Micro-needle double-activity factor hair transplantation hair follicle accurate activity promoting and positioning system

Abstract

The invention discloses a microneedle double-activity factor hair transplantation hair follicle accurate activity promoting and positioning system, and belongs to the technical field of medical diagnosis assistance, and the system comprises a grid drift correction module, a demand fusion prescription module and a residual error gating classification module. According to the system, firstly, multi-dimensional image acquisition and gridding processing are carried out on an operation area through a grid drift correction module, the basic confidence coefficient and the drift factor vector of each grid unit are obtained, drift correction is carried out on original hair follicle features according to the basic confidence coefficient and the drift factor vector, and correction features are obtained. And the demand fusion prescription module fuses the drift factor and the confidence coefficient, generates a preliminary grid-level active-active factor prescription parameter in combination with the activation income and the risk state, and optimizes the spatial distribution of the delivery parameter through spatial regularization to obtain a final delivery scheme. The residual gating classification module achieves differentiated delivery of high demand points, risk points and the like in the grids based on point position residual scores and heterogeneity gating, and the hair follicle activation promoting effect and safety are improved.

Description

Technical Field

[0001] This invention relates to the field of medical diagnostic auxiliary technology, and in particular to a microneedle dual-activation factor hair follicle precise activation positioning system. Background Technology

[0002] Existing technologies typically employ a "surgical area acquisition - treatment plan formulation - instrument execution" process for assisted treatment: first, information about the surgical area is acquired based on scalp imaging, local tissue condition observation, or multimodal imaging; then, a hair transplant or intervention plan is formulated through methods such as hair follicle identification, density assessment, and direction planning; finally, handheld instruments, semi-automatic devices, or robotic arms perform extraction, implantation, or delivery operations in a skin surface coordinate system. To achieve more precise operations, some plans also establish a surgical area coordinate mapping relationship, corresponding image coordinates to scalp positions, and calculate the target pose or operating parameters of the instruments accordingly; simultaneously, some plans use microneedle stimulation or local delivery of active factors as complementary interventions to support hair follicle activation during or after surgery, thereby improving hair follicle survival and hair regrowth outcomes. Therefore, the key to existing technologies lies in: assessing the state of hair follicles based on limited observational data and translating the assessment results into executable parameters and path control.

[0003] Based on the above ideas, existing publicly available solutions have been improved from the perspectives of "surgical area scanning and solution development" and "multimodal imaging and instrument pose control".

[0004] For example, Chinese invention patent CN113081088A discloses a method for microneedle hair follicle transplantation, which includes the following steps: S1, scanning the area to be transplanted: scanning the area to be transplanted with a scanner; S2, cleaning and disinfecting the scalp and hair: cleaning and disinfecting the patient's head, and disinfecting the area to be transplanted and the area from which hair follicles are extracted separately with alcohol; S3, developing a transplantation plan: developing a corresponding hair transplantation plan based on the data scanned in step S2; S4, extracting hair follicles and preserving them at low temperature: extracting healthy hair follicles from the back of the scalp with tweezers and preserving the hair follicles at low temperature.

[0005] For example, Chinese invention patent CN116421277A discloses a hair follicle transplantation control method, device, computer equipment, and storage medium, which includes: determining the subcutaneous features of hair follicles in ultrasound images; the subcutaneous features are used to reflect the growth state of hair follicles in subcutaneous tissue; determining the hair follicles to be transplanted in the transplantation area based on the skin surface features and subcutaneous features; and determining the target pose required for the end effector of the robotic arm to process the hair follicles to be transplanted based on the subcutaneous features of the hair follicles to be transplanted and the target skin surface coordinates.

[0006] The aforementioned comparative documents show that existing technologies can achieve surgical area information acquisition, plan generation, and instrument execution control to a certain extent, but their core still mainly relies on imaging and feature extraction results to estimate hair follicle status and operating parameters.

[0007] However, in implementing the aforementioned existing technical solutions, at least the following technical problems still exist: Existing technologies often rely on indirect characterizations extracted from limited observational data such as scalp endoscopy, two-dimensional / three-dimensional imaging, or ultrasound, including follicle density, hair shaft diameter, follicle opening visibility, hair follicle color characteristics, and subcutaneous features. These characterizations are easily affected by factors such as differences in acquisition magnification and light source / polarization conditions, scalp oil reflection, dyeing and perming residue, scabs and hair shaft obstruction, local edema caused by local anesthesia or intraoperative exudation, and skin texture stretching caused by changes in traction tension, resulting in significant drift and thus bias in the assessment of the need for activation. This bias further transmits to the process of determining delivery strategies and instrument parameters, making it difficult to accurately match the dosage, delivery depth, and delivery frequency of point-to-point activation intervention with the actual needs. This can easily lead to insufficient allocation to key activation areas resulting in poor activation effects, or excessive intervention in areas with poor stimulation tolerance, leading to the risk of local overstimulation.

[0008] On the other hand, when existing technologies discretize the evaluation results into grid or partition prescription parameters, parameter jumps are prone to occur near the grid boundary, forming unreasonable transition zones; and when a uniform strategy within the grid is adopted, it will mask the differences between points within the grid, making it difficult to simultaneously take into account grid stability and point refinement. Summary of the Invention

[0009] To address the technical problems of existing technologies, such as feature drift caused by the susceptibility of surgical area observation data to disturbances in acquisition conditions and tissue state, leading to biased assessment of hair follicle activation needs and further resulting in inaccurate allocation of dual-activation factor delivery parameters, and the difficulty in simultaneously ensuring grid stability and differentiated control of points within the grid due to boundary jumps in grid prescription parameters, this invention provides a microneedle dual-activation factor hair follicle precise activation positioning system: This system includes: The grid drift correction module is used to perform gridding on the surgical area and obtain the basic confidence and drift factor vector of each grid cell. Based on the drift factor vector, the original hair follicle feature vector of each grid cell is drifted and corrected to obtain the corrected hair follicle feature vector.

[0010] The demand fusion prescription module is used to analyze the hair follicle activation demand status of each grid unit based on the corrected hair follicle feature vector. At the same time, it combines the drift factor vector of each grid unit and the basic confidence to obtain the effective confidence, thereby generating preliminary grid-level dual-activation factor prescription parameters. Based on the preliminary grid-level dual-activation factor prescription parameters, spatial regularization processing is performed to obtain the final grid-level dual-activation factor prescription parameters.

[0011] The residual gating classification module is used to obtain the local residuals of each point in each grid cell relative to the grid cell mean and construct the point residual score. Based on the dispersion of the point residual score, the grid heterogeneity index is obtained and heterogeneity gating is performed.

[0012] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. The microneedle dual-activation factor hair follicle precision activation positioning system provided by this invention achieves high-precision identification of hair follicle status and determination of activation needs under complex and variable acquisition conditions through innovative collaboration of grid drift correction and dynamic confidence weighting modules. Unlike existing technologies that rely solely on a single perspective or static features for direct determination, this solution utilizes multi-perspective and multi-condition acquisition and effectively corrects data deviations caused by multi-source disturbances such as illumination, tension, and local occlusion through drift factor modeling. This processing method, combined with quality self-assessment and automatic resampling mechanisms, significantly improves the authenticity and reliability of the data, thereby ensuring the accuracy and traceability of subsequent prescription generation. Compared to traditional manual subjective evaluation or single-threshold filtering, it can significantly reduce misjudgments and omissions.

[0013] 2. This invention innovatively combines baseline confidence, drift factor, and spatial regularization global optimization to construct a delivery parameter field with adaptive confidence and spatial smoothness. Unlike existing technologies that suffer from parameter fragmentation, boundary abrupt changes, or averaging, this approach can dynamically adjust parameter weights in different regions, achieving spatially smooth optimization of delivery dose, depth, and frequency, effectively suppressing jump zones or delivery blind spots caused by occasional anomalies or boundary effects. The synergistic effect of spatial regularization and confidence fusion ensures both accurate and reliable local decision-making and consistency and coherence of global intervention strategies, significantly optimizing intraoperative experience and clinical controllability.

[0014] 3. This invention effectively overcomes the bottlenecks of existing technologies that rely on average delivery or rigid partitioning through innovative processing of grid heterogeneity gating and point residual classification. The system can automatically identify high-demand points, risk points, and ordinary points based on the activation and risk residuals of points within the grid, and dynamically adjust delivery parameters as needed to achieve differentiated intervention at the point level. Compared to traditional solutions that can only adjust the overall area or a single parameter, this invention can precisely focus on key targets within the same area, avoiding overstimulation of sensitive areas, significantly improving factor utilization efficiency and individualized therapeutic effects, and achieving high survival rates and rapid recovery. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the microneedle dual-activation factor hair transplant follicle precision activation and positioning system provided in the embodiments of this application; Figure 2 This application provides a flowchart of the surgical area gridded acquisition and basic confidence resampling gating process. Figure 3 A flowchart of grid-level prescription generation and spatial regularization for drift correction and effective confidence fusion provided in an embodiment of this application; Figure 4 A flowchart of grid heterogeneity gating and differentiated delivery allocation driven by local residuals provided in this application embodiment. Detailed Implementation

[0017] The technical solution provided in this application will now be described with reference to the accompanying drawings.

[0018] To facilitate understanding of the embodiments of this application, the following points will be explained first: First, in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates an "or" relationship between the preceding and following related objects, but it does not exclude the possibility of indicating an "and" relationship; the specific meaning can be understood in context. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c; a and b; a and c; b and c; or a and b and c. Here, a, b, and c can be single or multiple.

[0019] Second, the use of prefixes such as "first" and "second" in this application is solely for the purpose of distinguishing and describing different things belonging to the same category, and does not constrain the order, size, or quantity of things. For example, "first message" and "second message" are simply different messages, and there is no chronological, size, or priority relationship between them.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] like Figure 1 The diagram shown is a structural schematic of the microneedle dual-activation factor hair follicle precise activation positioning system provided in this application embodiment, including: a grid drift correction module, a demand fusion prescription module, a residual gating classification module, and an activation database.

[0023] The grid drift correction module is connected to the demand fusion prescription module, which in turn is connected to the residual gating classification module. The grid drift correction module, the demand fusion prescription module, and the residual gating classification module are all connected to the activation database. The activation database is used to store various parameters involved in the microneedle dual-activation factor hair follicle precise activation positioning system.

[0024] The grid drift correction module is used to perform gridding on the surgical area and obtain the basic confidence and drift factor vector of each grid cell. Based on the drift factor vector, the original hair follicle feature vector of each grid cell is drifted and corrected to obtain the corrected hair follicle feature vector.

[0025] The demand fusion prescription module is used to analyze the hair follicle activation demand status of each grid unit based on the corrected hair follicle feature vector. At the same time, it combines the drift factor vector of each grid unit and the basic confidence to obtain the effective confidence, thereby generating preliminary grid-level dual-activation factor prescription parameters. Based on the preliminary grid-level dual-activation factor prescription parameters, spatial regularization processing is performed to obtain the final grid-level dual-activation factor prescription parameters.

[0026] The residual gating classification module is used to obtain the local residuals of each point in each grid cell relative to the grid cell mean and construct the point residual score. Based on the dispersion of the point residual score, the grid heterogeneity index is obtained and heterogeneity gating is performed.

[0027] In this embodiment, to ensure consistent calculation methods, traceable and reusable parameters, and ease of preoperative initialization and intraoperative invocation across modules, a pre-established activation database is constructed. This activation database comprises storage media and a database management component, and can be deployed on a local industrial control computer, edge computing device, or cloud server. It interacts with the grid drift correction module, the demand fusion prescription module, and the residual gating classification module through a unified data interface.

[0028] The above-mentioned activation database is established as follows: when the system is first deployed, a system-level basic parameter table and a model parameter table are generated based on pre-calibrated trials, historical case data statistics and clinical rule configuration. In subsequent use, the data generated by each preoperative collection and intraoperative delivery are recorded in a structured manner to form a continuously updated data set.

[0029] The system-level basic parameter table stores globally applicable configuration parameters, including but not limited to grid partitioning specifications (such as the area / number of each grid), gating thresholds (such as grid heterogeneity thresholds and baseline confidence thresholds), upper limits for delivery dose, upper and lower limits for delivery depth, upper and lower limits for delivery frequency, safety thresholds, and initial calibration parameters for the surgical area coordinate system and reference points. These parameters are generally defined by the R&D team / clinical experts and support manual adjustment or upgrade maintenance. The model parameter table stores parameters related to the algorithm model, such as drift factor feature definitions, drift compensation matrices, score fusion weights, effective confidence fusion rules, spatial regularization algorithm parameters (such as neighborhood weights and boundary determination rules), and grading determination parameters. This table can be automatically generated through calibration experiments / model training or manually modified based on multi-center case statistical experience, supporting the parameterized configuration of the core algorithm and model iteration.

[0030] like Figure 2 The flowchart of the surgical area gridded acquisition and basic confidence resampling gating provided in this application embodiment describes the complete link of the system from acquisition to quality gating: First, multiple images are acquired to establish a surgical area coordinate system. Under a unified coordinate reference, the surgical area is divided into grid units. Then, the image quality score of each grid unit is obtained, and the basic confidence of the grid unit is obtained accordingly. Next, the basic confidence of the grid unit is judged to be no lower than the lower limit threshold of the basic confidence: if the judgment result is yes, it means that the observation data of the grid meets the minimum reliability requirement, and no resampling processing is performed on the grid unit; if the judgment result is no, it means that the reliability of the grid data is insufficient, and resampling processing is required on the grid unit to ensure that the data quality entering the subsequent evaluation and prescription calculation is consistent and controllable.

[0031] Specifically, the surgical area is gridded and the basic confidence level and drift factor vector of each grid cell are obtained. The specific process is as follows: Multi-view or multi-condition image acquisition of the surgical area is performed, and a surgical area coordinate system is established for subsequent gridding and data localization. Multi-view or multi-condition image acquisition refers to acquiring multiple sets of image data covering the entire surgical area before or during surgery using multiple cameras or different shooting angles, lighting conditions, polarization, magnification, etc., from the same camera, in order to restore the real tissue structure as much as possible and overcome the blind spots and local occlusions of a single viewpoint. These acquired data are combined with anatomical landmarks or manually or automatically attached reference points (such as markers or special dot arrays), and a unified surgical area coordinate system is established through feature point registration, 3D reconstruction, or affine transformation algorithms. This coordinate system is used to ensure one-to-one correspondence and traceability of image data from different viewpoints and shooting times in space, providing a global reference system for subsequent gridding and precise point localization.

[0032] Based on the aforementioned surgical area coordinate system, the surgical area is divided into grid cells, with each grid cell corresponding to an independent spatial partition for evaluation and processing. Within the surgical area coordinate system, the entire surgical area is divided into several grid cells according to the preset resolution, coverage area, and actual intervention requirements. Each grid cell is typically a regular square, rectangle, or hexagon, which not only completely covers the target area but also facilitates subsequent batch algorithm processing and boundary smoothing. Each grid cell corresponds to an independent spatial partition for evaluation and processing; that is, in subsequent processes, all feature extraction, data statistics, intervention parameter calculation, and delivery operations are completed independently using the grid as the basic unit. This partitioning method allows for refined difference control (e.g., using different dosage strategies for sparse and dense areas) while ensuring overall consistency.

[0033] Image quality scores are obtained by analyzing the quality features of the image corresponding to each grid cell. A baseline confidence level, reflecting the reliability and trustworthiness of the observation data for each grid cell, is then calculated based on this score. The baseline confidence level is used to determine whether to resample the grid cell. Image quality features refer to the quality parameters extracted from the image within the coverage area of ​​each grid cell, including but not limited to sharpness (edge ​​sharpness), signal-to-noise ratio, brightness uniformity, occlusion rate, reflectivity, scab coverage, and the proportion of overexposure or underexposure. These features are weighted and fused or rule-based to obtain an image quality score (usually normalized to the 0-1 range), used to quantify the reliability and usability of the current grid observation data.

[0034] The specific method for obtaining the above-mentioned basic confidence level is as follows: g is the grid number (a local region), g = {1, 2, 3, ..., G}, G is the total number of grid cells, q gω is the grid quality index vector, which is an n×1 column vector. Each component represents a different quality feature, such as sharpness, occlusion rate, and reflectivity. It is obtained directly from the image quality features through image algorithms. ω is the weight vector of each quality index, which is also an n×1 column vector and can be set by calibration or experience. This refers to combining multiple quality factors into a single quality score, where σ(·) is the Sigmoid function, which compresses any real number to (0, 1), allowing c to... g It has a confidence level meaning: the closer it is to 1, the more reliable it is.

[0035] Simultaneously, drift-related features are extracted from the corresponding images and clinical data of each grid cell and combined into a drift factor vector to reflect the current tissue state disturbances and observational uncertainties of each grid cell. Drift-related features refer to external disturbances and tissue state changes that affect the accuracy and stability of the observed data. These include, but are not limited to: magnification, angular offset, changes in light source brightness and polarization angle, scalp tension / deformation, regional edema, abnormal sebum secretion, and local tissue traction. These features can be obtained automatically or semi-automatically through multiple channels such as image analysis, sensor signals, operator annotation, and clinical auxiliary examinations. The drift factor vector encodes the above multidimensional drift features into a unified feature vector, which serves as the input to subsequent drift compensation, feature correction, data weighting, and confidence fusion algorithms, dynamically reflecting the degree to which each grid cell is currently affected by disturbances and uncertainties.

[0036] Furthermore, based on the basic confidence level, a decision is made as to whether to resample the grid cells. The specific decision-making process is as follows: The baseline confidence level of the grid cell is compared with the baseline confidence lower limit threshold. The baseline confidence lower limit threshold can be dynamically adjusted based on historical statistics or clinical requirements, and the specific value can be optimized according to the actual application scenario.

[0037] If the baseline confidence level of a grid cell is lower than the baseline confidence level threshold, it indicates a significant quality issue with the current image or observation data. In this case, resampling of the grid cell is required. Operators should be prompted to re-acquire images of the corresponding area, or supplementary acquisitions should be made using different angles and lighting conditions until the data quality meets the requirements. If the baseline confidence level of a grid cell is not lower than the baseline confidence level threshold, it indicates that the data quality meets the standards. In this case, resampling of the grid cell is not required, and the grid cell can be directly included in subsequent drift analysis, feature extraction, and delivery parameter calculation processes.

[0038] This process can automatically and objectively screen the reliability of collected data, preventing downstream analysis distortion caused by poor shooting conditions or occasional obstruction. By setting confidence thresholds, it can achieve dynamic control and personalized adjustment of the overall data quality of the system, improve the system's adaptability and robustness in different clinical environments, and the timely resampling mechanism can also reduce abnormal interventions and ineffective delivery in the later stages, optimizing the clinical operation experience and the final efficacy.

[0039] like Figure 3 The flowchart of grid-level prescription generation and spatial regularization, which integrates drift correction and effective confidence scores, provided in this application embodiment, describes the main process from drift modeling to prescription parameter output: First, drift factor vectors are obtained. On the one hand, a drift compensation matrix is ​​constructed based on drift information, and the original hair follicle feature vectors of the grid are drift corrected to obtain corrected hair follicle feature vectors. Then, the hair follicle activation demand state of each grid unit is analyzed based on the corrected hair follicle feature vectors, and the activation benefit state and stimulation risk state of each grid unit are output. On the other hand, drift information is also used to combine with the basic confidence score to obtain an effective confidence score, which reflects the acceptability of the observation conclusions of the grid. Subsequently, the above state variables and confidence scores are aggregated to generate preliminary grid-level dual-activation factor prescription parameters, and further processed through spatial regularization to obtain the final grid-level dual-activation factor prescription parameters, so as to suppress prescription mutations between grids, enhance spatial consistency, and retain the true boundaries of definite differences.

[0040] Specifically, the original hair follicle feature vector of each grid cell is shifted and corrected based on the drift factor vector. The specific correction process is as follows: The correspondence between the drift factor vector and the drift of hair follicle features is obtained through pre-calibrated experiments or historical data analysis. A drift compensation matrix is ​​then constructed to correct the weights of the original hair follicle features under different drift factors. The specific construction process involves acquiring standard hair follicle images under different lighting, angle, distance, and skin tension conditions, and statistically analyzing the changes in feature vectors under various drift factors. This leads to the following relationships: certain types of drift (such as increased reflectivity) typically result in a larger hair shaft diameter; certain types of drift (such as magnification changes) affect density per unit area; and certain types of drift (such as occlusion) lead to a decrease in the visibility of the hair follicle opening. The shift pattern is obtained by modeling these patterns. The drift compensation matrix is ​​not a simple linear matrix; it may include: feature dimension weight factors, drift sensitivity coefficients, and cross-influence terms (such as the combined effect of lighting and angle changes on color features). The purpose is to form a stable mechanism that inputs drift factors and outputs correction weights.

[0041] The original hair follicle feature vector of each grid cell is input into the drift compensation matrix. The corresponding correction weight is calculated based on the drift factor vector, so that each feature dimension is reasonably compensated or suppressed. For features that are greatly affected by drift (such as the increase in brightness caused by reflection), the correction weight will automatically reduce its contribution. For relatively stable features that are not affected by drift (such as statistical patterns in color texture), the correction process will maintain their original values. For features with non-uniform drift, adaptive piecewise correction can also be used to avoid over-correction or under-correction.

[0042] Specifically: f g The original hair follicle state feature vector (density, hair shaft diameter distribution, hair follicle opening visibility, erythema / vascular texture, keratinization / scaling texture, color index, etc.) is an n×1 column vector, where n is the number of features and d is the number of features. g The vector represents the drift factors (magnification deviation, polarization angle difference, reflectivity, occlusion rate, crusting rate, texture stretching / tension index, local edema characterization related to local anesthesia exudation, etc.), which are m×1 column vectors, where m is the number of drift factors, and W is the drift compensation matrix (obtained through calibration: the same real area at different d...). g The mapping of the features (the transformation of the features) is an n×m matrix, where n is the number of features (and f). g (same), m is the number of drift factors (and d) g Same), f gw The feature vector for correcting hair follicles is an n×1 column vector.

[0043] The final output corrected feature vector is closer to the actual tissue state, avoiding misinterpreting changes in camera conditions as changes in the hair follicle itself, thereby improving the accuracy of subsequent diagnosis, grading and delivery planning.

[0044] Furthermore, based on the corrected hair follicle feature vector analysis, the hair follicle activation demand status of each grid unit is determined. The specific analysis process is as follows: From the corrected hair follicle feature vector obtained from each grid cell, activation-related features are extracted and normalized and weighted to obtain the activation benefit state of each grid cell. Activation-related features refer to feature indicators that can reflect the activity level, regeneration potential, and response to exogenous stimuli of hair follicles in the target area. Common activation-related features include: hair follicle unit density (low density usually indicates a high demand for activation), miniaturization ratio (high miniaturization ratio indicates low activity states such as atrophy, weakness, and dormancy), hair shaft diameter distribution (thin / soft hair indicates weak activity), hair follicle opening visibility / opening morphology (invisible or closed hair follicle opening indicates metabolic / germination restriction), local blood flow / erythema (moderate congestion indicates good basal metabolism), etc. According to specific implementation needs, the above (or more) features are extracted from the corrected hair follicle feature vector. The above activation-related features are automatically detected by algorithms such as image segmentation, feature extraction, and threshold judgment. Through normalization (i.e., different features are standardized to a unified numerical range) and weighted fusion (different features are assigned different influences based on clinical experience or model weights), a single activation benefit state score is obtained, specifically: A g The activation benefit score (the larger the score, the more worthwhile it is to activate). The weight vector that maps the corrected hair follicle features to the activation benefit state is an n×1 column vector. It can be calibrated with experience and small samples, avoiding complex training dependencies and facilitating implementation. σ(·) is the Sigmoid function, which ensures that the output is in (0, 1), which is convenient for subsequent fusion with threshold and safety upper limit.

[0045] The activation benefit state is used to reflect the potential benefit of the grid cell to the stimulation of the dual activation factors; the activation benefit state score provides a data basis for the subsequent intelligent allocation and delivery scheme optimization of prescription parameters, avoiding excessive or insufficient intervention caused by subjective experience-based decision-making.

[0046] Risk-related features are extracted from the corrected hair follicle feature vectors of each grid cell and then normalized and weighted to obtain the stimulus risk state of each grid cell. Risk-related features are used to identify the potential risk of adverse reactions such as inflammation, tissue damage, or decreased tolerance when a grid cell is subjected to exogenous stimuli (e.g., microneedle delivery, active factor injection). Commonly used risk-related features include: local erythema / congestion / pigmentation (indicating inflammation or abnormally active capillaries), signs of skin barrier damage such as scaling, keratosis, and sebum plugs (indicating barrier fragility), significant perifollicular vascularity or edema (indicating abnormal interstitial fluid and sensitivity of the surgical area), and significant occlusion / contamination around the hair shaft or follicle (affecting delivery uniformity and safety). Similarly, the above (or more) risk-related features are extracted from the corrected feature vectors. These risk-related features are automatically detected using algorithms such as image segmentation, feature extraction, and threshold judgment, and then normalized and weighted to obtain a stimulus risk state score. Specifically: R g β is the stimulus risk state score, and β is the weight vector that maps the corrected hair follicle features to the stimulus risk state. It is an n×1 column vector that can be calibrated empirically and with small samples.

[0047] The stimulation risk state is used to characterize the stimulation sensitivity and safety risk of the grid unit during microneedle delivery and active factor injection. The stimulation risk state score enables the system to balance therapeutic benefits and safety when automatically allocating delivery parameters, reducing the probability of postoperative adverse reactions and achieving personalized delivery with controllable risks.

[0048] Specifically, the effective confidence level is obtained by combining the drift factor vector of each grid cell and the basic confidence level, and then the preliminary grid-level dual-activity factor prescription parameters are generated. The specific process is as follows: The effective confidence level is used to dynamically adjust the generation intensity of grid-level dual-activity factor prescription parameters; specifically, the drift state score is obtained based on the drift factor vector. γ is the weight vector that maps the corrected hair follicle features to the drift state. It is an m×1 column vector that can be calibrated using experience and small samples. The effective confidence level is then... The purpose of this is that when the observation conditions are reliable and the external disturbances are small, the system fully trusts the observation results and uses them to strongly drive the subsequent prescription parameters; when the observation conditions are poor or the drift is severe, the system actively reduces the generation intensity of prescription parameters, tends to a more conservative strategy, and reduces risks.

[0049] Preliminary grid-level dual-activation factor prescription parameters are generated based on the effective confidence level, activation benefit state, and stimulation risk state of each grid cell. The preliminary grid-level dual-activation factor prescription parameters include the dose baseline value, delivery depth baseline value, and delivery frequency baseline value of the grid cell. The calculation weights and safety boundaries of dose, delivery depth, and delivery frequency are dynamically adjusted according to the level of effective confidence, thereby achieving adaptive and personalized parameter allocation. At the same time, spatial regularization processing is performed on all grid cell parameters based on the preliminary grid-level dual-activation factor prescription parameters.

[0050] Actual drug administration (or stimulation, energy, operation time, etc.) needs to consider the maximum allowable value, the reliability of the current observation / edema / drift status, the level of local demand, and the level of local risk. Each factor cannot determine the final value independently; a multiplicative summation is required. A low value in any factor automatically reduces the budget. Therefore, the above dosage baseline value is... B max The upper limit of safety is determined by clinical experience or product safety criteria.

[0051] The above delivery depth benchmark value is h max The upper and lower limits of delivery depth are determined by the equipment and anatomical structure; clip(·) is the upper and lower limit amplitude operation to ensure that it is within the reachable safe range.g -R g When the need for activation is high and the risk is low, a deeper approach is recommended (e.g., to activate deep hair follicles); conversely, a shallower approach is recommended (e.g., for inflammation / keratosis / edema / poor tolerance). eh ×E g When drift or edema is indicated (such as local exudation, subcutaneous tightness, or positioning deviation), the depth automatically converges to a shallower level to prevent excessively deep injections that could lead to leakage, fibrosis, or redness and swelling; k h This is the demand-risk adjustment factor. It is typically determined by clinical experience, equipment testing, or regression analysis; k eh The adjustment coefficient for the depth of drift is fitted or calibrated from historical surgical data.

[0052] The above delivery frequency benchmark value is k n ×A g This indicates that the greater the demand, the more frequent the delivery can be (e.g., delivering medication / energy / stimuli in multiple deliveries), −k r ×R g Higher risk (prone to redness, fibrosis, intolerance) automatically reduces frequency to avoid overstimulation. -k e ×E g The greater the drift, the less frequent the injection should be, to prevent cross-channeling of exudate / injection and cumulative tissue damage; k n To enhance the frequency gain factor for activation demand, k r k is the risk suppression coefficient on frequency. e The coefficients represent the frequency suppression coefficients for drift; all coefficients can be obtained by fitting historical case regression, expert scores, and experimental data from the product development phase, n min n max The upper and lower bounds for delivery frequency are given by physician manuals, guidelines, and registration trial protocols.

[0053] Furthermore, based on the preliminary grid-level dual-activity factor prescription parameters, spatial regularization is performed on all grid cell parameters. The specific processing procedure is as follows: Based on the spatial adjacency relationships between grid cells, a global parameter optimization model is constructed. A weighted difference term for the dose reference values ​​of adjacent grid cells is introduced into the objective function, and the regularization weights are dynamically adjusted according to the effective confidence level of each grid cell. In this embodiment, spatial adjacency refers to the geometric topological connection between adjacent grid cells, which can typically be automatically established through spatial indexing, grid ID mapping, or two-dimensional / three-dimensional coordinate mapping. Each grid cell can form adjacency relationships with several directly adjacent grid cells (such as top, bottom, left, right, diagonal, or hexagonal honeycomb structures). This relationship is not only used to identify boundaries and internal grids but also provides a spatial information basis for subsequent parameter smoothing and global optimization. Through spatial adjacency relationships, the mutual influence of parameters in space can be achieved while ensuring regional independence. For example, abrupt changes in prescription parameters in space often correspond to observation errors or abnormal interference; reasonable use of adjacency relationships can suppress abnormal jumps and enhance overall consistency.

[0054] The core of spatial regularization lies in establishing a global parameter optimization model. The objective function of this model not only includes the term that the cell grid parameters should closely match the original suggested values ​​driven by the local grid data, but more importantly, it introduces a weighted difference term for the parameters of adjacent grids. Specifically: the local fitting term ensures that the prescription parameters of each grid after regularization do not deviate from the actual needs of the grid itself (such as the initial parameters driven by activation benefits, risks, and effective confidence); the neighborhood smoothing term minimizes the weighted differences of parameters among all adjacent grids, promoting a smoother spatial distribution of the parameter field and preventing isolated high / low doses or overly sharp boundary changes. This objective function is typically implemented using quadratic optimization (least squares) or graphical models / Markov random fields, achieving global smoothing while ensuring local accuracy.

[0055] The degree to which each grid cell is smoothed in spatial regularization depends not only on its geometric adjacency but also on the dynamic adjustment of the regularization weight based on its effective confidence level. Specifically: grid cells with high effective confidence levels indicate that their observation data is reliable and drift correction is sufficient; the system should place more trust in their local original parameters and assign them a smaller smoothing weight to make them less susceptible to being influenced by their neighbors and to highlight local characteristics. Grid cells with low effective confidence levels indicate that their local observations are unreliable; the system should adopt more neighborhood parameters and actively smooth them with a higher regularization weight to make their parameters more conform to the overall trend and avoid outliers disrupting the global distribution. This confidence-based adaptive regularization mechanism balances local authenticity with global consistency, ensuring that parameter optimization is both safe and accurate.

[0056] When the set smooth triggering conditions are met, spatial regularization can also be performed on the delivery depth reference value and / or delivery frequency reference value; after spatial regularization, the final grid-level dual-activity factor prescription parameters are obtained.

[0057] Under normal circumstances, spatial regularization is mainly performed on dose baseline values ​​to avoid large spatial fluctuations in delivered dose. However, for delivery depth and delivery frequency baseline values, the system allows setting smoothing trigger conditions. If the parameter difference between adjacent grids exceeds a preset threshold, or the effective confidence of a grid is lower than the safety lower limit, the system automatically performs local spatial smoothing operations on the corresponding region for that parameter. For example, regularization is only triggered when the spatial distribution of depth / frequency shows abnormal jumps, or when its local effective confidence is low and special protection is required clinically. The regularization form is similar to that of dose, but different weight settings and adjacency methods can be used (e.g., smoothing only high-risk boundaries, leaving internal points unprocessed). This flexible and adjustable multidimensional regularization mechanism can better adapt to the actual risks and clinical needs of different parameters during delivery.

[0058] After spatial regularization, the system obtains a set of final grid-level dual-activation factor prescription parameters (dose baseline, delivery depth baseline, and delivery frequency baseline). These parameters exhibit continuous spatial variation, natural boundary transitions, and fully reflect actual tissue needs and observational reliability. This processing effectively eliminates local dose peaks / troughs caused by individual anomalous observations, avoids complications from excessive activation or inhibition, optimizes intraoperative operational consistency, and improves the experience for medical staff and patients.

[0059] like Figure 4 The flowchart of grid heterogeneity gating and differentiated delivery allocation driven by local residuals provided in this application describes the logic from whether grid subdivision is needed to how to differentiate allocation: The system first obtains the local residuals of each point within each grid cell relative to the grid cell mean, then constructs the point residual score, and obtains the grid heterogeneity index based on the dispersion of the point residual score. Subsequently, gating is completed by determining whether the node grid heterogeneity index is lower than the grid heterogeneity threshold: if yes, it means that the differences between points within the grid are insufficient to support a finer-grained strategy, and no point-level classification is required; if no, it means that there are significant differences within the grid, and point-level classification is required. Furthermore, a classification label for each point is generated according to the point classification rules, and finally, differentiated dual-active factor delivery parameter allocation is performed based on the classification results, thereby achieving fine-grained delivery control at the point level within the grid without compromising grid-level stability.

[0060] Specifically, the local residuals of each point within each grid cell relative to the grid cell mean are obtained, and the point residual fractions are constructed. The specific process is as follows: The characteristic mean of each grid cell is determined based on the characteristic data of all points within each grid cell. The characteristic data of each point is then compared with the characteristic mean of its corresponding grid cell to obtain the local residual of each point relative to the grid cell mean.

[0061] The process of determining the feature mean is as follows: Each grid cell contains several sampling points. The system first collects feature data (such as hair follicle activity, inflammation characteristics, structural indicators, etc.) for all points within the grid, and calculates the mean for each feature across all points to form the feature mean of the grid cell. This allows the system to eliminate natural differences between individuals by using the overall state within the current grid as a reference. In this embodiment, the mean for each feature data at each sampling point is calculated independently within each grid cell, thus forming a feature mean vector containing multidimensional content. Each dimension's mean reflects the average level of the current grid as a whole for that feature. For example, after calculating the mean for hair follicle activity, inflammation characteristics, structural indicators, etc., these are used as a standard reference for the current grid, laying the foundation for subsequent local residual analysis and precise control at each point.

[0062] The local residual is obtained by subtracting the feature data of each point within the grid from the feature mean of the corresponding grid cell. This yields the local residual, which represents the degree of deviation of that point from the average level of the grid. It is a residual vector composed of multiple features, thus enabling quantitative identification of which points are more anomalous or require special attention compared to the average of the region.

[0063] The activation-related and risk-related residual components are extracted from the local residuals, and point residual scores are constructed for each point within the grid cell based on these components. Using predefined feature channels, the residual vector is physiologically decomposed into activation-related residual components reflecting activation-related features such as insufficient hair follicle activity, miniaturization, and density gaps, and risk residual components reflecting risk features such as erythema, inflammation, skin barrier fragility, and sebum plugs. Subsequently, the activation-related and risk-related residual components are normalized and weighted, respectively, to form individual activation-related and risk-related residual values. Point residual scores are then constructed based on the activation-related residual values ​​(as positive factors for activation demand) and the risk-related residual values ​​(as negative factors for suppressing delivery intensity). A higher point residual score indicates that the point has greater activation value and lower risk compared to the grid average, thus serving as a basis for subsequent point priority ranking and differentiated delivery parameter allocation.

[0064] Specifically, the heterogeneity index of the grid is obtained based on the dispersion of the point residual fractions, and heterogeneity gating is performed. The specific process is as follows: The grid heterogeneity index, reflecting the degree of fluctuation in residual distribution, is determined based on the residual scores of all points within a grid cell. Within a grid cell, the residual scores of all points may be highly concentrated or dispersed. In this embodiment, the variance of the residual scores is preferably used as the grid heterogeneity index to quantify the dispersion of residual distribution among points within the grid. Specifically, the grid heterogeneity index is defined as the statistical variance of the residual scores of all points within the grid (i.e., the average of the squared differences between the residual scores of all points and the mean of the grid). The higher the value, the greater the state difference among points within the grid, indicating a significant differentiation in activation needs and risks, requiring further subdivision and differentiation at the point level. If the grid heterogeneity index is low, it indicates that the residual scores of points within the region are relatively concentrated and the state is balanced, allowing for unified overall processing.

[0065] Performing heterogeneity gating refers to comparing the grid heterogeneity index with the grid heterogeneity threshold; the grid heterogeneity threshold refers to the upper limit of the grid heterogeneity index extracted from the activation database within a specified range.

[0066] When the grid heterogeneity index is lower than the grid heterogeneity threshold, it is determined that no point-level classification is required; when the grid heterogeneity index is not lower than the grid heterogeneity threshold, it is determined that point-level classification is required.

[0067] This gating mechanism avoids the inefficiency of performing complex classifications for all grids, and only refines the processing when it is truly necessary, thus achieving automatic diversion and intelligent cost reduction and efficiency improvement.

[0068] Furthermore, the locations are categorized, and the specific categorization process is as follows: Based on the residual scores of each point within the grid cell, a classification label is generated for each point according to the point classification rules. In this embodiment, a quantile threshold-based method is preferred, marking points with residual scores in the top few quantile intervals (e.g., the top 20%) as high-demand, low-risk points (core points), marking points with scores in the bottom few quantile intervals (e.g., the bottom 20%) as high-risk points, and classifying the remaining points as ordinary points. The above classification rules can be dynamically optimized based on clinical experience, statistical analysis results, or machine learning models to ensure that the point classification conforms to the data distribution characteristics and has medical relevance and operability.

[0069] The site grading rules are used to classify sites into different levels (such as core sites, high-risk sites, and ordinary sites) based on the residual scores of each site, and generate corresponding grading labels. The system then assigns different dual-active factor delivery parameters to each site within the grid cell according to the site's grading label. For example, higher doses or deeper delivery are assigned to core sites, while lower doses or shallower delivery are assigned to high-risk sites, thereby achieving differentiated and refined adjustment of delivery parameters to maximize efficacy and reduce risk.

[0070] In this embodiment, the site-level delivery parameter allocation is a further refinement based on the final grid-level dual-activity factor prescription parameters obtained after spatial regularization. The system first uses the dose reference value, delivery depth reference value, and delivery frequency reference value of each grid cell as the basic anchor points for the delivery parameters of all sites within that grid. Subsequently, based on the site residual score and classification results, the system fine-tunes the parameters for different categories of sites using a tiered correction coefficient. This achieves upward adjustment of delivery parameters for high-demand sites, downward adjustment of delivery parameters for high-risk sites, and maintenance of baseline parameters for ordinary sites, thereby completing differentiated delivery at the site level. The above correction process is always constrained by grid-level and global safety constraints, ensuring that the site-level allocation does not exceed safety boundaries, effectively balancing the continuity of parameter distribution, local adaptability, and overall system safety.

[0071] In a specific example embodiment, the preferred approach is to assign an upward correction factor of 1.2 to 1.4 to core points, maintain 1.0 for ordinary points, and assign a downward correction factor of 0.6 to 0.8 to risk points. Actual delivery parameters are calculated using the formula "point parameter = correction factor × grid baseline parameter," and limited to a safe range allowed by the grid parameters using the clip function. For parameters such as dose, delivery depth, and delivery frequency, the above method is used to achieve point-level differentiated allocation. This scheme ensures the continuity, safety, and accuracy of delivery parameter distribution, facilitating engineering implementation and clinical operation.

[0072] In a specific example embodiment, a dual-channel microneedle delivery and positioning device is used, which includes: an image acquisition component (with switchable modes of coaxial white light + cross-polarized light, and a fixed magnification lens), a surgical area reference point attachment component (for establishing coordinate mapping from the image to the scalp), a handheld single-needle delivery head (needle body connected to a lead screw-type depth limiting mechanism, driven by a stepper motor, with a depth resolution of 0.05 mm and an allowable delivery depth range of 0.8–2.2 mm), a dual-channel micro-infusion pump (A / B channels, with linear calibration of stepping pulse and output volume, a minimum controllable output of 0.1 μL, and tubing with pressure sensors for needle blockage / backflow warnings), and a control terminal and activation database (storing thresholds, calibration matrices, weight vectors, prescriptions, and logs).

[0073] In this embodiment, the surgical area is divided into 10mm × 10mm grid cells, with 16 delivery points planned within each grid; the system pre-stores the grid dose safety upper limit B in the activation database. max =120μL / grid, delivery depth upper and lower bounds h min =0.8mm, h max =2.2mm, frequency upper and lower bounds n min =1、n max=3, and the weight vector γ used for drift state calculation and the coefficient k used for prescription mapping. h k eh k n k r k e (These parameters can be obtained and solidified through product calibration experiments / historical case regression / expert rules).

[0074] After the surgery begins, the device acquires a white light image and a cross-polarization image of a target grid g, extracts image quality features, and obtains the basic confidence level c. g For example, the grid still exhibits some reflection and slight occlusion under polarization, but its clarity is good, resulting in an overall score of c. g =0.85. Simultaneously, the system extracts drift-related features from the image corresponding to this grid and the on-site record to form a drift factor vector d. g (e.g., polarization angle deviation, reflectance intensity shift, magnification deviation, local edema / compactness index, etc.), and calculate the drift state fraction: In this example, take γ=[0.8, 0.5, 0.2, 0.6], d g =[0.4, 0.3, 0.1, 0.4] (all are standardized drift intensities), then γᵀd g =0.73, thus obtaining E g =0.675, indicating that there is significant observational perturbation in this grid. Therefore, the effective confidence score is calculated according to c. eff =c g ×(1−E g ) Calculate and obtain c eff =0.85×(1−0.675)≈0.276. Here, the convergence of the multiplication of ceff means that even if the image itself looks acceptable (c... g (Higher), as long as the drift state indicates strong uncertainty (E) g If the intensity is too high, the system will automatically converge to a more conservative prescription generation intensity, reducing the risk of over- or over-stimulation caused by misjudgment.

[0075] The system then calculates the activation benefit state A of the grid based on the corrected hair follicle feature vector. g With stimulus risk state R g For example, after drift correction, the mesh exhibits sparse and partially miniaturized activation signals, while the risk of inflammation / keratosis is moderate to low, resulting in A. g =0.75, R g =0.25. Substituting this into the dose budget yields B. g=18.63 μL / grid. The device converts the total budget for this grid into a point budget: if 16 points are planned within the grid and there is no point-level redistribution, the fixed dose per point is approximately 18.63 / 16 ≈ 1.16 μL / point; if subsequent heterogeneous gating trigger points differ, then B will still be used. g As the upper limit of the total number of grids, the allocation of points is not allowed to exceed this limit.

[0076] In this example, the delivery depth parameter is k. h =1.0mm, k eh =0.6mm, then h g =clip(0.895, 0.8, 2.2) = 0.895mm. Its engineering meaning is: although the benefit of this mesh is greater than the risk (A... g -R g (Positive) tends towards deeper stimuli, but due to higher uncertainty such as drift (E) g (Higher depth), the depth is automatically pulled back to a shallower depth to reduce the risk of excessive depth due to misregistration or leakage / hardening due to abnormal tissue state.

[0077] In this example, we take k as the delivery frequency parameter. n =2.0、k r =1.5, k e =1.0, then n g =clip(1.45, 1, 3)=1.45; When the device executes, it maps continuous values ​​to discrete levels (e.g., rounding to the nearest integer or dividing by threshold), thereby selecting a single delivery or a low-frequency scheme, which also reflects the constraint strategy of being more conservative as the drift increases.

[0078] After obtaining the preliminary prescription parameters for all grids, the system performs spatial regularization on the parameters (in this example, prioritizing the dose reference value B). g Regularization is used to suppress abrupt changes at grid boundaries. Specifically, the system constructs a smoothing objective based on grid adjacency relationships, penalizing dose differences between adjacent grids, and introducing c... eff As a credibility weight: c eff High-resolution grids have more confidence in themselves, c eff Lower-quality grids are more easily pulled back by their neighbors. For example, if the eight neighboring grids of a given grid exhibit consistent moderate demand and more stable observations in the same region (c... eff (Higher), its initial B g The values ​​are mostly between 22 and 26 μL, so the B of the mesh after spatial regularization is... g It can be gently increased from 18.65 μL to approximately 20 μL (still affected by B). max (and risk constraints), thereby avoiding band discontinuities in clinical implementation due to dose abrupt changes between adjacent cells.

[0079] Finally, the device sends the normalized prescription parameters to the execution end: the control terminal converts the target dose at each point into the number of step pulses of the dual-channel pump based on the pump calibration relationship (e.g., 200 pulses in channel A ≈ 1 μL, 200 pulses in channel B ≈ 1 μL); according to h g Set the lead screw depth mechanism to 0.895mm; press n g The gear selector controls the delivery rounds of the grid. During delivery, a pressure sensor monitors pipeline pressure changes in real time. If an abnormal rise exceeding a threshold occurs, it indicates a potential risk of needle blockage / backflow, and delivery at that point is automatically paused. Simultaneously, the point is marked as "execution abnormal" and written to the activation database. After delivery is complete, the c-axis of the grid is... g d g E g c eff A g R g B g h g n g The execution logs are also stored in the database for postoperative review and subsequent parameter calibration. This allows the multiplicative convergence concept of maximum allowable value × confidence level × requirement × (1 − risk) to be truly applied to executable delivery dose, depth, and frequency control under specific equipment constraints.

[0080] The various features and processes described above can be used independently of each other or can be combined in various ways. All possible combinations and sub-combinations are intended to fall within the scope of this disclosure. Furthermore, certain method or process blocks may be omitted in some embodiments. The methods and processes described herein are not limited to any particular order, and the blocks or states associated with them may be performed in other suitable orders. For example, the described blocks or states may be performed in an order different from the order specifically disclosed, or multiple blocks or states may be combined in a single block or state. Example blocks or states may be performed serially, in parallel, or in some other manner. Blocks or states may be added to or removed from the disclosed example embodiments. The exemplary systems and components described herein may be configured differently from those described. For example, elements may be added to, removed from, or rearranged compared to the disclosed example embodiments.

[0081] The various operations of the example methods described herein can be performed at least in part by an algorithm. This algorithm can be contained in program code or instructions stored in memory (e.g., the aforementioned non-transitory computer-readable storage medium). Such an algorithm may include a machine learning algorithm. In some embodiments, the machine learning algorithm may not be explicitly programmed into the computer to perform the function, but can learn from training data to create a predictive model that performs the function.

[0082] The various operations of the example methods described herein can be performed, at least in part, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors can constitute the engine of a processor implementation that operates to perform one or more of the operations or functions described herein.

[0083] Similarly, the methods described herein can be implemented at least in part by a processor, where one or more specific processors are examples of hardware. For example, at least some operations of a method can be performed by one or more processors or an engine implemented by a processor. Furthermore, one or more processors can also be operated to support the performance of related operations in a “cloud computing” environment or as “Software as a Service” (SaaS). For example, at least some operations can be performed by a set of computers (as an example of a machine including processors), where these operations are accessible via a network (e.g., the Internet) and via one or more suitable interfaces (e.g., application programming interfaces (APIs)).

[0084] The performance of certain operations can be distributed across processors, residing not only within a single machine but also deployed across multiple machines. In some example embodiments, the processor or processor-implemented engine may reside in a single geographic location (e.g., within a home environment, office environment, or server cluster). In other example embodiments, the processor or processor-implemented engine may be distributed across multiple geographic locations.

[0085] In this specification, multiple instances may implement components, operations, or structures described as single instances. Although individual operations of one or more methods are shown and described as separate operations, one or more of the separate operations may be performed simultaneously and do not need to be performed in the order shown. Structures and functions presented as separate components in the example configuration may be implemented as composite structures or components. Similarly, structures and functions presented as single components may be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of this document.

[0086] While an overview of the subject matter has been described with reference to specific example embodiments, various modifications and changes can be made to these embodiments without departing from the broader scope of embodiments of this disclosure. Such embodiments of the subject matter are referred to herein, individually or collectively, by the term "invention," and are used for convenience only and are not intended to limit the scope of this application to any single disclosure or concept, should more than one disclosure or concept be disclosed in fact.

[0087] The embodiments described herein have been described in sufficient detail to enable those skilled in the art to practice the disclosed teachings. Other embodiments may be used and derived therefrom, such that structural and logical substitutions and changes may be made without departing from the scope of this disclosure. Therefore, the detailed description should not be construed as limiting, and the scope of the various embodiments is defined only by the appended claims and the full scope of their equivalents.

Claims

1. A microneedle double-active factor hair transplantation hair follicle precise activation positioning system, characterized in that, Comprise: A grid drift correction module for grid processing of the operation area and obtaining a basic confidence and a drift factor vector of each grid unit, and correcting the original hair follicle feature vector of each grid unit based on the drift factor vector to obtain a corrected hair follicle feature vector; A demand fusion prescription module for analyzing the hair follicle activation demand state of each grid unit according to the corrected hair follicle feature vector, obtaining an effective confidence by combining the drift factor vector and the basic confidence of each grid unit, and then generating a preliminary grid-level dual activation factor prescription parameter, and performing spatial regularization processing based on the preliminary grid-level dual activation factor prescription parameter to obtain a final grid-level dual activation factor prescription parameter; A residual gate classification module for obtaining the local residual of each point in each grid unit relative to the grid unit mean value and constructing a point residual score, and obtaining a grid heterogeneity index based on the discrete degree of the point residual score and performing heterogeneity gating.

2. The micro-needle dual active factor hair follicle precise activation positioning system of claim 1, wherein: The specific process of grid processing of the operation area and obtaining the basic confidence and the drift factor vector of each grid unit is as follows: Multi-angle or multi-condition image acquisition is performed on the operation area, and an operation area coordinate system is established for subsequent grid division and data positioning; Based on the operation area coordinate system, the operation area is divided into grid units, each of which corresponds to a spatial partition for independent evaluation and processing; Image quality scores are obtained through the quality characteristics of the images corresponding to each grid unit, and the basic confidence reflecting the reliability and credibility of the observation data of each grid unit is obtained according to the image quality scores, and it is determined whether to perform resampling processing on the grid unit based on the basic confidence; Meanwhile, drift-related features are extracted from the images and clinical data corresponding to each grid unit and combined into a drift factor vector reflecting the current tissue state disturbance and observation uncertainty of each grid unit.

3. The micro-needle dual active factor hair follicle precise activation positioning system of claim 2, wherein: The specific determination process of determining whether to perform resampling processing on the grid unit based on the basic confidence is as follows: The basic confidence of the grid unit is compared with the lower limit threshold of the basic confidence; If the basic confidence of a certain grid unit is lower than the lower limit threshold of the basic confidence, it is determined to perform resampling processing on the grid unit; If the basic confidence of a certain grid unit is not lower than the lower limit threshold of the basic confidence, it is determined not to perform resampling processing on the grid unit.

4. The micro-needle dual active factor hair follicle precise activation positioning system of claim 1, wherein: The specific correction process of correcting the original hair follicle feature vector of each grid unit based on the drift factor vector is as follows: The corresponding relationship between the drift factor vector and the hair follicle feature drift is obtained through pre-calibration experiments or historical data analysis, and a drift compensation matrix that can set the correction weight of the original hair follicle feature under different drift factors is constructed; The original hair follicle feature vector of each grid unit is input into the drift compensation matrix, and the corrected hair follicle feature vector that can accurately reflect the true tissue state is obtained by combining the drift factor vector.

5. The micro-needle dual active factor hair follicle precise activation positioning system of claim 1, wherein: The specific analysis process of analyzing the hair follicle activation demand state of each grid unit according to the corrected hair follicle feature vector is as follows: Activation-related features are extracted from the corrected hair follicle feature vector of each grid unit, and the activation benefit state of each grid unit is obtained through normalization and weighted fusion; The pro-activation benefit state is used to reflect the potential benefit degree of the grid cell to the dual-active factor stimulation; Risk-related features are extracted from the corrected hair follicle feature vectors of each grid cell, and a stimulation risk state of each grid cell is obtained by normalization and weighted fusion. The stimulation risk state is used to represent the stimulation sensitivity and safety risk of the grid cell in the process of microneedle delivery and active factor injection.

6. The micro-needle dual active factor hair follicle precise activation positioning system of claim 1, wherein: The effective confidence is obtained based on the drift factor vector and the basic confidence of each grid cell, and a preliminary grid-level dual-active factor prescription parameter is generated, and the specific process is as follows: The effective confidence is used to dynamically adjust the generation strength of the grid-level dual-active factor prescription parameter. A preliminary grid-level dual-active factor prescription parameter is generated based on the effective confidence, the pro-activation benefit state, and the stimulation risk state of each grid cell. The preliminary grid-level dual-active factor prescription parameter includes the dose reference value, the delivery depth reference value, and the delivery frequency reference value of the grid cell. Meanwhile, spatial regularization processing is performed on all grid cell parameters based on the preliminary grid-level dual-active factor prescription parameter.

7. The micro-needle dual active factor hair follicle precise activation positioning system of claim 6, wherein: The specific processing process of the spatial regularization processing based on the preliminary grid-level dual-active factor prescription parameter is as follows: Based on the spatial adjacency relationship between the grid cells, a global parameter optimization model is constructed, a weighted difference term of adjacent grid dose reference values is introduced into the objective function, and the regularization weight is dynamically adjusted according to the effective confidence of each grid cell. When the set smoothing trigger condition is met, spatial regularization processing can also be performed on the delivery depth reference value and / or the delivery frequency reference value. The final grid-level dual-active factor prescription parameter is obtained after spatial regularization processing.

8. The micro-needle dual active factor hair follicle precise activation positioning system of claim 1, wherein: The specific process of obtaining the local residual of each point relative to the grid cell mean and constructing the point residual score is as follows: The feature mean of each grid cell is determined according to the feature data of all points in each grid cell, the feature data of each point is compared with the feature mean of the corresponding grid cell to which it belongs, and the local residual of each point relative to the grid cell mean is obtained. Activation-related residual components and risk-related residual components are extracted from the local residual, and the point residual score of each point in the grid cell is constructed based on the activation-related residual components and the risk-related residual components.

9. The micro-needle dual active factor hair follicle precise activation positioning system of claim 1, wherein: The specific process of obtaining the grid heterogeneity index based on the dispersion degree of the point residual score and performing heterogeneity gating is as follows: The grid heterogeneity index reflecting the fluctuation degree of residual distribution is determined based on the point residual score of all points in the grid cell. The heterogeneity gating is performed by comparing the grid heterogeneity index with a grid heterogeneity threshold. When the grid heterogeneity index is lower than the grid heterogeneity threshold, it is determined that no classification at the point level is needed. When the grid heterogeneity index is not lower than the grid heterogeneity threshold, it is determined that classification at the point level is needed.

10. The micro-needle dual active factor hair follicle precise activation positioning system of claim 9, wherein: The specific classification process of the point level classification is as follows: Based on the point residual score of each point in the grid cell, a hierarchical identification of each point is generated according to a point grading rule. The point grading rule is used to map each point to a point grading result for delivery parameter allocation. According to the point grading result, differentiated dual-active factor delivery parameter allocation is performed on each point in the grid cell.

Images