Standardized management method, device and equipment for clinical pathway of flora transplantation and medium

Abstract

The invention relates to a flora transplantation clinical pathway standardization management method, device and equipment and a medium. The method comprises the following steps: on the basis of clinical data of a patient and screening data of a donor, executing flora suitability evaluation by using a deep learning model and selecting an adaptive donor; obtaining a flora sample of an adaptive donor, and carrying out high-throughput sequencing; generating a donor flora quality evaluation score according to a preset flora composition diversity index based on a sequencing result; screening the flora samples of which the quality evaluation scores are higher than a preset quality threshold value to prepare a flora preparation; detecting activity parameters and diversity parameters of the flora preparation, and performing colonization potential classification prediction by combining clinical data of the patient and adopting a support vector machine algorithm so as to output a potential colonization success rate; and when the potential colonization success rate exceeds a preset success rate threshold, outputting a clinical transplantation scheme through a preset transplantation scheme generation algorithm based on the flora preparation and the clinical data of the patient. According to the method, through multi-stage intelligent decision making, the safety and the effectiveness are improved.

Description

Technical Field

[0001] This invention belongs to the fields of biomedical engineering and medical artificial intelligence, and in particular relates to standardized management methods, devices, equipment and media for clinical pathways of microbiome transplantation. Background Technology

[0002] With the development of microbiome medicine and precision medicine technologies, fecal microbiota transplantation (FMT) has gradually become an effective means of treating diseases related to intestinal microbiota imbalance. This technology transplants the functional microbiota of a healthy donor into the patient's gut, achieving microbiota reconstruction and disease intervention. Currently, clinical practice mainly adopts an experience-driven, manual decision-making model, the core process of which typically includes: manually matching donor-patient pairs, screening samples based on limited indicators (such as microbiota concentration), standardized preparation of microbiota preparations, and performing the transplantation procedure according to clinical guidelines.

[0003] However, existing technologies have the following drawbacks: Donor suitability assessment is subjective, traditionally relying on physician experience to match donors and patients, lacking quantitative correlation analysis between multidimensional clinical data (such as host immune status and gut microenvironment) and donor microbiota function (such as metabolic genes and beneficial bacteria abundance), leading to insufficient transplantation accuracy; Microbiota quality control is one-sided, with sample screening relying solely on a single diversity indicator (such as the Shannon index), neglecting key parameters such as microbiota activity and functional gene integrity, resulting in a disconnect between formulation quality assessment results and clinical colonization effects, and a lack of colonization potential prediction. Traditional methods cannot quantitatively assess the colonization success rate of microbiota preparations in specific patients before transplantation, and postoperative effects rely on follow-up verification, easily leading to ineffective transplantation and waste of medical resources; Dynamic adjustment of protocols is lagging, with clinical transplantation protocols rigidly implemented, lacking a dynamic optimization mechanism based on real-time activity parameters and patient response data, making it difficult to cope with colonization fluctuations caused by individual differences. Summary of the Invention

[0004] Therefore, it is necessary to provide a standardized management method, device, equipment, and medium for clinical pathways of microbiome transplantation that can solve the above problems.

[0005] Firstly, this application provides a standardized management method for clinical pathways in microbiome transplantation, including:

[0006] Based on the patient's clinical data and the donor screening data, a deep learning model was used to perform microbial compatibility assessment, and a suitable donor was selected based on the microbial compatibility assessment results.

[0007] Obtain bacterial community samples from suitable donors and perform high-throughput sequencing on the bacterial community samples to obtain sequencing results;

[0008] Based on the sequencing results, a donor microbial community quality assessment score is generated according to a preset microbial community composition diversity index.

[0009] Microbial samples with quality assessment scores higher than a preset quality threshold are selected, and the selected microbial samples are processed using formulation equipment to obtain microbial formulations.

[0010] The activity and diversity parameters of the microbial preparation were detected, and based on the activity and diversity parameters and the patient's clinical data, the colonization potential was classified and predicted using the support vector machine algorithm, and the potential colonization success rate was output.

[0011] When the potential colonization success rate exceeds the preset success rate threshold, a clinical transplantation plan is output based on the activity and diversity parameters of the microbial preparation and the patient's clinical data, using a preset transplantation plan generation algorithm.

[0012] In one embodiment, based on the patient's clinical data and donor screening data, a deep learning model is used to perform microbiome fit assessment, and a suitable donor is selected based on the microbiome fit assessment results, including:

[0013] Multi-dimensional feature extraction is performed on the patient's clinical data to generate a first feature matrix, which includes disease type, gut environment parameters and immune status indicators.

[0014] Dynamic weight analysis was performed on the donor screening data to generate a second feature matrix, which includes the distribution of microbial abundance, metabolic functional genes and donor health rating.

[0015] The first feature matrix and the second feature matrix are fused by a convolutional neural network to output a suitability score.

[0016] Select donors whose fit score is higher than the preset fit threshold as the selected fit donors.

[0017] In one embodiment, based on sequencing results, a donor microbial community quality assessment score is generated according to a preset microbial community composition diversity index, including:

[0018] Based on the sequencing results, several pre-defined microbial community diversity indicators were calculated, including the Shannon diversity index, the Simpson diversity index, the microbial community evenness index, and the abundance ratio of core beneficial bacteria genera.

[0019] The pre-defined gradient boosting decision tree model is used to perform weighted aggregation calculations on the diversity indices of multiple bacterial communities, and output the quality assessment score of the donor bacterial community.

[0020] In one embodiment, based on activity parameters and diversity parameters, combined with the patient's clinical data, a support vector machine algorithm is used to perform colonization potential classification prediction, outputting the potential colonization success rate, including:

[0021] Time series normalization is performed on the activity parameters to generate an activity feature vector;

[0022] The diversity parameters are weighted using the entropy weighting method to generate a diversity feature vector.

[0023] Principal component dimensionality reduction decomposition is performed on the patient's clinical data to generate clinical feature vectors;

[0024] The activity feature vector, diversity feature vector and clinical feature vector are combined into tensors to construct a multidimensional heterogeneous feature space;

[0025] Using the support vector machine algorithm, the multidimensional heterogeneous feature space is mapped to a high-dimensional space through the radial basis kernel function;

[0026] Solve for the optimal decision hyperplane parameters in high-dimensional space, and calculate the classification decision function value based on the optimal decision hyperplane parameters;

[0027] The classification decision function value is used to generate the potential colonization success rate through a probability transformation function.

[0028] In one embodiment, after generating the clinical transplantation protocol, the process further includes:

[0029] Extract the execution elements of a clinical transplantation protocol;

[0030] The execution elements, activity and diversity parameters of the microbial preparation, and clinical data are recombined into a feature vector to be predicted according to the preset conversion rules.

[0031] The predicted transplantation effect is obtained by performing transplantation effect prediction calculation on the feature vector to be predicted using a preset prediction model.

[0032] In one embodiment, after the clinical transplantation protocol is executed, the method further includes:

[0033] Obtain clinical data from patients after the implementation of clinical transplant protocols;

[0034] Multi-dimensional feature extraction is performed on clinical data after the implementation of clinical transplantation protocols to generate a third feature matrix;

[0035] The deviation between the third feature matrix and the predicted transplantation effect is calculated to generate a deviation index.

[0036] When the deviation index exceeds the preset deviation threshold, the execution elements of the clinical transplantation plan are adjusted based on the preset correction rules. The execution elements include at least one of the following: the dosage of the microbial preparation, the transplantation frequency, and the adjuvant therapy drugs.

[0037] In one embodiment, after the clinical transplantation protocol is generated but before its execution, the method further includes:

[0038] Acquire patients' clinical data before microbiota transplantation and perform tensor fusion with the parameter information of microbiota preparations and the execution elements of clinical transplantation protocols to generate a risk feature set;

[0039] A risk feature set is generated by processing the pre-set transplantation decision model, and the operational feasibility score and immune compatibility score are output.

[0040] If the feasibility score is lower than the preset operation threshold, or the immune compatibility score is lower than the preset compatibility threshold, a transplantation blocking instruction and alarm event log will be generated.

[0041] Secondly, this application also provides a standardized management device for clinical pathways of microbiota transplantation, comprising:

[0042] The donor matching assessment module is used to perform microbiome compatibility assessment based on the patient's clinical data and the donor screening data, using a deep learning model, and select a suitable donor based on the microbiome compatibility assessment results.

[0043] The sample acquisition and sequencing module is used to acquire bacterial community samples suitable for donors and perform high-throughput sequencing on the bacterial community samples to obtain sequencing results.

[0044] The sample quality scoring module is used to generate a donor microbial community quality assessment score based on sequencing results and according to preset microbial community composition diversity indicators.

[0045] The microbial community screening preparation module is used to screen out microbial community samples with quality assessment scores higher than a preset quality threshold, and to use preparation equipment to prepare microbial community samples to obtain microbial community preparations.

[0046] The colonization potential prediction module is used to detect the activity and diversity parameters of the microbial preparation, and based on the activity and diversity parameters, combined with the patient's clinical data, it uses the support vector machine algorithm to perform colonization potential classification prediction and output the potential colonization success rate.

[0047] The transplantation protocol generation module is used to output a clinical transplantation protocol based on the activity and diversity parameters of the microbial preparation and the patient's clinical data, using a preset transplantation protocol generation algorithm when the potential colonization success rate exceeds a preset success rate threshold.

[0048] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-mentioned standardized management method for the clinical pathway of microbiota transplantation.

[0049] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, performs the steps of the above-described standardized management method for clinical pathways of microbiome transplantation.

[0050] The aforementioned standardized management methods, devices, equipment, and media for the clinical pathway of microbiome transplantation utilize deep learning models to perform microbiome fit assessment. Based on patient clinical data and donor screening data, quantitative analysis is performed to overcome the subjective limitations of traditional manual experience-based matching. This achieves intelligent fusion and fit scoring of multidimensional characteristics of donors and patients, significantly improving donor fit accuracy. A quality assessment score is generated based on high-throughput sequencing results and pre-set microbiome composition diversity indicators. High-quality microbiome samples are screened comprehensively using multi-dimensional parameters, addressing the problem of one-sided quality control due to reliance on a single indicator. After preparing the microbiome preparation, by detecting activity and diversity parameters and combining them with patient clinical data, a support vector machine algorithm is used to perform colonization potential classification prediction to output the potential colonization success rate. This compensates for the lack of colonization potential prediction in traditional methods, avoiding ineffective transplantation and resource waste. When the potential colonization success rate exceeds a pre-set success rate threshold, a pre-set transplantation protocol generation algorithm outputs a clinical transplantation protocol, achieving standardized protocol generation based on real-time data and intelligent decision-making. This alleviates the problem of lag in dynamic protocol adjustment and improves the safety and effectiveness of microbiome transplantation. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart of the standardized management method for clinical pathways of microbiome transplantation according to the present invention;

[0053] Figure 2 This is a structural diagram of the standardized management device for clinical pathways of microbiome transplantation according to the present invention;

[0054] Figure 3 This is a structural diagram of a standardized management device for clinical pathways of microbiome transplantation in one embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] In one embodiment, such as Figure 1As shown, a standardized management method for clinical pathways in microbiome transplantation is provided. This embodiment illustrates the application of this method to a terminal, but it is understood that the method can also be applied to a server, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In the implementation environment, the hardware architecture includes a medical terminal (such as a hospital clinical workstation, equipped with input devices and a display for collecting patient clinical data and donor screening data), a server (integrated with a GPU accelerator for running deep learning models and support vector machine algorithms, processing microbiome fitness assessment, quality scoring, and colonization potential prediction calculations), a formulation device (an automated microbiome sample preparation unit for standardizing the screened samples to generate microbiome formulations), and a high-throughput sequencing device (connected to the server for sequencing result analysis). In this application scenario, when a medical institution needs to perform microbiota transplantation on patients with gut microbiota imbalance, the terminal collects the patient's clinical data and donor screening data, and transmits them to the server via the network. The server calls a deep learning model to fuse the abundance distribution of donor microbiota and metabolic gene data, generates a suitability score and a quality assessment score, and then feeds it back to the terminal to guide the preparation of microbiota preparations. Simultaneously, the server performs colonization potential prediction based on activity and diversity parameters. If the success rate exceeds a threshold, it outputs a clinical transplantation plan to the terminal, guiding physicians to execute the plan and making dynamic adjustments (such as adjusting the transplantation frequency based on postoperative deviation indicators), achieving intelligent decision-making throughout the entire process and improving transplantation safety and effectiveness. In this embodiment, the method includes the following steps:

[0057] S01, based on the patient's clinical data and the donor screening data, uses a deep learning model to perform microbiome fit assessment, and selects a suitable donor based on the microbiome fit assessment results.

[0058] The clinical data encompasses patient disease type, gut environment parameters, and immune status indicators. Screening data includes donor microbiota abundance distribution, metabolic function genes, and health rating. Fit assessment quantifies the degree of match between the patient's clinical status and the donor microbiota functional characteristics using deep feature fusion technology. In implementation, multi-dimensional feature extraction is performed on the patient's clinical data to generate a first feature matrix representing the host microenvironment state; simultaneously, dynamic weight analysis is performed on the donor screening data to generate a second feature matrix representing microbiota function. The two matrices are fused using convolutional neural networks: convolutional kernels perform cross-matrix convolution operations in the feature dimension space to extract deep features such as the implicit association between the patient's immune status and donor metabolic genes, and the synergistic patterns between gut environment parameters and microbiota abundance distribution, outputting a quantitative fit score. The selection process automatically selects the optimal donor based on a preset fit threshold (e.g., score > 0.82), replacing the traditional experience-driven manual matching model and transforming fit assessment from subjective experience-based judgment to objective quantitative decision-making.

[0059] S02, obtain a bacterial community sample suitable for the donor, and perform high-throughput sequencing on the bacterial community sample to obtain the sequencing results.

[0060] After selecting a suitable donor, a gut microbiota sample (usually a sterile fecal sample) is obtained from the donor and high-throughput sequencing is performed. This can be done using metagenomic sequencing or 16S rRNA gene sequencing technology, and the sample DNA is sequenced in large-scale parallel using a next-generation sequencing platform such as Illumina NovaSeq to generate raw data (FASTQ format) containing microbial species composition, gene function, and metabolic pathways. High-throughput sequencing is a digital technique for fully characterizing the composition and function of the microbial community. The sequencing results refer to a standardized sequence dataset after quality control (including removal of low-quality reads and chimera filtering), providing a multi-dimensional omics foundation for subsequent quality assessment modules (covering key parameters such as diversity index calculation, core genus abundance analysis, and functional gene integrity verification). This addresses the problem of one-sided sample quality assessment caused by traditional methods relying on single culture or microscopic examination, and provides a quantitative input source of microbial composition diversity indicators for gradient boosting decision tree models.

[0061] S03 generates a donor microbial community quality assessment score based on sequencing results and a preset microbial community composition diversity index.

[0062] Based on the standardized sequence dataset generated by high-throughput sequencing, pre-defined microbial community diversity indicators (such as Shannon diversity index, Simpson diversity index, microbial community evenness index, and abundance ratio of core beneficial bacteria genera) are calculated. These microbial community diversity indicators are a set of quantitative parameters reflecting the stability and functional integrity of the microbial community structure. The quality assessment score refers to the comprehensive evaluation score generated by non-linear weighted aggregation of the above indicators through a gradient boosting decision tree model (GBDT). In implementation, an indicator training set containing historical high-quality microbial community samples is constructed. The GBDT model autonomously learns the implicit association between each indicator and the clinical colonization effect (e.g., the weight coefficient of the abundance of core beneficial bacteria genera on the transplantation success rate reaches 0.62). The diversity indicators of the current donor sample are weighted according to feature importance, and the output quality assessment score ranges from 0 to 1. This score comprehensively characterizes the functional robustness and host adaptation potential of the microbial community preparation. Compared with the traditional single-indicator screening method, it improves the detection rate of high-quality samples and provides an objective quantitative basis for subsequent preparation.

[0063] S04: Select bacterial samples with quality assessment scores higher than the preset quality threshold, and use formulation equipment to prepare and process the selected bacterial samples to obtain bacterial formulations.

[0064] Based on the quality assessment score, sample screening is performed using a preset quality threshold (a dynamically configured threshold value according to different disease types, such as >0.85 for inflammatory bowel disease and >0.78 for metabolic syndrome). After screening out microbial samples with scores higher than the threshold, the samples are standardized using automated preparation equipment (including centrifugation, anaerobic environment maintenance, and lyophilization protectant addition units). For example, sample heterogeneity is eliminated by homogenization with buffer, active microorganisms are enriched by centrifugation under nitrogen protection (centrifugation force 800×g for 10 minutes), trehalose-glycerol composite lyophilization protectant is added, and the samples are deep-frozen at -80℃ to generate a microbial preparation that meets clinical transplantation standards.

[0065] S05, detect the activity and diversity parameters of the microbial preparation, and based on the activity and diversity parameters, combined with the patient's clinical data, use the support vector machine algorithm to perform colonization potential classification prediction and output the potential colonization success rate.

[0066] The study involved detecting the activity parameters (a set of dynamic indicators characterizing the metabolic activity of the microbial community, such as the survival rate of the microbial community and the time series of ATP concentration) and diversity parameters (static indicators reflecting the stability of the microbial community structure, including the proportion of core beneficial bacteria genera and species evenness) of the microbial community preparation. The activity parameters were normalized over time to generate an activity feature vector, and the diversity parameters were weighted using the entropy weighting method to generate a diversity feature vector. Simultaneously, principal component analysis was performed on the patient's clinical data to generate a clinical feature vector. These feature vectors were then used to construct a multidimensional heterogeneous feature space through tensor combination. The Support Vector Machine (SVM) algorithm was used to map this space to a high-dimensional space using the radial basis function kernel function. The optimal decision hyperplane parameters were solved, and the classification decision function value was calculated. The potential colonization success rate (a quantitative probability value, ranging from 0 to 1, of successful colonization of the microbial community preparation in the patient's gut) was output through a probability transformation function. The safety and effectiveness of microbial community transplantation were improved through high-dimensional decision boundary optimization using SVM.

[0067] S06, When the potential colonization success rate exceeds the preset success rate threshold, a clinical transplantation plan is output based on the activity and diversity parameters of the microbial preparation and the patient's clinical data, using a preset transplantation plan generation algorithm.

[0068] Specifically, when the potential colonization success rate exceeds a preset success rate threshold, the algorithm (a hybrid decision-making system based on a rule engine and machine learning model) generates a transplantation protocol based on the activity and diversity parameters of the microbial preparation and the patient's clinical data. This algorithm performs tensor fusion of the microbial preparation characteristics and clinical data to construct a multidimensional feature space. It learns the mapping relationship of historical successful transplantation cases (such as the nonlinear relationship between microbial activity and dosage ratio) through a gradient boosting decision tree model to generate a clinical transplantation protocol (a set of execution elements including microbial preparation dosage, transplantation frequency, and adjuvant therapy drug combination). Through intelligent closed-loop decision-making throughout the entire process, the accuracy of protocol adaptation is improved, and the resource waste and safety risks caused by the rigid execution of traditional protocols are resolved.

[0069] The aforementioned standardized management method for clinical pathways in microbiome transplantation utilizes deep learning models to perform microbiome suitability assessments based on patient clinical data and donor screening data to select suitable donors. This addresses the subjective limitations of traditional manual experience-based matching, achieving intelligent fusion of multidimensional characteristics between donors and patients. After obtaining microbiome samples from suitable donors and performing high-throughput sequencing, a donor microbiome quality assessment score is generated based on the sequencing results and preset microbiome composition diversity indicators. High-quality microbiome samples are comprehensively screened through multidimensional parameters, overcoming the one-sided microbiome quality control problem caused by relying on a single indicator. Microbiome preparations are prepared using formulation equipment, and their activity and diversity parameters are detected. Combined with patient clinical data, a support vector machine algorithm is used to perform colonization potential classification and prediction to output the potential colonization success rate, compensating for the lack of colonization potential prediction in traditional methods and avoiding ineffective transplantation. When the potential colonization success rate exceeds a preset success rate threshold, a clinical transplantation plan is output based on microbiome preparation parameters and patient data through a preset transplantation plan generation algorithm. This achieves dynamic optimization decision-making based on real-time activity and diversity parameters, dynamically adjusting the solution to address lag issues and comprehensively improving the safety and effectiveness of microbiome transplantation.

[0070] In one embodiment, based on the patient's clinical data and donor screening data, a deep learning model is used to perform microbiome fit assessment, and a suitable donor is selected based on the microbiome fit assessment results, including:

[0071] S11 performs multi-dimensional feature extraction on the patient's clinical data to generate a first feature matrix, which includes disease type, intestinal environment parameters and immune status indicators.

[0072] S12, Perform dynamic weight analysis on the donor screening data to generate a second feature matrix, which includes the distribution of microbial abundance, metabolic function genes and donor health rating.

[0073] S13, the first feature matrix and the second feature matrix are fused by a convolutional neural network to output the fit score;

[0074] S14, select donors whose fit score is higher than the preset fit threshold as selected fit donors.

[0075] For example, multi-dimensional feature extraction is performed on the patient's clinical data to generate a first feature matrix containing disease type, intestinal environment parameters, and immune status indicators (e.g., multi-dimensional feature extraction is performed on the patient's clinical data (preoperative data, including disease type (e.g., ICD-10 coding), intestinal environment parameters (pH value: detection method refers to GB / T 27404-2008 "Laboratory Quality Control Standard for Food Physicochemical Testing", sample dilution ratio 1:10), immune status indicators (proportion of regulatory T cells: flow cytometry detection, instrument model BD FACSCanto II)), and a 32-dimensional first feature matrix is ​​generated through One-Hot coding (disease type) and Z-score standardization (continuous indicators). The comprehensive state of the host microenvironment is quantitatively characterized through feature engineering. Dynamic weight analysis is performed on the donor screening data to generate a second feature matrix containing microbial abundance distribution, metabolic function genes, and donor health rating (donor screening data (microbial abundance distribution: 16S)). rRNA sequencing annotation results, metabolic function genes: KEGG pathway annotation completeness, donor health rating: according to the "Expert Consensus on Donor Screening and Management in Chinese Microbiome Transplantation (2023 Edition)" classification) Dynamic weight analysis was performed—based on the gradient descent algorithm of 1200 historical successful transplantation cases, the weights of each feature (microbiome abundance distribution weight 0.42, metabolic function gene weight 0.38, donor health rating weight 0.20) were calculated, generating a 24-dimensional second feature matrix), where the dynamic weight analysis adaptively adjusted the contribution weight of each microbiome functional feature based on historical transplantation effect data. A convolutional neural network (CNN) with three layers fusing two matrices, configured as follows: convolutional layer 1 (3×3 kernels, 64 elements, ReLU activation), convolutional layer 2 (3×3 kernels, 128 elements, ReLU activation), and a fully connected layer (output dimension 1, Sigmoid activation); optimized using cross-entropy loss function to output a suitability score in the 0-1 range. This fusion of the two matrices utilizes convolutional kernels to perform cross-matrix convolution operations in the feature dimension space, extracting deep features such as the implicit association between patient immune status and donor metabolic genes, and the synergistic patterns between gut environmental parameters and microbiome abundance distribution, outputting a quantified suitability score. During the selection phase, the optimal donor is selected based on a preset suitability threshold (e.g., score > 0.82), replacing the traditional experience-driven manual matching method. When treating patients with inflammatory bowel disease, the first feature matrix extracts their disease subtype (Crohn's disease), intestinal pH (6.8), fecal calprotectin concentration (320 μg / g), and the proportion of regulatory T cells (12%); the second feature matrix dynamically weights and calculates the abundance of Bacteroidetes phylum (32%), the integrity of short-chain fatty acid synthesis gene clusters (0.91), and the donor health index (Grade A).CNN extracts features in the immune-metabolic cross dimension of the dual matrix using 3×3 convolution kernels. It found that the matching weight between the patient's low Treg ratio and the donor's high butyrate synthesis gene reached 0.76, generating an fitness score of 0.85, which exceeded the threshold and was selected as a suitable donor.

[0076] In one embodiment, based on sequencing results, a donor microbial community quality assessment score is generated according to a preset microbial community composition diversity index, including:

[0077] S21, based on sequencing results, calculates several preset bacterial community composition diversity indicators, including Shannon diversity index, Simpson diversity index, bacterial community evenness index, and abundance ratio of core beneficial bacterial genera.

[0078] S22 uses a pre-defined gradient boosting decision tree model to perform weighted aggregation calculations on the calculated diversity indicators of multiple microbial communities, and outputs a donor microbial community quality assessment score.

[0079] Specifically, the sequencing results are standardized sequence datasets generated by high-throughput sequencing. After quality control (including removal of low-quality reads and chimera filtering), they are used to quantitatively analyze the microbial community composition. Preset microbial community diversity indicators include the Shannon diversity index (which quantifies species richness and evenness, calculated as -∑(p...). i ·ln(p i )), where p i The relative abundance of the i-th species), and the Simpson diversity index (which measures the concentration of dominant species and is calculated as 1 - ∑(p i 2The evenness index of the bacterial community, E = H / lnS (reflecting the evenness of species distribution, calculated as Shannon index divided by ln(number of species)), and the abundance ratio of core beneficial bacterial genera (e.g., core beneficial bacterial genera: Bifidobacterium, Lactobacillus, Faecalibacterium, calculating the proportion of their total abundance to the total bacterial population). Based on this sequencing result, these diversity indicators can be calculated in parallel using bioinformatics tools (such as QIIME2 or mothur). For example, after performing OTU clustering and species annotation on the sample sequences, a Python script can be used to automatically generate numerical matrices for each indicator. Gradient Boosting Decision Tree (GBDT) is an ensemble machine learning algorithm. It is initialized using a historical high-quality microbial community training set (containing indicator data from ≥1000 successful clinical colonization cases). Parameters are set as follows: 100 trees, tree depth 8, learning rate 0.1, and mean squared error (MSE) loss function. Through feature importance analysis, the weights of each indicator are determined (abundance ratio of core beneficial bacteria genera 0.42, Shannon index 0.25, Simpson index 0.20, and evenness index 0.13). After weighted aggregation, a quality assessment score in the 0-1 range is output. This score comprehensively characterizes the functional robustness and host adaptability potential of the microbial community formulation (e.g., a score >0.85 indicates a high-quality sample that can be directly used for formulation preparation). This addresses the limitations of traditional methods that rely on a single indicator.

[0080] In one embodiment, based on activity parameters and diversity parameters, combined with the patient's clinical data, a support vector machine algorithm is used to perform colonization potential classification prediction, outputting the potential colonization success rate, including:

[0081] S31.1 Perform time series normalization on the activity parameters to generate an activity feature vector;

[0082] S31.2, perform entropy weighting calculation on the diversity parameters to generate a diversity feature vector;

[0083] S31.3 Perform principal component dimensionality reduction decomposition on the patient's clinical data to generate clinical feature vectors;

[0084] S32 combines the activity feature vector, diversity feature vector and clinical feature vector into tensors to construct a multidimensional heterogeneous feature space;

[0085] S33 utilizes the support vector machine algorithm to map the multidimensional heterogeneous feature space to a high-dimensional space through the radial basis kernel function;

[0086] S34, Solve for the optimal decision hyperplane parameters in high-dimensional space, and calculate the classification decision function value based on the optimal decision hyperplane parameters;

[0087] S35, the classification decision function value is used to generate the potential colonization success rate through a probability transformation function.

[0088] For example, the activity parameters are a set of dynamic indicators characterizing the metabolic activity of the microbial preparation (including microbial survival rate (detected by plate counting method, culture medium is TSA medium, anaerobic culture at 37℃ for 48h), ATP concentration (luciferin-luciferase method, detection kit No. Thermo Fisher A22066, detection wavelength 560nm)). These parameters are normalized over time (e.g., Z-score normalization is performed on the 24h activity parameter time series, converting the time series data of each activity parameter into a normalized value with a mean of 0 and a standard deviation of 1, and extracting the main fluctuation characteristics through principal component analysis) to generate an activity feature vector. The diversity parameters are static indicators reflecting the stability of the microbial community structure (including the proportion of core beneficial bacteria genera and species evenness). These parameters are calculated using an entropy weighting method (based on information entropy theory, the information entropy value of each diversity parameter is calculated, and weights are assigned according to the entropy value; the entropy value is...). Weight is For example, when the entropy weight of the core beneficial bacteria genus is high, the weight is set to 0.4; when the entropy weight of the evenness is low, the weight is set to 0.2. This generates a diversity feature vector. The patient's clinical data (including disease type, immune indicators, and intestinal environment parameters) undergoes principal component analysis (using the principal component analysis algorithm, retaining the top k principal components with a cumulative variance contribution rate > 85% to reduce data dimensionality; the default k = 5) to generate a clinical feature vector. The activity feature vector, diversity feature vector, and clinical feature vector are combined using tensors (e.g., using tensor operations in the NumPy library to construct a three-dimensional feature space matrix) to construct a multidimensional heterogeneous feature space. Using the Support Vector Machine (SVM) algorithm, through the radial basis function (K(x...)... i x j )=exp(-γ||x i -x j || 2 The kernel function parameter γ is set to 0.1, mapping the multidimensional heterogeneous feature space to a high-dimensional space. In the high-dimensional space, the optimal decision hyperplane parameters are solved (using a sequential minimum optimization algorithm to optimize the Lagrange multipliers and solve for the margin-maximizing hyperplane, such as calculating the hyperplane's normal vector and bias term). The classification decision function value is then calculated based on the optimal decision hyperplane parameters (based on the decision function f(x) = sign(∑α). i y i K(x i ,x)+b), where α i For support vector weights, y i(For category labels); The classification decision function value is mapped to the [0,1] interval using a probability transformation function (using the Platt scaling method, the decision function value is mapped to the [0,1] interval, the formula is P(x=1|x)=1 / (1+exp(Af(x)+B)), where A and B are model training parameters, optimized based on 1000 labeled samples, A=-0.82, B=0.35), and the potential colonization success rate is output. (That is, the probability that the microbial preparation will successfully colonize the patient's intestine, ranging from 0 to 1), and this success rate is used for subsequent transplantation decisions (e.g., triggering protocol generation when the success rate > 0.75).

[0089] In one embodiment, after generating the clinical transplantation protocol, the process further includes:

[0090] S41, Extract the execution elements of the clinical transplantation protocol;

[0091] S42, according to the preset conversion rules, reorganize the execution elements, the activity parameters and diversity parameters of the microbial preparation and the clinical data into a feature vector to be predicted;

[0092] S43, the transplantation effect prediction calculation is performed on the feature vector to be predicted by the preset prediction model to obtain the predicted transplantation effect.

[0093] Specifically, the execution elements are the core components of the clinical transplantation protocol (including the dosage of microbial agents, transplantation frequency, and combination of adjuvant therapy drugs, such as dosage units of CFU / mL and frequency of weekly doses). These elements are automatically extracted from the protocol text using parsing algorithms (such as a regular expression-based element extractor). The feature vector to be predicted is generated by using preset transformation rules (the rule engine is based on a predefined logical mapping of a historical successful case database, such as mapping dosage values ​​to a normalized interval of [0,1]. Activity parameters and diversity parameters are combined with clinical data for tensor recombination. Tensor combination operations are performed using the NumPy library to generate multidimensional feature vectors and construct a high-dimensional data structure). The preset prediction model is a machine learning model trained using the Gradient Boosting Decision Tree (GBDT) algorithm (the model is initialized using a historical transplantation case dataset, with a tree depth of 10 and a learning rate of 0.05). This model performs transplantation effect prediction calculations on the input feature vector to be predicted (calculating predicted values ​​through forward propagation, such as outputting the colonization improvement rate or symptom relief probability) to obtain the predicted transplantation effect (a quantitative score in the range of 0-1, representing the expected clinical efficacy).

[0094] In one embodiment, after the clinical transplantation protocol is executed, the method further includes:

[0095] S51, Obtain clinical data of patients after the implementation of clinical transplantation protocols;

[0096] S52 performs multi-dimensional feature extraction on clinical data after the implementation of the clinical transplantation protocol, generating a third feature matrix;

[0097] S53, calculate the deviation between the third feature matrix and the predicted transplantation effect to generate a deviation index;

[0098] S54. When the deviation index exceeds the preset deviation threshold, the execution elements of the clinical transplantation plan are adjusted based on the preset correction rules. The execution elements include at least one of the following: the dosage of the microbial preparation, the transplantation frequency, and the adjuvant therapy drug.

[0099] For example, clinical data of patients after the implementation of a clinical transplantation protocol is acquired, which is a real-time postoperative monitoring dataset collected through a medical terminal (such as an electronic health record system) (including symptom scores, inflammatory biomarkers such as fecal calprotectin concentration, and gut microbiota colonization detection results); multi-dimensional feature extraction is performed on the above clinical data (using feature engineering algorithms, such as dimensionality reduction techniques based on principal component analysis, to extract key dimensions such as symptom relief rate, immune response strength, and gut microbiota remodeling efficiency), generating a third feature matrix (a multi-dimensional numerical array, where rows represent time point sequences and columns represent standardized feature vectors); the deviation between the third feature matrix and the predicted transplantation effect (i.e., the expected efficacy score previously generated by the prediction model) is calculated (using the Euclidean distance algorithm to calculate the deviation value between the actual feature vector and the predicted vector in the feature space, the formula is...). Generate a deviation index (a value ranging from 0 to 1, with higher values ​​indicating a more significant deviation from expectations); when the deviation index exceeds a preset deviation threshold (e.g., a threshold of 0.3, based on statistical analysis of 200 postoperative cases, this threshold corresponds to a clinical efficacy difference ≥20% significance level, P<0.05), adjust the execution elements of the clinical transplantation protocol based on preset correction rules (predefined logic in the rule engine, such as an if-else decision tree), including the dosage of the gut microbiota preparation (e.g., if the deviation >0.3 and the colonization rate <50%, increase the dosage by 10^8 CFU / mL (based on the pharmacokinetic model, this amount can increase the gut microbiota concentration by 15%). 20%, and no safety risks); if the colonization rate is >90% but the symptoms do not improve, reduce the dose by 5×10^7 CFU / mL;), transplantation frequency (e.g., if the deviation is >0.3 and the colonization duration is <7 days, adjust the transplantation frequency from twice a week to three times a week (based on the gut microbiota colonization half-life data, increasing the frequency can extend the colonization time to 10-14 days) and adjuvant therapy drugs (e.g., if the deviation is >0.3 and accompanied by immune rejection (e.g., IL-6 concentration >10 pg / mL), add at least one of the following: mesalazine, dose 1g / time, 3 times a day, refer to the "Consensus on Diagnosis and Treatment of Inflammatory Bowel Disease (2022)").

[0100] In one embodiment, after the clinical transplantation protocol is generated but before its execution, the method further includes:

[0101] S61: Obtain the patient's clinical data before microbiota transplantation and perform tensor fusion with the parameter information of microbiota preparations and the execution elements of the clinical transplantation protocol to generate a risk feature set.

[0102] S62 generates a risk feature set by processing a preset transplantation decision model, and outputs an operational feasibility score and an immune compatibility score.

[0103] S63. If the operation feasibility score is lower than the preset operation threshold, or the immune compatibility score is lower than the preset compatibility threshold, then a transplantation blocking instruction and alarm event log will be generated.

[0104] Specifically, clinical data (including heart rate, blood pressure, and immune indicators (T cell subset ratio)) within 24 hours before transplantation are obtained from the patient and combined with microbiome preparation parameters (activity parameters, dosage) and protocol execution elements (transplantation pathway: such as gastroscopy / colonoscopy) through tensor fusion (3×12 dimensions) to generate a risk feature set, characterizing the host-microbiome-protocol interaction risk. The pre-defined transplantation decision model is a machine learning model trained using the random forest algorithm (the model was trained on 1800 transplantation cases (including 300 cases of risk events), with 200 trees, a tree depth of 12, a feature subset ratio of 0.7, and an optimization metric of F1 score (optimal F1 = 0.91)). When processing the risk feature set, multi-task learning is performed: a feasibility score (quantifying the safety of the surgical procedure; feasibility score (0-1): includes 3 dimensions: instrument compatibility (e.g., the matching degree between the endoscope channel diameter and the preparation delivery tube, weight 0.4); patient position tolerance (based on BMI classification, weight 0.3); and estimated operation time (<30 minutes is considered excellent, weight 0.3)) and an immunocompatibility score (quantifying the risk of host immune response). The scale is 0-1, containing two dimensions (donor-patient HLA typing matching, weight 0.6; preoperative IL-10 concentration (≥5 pg / mL is preferred), weight 0.4); preset operation thresholds and compatibility thresholds (operation feasibility threshold 0.6 (operation failure rate ≥35% when below this value), immune compatibility threshold 0.55 (rejection rate ≥40% when below this value)). When the operation feasibility score is lower than the operation threshold (e.g., obstruction of endoscopic transplantation pathway) or the immune compatibility score is lower than the compatibility threshold (e.g., high rejection risk), a transplantation blocking command is generated (terminating the preparation delivery device operation via API interface) and an alarm event log is generated (including timestamp, risk feature set snapshot and decision model output, stored in the hospital event management system), effectively avoiding clinical risks.

[0105] The standardized management method for the aforementioned microbiome transplantation clinical pathway utilizes a deep learning model to perform microbiome fit assessment and select suitable donors based on patient clinical data and donor screening data. This overcomes the subjective limitations of traditional manual experience-based matching, achieving quantitative fusion of patient disease type, intestinal environment parameters, and immune status indicators with donor microbiome abundance distribution and metabolic function genes, thus improving fit accuracy. After obtaining microbiome samples from suitable donors and performing high-throughput sequencing, a gradient boosting decision tree model is used to generate donor microbiome quality assessment scores based on the sequencing results and preset microbiome composition diversity indicators (such as Shannon diversity index, Simpson diversity index, microbiome evenness index, and abundance ratio of core beneficial bacteria genera). High-quality samples are screened through multi-dimensional parameter weighted aggregation, addressing the problem of one-sided microbiome quality control caused by relying on a single indicator. The activity and diversity parameters of the microbiome preparation are detected, and a support vector machine algorithm is used to perform colonization potential assessment in conjunction with patient clinical data. Force classification prediction maps the tensor combination of activity feature vectors, diversity feature vectors, and clinical feature vectors to a high-dimensional space to solve for the optimal decision hyperplane, outputting the potential colonization success rate for preoperative quantitative assessment, avoiding ineffective transplantation and resource waste. When the potential colonization success rate exceeds a preset success rate threshold, a clinical transplantation plan is generated based on the activity parameters and diversity parameters of the microbial preparation and patient data through a preset transplantation plan generation algorithm. After the plan is generated, execution elements (such as microbial preparation dosage, transplantation frequency, and adjuvant therapy drugs) are extracted and recombined into a feature vector to be predicted for transplantation effect prediction calculation. After execution, patient clinical data is acquired and deviation indicators are generated through deviation calculation to dynamically adjust plan elements. The solution dynamically adjusts for lag issues. By generating a risk feature set before plan execution, the feasibility score and immunocompatibility score are output, and a transplantation blocking command is generated to avoid risks. Thus, through multi-stage intelligent closed-loop decision-making, the safety and effectiveness of microbial transplantation are significantly improved.

[0106] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0107] Based on the same inventive concept, this application also provides a microbiota transplantation clinical pathway standardization management device for implementing the aforementioned standardized management method for microbiota transplantation clinical pathways. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more embodiments of the microbiota transplantation clinical pathway standardization management device provided below can be found in the limitations of the microbiota transplantation clinical pathway standardization management method described above, and will not be repeated here.

[0108] In one exemplary embodiment, such as Figure 2 As shown, a standardized management device for clinical pathways of microbiota transplantation is provided, comprising:

[0109] The donor matching assessment module 101 is used to perform microbiome compatibility assessment based on the patient's clinical data and the donor screening data, using a deep learning model, and to select a suitable donor based on the microbiome compatibility assessment results.

[0110] The sample acquisition and sequencing module 102 is used to acquire microbial community samples of the suitable donor and perform high-throughput sequencing on the microbial community samples to obtain sequencing results.

[0111] The sample quality scoring module 103 is used to generate a donor microbial community quality assessment score based on the sequencing results and according to a preset microbial community composition diversity index.

[0112] The microbial community screening preparation module 104 is used to screen out microbial community samples with quality assessment scores higher than a preset quality threshold, and to use preparation equipment to prepare and process the screened microbial community samples to obtain microbial community preparations.

[0113] The colonization potential prediction module 105 is used to detect the activity parameters and diversity parameters of the microbial preparation, and based on the activity parameters and diversity parameters, combined with the patient's clinical data, it uses the support vector machine algorithm to perform colonization potential classification prediction and output the potential colonization success rate.

[0114] The transplantation protocol generation module 106 is used to output a clinical transplantation protocol based on the activity and diversity parameters of the microbial preparation and the patient's clinical data when the potential colonization success rate exceeds the preset success rate threshold, through a preset transplantation protocol generation algorithm.

[0115] In one embodiment, the donor adaptation evaluation module 101 is further configured to:

[0116] Multi-dimensional feature extraction is performed on the patient's clinical data to generate a first feature matrix, which includes disease type, gut environment parameters and immune status indicators.

[0117] Dynamic weight analysis was performed on the donor screening data to generate a second feature matrix, which includes the distribution of microbial abundance, metabolic functional genes and donor health rating.

[0118] The first feature matrix and the second feature matrix are fused by a convolutional neural network to output a suitability score.

[0119] Select donors whose fit score is higher than the preset fit threshold as the selected fit donors.

[0120] In one embodiment, the sample quality scoring module 103 is further configured to:

[0121] Based on the sequencing results, several pre-defined microbial community diversity indicators were calculated, including the Shannon diversity index, the Simpson diversity index, the microbial community evenness index, and the abundance ratio of core beneficial bacteria genera.

[0122] The pre-defined gradient boosting decision tree model is used to perform weighted aggregation calculations on the diversity indices of multiple bacterial communities, and output the quality assessment score of the donor bacterial community.

[0123] In one embodiment, the colonization potential prediction module 105 is further configured to:

[0124] Time series normalization is performed on the activity parameters to generate an activity feature vector;

[0125] The diversity parameters are weighted using the entropy weighting method to generate a diversity feature vector.

[0126] Principal component dimensionality reduction decomposition is performed on the patient's clinical data to generate clinical feature vectors;

[0127] The activity feature vector, diversity feature vector and clinical feature vector are combined into tensors to construct a multidimensional heterogeneous feature space;

[0128] Using the support vector machine algorithm, the multidimensional heterogeneous feature space is mapped to a high-dimensional space through the radial basis kernel function;

[0129] Solve for the optimal decision hyperplane parameters in high-dimensional space, and calculate the classification decision function value based on the optimal decision hyperplane parameters;

[0130] The classification decision function value is used to generate the potential colonization success rate through a probability transformation function.

[0131] In one embodiment, a scheme prediction and adjustment module 107 is further included, for:

[0132] Extract the execution elements of a clinical transplantation protocol;

[0133] The execution elements, activity and diversity parameters of the microbial preparation, and clinical data are recombined into a feature vector to be predicted according to the preset conversion rules.

[0134] The predicted transplantation effect is obtained by performing transplantation effect prediction calculation on the feature vector to be predicted using a preset prediction model.

[0135] In one embodiment, the scheme prediction adjustment module 107 is further configured to:

[0136] Obtain clinical data from patients after the implementation of clinical transplant protocols;

[0137] Multi-dimensional feature extraction is performed on clinical data after the implementation of clinical transplantation protocols to generate a third feature matrix;

[0138] The deviation between the third feature matrix and the predicted transplantation effect is calculated to generate a deviation index.

[0139] When the deviation index exceeds the preset deviation threshold, the execution elements of the clinical transplantation plan are adjusted based on the preset correction rules. The execution elements include at least one of the following: the dosage of the microbial preparation, the transplantation frequency, and the adjuvant therapy drugs.

[0140] In one embodiment, the scheme prediction adjustment module 107 is further configured to:

[0141] Acquire patients' clinical data before microbiota transplantation and perform tensor fusion with the parameter information of microbiota preparations and the execution elements of clinical transplantation protocols to generate a risk feature set;

[0142] A risk feature set is generated by processing the pre-set transplantation decision model, and the operational feasibility score and immune compatibility score are output.

[0143] If the feasibility score is lower than the preset operation threshold, or the immune compatibility score is lower than the preset compatibility threshold, a transplantation blocking instruction and alarm event log will be generated.

[0144] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the standardized management method for the clinical pathway of microbiota transplantation as described above.

[0145] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0146] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0147] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A standardized management method for clinical pathways of microbiome transplantation, characterized in that, The method includes: Based on the patient's clinical data and the donor screening data, a deep learning model is used to perform microbial compatibility assessment, and a suitable donor is selected based on the microbial compatibility assessment results. Obtain a microbial community sample from the adapted donor and perform high-throughput sequencing on the microbial community sample to obtain sequencing results; Based on the sequencing results, a donor microbial community quality assessment score is generated according to a preset microbial community composition diversity index. Microbial samples with quality assessment scores higher than a preset quality threshold are selected, and the selected microbial samples are prepared using a formulation device to obtain a microbial formulation. The activity and diversity parameters of the bacterial culture preparation were detected, and based on the activity and diversity parameters and the patient's clinical data, a support vector machine algorithm was used to perform colonization potential classification and prediction, and the potential colonization success rate was output. When the potential colonization success rate exceeds a preset success rate threshold, a clinical transplantation plan is output based on the activity and diversity parameters of the microbial preparation and the patient's clinical data, using a preset transplantation plan generation algorithm.

2. The method according to claim 1, characterized in that, The process of using a deep learning model to perform microbial compatibility assessment based on patient clinical data and donor screening data, and selecting a suitable donor based on the microbial compatibility assessment results, includes: Multi-dimensional feature extraction is performed on the patient's clinical data to generate a first feature matrix, which includes disease type, intestinal environment parameters and immune status indicators. Dynamic weight analysis is performed on the donor screening data to generate a second feature matrix, which includes microbial abundance distribution, metabolic function genes and donor health rating. The first feature matrix and the second feature matrix are fused by a convolutional neural network to output a suitability score. Select donors whose fit score is higher than the preset fit threshold as the selected fit donors.

3. The method according to claim 1, characterized in that, The step of generating a donor microbial community quality assessment score based on the sequencing results and according to a preset microbial community composition diversity index includes: Based on the sequencing results, several preset bacterial community composition diversity indices are calculated, including the Shannon diversity index, the Simpson diversity index, the bacterial community evenness index, and the abundance ratio of core beneficial bacterial genera. The calculated diversity indices of the multiple microbial communities are weighted and aggregated using a pre-defined gradient boosting decision tree model, and the quality assessment score of the donor microbial community is output.

4. The method according to claim 1, characterized in that, Based on the activity parameters and diversity parameters, combined with the patient's clinical data, a support vector machine algorithm is used to perform colonization potential classification and prediction, outputting the potential colonization success rate, including: The activity parameters are subjected to time series normalization to generate an activity feature vector; The diversity parameters are weighted using the entropy weighting method to generate a diversity feature vector; Principal component dimensionality reduction decomposition was performed on the patient's clinical data to generate clinical feature vectors; The activity feature vector, diversity feature vector and clinical feature vector are combined by tensor to construct a multidimensional heterogeneous feature space; Using the support vector machine algorithm, the multidimensional heterogeneous feature space is mapped to a high-dimensional space through the radial basis kernel function; Solve for the optimal decision hyperplane parameters in the high-dimensional space, and calculate the classification decision function value based on the optimal decision hyperplane parameters; The classification decision function value is used to generate the potential colonization success rate through a probability transformation function.

5. The method according to claim 1, characterized in that, After generating the clinical transplantation protocol, the following is also included: Extract the execution elements of the clinical transplantation protocol; The execution elements, the activity and diversity parameters of the microbial preparation, and the clinical data are recombined into a feature vector to be predicted according to a preset conversion rule. The transplantation effect is predicted by performing a transplantation effect prediction calculation on the feature vector to be predicted using a preset prediction model.

6. The method according to claim 5, characterized in that, After the clinical transplantation protocol is executed, the method further includes: Obtain clinical data from patients after the implementation of clinical transplant protocols; Multi-dimensional feature extraction is performed on the clinical data after the implementation of the aforementioned clinical transplantation protocol to generate a third feature matrix; The deviation between the third feature matrix and the predicted transplantation effect is calculated to generate a deviation index. When the deviation index exceeds the preset deviation threshold, the execution elements of the clinical transplantation plan are adjusted based on the preset correction rules. The execution elements include at least one of the following: the dosage of the microbial preparation, the transplantation frequency, and the adjuvant therapy drug.

7. The method according to claim 5, characterized in that, After the clinical transplantation protocol is generated but before its execution, the method further includes: Acquire the patient's clinical data before microbiota transplantation, and perform tensor fusion with the parameter information of the microbiota preparation and the execution elements of the clinical transplantation protocol to generate a risk feature set; The generated risk feature set is processed by a preset transplantation decision model, and the operation feasibility score and immune compatibility score are output. If the feasibility score of the operation is lower than the preset operation threshold, or the immune compatibility score is lower than the preset compatibility threshold, then a transplantation blocking instruction and an alarm event log are generated.

8. A standardized management device for clinical pathways of microbiome transplantation, characterized in that, The device includes: The donor fit assessment module is used to perform microbiome fit assessment based on the patient's clinical data and the donor screening data, using a deep learning model, and to select a suitable donor based on the microbiome fit assessment results. The sample acquisition and sequencing module is used to acquire the bacterial community sample of the adaptor donor and perform high-throughput sequencing on the bacterial community sample to obtain sequencing results. The sample quality scoring module is used to generate a donor microbial community quality assessment score based on the sequencing results and according to a preset microbial community composition diversity index. The microbial community screening preparation module is used to screen out microbial community samples whose quality assessment scores are higher than a preset quality threshold, and to use preparation equipment to prepare the screened microbial community samples to obtain microbial community preparations. The colonization potential prediction module is used to detect the activity parameters and diversity parameters of the microbial preparation, and based on the activity parameters and diversity parameters, combined with the patient's clinical data, uses a support vector machine algorithm to perform colonization potential classification prediction and output the potential colonization success rate. The transplantation protocol generation module is used to output a clinical transplantation protocol based on the activity and diversity parameters of the microbial preparation and the patient's clinical data when the potential colonization success rate exceeds a preset success rate threshold, using a preset transplantation protocol generation algorithm.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images