Blood concentration detection method

Abstract

The invention discloses a plasma concentration detection method, and belongs to the field of biomedical detection.The plasma concentration detection method comprises the steps that a saliva sample and a mesenchymal fluid sample of a to-be-detected object are synchronously collected through non-invasive equipment and minimally invasive equipment, and after centrifugation and membrane filtration pretreatment, the saliva sample and the mesenchymal fluid sample are obtained; a multi-channel biosensor is used for detecting drug metabolite concentrations in saliva and interstitial fluid respectively, a blood drug concentration predicted value is further calculated through a composite calibration model, the model is constructed based on a random forest algorithm, saliva and interstitial fluid concentration data are input, the blood drug concentration is output, and dynamic correction is conducted through species characteristic parameters. Wherein species parameters are obtained in real time through near infrared spectrum, weight coefficients (alpha and beta) are optimized in combination with a multiple regression equation, and finally high-precision prediction is achieved through a weighting algorithm. The method has the advantages of non-invasiveness, cross-species universality, dynamic adaptive correction and the like, and is suitable for clinical monitoring, veterinary medicine and pharmacokinetic research.

Description

Technical Field

[0001] The present invention belongs to the field of biomedical detection, and particularly relates to a method for detecting blood drug concentration. Background Art

[0002] Blood drug concentration detection is an important part of clinical pharmacology and individualized medication. Traditional detection methods mainly rely on invasive blood collection, which has disadvantages such as poor patient compliance, complex operation, and high infection risk. In recent years, non-invasive detection technologies (such as saliva detection) have gradually developed, but their accuracy is limited by the non-linear relationship between the concentrations of drug metabolites in body fluids and blood and individual physiological differences. In addition, existing technologies are mostly limited to a single species (such as humans), and it is difficult to meet the needs of veterinary medicine or cross-species pharmacokinetic studies. Summary of the Invention

[0003] In view of the above pain points, the present invention provides a method for detecting blood drug concentration. By synchronously collecting saliva and interstitial fluid samples, and after pretreatment, using a multi-channel biosensor to detect the concentration of drug metabolites, and combining with a composite calibration model for dynamic weighted calculation, high-precision cross-species blood drug concentration prediction is achieved.

[0004] The solution of the present invention is as follows:

[0005] A method for detecting blood drug concentration, characterized by comprising the following steps:

[0006] S1. Synchronously collect saliva samples and interstitial fluid samples of the object to be measured using non-invasive devices and minimally invasive devices; the non-invasive device includes a saliva collection swab, and the minimally invasive device includes a microdialysis probe and an interstitial fluid patch, and the object to be measured includes humans, dogs, cats, pigs, and other mammals;

[0007] S2. Pretreat the saliva samples and interstitial fluid samples to remove impurities and isolate the target analytes;

[0008] S3. Simultaneously detect the first concentration value of the drug metabolite in the processed saliva sample and the second concentration value of the drug metabolite in the interstitial fluid sample through a multi-channel biosensor;

[0009] S4. Input the first concentration value and the second concentration value into a preset composite calibration model, and calculate the corresponding blood drug concentration prediction value through a weighted algorithm; the composite calibration model is established based on a database of the corresponding relationships between blood drug concentrations and saliva and interstitial fluid drug concentrations of different species.

[0010] Preferably, the pretreatment includes centrifugal separation at a rotation speed of 3000 - 5000 rpm for 5 - 8 minutes, and membrane filtration treatment with a pore size of 0.22 - 0.45 μm.

[0011] Preferably, the multi-channel biosensor adopts chemiluminescence immunoassay, comprising:

[0012] 3-1. Alkaline phosphatase-labeled probe for saliva samples;

[0013] 3-2. Horseradish peroxidase-labeled probe for interstitial fluid samples;

[0014] 3-3. Dual-wavelength (450nm±10nm and 630nm±10nm) photoelectric detection module.

[0015] Preferably, the method for establishing the composite calibration model includes:

[0016] 4-1. Collect time-series blood, saliva, and interstitial fluid samples from different species after drug administration;

[0017] 4-2. Determine the actual and accurate value of the drug concentration in each sample by high performance liquid chromatography-mass spectrometry;

[0018] 4-3. A nonlinear mapping model of saliva-interstitial fluid concentration and blood drug concentration was constructed using the random forest algorithm, as follows:

[0019] Assume that the input variable is the concentration of drug metabolites in saliva samples C saliva and interstitial fluid sample drug metabolite concentration C interstitial , the output variable is the blood drug concentration C plasma ;

[0020] The random forest consists of m decision trees. For each decision tree T i (i=1,2,…,m), randomly extract n groups of data from the database with replacement as the training set; at each node, all the features of the input variables (i.e., C saliva and C interstitial ) randomly select k features for optimal division to determine the branching rules of the decision tree;

[0021] For a new sample input x=(c saliva ,c interstitial ), each decision tree T i Will give a predicted value y based on its own branching rules i , the final random forest prediction value y rf is the average value of all decision tree predictions, and the calculation formula is:

[0022]

[0023] 4-4. The model is calibrated using species-specific parameters including body surface area (BSA), metabolic rate (BMR), and epidermal thickness (EKT). Model parameter calibration is required to be performed every 24 hours.

[0024] Preferably, the species characteristic parameters are obtained by real-time monitoring through near-infrared spectroscopy with a wavelength range of 900 - 1700 nm, including:

[0025] 5 - 1. The body surface area BSA is calculated by scanning the outer contour of the object to be measured through near-infrared spectroscopy and combining the geometric parameters of body length and body width, where BSA = k1 × M 0.67 , M is the body weight, and k1 is an empirical coefficient; or the standard body surface area of the target species is matched through a preset cross-species body type database;

[0026] 5 - 2. The basal metabolic rate BMR is indirectly deduced based on the water content in the dermis layer and subcutaneous fat density detected by near-infrared spectroscopy, combined with the species-specific basal metabolism formula, where BMR = k2 × M a , k2 and a are species-specific coefficients;

[0027] 5 - 3. The epidermal keratin layer thickness EKT is calculated by analyzing the reflection signal of near-infrared spectroscopy within the detection depth range of 0.1 - 0.3 mm to calculate the thickness value of the epidermal keratin layer.

[0028] Preferably, the model correction is specifically as follows:

[0029] Define the comprehensive characteristic parameter vector S = (BSA, BMR, EKT), and correct the prediction result of the random forest model according to the species characteristic parameters monitored in real time; let the correction coefficient be γ(S), which is a function of S and is obtained through multiple linear regression equations for the data in the database:

[0030] γ(S) = a0 + a1 × BSA + a2 × BMR + a3 × EKT

[0031] where a0, a1, a2, a3 are regression coefficients; the model parameters are corrected once every 24 hours, and the original database is expanded using the newly collected data within these 24 hours, and then the regression coefficients a0, a1, a2, a3 are refitted using the least squares method; the corrected predicted value y corrected is the random forest predicted value multiplied by the correction coefficient, that is:

[0032] y corrected = γ(S) × y rf .

[0033] Preferably, the corresponding predicted value of blood drug concentration calculated by the weighted algorithm includes the following content:

[0034] Based on the corrected predicted value y corrected , the predicted value of blood drug concentration C plasma is calculated using the weighted algorithm, and the calculation formula is:

[0035] C plasma = α·C saliva + β·C interstitial

[0036] wherein, the weighting coefficients α and β are determined by the following rules:

[0037] Effect of body surface area BSA on the weighting coefficients: The body surface area affects the volume of drug distribution. Through regression analysis, it is determined that α is negatively correlated with the body surface area and β is positively correlated. Let α = f1(BSA) and β = f2(BSA), and the specific functional forms are obtained through polynomial regression fitting.

[0038] f1(BSA) = b0 + b1×BSA + b2×BSA 2 ,

[0039] f2(BSA) = c0 + c1×BSA + c2×BSA 2 ,

[0040] where b0, b1, b2, c0, c1, c2 are regression coefficients;

[0041] Effect of metabolic rate BMR on the weighting coefficients: The metabolic rate affects the drug clearance rate. By adjusting the weights of the metabolic parameters in the model, α and β are optimized. Let the adjustment functions of α and β with respect to the metabolic rate be g1(BMR) and g2(BMR) respectively:

[0042] g1(BMR) = d0 + d1×BMR,

[0043] g2(BMR) = e0 + e1×BMR,

[0044] where d0, d1, e0, e1 are regression coefficients;

[0045] Effect of epidermal keratin layer thickness EKT on the weighting coefficients: The epidermal keratin layer thickness affects the interstitial fluid penetration delay. By correcting the osmotic kinetics model, the time-dependent weight of β is adjusted. Let the adjustment function of β with respect to the epidermal keratin layer thickness and time t be h(EKT, t);

[0046] Finally, considering all factors, α and β are:

[0047] α = α0×f1(BSA)×g1(BMR)

[0048] β = β0×f2(BSA)×g2(BMR)×h(EKT, t)

[0049] where α0 and β0 are initial weight coefficients, which are determined by cross-validation; by obtaining the species characteristic parameters of the object to be measured in real time, the values of α and β are dynamically optimized, and then the predicted value C of the blood drug concentration is obtained.plasma .

[0050] Compared with the prior art, the advantages of the present invention are:

[0051] (1) High-precision composite calibration model: Based on the random forest algorithm, a nonlinear mapping relationship between saliva-interstitial fluid and blood drug concentration is constructed, and the model is dynamically corrected through species characteristic parameters (body surface area, metabolic rate, etc.), significantly improving the prediction accuracy;

[0052] (2) Cross-species universality: covering humans and mammals such as dogs, cats, and pigs, and combining near-infrared spectroscopy to obtain species physiological parameters in real time, achieve model adaptive adjustment, and broaden application scenarios;

[0053] (3) Efficient pretreatment and detection technology: Standardized centrifugation and membrane filtration processes and dual-wavelength multi-channel biosensors (chemiluminescence immunoassay) are used to shorten the detection cycle and improve detection sensitivity and specificity. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Schematic diagram of the process flow for the blood drug concentration detection method. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0056] Example: This example uses cats as the test object

[0057] S1. Sample collection.

[0058] A healthy adult short-haired cat weighing M = 4 kg and fasting was selected as the test subject. A saliva sample was first collected using a non-invasive device: a customized soft-bristled saliva collection swab (consistent with the oral physiology of cats) was used to gently rub the cat's cheek mucosa for 10 seconds to collect 0.2 mL of saliva, avoiding contamination with food residue or blood.

[0059] Minimally invasive sample collection was performed simultaneously: a microdialysis probe (0.3mm outer diameter, 0.2mm implantation depth) was minimally invasively implanted subcutaneously in the hairless area between the shoulder blades. The probe was covered with an interstitial fluid patch to absorb the infiltrated interstitial fluid. A 0.15mL interstitial fluid sample was collected within 30 minutes. Strict aseptic procedures were followed to avoid injury to the animal.

[0060] Real-time monitoring of species characteristic parameters:

[0061] Body surface area (BSA): The whole body contour of the cat was scanned by a near-infrared spectrometer (wavelength range 900 - 1700 nm) to obtain geometric parameters such as body length (40 cm from the tip of the nose to the base of the tail) and body width (15 cm at the widest part of the scapula); the special formula for felines BSA = k1×M was used 0.67 , where the empirical coefficient k1 = 0.00886 (calibrated based on the feline body size database), and BSA = 0.00886×4 was calculated 0.67 ≈0.072 m 2 ;

[0062] Basal metabolic rate (BMR): The water content in the dermis layer (60%) and subcutaneous fat density (0.95 g / cm 3 ) were detected by near-infrared spectroscopy, and combined with the basic metabolic formula for cats BMR = 70×M 0.75 (in line with the metabolic rules of mammals, and the species-specific coefficient was verified by experiments), and BMR = 70×4 was calculated 0.75 ≈

[0063] 138.6 kcal / day;

[0064] Epidermal keratin layer thickness (EKT): Through the analysis of the reflection signal attenuation characteristics of near-infrared spectroscopy in the shallow layer of the skin (depth 0.1 - 0.3 mm), a thickness calculation model was established based on the Lambert-Beer law, and EKT = 0.06 mm was measured, reflecting the barrier function state of the cat's back skin.

[0065] S2, Sample pretreatment.

[0066] The saliva and interstitial fluid samples were respectively transferred into 1.5 mL centrifuge tubes and centrifuged at 4000 rpm for 6 min in a refrigerated centrifuge (temperature 4°C); during centrifugation, heavy impurities such as cell debris and mucus precipitated to the bottom of the tube, and the supernatant contained the target analytes (drug metabolites); subsequently, negative pressure filtration was carried out using a hydrophilic polyethersulfone filter membrane with a pore size of 0.3 μm to remove residual fine particles and protein polymers, ensuring that the clarity of the sample met the requirements of subsequent detection; the pretreated samples were transferred to sterile test tubes and immediately subjected to concentration detection or cryopreserved (-20°C).

[0067] S3, Concentration detection.

[0068] A multi-channel chemiluminescence immunoassay system was used to process the two samples in separate channels:

[0069] Saliva sample detection: Add 50 μL of the pretreated saliva sample to the detection cup, add a drug metabolite-specific probe labeled with alkaline phosphatase (ALP) (probe concentration 10 μg / mL, optimized for the structure of feline drug metabolites), incubate at 37 °C for 15 min to form an antigen-antibody-enzyme labeled complex; after the reaction, add a chemiluminescent substrate (4-methylumbelliferyl phosphate, MUP), and detect the fluorescence signal through a dual-wavelength photoelectric detection module (excitation light 450 nm ± 10 nm, emission light 630 nm ± 10 nm), and obtain the first concentration value C of the drug metabolite in saliva through standard curve fitting saliva = 12.5 ng / mL;

[0070] Interstitial fluid sample detection: Take 30 μL of the interstitial fluid sample, add a homologous probe labeled with horseradish peroxidase (HRP) (probe concentration 15 μg / mL, adapted to the complex matrix environment of interstitial fluid), incubate and react under the same conditions, and measure the second concentration value C through the same detection module interstitial = 15.2 ng / mL; The specificity of the two probes was verified by cross-reaction to ensure that the detection results are not interfered by the sample matrix.

[0071] S4. Plasma drug concentration.

[0072] 1. Establishment of a composite calibration model:

[0073] 1-1. Construction of historical data: In the early stage, 80 cats were collected for blood (collected through the jugular vein), saliva, and interstitial fluid samples at 0.5, 1, 2, 4, 6, and 8 h after oral administration of a certain drug (such as an antibiotic). The actual plasma drug concentration value (accuracy ±0.1 ng / mL) was determined by high-performance liquid chromatography-mass spectrometry (HPLC-MS) technology, and a time series database containing C saliva 、C interstitial 、C plasma was established to cover the entire process of drug absorption, distribution, metabolism, and excretion; [[ID=2**]]

[0074] 1-2. Random forest model training: The model contains 40 decision trees. Each tree randomly and with replacement selects 60 groups of data from the database as the training set through the bootstrap method, and the remaining 20 groups are used for verification. When dividing at each node, all features are randomly selected from 2 input features (C saliva 、C interstitial ) for optimal division (Gini index as the splitting criterion) to ensure the diversity and generalization ability of the model; for a newly input sample x = (c saliva , c interstitial ), each tree outputs a predicted value y i , and finally the average is taken to obtain:

[0075]

[0076] 1-3. Species Characteristic Correction: Define a comprehensive eigenvector S = (BSA, BMR, EKT), where body surface area (BSA) directly affects the volume of drug distribution in the body. A larger body surface area results in a wider distribution of the drug and potentially slower equilibrium between blood and body fluid concentrations. BMR reflects liver metabolism and renal excretion capacity. A higher BMR indicates faster drug clearance and a more complex dynamic equilibrium between drug concentrations in body fluids and blood. Epidermal stratum corneum thickness (EKT) influences the penetration efficiency of interstitial fluid collection. Increased EKT thickness may cause interstitial fluid concentrations to lag behind blood concentrations.

[0077] The correction coefficient γ(S) is established by multiple linear regression:

[0078] γ(S)=a0+a1×BSA+a2×BMR+a3×EKT

[0079] The regression coefficients are a0 = 1.02, a1 = 0.15, a2 = -0.0003, and a3 = 0.4 (obtained through least squares fitting of historical data; the model prediction error was reduced by 18% after correction). The positive correlation term for BSA is: the larger the body surface area, the more dispersed the drug distribution, and the predicted value needs to be slightly adjusted upward to correct for the influence of distribution volume. The negative correlation term for BMR is: the higher the metabolic rate, the faster the drug clearance, and the predicted value needs to be slightly adjusted downward with increasing metabolic rate. The positive correlation term for EKT is: when the stratum corneum is thicker, the interstitial fluid drug concentration may be low, and the correction coefficient increases with increasing thickness.

[0080] 1-4. Real-time correction calculation:

[0081] Substitute the characteristic parameters of the cat in this embodiment into:

[0082] BSA=0.072m 2 , BMR=138.6kcal / day, EKT=0.06mm

[0083] γ(S)=1.02+0.15×0.072-0.0003×138.6+0.4×0.06=0.99

[0084] That is, the corrected prediction value is 0.99 times that of the random forest prediction value, reflecting the targeted adjustment of the physiological characteristics of cats;

[0085] The corrected predicted value is: corrected =γ(S)×y rf .

[0086] 2. Dynamic optimization of weighted algorithm:

[0087] The weighted algorithm integrates saliva (C saliva) and interstitial fluid (C interstitial ) concentration, dynamically assign weights in combination with species characteristics, and the core logic is as follows:

[0088] 2-1. Influence of body surface area (BSA) on weight:

[0089] The body surface area is positively correlated with the drug distribution volume. The larger the distribution volume, the less the drug concentration in saliva may be affected by tissue binding, and the weight of interstitial fluid should increase relatively because it is closer to blood vessels. Obtained by polynomial regression fitting:

[0090] α = f1(BSA) = b0 + b1×BSA + b2×BSA 2 = 0.5 - 0.4×BSA + 0.2×BSA 2

[0091] β = f2(BSA) = c0 + c1×BSA + c2×BSA 2 = 0.5 + 0.3×BSA - 0.1×BSA 2

[0092] For this example, BSA = 0.072m 2 , it is calculated that α decreases with the increase of BSA (negative correlation), and β increases with the increase of BSA (positive correlation), which conforms to the physiological mechanism of "the larger the body surface area, the higher the weight of interstitial fluid";

[0093] 2-2. Influence of metabolic rate (BMR) on weight:

[0094] The metabolic rate directly affects the drug clearance rate. A high metabolic rate will accelerate the distribution and clearance of drugs from blood to body fluids, and the weights need to be dynamically adjusted to reflect the differences in clearance kinetics. The linear regression function is defined as:

[0095] g1(BMR) = d0 + d1×BMR = 1 - 0.0005×BMR

[0096] The weight of saliva decreases with the increase of metabolic rate;

[0097] g2(BMR) = e0 + e1×BMR = 1 + 0.0006×BMR

[0098] The weight of interstitial fluid increases with the increase of metabolic rate;

[0099] In this example, BMR = 138.6 kcal / day, and it is calculated that g1 = 0.93 and g2 = 1.08, indicating that at a high metabolic rate, the contribution of interstitial fluid concentration to blood drug concentration is higher, which conforms to the law of "interstitial fluid is closer to blood dynamics during fast clearance";

[0100] 2-3. Time-dependent correction of epidermal keratin layer thickness (EKT):

[0101] For cats, the EKT is relatively thin (0.06 mm), and the interstitial fluid penetration delay is negligible. Therefore, the time-dependent function h(EKT, t) is not enabled for the time being, and only basic weight correction is adopted. The initial weights α0 = β0 = 0.5 are determined through cross-validation to ensure the minimization of the prediction error of the model on the training set and the validation set;

[0102] 2-4. Final calculation:

[0103] Considering the above factors, the weighting coefficient is the product of the initial weight and each influence function:

[0104] α = α0 × f1(BSA) × g1(BMR) = 0.5 × 0.47 × 0.93 = 0.22

[0105] β = β0 × f2(BSA) × g2(BMR) = 0.5 × 0.53 × 1.08 = 0.29

[0106] Finally, calculate the predicted value C of the blood drug concentration plasma :

[0107] C plasma = α·C saliva + β·C interstitial = 0.22 × 12.5 + 0.29 × 15.2 = 6.46 ng / mL

[0108] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A method for detecting blood drug concentration, characterized in that, It includes the following steps: S1. Synchronously collect saliva samples and interstitial fluid samples of the object to be measured by using non-invasive devices and minimally invasive devices; the non-invasive device includes a saliva collection swab, and the minimally invasive device includes a microdialysis probe and an interstitial fluid patch. The object to be measured includes humans, dogs, cats, pigs, and other mammals; S2. Pretreat the saliva samples and interstitial fluid samples to remove impurities and isolate the target analytes; S3. Simultaneously detect the first concentration value of the drug metabolite in the pretreated saliva sample and the second concentration value of the drug metabolite in the interstitial fluid sample by using a multi-channel biosensor; S4. Input the first concentration value and the second concentration value into a preset composite calibration model, and calculate the corresponding predicted blood drug concentration value through a weighted algorithm; The composite calibration model is established based on a database of the corresponding relationships between blood drug concentrations and saliva and interstitial fluid drug concentrations of different species.

2. The blood drug detection method according to claim 1, wherein The pretreatment includes centrifugal separation at a rotation speed of 3000 - 5000 rpm for 5 - 8 minutes and membrane filtration treatment with a pore size of 0.22 - 0.45 μm.

3. The blood drug detection method according to claim 1, wherein The multi-channel biosensor uses chemiluminescence immunoassay and includes: 3-1. An alkaline phosphatase-labeled probe for saliva samples; 3-2. A horseradish peroxidase-labeled probe for interstitial fluid samples; 3-3. A dual-wavelength (450 nm ± 10 nm and 630 nm ± 10 nm) photoelectric detection module.

4. A blood drug detection method according to claim 1, characterized in that, The method for establishing the composite calibration model includes: 4-1. Collect sequential blood, saliva, and interstitial fluid samples of different species after drug administration; 4-2. Determine the actual accurate values of the drug concentrations in each sample by using high-performance liquid chromatography-mass spectrometry; 4-3. Use the random forest algorithm to construct a non-linear mapping model between saliva-interstitial fluid concentration and blood drug concentration, specifically as follows: Let the input variables be the concentration C of the drug metabolite in the saliva sample saliva and the concentration C of the drug metabolite in the interstitial fluid sample interstitial , and the output variable be the blood drug concentration C plasma ; A random forest consists of m decision trees. For each decision tree T i (i = 1, 2, …, m), n groups of data are randomly sampled with replacement from the database as the training set; at each node, k features are randomly selected from all the features of the input variables (i.e., C saliva and C interstitial ) for optimal partitioning to determine the branching rules of the decision tree; For the new sample input x = (c saliva , c interstitial ), each decision tree T i will give a predicted value y i according to its own branching rules. Finally, the predicted value y rf of the random forest is the average of all decision tree predicted values, and the calculation formula is: 4-4. Calibrate the model through species characteristic parameters including body surface area BSA, metabolic rate BMR, and epidermal thickness EKT, and require model parameter calibration to be performed once every 24 hours.

5. The blood drug detection method according to claim 4, characterized in that, The species characteristic parameters are obtained through real-time monitoring by near-infrared spectroscopy with a wavelength range of 900 - 1700 nm and include: 5-1. The body surface area BSA is calculated by scanning the outer contour of the object to be measured through near-infrared spectroscopy and combining the geometric parameters of body length and body width, where BSA = k1 × M 0.67 , M is the body weight, and k1 is an empirical coefficient; or the standard body surface area of the target species is matched through a preset cross-species body size database; 5-2. The metabolic rate BMR is indirectly derived based on the water content in the dermis layer detected by near-infrared spectroscopy, the subcutaneous fat density, and combined with the species-specific basal metabolism formula, where BMR = k2 × M a , and k2 and a are species-specific coefficients; 5-3. The epidermal stratum corneum thickness EKT is calculated by analyzing the reflection signals of near-infrared spectroscopy within a detection depth range of 0.1 - 0.3 mm to obtain the thickness value of the epidermal stratum corneum.

6. A blood drug detection method according to claim 4, characterized in that, The specific model calibration is as follows: Define the comprehensive characteristic parameter vector S = (BSA, BMR, EKT), and correct the prediction result of the random forest model according to the species characteristic parameters monitored in real time; let the correction coefficient be γ(S), which is a function of S and is obtained through a multiple linear regression equation for the data in the database: γ(S) = a0 + a1×BSA + a2×BMR + a3×EKT where a0, a1, a2, and a3 are regression coefficients; the model parameters are corrected every 24 hours. The original database is augmented with the newly collected data within these 24 hours, and then the least squares method is used again to fit the regression coefficients a0, a1, a2, and a3; the predicted value y after correction corrected is the predicted value of the random forest multiplied by the correction coefficient, that is: y corrected = γ(S) × y rf .

7. A blood drug detection method according to claim 1, characterized in that, The calculation of the corresponding predicted blood drug concentration value through the weighted algorithm includes the following content: Based on the corrected predicted value y corrected , a weighted algorithm is used to calculate the predicted value C of the blood drug concentration plasma , and the calculation formula is: C plasma = α·C saliva + β·C interstitial Among them, the weighting coefficients α and β are determined by the following rules: Effect of body surface area (BSA) on the weighting coefficients: Body surface area affects the volume of drug distribution. Through regression analysis, it is determined that α is negatively correlated with body surface area and β is positively correlated. Let α = f1(BSA) and β = f2(BSA), and the specific functional forms are obtained through polynomial regression fitting. f1(BSA) = b0 + b1×BSA + b2×BSA 2 , f2(BSA) = c0 + c1×BSA + c2×BSA 2 , where b0, b1, b2, c0, c1, c2 are regression coefficients; Effect of basal metabolic rate (BMR) on the weighting coefficients: Basal metabolic rate affects the drug clearance rate. By adjusting the weights of the metabolic parameters in the model, α and β are optimized. Let the adjustment functions of α and β with respect to basal metabolic rate be g1(BMR) and g2(BMR) respectively: g1(BMR) = d0 + d1 × BMR, g2(BMR) = e0 + e1 × BMR, where d0, d1, e0, e1 are regression coefficients; Effect of epidermal keratin thickness (EKT) on the weighting coefficients: Epidermal keratin thickness affects the interstitial fluid penetration delay. By correcting the osmotic kinetic model, the time-dependent weight of β is adjusted. Let the adjustment function of β with respect to epidermal keratin thickness and time t be h(EKT, t); Finally, considering all factors, α and β are: α = α0 × f1(BSA) × g1(BMR) β = β0 × f2(BSA) × g2(BMR) × h(EKT, t) where α0 and β0 are initial weight coefficients determined by cross-validation; by obtaining the species characteristic parameters of the object to be measured in real time, the values of α and β are dynamically optimized, and then the predicted value C of the blood drug concentration is obtained plasma 。

Images