Intelligent sample adding judgment method and intelligent test tube rack

Abstract

The invention relates to an intelligent sample adding judgment method and an intelligent test tube rack. Initializing the intelligent test tube rack, clearing the last sample adding data, and recovering the initial state; selecting a sample adding mode, wherein the sample adding mode comprises test tube positions, reagent types and sample adding amount information of each step of sample adding; sample adding is guided according to the sample adding mode, the controller controls the corresponding test tube indicating lamp to be turned on, and the display screen displays sample adding information; single-step judgment: judging whether the reagent type and the sample adding action are correct or not through an RFID reader-writer and an infrared probe, and judging whether the sample adding action accords with a preset similarity threshold or not by utilizing a deep learning model; sequence judgment: judging that the whole sample adding process is correct if all single-step sample adding is correct and the sequence accords with a preset rule, or else, giving an alarm for reminding; and storing and analyzing the sample adding data, uploading the data to a control platform for storage and analysis, and realizing data tracking and mining. The device improves the experiment efficiency and the result reliability, is suitable for sample adding operation of modern laboratories, and has a good application prospect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent sample adding of laboratory test tubes, in particular to an intelligent sample adding determination method and an intelligent test tube rack. BACKGROUND

[0002] In modern laboratories, test tubes, as one of the basic experimental instruments, are widely used in biology, chemistry, medicine and other fields. Traditional test tube racks have only basic functions of placing test tubes and cannot meet the needs of modern laboratories for sample management, data recording and environmental monitoring. The test tube racks in the prior art do not have intelligent functions, cannot realize intelligent determination of sample adding sequence and sample adding type, cannot store and display sample adding data, cannot determine the correctness of sample adding, are prone to sample adding errors and difficult to find, and affect the test results. SUMMARY

[0003] The present application aims to provide an intelligent sample adding determination method and an intelligent test tube rack to solve the problems of low intelligence level of test tube racks, difficult to find sample adding errors and complicated sample adding steps prone to errors in the prior art.

[0004] To achieve the above-mentioned purpose, the present application provides the following technical solution: an intelligent sample adding determination method, comprising the following steps,

[0005] Initializing the intelligent test tube rack, clearing the last sample adding data and restoring the initial state;

[0006] Selecting a sample adding mode, including information such as test tube position, reagent type and sample adding amount of each step;

[0007] Guiding sample adding according to the sample adding mode, the controller controls the corresponding test tube indicator light to light up and the display screen displays the sample adding information;

[0008] Single-step determination, determining whether the reagent type and sample adding action are correct through the RFID reader and infrared probe, and determining whether the sample adding action meets the preset similarity threshold value by using the deep learning model;

[0009] Sequence determination, if all single-step sample adding is correct and the sequence meets the preset rule, the whole sample adding process is determined to be correct, otherwise it is wrong and an alarm is given for reminding;

[0010] Storing and analyzing sample adding data, uploading the data to the control platform for storage and analysis, and realizing data tracking and mining.

[0011] As a further improvement of the present application, the deep learning model comprises:

[0012] Collecting sample adding operation videos, collecting and storing video data at a frame rate F;

[0013] Preprocessing video data, adjusting image size to size D, normalizing pixel value to range R, enhancing image contrast;

[0014] Training a convolutional neural network model using preprocessed image data, calculating gradients using backpropagation algorithm through forward propagation calculation, and updating model parameters using an optimizer;

[0015] Feature extraction is performed on the new sample operation image, and the feature vector Fv_img of the image is extracted, with a dimension of N;

[0016] Classify and determine the feature vector, calculate the probability P_i = exp(Fv_img·W_i) / Σ(exp(Fv_img·W_j)) belonging to each category, if the probability exceeds the confidence threshold P, it is determined that the action is of this category; otherwise, it is determined as unknown or abnormal action.

[0017] As a further improvement of the present application, the multi-sensor data fusion algorithm steps are as follows,

[0018] Collecting sensor data, including RFID tag information on reagent bottles read by RFID reader, occlusion state of test tube hole detected by infrared transceiver probe, and temperature and humidity data of experimental environment collected by temperature and humidity sensor;

[0019] Preprocessing the collected sensor data, decoding and format converting the RFID data, smoothing the binary state signal of the infrared transceiver probe, and linear fitting and normalizing the temperature and humidity data;

[0020] Extracting feature vectors of each sensor data, forming RFID feature vector Fv_rfid with dimension N1, infrared feature vector Fv_ir with dimension N2, and environment feature vector Fv_env with dimension N3;

[0021] Fusing each feature vector, using a weighted fusion algorithm, according to the reliability and correlation of each sensor data, respectively assigning different weight coefficients (W1, W2, W3) to RFID features, infrared features and environment features, to obtain the fused comprehensive feature vector Fv_fused = [Fv_rfid×W1, Fv_ir×W2, Fv_env×W3];

[0022] According to the fused comprehensive feature vector, input the fused comprehensive feature vector into the classification model based on support vector machine, and judge the sample state category to which the feature vector belongs.

[0023] As a further improvement of the present application, the sample guiding optimization recognition process based on reinforcement learning includes the following steps:

[0024] Initialize the intelligent test tube rack and observe the current sample adding environment state, and build an initial state vector (S0, dimension N4);

[0025] According to the current state vector, a suitable sample adding guide action is selected from a predefined action space, the expected return value O of each possible action is calculated, and the action with the maximum O value is selected as the current guide action;

[0026] After the selected sample adding guide action is executed, the experimenter adds samples according to the guide, and the relevant information is collected and recorded by the sensor in real time;

[0027] According to the sample adding result, the corresponding reward value is calculated, and the state, action and reward information of this time are stored in the experience replay buffer;

[0028] A batch of historical samples are randomly extracted from the experience replay, the parameters of the sample adding guide strategy network are updated by using the reinforcement learning algorithm, and the network weight is adjusted by minimizing the mean square error loss function L = Σ (Q_pred-Q_true) 2 between the predicted action value and the actual reward;

[0029] The above steps are repeated, and as the sample adding experiment continues, the intelligent test tube rack continuously observes the new state, executes the guide action, obtains the reward and updates the strategy, forming a stable and efficient sample adding guide strategy.

[0030] As a further improvement of the present application, the initialization of the intelligent test tube rack includes pressing the reset button (10) to clear the last sample adding data, and restoring the state of the intelligent test tube rack to the initial state to prepare for new sample adding operation.

[0031] Another aspect of the present application also provides an intelligent test tube rack, which comprises a rack body, a matched pipette, a matched reagent box and a control platform, the rack body is provided with test tube placing holes, state indicator lights, a liquid crystal display screen, an RFID reader / writer, a power indicator light, a power switch, a battery compartment, a reset button, a controller, a buzzer, a temperature and humidity sampler and an infrared reflection probe; the matched pipette is provided with an RFID chip, and the matched reagent box is provided with an RFID reader / writer; the control platform communicates with the intelligent test tube rack through a wireless networking mode.

[0032] As a further improvement of the present application, the rack body is made of high-strength, corrosion-resistant engineering plastic or aluminum alloy material, and is designed as a multi-layer structure, each layer can be independently controlled and monitored, and the layers are connected through a wireless communication module.

[0033] As a further improvement of the present application, the test tube placing hole is internally provided with a soft silicone rubber gasket, and the periphery is provided with an infrared reflection probe for detecting the sample adding condition of the matched pipette to the test tube

[0034] As a further improvement of the present application, the selected sample adding mode is selected by a preset sample adding mode through a button on the control platform or the intelligent test tube rack, and the sample adding mode can be imported into the control platform in advance according to the sample adding mode table, and the control platform sends the relevant sample adding mode to each intelligent test tube rack.

[0035] As a further improvement of the present application, the storage and analysis of the sample adding data includes uploading the collected sample adding data to the control platform through a wireless network, storing the data by the control platform, realizing the full life cycle tracking of sample adding, the traceability of sample adding record, compressing and storing the data by using a data compression algorithm, and deeply processing the data by using a smart data analysis and mining algorithm to mine valuable information.

[0036] The intelligent test tube rack of the present application realizes automatic recognition, real-time monitoring and accurate determination of the sample adding process by integrating RFID reader, temperature and humidity sampler, infrared transceiver probe and other elements with PLC controller, greatly improving the intelligent level of sample adding operation. Deep learning algorithm combined with multi-sensor data fusion technology ensures high precision and reliability of sample adding determination, effectively avoids sample adding errors, and ensures the accuracy of experimental results.

[0037] The sample adding guide optimization process based on reinforcement learning can automatically adjust the guide strategy according to the operation habits of the experimenters and different sample adding scenes, improve the adaptability and accuracy of the guide, and thus speed up the experimental process and improve the experimental efficiency. At the same time, the sample adding data is transmitted to the control platform through wireless network for storage and analysis, realizing real-time tracking and traceability of the data, providing strong data support for laboratory management, and facilitating experimenters to check historical records at any time, conduct data statistics and analysis.

[0038] In addition, the intelligent test tube rack has good expandability and can be flexibly customized and expanded according to the specific needs of the laboratory, such as increasing the number of test tube placement, expanding the types of supporting reagents, etc. Its operation is simple, the user interface is friendly, and the dependence on the professional experience of the operator is reduced. It can also monitor the environmental temperature and humidity in real time, ensure that the sample adding determination is not affected by environmental changes, improve the system stability, and reduce the experimental errors caused by environmental factors. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 Flowchart of the intelligent sample adding determination method of the present application;

[0040] Figure 2 Networking schematic diagram of the intelligent test tube rack of the present application;

[0041] Figure 3 Structure diagram of the intelligent test tube rack of the present application;

[0042] Figure 4The side view of the intelligent test tube rack of the present application;

[0043] Figure 5 The structural schematic diagram of the intelligent test tube rack matching equipment;

[0044] Figure 6 The front view of the intelligent test tube rack matching equipment;

[0045] Figure 7 The schematic diagram of the element layout of the intelligent test tube rack;

[0046] Figure 8 The wiring diagram principle diagram of the control element of the intelligent test tube rack;

[0047] Figure 9 The schematic diagram of the intelligent sample adding determination method

[0048] Figure 10 The logic diagram of the intelligent sample adding determination

[0049] Figure 11 The schematic diagram of the intelligent sample adding sequence determination flow DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0051] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of units is only a logical function division. In actual implementation, another division mode can be adopted, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed units can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0052] The present application also provides an intelligent sample adding determination method, comprising the following steps:

[0053] Step 1, initializing the intelligent test tube rack, pressing the reset button 10 to clear the last sample adding data, including sample adding record, error information and the like, restoring the state of the intelligent test tube rack to the initial state, and preparing for new sample adding operation.

[0054] Step 2, select the sample mode, select the preset sample mode through the button on the control platform or the intelligent test tube rack, and the sample mode can be set in advance according to the sample mode table and then introduced into the control platform. The control platform will issue relevant sample modes to each intelligent test tube rack, including the test tube position, reagent type, sample amount and other information of each step.

[0055] Step 3, guide the sample adding according to the sample mode, the controller 11 controls the test tube state indicating light 3 of the single step to be added to light up according to the selected sample mode, and the corresponding test tube liquid crystal display screen 5 displays the reagent and dose to be added, guiding the experimenters to perform sample adding operation.

[0056] Among them, determining whether the sample adding action is correct includes single step determination and sequential determination:

[0057] Single step determination: when the experimenter performs sample adding operation, first read the RFID chip 16 information on the matched pipette 15 through the RFID reader and writer 6 to obtain the current sample reagent type; at the same time, detect whether the matched pipette 15 performs sample adding operation to the test tube in the corresponding test tube placement hole 2 through the infrared reflection probe 14. Compare the obtained reagent type information with the preset reagent type information of the current sample adding step. If they are consistent, the reagent type is correct; if they are not consistent, the reagent type is incorrect. If the reagent type is correct and the sample adding action is detected, further identify and determine the sample adding action through the deep learning algorithm, use the pre-trained convolutional neural network (CNN) model to extract the features of the sample adding operation image collected by the camera to obtain the image feature vector (Fv_img), and calculate the similarity with the preset correct sample adding action image feature vector (Fv_img_std). The similarity formula is: Sim = cos (Fv_img, Fv_img_std). If the similarity reaches the preset threshold (S), the current sample adding action is correct; otherwise, the current sample adding action is incorrect.

[0058] Sequential determination: after all single step sample adding actions are completed, according to the determination results of each single step sample adding action and the sample adding sequence information, the entire sample adding process is determined through the preset sample adding sequence rule. If all single step sample adding actions are correct and the sample adding sequence meets the preset rule, the entire sample adding process is correct; otherwise, the entire sample adding process is incorrect, and an alarm is issued to stop the next sample adding guidance until the sample adding error is corrected.

[0059] The sample adding data is stored and analyzed, the collected sample adding data including sample adding time, sample adding amount, sample adding sequence, sample adding result, etc. is uploaded to the control platform through the wireless network, the control platform stores the data, realizes sample adding full life cycle tracking and sample adding record checking and tracing back. Data compression algorithm is used for compressed storage of data, and intelligent data analysis and mining algorithm is used for deep processing of data to mine valuable information such as sample adding error rate and experimental efficiency, thereby providing decision support for laboratory management, etc.

[0060] The steps of the deep learning algorithm are as follows:

[0061] The sample adding operation video is collected, the camera on the intelligent test tube rack is used to collect the sample adding operation video at a frame rate F, and the video data is stored in the local storage.

[0062] The collected video data is preprocessed, including adjusting the image size to a size D, normalizing the pixel value to a range R, and enhancing the image contrast to improve the image quality and facilitate subsequent model processing.

[0063] The preprocessed image data is used to train a convolutional neural network (CNN) model. The features of the image are extracted through convolutional layers, pooling layers and fully connected layers. In the initial stage, the model is randomly initialized with weights. The loss value Loss between the predicted result and the true label is calculated through forward propagation. Loss = Σ (y_true-y_pred)^2. The gradient is calculated using the backpropagation algorithm, and the model parameters are updated using the optimizer O. After multiple iterations of training, the model gradually learns the key features in the image, such as the shape of the pipette, the label of the reagent bottle, and the liquid level change of the test tube, and establishes a correlation with the class label. The training process lasts for a duration T, and the accuracy of the model on the validation set reaches a satisfactory level A.

[0064] The features of the new sample adding operation image are extracted, and the preprocessed image is input into the trained CNN model to extract the feature vector Fv_img with a dimension of N.

[0065] The feature vector is classified and judged, and the extracted feature vector is input into the Softmax classifier to calculate the probability distribution of the feature vector belonging to each pre-set sample adding action category. The probability calculation formula is: P_i = exp(Fv_img·W_i) / Σ(exp(Fv_img·W_j)), where W_i is the weight vector of category i. If the probability value of a certain category exceeds the set confidence threshold P, it is determined that the current sample adding action is the sample adding operation corresponding to that category; otherwise, if the probabilities of all categories are lower than the threshold, it is determined as an unknown or abnormal sample adding action, and the judgment result is fed back to the experimental personnel through the display element of the intelligent test tube rack, and recorded in the data storage module.

[0066] In further embodiments, the steps of the multi-sensor data fusion algorithm are as follows:

[0067] Collecting sensor data, including RFID tag information such as tag ID, reagent name, concentration, etc. read by RFID reader 6 on reagent bottles, binary signals representing the shielding state of test tube holes detected by infrared transceiver probe 14, 0 for unshielded, 1 for shielded, temperature and humidity data of the experimental environment collected by the temperature and humidity sensor, temperature unit °C, humidity unit %RH.

[0068] Preprocessing the collected sensor data, decoding and formatting the RFID data, extracting key information fields such as reagent name, concentration, etc.; smoothing the binary state signal of the infrared transceiver probe 14 to filter out temporary fluctuations caused by external interference; linear fitting and normalization of temperature and humidity data to convert the numerical range to the range R1 to adapt to the requirements of the subsequent fusion algorithm.

[0069] Extracting feature vectors of each sensor data, for RFID data, extracting features such as label ID update time difference, reagent name change, etc. before and after sampling, forming RFID feature vector Fv_rfid with dimension N1; for infrared transceiver probe 14 data, counting the number of times and duration of shielding and release within a unit time, etc. features, constructing infrared feature vector Fv_ir with dimension N2; extracting statistical features such as mean, variance, and change rate of temperature and humidity from temperature and humidity data, generating environment feature vector Fv_env with dimension N3.

[0070] Fusing each feature vector, using a weighted fusion algorithm, according to the reliability and correlation of each sensor data, respectively assigning different weight coefficients W1, W2, W3 to RFID features, infrared features, and environmental features. After multiplying each feature vector by the corresponding weight, the combined feature vector Fv_fused is obtained, with dimension N1+N2+N3. The fusion formula is: Fv_fused = [Fv_rfid × W1, Fv_ir × W2, Fv_env × W3].

[0071] According to the fused comprehensive feature vector, the fused comprehensive feature vector is input into a classification model based on a support vector machine (SVM). The SVM model determines the sample addition state category to which the feature vector belongs according to a pre-set classification boundary, such as "sample addition correct", "sample addition error-reagent error", "sample addition error-sequence error", and the like. The determination result is fed back to the experimental personnel in real time through a display and alarm system of the intelligent test tube rack, and is uploaded to a control platform for recording and storage. If the determination result is sample addition error, the system immediately suspends the subsequent sample addition guiding process until the experimental personnel corrects the error operation according to the prompt, re-triggers the sample addition action, and the determination result is correct, and then the subsequent sample addition guiding steps are continued.

[0072] In a further embodiment, the sample addition guiding optimization identification process based on reinforcement learning includes the following steps:

[0073] An intelligent test tube rack is initialized, and a current sample addition environment state is observed, including the current residual amount of reagents in each test tube in milliliters, a completed sample addition step sequence, characteristic parameters of past operation habits of the experimental personnel, such as average sample addition speed, common error type frequency, and the like, and information such as temperature and humidity of the laboratory environment, an initial state vector S0 is constructed, with a dimension of N4, for describing the starting condition of the sample addition task.

[0074] According to the current state vector S1, a suitable sample addition guiding action is selected from a pre-defined action space. The action space includes various possible guiding strategies, such as different combination modes of lighting specific test tube corresponding indicator lights, displaying specific sample addition step detailed instructions on the display screen, such as font size, color, flashing frequency, and the like, issuing voice prompts of different frequencies and lengths, and the like. When selecting an action, the learned strategy network is combined to calculate the expected return value O value of each possible action, O(s, a) = E[r_t + γ*maxO(s', a')], where s is the current state, a is the current action, r_t is the current reward, γ is the discount factor, s' is the next state, and a' is the next action. The action with the maximum O value is selected as the current guiding action.

[0075] After the selected sample addition guiding action is performed, the experimental personnel performs sample addition operation according to the guidance. Related information such as sample addition time and reagent amount during the operation is collected and recorded in real time by the sensor, for subsequent reward calculation and strategy updating.

[0076] After the experimenters complete the sample adding operation, the corresponding reward value is calculated according to the sample adding result. If the sample adding is correct and meets the preset sample adding mode, a positive reward V+ is given, and additional reward points V1+ are given according to the smoothness of the sample adding operation, such as whether the sample adding speed is within a reasonable range, whether there is additional operation adjustment, etc. If the sample adding is incorrect, a negative reward V- is calculated according to the error type, such as reagent error, sequence error, measurement error, etc. The reward calculation formula is: Reward = V+ + V1+ when correct, Reward = V- when incorrect. The current state, action and reward information are stored in the experience replay buffer.

[0077] A batch of historical samples are randomly extracted from the experience replay, and the parameters of the sample adding guidance policy network are updated using reinforcement learning algorithms such as deep Q network, DQN. By minimizing the mean square error loss function between the predicted action value and the actual reward, the network weights are adjusted, and the policy network gradually learns the rules of selecting the optimal action in different states to maximize the long-term cumulative reward. The loss function formula is: L = Σ(Q_pred - Q_true)^2, where Q_pred represents the predicted action value and Q_true represents the actual reward. The update process uses an optimization algorithm, and the learning rate is set to L. Network parameter update is performed once every C sample adding operations.

[0078] The above steps are repeated, and as the sample adding experiment continues, the intelligent test tube rack continuously observes new states, performs guidance actions, obtains rewards and updates the policy. Through continuous trial and error and learning, the reinforcement learning model gradually converges to form a stable and efficient sample adding guidance policy. For example, through multiple iterations, the system learns the operation habits of different experimenters and sample adding scenarios, and automatically adjusts the intensity and method of guidance actions, such as reducing the frequency of voice prompts for experimenters with high operation proficiency, and only reminding them at key steps; while for novice experimenters, the diversity and prompting of guidance actions are increased to ensure the smooth progress of the sample adding process. The specific time of iteration optimization for a period of T1 depends on the experiment frequency and sample quantity, and the accuracy of sample adding guidance can be stabilized above level A1, significantly improving the sample adding efficiency and accuracy, and reducing the sample adding error rate caused by improper guidance.

[0079] As shown in Figure 2 , the intelligent test tube rack accesses the control platform through a wireless networking mode to realize management of the intelligent test tube rack. The matching reagent kit interacts with the intelligent test tube rack through wireless transmission to provide reagent information for intelligent sample adding judgment of the intelligent test tube rack.

[0080] As shown in Figures 3-4As shown, the intelligent test tube rack includes a rack body 1, a matching pipette 15, a matching reagent box 17, and a control platform. The rack body 1 is provided with a test tube placing hole 2 for placing a test tube; a state indicating lamp 3 is used to display the working state of the test tube rack; a liquid crystal display screen 5 is used to display the test tube quantity, state information, and environmental parameters; an RFID reader / writer 6 is used to read the RFID chip 16 on the matching pipette 15 and the RFID tag information on the matching reagent box 17, to realize the identity recognition and management of the pipette and the reagent box; a power indicator 7 is used to display the power state; a power switch 8 is used to control the power on / off; a battery compartment 9 provides a backup power supply for the test tube rack; a reset button 10 is used to manually reset the controller 11; the controller 11 as a core component, connects and controls all the above components, to realize the intelligent management of the test tube rack; a buzzer 12 is used to issue a sound prompt to remind the user to operate or alarm; a temperature and humidity sampler 13 is used to collect environmental parameters to ensure the suitability of the test tube storage environment; an infrared reflection probe 14 is used to detect the placement state of the test tube in the test tube placing hole 2. The matching pipette 15 and the matching reagent box 17 respectively interact with the test tube rack through the RFID chip 16 and the RFID reader / writer 18. The control platform communicates with the controller 11 of the test tube rack through a wireless networking mode, to realize remote monitoring and management. The components work cooperatively, the state of the test tube in the test tube placing hole 2 is detected by the infrared reflection probe 14 and fed back to the controller 11, the controller 11 controls the state indicating lamp 3 to display the state according to the detection result, and displays the related information on the liquid crystal display screen 5; the RFID reader / writer 6 reads the information of the matching pipette 15 and the reagent box 17, to help manage the experimental supplies; the temperature and humidity sampler 13 monitors the environmental parameters in real time, once the preset threshold is exceeded, the controller 11 will trigger the buzzer 12 to alarm and display the abnormal state; the power switch 8 controls the power on / off, the battery compartment 9 provides a backup power supply to ensure that the device can still maintain basic functions when power is off; the reset button 10 can manually reset the controller 11 when the system is abnormal, to restore the normal operation of the system. Through this design, the intelligent test tube rack realizes the intelligent management of the test tube, environmental monitoring, information interaction, and remote monitoring, etc., to improve the efficiency and safety of laboratory work.

[0081] As shown, Figures 5-6 The intelligent test tube rack is used with the matching pipette 15 and the matching reagent box 17. The matching pipette 15 is provided with an RFID chip 16, and the matching reagent box 17 is provided with an RFID reader / writer 218.

[0082] In a further embodiment, the frame 1 is used to place test tubes and has test tube placement holes 2. Data acquisition components such as a temperature and humidity sampler 13, an infrared beam detector 14, and an RFID reader / writer 16 are arranged on the frame 1 to collect sample addition data and provide it to the data processing and control unit. A controller 11 is arranged on the frame 1 to control and process relevant data information of the intelligent test tube rack, serving as the data processing and control unit for the intelligent test tube rack. Status indicator lights 3, power indicator lights 7, and an LCD screen 5 are arranged on the frame 1 to display and remind users of the sample addition status of the test tubes. A communication module is arranged on the controller 11 to enable network communication between the intelligent test tube rack controller 11 and the control platform and the matching reagent kit 17.

[0083] Intelligent test tube rack example: Relevant intelligent monitoring equipment is arranged on rack 1, such as... Figures 2-7 As shown, an infrared beam probe 14 is arranged on the test tube placement hole 2 to detect the sample addition of the matching pipette 15 into the test tube. The status indicator 3 indicates the sample addition status. The button 4 is used for manual adjustment and selection of relevant sample addition information. The LCD screen 5 is used to display the sample addition information and status of the corresponding test tube. The RFID reader 16 is used to read the reagent information of the matching pipette 15. The power indicator 7, power switch 8, and battery compartment 9 are the power supply for the intelligent test tube rack and related control and indication devices to power the intelligent test tube rack. The controller 11 collects relevant information and controls and determines the sample addition order. The buzzer 12 is the alarm for the intelligent test tube rack. The temperature and humidity sampler 13 collects the temperature and humidity data of the test environment.

[0084] The element is in accordance with Figure 6 As shown, the tubes are arranged on test tube rack 1, according to... Figure 7 The wiring diagram of the components shown is connected to the controller. The controller 11 controls all components, reads relevant information, and controls the operation of the components. In this embodiment, the RFID reader 16 is powered by a power supply and controlled by the PLC controller.

[0085] When the matching pipette 15 takes a sample from the matching reagent kit 17, the RFID reader 218 arranged on the matching reagent kit 17 writes the reagent information into the RFID chip 16 on the matching pipette 15, so that the RFID reader 16 can read the relevant reagent information when the matching pipette 15 adds a sample to the smart test tube rack.

[0086] When the matching pipette 15 adds samples into the test tube, the infrared through-beam probe 14 arranged on the test tube well 2 identifies the sample addition action, the RFID reader 16 reads the relevant reagent information stored in the RFID chip 16, and the relevant information is transmitted to the controller, thus enabling the collection and storage of relevant data of the smart test tube rack.

[0087] Furthermore, the intelligent test tube rack, equipped with an intelligent sample addition determination method, can intelligently guide the sample addition order and intelligently determine the sample addition results.

[0088] As shown in Figures 8-11 , the intelligent sample adding judgment system includes initialization, sample adding mode selection, sample adding guidance, sample adding step judgment, and data storage analysis.

[0089] Initialization, initialize the intelligent test tube rack data, and prepare for the sample adding process.

[0090] Sample adding mode selection, select the preset sample adding mode, and select the mode. After that, the subsequent sample adding is performed according to the sample adding mode. The sample adding mode can be set in advance according to "Table 1 Sample Adding Mode Table" and imported into the control platform. The control platform will issue the relevant sample adding mode to each intelligent test tube rack. When performing sample adding, the corresponding sample adding mode is selected. The subsequent intelligent test tube rack will guide and intelligently judge the sample adding steps according to the selected sample adding mode.

[0091] Sample adding guidance, according to the selected sample adding mode, the intelligent test tube rack determines the sample adding test tube and reagent needed for this step by judging the sample adding sequence. The corresponding test tube side LED indicator light is turned on. The display screen displays the reagent and dosage that need to be added, guiding the sample adding step execution.

[0092] Sample adding step judgment, which is divided into single-step judgment and sequential judgment. The judgment logic is shown in Figure 9 .

[0093] Single-step judgment needs to determine whether the single-step sample adding is executed, whether the sample adding step is consistent with the set step, and whether the sample adding reagent is consistent with the set reagent. Only when the single-step sample adding is executed, the sample adding step is correct, and the sample adding reagent is correct, it is determined that the single-step sample adding is correct. Among them, whether the single-step sample adding is executed needs to determine whether the reagent information collected by the RFID patch is consistent with the set reagent information, and the corresponding test tube infrared reflection probe collects that the sample adding has been executed. Only when both meet the requirements, it is determined that the single-step sample adding has been executed.

[0094] After the single-step sample adding judgment is correct, the next step of sample adding is guided according to the sample adding sequence, and the related indicator light is turned on.

[0095] Sequential judgment is the overall judgment of the entire sample adding work. Only when all single-step sample adding is correct, the sample adding sequence is correct, the sample adding measurement is correct, and the sample adding step is executed, it is determined that the sample adding sequence is correct. Among them, the sample adding measurement judgment is realized by converting the measurement into the judgment of the corresponding sample adding times.

[0096] When the single-step sample adding or sequential sample adding is incorrect, the intelligent test tube rack will issue an alarm to remind that the sample adding process is incorrect, stop the next step of sample adding guidance, and until the sample adding error is corrected, the alarm can be eliminated.

[0097] The data storage analysis is to send the collected data to the control platform through the wireless network, and the platform stores the data to realize the whole life cycle tracking of sample adding and the traceability of sample adding record.

[0098] The intelligent sample adding judgment embodiment is shown in FIG. 1, which is a sample adding judgment process. The sample adding mode 1 in Table 1 is set as the sample adding mode 1, and the specific implementation of the intelligent sample adding judgment is described. Figure 8 First, the intelligent test tube rack is initialized, and the reset button 10 on the test tube rack is pressed to clear the sample adding data.

[0099] The sample adding mode 1 is selected through the control platform or the button on the intelligent test tube rack, and the controller 11 reads the data corresponding to the sample adding mode 1.

[0100] The controller 11 controls the single-step test tube 1 state indicator light 3 to light up according to the sample adding sequence of the sample adding mode 1, and the corresponding test tube liquid crystal display screen 5 displays the reagent and the reagent dose A, 1ml, guiding the experimenters to add samples.

[0101] The experimenters perform single-step sample adding operation according to the sample adding mode, and when the reagent A is taken according to the sample adding guidance, the RFID reader / writer 218 on the matched reagent box 17A writes the reagent A information into the RFID chip 16 of the matched pipette 15, and the RFID reader / writer 16 arranged on the intelligent test tube rack reads the reagent A information when the pipette 15 adds samples in the test tube 1. At the same time, the infrared opposite radiation probe 14 on the test tube 1 corresponding test tube hole identifies the sample adding action, and determines that the test tube 1 has been added. The intelligent test tube rack collects relevant information, and the controller 11 compares the sample adding reagent information of the test tube with the set single-step sample adding information. If they are consistent, it is determined that the single-step sample adding is correct, and the next single-step sample adding is guided. If they are inconsistent, an alarm is given to remind the sample adding error.

[0102] After the single-step sample adding is correct, the intelligent test tube rack intelligently guides the second single-step sample adding, the test tube 2 corresponding indicator light 3 is lit up, and the corresponding liquid crystal display screen 5 displays the sample adding information reagent A, 2ml. After completing the sample adding once, the display screen displays that the sample adding has been completed once. After the second sample adding is completed, it is determined that the single-step sample adding is completed, and the intelligent test tube rack guides the next single-step sample adding. If the sample adding is only performed once, and other test tubes are added, an alarm will be given.

[0103]

[0104] ​After the tester completes all single-step dispensing of step one, the intelligent test tube rack determines that the dispensing of step one is correct, and step one is completed, then enters step two. After completing all single-step dispensing of step two, the dispensing of step two is completed, and the process is repeated until all steps are completed. When all dispensing steps are correct, the dispensing sequence is normal, and all dispensing steps have been executed, the intelligent test tube rack determines that the dispensing sequence is correct. If an error occurs in any dispensing step or single step, the intelligent test tube rack will issue an alarm and stop the next dispensing guide.

[0105] After all dispensing steps are completed according to the set mode, the intelligent test tube rack prompts that the dispensing is normal, and the test is completed.

[0106] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit. The above is only an embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process conversion using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

[0107] The specific embodiments of the application are described in detail above, but they are only examples. The present application is not limited to the above-described specific embodiments. Any equivalent modification or substitution of the present application by those skilled in the art is also within the scope of the present application. Therefore, any equivalent transformation, modification, improvement, etc. made without departing from the spirit and principle range of the present application should be included in the scope of the present application.

Claims

1. A smart sample loading decision method, characterized in that: The method comprises the following steps: Initialize the intelligent test tube rack, clear the last time sample data, and restore the initial state; Select the sample adding mode, including the test tube position, reagent type and sample adding amount information of each step; According to the sample adding mode, guide the sample adding, the controller controls the corresponding test tube indicating light to light up, and the display screen displays the sample adding information; Single step judgment, judge whether the reagent type and sample adding action are correct through RFID reader and infrared probe, and judge whether the sample adding action meets the preset similarity threshold value by using deep learning model; Sequential judgment, if all single step sample adding is correct and the sequence meets the preset rule, the whole sample adding process is correct, otherwise it is wrong and an alarm is given; Store and analyze the sample adding data, upload the data to the control platform for storage and analysis, and realize data tracking and mining.

2. The method of claim 1, wherein: The deep learning model comprises: Collect sample adding operation video, collect and store video data at frame rate F; Preprocess the video data, adjust the image size to size D, normalize the pixel value to range R, and enhance the image contrast; Train the convolutional neural network model using the preprocessed image data, calculate the gradient by forward propagation and back propagation algorithm, and update the model parameters using the optimizer; Feature extraction is performed on the new sample adding operation image to extract the feature vector Fv_img of the image with a dimension of N; Classify and determine the feature vector, calculate the probability of belonging to each category: P_i=exp(Fv_img·W_i) / Σ(exp(Fv_img·W_j)); Where W_i is the weight vector of category i, W_j is the weight vector of category j, if the probability exceeds the confidence threshold P, it is determined that the action is of this category; otherwise, it is determined as unknown or abnormal action.

3. The method of claim 2, wherein: The multi-sensor data fusion algorithm steps are as follows, Collect sensor data, including RFID tag information on reagent bottles read by RFID reader, shielding state of test tube hole detected by infrared transceiver probe, and temperature and humidity data of experimental environment collected by temperature and humidity sensor; Preprocess the collected sensor data, decode and format convert the RFID data, smooth the binary state signal of the infrared transceiver probe, and linearly fit and normalize the temperature and humidity data; Extract feature vectors of each sensor data to form RFID feature vector Fv_rfid with dimension N1, infrared feature vector Fv_ir with dimension N2, and environment feature vector Fv_env with dimension N3; Fuse the feature vectors, use weighted fusion algorithm, and give different weight coefficients (W1, W2, W3) to RFID features, infrared features and environment features according to the reliability and correlation of each sensor data to obtain the fused comprehensive feature vector Fv_fused=[Fv_rfid×W1, Fv_ir×W2, Fv_env×W3]; Determine according to the fused comprehensive feature vector, input the fused comprehensive feature vector into the classification model based on support vector machine to judge the sample adding state category to which the feature vector belongs.

4. The intelligent sample loading determination method of claim 3, wherein: The sample adding guide optimization recognition process based on reinforcement learning comprises the following steps: Initialize the intelligent test tube rack and observe the current sample adding environment state, and build an initial state vector (S0, dimension N4); According to the current state vector, select a suitable sample adding guide action from the predefined action space, calculate the expected return value O of each possible action, and select the action with the maximum O value as the current guide action; After executing the selected sample adding guide action, the experimenter adds samples according to the guide, and the relevant information is collected and recorded by the sensor in real time; According to the sample adding result, the corresponding reward value is calculated, and the state, action and reward information of this time are stored in the experience replay buffer. A batch of historical samples are randomly extracted from the experience replay, and the parameters of the augmented guidance policy network are updated using reinforcement learning algorithms by minimizing the mean square error loss function L = Σ(Q_pred - Q_true) between the predicted action value and the actual reward 2 , adjust the network weights; Repeat the above steps, and as the sample adding experiment continues, the intelligent test tube rack continuously observes the new state, executes the guide action, obtains the reward, and updates the strategy to form a stable and efficient sample adding guide strategy.

5. The method of claim 1, wherein: It also includes a reset button (10) to clear the last sample adding data and restore the state of the intelligent test tube rack to the initial state for new sample adding operation.

6. The method of claim 1, wherein: The selected sample adding mode is selected by a button on the control platform or the intelligent test tube rack to select the pre-set sample adding mode. The sample adding mode can be set in advance according to the sample adding mode table and imported into the control platform. The control platform sends the relevant sample adding mode to each intelligent test tube rack.

7. The method of claim 1, wherein: The storage and analysis of sample adding data includes uploading the collected sample adding data to the control platform through a wireless network, storing the data by the control platform, realizing the full life cycle tracking of sample adding, and making the sample adding record traceable and backtraceable. Data compression algorithm is used for compressed storage, and intelligent data analysis and mining algorithm is used for deep processing of data to mine valuable information.

8. A smart test tube rack for implementing the smart pipetting method of any one of claims 1 to 7, characterized in that: It includes a rack body (1), a matching pipette (15), a matching reagent box (17) and a control platform. The rack body (1) is provided with test tube placing holes (2), state indicator lights (3), liquid crystal display screens (5), RFID readers / writers (6), power indicator lights (7), power switches (8), battery compartments (9), reset buttons (10), controllers (11), buzzers (12), temperature and humidity samplers (13), infrared reflection probes (14); The matching pipette (15) is provided with an RFID chip (16), and the matching reagent box (17) is provided with an RFID reader / writer (18); The control platform communicates with the intelligent test tube rack through a wireless networking mode.

9. The intelligent tube rack of claim 8, wherein: The rack body (1) is made of high-strength, corrosion-resistant engineering plastic or aluminum alloy material and is designed as a multi-layer structure, each layer can be independently controlled and monitored, and the layers are connected through a wireless communication module.

10. The intelligent test tube rack of claim 8, wherein: The test tube placing hole (2) is provided with a soft silicone rubber gasket inside and infrared reflection probes (14) around it for detecting the sample adding of the matching pipette (15) into the test tube.

Images