal convolutional network model, and obtaining a specific type of the object to be detected according to an output result of the temporal convolutional network model; Wherein, the detection signal includes a spectral signal, and the temporal convolutional network model is obtained after training based on detection signal samples with specific type labels.

[0006] Furthermore, handles are provided at both ends of the medical instrument test tube rack, and an opening is provided on one side of the tube groove.

[0007] In some embodiments, the sampling needle includes a bracket, a first motor is provided at the end of the bracket, and an output end of the first motor is connected to a first screw rod provided at one side of the bracket; a slider is provided on the first screw rod; A guide rail is provided on the other side of the bracket, a sampling pump is installed on the guide rail, a needle tube is connected to the lower end of the sampling pump, and a needle is provided at the end of the needle tube; The sliding block is connected to the sampling pump.

[0008] Furthermore, the sampling pump is provided with a connecting block, the connecting block is sleeved on the needle tube, and a spring is sleeved on the needle tube between the connecting block and the sampling pump.

[0009] The embodiment of the present invention also discloses a cardiovascular and cerebrovascular drug development and detection method, comprising: Obtaining the drug to be tested and the testing reagent and mixing them; The mixed drug to be detected and the detection reagent are transferred to an incubation plate for sufficient reaction to generate a first mixed drug; Transferring the first mixed reagent to a magnetic separation recovery disk, and adding an excitation light liquid to generate a second mixed reagent; Detecting the second mixed medicine by an optical component to generate a detection signal; The detection signal is input into a preset temporal convolutional network model, and the specific type of the object to be detected is obtained according to the output result of the temporal convolutional network model; wherein the detection signal includes a spectral signal, and the temporal convolutional network model is obtained after training based on detection signal samples with specific type labels.

[0010] Furthermore, the detection signal is input into a preset time convolutional network model, and the specific type of the object to be detected is obtained according to the output result of the time convolutional network model, including: Input each detection signal into the corresponding temporal convolution subnetwork in the preset temporal convolution network model; Among them, the temporal convolutional network model includes multiple temporal convolutional sub-networks, and the multiple temporal convolutional sub-networks respectively extract features of corresponding detection signals and then perform feature fusion, and use the fused features as the features extracted by the temporal convolutional network model.

[0011] Furthermore, the step of inputting the detection signal into a preset temporal convolutional network model and obtaining the specific type of the object to be detected according to the output result of the temporal convolutional network model comprises: Inputting the detection signal into a fully convolutional network layer based on dilated convolution of the temporal convolutional network model, performing feature extraction on the detection signal using the fully convolutional network layer, and outputting a feature vector corresponding to the detection signal; The feature vector is input into the fully connected layer of the temporal convolutional network model, the output result of the fully connected layer is input into the logistic regression layer of the temporal convolutional network model, the predicted probability corresponding to the specific type is output, and the specific type of the object to be tested is obtained according to the predicted probability.

[0012] Furthermore, the model training step is also included, specifically including: Acquire multiple detection signal samples and specific labels corresponding to each detection signal sample; A combination of a detection signal and a specific type label corresponding to each detection signal sample is used as a training sample, thereby obtaining a plurality of training samples, and the temporal convolutional network model is trained using the plurality of training samples.

[0013] Furthermore, the training of the temporal convolutional network model using the multiple training samples includes: Input any sample detection signal into the temporal convolutional network model, and output the predicted probability of the specific type corresponding to the sample detection signal; Calculating a loss value according to the predicted probability corresponding to the sample detection signal and the specific type label of the sample detection signal using a preset loss function; If the loss value is less than a preset threshold or the number of training times reaches a preset number, the training of the temporal convolutional network model is completed.

[0014] Furthermore, the step of obtaining a plurality of detection signal samples and a specific label corresponding to each detection signal sample further includes: Performing data enhancement processing on each detection signal sample using a sliding window of a preset time length, so that each detection signal sample generates a plurality of detection signal samples after data enhancement processing; The combination of the detection signal and the specific type label corresponding to each detection signal sample is used as a training sample, specifically: The combination of the detection signal and the specific type label corresponding to each detection signal sample obtained after the data enhancement process is used as a training sample.

[0015] Compared with the prior art, the present invention has at least the following beneficial technical effects: The present invention ensures that there are enough samples to be tested through a sample loading unit, samples are taken through a manipulator and a sampling needle, and the samples are transferred to an incubation tray for sufficient reaction; after the sufficient reaction, the medicines in the incubation tray are transferred to a magnetic separation recovery tray for detection; in this process, on the one hand, samples to be tested are continuously added to the incubation tray, and samples with good reactions are transferred out to achieve continuous circulation, and at the same time, optical components are used to detect in the magnetic separation recovery tray, and the reagents after detection are recovered to form a cyclic detection; in addition, after each detection by the optical component, the signal is output to the analysis unit as the input signal of the analysis unit, and the analysis unit performs analysis through a network model, and the process does not require human intervention, and the detection is performed cyclically. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a structural schematic diagram of a cardiovascular and cerebrovascular drug development and testing device according to some embodiments of this specification; Figure 2 It is a schematic diagram of the structure of a drug test tube rack of a cardiovascular drug research and development testing device according to some embodiments of this specification; Figure 3 It is a schematic diagram of the structure of a sampling needle of a cardiovascular and cerebrovascular drug development and testing device according to some embodiments of this specification; Figure 4 It is a structural schematic diagram of a first manipulator of a cardiovascular and cerebrovascular drug development and testing device according to some embodiments of this specification; Figure 5 It is a structural schematic diagram of a second manipulator of a cardiovascular and cerebrovascular drug development and testing device according to some embodiments of this specification; Figure 6 It is a flowchart of the steps of a cardiovascular and cerebrovascular drug development and detection method according to some embodiments of this specification; Figure 7 It is a flowchart of the steps of model training of a cardiovascular and cerebrovascular drug development and detection method according to some embodiments of this specification; Figure 8 is a block diagram of a computer device embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] Temporal Convolutional Networks (TCNs) are a method of applying Convolutional Neural Networks (CNNs) to time series data. It exploits the parallelism and parameter sharing of convolution operations to capture temporal patterns in time series. TCNs have the following main features: One-dimensional convolution: TCNs use one-dimensional convolution operations, which enables it to capture and learn time series. Unlike traditional RNNs (recurrent neural networks), TCNs can handle long-term dependencies more efficiently by parallelizing computations. Fixed receptive field: TCNs can increase the receptive field by increasing the depth of the convolution layer to capture longer temporal patterns. In contrast, the receptive field of traditional RNNs usually decreases over time. Residual connections: In order to avoid gradient vanishing and difficulties in training, TCNs introduce residual connections, which add the output of the convolution layer to its input. This makes it easier for information to flow in the network and helps with network training and convergence. Flexible network structure: The network structure of TCNs is very flexible and can be designed according to the needs of specific tasks. For example, convolution kernels of different sizes and depths, different pooling methods, batch normalization, etc. can be used.

[0019] like Figures 1 to 3 The cardiovascular drug development and detection equipment of the present invention shown in the figure may specifically include: a sample loading unit 1, which is loaded with a drug test tube rack 11, and the drug test tube rack 11 is provided with a test tube 113 for loading the drug to be detected; a first manipulator 4 is connected to a sampling needle 2, and is used to sample the detection reagent from the reagent kit 3 and the drug to be detected from the test tube 113 to the incubation tray 5, so that it can react fully to generate a first mixed drug; a second manipulator 7 is used to transfer the first mixed drug in the incubation tray 5 to the magnetic separation recovery tray 6; an optical component 8 is used to detect the second mixed drug in the magnetic separation recovery tray 6, wherein the second mixed drug is the first mixed drug with the excitation light liquid added; an analysis unit 9 is used to receive the detection signal of the optical component 8, and input the detection signal into a preset time convolution network model, and obtain the specific type of the object to be detected according to the output result of the time convolution network model; wherein the detection signal includes a spectral signal, and the time convolution network model is obtained after training based on the detection signal sample with a specific type label.

[0020] In one embodiment, the device of the present invention is composed of components such as a sample loading unit 1, a manipulator, an optical component 8 and an analysis unit 9. The components cooperate with each other to realize functions such as sample loading, mixing, detection and analysis. The device has a compact structure and is easy to operate, and can efficiently meet the needs of drug development and detection. Through the operation of the first manipulator 4 and the second manipulator 7, the entire process from sampling to incubation and transfer to the magnetic separation recovery disk 6 is automated, reducing the tediousness of manual operation and improving work efficiency and accuracy. The device uses the optical component 8 to detect the second mixed agent in the magnetic separation recovery disk 6, collects spectral signals, and analyzes the detection signals through a preset time convolution network model, which can quickly obtain The specific type of the object to be tested is obtained, which solves the problem that there is a large lack of specific parameters for this type of drug at present, which is not conducive to the formulation of clinical treatment plans; the method based on spectral analysis and time convolution network can improve the accuracy and reliability of detection: the equipment of the present invention has automated operation and high-precision detection technology, which can greatly shorten the time for drug research and development detection and improve work efficiency. At the same time, it reduces the input of human resources and the waste of reagents, and can effectively save costs; the structural design of the equipment is flexible and can be expanded and improved as needed. By adding sample loading units, manipulators and other components, it can meet the needs of larger-scale drug research and development detection, and its scalability provides the possibility for further improving the function and adaptability of the equipment.

[0021] In one embodiment, handles 114 are provided at both ends of the drug test tube rack 11 to facilitate taking and placing the drug test tube rack 11 , and an opening 112 is provided on one side of the tube slot 111 , through which one can visually check whether the drugs in the test tube are normal.

[0022] In one embodiment, the sampling needle 2 includes a bracket 200, a first motor 201 is provided at the end of the bracket, and the output end of the first motor 201 is connected to a first screw rod 202 arranged on one side of the bracket 200; a slider 203 is provided on the first screw rod 202; a guide rail 204 is provided on the other side of the bracket 200, and a sampling pump 205 is installed on the guide rail 204, and a needle tube 206 is connected to the lower end of the sampling pump 205, and a needle 207 is provided at the end of the needle tube 206; the slider 203 is connected to the sampling pump 205. The sampling needle 2 is connected to the first translational manipulator 4, and the sampling needle 2 cooperates with the first motor 201, the first screw rod 202, the guide rail 204 and the slider 203 to realize sampling by the needle tube 206 driving the needle 207 to move upward or downward. The first translational manipulator 4 is connected to the sampling needle 2, which has a high sampling speed and can save labor costs. The sampling needle 2 is composed of a bracket 200, a first motor 201, a first screw rod 202, a slider 203, a guide rail 204, a sampling pump 205, a needle tube 206 and a needle 207, and has a simple and clear structure, which is easy to manufacture and maintain. The design of the guide rail 204 and the slider 203 can make the movement of the needle 207 precisely controllable, and the sampling operation can be performed accurately.

[0023] In one embodiment, a connection block 208 is provided on the sampling pump 205, and the connection block 208 is sleeved on the needle tube 206. A spring 209 is sleeved on the needle tube 206 between the connection block 208 and the sampling pump 205. Through the cooperation of the first motor 201 and the first screw rod 202, the needle tube 206 drives the needle head 207 to move upward or downward through the movement of the slider 203, so as to realize the function of automatic sampling; through the connection block 208 being sleeved on the needle tube 206, and the spring 209 being provided on the needle tube, it can ensure that the sampling needle 2 is firm and stable, and avoid sample leakage or contamination.

[0024] In one embodiment, Figure 4As shown, the first manipulator 4 is a translation manipulator, which includes a first base 401, on which a support column 402 is provided; a cross bar 404 and a second motor 403 are provided at the upper end of the support column 402, for driving the cross bar 404 to rotate relative to the support column 402; a second screw is provided in the cross bar 404, and the second screw is connected to a third motor 405 provided at one end of the cross bar 404, and a horizontal moving block 406 connected to the second screw is also provided on the cross bar 404; the third motor 405 drives the horizontal moving block 406 to move horizontally on the cross bar 404 through two screws, and the horizontal moving block 406 is connected to a mounting block 407, and the sampling needle 2 is mounted on the mounting block 407, thereby driving the sampling needle 2 to move horizontally, and the sampling needle 2 itself can move horizontally, so that it can achieve all-round movement in three-dimensional directions, which is beneficial to sampling.

[0025] In one embodiment, Figure 5 As shown, the second machine manipulator 7 includes a second base 701, on which two slide rails 702 are provided, and a movable base 703 is provided on the slide rails. The movable base is connected to a multi-directionally movable robot arm 704, and the robot arm 704 is equipped with a clamping member 705 for transferring the first mixed drug from the incubation plate 5 to the magnetic separation recovery plate 6.

[0026] In an embodiment of the present invention, by loading the sample to be detected in the sample loading unit 1, it can be ensured that there are sufficient samples for detection, sampling is performed by the manipulator and the sampling needle 2, and the sample is transferred to the incubation tray 5 for sufficient reaction; after sufficient reaction, the medicines in the incubation tray 5 are transferred to the magnetic separation recovery tray 6 for detection; in this process, on the one hand, the sample to be detected is continuously added to the incubation tray 5, and the sample with good reaction is transferred out to achieve continuous circulation, and similarly, the optical component 8 is used to detect in the magnetic separation recovery tray 6, and the reagents after detection are recovered to form a cyclic detection; in addition, after each detection by the optical component 8, the signal is output to the analysis unit 9 as the input signal of the analysis unit 9, and the analysis unit 9 performs analysis through the network model, and the process does not require human intervention, and the detection is performed cyclically.

[0027] In the embodiment of the present invention, Figure 6 As shown, a cardiovascular drug development and detection method is also disclosed, comprising: Step S601, obtaining a drug to be detected and a detection reagent, and mixing them; Step S602, transferring the mixed drug to be detected and the detection reagent to an incubation plate for sufficient reaction to generate a first mixed drug; Step S603, transferring the first mixed reagent to a magnetic separation recovery disk, and adding an excitation light liquid to generate a second mixed reagent; Step S604: detecting the second mixed medicine through an optical component to generate a detection signal; Step S605: input the detection signal into a preset time convolutional network model, and obtain the specific type of the object to be tested according to the output result of the time convolutional network model; wherein, the detection signal includes a spectral signal, and the time convolutional network model is obtained after training based on the detection signal sample with a specific type label.

[0028] The entire process of sample sampling, reaction to recovery is circulated through the sampling needle, incubation plate and magnetic separation recovery plate, and the trained network model is used to achieve automated detection, accurate detection and improved detection efficiency.

[0029] In a specific embodiment, the detection signal is input into a preset temporal convolutional network model, and the specific type of the object to be detected is obtained according to the output result of the temporal convolutional network model, including: Each detection signal is respectively input into the corresponding time convolution subnetwork in the preset time convolution network model, and the specific type of the object to be tested is obtained according to the output result of the time convolution network model. The time convolution network model includes multiple time convolution subnetworks, and the multiple time convolution subnetworks respectively extract the features of the corresponding detection signals and then perform feature fusion, and the fused features are used as the features extracted by the time convolution network model.

[0030] The features of the corresponding detection signals are extracted through multiple temporal convolutional sub-networks and then fused. The fused features are used as the features extracted by the temporal convolutional network model. Since the features of multiple channels are fused, the accuracy of specific detection and recognition can be improved. At the same time, the temporal convolutional network model can automatically eliminate redundant information, further improving the detection accuracy and efficiency.

[0031] In a specific embodiment, the detection signal is input into a preset temporal convolutional network model, and the specific type of the object to be tested is obtained according to the output result of the temporal convolutional network model, including: inputting the detection signal into a full convolutional network layer based on dilated convolution of the temporal convolutional network model, performing feature extraction on the detection signal using the full convolutional network layer, and outputting a feature vector corresponding to the detection signal; inputting the feature vector into a fully connected layer of the temporal convolutional network model, inputting the output result of the fully connected layer into the logistic regression layer of the temporal convolutional network model, outputting the predicted probability corresponding to the specific type, and obtaining the specific type of the object to be tested according to the predicted probability.

[0032] In a specific embodiment, the present invention uses a full convolutional network based on dilated convolution to construct a temporal convolutional network. The processing flow of inputting the detection signal to be detected into the temporal convolutional network model can be implemented by the following method: First, the detection signal is input from the input layer of the temporal convolutional network, and the detection signal is feature extracted through the full convolutional network layer based on dilated convolution, and the feature vector corresponding to the detection signal is output. Secondly, the obtained feature vector is input into the fully connected layer of the temporal convolutional network model, and the classification is performed through the fully connected layer. The output result of the fully connected layer is input into the logistic regression layer of the convolutional neural network model, and the prediction probability corresponding to the specific type is output.

[0033] As a preferred solution, the full convolution network based on dilated convolution can be implemented with three layers of full convolution layers with dilation coefficients of 1, 2, and 4 respectively. Between the convolution layers, the dilation coefficient is doubled layer by layer, so that the receptive field grows exponentially and the sequence learning ability is enhanced. The logistic regression layer can be implemented with a softmax classifier. The extracted one-dimensional feature vector is input into the softmax classifier to obtain the prediction probability corresponding to the specific type. Between each full convolution layer, ReLU (The Rectified Linear Unit) can be used as the activation function, which is characterized by fast convergence and simple gradient calculation.

[0034] Among them, the dilated convolution formula is as follows: in, is the p-th feature of the t-th element in the n-th dilated convolutional layer, is the td·i-th feature of the q-th element in the n-1-th dilated convolutional layer, i is a specific position index within the convolution kernel, t−d⋅i is the adjusted position index used to determine the specific position where the weight acts on the input feature, and d is the dilation coefficient of the current layer; is the convolution kernel weight used in the nth dilated convolutional layer to generate the output of the current layer n from the output of the previous layer n−1, n is the layer where the weight is located, p is the weight used to calculate the pth output feature of the current layer, i is the position index in the convolution kernel, which is used to indicate the position of this particular weight in the convolution operation, is the bias parameter for the pth output feature in the nth dilated convolutional layer, k is the convolution kernel size, for The selection set of , q is The elements in a collection.

[0035] In order to better realize the feature learning of long sequences, in the specific implementation process, each temporal convolution subnetwork can be implemented by stacking 16 full convolution modules. Each module is implemented by 3 layers of full convolution layers based on dilated convolution, where the convolution kernel size, dilation coefficient and receptive field of each layer satisfy the following formula: Among them, Rm is the receptive field size of the mth layer, Rm-1 is the receptive field size of the previous layer (i.e., the m-1 layer), km is the convolution kernel size of the mth layer, and d is the expansion coefficient of the current layer; For example, the convolution kernel size of each layer of the dilated convolution is 9, the dilation coefficient of the first layer is 1, the second layer is 2, and the third layer is 4; then the receptive field sizes of the first to third layers are 9, 26, and 111, respectively.

[0036] In order to increase the convergence speed and improve the detection performance, a BN (Batchnormalization) layer can be used before the first convolutional layer. To reduce the risk of overfitting, a dropout layer can be added after each temporal convolutional subnetwork module, and the dropout probability can be set to 0.2.

[0037] By extracting features through the fully convolutional layer of the preset temporal convolutional network model and outputting the predicted probability corresponding to the specific type through the logistic regression layer, the specific type can be accurately classified, ensuring the efficiency and accuracy of the detection results.

[0038] In one embodiment, before the detection signal is input into a preset temporal convolutional network model, it also includes: obtaining multiple detection signal samples and specific labels corresponding to each detection signal sample; taking the combination of the detection signal and the specific type label corresponding to each detection signal sample as a training sample, thereby obtaining multiple training samples, and using the multiple training samples to train the temporal convolutional network model.

[0039] In some embodiments, Figure 7 As shown, it also includes model training steps, including: Step S701, obtaining a plurality of detection signal samples and a specific label corresponding to each detection signal sample; Step S702: taking a combination of a detection signal and a specific type label corresponding to each detection signal sample as a training sample, thereby obtaining a plurality of training samples, and using the plurality of training samples to train the temporal convolutional network model.

[0040] For example, first, multiple detection signal samples are obtained, and the specific type corresponding to each detection signal sample in the multiple detection signal samples is obtained, and the specific type determined for each detection signal is used as the label of the detection signal. For example, the detection signal under the specific state corresponding to different spectra is continuously collected by an optical component. Optionally, a multi-channel medical-grade EEG measurement device is used to collect spectral signals. Secondly, the combination of the detection signal and the specific type label corresponding to each detection signal sample is used as a sample to obtain multiple training samples. The detection signal in each sample is input into the constructed time convolution network model, and the convolution parameters and bias parameters of the time convolution network model are adjusted according to the output results to realize the training process of the convolution neural network model, thereby obtaining the above-mentioned preset time convolution network model. Multiple training samples are obtained by obtaining multiple detection signal samples and the specific type label corresponding to each detection signal sample, and the time convolution network model is trained using multiple training samples, so that the detection signal input into the time convolution network model for detection can obtain an accurate specific type.

[0041] In one embodiment, a convolutional neural network model is trained using multiple training samples, including: inputting any sample detection signal into the temporal convolutional network model, and outputting the predicted probability of the specific type corresponding to the sample detection signal; using a preset loss function to calculate the loss value according to the predicted probability corresponding to the sample detection signal and the specific type label of the sample detection signal; if the loss value is less than a preset threshold or the number of training times reaches a preset threshold, the preset neural network training is completed.

[0042] Select any detection signal (i.e., sample detection signal) from the detection signal sample, input it into the preset temporal convolutional network model, extract the feature vector through the full convolution layer, perform classification through the full connection layer, and output the prediction probability of the specific type through the logistic regression layer. According to the prediction probability and the label of the specific type of the sample detection signal, calculate the loss value corresponding to the loss function and update the parameters. In the specific implementation process, the multi-classification cross entropy function can be selected as the loss function Loss: Where the base of log is e, M is the number of categories, and y ic is a sign function (0 or 1), which takes 1 when the true category of sample i is equal to c, and takes 0 otherwise; Pic is the predicted probability that the observed sample i belongs to category c; N is the number of samples.

[0043] On the one hand, it is possible to determine whether the loss value is less than a preset threshold value, if so, the training process is terminated, if not, a new sample detection signal is selected for training; on the other hand, it is possible to determine that the training times reach a preset times, and then stop the training.

[0044] In the specific implementation process, the training can be terminated after the entire training data set is iterated for a preset number of times, such as 30 times, to prevent overfitting problems caused by excessive training. The He initialization method can be used to control the variance of each layer parameter, and the Adam optimizer with a learning rate of 10-4 can be used to train the neural network model.

[0045] In some embodiments, a sliding window of a preset time length is used to perform data enhancement processing on each detection signal sample, so that each detection signal sample generates multiple detection signal samples after data enhancement processing; the combination of the detection signal and the specific type label corresponding to each detection signal sample is used as a training sample, specifically: the combination of the detection signal and the specific type label corresponding to each detection signal sample obtained after the data enhancement processing is used as a training sample.

[0046] After the detection signal is measured and obtained, it is necessary to divide the samples. Since the detection signal collection process is cumbersome, the sample size of the collected detection signal is often limited. In an embodiment of the present invention, a sliding window method is used for data enhancement, and the time length of the sliding window can be set according to demand. For example, the detection signal collected from a subject in a specific state lasts for 60 minutes (3600 seconds). With 10 seconds as the window and 1 second as the sliding step, a total of (3600-10) / 1+1=3591 samples can be generated. 3591*1=3591 samples can be divided out once per dose. Accordingly, the samples used for training are integrated according to the detection signal after data enhancement processing and its corresponding label. By using a sliding window of preset duration to perform data enhancement processing on each detection signal sample, a large number of training samples are obtained using a small number of collected samples, which simplifies the detection signal collection process and makes the sample acquisition process easier.

[0047] In one embodiment, the detection signal is preprocessed, including inputting the detection signal into a preset time convolution network model. Specifically, the detection signal after signal preprocessing is input into a preset time convolution network model; the signal preprocessing includes: DC component elimination and noise reduction processing.

[0048] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0049] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0050] like Figure 8 As shown, in an embodiment of the present invention, the present invention further provides a computer device, the above-mentioned computer device 12 is manifested in the form of a general computing device, and the components of the computer device 12 may include but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0051] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. For example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, an Audio Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0052] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device, including volatile and non-volatile media, removable and non-removable media.

[0053] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 31 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write to non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). Although Figure 8 Not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical medium) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 42, which are configured to perform the functions of various embodiments of the present invention.

[0054] A program / utility 41 having a set (at least one) of program modules 42 may be stored in, for example, a memory, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules 42, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0055] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, cameras, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed through an input / output (I / O) interface 22. Furthermore, the computer device 12 may also communicate with one or more networks (e.g., local area networks (LANs)), wide area networks (WANs), and / or public networks (e.g., the Internet) through a network adapter 20. Figure 8 As shown, the network adapter 21 communicates with other modules of the computer device 12 via the bus 18. It should be understood that although Figure 8 Not shown, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units 16, external disk drive arrays, RAID systems, tape drives, and data backup storage systems 34, etc.

[0056] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the graphic edge matching recognition and comparison method provided in the embodiment of the present invention.

[0057] That is, when the processing unit 16 executes the above program, it realizes: obtaining the drug to be detected and the detection reagent, and mixing them; transferring the mixed drug to be detected and the detection reagent to an incubation plate for sufficient reaction to generate a first mixed reagent; transferring the first mixed reagent to a magnetic separation recovery plate, and adding an excitation light liquid to generate a second mixed reagent; detecting the second mixed reagent through an optical component to generate a detection signal; inputting the detection signal into a preset time convolution network model, and obtaining the specific type of the object to be detected according to the output result of the time convolution network model.

[0058] Finally, it should be noted that the embodiments disclosed in the present invention are only preferred embodiments of the present invention, which 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 aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A cardiovascular drug development and testing device, characterized in that: include: A sample loading unit, which is loaded with a medicine test tube rack, on which a test tube for loading a medicine to be tested is provided; A first manipulator connected to a sampling needle, the first manipulator is used to transfer the detection reagent from the test kit to the incubation tray, and to transfer the drug to be detected from the test tube to the incubation tray to obtain a first mixed drug; The second manipulator is used to transfer the first mixed medicine in the incubation plate to the magnetic separation recovery plate; An optical component, used for detecting a second mixed reagent in the magnetic separation recovery disk, wherein the second mixed reagent is the first mixed reagent to which an excitation light liquid is added; An analysis unit, used for receiving a detection signal of the optical component, and inputting the detection signal into a preset temporal convolutional network model, and obtaining a specific type of the object to be detected according to an output result of the temporal convolutional network model; The detection signal includes a spectral signal, and the temporal convolutional network model is obtained after training based on detection signal samples with specific type labels.

2. The cardiovascular drug development and testing equipment according to claim 1, characterized in that: The two ends of the medicine test tube rack are provided with handles, and one side of the tube groove is provided with an opening.

3. The cardiovascular drug development and testing equipment according to claim 1, characterized in that: The sampling needle comprises a bracket, a first motor is disposed at the end of the bracket, an output end of the first motor is connected to a first screw rod disposed at one side of the bracket; a slider is disposed on the first screw rod; A guide rail is provided on the other side of the bracket, a sampling pump is installed on the guide rail, a needle tube is connected to the lower end of the sampling pump, and a needle is provided at the end of the needle tube; The sliding block is connected to the sampling pump.

4. The cardiovascular drug development and testing equipment according to claim 3, characterized in that: The sampling pump is provided with a connecting block, the connecting block is sleeved on the needle tube, and a spring is sleeved on the needle tube between the connecting block and the sampling pump.

5. A cardiovascular drug development and detection method, implemented using the cardiovascular drug development and detection device according to any one of claims 1 to 4, characterized in that: The steps include: Obtaining the drug to be tested and the testing reagent and mixing them; Transferring the mixed drug to be detected and the detection reagent to an incubation plate for sufficient reaction to generate a first mixed drug; Transferring the first mixed reagent to a magnetic separation recovery disk, and adding an excitation light liquid to generate a second mixed reagent; Detecting the second mixed medicine by an optical component to generate a detection signal; The detection signal is input into a preset temporal convolutional network model, and the specific type of the object to be detected is obtained according to the output result of the temporal convolutional network model.

6. The method according to claim 5, characterized in that The step of inputting the detection signal into a preset time convolutional network model, and obtaining the specific type of the object to be detected according to the output result of the time convolutional network model, comprises: Input each detection signal into the corresponding temporal convolution subnetwork in the preset temporal convolution network model; Among them, the temporal convolutional network model includes multiple temporal convolutional sub-networks, and the multiple temporal convolutional sub-networks respectively extract the features of the corresponding detection signals and then perform feature fusion, and the fused features are used as the features extracted by the temporal convolutional network model.

7. The method according to claim 5, characterized in that The step of inputting the detection signal into a preset temporal convolutional network model and obtaining the specific type of the object to be detected according to the output result of the temporal convolutional network model comprises: Inputting the detection signal into a fully convolutional network layer based on dilated convolution of the temporal convolutional network model, performing feature extraction on the detection signal using the fully convolutional network layer, and outputting a feature vector corresponding to the detection signal; The feature vector is input into the fully connected layer of the temporal convolutional network model, the output result of the fully connected layer is input into the logistic regression layer of the temporal convolutional network model, the predicted probability corresponding to the specific type is output, and the specific type of the object to be tested is obtained according to the predicted probability.

8. The method according to claim 5, characterized in that It also includes model training steps, including: Acquire multiple detection signal samples and specific labels corresponding to each detection signal sample; A combination of a detection signal and a specific type label corresponding to each detection signal sample is used as a training sample, thereby obtaining a plurality of training samples, and the temporal convolutional network model is trained using the plurality of training samples.

9. The method according to claim 8, characterized in that Training the temporal convolutional network model using the multiple training samples includes: Input any sample detection signal into the temporal convolutional network model, and output the predicted probability of the specific type corresponding to the sample detection signal; Calculating a loss value according to the predicted probability corresponding to the sample detection signal and the specific type label of the sample detection signal using a preset loss function; If the loss value is less than a preset threshold or the number of training times reaches a preset number, the training of the temporal convolutional network model is completed.

10. The method according to claim 9, characterized in that The step of obtaining a plurality of detection signal samples and a specific label corresponding to each detection signal sample further includes: Performing data enhancement processing on each detection signal sample using a sliding window of a preset time length, so that each detection signal sample generates a plurality of detection signal samples after data enhancement processing; The combination of the detection signal and the specific type label corresponding to each detection signal sample is used as a training sample, specifically: The combination of the detection signal and the specific type label corresponding to each detection signal sample obtained after the data enhancement process is used as a training sample.