Blood sample analysis method and blood sample detection equipment

Abstract

The invention discloses a blood sample analysis method and blood sample detection equipment, and the blood sample analysis method comprises the steps: collecting a blood sample image through an image collection module arranged on the blood sample detection equipment, and carrying out the preprocessing and standardization processing of the blood sample image; constructing a CTC sample database and an AI model, inputting the blood sample image into the trained AI model, performing AI identification on the blood sample image, and generating an identification result; comparing the identification result with data in a CTC sample database to obtain a comparison result; and performing feature analysis on the comparison result to obtain a detection result, and generating a detection report. The blood sample analysis method is mainly based on an AI technology, AI identification is carried out on an obtained blood sample image, the blood sample image is detected and analyzed through an AI algorithm, and the blood sample detection efficiency and accuracy are improved. The blood sample detection equipment can be used for automatically extracting and purifying a blood sample, efficiently completing sample treatment before detection, reducing errors in the treatment process and improving the efficiency and the detection accuracy.

Description

Technical Field

[0001] This application relates to the field of blood sample testing technology, and in particular to a blood sample analysis method and blood sample testing equipment. Background Technology

[0002] In routine experiments, whole blood sample sorting involves multiple steps. With large sample volumes, inaccurate results or low efficiency may occur. Traditional manual operation relies on operators manually processing samples, sorting cells, and analyzing them. This is not only cumbersome but also susceptible to human error, such as pipetting errors and cell identification deviations, leading to significant fluctuations in results. When sample volumes surge, manual operation speed cannot meet timeliness requirements, and prolonged high-intensity work can easily cause fatigue, further increasing the risk of operational errors and resulting in missed or misjudged key data. Furthermore, the standardization of manual operation is difficult to guarantee; differences in the skill levels of different operators can lead to inconsistent and incomparable results, severely limiting the reliability and application scope of CTC detection. Therefore, developing a fully automated CTC instrument has become an urgent need to improve detection efficiency and accuracy and promote the development of CTC detection technology. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a blood sample analysis method that can solve the problems of inaccurate analysis results and low analysis efficiency in the prior art.

[0004] This application also proposes a blood sample testing device that uses the above-mentioned blood sample analysis method, which can solve the problems of poor sample processing and blood sample analysis efficiency and inaccurate results.

[0005] The blood sample analysis method according to the first aspect of this application includes the following steps: Blood sample images are acquired by an image acquisition module installed on the blood sample testing equipment, and the blood sample images are preprocessed and standardized. A CTC sample database and an AI model are constructed. Preprocessed and standardized blood sample images are input into the trained AI model to perform AI recognition on the blood sample images and generate recognition results. The identification results are compared with the data in the CTC sample database to obtain the comparison results; Feature analysis is performed on the comparison results to obtain the detection results and generate a detection report.

[0006] The blood sample analysis method according to the embodiments of this application has at least the following beneficial effects: By installing an image acquisition module on the blood sample testing equipment, blood sample images can be acquired efficiently and accurately. Preprocessing the acquired images removes interference noise and enhances the CTC (Cellular Toxic Organism) detail features, improving the efficiency and accuracy of blood sample testing. Standardizing the acquired images eliminates image differences caused by different imaging devices and parameters, further improving detection accuracy. By constructing a CTC sample database and an AI model, AI is used for comparison and analysis of blood sample data, significantly improving the efficiency and accuracy of blood sample testing.

[0007] According to some embodiments of this application, constructing the CTC sample database includes the following steps: Collect CTC images from clinical whole blood samples, input the CTC images into a CTC sample database, and perform preprocessing and standardization on the CTC images; The preprocessing steps are as follows: First, the CTC image is denoised to preserve CTC detail features and reduce interference; then, image enhancement is performed to highlight the CTC contour and fluorescence signal. The standardization process is as follows: the size and pixel values ​​of the CTC images input into the CTC sample database are adjusted to uniform values; Expand the CTC sample database to increase the dataset size; The CTC images are classified into multiple first sample sets based on different tumor types; each first sample set is further classified into multiple second sample sets based on different tumor morphologies; each second sample set is further classified into multiple third sample sets based on different clinical stages. Images of normal blood cells were collected and entered into the CTC sample database to establish a comparison sample set.

[0008] According to some embodiments of this application, the detection results and the blood sample image are synchronized to the CTC sample database.

[0009] According to some embodiments of this application, the AI ​​recognition includes the following steps: By using an attention mechanism module installed on the blood sample testing device, the system automatically focuses on the key distinguishing regions of CTCs, reducing interference from irrelevant features; By using a multi-feature fusion module installed on the blood sample detection device, the morphological and molecular features of CTCs in the blood sample image are extracted simultaneously. The morphological features include the cell outline, the ratio of cell nucleus to cytoplasm, and the edge smoothness of CTCs. The molecular features include the fluorescence intensity and spatial distribution density of fluorescently labeled proteins. The extracted morphological and molecular features of CTC were compared with data in the database.

[0010] According to some embodiments of this application, constructing the AI ​​model includes the following steps: Using a pre-trained ResNet-50 model as the basic architecture, the parameters of the first 10 layers were frozen, and the subsequent layers and the newly added multi-feature fusion module and attention mechanism module were fine-tuned; the kernels of the first 10 convolutional layers were frozen to retain the general feature extraction capability. The difference between the model's predicted values ​​and manually labeled values ​​is calculated using the cross-entropy loss function, and the learning rate is dynamically adjusted using the Adam optimizer. Mini-batch stochastic gradient descent is used for training, with 50 training epochs. An early stopping mechanism is also introduced, which stops training when the validation set loss does not decrease for 10 consecutive training epochs to avoid model overfitting.

[0011] According to some embodiments of this application, the construction of the AI ​​model further includes the following steps: By analyzing the model's prediction results on the validation set using a confusion matrix, and increasing the number of samples in the corresponding category for CTC samples with high misclassification, the model is retrained accordingly. Model pruning techniques are employed to remove redundant neurons and connections in the network, reducing the number of model parameters (by more than 30%) while maintaining model accuracy, thereby improving the model's inference speed on the device and ensuring real-time detection requirements. Finally, 5-fold cross-validation is used to evaluate the model's performance, ensuring its stability across different subsets of data.

[0012] A blood sample testing device according to a second aspect of this application, used to implement the above-described blood sample analysis method, includes: A workbench is provided with multiple rows of test tube slots, which are used to place test tubes for different processes. A pipetting mechanism, which is disposed above the workbench, is used to transfer liquids from test tube troughs on the workbench; A centrifugation mechanism is provided on one side of the workbench. The centrifugation mechanism includes a centrifuge and a robotic arm. The robotic arm is used to pick up test tubes from the test tube tank and place them onto the centrifuge. A first shifting mechanism is connected to the worktable; The second shifting mechanism is connected to the robotic arm.

[0013] The blood sample testing device according to the embodiments of this application has at least the following beneficial effects: The embodiments of this application reduce manual intervention through automated operation, greatly improving the speed of sample processing. For example, traditional manual sample processing may take several hours or even days, while the CTC fully automated pretreatment equipment can process multiple samples in a short time, improving detection efficiency.

[0014] This application's embodiments utilize automated equipment to precisely control operating conditions, such as time, solvent volume, and centrifugation parameters, reducing human error. In a laboratory environment, human factors can lead to deviations in sample processing results, while CTC equipment, through programmed control, ensures consistency in each processing step. For example, in drug development, the detection of drug metabolites requires highly accurate sample processing; CTC equipment can guarantee that the repeatability error of extraction and purification is within an extremely small range.

[0015] The blood sample testing equipment of this application embodiment can reduce direct contact between operators and samples when processing some toxic and hazardous samples, such as chemical reagents and biological samples. For example, when processing environmental samples containing heavy metals, the CTC equipment can be operated in a closed environment, avoiding operator contact with hazardous substances and reducing occupational health risks.

[0016] The blood sample testing equipment of this application embodiment can reduce labor costs and reagent waste. For example, automated equipment can reduce reagent costs by precisely controlling reagent dosage and avoiding excessive reagent use. At the same time, due to the increased processing efficiency, the laboratory can process more samples, improving equipment utilization and reducing the processing cost per sample.

[0017] In this embodiment, the system also includes a fluorescence scanning module, an optical imaging module, a multi-feature fusion module, an attention mechanism module, and an analysis and control module.

[0018] In this embodiment of the application, the workbench includes a first tabletop and a second tabletop, with a gap between the first tabletop and the second tabletop. The first test tube slot, the second test tube slot, the third test tube slot, and the fourth test tube slot are disposed on the first tabletop, and the fifth test tube slot and the sixth test tube slot are disposed on the second tabletop. A magnet is disposed on the gap, and the fifth test tube slot is disposed on one end of the second tabletop near the gap.

[0019] In this embodiment of the application, a pipette tip is detachably provided on the pipetting mechanism, and a pipette tip groove is provided on the worktable for placing the pipette tip.

[0020] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0021] The accompanying drawings are used to provide a further understanding of the technical solutions disclosed in this application and form part of the specification. They are used together with the embodiments disclosed in this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions disclosed in this application.

[0022] Figure 1 This is a schematic diagram of the overall structure of an embodiment of this application; Figure 2 This is a schematic diagram of the workbench structure in an embodiment of this application; Figure 3 This is a schematic diagram of the pipetting mechanism in an embodiment of this application; Figure 4 This is a schematic diagram of the frame structure according to an embodiment of this application; Figure 5 This is a schematic diagram of the hardware working environment in an embodiment of this application.

[0023] Reference numerals: Blood sample testing equipment 100; Workbench 110; First test tube trough 111; Second test tube trough 112; Third test tube trough 113; Fourth test tube trough 114; Fifth test tube trough 115; Sixth test tube trough 116; Seventh test tube trough 117; First work surface 118; Second work surface 119; Sample test tube 1110; Reaction test tube 1120; Pipetting mechanism 120; Pipetting body 121; Pipette tip 122; Connecting part 123; Third guide rail 124; Third drive mechanism 125; Tip slot 126; Centrifugation mechanism 130; Centrifuge 131; Robotic arm 132; Frame 140; First guide rail 161; First slider 162; First shifting mechanism 160; Second shifting mechanism 170; Second guide rail 171; Gap 190; Magnet 191. Detailed Implementation

[0024] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0025] In the description of this application, it should be understood that the descriptions of orientation, such as up, down, front, back, left, right, etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. If "first" or "second" is mentioned, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of technical features indicated, or implicitly indicating the order of the technical features indicated.

[0026] This application provides a blood sample analysis method and a blood sample testing device. The blood sample analysis method is mainly based on AI technology, which performs AI recognition on the acquired blood sample images and uses AI algorithms to detect and analyze the blood sample images, thereby improving the efficiency and accuracy of blood sample testing. The blood sample testing device can automatically extract and purify blood samples, efficiently complete the sample processing before testing, reduce errors in the processing process, and improve efficiency and detection accuracy.

[0027] like Figure 1 As shown, Figure 1 The image below is a schematic diagram of the overall structure of a blood sample testing device 100 according to an embodiment of this application. The blood sample testing device 100 can be used to implement a blood sample analysis method according to another embodiment of this application. The blood sample testing device 100 includes a workbench 110, a pipetting mechanism 120, a centrifugation mechanism 130, a frame 140, a first transfer mechanism 160, and a second transfer mechanism 170. A first test tube slot 111, a second test tube slot 112, a third test tube slot 113, a fourth test tube slot 114, a fifth test tube slot 115, a sixth test tube slot 116, and a seventh test tube slot 117 are sequentially arranged on the workbench 110. The workbench 110 is used for blood sample processing. The pipetting mechanism 120 is arranged above the workbench 110 and is used to transfer the liquid in each test tube slot on the workbench 110. like Figure 1 As shown, the first test tube 111 contains a sample test tube 1110, the second test tube 112 contains a reaction test tube 1120, the third test tube 113 contains a reagent test tube, the fourth test tube 114 contains a recovery test tube, the fifth test tube 115 contains a purification test tube, the sixth test tube 116 contains a finished product test tube, and the seventh test tube 116 contains a washing test tube.

[0028] The pipetting mechanism 120 includes a pipetting body 121 and a pipetting tip 122. The pipetting tip 122 is a hollow structure and is connected to a negative pressure generating device. The negative pressure generating device can be located inside the pipetting body 121 or connected to the outside of the pipetting mechanism 120. Preferably, the negative pressure generating device is a miniature vacuum pump integrated inside the pipetting body 121.

[0029] The pipetting mechanism 120 is provided with a third guide rail 124, and the pipetting body 121 is slidably connected to the third guide rail 124. A third drive mechanism 125, which is a cylinder, is connected to the pipetting body 121. The pipetting body 121 can move up and down under the drive of the third drive mechanism 125.

[0030] The pipette tip 122 is detachably connected to the pipette body 121, which facilitates the replacement of the pipette tip 122 during and after sample processing, prevents residual liquid in the pipette tip 122 from causing sample contamination, and improves detection accuracy.

[0031] Furthermore, the pipetting body 121 is provided with a connecting part 123, and the upper part of the pipetting tip 122 has an opening. The pipetting tip 122 is fitted onto the connecting part 123 through the opening. Specifically, the opening of the pipetting tip 122 can be made of an elastic structure or material to facilitate fitting the pipetting tip 122 onto the connecting part 123 and to facilitate removing the pipetting tip 122.

[0032] Furthermore, the workbench 110 may be provided with a pipette tip groove 126, which is used to place replaceable pipette tips 122. By placing the pipette tip 122 in the pipette tip groove 126, after the pipette tip on the connecting part 123 is removed, the pipetting body 121 only needs to move above the pipette tip groove 126 and then press down to put the pipette tip 122 on the pipette tip groove 126 into the connecting part 123, which is simple and convenient and improves work efficiency.

[0033] Furthermore, the connecting part 123 can automatically detach the pipette tip 122 without manual replacement. The replacement structure adopts existing technology, and for details, please refer to patent document CN113980804A, which will not be elaborated here.

[0034] Furthermore, the pipetting mechanism 120 is provided with a guide rail 124, the pipetting body 121 is movably connected to the guide rail 124, and the pipetting body 121 is connected to a pipetting cylinder 125, which can move up and down under the drive of the pipetting cylinder 125.

[0035] The centrifugation mechanism 130 is located on one side of the workbench 110. The centrifugation mechanism 130 includes a centrifuge 131 and a robotic arm 132. The robotic arm 132 is used to pick up test tubes from the test tube slot and place them onto the centrifuge 131. The robotic arm 132 is equipped with a miniature vacuum pump to create negative pressure at the picking end of the robotic arm 132, thereby picking up the test tubes. The centrifuge 131 is equipped with a fixing slot, which is used to restrict the test tubes placed on the centrifuge 131 so that the test tubes do not move during the centrifugation process.

[0036] The second shifting mechanism 170 is mounted on the frame 140 and is connected to the robotic arm 132. It is used to drive the robotic arm 132 to move, assist the robotic arm 132 in grasping the test tube, and place the test tube on the centrifuge 131.

[0037] The first shifting mechanism 160 is located at the bottom of the frame 140 and is connected to the worktable.

[0038] The first shifting mechanism 160 includes a first guide rail 161 disposed at the bottom of the frame 140 and a first slider 162 slidably connected to the first guide rail 161. A first driving mechanism is connected to the first slider 162. The first driving mechanism can be a cylinder, a hydraulic cylinder, a motor rack and pinion mechanism, etc., preferably a cylinder. The first driving mechanism is used to push the first slider 162 to move on the first guide rail 161. The worktable 110 is disposed on the slider 162. The first shifting mechanism 160 can drive the worktable 110 to move back and forth, which facilitates the pipetting mechanism 120 to aspirate and dispense liquid from different test tube tanks.

[0039] The second shifting mechanism 170 includes a second guide rail 171 mounted on the frame 140 and a second slider connected to the robotic arm 132. The robotic arm 132 is connected to a second drive mechanism, which can be a cylinder, a hydraulic cylinder, a motor gear rack mechanism, etc., preferably a motor gear rack mechanism. The second drive mechanism is used to drive the robotic arm 132 to move on the second guide rail 171.

[0040] The workbench 110 includes a first work surface 118 and a second work surface 119, with a gap 190 between them. A first test tube slot 111, a second test tube slot 112, a third test tube slot 113, and a fourth test tube slot 114 are located on the first work surface 118, while a fifth test tube slot 115, a sixth test tube slot 116, and a seventh test tube slot 117 are located on the second work surface 119. A magnet 191 is installed in the gap 190, and the fifth test tube slot 115 is located on the second work surface 119 near the gap 190. Magnetic beads are placed inside the purification test tubes, and the magnet 190 provides magnetic force for the reaction of the magnetic beads within the purification test tubes, thereby achieving rapid separation, washing, and enrichment of the magnetic beads.

[0041] The blood sample detection device in this application embodiment also includes a fluorescence scanning module, an optical imaging module, a multi-feature fusion module, an attention mechanism module, and an analysis and control module.

[0042] like Figure 5 As shown, Figure 5This is a schematic diagram of the hardware operating environment structure of a blood sample analysis method according to a first aspect embodiment of this application. The hardware operating environment includes a processor 210, a user interface 220, an image acquisition module 230, a memory 240, a network interface 250, and a communication line 260. The user interface 220, image acquisition module 230, memory 240, network interface 250, and processor 210 are connected via the communication line 260. The processor 210 may be a CPU, etc.; the user interface 220 may include a display screen, mouse, keyboard, data cable interface, etc.; the network interface 250 may include a data cable interface, Wi-Fi interface, Bluetooth interface, etc.; the image acquisition module 230 may include a fluorescence scanning module and an optical imaging module; and the memory 240 may include an operating system, a data interface control program, a network connection program, and an AI-based image analysis program.

[0043] The image acquisition module 230 includes a fluorescence scanning module and an optical imaging module.

[0044] A blood sample analysis method according to an embodiment of this application includes the following steps: Blood sample images are acquired by an image acquisition module installed on the blood sample testing equipment, and the blood sample images are preprocessed and standardized. A CTC sample database and an AI model are constructed. Preprocessed and standardized blood sample images are input into the trained AI model to perform AI recognition on the blood sample images and generate recognition results. The identification results are compared with the data in the CTC sample database to obtain the comparison results; Feature analysis is performed on the comparison results to obtain the detection results and generate a detection report.

[0045] In this embodiment of the application, the construction of the CTC sample database includes the following steps: Collect CTC images from clinical whole blood samples, input the CTC images into a CTC sample database, and perform preprocessing and standardization on the CTC images; The preprocessing steps are as follows: First, the CTC image is denoised to preserve CTC detail features and reduce interference; then, image enhancement is performed to highlight the CTC contour and fluorescence signal. The standardization process is as follows: the size and pixel values ​​of the CTC images input into the CTC sample database are adjusted to uniform values; Expand the CTC sample database to increase the dataset size; The CTC images are classified into multiple first sample sets based on different tumor types; each first sample set is further classified into multiple second sample sets based on different tumor morphologies; each second sample set is further classified into multiple third sample sets based on different clinical stages. Deep mining was performed on the CTC images in each third sample set to extract and record the morphological and molecular features of CTCs in the CTC images. The morphological features of CTCs include cell outline, cell size, nucleus shape, cytoplasmic texture, nucleus-to-cytoplasm ratio, and cell edge smoothness. The molecular features of CTCs include the expression intensity, distribution pattern, and spatial distribution density of specific proteins. The specific proteins include cytokeratin and epithelial cell adhesion molecules. The expression intensity of the specific proteins is mainly identified by detecting the fluorescence intensity of fluorescently labeled proteins.

[0046] Images of normal blood cells were collected and entered into the CTC sample database to establish a comparison sample set.

[0047] In this embodiment of the application, the detection results and the blood sample image are synchronized to the CTC sample database.

[0048] In this embodiment of the application, the AI ​​recognition includes the following steps: By using an attention mechanism module installed on the blood sample testing device, the system automatically focuses on the key distinguishing regions of CTCs, reducing interference from irrelevant features; By using a multi-feature fusion module installed on the blood sample detection device, the morphological and molecular features of CTCs in the blood sample image are extracted simultaneously. The morphological features include the cell outline, the ratio of cell nucleus to cytoplasm, and the edge smoothness of CTCs. The molecular features include the fluorescence intensity and spatial distribution density of fluorescently labeled proteins. The extracted morphological and molecular features of CTC were compared with data in the database.

[0049] In this embodiment of the application, the construction of the AI ​​model includes the following steps: We used a pre-trained ResNet-50 model as the basic architecture, froze the parameters of the first 10 network layers, and fine-tuned the subsequent network layers and the newly added multi-feature fusion module and attention mechanism module. The difference between the model's predicted values ​​and manually labeled values ​​is calculated using the cross-entropy loss function, and the learning rate is dynamically adjusted using the Adam optimizer. Mini-batch stochastic gradient descent is used for training, with 50 training epochs. An early stopping mechanism is introduced, dividing the dataset into training, validation, and test sets in a ratio of 7:2:1. Training stops when the validation set has 10 consecutive losses and the number of training epochs has not decreased, thus preventing model overfitting.

[0050] In this embodiment of the application, the construction of the AI ​​model further includes the following steps: By analyzing the model's prediction results on the validation set using a confusion matrix, and increasing the number of samples in the corresponding category for CTC samples with high misclassification, the model is retrained accordingly. Model pruning techniques are employed to remove redundant neurons and connections in the network, reducing the number of model parameters while maintaining model accuracy, thereby improving the model's inference speed on the device and ensuring real-time detection requirements. Finally, five-fold cross-validation is used to evaluate the model's performance and ensure its stability on different subsets of data.

[0051] Model pruning is one of the core methods of model compression. It optimizes model performance by removing redundant parameters or structures from neural networks. It is commonly used to improve inference speed, reduce storage usage, and lower energy consumption, while maintaining model accuracy as much as possible. In this embodiment, the multi-feature fusion module and attention mechanism module can extract and identify various morphological and molecular features of CTCs. At the same time, through AI algorithms, the device can deeply learn the complex features and morphology of CTCs, further improving detection sensitivity and specificity, reducing false positive and false negative results. Combined with a large amount of high-quality CTC image data, the AI ​​model is fully trained and optimized to enhance the model's recognition ability and provide more accurate and reliable technical support for CTC detection.

[0052] The method of using the blood sample testing device in this embodiment is as follows: Before starting the equipment, at least 4 mL of fresh blood / bone marrow sample is placed in sample tube 1110, and reaction tube 1120 is empty. The reagent tube contains the reagents required for the reaction. The reagents in the reagent tube can be 1X RBC, CTC buffer A, etc. The reagent tube can be set in multiple rows to hold different reagents. The purification reagent contains magnetic bead reagent, which can be magnetic beads.

[0053] After the device program is started, the pipetting mechanism will transfer the sample from the blood collection tube and the 1X RBCs from the reagent tube to the reaction tube for reaction. After the reaction, the device will move the reaction tube to the centrifuge for centrifugation. After centrifugation, the device will discard the supernatant after the reaction. The supernatant is the "red blood cell fluid after lysis". After the blood sample testing device repeats the above steps once, it will transfer CTC buffer A from the reagent tube to the reaction tube and resuspend and mix the sample.

[0054] After mixing, the sample is transferred to a purification tube by the pipetting mechanism 120 and reacted with magnetic beads. This reaction step purifies the sample by removing impurities through the specific adsorption of target cells by the magnetic beads. Two methods can be selected for this step: negative selection and positive selection. In negative selection, magnetic beads adsorb positive cells, and the supernatant is recovered to obtain the target cells. In positive selection, magnetic beads adsorb positive cells, and the magnetic bead solution is recovered to obtain the target cells. After the magnetic bead reaction, the pipetting mechanism 120 transfers the sample to a finished product tube.

[0055] After the equipment program finishes running, the finished tube is removed manually and made into a glass slide for further testing and analysis.

[0056] The steps for preparing a glass slide are as follows: 1. Centrifuge the finished product tubes, wash them with PBS solution, and then centrifuge again; 2. After centrifugation, add fixative to fix the cells, and then centrifuge again; 3. Repeat step ② once; 4. Resuspend the cells in an appropriate amount of fixative, drop them onto an adhesive glass slide, and dry them. 5. Re-fix the cells using aldehyde solutions; 6. Senervate the cells with 2X SSC solution; 7. Use gradient solutions of ethanol to perform gradient dehydration on cells; 8. Add the target chromosome probe and perform hybridization on an immunohybridizer; 9. After probe hybridization, wash the cells with washing solution to remove excess probes; 10. Incubate the cells with immunofluorescence antibodies; 11. Wash the cells with PBST to remove excess antibodies; 12. Add a blue fluorescent dye to the nucleus; 13. Cover with a cover slip and seal with adhesive to complete the preparation of the glass slide.

[0057] After the slide is prepared, it is placed in the blood sample testing equipment for CTC image processing and AI recognition to generate a test report.

[0058] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0059] The embodiments of this application have been described in detail above with reference to the accompanying drawings. However, this application is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of this application. Furthermore, unless otherwise specified, the embodiments and features described in the embodiments of this application can be combined with each other.

Claims

1. A blood sample analysis method, characterized in that, Includes the following steps: Blood sample images are acquired by an image acquisition module installed on the blood sample testing equipment, and the blood sample images are preprocessed and standardized. A CTC sample database and an AI model are constructed. Preprocessed and standardized blood sample images are input into the trained AI model to perform AI recognition on the blood sample images and generate recognition results. The identification results are compared with the data in the CTC sample database to obtain the comparison results; Feature analysis is performed on the comparison results to obtain the detection results and generate a detection report.

2. The blood sample analysis method according to claim 1, characterized in that, The construction of the CTC sample database includes the following steps: Collect CTC images from clinical whole blood samples, input the CTC images into a CTC sample database, and perform preprocessing and standardization on the CTC images; The preprocessing steps are as follows: First, the CTC image is denoised to preserve CTC detail features and reduce interference; then, image enhancement is performed to highlight the CTC contour and fluorescence signal. The standardization process is as follows: the size and pixel values ​​of the CTC images input into the CTC sample database are adjusted to uniform values; Expand the CTC sample database to increase the dataset size; The CTC images are classified into multiple first sample sets based on different tumor types; each first sample set is further classified into multiple second sample sets based on different tumor morphologies; each second sample set is further classified into multiple third sample sets based on different clinical stages. Images of normal blood cells were collected and entered into the CTC sample database to establish a comparison sample set.

3. The blood sample analysis method according to claim 2, characterized in that, The detection results and the blood sample image are synchronized to the CTC sample database.

4. The blood sample analysis method according to claim 1, characterized in that, The AI ​​recognition includes the following steps: By using an attention mechanism module installed on the blood sample testing device, the system automatically focuses on the key distinguishing regions of CTCs, reducing interference from irrelevant features; By using a multi-feature fusion module installed on the blood sample detection device, the morphological and molecular features of CTCs in the blood sample image are extracted simultaneously. The morphological features include the cell outline, the ratio of cell nucleus to cytoplasm, and the edge smoothness of CTCs. The molecular features include the fluorescence intensity and spatial distribution density of fluorescently labeled proteins. The extracted morphological and molecular features of CTC were compared with data in the database.

5. The blood sample analysis method according to claim 4, characterized in that, The construction of the AI ​​model includes the following steps: We used a pre-trained ResNet-50 model as the basic architecture, froze the parameters of the first 10 network layers, and fine-tuned the subsequent network layers and the newly added multi-feature fusion module and attention mechanism module. The difference between the model's predicted values ​​and manually labeled values ​​is calculated using the cross-entropy loss function, and the learning rate is dynamically adjusted using the Adam optimizer. Mini-batch stochastic gradient descent is used for training, with 50 training epochs. An early stopping mechanism is also introduced, which stops training when the validation set loss does not decrease for 10 consecutive training epochs to avoid model overfitting.

6. The blood sample analysis method according to claim 5, characterized in that, The construction of the AI ​​model also includes the following steps: By analyzing the model's prediction results on the validation set using a confusion matrix, and increasing the number of samples in the corresponding category for CTC samples with high misclassification, the model is retrained accordingly. Model pruning techniques are employed to remove redundant neurons and connections in the network, reducing the number of model parameters while maintaining model accuracy, thereby improving the model's inference speed on the device and ensuring real-time detection requirements. Finally, 5-fold cross-validation is used to evaluate the model's performance and ensure its stability on different subsets of data.

7. A blood sample testing device, characterized in that, For implementing the blood sample analysis method of claim 1, comprising: A workbench used for processing blood samples; The first test tube trough is used to place sample test tubes; The second test tube trough is used to hold reaction test tubes; The third test tube trough is used to hold reagent test tubes; The fourth test tube trough is used to hold recycled test tubes; The fifth test tube trough is used to hold purified test tubes; The sixth test tube trough is used to hold finished product test tubes; A pipetting mechanism, which is disposed above the workbench, is used to transfer liquids from test tube troughs on the workbench; A centrifugation mechanism is provided on one side of the workbench. The centrifugation mechanism includes a centrifuge and a robotic arm. The robotic arm is used to pick up test tubes from the test tube tank and place them onto the centrifuge. A first shifting mechanism is connected to the worktable; The second shifting mechanism is connected to the robotic arm.

8. The blood sample testing device according to claim 7, characterized in that, It also includes a fluorescence scanning module, an optical imaging module, a multi-feature fusion module, an attention mechanism module, and an analysis and control module.

9. The blood sample testing device according to claim 7, characterized in that, The workbench includes a first tabletop and a second tabletop, with a gap between the first tabletop and the second tabletop. The first test tube slot, the second test tube slot, the third test tube slot, and the fourth test tube slot are disposed on the first tabletop, and the fifth test tube slot and the sixth test tube slot are disposed on the second tabletop. A magnet is disposed in the gap, and the fifth test tube slot is disposed on one end of the second tabletop near the gap.

10. The blood sample testing device according to claim 7, characterized in that, The pipetting mechanism is detachably equipped with a pipette tip, and the worktable is provided with a pipette tip groove for placing the pipette tip.

Images