Blood layering identification method and sample storage equipment

Abstract

The invention relates to the technical field of blood layering identification, particularly provides a blood layering identification method and a sample storage device, and aims to solve the problem that a boundary identification result is unstable due to the fact that an existing method is extremely sensitive to external interference factors such as uneven environment illumination and light reflection of a test tube wall and cannot effectively cope with color errors of deterministic hierarchies. In order to achieve the purpose, the blood layering recognition method comprises the steps that an original image of a centrifuged to-be-recognized blood sample is obtained; preprocessing the original image to obtain an illumination invariance image, and obtaining boundary position coordinates of all blood layers of the blood sample to be identified according to the illumination invariance image; according to the method, the image is preprocessed at the front end, the influence caused by external interference factors is eliminated from the source, and the robustness is improved; high-dimensional feature analysis is fused at the middle end and region constraint is carried out at the rear end, so that external interference factors can be removed out of a search range, boundary identification is only carried out in a clean region, and the anti-interference capability is improved.

Description

Technical Field

[0001] This invention relates to the field of blood stratification identification technology, specifically providing a blood stratification identification method and sample storage device. Background Technology

[0002] Automated visual recognition of blood sample stratification is a key technology in in vitro diagnostic equipment and laboratory automation systems. Centrifuged blood samples have a complex multilayered structure, generally including a plasma layer (yellow), a PBMC layer (white membrane layer), a white layer (such as a separating gel or a specific cell layer), and a red blood cell layer (dark red).

[0003] Current methods for stratified blood sample identification mostly rely on scanning the grayscale or color thresholds of a global image directly from top to bottom. The PBMC layer (white membrane layer) has weak color characteristics and is extremely thin. In industrial environments, due to the difficulty in maintaining perfectly consistent lighting, strong reflected light and shadows from test tube walls and racks can disrupt the weak PBMC layer signal, making it unidentifiable. Furthermore, the PBMC layer, sandwiched between the relatively transparent plasma layer and the bright, glaring white layer, is easily "swallowed up" by the strong halo of the underlying white layer, or its boundaries are blurred due to the varying shades of the plasma above. In addition, the plasma layer (yellow layer) itself may exhibit color shifts due to individual differences or slight hemolysis, making it difficult for traditional identification methods to find its boundaries, leading to overall positioning failure. Summary of the Invention

[0004] The present invention aims to solve the above-mentioned technical problems, namely, to solve the problem that the existing methods are extremely sensitive to external interference factors such as uneven ambient lighting and test tube wall reflection, and cannot effectively deal with the color error of the deterministic level itself, resulting in unstable boundary line recognition results.

[0005] In a first aspect, the present invention provides a blood stratification identification method, the identification method comprising: acquiring an original image of a centrifuged blood sample to be identified; preprocessing the original image to obtain an illumination-invariant image; and obtaining the boundary coordinates of all blood strata of the blood sample to be identified based on the illumination-invariant image.

[0006] In a specific implementation of the above-mentioned blood layer recognition method, "preprocessing the original image to obtain an illumination-invariant image" includes: using an image decomposition model to decompose the original image into an illumination component image and an object intrinsic reflection component image; suppressing or removing the illumination component image, retaining the object intrinsic reflection component image, and using the object intrinsic reflection component image as an illumination-invariant image.

[0007] In a specific implementation of the above-described blood stratification identification method, the blood stratification includes easily identifiable blood layers and difficult-to-identify blood layers; "obtaining the boundary coordinates of all blood layers of the blood sample to be identified based on the illumination-invariant image" includes: performing adaptive anchor point identification on the illumination-invariant image to obtain the first and second coarse boundary coordinates of the easily identifiable blood layers; delineating a constraint search region based on the first and second coarse boundary coordinates; and performing local feature diagnosis on the constraint search region to determine multiple boundary coordinates; wherein the multiple boundary coordinates include: the boundary coordinates of the difficult-to-identify blood layers; and / or the boundary coordinates between the easily identifiable blood layers and the difficult-to-identify blood layers.

[0008] In a specific implementation of the above-mentioned blood layer recognition method, "performing adaptive anchor point recognition on the illumination-invariant image to obtain the first and second coarse boundary coordinates of the easily identifiable blood layer" includes: extracting features from the illumination-invariant image to obtain a fused feature set; wherein, the fused feature set includes the color features and vertical spatial position information of each pixel in the illumination-invariant image; and performing adaptive anchor point recognition on the fused feature set to obtain the first and second coarse boundary coordinates of the easily identifiable blood layer.

[0009] In a specific implementation of the above-mentioned blood layer recognition method, "performing local feature diagnosis on the constrained search region to determine multiple boundary position coordinates" includes: sequentially performing local feature enhancement and vertical gradient analysis on the constrained search region to identify multiple signal abrupt change points in all blood layers within the constrained search region; associating and verifying the hierarchical topological order of the multiple signal abrupt change points to determine multiple image jump point sequences; and determining multiple boundary position coordinates based on the position coordinates of the multiple image jump point sequences.

[0010] In a specific embodiment of the above-described blood layering identification method, the easily identifiable layer includes a plasma layer and a red blood cell layer, and the difficult-to-identify blood layer includes a PBMC layer and a white blood cell layer. The plasma layer, the PBMC layer, the white blood cell layer, and the red blood cell layer are arranged sequentially from top to bottom. The first coarse boundary coordinate is the coarse boundary coordinate between the plasma layer and the PBMC layer, and the second coarse boundary coordinate is the coarse boundary coordinate between the white blood cell layer and the red blood cell layer. "Determining multiple boundary coordinates based on the position coordinates of multiple image transition point sequences" includes: the position coordinate of the first transition point is the actual boundary coordinate between the plasma layer and the PBMC; the position coordinate of the second transition point is the actual boundary coordinate between the PBMC and the white blood cell layer; and the position coordinate of the third transition point is the actual boundary coordinate between the white blood cell layer and the red blood cell layer.

[0011] In a specific implementation of the above-mentioned blood layer identification method, "obtaining the boundary coordinates of all blood layers of the blood sample to be identified based on the illumination-invariant image" further includes: performing adaptive anchor point identification on the illumination-invariant image to obtain the first coarse boundary coordinates, the second coarse boundary coordinates, and the third actual boundary coordinates of the easily identifiable blood layers.

[0012] In a specific embodiment of the above-described blood layering identification method, the easily identifiable layer includes a plasma layer, a red blood cell layer, and a separating gel layer, and the difficult-to-identify blood layer includes a PBMC layer. The plasma layer, the PBMC, the separating gel layer, and the red blood cell layer are arranged sequentially from top to bottom. The first coarse boundary coordinate is the coarse boundary coordinate between the plasma layer and the PBMC layer, the second coarse boundary coordinate is the coarse boundary coordinate between the PBMC layer and the separating gel layer, and the third actual boundary coordinate is the actual boundary coordinate between the separating gel layer and the red blood cell layer. "Determining multiple boundary coordinates based on the position coordinates of multiple image jump point sequences" includes: the position coordinate of the first jump point is the actual boundary coordinate between the plasma layer and the PBMC; the position coordinate of the second jump point is the actual boundary coordinate between the PBMC and the separating gel layer.

[0013] In a specific implementation of the above-mentioned blood layer recognition method, the control method further includes: after obtaining multiple boundary position coordinates, performing sequential verification and thickness verification on the multiple boundary position coordinates.

[0014] In a second aspect, the present invention also provides a sample storage device, the sample storage device including a controller configured to perform the blood stratification identification method described above.

[0015] When using the above technical solution, the blood layer identification method of the present invention acquires the original image of the centrifuged blood sample to be identified; preprocesses the original image to obtain an illumination-invariant image, and obtains the boundary coordinates of all blood layers of the blood sample to be identified based on the illumination-invariant image. By preprocessing the original image at the front end, the present invention can eliminate the influence of tube wall reflection, shadows, and uneven ambient light from the source, maximizing the preservation of the inherent color and contrast of each blood layer and improving robustness; by fusing high-dimensional feature analysis in the middle and performing region constraints at the back end, noise generated by tube wall reflection, top and bottom, etc., can be removed from the search range, and boundary line identification is performed only in clean areas, improving anti-interference ability. Attached Figure Description

[0016] The preferred embodiments of the present invention are described below with reference to the accompanying drawings, in which: Figure 1 This is a flowchart of the main steps of the blood layer identification method of the present invention; Figure 2 This is a detailed flowchart of the steps of Embodiment 1 of the blood layer identification method of the present invention; Figure 3 This is a distribution diagram of the blood layer in Embodiment 1 of the present invention; Figure 4 This is a distribution diagram of the blood layer in Embodiment 2 of the present invention; The layers are: 1. Plasma layer; 2. PBMC layer; 3. White layer; 4. Red blood cell layer; 5. Separating gel layer. Detailed Implementation

[0017] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art can make adjustments as needed to adapt to specific application scenarios.

[0018] It should be noted that although the steps of the control method of the present invention are described in a specific order in this application, this order is not restrictive. Those skilled in the art can perform the steps in different orders without departing from the basic principles of the present invention. Furthermore, ordinal numbers such as "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0019] In medical testing, blood undergoes centrifugation and will physically separate into layers based on specific gravity. This invention is a control method for identifying the blood layering in centrifuged blood samples.

[0020] like Figure 1As shown, the identification method of the present invention includes: S1, Obtain the original image of the centrifuged blood sample to be identified.

[0021] In some embodiments, the acquisition of the original image is performed using a visual recognition device, such as an industrial camera.

[0022] S2, preprocess the original image to obtain an illumination-invariant image.

[0023] S3, based on the illumination invariance image, obtain the coordinates of the boundary positions of all blood layers in the blood sample to be identified.

[0024] In step S2, "preprocessing the original image to obtain an illumination-invariant image" includes: S21, using an image decomposition model, the original image is decomposed into an illumination component image and an object intrinsic reflection component image; S22, suppress or remove the illumination component image, retain the object's intrinsic reflection component image, and use the object's intrinsic reflection component image as the illumination invariant image.

[0025] In some embodiments, the image decomposition model is, for example, based on the Retinex model, S (original image) = R (image of the object's intrinsic reflectance component) × L (image of the illumination component), where the image of the object's intrinsic reflectance component is the true color image of the blood, and the image of the illumination component contains reflections and shadows. L is estimated using an algorithm, removed from S, and R is ultimately retained or reconstructed. This step, through data source purification and illumination equalization preprocessing of the original image, can eliminate the influence of reflections from the test tube wall, shadows, and uneven ambient light at the source, maximizing the preservation of the intrinsic color and contrast of each layer of blood, resulting in an illumination-invariant image.

[0026] In step S3, "obtaining the boundary coordinates of all blood strata in the blood sample to be identified based on the illumination invariance image" includes: S31, Adaptive anchor point recognition is performed on the illumination-invariant image to obtain the coordinates of the first and second coarse boundary positions of the blood layer, which are easily determined. S32, Determine the constraint search area based on the coordinates of the first coarse boundary position and the coordinates of the second coarse boundary position; S33, perform local feature diagnosis on the constrained search region to determine the coordinates of multiple boundary positions; wherein, the coordinates of multiple boundary positions include: It is difficult to determine the coordinates of the boundary of the blood layer; The coordinates of the boundary between easily identifiable and difficult-to-identify blood layers.

[0027] In some embodiments, the blood layering includes an easily identifiable blood layer and a difficult-to-identify blood layer. Specifically, to ensure the delineation of the constrained search region, the easily identifiable blood layer comprises at least two blood layers with stable color features, and the corresponding coarse boundary coordinates include a first coarse boundary coordinate and a second coarse boundary coordinate. Obtaining the coarse boundary coordinates of the easily identifiable blood layer serves to coarsely locate the area covered by its boundary, thereby defining the constrained search region for the difficult-to-identify blood layer. Then, the boundary lines within the constrained search region are precisely located to determine the boundary coordinates of the difficult-to-identify blood layer and / or the boundary coordinates between the easily identifiable and difficult-to-identify blood layers.

[0028] Further, in step S31, "adaptive anchor point identification is performed on the illumination-invariant image to obtain the coordinates of the first and second coarse boundary positions of the blood layer, which are easily determined" includes: S311, extract features from the illumination-invariant image to obtain a fused feature set; wherein, the fused feature set includes the color features and vertical spatial location information of each pixel in the illumination-invariant image; S312, Adaptive anchor point identification is performed on the fused feature set to obtain the coordinates of the first and second coarse boundary positions of the blood layer, which are easily determined. S313, adaptive anchor point recognition is performed on the fused feature set to obtain the coordinates of the first coarse boundary position, the second coarse boundary position, and the third actual boundary position of the blood layer, which are easily determined.

[0029] In some embodiments, the purpose of this step is to establish structural anchor points with color error tolerance on a clean data source. Specifically, an illumination-invariant image is input into a feature extraction module, and the color features and vertical spatial location information of each pixel are extracted by an embedded processor or a host computer system to form a fused feature set. This feature set integrates high-dimensional color features and the vertical spatial location information of the pixel (i.e., 4D or higher-dimensional features). Based on the above fused feature set, the system performs high-dimensional analysis to identify easily identifiable blood layers and the coordinates of their boundaries. Because the input image has been purified and vertical spatial location information has been fused, this step can adaptively cope with color errors (such as mild hemolysis) caused by biological factors in easily identifiable blood layers, greatly improving the anchor points' resistance to color errors.

[0030] Furthermore, steps S312 and S313 are parallel, addressing two different blood stratification scenarios. In step S312, the easily identifiable blood layers are plasma layer 1 and erythrocyte layer 4. The boundary coordinates are obtained as the coarse boundary coordinates corresponding to the boundary line between plasma layer 1 and erythrocyte layer 4 and the difficult-to-determine blood layer, namely, the first and second coarse boundary coordinates. In step S313, when using blood tubes containing thixotropic separating gel, after centrifugation, the separating gel forms a significantly thick, grayish-white, and fixed barrier layer. At this point, the easily identifiable layer includes plasma layer 1, erythrocyte layer 4, and separating gel layer 5. In addition to the coarse boundary coordinates corresponding to the boundary line between plasma layer 1 and separating gel layer 5 and the difficult-to-determine blood layer, namely the first and second coarse boundary coordinates, the actual boundary coordinates corresponding to the boundary line between separating gel layer 5 and erythrocyte layer 4 are also obtained, namely, the third actual boundary coordinates.

[0031] Furthermore, in step S33, "performing local feature diagnosis on the constrained search region to determine the coordinates of multiple boundary positions" includes: S331 performs local feature enhancement and longitudinal gradient analysis on the constrained search region to identify multiple signal abrupt change points in all blood layers within the constrained search region.

[0032] In some embodiments, local histogram equalization or contrast stretching methods are used to enhance local features in the constraint search region. For example, for a difficult-to-determine blood layer whose constraint search region includes PBMC layer 2 and white layer 3, given the characteristics of PBMC layer 2 being extremely thin and dark, and the underlying white layer 3 being extremely bright, a local enhancement algorithm is used to forcibly increase the grayscale dynamic range between PBMC layer 2 (dark signal) and white layer 3 (strong signal), constructing a feature space suitable for high-sensitivity analysis. As another example, for blood containing separating gel layer 5, the color difference between PBMC layer 2 (usually milky white or light gray) and separating gel layer 5 (usually white or off-white) is extremely small, belonging to a "low contrast" region. Forcibly stretching the grayscale distribution within the constraint search region distinguishes the "grayscale signal" of PBMC layer 2 from the "bright signal" of separating gel layer 5, improving recognition accuracy.

[0033] Vertical gradient analysis, also known as vertical signal scanning and gradient calculation, extracts the grayscale / color projection curve of the constrained search region along the vertical direction (Y-axis) and calculates the first derivative (gradient) of this curve. Technical principle: Instead of relying on the absolute brightness value of a single pixel, it analyzes the rate of change of brightness along the Y-axis to identify abrupt changes in the signal.

[0034] The above steps, by first enhancing local features and then performing gradient analysis, can improve the accuracy of blood identification.

[0035] S332, perform hierarchical topological order association and verification on multiple signal abrupt change points to determine multiple image jump point sequences.

[0036] In some embodiments, by associating and verifying multiple signal mutation points in a hierarchical topological order, noise can be eliminated, persistent and coherent signal mutation points can be retained, and an ordered and coherent sequence of image jump points can be output, i.e. the boundary lines between adjacent blood layers, thereby achieving precise positioning of the boundary lines between multiple blood layers.

[0037] S333, determine multiple boundary position coordinates based on the position coordinates of multiple image jump point sequences.

[0038] The identification method of the invention will be described in detail below with reference to specific embodiments.

[0039] Example 1 like Figure 3 As shown, the blood layers, from top to bottom, include plasma layer 1, PBMC layer 2, white blood cell layer 3, and red blood cell layer 4. The easily identifiable layers include plasma layer 1 and red blood cell layer 4, while the difficult-to-identify blood layers include PBMC layer 2 and white blood cell layer 3.

[0040] In steps S312, S32, and S33, the first coarse boundary coordinates are the coarse boundary coordinates between plasma layer 1 and PBMC layer 2, and the second coarse boundary coordinates are the coarse boundary coordinates between white layer 3 and red blood cell layer 4. The constrained search area includes PBMC layer 2, white layer 3, the boundary line between PBMC layer 2 and white layer 3, the boundary line between plasma layer 1 and PBMC layer 2, and the boundary line between white layer 3 and red blood cell layer 4. Among them, the boundary coordinates of the difficult-to-determine blood layer are the position coordinates of the boundary line between PBMC layer 2 and white layer 3, and the boundary coordinates between the easily determined blood layer and the difficult-to-determine blood layer are the position coordinates of the boundary line between plasma layer 1 and PBMC layer 2 and the boundary line between white layer 3 and red blood cell layer 4.

[0041] In step S332, the boundary localization based on the topological sequence is based on the corrected physical sequence (up → yellow → PBMC → white → red → down). The system searches for a sequence of abrupt change points on the gradient map that conform to the topological structure of "stable-dark valley-bright peak-deep valley", thereby accurately locating the boundaries of each layer.

[0042] In step S333, "determining multiple boundary position coordinates based on the position coordinates of multiple transition points" includes: The coordinates of the first jump point are the actual boundary coordinates between plasma layer 1 and PBMC; The coordinates of the second jump point are the actual boundary coordinates between PBMC and white layer 3. The coordinates of the third jump point are the actual boundary coordinates between the white layer 3 and the red cell layer 4.

[0043] In this embodiment, the first boundary line (plasma layer 1 / PBMC layer 2) is used to locate the "first downward transition point". The coordinates of the position corresponding to the first boundary line are the coordinates of the first transition point. Feature description: The signal undergoes a significant negative gradient abrupt change from the relatively stable region of the upper plasma layer 1, entering the low grayscale region of the PBMC layer 2.

[0044] Second boundary line (PBMC layer 2 / white layer 3): Locating the "abrupt upward transition point". The coordinates of the location corresponding to the second boundary line are the coordinates of the second transition point. Feature description: The signal undergoes a large positive gradient abrupt change from the "dark valley" of PBMC layer 2, rising sharply to the bright peak of white layer 3. This gradient peak corresponds to the lower boundary of PBMC layer 2.

[0045] The third dividing line (white layer 3 / red cell layer 4): locates the "second downward jump point". The coordinates of the position corresponding to the third dividing line are the coordinates of the third jump point. Feature description: The signal undergoes another negative abrupt change from the bright plateau of white layer 3, and drops to the dark area of ​​the bottom red cell layer 4.

[0046] In the above steps, the first dividing line is identified to determine the starting position of PBMC layer 2, the second dividing line is identified to distinguish the target cell layer and the interference layer (i.e., PBMC layer 2 and white layer 3) and the thickness of PBMC layer 2, and the third dividing line is identified to determine the ending position of white layer 3. At the same time, it is also to verify whether the overall layer sequence logic is correct (to prevent red blood cell layer 4 from being misjudged as part of white layer 3), thus achieving fine segmentation of two layers.

[0047] Example 2 like Figure 4 As shown, the blood layer comprises, from top to bottom, a plasma layer 1, PBMCs, a separating gel layer 5, and a red blood cell layer 4. The easily identifiable layers include plasma layer 1, red blood cell layer 4, and separating gel layer 5, while the difficult-to-identify blood layer includes PBMC layer 2. In this embodiment, because when using blood tubes containing thixotropic separating gel, after centrifugation, the separating gel forms a significantly thick, grayish-white, and fixed-position barrier layer, the separating gel layer 5 has a very strong color characteristic and is very easy to identify.

[0048] In steps S312, S32, and S33, the first coarse boundary coordinates are the coarse boundary coordinates between plasma layer 1 and PBMC layer 2, the second coarse boundary coordinates are the coarse boundary coordinates between PBMC layer 2 and separating gel layer 5, and the third actual boundary coordinates are the actual boundary coordinates between separating gel layer 5 and red blood cell layer 4. The constrained search area includes PBMC layer 2, the boundary line between plasma layer 1 and PBMC layer 2, and the boundary line between PBMC layer 2 and separating gel layer 5. Since the difficult-to-determine layer only includes PBMC layer 2, there is correspondingly only the definition of the boundary coordinates between the easily determined blood layer and the difficult-to-determine blood layer, that is, the position coordinates of the boundary line between plasma layer 1 and PBMC layer 2 and the position coordinates of the boundary line between PBMC layer 2 and separating gel layer 5.

[0049] In step S333, "determining multiple boundary position coordinates based on the position coordinates of multiple transition points" includes: The coordinates of the first jump point are the actual boundary coordinates between plasma layer 1 and PBMC; The coordinates of the second jump point are the actual boundary coordinates between PBMC and the separating adhesive layer 5.

[0050] In this embodiment, the system identifies three "strong feature layers," with the separating gel layer 5 serving as a powerful intermediate anchor point, further subdividing the original large search area. Searching for the PBMC layer 2 only requires locating it in the region between the plasma layer 1 and the separating gel layer 5, which is more precise and less prone to interference than searching between just two layers (plasma and red blood cells). Furthermore, since the easily identifiable layers include three blood layers, the coordinates of the third actual boundary position identified between two easily identifiable layers are the actual boundary position coordinates between those two layers.

[0051] The control method of the present invention further includes: After obtaining the coordinates of multiple boundary positions, sequential verification and thickness verification are performed on the multiple boundary position coordinates.

[0052] In some embodiments, the purpose of the verification is to identify whether all boundary position coordinates conform to the new physical layer sequence. Specifically, it is assumed that the origin of the coordinate system is at the upper left corner of the image, where the Y-axis direction is vertically downward as the positive direction. In the sequence verification, all boundary position coordinates must satisfy: Y1 < Y2 < Y3, that is, ensuring that plasma layer 1 is above PBMC layer 2, and PBMC layer 2 is above white layer 3. Here, Y1 is the Y value of the position coordinate of the first transition point, Y2 is the Y value of the position coordinate of the second transition point, and Y3 is the Y value of the position coordinate of the third transition point. In the thickness verification, the thickness of PBMC layer 2 and the thickness of white layer 3 are calculated respectively to see if they are within a preset range value. If they are within the preset range, the logical verification is successful.

[0053] Based on the various embodiments described above, the present invention also provides a sample storage device, the sample storage device including a controller, the controller being configured to execute the blood stratification identification method described above.

[0054] like Figure 2 As shown, the detailed steps of Embodiment 1 of the blood layer identification method of the present invention are as follows: S101, Obtain the original image of the centrifuged blood sample to be identified; S102, preprocess the original image to obtain an illumination-invariant image; S103, extract features from the illumination-invariant image to obtain a fused feature set; S104, Adaptive anchor point identification is performed on the fused feature set to obtain the coordinates of the first and second coarse boundary positions of the blood layer, which are easily determined. S105, local feature enhancement and longitudinal gradient analysis are performed on the constrained search region to identify multiple signal abrupt change points in all blood layers within the constrained search region; S106, perform hierarchical topological order association and verification on multiple signal abrupt change points to determine multiple image jump point sequences; S107, Determine multiple boundary position coordinates based on the position coordinates of multiple image jump point sequences; S108 performs sequential verification and thickness verification on the coordinates of multiple boundary positions.

[0055] Based on the various embodiments described above, the blood layer identification method of the present invention acquires the original image of the centrifuged blood sample to be identified; preprocesses the original image to obtain an illumination-invariant image, and obtains the boundary coordinates of all blood layers in the blood sample to be identified based on the illumination-invariant image. By preprocessing the original image at the front end, the present invention can eliminate the influence of tube wall reflections, shadows, and uneven ambient light from the source, maximizing the preservation of the inherent color and contrast of each blood layer and improving robustness. By fusing high-dimensional feature analysis in the middle and performing region constraints at the back end, noise generated by tube wall reflections, top and bottom, etc., can be removed from the search range, allowing boundary line identification only in clean areas, thus improving anti-interference capability.

[0056] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for blood stratification identification, characterized in that, The identification method includes: Obtain the original image of the blood sample to be identified after centrifugation; The original image is preprocessed to obtain an illumination-invariant image; The coordinates of the boundary positions of all blood layers in the blood sample to be identified are obtained from the illumination invariant image.

2. The blood layering identification method according to claim 1, characterized in that, "Preprocessing the original image to obtain an illumination-invariant image" includes: Using an image decomposition model, the original image is decomposed into an illumination component image and an object intrinsic reflection component image; Suppress or remove the illumination component image, retain the object's intrinsic reflection component image, and use the object's intrinsic reflection component image as an illumination-invariant image.

3. The blood stratification identification method according to claim 1, characterized in that, The blood stratification includes easily identifiable blood layers and difficult-to-identify blood layers; "Obtaining the boundary coordinates of all blood strata in the blood sample to be identified based on the illumination-invariant image" includes: Adaptive anchor point identification is performed on the illumination-invariant image to obtain the first and second coarse boundary coordinates of the easily identifiable blood layer. The constraint search region is defined based on the coordinates of the first coarse boundary position and the coordinates of the second coarse boundary position. Local feature diagnosis is performed on the constrained search region to determine multiple boundary coordinates; wherein the multiple boundary coordinates include: The coordinates of the difficult-to-determine boundary position of the blood layer; and / or The coordinates of the boundary between the easily identifiable blood layer and the difficult-to-identify blood layer.

4. The blood layering identification method according to claim 3, characterized in that, "Performing adaptive anchor point identification on the illumination-invariant image to obtain the first and second coarse boundary coordinates of the easily determinable blood layer" includes: Feature extraction is performed on the illumination-invariant image to obtain a fused feature set; wherein, the fused feature set includes the color features and vertical spatial position information of each pixel in the illumination-invariant image; Adaptive anchor point identification is performed on the fused feature set to obtain the coordinates of the first and second coarse boundary positions of the easily identifiable blood layer.

5. The blood stratification identification method according to claim 4, characterized in that, "Performing local feature diagnosis on the constrained search region to determine the coordinates of multiple boundary locations" includes: Local feature enhancement and longitudinal gradient analysis are performed sequentially on the constrained search region to identify multiple signal abrupt change points in all blood layers within the constrained search region. The hierarchical topological order of multiple signal abrupt change points is correlated and verified to determine multiple image jump point sequences; Based on the position coordinates of the multiple image transition point sequences, the coordinates of the multiple boundary positions are determined.

6. The blood layering identification method according to claim 5, characterized in that, The easily identifiable layer includes a plasma layer (1) and a red blood cell layer (4), and the difficult-to-identify blood layer includes a PBMC layer (2) and a white layer (3). The plasma layer (1), the PBMC layer (2), the white layer (3) and the red blood cell layer (4) are arranged sequentially from top to bottom. The first coarse boundary coordinate is the coarse boundary coordinate between the plasma layer (1) and the PBMC layer (2), and the second coarse boundary coordinate is the coarse boundary coordinate between the white layer (3) and the red blood cell layer (4). "Determining multiple boundary position coordinates based on the position coordinates of multiple image transition point sequences" includes: The coordinates of the first jump point are the actual boundary coordinates between the plasma layer (1) and the PBMC; The coordinates of the second jump point are the actual boundary coordinates between the PBMC and the white layer (3); The coordinates of the third jump point are the actual boundary coordinates between the white layer (3) and the red cell layer (4).

7. The blood stratification identification method according to claim 3, characterized in that, "Obtaining the boundary coordinates of all blood strata in the blood sample to be identified based on the illumination-invariant image" also includes: Adaptive anchor point identification is performed on the illumination-invariant image to obtain the first coarse boundary coordinates, the second coarse boundary coordinates, and the third actual boundary coordinates of the easily identifiable blood layer.

8. The blood layering identification method according to claim 7, characterized in that, The easily identifiable layer includes a plasma layer (1), a red blood cell layer (4), and a separating gel layer (5), and the difficult-to-identify blood layer includes a PBMC layer (2). The plasma layer (1), the PBMC, the separating gel layer (5), and the red blood cell layer (4) are arranged sequentially from top to bottom. The first coarse boundary coordinate is the coarse boundary coordinate between the plasma layer (1) and the PBMC layer (2), the second coarse boundary coordinate is the coarse boundary coordinate between the PBMC layer (2) and the separating gel layer (5), and the third actual boundary coordinate is the actual boundary coordinate between the separating gel layer (5) and the red blood cell layer (4). "Determining multiple boundary position coordinates based on the position coordinates of multiple image transition point sequences" includes: The coordinates of the first jump point are the actual boundary coordinates between the plasma layer (1) and the PBMC; The coordinates of the second jump point are the actual boundary coordinates between the PBMC and the separating adhesive layer (5).

9. The blood stratification identification method according to claim 1, characterized in that, The control method further includes: After obtaining multiple boundary position coordinates, sequential verification and thickness verification are performed on the multiple boundary position coordinates.

10. A sample storage device, characterized in that: The sample storage device includes a controller configured to perform the blood stratification identification method according to any one of claims 1 to 9.

Images