Gel preparation centrifugal device and PRP quality optimization method based on same

Abstract

The invention discloses a gel preparation centrifugal device and a PRP quality optimization method based on the same. The method comprises the following steps: acquiring hematocrit data, platelet concentration data and blood viscosity data, and normalizing; performing single-index feature mapping on the normalized data to obtain hematocrit features, blood viscosity features and platelet concentration features, and extracting blood physicochemical features; extracting cross-domain residual features according to the physical and chemical features of the blood, and then extracting centrifugation-physical and chemical coupling features; according to the centrifugation-physicochemical coupling features, the calibrated coupling features are extracted, and then centrifugation parameter features are extracted; extracting a centrifugal parameter weight matrix according to the physical and chemical characteristics of the blood and the centrifugal parameter characteristics, and then calculating an optimal centrifugal parameter combination; and the optimal centrifugal parameter combination is sent to a centrifugal device, centrifugal operation is carried out, and parameter calibration and real-time adjustment are carried out on the rotating speed, time and temperature. The centrifugal parameters are individually adjusted based on the physical and chemical properties of the blood so as to improve the PRP extraction quality.

Description

Technical Field

[0001] This invention relates to the field of blood gel preparation technology, and in particular to a gel preparation centrifugation device and a PRP quality optimization method based on the device. Background Technology

[0002] In the fields of medical aesthetics and regenerative medicine, it is usually necessary to separate and purify various components in the blood. The quality of the operation will directly affect the purity and bioactivity of the product, and thus affect its skin revitalization effect and tissue repair safety in the subsequent autologous injection filling process.

[0003] For platelet-rich plasma (PRP), the current centrifugation purification process suffers from insufficient adaptability of preset fixed centrifugation parameters. Specifically, there are significant individual differences in the physicochemical properties of blood, such as viscosity and hematocrit, among different blood donors. Constant operating parameters such as centrifugation speed, reaction time, and temperature control precision cannot match the dynamic characteristics of different blood samples. The purification process lacks flexibility and makes it difficult to achieve targeted purification of PRP. Ultimately, this results in the platelet concentration and growth factor bioactivity in the PRP enrichment layer failing to meet clinically ideal standards.

[0004] Existing solutions to this problem mainly rely on manual experience adjustments, but they still have the following drawbacks: manual experience adjustments are easily affected by the operator's professional level and subjective judgment, making it difficult to guarantee the accuracy and consistency of parameter adjustments, and making it impossible to stably output high-quality PRP enrichment layers; manual experience adjustments lack the step of establishing a correlation model between centrifugation parameters and PRP enrichment layer quality, making it difficult to efficiently process multi-dimensional physicochemical index data of blood samples, resulting in large errors in the separation and purification process.

[0005] Therefore, the industry needs a technical solution that can improve the output quality of PRP enrichment layers to address the differences in the physicochemical properties of different blood samples. Summary of the Invention

[0006] In view of this, the present invention aims to provide a gel preparation centrifugation device and a PRP quality optimization method based on the device, so as to solve the problem that the purification effect of PRP enrichment layer is not good due to the difficulty of personalizing and adapting to the differences in the physicochemical parameters of patients' blood in traditional methods.

[0007] A gel preparation centrifuge apparatus includes a main unit housing, a front unit housing fixedly mounted at the front of the main unit housing, a side unit housing fixedly mounted on the side of the main unit housing, a cover hinged to the top of the main unit housing, an outer shell fixedly mounted on the inner wall of the main unit housing, an inner shell fixedly mounted inside the outer shell, a cooling cavity formed between the outer shell and the inner shell, a sealing ring provided on the top of the inner shell, an angle rotor and a horizontal rotor disposed inside the inner shell, a hanging cup detachably mounted on the horizontal rotor, and the bottom of the angle rotor detachably mounted to the output end of a servo motor. The servo motor is fixedly mounted to the inner bottom wall of the main unit housing via a tripod, shock-absorbing pads, support legs, and bolts. The apparatus also includes a speed drive unit for controlling the centrifugation speed, a timing unit for controlling the centrifugation time, a temperature control unit for controlling the centrifugation temperature, and a parameter control module for calibrating and adjusting the above parameters in real time.

[0008] Furthermore, the bottom of the outer casing is connected to the main unit via triangular legs, the bottom hinge of the casing cover has an electric spring, the electric spring is movably connected to the inner bottom wall of the main unit via a mounting base, an ultraviolet lamp is embedded in the bottom wall of the casing cover, and an observation window is embedded in the surface of the casing cover.

[0009] Furthermore, a drive circuit board is embedded in the inner wall of the front chassis, a touch screen is embedded in the surface of the front chassis, an electronic lock is fixedly installed on the inner top wall of the front chassis, the latch part of the electronic lock is installed on the bottom wall of the cover, and the body part of the electronic lock is installed on the inner top wall of the front chassis.

[0010] Furthermore, a water collection box is slidably installed on the front side of the side unit, a heating plate is fixedly installed on the inner bottom wall of the side unit, heat insulation plates are provided on both sides of the heating plate, a test tube slot is opened inside the heating plate, a test tube holder is fixedly installed on the top of the heating plate, a partition plate is fixedly installed on the inner top wall of the side unit and on the side of the heat insulation plate, a test tube holder is embedded on the surface of the side unit and in the interval of the partition plate, and a fan is fixedly installed on the side wall of the side unit. A condenser and a compressor are fixedly installed at the rear of the side casing. The condenser and the compressor are connected by a pipe. The other end of the condenser is connected to a solenoid valve, and the other end of the solenoid valve is connected to the cooling chamber.

[0011] A method for optimizing the quality of PRP based on the above-mentioned gel preparation centrifugation device includes the following steps: A1: Collect hematocrit data, platelet concentration data, and blood viscosity data, and normalize them respectively to obtain normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data. A2: Based on the normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data, single-index feature mapping was performed to obtain hematocrit characteristics, blood viscosity characteristics, and platelet concentration characteristics. A3: Extract the physicochemical characteristics of blood based on hematocrit characteristics, blood viscosity characteristics, and platelet concentration characteristics; A4: Based on the physicochemical characteristics of blood, extract cross-domain residual features, and based on the cross-domain residual features, extract centrifugation-physicochemical coupling features; A5: Based on the centrifugation-physicochemical coupling characteristics, extract the calibrated coupling characteristics, and based on the calibrated coupling characteristics, extract the centrifugation parameter characteristics; A6: Based on the physicochemical characteristics of blood and the characteristics of centrifugation parameters, extract the centrifugation parameter weight matrix, and calculate the optimal combination of centrifugation parameters based on the centrifugation parameter weight matrix; A7: Send the optimal combination of centrifugation parameters to the parameter control module of the gel preparation centrifuge device, start the centrifugation operation, and perform parameter calibration and real-time adjustment on the speed drive unit, timing unit, and temperature control unit; when the centrifugation speed, centrifugation time, and centrifugation temperature all reach the preset standards, stop the centrifugation operation and collect the PRP enrichment layer.

[0012] Furthermore, step A1 includes: A11: Collect hematocrit data and platelet concentration data of blood samples using a fully automated blood analyzer; A12: Blood viscosity data of blood samples were collected using a rotational viscometer, including low shear rate viscosity values, medium shear rate viscosity values, and high shear rate viscosity values. A13: The hematocrit data, blood viscosity data, and platelet concentration data were processed using the linear normalization method to obtain normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data, respectively.

[0013] Furthermore, step A3 includes: A31: Based on hematocrit characteristics, blood viscosity characteristics, and platelet concentration characteristics, gate screening characteristics are calculated through an association weight gating mechanism; A32: Based on the gating screening features, feature constraint enhancement is performed through the blood clinical constraint function to obtain constraint-enhanced features; A33: Based on the constraint enhancement features, the blood physicochemical features are obtained by combining global average pooling with a fully connected layer for feature dimensionality reduction and integration.

[0014] It should be further explained that the three core indicators in the blood physicochemical feature extraction process—hematocrit data, blood viscosity data, and platelet concentration data—have differentiated characteristics. Hematocrit data dominates the stratification boundary between red blood cells and plasma during centrifugation and exhibits strong feature stability. Blood viscosity data includes multiple values ​​at low, medium, and high shear rates, exhibiting large feature fluctuations and a non-linear correlation with centrifugation sedimentation rate. Platelet concentration data serves as the benchmark for PRP enrichment rate, showing a positive correlation with purification effect and moderate sensitivity. Therefore, it is necessary to distinguish the degree of correlation between the three indicator features and the PRP purification effect to avoid weakening of strongly correlated features and introducing interference from weakly correlated features due to unreasonable allocation of feature weights, which would affect the accuracy of subsequent feature modeling and the PRP purification effect.

[0015] This invention first addresses the differential characteristics of three types of data by employing different single-index feature mapping methods to process the normalized data of each type. Specifically, for hematocrit data, batch normalization combined with convolutional layer processing is used to ensure feature stability while extracting features. For blood viscosity data, ReLU function combined with convolutional layer processing is used to adapt to its fluctuation characteristics under multiple shear rates and enhance effective features. For platelet concentration data, depthwise separable convolutional layer processing is used to reduce computational complexity while extracting features, providing high-quality basic features for subsequent weight calculation. Subsequently, the core of the associated weight gating mechanism is entered. First, a multilayer perceptron combined with different activation functions is used to calculate the corresponding weights for the three types of features. Hematocrit, as the core basis for centrifugation stratification, has strong data stability and directly determines the separation boundary between red blood cells and plasma. A Sigmoid activation function is used to stabilize the corresponding weights within a reasonable range, matching their linear correlation with the PRP stratification effect. Platelet concentration, as the benchmark for calculating PRP enrichment rate, generally shows a stable positive correlation with the final purification effect. Similarly, a Sigmoid activation function is used to ensure that the weights reflect their contribution to the enrichment rate, avoiding the underestimation of the benchmark feature. Blood viscosity includes multiple sets of data at low, medium, and high shear rates. Different shear rate viscosity values ​​directly affect the sedimentation rate of blood cells. For example, high shear rate viscosity values ​​are related to the need for rapid sedimentation, while medium shear rate viscosity values ​​determine sedimentation uniformity. These characteristics fluctuate greatly and show a non-linear correlation with centrifugation parameters. A hyperbolic tangent function is used to adapt to these characteristics. The strong fluctuation characteristics allow the weights to dynamically calibrate the priority of viscosity features at different shear rates, solving the problem that traditional methods using fixed weights cannot adapt to the characteristics of the three types of indicators, leading to blurred stratification boundaries or mismatched sedimentation velocities. Then, each feature is multiplied element-wise with its corresponding weight, amplifying the effect of strongly correlated features while suppressing weakly correlated interference features, such as viscosity noise under abnormal shear rates or random fluctuations in platelet concentration. Finally, the processed three types of features are concatenated, and an attention mechanism is used to calculate the fusion weight, rather than simply adding features. This ensures that the fused gating and screening features comprehensively and accurately reflect the core information directly related to PRP purification effects. By progressively addressing and cooperating with each other around the actual needs of PRP purification, this method effectively solves the problems of poor feature selection targeting and inaccurate differentiation of correlation in this scenario, allowing the screened features to directly provide effective support for the subsequent optimization of centrifugation parameters, fitting the actual application scenario.

[0016] Existing technologies for processing multi-indicator feature screening typically employ equal weighting or simply concatenate all features before inputting them into subsequent modules. While some technologies use fixed weights, these weights are preset and cannot be dynamically adjusted based on the actual characteristics of the three types of data and their correlation with purification effects. This results in the inability to fully leverage the role of strongly correlated features, while weakly correlated features or even noisy features interfere with subsequent modeling, ultimately affecting feature quality and the accuracy of PRP purification. In contrast, the correlation weight gating mechanism of this invention has the core advantage of dynamically calculating corresponding weights based on the inherent characteristics of the three types of blood indicators, rather than using fixed weights or equal weighting. It also incorporates an attention mechanism to achieve adaptive fusion of correlation information among the three types of features. This ensures the effective extraction of core features of each indicator and distinguishes the correlation between various features and PRP purification effects, maximizing the role of strongly correlated features and effectively suppressing interference information. The resulting gated screening features are of higher quality and more targeted, laying a solid foundation for subsequent coupled modeling of features and centrifugation parameters and the generation of optimal centrifugation parameters, thereby improving the purity and bioactivity of the final PRP enrichment layer.

[0017] It should be further explained that the three core indicators—hematocrit, blood viscosity, and platelet concentration—all have clearly defined clinical safety and effective threshold ranges. Moreover, there are individual differences in blood indicators among different patients, and the range of characteristic fluctuations varies. Although the gated features have strengthened the strong correlation information, they may still deviate from the clinically reasonable range due to individual blood abnormalities (such as excessively high / low hematocrit or blood viscosity exceeding the normal range) or minor distortions in the feature extraction process. If they are directly used for subsequent centrifugation parameter modeling, the generated centrifugation parameters may not be suitable for clinical needs. At the same time, it is necessary to retain the reasonable differences in individual blood while constraining the features to avoid excessive constraints that lead to feature distortion.

[0018] This invention first calculates clinical constraint weight coefficients based on gating screening features. A multilayer perceptron is used to perform deep analysis of the gating screening features, and the analysis results are normalized using a Softmax function to obtain the clinical constraint weight coefficients. The constraint strength of three types of indicator features is then assigned: higher constraint weights are assigned to features like blood viscosity, which fluctuates greatly and is sensitive to centrifugation and sedimentation, while moderate weights are assigned to relatively stable features like hematocrit, ensuring more targeted constraints and avoiding insufficient or excessive constraints caused by applying a uniform constraint standard to all features. Subsequently, the clinical constraint function is calculated. First, the clinical threshold intervals of hematocrit, blood viscosity, and platelet concentration are concatenated and mapped into a vector of the same dimension as the gating screening features using a multilayer perceptron, ensuring that the constraint benchmark accurately matches clinical standards. Then, the difference between the gating screening features and this mapped vector is calculated to obtain the feature deviation from the clinical range. The degree of variation is determined by performing element-wise multiplication of the variance of the gated screening features. This leverages the variance to dynamically adapt to the individual characteristic fluctuations of different patients. For example, for individuals with large blood viscosity variance, the constraint flexibility of the fluctuating features is enhanced to avoid suppressing reasonable individual differences due to fixed constraints. The LeakyReLU function is used to perform nonlinear activation on the above results, effectively preserving the effective information in the features while suppressing abnormal features with excessive deviation. Finally, gradient penalty coefficient and L2 regularization are combined for gradient penalty to further suppress extreme abnormal features and prevent features from deviating excessively from the clinical range, thus affecting the safety of subsequent modeling. The resulting constrained enhanced features not only pull the deviating features back to the reasonable range through clinical constraints, ensuring the clinical adaptability of subsequent centrifugation parameter modeling, but also preserve reasonable differences in individual blood through dynamic weights and variance adaptation, ensuring that the features still reflect individual blood characteristics and providing support for the generation of personalized centrifugation parameters.

[0019] Compared to existing technologies, this invention dynamically allocates constraint strength through clinical constraint weight coefficients to match the clinical importance of different indicator features. On the other hand, it achieves adaptive adaptation to individual feature fluctuations through multi-mechanism coupling of clinical threshold interval mapping, variance adaptation, nonlinear activation, and gradient penalty, while effectively suppressing extreme abnormal features. This avoids feature distortion caused by fixed threshold pruning and the lack of specificity of uniform constraints.

[0020] Furthermore, step A4 includes: A41: Based on the physicochemical characteristics of blood, perform cross-domain feature initialization to obtain cross-domain residual features; A42: Based on the cross-domain residual characteristics, the coupling relationship between blood physicochemical characteristics and centrifugation parameters, as well as the coupling relationship within centrifugation parameters, is captured through a multimodal interaction mechanism to obtain centrifugation-physicochemical coupling characteristics.

[0021] It should be further explained that blood physicochemical characteristics and centrifugation parameters belong to different data modalities. Blood physicochemical characteristics directly determine the direction of centrifugation parameter adaptation. For example, high hematocrit requires higher rotation speed to achieve stratification. In addition, there are also strict interaction constraints within centrifugation parameters (rotation speed-time, time-temperature). For example, high rotation speed needs to be matched with shorter centrifugation time to avoid platelet rupture, and a temperature of around 37°C needs to be coordinated with rotation speed to ensure activity. Cross-modal data has dimensional and semantic gaps. Traditional methods are difficult to simultaneously and accurately capture the cross-modal coupling relationship between blood physicochemical characteristics and centrifugation parameters, as well as the interaction coupling relationship within centrifugation parameters. This can easily lead to a mismatch between parameter combinations and blood characteristics, ultimately resulting in incomplete stratification, platelet inactivation, and substandard purity, which affects the PRP purification effect.

[0022] To address the aforementioned issues, this invention first maps the centrifugation parameter threshold range to centrifugation parameter threshold range features using a multilayer perceptron. Secondly, it calibrates the cross-domain residual features and centrifugation parameter threshold range features using a self-attention mechanism before multiplying them element-wise. The blood-side self-attention mechanism highlights core blood features strongly correlated with PRP purification, such as features close to the optimal stratification range in hematocrit and the medium shear rate feature determining sedimentation velocity in blood viscosity. The device-side self-attention mechanism focuses on clinically safe and effective ranges suitable for PRP purification from the parameter threshold range, such as avoiding extremely high rotation speeds in the rotation speed range and selecting duration ranges that balance stratification and activity in the time range. The multiplication of these two calibrated mechanisms essentially achieves the docking of key blood features with effective device parameters, rather than indiscriminate coupling, thus solving the problem of traditional methods blindly matching all blood features with parameter ranges, leading to distorted coupling relationships. Next, the above multiplication result is multiplied element-wise again with the result of "the concatenation of cross-domain residual features and centrifugation parameter threshold interval features and then processed by a multilayer perceptron". The hyperbolic tangent function is then activated to deepen the dual correlation between cross-modal coupling and parameter internal coupling: the concatenation operation is used to integrate the collaborative information of blood features-parameters and parameters-parameters, while the multilayer perceptron can uncover the nonlinear relationship in this correlation. For example, the adaptation ratio of rotation speed and time is nonlinear under different blood viscosities. The hyperbolic tangent function can limit the coupling result to a reasonable range and avoid the occurrence of extreme coupling values, such as dangerous combinations of ultra-high speed and long time. Finally, the activated features are processed by the residual block to obtain centrifugation-physicochemical coupling features, which retain the core original features of the blood side and the equipment side. This solves the technical problem of cross-modal coupling and matches the scenario logic in PRP purification where blood characteristics determine equipment parameters and the internal collaboration of equipment parameters ensures the effect. Existing technologies for processing the correlation between blood features and centrifugation parameters suffer from a core deficiency: they fail to address the characteristics of cross-modal and dual-coupling scenarios. One approach uses linear formulas to establish a single correlation, capturing only a simple linear relationship between blood features and a single parameter while ignoring internal interaction constraints, leading to unreasonable parameter combinations. Another approach concatenates blood features and parameter ranges before inputting them into the model without modal alignment and attention calibration, resulting in severe interference from noise features and invalid parameters during coupling, making it impossible to accurately capture the core coupling relationship. The multimodal interaction mechanism of this invention offers three key advantages over existing technologies: first, it achieves cross-domain alignment through modal mapping, resolving the pain point of ineffective interaction between data from different modalities; second, it achieves cross-domain alignment through attention calibration; third, it achieves cross-domain alignment through modal mapping; fourth, it achieves cross-domain alignment through attention calibration; fifth, it achieves cross-domain alignment through modal mapping; sixth, it achieves cross-domain alignment through attention calibration; seventh, it achieves cross-domain alignment through modal mapping; and eighth, it achieves cross-domain alignment through modal mapping. The mechanism employs intention-calibration and hierarchical coupling to capture the cross-modal coupling of blood physicochemical characteristics and centrifugation parameters, as well as the internal coupling of centrifugation parameters, thus adapting to the complex requirements of PRP purification. Thirdly, the entire mechanism does not require pre-set fixed association rules; it can adaptively learn the individual differences in the blood of different patients through multilayer perceptrons and attention mechanisms, avoiding the problem that traditional fixed rules cannot adapt to personalized needs. The resulting centrifugation-physicochemical coupling features not only fully preserve the individual specificity of blood physicochemical characteristics but also accurately integrate the interactive constraint information of centrifugation parameters, providing a high-quality core basis for the subsequent generation of the optimal centrifugation parameter combination. This directly improves the accuracy of parameter optimization, ensuring that the final output PRP enrichment layer possesses both high purity and high bioactivity.

[0023] Furthermore, step A5 includes: A51: Based on the centrifugal-physicochemical coupling characteristics, dynamic correction of coupling distortion is performed to obtain the calibrated coupling characteristics; A52: Based on the calibrated coupling characteristics, coupling modeling is completed through a global attention network to obtain centrifugal parameter characteristics.

[0024] It should be further explained that after generating the centrifugation-physicochemical coupling features, it is necessary to extract the core information that can directly guide the generation of parameters. This invention calculates the calibration constraint weights by combining depthwise separable convolution, batch normalization, global average pooling, and multilayer perceptron with the sigmoid function. Then, the weights are summed with the identity matrix and multiplied element-wise with the original coupling features to complete the coupling distortion correction. Among them, depthwise separable convolution can extract the core information in the coupling features that is strongly related to PRP purification with light computation. Batch normalization can stabilize the feature distribution. Global average pooling is used to suppress local noise generated during the coupling process, so that the final generated calibration constraint weights have dynamic adaptability. For the high-speed features of high hematocrit coupling, the weights will be moderately suppressed to avoid exceeding the equipment safety threshold. For the medium-time features of normal viscosity coupling, the weights will remain stable to retain reasonable correlation. This corrects the coupling distortion and ensures that the calibrated coupling features meet clinical safety requirements and do not deviate from individual blood characteristics. Secondly, based on the calibrated coupling features, the centrifugation parameter features are extracted through an attention mechanism combined with the GeLU function. The attention mechanism adaptively strengthens the core coupling relationships related to the core requirements of PRP purification from the calibrated coupling features, such as the hierarchical adaptation relationship between hematocrit and rotation speed, the synergistic relationship between blood viscosity and sedimentation time, and the relationship between platelet concentration and temperature activity assurance, while weakening secondary correlations. The GeLU function can perform nonlinear optimization on the extracted features, adapting to the nonlinear correlation between blood features and parameter generation in PRP purification, ensuring that the final centrifugation parameter features can accurately reflect individual blood characteristics and provide direct guidance for subsequent parameter generation.

[0025] Furthermore, step A6 includes: A61: The physicochemical characteristics of blood and the centrifugation parameter characteristics are spliced ​​together to obtain the fused features; the fused features are then input into a multilayer perceptron for calculation, and the calculation results are processed by the Softmax function to obtain a centrifugation parameter weight matrix containing centrifugation speed weight, centrifugation time weight, and centrifugation temperature control accuracy weight. A62: Perform element-wise multiplication of the fusion features with the centrifugation parameter weight matrix, and input the result into the linear layer to obtain the preliminary centrifugation parameters; then constrain the preliminary centrifugation parameters within the preset centrifugation parameter threshold range through the interval clipping function to obtain the optimal combination of centrifugation parameters, including the optimal centrifugation speed, optimal centrifugation time, and optimal centrifugation temperature.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) A new centrifuge structure was designed. Cold air is supplied to the cold air chambers of the outer shell and inner shell through the condenser, compressor and solenoid valve. The cold air is conducted into the inner shell to ensure that the blood in the test tube on the angle rotor is in a low temperature state and to ensure that the blood medium will not deteriorate. The servo motor can drive the test tube on the angle rotor to centrifuge at high speed. Centrifugal force is used to centrifuge and settle the mixture. That is, the difference in sedimentation speed of different substances in the centrifugal force field is used to achieve separation, concentration or purification. By inserting the test tube into the test tube holder one, the test tube is heated by the heating plate. The plasma protein is instantly denatured, aggregated and coagulated by the high temperature, thereby solidifying the sample. By inserting the test tube into the test tube holder two, the test tube is blown by the fan to achieve rapid cooling of the test tube.

[0027] (2) In response to the core problems of traditional fixed centrifugation parameters being unable to adapt to the physicochemical properties of different patients' blood and the lack of precision and consistency in manual experience adjustment, this invention addresses the issue by specifically processing the physicochemical data of blood samples, such as hematocrit, blood viscosity, and platelet concentration, to establish a dynamic correlation between blood physicochemical characteristics and centrifugation parameters. It optimizes parameter combinations in conjunction with clinical safety constraints and individual differences, and completes the PRP enrichment layer purification operation by controlling the centrifugation speed, action time, and temperature control precision. This invention can autonomously adapt to the dynamic blood characteristics of different individuals, effectively avoiding problems such as incomplete stratification and platelet inactivation caused by parameter mismatch, significantly improving the purity of the PRP enrichment layer and the biological activity of growth factors, while ensuring the stability and consistency of the purification process. This provides safe, efficient, and personalized technical support for skin revitalization and tissue repair through clinical injection filling in medical aesthetics.

[0028] (3) In response to the problem that the three core indicators in the extraction of blood physicochemical features have large differences in characteristics, and the traditional fixed weights lead to poor feature selection and inaccurate differentiation of correlation, this invention first adopts differentiated single-indicator feature mapping according to the characteristics of the indicators. For hematocrit data, batch normalization combined with convolutional layers is used to ensure stability. For blood viscosity data, ReLU function combined with convolutional layers is used to adapt to fluctuations. For platelet concentration data, depth-separable convolutional layers are used to balance extraction and efficiency. Then, through the correlation weight gating mechanism, different activation functions are combined to dynamically calculate the weights of the three types of features. After the fusion of weights and features and attention weighting, accurate gating features are obtained, which are adapted to the characteristics of each indicator, effectively amplify strong correlation features and suppress interference, solve the problems of blurred layer boundaries and mismatched sedimentation speed, significantly improve feature quality, provide effective support for subsequent centrifugation parameter optimization, and ensure the accuracy and stability of PRP purification.

[0029] (4) To address the issue that gating features may deviate from the clinical range after screening, and the need to balance clinical constraints with individual blood differences, as well as the shortcomings of traditional uniform constraints, this invention first generates dynamic clinical constraint weight coefficients based on gating features by using multilayer perceptron analysis combined with the Softmax function. The constraint intensity is allocated according to the characteristics of the indicators. Then, the clinical threshold range is mapped to the adaptation dimension vector through the blood clinical constraint function. The feature variance is combined to adapt to individual fluctuations. After LeakyReLU activation and gradient penalty, the constraint-enhanced features are obtained, avoiding the drawbacks of uniform constraints. This not only pulls the deviating features back to the clinical range, ensuring the clinical adaptability of subsequent modeling, but also preserves reasonable individual differences, providing support for the generation of personalized centrifugation parameters. At the same time, it suppresses extreme anomalies, improving feature safety and accuracy.

[0030] (5) Addressing the gap between blood physicochemical characteristics and centrifugation parameters across modalities, traditional methods struggle to simultaneously and accurately capture both the cross-modal coupling relationship between blood physicochemical characteristics and centrifugation parameters, as well as the internal interactive coupling relationship within the centrifugation parameters. This often leads to a mismatch between parameter combinations and blood characteristics. This invention achieves multiple technical effects through an interaction mechanism adapted to cross-modal characteristics and a core feature retention design: effectively breaking through the dimensional and semantic barriers of cross-modal data, accurately capturing the cross-modal correlation between blood physicochemical characteristics and centrifugation parameters, and the internal synergistic coupling of rotational speed, time, and temperature. This solves the problem of blind coupling caused by traditional methods. It addresses the issue of poor parameter adaptability; it enables precise matching of centrifugation parameter combinations with individual blood characteristics, such as matching high rotation speeds to high-hematocrit samples and optimal sedimentation times to high-viscosity samples; it significantly improves industry pain points such as incomplete stratification, platelet inactivation, and substandard PRP purity, ensuring that the centrifugation process fully preserves platelet activity and growth factor bioactivity, while possessing strong individual adaptability, dynamically responding to differences in blood indicators among different patients, ensuring that various blood samples can stably output high-quality PRP enrichment layers, and greatly improving the safety and effectiveness in clinical applications of medical aesthetics. Attached Figure Description

[0031] Figure 1 This is a side-view three-dimensional structural diagram of Embodiment 1 of the present invention; Figure 2 This is a rear-view stereoscopic structural diagram of Embodiment 1 of the present invention; Figure 3 This is a three-dimensional structural diagram of Embodiment 1 of the present invention without a lid; Figure 4 This is a schematic diagram of the three-dimensional structure after sectioning in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the right side structure after removing the outer shell in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram of the left side structure after removing the outer shell in Embodiment 1 of the present invention; Figure 7 This is a bottom view of the structure of Embodiment 1 of the present invention after the outer shell has been removed; Figure 8 This is a schematic diagram of the outer shell, inner shell, and angular rotor of Embodiment 1 of the present invention; Figure 9 This is a bottom view of the outer shell, inner shell, and servo motor of Embodiment 1 of the present invention; Figure 10 This is a schematic diagram of the angle rotor and servo motor in Embodiment 1 of the present invention; Figure 11 This is a schematic diagram of the three-dimensional structure of the horizontal rotor in Embodiment 1 of the present invention; Figure 12This is a schematic diagram of the three-dimensional structure of the hanging cup according to Embodiment 1 of the present invention; In the diagram: 1. Main unit chassis; 2. Front chassis; 3. Side chassis; 4. Cover; 5. Outer shell; 6. Inner shell; 7. Angle rotor; 8. Servo motor; 9. Tripod; 10. Shock-absorbing pad; 11. Support leg; 12. Triangular leg; 13. Electric spring; 14. Drive circuit board; 15. Touch screen; 16. Electronic lock; 17. Observation window; 18. Water collection box; 19. Heating plate; 20. Heat insulation plate; 21. Test tube holder one; 22. Partition plate; 23. Test tube holder two; 24. Fan; 25. Condenser; 26. Compressor; 27. Solenoid valve; 28. Sealing ring; Figure 13 This is a flowchart illustrating the PRP quality optimization method of Embodiment 2 of the present invention. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings, but this is not intended to limit the present invention in any way. Any modifications or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention.

[0033] Example 1: Please see Figure 1-3 This embodiment provides a gel preparation centrifugation device, including a main unit box 1, a front unit box 2 fixedly installed at the front of the main unit box 1, a side unit box 3 fixedly installed on the side of the main unit box 1, and a box cover 4 hinged to the top of the main unit box 1.

[0034] Please see Figure 4-9 An outer shell 5 is fixedly installed on the inner wall of the main unit casing 1. An inner shell 6 is fixedly installed inside the outer shell 5, forming a cooling cavity between the outer shell 5 and the inner shell 6. A sealing ring 28 is provided on the top of the inner shell 6 to ensure a seal between the inner shell 6 and the casing cover 4, preventing gas leakage. An angle rotor 7 and a horizontal rotor are provided inside the inner shell 6. A test tube groove is provided on the angle rotor 7 to facilitate the tilted installation of test tubes. Please refer to [link / reference]. Figure 11-12 A hanging cup is detachably mounted on the horizontal rotor, and a cover is mounted on the hanging cup. The bottom of the angle rotor 7 is detachably mounted to the output end of the servo motor 8. The servo motor 8 is fixed to the inner bottom wall of the main unit 1 via a tripod 9, shock-absorbing pads 10, support legs 11, and bolts. The servo motor 8 can drive the angle rotor 7 to rotate at high speed, achieving centrifugal separation. At the same time, the vibration generated during operation is absorbed by the shock-absorbing pads 10, reducing noise generation.

[0035] Please see Figure 4-10The bottom of the outer shell 5 is connected to the main unit 1 via a triangular leg 12. The bottom hinge of the cover 4 has an electric spring 13. The electric spring 13 is movably connected to the inner bottom wall of the main unit 1 via a mounting base. The cover 4 can be automatically opened by the electric spring 13 to reduce bacterial contact. The bottom wall of the cover 4 is embedded with an ultraviolet lamp, which is used to sterilize the inner shell 6 when the equipment is not in use. The surface of the cover 4 is embedded with an observation window 17 for viewing during centrifugation.

[0036] Please see Figure 4-10 The front chassis 2 has a drive circuit board 14 embedded in its inner wall, a touch screen 15 embedded in its surface, and an electronic lock 16 fixedly installed on its inner top wall. The latch of the electronic lock 16 is installed on the bottom wall of the cover 4, and the body of the electronic lock 16 is installed on the inner top wall of the front chassis 2. When the cover 4 is closed, the electronic lock 16 can lock the cover 4.

[0037] Please see Figure 4-10 A water collection box 18 is slidably installed on the front side of the side casing 3. A heating plate 19 is fixedly installed on the inner bottom wall of the side casing 3 for heating the test tubes. Insulation plates 20 are provided on both sides of the heating plate 19 to reduce heat loss to both sides. Test tube slots are opened inside the heating plate 19. A test tube holder 21 is fixedly installed on the top of the heating plate 19. A partition plate 22 is fixedly installed on the inner top wall of the side casing 3 and on the side of the insulation plate 20. A second test tube holder 23 is embedded on the surface of the side casing 3 and in the space between the partition plates 22. A fan 24 is fixedly installed on the side wall of the side casing 3. The operation of the fan 24 can provide air cooling for the test tubes inserted in the second test tube holder 23.

[0038] Please see Figure 4-7 A condenser 25 and a compressor 26 are fixedly installed at the rear of the side casing 3. The condenser 25 and the compressor 26 are connected by a pipe. The other end of the condenser 25 is connected to a solenoid valve 27. The other end of the solenoid valve 27 is connected to the cooling chamber. The refrigeration principle here is the same as that of an air conditioner and a refrigerator.

[0039] Please see Figure 1-10 The servo motor 8, electric spring 13, drive circuit board 14, touch screen 15, electronic lock 16, heating plate 19, fan 24, condenser 25, compressor 26 and solenoid valve 27 are all existing technologies. Their working principle, size and model are not related to the function of this application, so they will not be described in detail. When using them, they can be selected according to the driving conditions.

[0040] The centrifuge device provided in this embodiment also includes a speed drive unit for controlling the centrifugation speed, a timing unit for controlling the centrifugation time, a temperature control unit for controlling the centrifugation temperature, and a parameter control module for calibrating and adjusting the above parameters in real time.

[0041] Working principle: The box cover 4 is automatically opened by the electric spring 13, the test tube is inserted into the angle rotor 7, and then the box cover 4 is automatically closed; the test tube on the angle rotor 7 can be centrifuged at high speed by the servo motor 8, and the mixture is centrifuged and settled by the centrifugal force. That is, the difference in the settling speed of different substances in the centrifugal force field is used to achieve separation, concentration or purification. At the same time, cold air is supplied to the cold air chambers of the outer shell 5 and the inner shell 6 through the condenser 25, compressor 26 and solenoid valve 27. The cold air is conducted into the inner shell 6 to ensure that the blood in the test tube on the angle rotor 7 is in a low temperature state and to ensure that the blood medium does not deteriorate. By inserting the test tube into the test tube holder 21, the test tube is heated by the heating plate 19. The plasma protein is instantly denatured, aggregated and coagulated by the high temperature, thereby solidifying the sample. By inserting the test tube into the test tube holder 23, the test tube is cooled down quickly by the fan 24.

[0042] Example 2: Please see Figure 13 This embodiment provides a PRP quality optimization method, which is based on the gel preparation centrifugation device in Embodiment 1 and includes the following steps: A1: Collect hematocrit data, platelet concentration data, and blood viscosity data, and normalize them respectively to obtain normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data; specifically including: A11: Collect hematocrit data and platelet concentration data of blood samples using a fully automated blood analyzer; A12: Blood viscosity data of blood samples were collected using a rotational viscometer, including low shear rate viscosity values, medium shear rate viscosity values, and high shear rate viscosity values. A13: The hematocrit data, blood viscosity data, and platelet concentration data were processed using the linear normalization method to obtain normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data, respectively.

[0043] A2: Based on the normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data, single-index feature mapping is performed to obtain the hematocrit characteristics, blood viscosity characteristics, and platelet concentration characteristics. The calculation method is as follows: ; in, This is a characteristic of hematocrit. For batch normalization, It is a convolutional layer. The data represents the normalized hematocrit. The blood viscosity characteristics are shown at low, medium, and high shear rates, respectively. For ReLU function, These represent the low shear rate viscosity, medium shear rate viscosity, and high shear rate viscosity values ​​in the normalized blood viscosity data. Characteristic of platelet concentration. It is a depth-separable convolutional layer. The data represents the normalized platelet concentration.

[0044] A3: Extract the physicochemical characteristics of blood based on hematocrit, blood viscosity, and platelet concentration characteristics; specifically including: A31: Based on hematocrit characteristics, blood viscosity characteristics, and platelet concentration characteristics, a weighted gating mechanism is used to calculate the gating selection features. The calculation method is as follows: ; in, Hematocrit weighting, For the Sigmoid function, It is a multilayer perceptron. As the weight for blood viscosity, It is the hyperbolic tangent function. For splicing operations, Platelet concentration weighting, For association weights, For attention mechanisms, For Hadama accumulation, For gating and filtering features; A32: Based on the gating screening features, feature constraint enhancement is performed using a blood clinical constraint function to obtain constraint-enhanced features. The calculation method is as follows: ; in, This represents the clinical constraint weighting coefficient. For the Softmax function, To constrain and enhance features, To sum element by element, The clinical constraint function for blood is calculated as follows: ; in, For the LeakyReLU function, These are the clinical threshold ranges for hematocrit, blood viscosity, and platelet concentration, respectively. For variance calculation, The gradient penalty coefficient is... For L2 regularization; A33: Based on the constraint enhancement features, feature dimensionality reduction and integration are performed using global average pooling combined with a fully connected layer to obtain the blood physicochemical features. The calculation method is as follows: ; in, For the physicochemical characteristics of blood, For linear layers, This is global average pooling.

[0045] In other embodiments, for scenarios where significant local key information exists in the constraint enhancement features, such as when some patients' blood indicators are close to clinical thresholds, the present invention also provides a calculation method that can fuse global and local features to replace step A33. The calculation method is as follows: ; in, This is for global max pooling.

[0046] A4: Based on the physicochemical characteristics of blood, extract cross-domain residual features; based on the cross-domain residual features, extract centrifugation-physicochemical coupling features; specifically including: A41: Based on the physicochemical characteristics of blood, cross-domain feature initialization is performed to obtain cross-domain residual features, calculated as follows: ; in, For cross-domain mapping features, For cross-domain residual characteristics, For residual blocks; A42: Based on the cross-domain residual characteristics, the coupling relationship between blood physicochemical characteristics and centrifugation parameters, as well as the coupling relationship within the centrifugation parameters, is captured through a multimodal interaction mechanism to obtain the centrifugation-physicochemical coupling characteristics. The calculation method is as follows: ; in, This represents the threshold range characteristics of centrifugation parameters. This refers to the threshold range for centrifugation speed, centrifugation time, and centrifugation temperature. For self-attention mechanism, It exhibits centrifugal-physicochemical coupling characteristics.

[0047] A5: Based on the centrifugation-physicochemical coupling characteristics, extract the calibrated coupling characteristics, and based on the calibrated coupling characteristics, extract the centrifugation parameter characteristics; specifically including: A51: Based on the centrifugal-physicochemical coupling characteristics, dynamic correction of coupling distortion is performed to obtain the calibrated coupling characteristics. The calculation method is as follows: ; in, To calibrate the constraint weights, For the calibrated coupling characteristics, It is the identity matrix; A52: Based on the calibrated coupling characteristics, coupling modeling is performed using a global attention network to obtain the centrifugal parameter characteristics, calculated as follows: ; in, For centrifugation parameter characteristics, This refers to the GeLU function.

[0048] In this embodiment, the parameter settings for the depthwise separable convolutional layer, batch normalization layer, global average pooling layer, multilayer perceptron, identity matrix, and global attention network involved in step A5 are as follows: Depthwise separable convolutional layer: The input feature dimension is consistent with the dimension of the centrifugal-physicochemical coupling feature, set to a 64-channel, 16×16 feature map; a 3×3 convolutional kernel is used, the stride is set to 1, and the edge padding mode is set to "same" mode; Batch normalization layer: Momentum parameter set to 0.9, epsilon parameter set to 1e-5; Global average pooling layer: The pooling window size is consistent with the feature map size output by the batch normalization layer, which is 16×16. Multilayer perceptron: The input dimension is a 64-dimensional vector output by the global average pooling layer. One hidden layer is set with 32 neurons. Linear activation is used. The output dimension is consistent with the number of channels of the centrifugal-physical coupling feature. Identity matrix: Dimensions set to 64×64; Attention mechanism: The input dimension is consistent with the number of channels of the calibrated coupled features, which is 64 dimensions; 8 attention heads are set to evenly divide the input feature dimension into 8 parts, each with a dimension of 8 dimensions; the key vector and query vector are both 8-dimensional, and the value vector is 64-dimensional. The attention weights are calculated using the scaled dot product method.

[0049] A6: Based on the physicochemical characteristics of blood and the characteristics of centrifugation parameters, extract the centrifugation parameter weight matrix, and calculate the optimal combination of centrifugation parameters based on the centrifugation parameter weight matrix; specifically including: A61: The physicochemical characteristics of blood and the centrifugation parameter characteristics are spliced ​​together to obtain the fused features; the fused features are then input into a multilayer perceptron for calculation, and the calculation results are processed by the Softmax function to obtain a centrifugation parameter weight matrix containing centrifugation speed weight, centrifugation time weight, and centrifugation temperature control accuracy weight. A62: Perform element-wise multiplication of the fusion features with the centrifugation parameter weight matrix, and input the result into the linear layer to obtain the preliminary centrifugation parameters; then constrain the preliminary centrifugation parameters within the preset centrifugation parameter threshold range through the interval clipping function to obtain the optimal combination of centrifugation parameters, including the optimal centrifugation speed, optimal centrifugation time, and optimal centrifugation temperature.

[0050] A7: Send the optimal combination of centrifugation parameters to the parameter control module of the gel preparation centrifuge device, start the centrifugation operation, and perform parameter calibration and real-time adjustment on the speed drive unit, timing unit, and temperature control unit; when the centrifugation speed, centrifugation time, and centrifugation temperature all reach the preset standards, stop the centrifugation operation and collect the PRP enrichment layer.

[0051] In this embodiment, a blood sample was collected from a 31-year-old female patient seeking cosmetic procedures. Although the physicochemical indicators of the blood sample were within the normal clinical range, they were significantly biased towards the borderline high values. Traditional fixed-parameter centrifugation methods could not avoid the potential purification risks. The method of this invention can output the optimal centrifugation parameters to overcome the above-mentioned defects. The specific experimental process and results are as follows: Blood sample collection stage: The patient's hematocrit was 49% and platelet concentration was [data missing], obtained using a fully automated blood analyzer. Blood viscosity data were collected using a rotational viscometer, with the low shear rate being... The shear rate is High shear rate The above hematocrit data, blood viscosity data, and platelet concentration data were preprocessed using linear normalization to obtain normalized data that met the model input requirements. Optimal centrifugation parameter calculation stage: The preprocessed normalized data is input into the model of this invention. After single-index feature mapping, correlation weight gating mechanism screening, blood clinical constraint function enhancement, multimodal interaction mechanism coupling, and centrifugation parameter feature extraction, it enters the centrifugation parameter weight matrix calculation and optimal parameter generation stage. The first step is to splice the physicochemical characteristics of blood with the centrifugation parameter characteristics to obtain fused features. After inputting the features into a multilayer perceptron for processing, the features are processed by Softmax to obtain a weight matrix with centrifugation speed weight of 0.32, centrifugation time weight of 0.35, and centrifugation temperature control accuracy weight of 0.33. The second step involves multiplying the fusion features and weight matrix element by element and inputting them into a linear layer to obtain preliminary centrifugation parameters. These parameters are then constrained by an interval pruning function within a preset threshold range (speed 2000-3500 rpm, time 5-15 minutes, temperature 35-38℃). The final output is the optimal combination of centrifugation parameters: optimal centrifugation speed 2800 rpm, optimal centrifugation time 8 minutes, and optimal centrifugation temperature 36.5℃.

[0052] Most existing traditional centrifugation methods use fixed clinical parameters: centrifugation speed of 3000 rpm, centrifugation time of 10 minutes, and centrifugation temperature of 37°C. These parameters are general settings for ordinary blood samples, which can lead to excessive shear force and prolonged exposure of platelets to a high-shear environment, resulting in decreased activity. They are not suitable for the special samples with high viscosity and high hematocrit in this embodiment.

[0053] After the blood samples in this embodiment were processed using the above-mentioned optimal centrifugation parameters, the purity of the PRP enrichment layer reached 91% and the platelet activity reached 89%, which meets clinical requirements. However, after centrifugation using traditional fixed parameters, the purity of the PRP enrichment layer was only 78% and the platelet activity was only 72%.

[0054] The neural network modules involved in this invention include multilayer perceptrons, convolutional layers, attention mechanisms, residual blocks, etc. The specific training process is as follows: Training data consists of hematocrit data, blood viscosity data (including low, medium, and high shear rate values), and platelet concentration data collected by a fully automated blood analyzer and a rotational viscometer, and preprocessed using linear normalization. Corresponding clinically optimal centrifugation parameter combinations (including centrifugation speed, reaction time, and temperature control accuracy) or quality indicators such as PRP enrichment layer purity and platelet activity are used as supervision labels. The mean squared error loss function is selected to measure the relationship between the centrifugation parameter characteristics output by the model and the actual values. The model employs label differentiation, combined with L2 regularization and batch normalization mechanisms to suppress overfitting and improve generalization ability. The optimizer uses the Adam optimizer, which adaptively adjusts the learning rate and calculates the gradients of parameters for each module during training based on the backpropagation algorithm, iteratively updating the weight parameters of components such as the multilayer perceptron, convolutional layers, and attention mechanisms. The training process uses mini-batch stochastic gradient descent, inputting training data into the model in preset batches. Through an iterative process of forward propagation to calculate the output results and backpropagation to update parameters, the model training is completed when the loss function value on the validation set stabilizes and meets the preset convergence conditions.

[0055] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0056] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0057] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A gel preparation centrifuge apparatus, comprising a main unit housing (1), a front housing (2) fixedly mounted at the front of the main unit housing (1), a side housing (3) fixedly mounted on the side of the main unit housing (1), and a housing cover (4) hinged to the top of the main unit housing (1), characterized in that: The inner wall of the main unit (1) is fixedly installed with an outer shell (5), and an inner shell (6) is fixedly installed inside the outer shell (5). A cooling cavity is formed between the outer shell (5) and the inner shell (6). A sealing ring (28) is provided on the top of the inner shell (6). An angle rotor (7) and a horizontal rotor are provided inside the inner shell (6). A hanging cup is detachably installed on the horizontal rotor. The bottom of the angle rotor (7) is detachably installed with the output end of the servo motor (8). The servo motor (8) is fixedly installed with the inner bottom wall of the main unit (1) through a tripod (9), a shock-absorbing pad (10), a support leg (11), and bolts. The device also includes a speed drive unit for controlling the centrifugal speed, a timing unit for controlling the centrifugal time, a temperature control unit for controlling the centrifugal temperature, and a parameter control module for calibrating and adjusting the above parameters in real time.

2. The gel preparation centrifuge apparatus according to claim 1, characterized in that: The bottom of the outer shell (5) is connected to the main unit (1) via a triangular leg (12). The bottom hinge of the cover (4) has an electric spring (13). The electric spring (13) is movably connected to the inner bottom wall of the main unit (1) via a mounting base. The bottom wall of the cover (4) is embedded with an ultraviolet lamp. The surface of the cover (4) is embedded with an observation window (17).

3. The gel preparation centrifuge apparatus according to claim 1, characterized in that: The inner wall of the front chassis (2) is embedded with a drive circuit board (14), the surface of the front chassis (2) is embedded with a touch screen (15), the inner top wall of the front chassis (2) is fixedly installed with an electronic lock (16), the latch part of the electronic lock (16) is installed on the bottom wall of the cover (4), and the body part of the electronic lock (16) is installed on the inner top wall of the front chassis (2).

4. The gel preparation centrifuge apparatus according to claim 1, characterized in that: A water collection box (18) is slidably installed on the front side of the side housing (3). A heating plate (19) is fixedly installed on the inner bottom wall of the side housing (3). Heat insulation plates (20) are provided on both sides of the heating plate (19). A test tube slot is opened inside the heating plate (19). A test tube holder (21) is fixedly installed on the top of the heating plate (19). A partition plate (22) is fixedly installed on the inner top wall of the side housing (3) and on the side of the heat insulation plate (20). A test tube holder (23) is embedded on the surface of the side housing (3) and in the interval of the partition plate (22). A fan (24) is fixedly installed on the side wall of the side housing (3). A condenser (25) and a compressor (26) are fixedly installed at the rear of the side casing (3). The condenser (25) and the compressor (26) are connected by a pipe. The other end of the condenser (25) is connected to a solenoid valve (27). The other end of the solenoid valve (27) is connected to the cooling cavity.

5. A method for optimizing the quality of PRP based on the gel preparation centrifuge apparatus as described in any one of claims 1-4, characterized in that, Includes the following steps: A1: Collect hematocrit data, platelet concentration data, and blood viscosity data, and normalize them respectively to obtain normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data. A2: Based on the normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data, single-index feature mapping was performed to obtain hematocrit characteristics, blood viscosity characteristics, and platelet concentration characteristics. A3: Extract the physicochemical characteristics of blood based on hematocrit characteristics, blood viscosity characteristics, and platelet concentration characteristics; A4: Based on the physicochemical characteristics of blood, extract cross-domain residual features, and based on the cross-domain residual features, extract centrifugation-physicochemical coupling features; A5: Based on the centrifugation-physicochemical coupling characteristics, extract the calibrated coupling characteristics, and based on the calibrated coupling characteristics, extract the centrifugation parameter characteristics; A6: Based on the physicochemical characteristics of blood and the characteristics of centrifugation parameters, extract the centrifugation parameter weight matrix, and calculate the optimal combination of centrifugation parameters based on the centrifugation parameter weight matrix; A7: Send the optimal combination of centrifugation parameters to the parameter control module of the gel preparation centrifuge device, start the centrifugation operation, and perform parameter calibration and real-time adjustment on the speed drive unit, timing unit, and temperature control unit; when the centrifugation speed, centrifugation time, and centrifugation temperature all reach the preset standards, stop the centrifugation operation and collect the PRP enrichment layer.

6. The PRP quality optimization method according to claim 5, characterized in that, Step A1 includes: A11: Collect hematocrit data and platelet concentration data of blood samples using a fully automated blood analyzer; A12: Blood viscosity data of blood samples were collected using a rotational viscometer, including low shear rate viscosity values, medium shear rate viscosity values, and high shear rate viscosity values. A13: The hematocrit data, blood viscosity data, and platelet concentration data were processed using the linear normalization method to obtain normalized hematocrit data, normalized blood viscosity data, and normalized platelet concentration data, respectively.

7. The PRP quality optimization method according to claim 4, characterized in that, The A3 step includes: A31: Based on hematocrit characteristics, blood viscosity characteristics, and platelet concentration characteristics, gate screening characteristics are calculated through an association weight gating mechanism; A32: Based on the gating screening features, feature constraint enhancement is performed through the blood clinical constraint function to obtain constraint-enhanced features; A33: Based on the constraint enhancement features, the blood physicochemical features are obtained by combining global average pooling with a fully connected layer for feature dimensionality reduction and integration.

8. The PRP quality optimization method according to claim 5, characterized in that, The A4 step includes: A41: Based on the physicochemical characteristics of blood, perform cross-domain feature initialization to obtain cross-domain residual features; A42: Based on the cross-domain residual characteristics, the coupling relationship between blood physicochemical characteristics and centrifugation parameters, as well as the coupling relationship within centrifugation parameters, is captured through a multimodal interaction mechanism to obtain centrifugation-physicochemical coupling characteristics.

9. The PRP quality optimization method according to claim 6, characterized in that, The A5 step includes: A51: Based on the centrifugal-physicochemical coupling characteristics, dynamic correction of coupling distortion is performed to obtain the calibrated coupling characteristics; A52: Based on the calibrated coupling characteristics, coupling modeling is completed through a global attention network to obtain centrifugal parameter characteristics.

10. The PRP quality optimization method according to claim 7, characterized in that, Step A6 includes: A61: The physicochemical characteristics of blood and the centrifugation parameter characteristics are concatenated to obtain fused features; the fused features are then input into a multilayer perceptron for computation, and the computation result is processed by a Softmax function to obtain a centrifugation parameter weight matrix containing centrifugation speed weight, centrifugation time weight, and centrifugation temperature control accuracy weight; A62: The fused features and the centrifugation parameter weight matrix are multiplied element-wise, and the computation result is input into a linear layer to obtain preliminary centrifugation parameters; then, the preliminary centrifugation parameters are constrained within a preset centrifugation parameter threshold range by an interval pruning function to obtain the optimal combination of centrifugation parameters, including optimal centrifugation speed, optimal centrifugation time, and optimal centrifugation temperature.

Images