A method for constructing a cloud platform for intravenous therapy management

Abstract

The present invention relates to the technical field of cloud platform construction, and in particular to a method for constructing an intravenous treatment management cloud platform. The method comprises the following steps: performing gene expression pathway enrichment analysis on the patient's venous system through gene sequencing technology, thereby obtaining gene expression data; performing gene expression spectrum analysis on the gene expression data and performing intravenous treatment response gene labeling, thereby obtaining a treatment response gene labeling set; using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data; performing node load balancing weight scaling based on resource scheduling system data and gene expression data, thereby obtaining resource balancing scaling data; performing multi-scale pyramid feature refinement on gene expression data, thereby obtaining gene image feature refinement data. The present invention realizes the method for constructing an intravenous treatment management cloud platform through node load balancing weight scaling and GPU heterogeneous computing acceleration processing.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud platform construction, and in particular to a method for constructing an intravenous therapy management cloud platform. Background Art

[0002] Due to the aging population and the increase in the number of patients with chronic diseases, the demand for intravenous therapy is increasing; a large amount of patient data and treatment information needs to be effectively managed. Cloud computing provides powerful computing and storage capabilities for processing large amounts of data; cloud platforms can be used to store, process and analyze the large amounts of data generated in intravenous therapy. In the traditional construction method of intravenous therapy management cloud platform, in the context of large-scale data processing and analysis, due to the failure to accurately estimate the system load, over-configuration, untimely resource release or inaccurate scheduling, resources are wasted; some existing systems are currently not adaptable enough to heterogeneous computing, and it is difficult to dynamically select the best heterogeneous computing resources according to different computing loads, which may result in the failure to fully utilize the potential of GPU acceleration. Summary of the invention

[0003] Based on this, it is necessary to provide a method for constructing an intravenous therapy management cloud platform to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a method for constructing an intravenous therapy management cloud platform is provided, the method comprising the following steps:

[0005] Step S1: performing gene expression pathway enrichment analysis on the patient's venous system through gene sequencing technology to obtain gene expression data; performing gene expression spectrum analysis on the gene expression data and performing intravenous treatment response gene markers to obtain a treatment response gene marker set;

[0006] Step S2: Using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data; performing node load balancing weight scaling based on the resource scheduling system data and gene expression data, thereby obtaining resource balancing scaling data;

[0007] Step S3: performing multi-scale pyramid feature refinement on the gene expression data to obtain gene image feature refinement data; performing channel and spatial attention adjustment on the gene image feature refinement data to obtain image feature enhancement data; performing GPU heterogeneous computing acceleration processing on the image feature enhancement data according to resource balanced scaling data to obtain parallel computing performance indicators;

[0008] Step S4: real-time monitoring of biological nodes of the patient through the sensor network and extraction of time and event-driven treatment parameters, thereby obtaining real-time physiological parameter change trend data;

[0009] Step S5: extracting personalized intravenous therapy management parameters based on the treatment response gene marker set and the real-time physiological parameter change trend data, thereby obtaining personalized intravenous therapy management parameters;

[0010] Step S6: Build a cloud platform using parallel computing performance indicators and personalized intravenous therapy management parameters to obtain cloud platform construction data.

[0011] The present invention obtains gene expression data through high-throughput genome sequencing, combined with gene expression spectrum analysis technology, to accurately and comprehensively understand the genetic information of the patient's venous system; it provides a personalized basis for subsequent intravenous treatment management, making the treatment plan more accurate and effective. Through cloud computing technology, an elastic computing resource scheduling system is built to realize intelligent management of computing resources; the statistics of resource scheduling system data help optimize the use of computing resources, improve the flexibility and performance of the system, and ensure that the computing tasks of gene expression data can be efficiently processed during peak and trough periods. By accelerating the processing of gene expression data through GPU heterogeneous computing, the data processing speed can be significantly improved, and the identification process of intravenous treatment response gene markers can be accelerated; GPU heterogeneous computing accelerated processing helps to evaluate system performance, ensure that GPU accelerated computing plays the maximum benefit in the entire process, and improve computing efficiency. Through the sensor network, real-time monitoring of the patient's physiological parameter change trend can obtain rich biological information. This helps to timely discover changes in the patient's physiological state and provide real-time data support for personalized treatment; time and event-driven treatment parameter extraction reduces the data transmission demand for the central cloud, and improves real-time and data privacy. Based on the treatment response gene marker set and real-time physiological parameter change trend data, the system can extract personalized intravenous treatment management parameters; these parameters are adjusted according to the patient's molecular characteristics and physiological state, making the treatment more accurate, maximizing the treatment effect, and reducing side effects. A cloud platform with high performance and personalized characteristics is constructed using parallel computing performance indicators and personalized intravenous treatment management parameters; it provides a powerful tool for the medical team to support large-scale data processing and the formulation of personalized treatment plans, and promotes the cutting-edge development of intravenous treatment management. Therefore, the present invention provides a method for constructing an intravenous treatment management cloud platform, which improves the system's performance, flexibility, and resource utilization efficiency through node load balancing weight scaling and GPU heterogeneous computing acceleration processing.

[0012] Preferably, step S1 comprises the following steps:

[0013] Step S11: performing gene expression pathway enrichment analysis on the patient's venous system through gene sequencing technology to obtain gene expression data;

[0014] Step S12: performing sparse calculation of gene expression amount on the gene expression data, thereby obtaining gene expression amount data; performing differential expression analysis on the gene expression amount data, thereby obtaining differentially expressed gene set data;

[0015] Step S13: constructing a gene expression network using the differentially expressed gene set data, thereby obtaining gene expression network data;

[0016] Step S14: performing functional enrichment analysis on the differentially expressed gene set data according to the gene expression network data, thereby obtaining enriched functional annotation data; performing key pathway mining based on the enriched functional annotation data, thereby obtaining key pathway gene set data;

[0017] Step S15: Generate a gene expression profile based on the gene expression network data and the key pathway gene set data, thereby obtaining gene expression profile data;

[0018] Step S16: performing gene expression pattern classification on the gene expression profile data, thereby obtaining expression pattern classification data; performing heterogeneity identification on the expression pattern classification data, thereby obtaining heterogeneous gene expression data;

[0019] Step S17: Perform intravenous treatment response gene labeling based on the heterogeneous gene expression data and the differentially expressed gene set data, thereby obtaining a treatment response gene marker set.

[0020] The present invention can comprehensively and accurately obtain the gene expression data of the patient's venous system through high-throughput genome sequencing; it is beneficial to deeply understand the patient's genomic characteristics and provide a comprehensive genetic information basis for subsequent treatment management. Through sparse calculation, the main information of gene expression can be effectively extracted, the redundancy of data can be reduced, and a more efficient data set can be provided for subsequent analysis; differential expression analysis identifies genes whose expression levels change significantly under different conditions, providing key data for understanding the biological differences of the patient's venous system. By constructing a gene expression network, the association between genes can be revealed, which helps to understand the interaction of genes in the patient's venous system; it helps to discover potential regulatory mechanisms and biological processes. Through functional enrichment analysis, the biological functions and pathways involved in the differentially expressed gene set can be identified, providing functional annotations for in-depth understanding of the patient's venous system; by mining key pathways, potential therapeutic targets and biological pathways that affect the treatment effect can be found, providing important clues for the formulation of subsequent treatment plans. Generating gene expression profiles helps to systematically understand the expression patterns of genes in the patient's venous system and provide an important bioinformatics basis for the formulation of personalized treatment plans. By classifying gene expression profiles, we can find differences in expression patterns between patients and provide guidance for individualized treatment. Identifying heterogeneous gene expression helps understand biological differences within patient groups and provides a basis for individualized prediction of treatment response. By analyzing heterogeneous gene expression data and differentially expressed gene sets, we can determine the treatment response gene marker set, which provides a biological basis for developing personalized intravenous treatment plans, and is expected to improve treatment effects and reduce adverse reactions.

[0021] Preferably, in step S12, the gene expression level is sparsely calculated, wherein the formula for sparse calculation of gene expression level is specifically:

[0022] ;

[0023] In the formula, represents the element value in the gene expression matrix, represents the number of rows in the gene expression matrix, represents the number of columns in the gene expression matrix, No. The initial expression value of the gene. represents the natural base, Indicates gene The initial expression value of the row, A measure of the trend of change. Indicates The sparseness of the row gene, represents the index of the correlation coefficient of the gene, Indicates The association weight parameter of the row gene, Indicates The expression adjustment coefficient of the gene is Indicates Biological factors affecting genes, Indicates The interaction parameters of the genes are represents the gene sampling time variable, Indicates The regulatory factor of the line Characteristic variables representing the values ​​of trigonometric functions.

[0024] The present invention constructs a gene expression sparse calculation formula for performing gene expression sparse calculation on gene expression data; in the formula Logarithmic transformation of some gene expression values ​​helps to smooth the data distribution and reduce the range of expression values, while the overall trend of gene expression is retained through cube root operation, which helps to maintain the stability of expression data; The periodic component of gene expression is partially added, which helps to simulate the periodic changes of gene expression. The sine function can simulate this periodic change and improve the biological rationality of the model; The elements in the gene expression matrix tend to be gradually more sparse in part, and the limit operation processes the sparse measure so that the model can take into account the sparsity of gene expression and better simulate the actual gene expression situation; Some of them comprehensively evaluate the relationship between genes. Through square and square root operations, multiple aspects of gene association are integrated, which helps to capture the complex relationship between genes and improve the expressiveness of the model; The partial introduction of second-order derivative operations, combined with the characteristic variables of the regulatory factor wᵢ and the trigonometric function, helps to simulate the dynamic adjustment of gene expression and the trend of changes over time. The second-order derivative operation can capture the acceleration information of the gene expression value. Combining the regulatory factors and trigonometric functions helps to better simulate the dynamic changes of gene expression and make the model more flexible and adaptable.

[0025] Preferably, step S2 comprises the following steps:

[0026] Step S21: Using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data;

[0027] Step S22: monitoring the resource scheduling system based on the resource scheduling system data, thereby obtaining resource utilization data;

[0028] Step S23: assigning task priority weights according to the gene expression data, thereby obtaining task priority weight data;

[0029] Step S24: performing adaptive load balancing according to the resource utilization data and the task priority weight data, thereby obtaining node load balancing data;

[0030] Step S25: Automatically scaling records are performed according to the node load balancing data, thereby obtaining automatic scaling record data;

[0031] Step S26: Perform resource balancing scaling processing based on the automatic scaling record data and the node load balancing data, thereby obtaining resource balancing scaling data.

[0032] The present invention uses cloud computing technology, and the system can realize dynamic resource allocation, flexibly adjust computing resources according to the changes in actual computing load, ensure that sufficient resources can be provided when the load is high, and release excess resources when the load is low, thereby improving the overall computing efficiency. By monitoring the resource scheduling system data, the system can understand the use of computing resources in real time; it is helpful to timely discover potential performance bottlenecks or resource waste, and take corresponding optimization measures to improve resource utilization efficiency. By determining the priority weight of the task through gene expression data, the system can personalize the task according to the patient's gene characteristics; it is helpful to improve the personalization of the treatment plan and better meet the needs of the patient. Combined with resource utilization and task priority weight data, the system can realize adaptive load balancing, ensure that different tasks are evenly distributed on the computing nodes; it is helpful to avoid excessive load or idleness of computing nodes and improve the overall computing efficiency. Based on the node load balancing data, the system can realize automatic scaling records and dynamically adjust the system scale according to actual needs; it is helpful to avoid waste of system resources, and quickly expand when needed to meet high load requirements, and improve the elasticity and adaptability of the system. By integrating the automatic scaling record data and node load balancing data, the system can calculate the comprehensive resource utilization index; this index comprehensively considers factors such as the system's dynamic adjustment and load balancing optimization, provides a comprehensive evaluation of system performance, and helps further optimization and adjustment.

[0033] Preferably, step S3 comprises the following steps:

[0034] Step S31: performing gene imaging processing on the gene expression data to obtain gene expression image data;

[0035] Step S32: performing model treatment response migration processing on the gene expression image data and the treatment response gene marker set, thereby obtaining a convolutional neural network optimization model;

[0036] Step S33: using a convolutional neural network optimization model to perform multi-scale pyramid difference remapping on the gene expression image data, thereby obtaining gene image feature refinement data;

[0037] Step S34: using a convolutional neural network optimization model to perform channel attention adjustment on the gene image feature refinement data, thereby obtaining key channel feature mapping data;

[0038] Step S35: performing spatial attention adjustment on the key channel feature mapping data to obtain image feature enhancement data;

[0039] Step S36: performing parallel computing task division on the image feature enhancement data, thereby obtaining parallel computing task data; performing heterogeneous computing memory mapping on the image feature enhancement data, thereby obtaining GPU memory mapping data;

[0040] Step S37: Perform GPU heterogeneous computing acceleration processing according to the parallel computing task data and the GPU memory mapping data, so as to obtain GPU heterogeneous computing acceleration data; perform GPU heterogeneous computing acceleration processing on the GPU heterogeneous computing acceleration data according to the resource balancing scaling data, so as to obtain parallel computing performance indicators.

[0041] The present invention helps to convert raw data into a more intuitive and visual form by processing gene expression data in an image form. Integrating the treatment response gene marker set into the model helps the model to better learn the patterns related to the treatment response in the gene expression image and improve the transfer learning ability of the model; through model migration processing, the learned knowledge can be shared between different data sets, thereby reducing the training time on the new data set and improving the generalization performance of the model. Multi-scale pyramid difference remapping helps to capture features at different scales, and channel and spatial attention adjustment helps to extract and highlight key information, thereby making the information of gene images richer; the processed feature map is more representative and can better express the abstract patterns in the gene image, providing more targeted data for subsequent computing tasks. By reasonably dividing parallel computing tasks, multiple computing tasks can be carried out simultaneously, fully utilizing computing resources and improving data processing speed; heterogeneous computing memory mapping helps to use GPU memory more effectively, reduce data transmission bottlenecks, and improve overall computing performance. GPU heterogeneous computing can give full play to the parallel computing capabilities of GPUs, significantly improve the processing speed of genetic image data, and accelerate the response time of the intravenous treatment management cloud platform. Through performance indicator analysis, the effect of GPU heterogeneous computing can be comprehensively evaluated, providing a reference for subsequent optimization and ensuring that the system can maintain efficient and stable performance under different loads.

[0042] Preferably, in step S37, the GPU heterogeneous computing acceleration processing, wherein the formula for the GPU heterogeneous computing acceleration processing is specifically:

[0043] ;

[0044] In the formula, Indicates GPU heterogeneous computing acceleration data, Indicates the number of parallel computing tasks, Indicates the sequence number of the parallel computing task, Indicates parallel computing tasks, Indicates GPU memory mapping data for parallel computing tasks, represents the performance parameter associated with the parallel computing task, Indicates The complexity measurement parameters of parallel computing tasks, represents pi, Indicates The phase angle parameters of the parallel computing tasks, Indicates The computing resource consumption parameters of a parallel computing task.

[0045] The present invention constructs a calculation formula for GPU heterogeneous computing acceleration processing, which is used to perform GPU heterogeneous computing acceleration processing according to parallel computing task data and GPU memory mapping data; in the formula The part represents the change rate of the GPU memory mapping data of the parallel computing task. By calculating this change rate, the memory access mode can be optimized and the data access efficiency can be improved, thereby reducing the data transmission time and accelerating the computing task; The part represents the impact of the performance parameter associated with the parallel computing task on GPU acceleration. This logarithmic term converts the performance parameter into the corresponding acceleration, which helps evaluate the contribution of each task to the overall computing performance; Phase angle parameters and computing resource consumption parameters are partially introduced. Through the sine function, a mechanism for balancing the phase angle and computing resource consumption parameters is provided, which helps to adjust the execution order and resource allocation of tasks to maximize the use of GPU computing resources; the square operation of the entire formula helps to emphasize the impact of each task on the overall acceleration data, helps to eliminate the adverse effects of individual tasks on the overall performance, and makes the entire system more stable.

[0046] Preferably, step S32 includes the following steps:

[0047] Step S321: performing convolutional neural image training on the gene expression image data to obtain convolutional neural image training data, wherein the convolutional neural image training data includes convolutional neural image weight data and a convolutional neural network model;

[0048] Step S322: performing feature mapping on the gene expression image data using the convolutional neural image weight data, thereby obtaining gene image feature data;

[0049] Step S323: Associating the image features with the gene markers of the gene image feature data and the treatment response gene marker set, thereby obtaining feature marker association data;

[0050] Step S324: Use the feature tag association data to perform transfer learning processing on the convolutional neural network model to obtain a convolutional neural network optimization model.

[0051] The present invention can identify the pattern of gene expression through CNN training, which is helpful to discover specific features related to intravenous treatment, such as vascular structure or tissue morphology; through training, the weight data of the convolutional neural network is obtained, and these weights reflect the degree of importance of the model to different features, providing valuable information for subsequent analysis; the convolutional neural network model formed during the training process can be used in subsequent data processing to provide a basis for the prediction of treatment response. Using the convolutional neural image weight data, feature mapping can generate more abstract and advanced feature representations, strengthen the capture of important structures and patterns in gene expression images; during the feature mapping process, the dimension reduction and denoising of the original data can be achieved, and the efficiency and accuracy of subsequent analysis can be improved. Determining the correlation between gene image features and treatment response gene markers helps to understand which image features are closely related to treatment response; the correlation data provides a basis for establishing a prediction model, improves the prediction performance of treatment response, and helps to formulate a more personalized treatment plan. Using the previously trained network weights, the network is more easily adapted to the current task, and the convergence speed and performance of the model are improved; for the situation where the gene expression image data may be less, transfer learning helps to build a high-performance model on a small number of samples and improve the generalization ability of the model.

[0052] Preferably, step S33 includes the following steps:

[0053] Step S331: constructing a multi-scale pyramid for the gene expression image data using a convolutional neural network optimization model, thereby obtaining multi-scale feature pyramid data;

[0054] Step S332: performing a pyramid pooling operation on the multi-scale feature pyramid data to obtain a pyramid pooling feature vector; performing attention weight calculation on the pyramid pooling feature vector to obtain attention weight data;

[0055] Step S333: performing feature weighted fusion on the multi-scale feature pyramid data and the attention weight data to obtain weighted fusion feature data; performing scale splicing on the weighted fusion feature data to obtain multi-scale image feature data;

[0056] Step S334: performing deconvolution processing and upsampling processing using the multi-scale image feature data, thereby obtaining high-resolution feature mapping data;

[0057] Step S335: extract mapping differences from the high-resolution feature mapping data and the gene image feature data, thereby obtaining gene image feature refinement data.

[0058] The convolutional neural network of the present invention can construct a multi-scale pyramid for gene expression images, thereby capturing feature information at different scales; it helps to fully understand the complex patterns of gene expression and improve the sensitivity of the model to gene correlation. The pyramid pooling operation helps to integrate multi-scale features and improve the perception of global information; it helps the model to better understand the key features at different scales in the gene expression image; by calculating the attention weight of the pyramid pooling feature vector, the model can automatically learn and focus on important areas in the gene expression image, enhance the perception of key features, and improve the model's understanding of the gene expression pattern. Feature weighted fusion enables the model to better integrate information at different scales by considering the weights of multi-scale image features; it helps the system to more comprehensively understand the patterns and laws in the gene expression image; scale splicing fuses the weighted fused features with the original multi-scale image features, further enhancing the model's understanding of the complex structure of the gene expression image and improving the capture of the correlation of features at different scales. Deconvolution and upsampling processing help to restore low-resolution feature maps to high resolution, improve the detail presentation of the gene expression image, and are of key significance for more accurately locating and analyzing the spatial distribution of gene expression. Mapping difference extraction helps to extract subtle changes and differences in gene image features by comparing high-resolution feature mapping data with gene image feature data, thereby describing important information in gene expression images in a more refined manner.

[0059] Preferably, step S34 includes the following steps:

[0060] Step S341: introducing a channel attention module into the convolutional neural network optimization model, thereby obtaining a convolutional neural network channel model;

[0061] Step S342: using the convolutional neural network channel model to perform global average pooling on the gene image feature refinement data, thereby obtaining global average pooling layer data;

[0062] Step S343: inputting the global average pooling layer data into a preset channel weight generator for channel weight learning, thereby obtaining channel weight data;

[0063] Step S344: weighting the gene image feature refinement data by the treatment response contribution channel using the channel weight data, thereby obtaining contribution channel weight data;

[0064] Step S345: Adjust the contribution channel weight data according to the gradient descent optimization algorithm and the convolutional neural network channel model to obtain key channel feature mapping data.

[0065] The channel attention module of the present invention can automatically learn and adjust the weights of different channels, so that the network pays more attention to the gene features that are important for intravenous treatment response; such optimization helps to improve the sensitivity of the model to gene images, thereby more accurately capturing the treatment response related information in the patient's gene expression spectrum. Through global average pooling, the high-dimensional feature mapping of the gene image can be effectively converted into fixed-dimensional features, reducing the data dimension and reducing the computational burden of the model; it helps to improve the generalization ability of the model, prevent overfitting, and simplify the complexity of subsequent channel weight learning. Through the channel weight learner, the system can automatically learn the weight of each channel in the gene image, thereby accurately capturing the contribution of each channel to the patient's treatment response; this adaptive channel weight learning mechanism helps to better understand the feature expression of different channels in the gene image, and provide more targeted information for subsequent processing. By introducing the treatment response contribution channel empowerment, the system can assign corresponding weights to each channel according to its contribution to the treatment response; it helps to focus on the channels that have a significant impact on the treatment response, thereby improving the sensitivity and interpretability of the model. Using the gradient descent optimization algorithm to adjust the contributing channel weights can further optimize the performance of the model; by continuously updating the channel weights based on the gradient information, the model can be better adapted to the training data, improving the generalization ability of the model while ensuring that the model's predictions of treatment responses have higher accuracy and stability; this iterative adjustment process helps the model to continuously optimize itself and gradually converge to a more effective parameter configuration.

[0066] Preferably, step S4 comprises the following steps:

[0067] Step S41: obtaining treatment parameters and real-time physiological parameters of the patient through the sensor network; performing interactive impact analysis on the treatment parameters and the real-time physiological parameters of the patient, thereby obtaining interactive impact analysis data;

[0068] Step S42: performing physiologically relevant edge computing on the treatment parameters to obtain physiologically relevant edge computing data; performing event-driven analysis and marking on the physiologically relevant edge computing data to obtain event-driven analysis data;

[0069] Step S43: performing time series analysis on the treatment parameters to obtain periodic variation data; performing time window cluster analysis on the periodic variation data to obtain time window cluster data;

[0070] Step S44: performing time-driven fusion of treatment parameters according to the event-driven analysis data and the time window clustering data, thereby obtaining time-driven data of treatment parameters;

[0071] Step S45: Predicting the real-time physiological parameter change trend of the treatment parameters and the patient's real-time physiological parameters according to the interactive impact analysis data and the treatment parameter time-driven data, thereby obtaining real-time physiological parameter change trend data.

[0072] The present invention obtains the patient's real-time physiological parameters, including key indicators of heart rate and blood pressure, which helps doctors understand the patient's physiological state more accurately; at the same time, obtaining treatment parameters, such as drug use and treatment plans, provides a basis for personalized treatment. Edge computing allows physiological-related calculations to be performed on the device side, reducing data transmission delays and providing more real-time feedback; event-driven analysis can identify key events and further strengthen the real-time processing of important information. Physiological-related edge computing can reduce the data transmission requirements of the central cloud platform and reduce bandwidth pressure; event-driven analysis refines sensor data and only transmits key events, reducing the burden of data transmission. Time series analysis helps to identify the changing trends of treatment parameters, including long-term trends and periodic changes; this provides doctors with a deeper understanding of treatment effects and patient conditions. Time window cluster analysis can divide patients into different groups according to the change patterns of treatment parameters; it helps to optimize personalized treatment plans and provide patients with more accurate medical services. Fusion of event-driven analysis data with time window cluster data can more comprehensively understand the relationship between the temporal changes of treatment parameters and the occurrence of events, helping the medical team to better understand the patient's condition. By combining the patient's physiological parameters, treatment parameters and their temporal changes, we can achieve personalized predictions of future changes in the patient's physiological state, providing doctors with earlier intervention opportunities.

[0073] The present invention obtains gene expression data through high-throughput genome sequencing, combined with gene expression spectrum analysis technology, to accurately and comprehensively understand the genetic information of the patient's venous system; it provides a personalized basis for subsequent intravenous treatment management, making the treatment plan more accurate and effective. Through cloud computing technology, an elastic computing resource scheduling system is built to realize intelligent management of computing resources; the statistics of resource scheduling system data help optimize the use of computing resources, improve the flexibility and performance of the system, and ensure that the computing tasks of gene expression data can be efficiently processed during peak and trough periods. By accelerating the processing of gene expression data through GPU heterogeneous computing, the data processing speed can be significantly improved, and the identification process of intravenous treatment response gene markers can be accelerated; GPU heterogeneous computing accelerated processing helps to evaluate system performance, ensure that GPU accelerated computing plays the maximum benefit in the entire process, and improve computing efficiency. Through the sensor network, real-time monitoring of the patient's physiological parameter change trend can obtain rich biological information. This helps to timely discover changes in the patient's physiological state and provide real-time data support for personalized treatment; time and event-driven treatment parameter extraction reduces the data transmission demand for the central cloud, and improves real-time and data privacy. Based on the treatment response gene marker set and real-time physiological parameter change trend data, the system can extract personalized intravenous treatment management parameters; these parameters are adjusted according to the patient's molecular characteristics and physiological state, making the treatment more accurate, maximizing the treatment effect, and reducing side effects. A cloud platform with high performance and personalized characteristics is constructed using parallel computing performance indicators and personalized intravenous treatment management parameters; it provides a powerful tool for the medical team to support large-scale data processing and the formulation of personalized treatment plans, and promotes the cutting-edge development of intravenous treatment management. Therefore, the present invention provides a method for constructing an intravenous treatment management cloud platform, which improves the system's performance, flexibility, and resource utilization efficiency through node load balancing weight scaling and GPU heterogeneous computing acceleration processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 A schematic diagram of the steps of a method for building an intravenous therapy management cloud platform;

[0075] Figure 2 for Figure 1 Detailed implementation steps of step S1 in FIG.

[0076] Figure 3 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0077] Figure 4 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0078] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0079] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0080] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0081] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0082] To achieve this, please refer to Figures 1 to 4 , a method for constructing an intravenous therapy management cloud platform, the method comprising the following steps:

[0083] Step S1: performing gene expression pathway enrichment analysis on the patient's venous system through gene sequencing technology to obtain gene expression data; performing gene expression spectrum analysis on the gene expression data and performing intravenous treatment response gene markers to obtain a treatment response gene marker set;

[0084] Step S2: Using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data; performing node load balancing weight scaling based on the resource scheduling system data and gene expression data, thereby obtaining resource balancing scaling data;

[0085] Step S3: performing multi-scale pyramid feature refinement on the gene expression data to obtain gene image feature refinement data; performing channel and spatial attention adjustment on the gene image feature refinement data to obtain image feature enhancement data; performing GPU heterogeneous computing acceleration processing on the image feature enhancement data according to resource balanced scaling data to obtain parallel computing performance indicators;

[0086] Step S4: real-time monitoring of biological nodes of the patient through the sensor network and extraction of time and event-driven treatment parameters, thereby obtaining real-time physiological parameter change trend data;

[0087] Step S5: extracting personalized intravenous therapy management parameters based on the treatment response gene marker set and the real-time physiological parameter change trend data, thereby obtaining personalized intravenous therapy management parameters;

[0088] Step S6: Build a cloud platform using parallel computing performance indicators and personalized intravenous therapy management parameters to obtain cloud platform construction data.

[0089] The present invention obtains gene expression data through high-throughput genome sequencing, combined with gene expression spectrum analysis technology, to accurately and comprehensively understand the genetic information of the patient's venous system; it provides a personalized basis for subsequent intravenous treatment management, making the treatment plan more accurate and effective. Through cloud computing technology, an elastic computing resource scheduling system is built to realize intelligent management of computing resources; the statistics of resource scheduling system data help optimize the use of computing resources, improve the flexibility and performance of the system, and ensure that the computing tasks of gene expression data can be efficiently processed during peak and trough periods. By accelerating the processing of gene expression data through GPU heterogeneous computing, the data processing speed can be significantly improved, and the identification process of intravenous treatment response gene markers can be accelerated; GPU heterogeneous computing accelerated processing helps to evaluate system performance, ensure that GPU accelerated computing plays the maximum benefit in the entire process, and improve computing efficiency. Through the sensor network, real-time monitoring of the patient's physiological parameter change trend can obtain rich biological information. This helps to timely discover changes in the patient's physiological state and provide real-time data support for personalized treatment; time and event-driven treatment parameter extraction reduces the data transmission demand for the central cloud, and improves real-time and data privacy. Based on the treatment response gene marker set and real-time physiological parameter change trend data, the system can extract personalized intravenous treatment management parameters; these parameters are adjusted according to the patient's molecular characteristics and physiological state, making the treatment more accurate, maximizing the treatment effect, and reducing side effects. A cloud platform with high performance and personalized characteristics is constructed using parallel computing performance indicators and personalized intravenous treatment management parameters; it provides a powerful tool for the medical team to support large-scale data processing and the formulation of personalized treatment plans, and promotes the cutting-edge development of intravenous treatment management. Therefore, the present invention provides a method for constructing an intravenous treatment management cloud platform, which improves the system's performance, flexibility, and resource utilization efficiency through node load balancing weight scaling and GPU heterogeneous computing acceleration processing.

[0090] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a method for constructing a cloud platform for intravenous therapy management of the present invention. In this example, the method for constructing a cloud platform for intravenous therapy management includes the following steps:

[0091] Step S1: performing gene expression pathway enrichment analysis on the patient's venous system through gene sequencing technology to obtain gene expression data; performing gene expression spectrum analysis on the gene expression data and performing intravenous treatment response gene markers to obtain a treatment response gene marker set;

[0092] In an embodiment of the present invention, a blood sample is collected from a patient to ensure the quality and integrity of the sample; a high-throughput genome sequencing technology is used to sequence the patient's venous system to obtain gene expression data; the gene expression data is analyzed to identify genes related to treatment response, and a treatment response gene marker set is formed.

[0093] Step S2: Using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data; performing node load balancing weight scaling based on the resource scheduling system data and gene expression data, thereby obtaining resource balancing scaling data;

[0094] In an embodiment of the present invention, cloud computing technology is used to build an elastic computing resource scheduling system to ensure that the system has elasticity and high availability; data of the elastic computing resource scheduling system is collected, including the status of computing nodes and task allocation; based on the resource scheduling system data, the utilization rate of elastic computing resources is calculated and recorded.

[0095] Step S3: performing multi-scale pyramid feature refinement on the gene expression data to obtain gene image feature refinement data; performing channel and spatial attention adjustment on the gene image feature refinement data to obtain image feature enhancement data; performing GPU heterogeneous computing acceleration processing on the image feature enhancement data according to resource balanced scaling data to obtain parallel computing performance indicators;

[0096] In the embodiment of the present invention, resource scheduling system data and gene expression data are used to perform GPU heterogeneous computing acceleration processing on the gene expression data; and performance index analysis is performed on the data after GPU heterogeneous computing acceleration, including computing speed and efficiency.

[0097] Step S4: real-time monitoring of biological nodes of the patient through the sensor network and extraction of time and event-driven treatment parameters, thereby obtaining real-time physiological parameter change trend data;

[0098] In an embodiment of the present invention, a biosensor network is deployed to monitor changes in physiological parameters of patients in real time; edge computing is performed at the monitoring nodes to process biological node data and reduce the burden on the cloud platform.

[0099] Step S5: extracting personalized intravenous therapy management parameters based on the treatment response gene marker set and the real-time physiological parameter change trend data, thereby obtaining personalized intravenous therapy management parameters;

[0100] In an embodiment of the present invention, the correlation between gene expression data and physiological parameters is deeply mined to ensure that the selected treatment response gene marker set is closely related to the patient's physiological condition; a dynamic weight adjustment mechanism is introduced to adjust the weight of the treatment response gene marker according to the patient's real-time physiological parameter change trend data.

[0101] Step S6: constructing a cloud platform using the parallel computing performance index and the personalized intravenous therapy management parameters, thereby obtaining cloud platform construction data;

[0102] In an embodiment of the present invention, parallel computing performance indicators are integrated with personalized treatment parameters to provide construction data for a cloud platform; a cloud platform is constructed based on the integrated data to ensure that the platform can provide personalized treatment management based on the patient's genetic information, physiological parameters, and treatment response data; data security and privacy protection are strengthened in the cloud platform to ensure the confidentiality of patient information; the constructed cloud platform is tested and performance tuning is performed to ensure efficient operation of the platform.

[0103] Preferably, step S1 comprises the following steps:

[0104] Step S11: performing gene expression pathway enrichment analysis on the patient's venous system through gene sequencing technology to obtain gene expression data;

[0105] Step S12: performing sparse calculation of gene expression amount on the gene expression data, thereby obtaining gene expression amount data; performing differential expression analysis on the gene expression amount data, thereby obtaining differentially expressed gene set data;

[0106] Step S13: constructing a gene expression network using the differentially expressed gene set data, thereby obtaining gene expression network data;

[0107] Step S14: performing functional enrichment analysis on the differentially expressed gene set data according to the gene expression network data, thereby obtaining enriched functional annotation data; performing key pathway mining based on the enriched functional annotation data, thereby obtaining key pathway gene set data;

[0108] Step S15: Generate a gene expression profile based on the gene expression network data and the key pathway gene set data, thereby obtaining gene expression profile data;

[0109] Step S16: performing gene expression pattern classification on the gene expression profile data, thereby obtaining expression pattern classification data; performing heterogeneity identification on the expression pattern classification data, thereby obtaining heterogeneous gene expression data;

[0110] Step S17: Perform intravenous treatment response gene labeling based on the heterogeneous gene expression data and the differentially expressed gene set data, thereby obtaining a treatment response gene marker set.

[0111] The present invention can comprehensively and accurately obtain the gene expression data of the patient's venous system through high-throughput genome sequencing; it is beneficial to deeply understand the patient's genomic characteristics and provide a comprehensive genetic information basis for subsequent treatment management. Through sparse calculation, the main information of gene expression can be effectively extracted, the redundancy of data can be reduced, and a more efficient data set can be provided for subsequent analysis; differential expression analysis identifies genes whose expression levels change significantly under different conditions, providing key data for understanding the biological differences of the patient's venous system. By constructing a gene expression network, the association between genes can be revealed, which helps to understand the interaction of genes in the patient's venous system; it helps to discover potential regulatory mechanisms and biological processes. Through functional enrichment analysis, the biological functions and pathways involved in the differentially expressed gene set can be identified, providing functional annotations for in-depth understanding of the patient's venous system; by mining key pathways, potential therapeutic targets and biological pathways that affect the treatment effect can be found, providing important clues for the formulation of subsequent treatment plans. Generating gene expression profiles helps to systematically understand the expression patterns of genes in the patient's venous system and provide an important bioinformatics basis for the formulation of personalized treatment plans. By classifying gene expression profiles, we can find differences in expression patterns between patients and provide guidance for individualized treatment. Identifying heterogeneous gene expression helps understand biological differences within patient groups and provides a basis for individualized prediction of treatment response. By analyzing heterogeneous gene expression data and differentially expressed gene sets, we can determine the treatment response gene marker set, which provides a biological basis for developing personalized intravenous treatment plans, and is expected to improve treatment effects and reduce adverse reactions.

[0112] As an example of the present invention, refer to Figure 2 As shown, in this example, step S1 includes:

[0113] Step S11: performing gene expression pathway enrichment analysis on the patient's venous system through gene sequencing technology to obtain gene expression data;

[0114] In an embodiment of the present invention, a biological tissue sample, which may be blood or vascular endothelial cells, is obtained from the patient's venous system; a DNA library is sequenced using high-throughput sequencing technology to generate a large number of short sequences or long reads representing the patient's genome; and quality control is performed on the sequencing data, including removal of low-quality sequences and sequencing adapters.

[0115] Step S12: performing sparse calculation of gene expression amount on the gene expression data, thereby obtaining gene expression amount data; performing differential expression analysis on the gene expression amount data, thereby obtaining differentially expressed gene set data;

[0116] In an embodiment of the present invention, the sequencing data is mapped to a reference genome to obtain the number of reads for each gene; the expression level of each gene is calculated using a sparse calculation formula for gene expression; the expression matrix is ​​sparsely processed to exclude low-expression genes to reduce the data dimension.

[0117] Step S13: constructing a gene expression network using the differentially expressed gene set data, thereby obtaining gene expression network data;

[0118] In the embodiment of the present invention, statistical methods are used to compare samples to find genes that are significantly expressed under different conditions; the differentially expressed genes are constructed into a network structure, in which nodes represent genes and edges represent interactions or correlations between them.

[0119] Step S14: performing functional enrichment analysis on the differentially expressed gene set data according to the gene expression network data, thereby obtaining enriched functional annotation data; performing key pathway mining based on the enriched functional annotation data, thereby obtaining key pathway gene set data;

[0120] In the embodiment of the present invention, bioinformatics tools are used to perform functional enrichment analysis on differentially expressed genes to find out the enriched functions of these genes in biological processes.

[0121] Step S15: Generate a gene expression profile based on the gene expression network data and the key pathway gene set data, thereby obtaining gene expression profile data;

[0122] In the embodiment of the present invention, by mining the key pathways in the gene expression network, particularly important pathways in the patient's venous system are found; the gene expression conditions of these key pathways are integrated to form the patient's gene expression profile.

[0123] Step S16: performing gene expression pattern classification on the gene expression profile data, thereby obtaining expression pattern classification data; performing heterogeneity identification on the expression pattern classification data, thereby obtaining heterogeneous gene expression data;

[0124] In the embodiment of the present invention, a clustering algorithm (such as K-means clustering, hierarchical clustering, etc.) is used to classify the gene expression profile to find gene clusters with similar expression patterns.

[0125] Step S17: performing intravenous treatment response gene labeling according to the heterogeneous gene expression data and the differentially expressed gene set data, thereby obtaining a treatment response gene marker set;

[0126] In the embodiment of the present invention, by analyzing the heterogeneity in the gene expression profile, genes with significant differences in gene expression under different conditions are found; according to the needs of intravenous treatment, gene markers related to treatment response are screened out from the heterogeneous genes.

[0127] Preferably, in step S12, the gene expression level is sparsely calculated, wherein the formula for sparse calculation of gene expression level is specifically:

[0128] ;

[0129] In the formula, represents the element value in the gene expression matrix, represents the number of rows in the gene expression matrix, represents the number of columns in the gene expression matrix, No. The initial expression value of the gene. represents the natural base, Indicates gene The initial expression value of the row, A measure of the trend of change. Indicates The sparseness of the row gene, represents the index of the correlation coefficient of the gene, Indicates The association weight parameter of the row gene, Indicates The expression adjustment coefficient of the gene is Indicates Biological factors affecting genes, Indicates The interaction parameters of the genes are represents the gene sampling time variable, Indicates The regulatory factor of the line Characteristic variables representing the values ​​of trigonometric functions.

[0130] The present invention constructs a gene expression sparse calculation formula for performing gene expression sparse calculation on gene expression data; in the formula Logarithmic transformation of some gene expression values ​​helps to smooth the data distribution and reduce the range of expression values, while the overall trend of gene expression is retained through cube root operation, which helps to maintain the stability of expression data; The periodic component of gene expression is partially added, which helps to simulate the periodic changes of gene expression. The sine function can simulate this periodic change and improve the biological rationality of the model; The elements in the gene expression matrix tend to be gradually more sparse in part, and the limit operation processes the sparse measure so that the model can take into account the sparsity of gene expression and better simulate the actual gene expression situation; Some of them comprehensively evaluate the relationship between genes. Through square and square root operations, multiple aspects of gene association are integrated, which helps to capture the complex relationship between genes and improve the expressiveness of the model; The partial introduction of second-order derivative operations, combined with the characteristic variables of the regulatory factor wᵢ and the trigonometric function, helps to simulate the dynamic adjustment of gene expression and the trend of changes over time. The second-order derivative operation can capture the acceleration information of the gene expression value. Combining the regulatory factors and trigonometric functions helps to better simulate the dynamic changes of gene expression and make the model more flexible and adaptable.

[0131] Preferably, step S2 comprises the following steps:

[0132] Step S21: Using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data;

[0133] Step S22: monitoring the resource scheduling system based on the resource scheduling system data, thereby obtaining resource utilization data;

[0134] Step S23: assigning task priority weights according to the gene expression data, thereby obtaining task priority weight data;

[0135] Step S24: performing adaptive load balancing according to the resource utilization data and the task priority weight data, thereby obtaining node load balancing data;

[0136] Step S25: Automatically scaling records are performed according to the node load balancing data, thereby obtaining automatic scaling record data;

[0137] Step S26: Perform resource balancing scaling processing based on the automatic scaling record data and the node load balancing data, thereby obtaining resource balancing scaling data.

[0138] The present invention uses cloud computing technology, and the system can realize dynamic resource allocation, flexibly adjust computing resources according to the changes in actual computing load, ensure that sufficient resources can be provided when the load is high, and release excess resources when the load is low, thereby improving the overall computing efficiency. By monitoring the resource scheduling system data, the system can understand the use of computing resources in real time; it is helpful to timely discover potential performance bottlenecks or resource waste, and take corresponding optimization measures to improve resource utilization efficiency. By determining the priority weight of the task through gene expression data, the system can personalize the task according to the patient's gene characteristics; it is helpful to improve the personalization of the treatment plan and better meet the needs of the patient. Combined with resource utilization and task priority weight data, the system can realize adaptive load balancing, ensure that different tasks are evenly distributed on the computing nodes; it is helpful to avoid excessive load or idleness of computing nodes and improve the overall computing efficiency. Based on the node load balancing data, the system can realize automatic scaling records and dynamically adjust the system scale according to actual needs; it is helpful to avoid waste of system resources, and quickly expand when needed to meet high load requirements, and improve the elasticity and adaptability of the system. By integrating the automatic scaling record data and node load balancing data, the system can calculate the comprehensive resource utilization index; this index comprehensively considers factors such as the system's dynamic adjustment and load balancing optimization, provides a comprehensive evaluation of system performance, and helps further optimization and adjustment.

[0139] As an example of the present invention, refer to Figure 3 As shown, in this example, step S2 includes:

[0140] Step S21: Using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data;

[0141] In an embodiment of the present invention, a virtual machine instance or container is created, and computing resources, memory, storage and network parameters are configured to ensure that these resources can meet the needs of the system; a resource scheduling system, such as Kubernetes and DockerSwarm, is deployed to manage and schedule cloud computing resources; and the system is configured to achieve elastic computing, that is, automatically expand or shrink computing resources according to changes in load.

[0142] Step S22: monitoring the resource scheduling system based on the resource scheduling system data, thereby obtaining resource utilization data;

[0143] In an embodiment of the present invention, the usage of cloud computing resources is monitored, including CPU utilization, memory usage, and network traffic; alarm rules are set so that notifications are received in a timely manner when resource usage is abnormal or exceeds a threshold, ensuring that problems can be responded to quickly.

[0144] Step S23: assigning task priority weights according to the gene expression data, thereby obtaining task priority weight data;

[0145] In an embodiment of the present invention, the priority weight of a task is calculated based on gene expression data, which involves analyzing the gene expression pattern to determine the biological importance of the task; the calculated task priority weight is associated with the task for use in subsequent resource scheduling.

[0146] Step S24: performing adaptive load balancing according to the resource utilization data and the task priority weight data, thereby obtaining node load balancing data;

[0147] In the embodiment of the present invention, the utilization of cloud computing resources, including the load and performance of each node, is continuously monitored; an adaptive algorithm is used to dynamically adjust the distribution of tasks according to the monitored resource utilization and task priority to achieve load balancing.

[0148] Step S25: Automatically scaling records are performed according to the node load balancing data, thereby obtaining automatic scaling record data;

[0149] In the embodiment of the present invention, automatic scaling rules are set based on indicators such as load and performance to define when automatic scaling is triggered; each time the system triggers automatic scaling, relevant information is recorded, including time, triggering reason, and scaling direction.

[0150] Step S26: performing resource balancing scaling processing based on the automatic scaling record data and the node load balancing data, thereby obtaining resource balancing scaling data;

[0151] In an embodiment of the present invention, recorded automatic scaling event information is summarized; node load balancing data is associated with automatic scaling records to form comprehensive resource utilization data; an algorithm is designed to calculate a comprehensive resource utilization index, which comprehensively considers factors such as load balancing and automatic scaling, and reflects the overall resource utilization of the system.

[0152] Preferably, step S3 comprises the following steps:

[0153] Step S31: performing gene imaging processing on the gene expression data to obtain gene expression image data;

[0154] Step S32: performing model treatment response migration processing on the gene expression image data and the treatment response gene marker set, thereby obtaining a convolutional neural network optimization model;

[0155] Step S33: using a convolutional neural network optimization model to perform multi-scale pyramid difference remapping on the gene expression image data, thereby obtaining gene image feature refinement data;

[0156] Step S34: using a convolutional neural network optimization model to perform channel attention adjustment on the gene image feature refinement data, thereby obtaining key channel feature mapping data;

[0157] Step S35: performing spatial attention adjustment on the key channel feature mapping data to obtain image feature enhancement data;

[0158] Step S36: performing parallel computing task division on the image feature enhancement data, thereby obtaining parallel computing task data; performing heterogeneous computing memory mapping on the image feature enhancement data, thereby obtaining GPU memory mapping data;

[0159] Step S37: Perform GPU heterogeneous computing acceleration processing according to the parallel computing task data and the GPU memory mapping data, so as to obtain GPU heterogeneous computing acceleration data; perform GPU heterogeneous computing acceleration processing on the GPU heterogeneous computing acceleration data according to the resource balancing scaling data, so as to obtain parallel computing performance indicators.

[0160] The present invention helps to convert raw data into a more intuitive and visual form by processing gene expression data in an image form. Integrating the treatment response gene marker set into the model helps the model to better learn the patterns related to the treatment response in the gene expression image and improve the transfer learning ability of the model; through model migration processing, the learned knowledge can be shared between different data sets, thereby reducing the training time on the new data set and improving the generalization performance of the model. Multi-scale pyramid difference remapping helps to capture features at different scales, and channel and spatial attention adjustment helps to extract and highlight key information, thereby making the information of gene images richer; the processed feature map is more representative and can better express the abstract patterns in the gene image, providing more targeted data for subsequent computing tasks. By reasonably dividing parallel computing tasks, multiple computing tasks can be carried out simultaneously, fully utilizing computing resources and improving data processing speed; heterogeneous computing memory mapping helps to use GPU memory more effectively, reduce data transmission bottlenecks, and improve overall computing performance. GPU heterogeneous computing can give full play to the parallel computing capabilities of GPUs, significantly improve the processing speed of genetic image data, and accelerate the response time of the intravenous treatment management cloud platform. Through performance indicator analysis, the effect of GPU heterogeneous computing can be comprehensively evaluated, providing a reference for subsequent optimization and ensuring that the system can maintain efficient and stable performance under different loads.

[0161] As an example of the present invention, refer to Figure 4 As shown, in this example, step S3 includes:

[0162] Step S31: performing gene imaging processing on the gene expression data to obtain gene expression image data;

[0163] In an embodiment of the present invention, the gene expression data is standardized to ensure that the expression values ​​of different genes are on a similar scale; a dimensionality reduction technique such as principal component analysis (PCA) or t-distributed neighbor embedding (t-SNE) is used to map the high-dimensional gene expression data to a lower-dimensional space; the reduced-dimensional data is reorganized into an image, and each sample can be regarded as a row or column of the image to form a matrix, and the pixel values ​​of the image can represent the expression level of the gene; an image processing tool or library, such as Matplotlib or OpenCV, is used to convert the matrix into an image and visualize it.

[0164] Step S32: performing model treatment response migration processing on the gene expression image data and the treatment response gene marker set, thereby obtaining a convolutional neural network optimization model;

[0165] In an embodiment of the present invention, gene expression image data is integrated with a treatment response gene marker set; transfer learning is performed using the weights of an existing model, and the model is fine-tuned to adapt it to the gene expression image data and treatment response markers; the data is trained using the model after transfer learning, and during the training process, the model parameters are adjusted by monitoring the performance of the verification data.

[0166] Step S33: using a convolutional neural network optimization model to perform multi-scale pyramid difference remapping on the gene expression image data, thereby obtaining gene image feature refinement data;

[0167] In an embodiment of the present invention, a convolutional neural network is used to extract feature maps of multiple scales, which is achieved by using convolution kernels of different sizes or through pooling layers; feature maps from different scales are fused to obtain more comprehensive information; and differential features in the image are emphasized to better capture the details of the gene expression image.

[0168] Step S34: using a convolutional neural network optimization model to perform channel attention adjustment on the gene image feature refinement data, thereby obtaining key channel feature mapping data;

[0169] In an embodiment of the present invention, a channel attention mechanism is used to emphasize channels that play a key role in the treatment response, which can be achieved by weighted summing the channels using attention weights; through channel attention adjustment, a weighted feature map is obtained to highlight the channels that play a key role in the model's prediction of the treatment response.

[0170] Step S35: performing spatial attention adjustment on the key channel feature mapping data to obtain image feature enhancement data;

[0171] In an embodiment of the present invention, a spatial attention mechanism is used to emphasize areas in the image that play a key role in predicting treatment response; through spatial attention adjustment, a weighted feature map is obtained to highlight the spatial areas that play a key role in model prediction.

[0172] Step S36: performing parallel computing task division on the image feature enhancement data, thereby obtaining parallel computing task data; performing heterogeneous computing memory mapping on the image feature enhancement data, thereby obtaining GPU memory mapping data;

[0173] In the embodiment of the present invention, the image feature enhancement data is divided into multiple tasks to achieve parallel computing; the image data is mapped to the GPU memory to accelerate the parallel computing tasks.

[0174] Step S37: performing GPU heterogeneous computing acceleration processing according to the parallel computing task data and the GPU memory mapping data, thereby obtaining GPU heterogeneous computing acceleration data; performing GPU heterogeneous computing acceleration processing on the GPU heterogeneous computing acceleration data according to the resource balancing scaling data, thereby obtaining parallel computing performance indicators;

[0175] In the embodiment of the present invention, GPU heterogeneous computing is used to accelerate the processing calculation formula to perform parallel computing tasks on the GPU to accelerate the training or reasoning process of the model; and the parallel computing performance indicators, including the acceleration ratio and the calculation speed, are analyzed to evaluate the effect of GPU heterogeneous computing.

[0176] Preferably, in step S37, the GPU heterogeneous computing acceleration processing, wherein the formula for the GPU heterogeneous computing acceleration processing is specifically:

[0177] ;

[0178] In the formula, Indicates GPU heterogeneous computing acceleration data, Indicates the number of parallel computing tasks, Indicates the sequence number of the parallel computing task, Indicates parallel computing tasks, Indicates GPU memory mapping data for parallel computing tasks, represents the performance parameter associated with the parallel computing task, Indicates The complexity measurement parameters of parallel computing tasks, represents pi, Indicates The phase angle parameters of the parallel computing tasks, Indicates The computing resource consumption parameters of a parallel computing task.

[0179] The present invention constructs a calculation formula for GPU heterogeneous computing acceleration processing, which is used to perform GPU heterogeneous computing acceleration processing according to parallel computing task data and GPU memory mapping data; in the formula The part represents the change rate of the GPU memory mapping data of the parallel computing task. By calculating this change rate, the memory access mode can be optimized and the data access efficiency can be improved, thereby reducing the data transmission time and accelerating the computing task; The part represents the impact of the performance parameter associated with the parallel computing task on GPU acceleration. This logarithmic term converts the performance parameter into the corresponding acceleration, which helps evaluate the contribution of each task to the overall computing performance; Phase angle parameters and computing resource consumption parameters are partially introduced. Through the sine function, a mechanism for balancing the phase angle and computing resource consumption parameters is provided, which helps to adjust the execution order and resource allocation of tasks to maximize the use of GPU computing resources; the square operation of the entire formula helps to emphasize the impact of each task on the overall acceleration data, helps to eliminate the adverse effects of individual tasks on the overall performance, and makes the entire system more stable.

[0180] Preferably, step S32 includes the following steps:

[0181] Step S321: performing convolutional neural image training on the gene expression image data to obtain convolutional neural image training data, wherein the convolutional neural image training data includes convolutional neural image weight data and a convolutional neural network model;

[0182] Step S322: performing feature mapping on the gene expression image data using the convolutional neural image weight data, thereby obtaining gene image feature data;

[0183] Step S323: Associating the image features with the gene markers of the gene image feature data and the treatment response gene marker set, thereby obtaining feature marker association data;

[0184] Step S324: Use the feature tag association data to perform transfer learning processing on the convolutional neural network model to obtain a convolutional neural network optimization model.

[0185] The present invention can identify the pattern of gene expression through CNN training, which is helpful to discover specific features related to intravenous treatment, such as vascular structure or tissue morphology; through training, the weight data of the convolutional neural network is obtained, and these weights reflect the degree of importance of the model to different features, providing valuable information for subsequent analysis; the convolutional neural network model formed during the training process can be used in subsequent data processing to provide a basis for the prediction of treatment response. Using the convolutional neural image weight data, feature mapping can generate more abstract and advanced feature representations, strengthen the capture of important structures and patterns in gene expression images; during the feature mapping process, the dimension reduction and denoising of the original data can be achieved, and the efficiency and accuracy of subsequent analysis can be improved. Determining the correlation between gene image features and treatment response gene markers helps to understand which image features are closely related to treatment response; the correlation data provides a basis for establishing a prediction model, improves the prediction performance of treatment response, and helps to formulate a more personalized treatment plan. Using the previously trained network weights, the network is more easily adapted to the current task, and the convergence speed and performance of the model are improved; for the situation where the gene expression image data may be less, transfer learning helps to build a high-performance model on a small number of samples and improve the generalization ability of the model.

[0186] In an embodiment of the present invention, a convolutional neural network structure is designed, including a convolutional layer, a pooling layer, and a fully connected layer; the gene expression image data is preprocessed, such as image normalization and size adjustment; the convolutional neural network is trained using a prepared data set, and the weights and biases are iteratively optimized so that the network can accurately learn the features of the gene expression image data; after the training is completed, the convolutional neural network model and its corresponding weights are saved for subsequent use. The gene expression image data to be feature mapped is preprocessed in the same way as during training; the features of the gene expression image data are extracted using the trained convolutional neural network model, which is achieved by obtaining the output of the intermediate layer or by global average pooling. The feature data extracted from the gene expression image is associated with the treatment response gene marker set, and the relationship between the feature and the marker is analyzed using statistical methods and correlation coefficients. Transfer learning of the convolutional neural network using feature marker association data helps the network better adapt to new tasks, namely treatment response prediction; after transfer learning, the entire convolutional neural network is optimized to improve its performance in the treatment response prediction task.

[0187] Preferably, step S33 includes the following steps:

[0188] Step S331: constructing a multi-scale pyramid for the gene expression image data using a convolutional neural network optimization model, thereby obtaining multi-scale feature pyramid data;

[0189] Step S332: performing a pyramid pooling operation on the multi-scale feature pyramid data to obtain a pyramid pooling feature vector; performing attention weight calculation on the pyramid pooling feature vector to obtain attention weight data;

[0190] Step S333: performing feature weighted fusion on the multi-scale feature pyramid data and the attention weight data to obtain weighted fusion feature data; performing scale splicing on the weighted fusion feature data to obtain multi-scale image feature data;

[0191] Step S334: performing deconvolution processing and upsampling processing using the multi-scale image feature data, thereby obtaining high-resolution feature mapping data;

[0192] Step S335: extract mapping differences from the high-resolution feature mapping data and the gene image feature data, thereby obtaining gene image feature refinement data.

[0193] The convolutional neural network of the present invention can construct a multi-scale pyramid for gene expression images, thereby capturing feature information at different scales; it helps to fully understand the complex patterns of gene expression and improve the sensitivity of the model to gene correlation. The pyramid pooling operation helps to integrate multi-scale features and improve the perception of global information; it helps the model to better understand the key features at different scales in the gene expression image; by calculating the attention weight of the pyramid pooling feature vector, the model can automatically learn and focus on important areas in the gene expression image, enhance the perception of key features, and improve the model's understanding of the gene expression pattern. Feature weighted fusion enables the model to better integrate information at different scales by considering the weights of multi-scale image features; it helps the system to more comprehensively understand the patterns and laws in the gene expression image; scale splicing fuses the weighted fused features with the original multi-scale image features, further enhancing the model's understanding of the complex structure of the gene expression image and improving the capture of the correlation of features at different scales. Deconvolution and upsampling processing help to restore low-resolution feature maps to high resolution, improve the detail presentation of the gene expression image, and are of key significance for more accurately locating and analyzing the spatial distribution of gene expression. Mapping difference extraction helps to extract subtle changes and differences in gene image features by comparing high-resolution feature mapping data with gene image feature data, thereby describing important information in gene expression images in a more refined manner.

[0194] In an embodiment of the present invention, a pre-trained convolutional neural network (CNN) is used to construct a multi-scale pyramid for a gene expression image; feature maps at different scales are extracted to form multi-scale pyramid data. The multi-scale pyramid data is processed using a pyramid pooling operation to obtain a pyramid pooling feature vector; an attention mechanism is applied to the pyramid pooling feature vector to calculate the attention weight of each feature. The multi-scale feature pyramid data is weighted fused using the attention weight data to obtain weighted fused feature data; the weighted fused feature data is scaled and spliced ​​with the original multi-scale feature pyramid data to form multi-scale image feature data. The multi-scale image feature data is deconvolved to expand the feature map size; the deconvolved data is upsampled to obtain high-resolution feature mapping data. The high-resolution feature mapping data and the gene image feature data are differentially extracted to obtain gene image feature refinement data.

[0195] Preferably, step S34 includes the following steps:

[0196] Step S341: introducing a channel attention module into the convolutional neural network optimization model, thereby obtaining a convolutional neural network channel model;

[0197] Step S342: using the convolutional neural network channel model to perform global average pooling on the gene image feature refinement data, thereby obtaining global average pooling layer data;

[0198] Step S343: inputting the global average pooling layer data into a preset channel weight generator for channel weight learning, thereby obtaining channel weight data;

[0199] Step S344: weighting the gene image feature refinement data by the treatment response contribution channel using the channel weight data, thereby obtaining contribution channel weight data;

[0200] Step S345: Adjust the contribution channel weight data according to the gradient descent optimization algorithm and the convolutional neural network channel model to obtain key channel feature mapping data.

[0201] The channel attention module of the present invention can automatically learn and adjust the weights of different channels, so that the network pays more attention to the gene features that are important for intravenous treatment response; such optimization helps to improve the sensitivity of the model to gene images, thereby more accurately capturing the treatment response related information in the patient's gene expression spectrum. Through global average pooling, the high-dimensional feature mapping of the gene image can be effectively converted into fixed-dimensional features, reducing the data dimension and reducing the computational burden of the model; it helps to improve the generalization ability of the model, prevent overfitting, and simplify the complexity of subsequent channel weight learning. Through the channel weight learner, the system can automatically learn the weight of each channel in the gene image, thereby accurately capturing the contribution of each channel to the patient's treatment response; this adaptive channel weight learning mechanism helps to better understand the feature expression of different channels in the gene image, and provide more targeted information for subsequent processing. By introducing the treatment response contribution channel empowerment, the system can assign corresponding weights to each channel according to its contribution to the treatment response; it helps to focus on the channels that have a significant impact on the treatment response, thereby improving the sensitivity and interpretability of the model. Using the gradient descent optimization algorithm to adjust the contributing channel weights can further optimize the performance of the model; by continuously updating the channel weights based on the gradient information, the model can be better adapted to the training data, improving the generalization ability of the model while ensuring that the model's predictions of treatment responses have higher accuracy and stability; this iterative adjustment process helps the model to continuously optimize itself and gradually converge to a more effective parameter configuration.

[0202] In an embodiment of the present invention, a channel attention module is added to the model, including calculating the importance weight of the channel, using the Sigmoid function to generate the weight; and multiplying the generated attention weight by the original convolutional neural network channel output. The feature map after the channel attention module is globally averaged pooled. This step uses the average value of the entire feature map as the output of each channel to reduce the spatial dimension of the data. The output of the global average pooling layer is input into a preset channel weight generator; the channel weight generator can be a fully connected neural network layer, and its goal is to learn to generate channel weights. The original gene image feature refinement data is weighted by the treatment response contribution channel using the channel weight data, including a channel-by-channel multiplication operation. The contribution channel weight data is weighted according to the gradient descent optimization algorithm and the convolutional neural network channel model, including calculating the loss function and back-propagating the gradient to update the channel weight.

[0203] Preferably, step S4 comprises the following steps:

[0204] Step S41: obtaining treatment parameters and real-time physiological parameters of the patient through the sensor network; performing interactive impact analysis on the treatment parameters and the real-time physiological parameters of the patient, thereby obtaining interactive impact analysis data;

[0205] Step S42: performing physiologically relevant edge computing on the treatment parameters to obtain physiologically relevant edge computing data; performing event-driven analysis and marking on the physiologically relevant edge computing data to obtain event-driven analysis data;

[0206] Step S43: performing time series analysis on the treatment parameters to obtain periodic variation data; performing time window cluster analysis on the periodic variation data to obtain time window cluster data;

[0207] Step S44: performing time-driven fusion of treatment parameters according to the event-driven analysis data and the time window clustering data, thereby obtaining time-driven data of treatment parameters;

[0208] Step S45: Predicting the real-time physiological parameter change trend of the treatment parameters and the patient's real-time physiological parameters according to the interactive impact analysis data and the treatment parameter time-driven data, thereby obtaining real-time physiological parameter change trend data.

[0209] The present invention obtains the patient's real-time physiological parameters, including key indicators of heart rate and blood pressure, which helps doctors understand the patient's physiological state more accurately; at the same time, obtaining treatment parameters, such as drug use and treatment plans, provides a basis for personalized treatment. Edge computing allows physiological-related calculations to be performed on the device side, reducing data transmission delays and providing more real-time feedback; event-driven analysis can identify key events and further strengthen the real-time processing of important information. Physiological-related edge computing can reduce the data transmission requirements of the central cloud platform and reduce bandwidth pressure; event-driven analysis refines sensor data and only transmits key events, reducing the burden of data transmission. Time series analysis helps to identify the changing trends of treatment parameters, including long-term trends and periodic changes; this provides doctors with a deeper understanding of treatment effects and patient conditions. Time window cluster analysis can divide patients into different groups according to the change patterns of treatment parameters; it helps to optimize personalized treatment plans and provide patients with more accurate medical services. Fusion of event-driven analysis data with time window cluster data can more comprehensively understand the relationship between the temporal changes of treatment parameters and the occurrence of events, helping the medical team to better understand the patient's condition. By combining the patient's physiological parameters, treatment parameters and their temporal changes, we can achieve personalized predictions of future changes in the patient's physiological state, providing doctors with earlier intervention opportunities.

[0210] In the embodiment of the present invention, firstly, the sensor network deployed on the patient is used to obtain the treatment-related parameters and the patient's real-time physiological parameters, which involves various sensor technologies, such as heart rate monitors, sphygmomanometers, and temperature sensors; the deployment of the sensor network needs to ensure the accuracy and real-time performance of the data. After obtaining the treatment parameters, physiologically relevant edge computing is performed on these parameters; this step usually includes applying mathematical models or machine learning algorithms to reveal the correlation between the treatment parameters and the patient's physiological state; for example, correlation analysis, regression analysis, or deep learning models can be used. After obtaining the data of physiologically relevant edge computing, event-driven analysis is performed to identify potential physiological events and trends; this may include a sudden increase in heart rate and an increase in blood pressure; these events are marked for subsequent analysis and processing. Time series analysis is performed on the treatment parameters to reveal their changing patterns over time; including periodic changes, trends, or sudden events; time series analysis may involve statistical methods and spectral analysis techniques. Combined with event-driven analysis data and time window clustering data, time-driven fusion of treatment parameters is performed; the purpose of this step is to comprehensively consider the physiological change trends of patients under specific events or time windows, and provide support for more accurate treatment parameter adjustment. Finally, based on the interaction analysis data and the time-driven data of treatment parameters, the real-time trend of changes in the patient's physiological parameters is predicted; this can be achieved using machine learning models and time series prediction algorithm techniques.

[0211] The present invention obtains gene expression data through high-throughput genome sequencing, combined with gene expression spectrum analysis technology, to accurately and comprehensively understand the genetic information of the patient's venous system; it provides a personalized basis for subsequent intravenous treatment management, making the treatment plan more accurate and effective. Through cloud computing technology, an elastic computing resource scheduling system is built to realize intelligent management of computing resources; the statistics of resource scheduling system data help optimize the use of computing resources, improve the flexibility and performance of the system, and ensure that the computing tasks of gene expression data can be efficiently processed during peak and trough periods. By accelerating the processing of gene expression data through GPU heterogeneous computing, the data processing speed can be significantly improved, and the identification process of intravenous treatment response gene markers can be accelerated; GPU heterogeneous computing accelerated processing helps to evaluate system performance, ensure that GPU accelerated computing plays the maximum benefit in the entire process, and improve computing efficiency. Through the sensor network, real-time monitoring of the patient's physiological parameter change trend can obtain rich biological information. This helps to timely discover changes in the patient's physiological state and provide real-time data support for personalized treatment; time and event-driven treatment parameter extraction reduces the data transmission demand for the central cloud, and improves real-time and data privacy. Based on the treatment response gene marker set and real-time physiological parameter change trend data, the system can extract personalized intravenous treatment management parameters; these parameters are adjusted according to the patient's molecular characteristics and physiological state, making the treatment more accurate, maximizing the treatment effect, and reducing side effects. A cloud platform with high performance and personalized characteristics is constructed using parallel computing performance indicators and personalized intravenous treatment management parameters; it provides a powerful tool for the medical team to support large-scale data processing and the formulation of personalized treatment plans, and promotes the cutting-edge development of intravenous treatment management. Therefore, the present invention provides a method for constructing an intravenous treatment management cloud platform, which improves the system's performance, flexibility, and resource utilization efficiency through node load balancing weight scaling and GPU heterogeneous computing acceleration processing.

[0212] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0213] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for constructing a cloud platform for intravenous therapy management, characterized in that: The following steps are involved: Step S1: Gene expression pathway enrichment analysis is performed on the patient's venous system by gene sequencing technology to obtain gene expression data; gene expression profile analysis is performed on the gene expression data and intravenous treatment response gene markers are performed to obtain a treatment response gene marker set, wherein step S1 is specifically as follows: Step S11: performing gene expression pathway enrichment analysis on the patient's venous system through gene sequencing technology to obtain gene expression data; Step S12: performing sparse calculation of gene expression on the gene expression data to obtain gene expression data; performing differential expression analysis on the gene expression data to obtain differentially expressed gene set data; wherein the formula for sparse calculation of gene expression in step S12 is specifically: In the formula, G ij represents the element value in the gene expression matrix, i represents the number of rows in the gene expression matrix, j represents the number of columns in the gene expression matrix, and x i The initial expression value of the gene in the i-th row, e represents the natural base, y j represents the initial expression value of the gene in row j, n represents the measurement value of the change trend, z i represents the sparseness of the gene in the i-th row, k represents the index of the correlation coefficient of the gene, a k represents the association weight parameter of the gene in the i-th row, b k represents the expression adjustment coefficient of the gene in the i-th row, c k represents the biological impact factor of the gene in the i-th row, d k represents the interaction parameter of the gene in the i-th row, t represents the gene sampling time variable, and w i represents the regulatory factor of the i-th row, v j Characteristic variables representing the values ​​of trigonometric functions; Step S13: constructing a gene expression network using the differentially expressed gene set data, thereby obtaining gene expression network data; Step S14: performing functional enrichment analysis on the differentially expressed gene set data according to the gene expression network data, thereby obtaining enriched functional annotation data; Mining key pathways based on enriched functional annotation data to obtain key pathway gene set data; Step S15: Generate a gene expression profile based on the gene expression network data and the key pathway gene set data, thereby obtaining gene expression profile data; Step S16: performing gene expression pattern classification on the gene expression profile data, thereby obtaining expression pattern classification data; performing heterogeneity identification on the expression pattern classification data, thereby obtaining heterogeneous gene expression data; Step S17: performing intravenous treatment response gene labeling according to the heterogeneous gene expression data and the differentially expressed gene set data, thereby obtaining a treatment response gene marker set; Step S2: Using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data; Node load balancing weight scaling is performed based on resource scheduling system data and gene expression data to obtain resource balancing scaling data; Step S2 includes the following steps: Step S21: Using cloud computing technology to build an elastic computing resource scheduling system, thereby obtaining resource scheduling system data; Step S22: monitoring the resource scheduling system based on the resource scheduling system data, thereby obtaining resource utilization data; Step S23: assigning task priority weights according to the gene expression data, thereby obtaining task priority weight data; Step S24: performing adaptive load balancing according to the resource utilization data and the task priority weight data, thereby obtaining node load balancing data; Step S25: Automatically scaling records are performed according to the node load balancing data, thereby obtaining automatic scaling record data; Step S26: performing resource balancing scaling processing based on the automatic scaling record data and the node load balancing data, thereby obtaining resource balancing scaling data; Step S3: Perform multi-scale pyramid feature refinement on the gene expression data to obtain gene image feature refinement data; perform channel and spatial attention adjustment on the gene image feature refinement data to obtain image feature enhancement data; perform GPU heterogeneous computing acceleration processing on the image feature enhancement data according to resource balanced scaling data to obtain parallel computing performance indicators, wherein step S3 is specifically as follows: Step S31: performing gene imaging processing on the gene expression data to obtain gene expression image data; Step S32: performing model treatment response migration processing on the gene expression image data and the treatment response gene marker set, thereby obtaining a convolutional neural network optimization model; Step S33: using a convolutional neural network optimization model to perform multi-scale pyramid difference remapping on the gene expression image data, thereby obtaining gene image feature refinement data; Step S34: using a convolutional neural network optimization model to perform channel attention adjustment on the gene image feature refinement data, thereby obtaining key channel feature mapping data; Step S35: performing spatial attention adjustment on the key channel feature mapping data to obtain image feature enhancement data; Step S36: performing parallel computing task division on the image feature enhancement data, thereby obtaining parallel computing task data; performing heterogeneous computing memory mapping on the image feature enhancement data, thereby obtaining GPU memory mapping data; Step S37: GPU heterogeneous computing acceleration processing is performed according to the parallel computing task data and the GPU memory mapping data, so as to obtain GPU heterogeneous computing acceleration data; GPU heterogeneous computing acceleration processing is performed on the GPU heterogeneous computing acceleration data according to the resource balancing scaling data, so as to obtain parallel computing performance indicators; wherein the GPU heterogeneous computing acceleration processing in step S37, wherein the formula for GPU heterogeneous computing acceleration processing is specifically: In the formula, A represents GPU heterogeneous computing acceleration data, m represents the number of parallel computing tasks, q represents the sequence number of parallel computing tasks, and p q represents the qth parallel computing task, X q Represents the GPU memory mapping data of the qth parallel computing task, Y q represents the performance parameter associated with the parallel computing task, z q represents the complexity metric parameter of the qth parallel computing task, π represents the circumference of a circle, and W q represents the phase angle parameter of the qth parallel computing task, V q Indicates the computing resource consumption parameter of the qth parallel computing task; Step S4: real-time monitoring of biological nodes of the patient through the sensor network and extraction of time and event-driven treatment parameters, thereby obtaining real-time physiological parameter change trend data; Step S5: extracting personalized intravenous therapy management parameters based on the treatment response gene marker set and the real-time physiological parameter change trend data, thereby obtaining personalized intravenous therapy management parameters; Step S6: Build a cloud platform using parallel computing performance indicators and personalized intravenous therapy management parameters to obtain cloud platform construction data.

2. The method for constructing a cloud platform for intravenous therapy management according to claim 1, characterized in that: Step S32 includes the following steps: Step S321: performing convolutional neural image training on the gene expression image data to obtain convolutional neural image training data, wherein the convolutional neural image training data includes convolutional neural image weight data and a convolutional neural network model; Step S322: performing feature mapping on the gene expression image data using the convolutional neural image weight data, thereby obtaining gene image feature data; Step S323: Associating the image features with the gene markers of the gene image feature data and the treatment response gene marker set, thereby obtaining feature marker association data; Step S324: Use the feature tag association data to perform transfer learning processing on the convolutional neural network model to obtain a convolutional neural network optimization model.

3. The method for constructing a cloud platform for intravenous therapy management according to claim 1, characterized in that: Step S33 includes the following steps: Step S331: constructing a multi-scale pyramid for the gene expression image data using a convolutional neural network optimization model, thereby obtaining multi-scale feature pyramid data; Step S332: performing a pyramid pooling operation on the multi-scale feature pyramid data to obtain a pyramid pooling feature vector; performing attention weight calculation on the pyramid pooling feature vector to obtain attention weight data; Step S333: performing feature weighted fusion on the multi-scale feature pyramid data and the attention weight data to obtain weighted fusion feature data; performing scale splicing on the weighted fusion feature data to obtain multi-scale image feature data; Step S334: performing deconvolution processing and upsampling processing using the multi-scale image feature data, thereby obtaining high-resolution feature mapping data; Step S335: extract mapping differences from the high-resolution feature mapping data and the gene image feature data, thereby obtaining gene image feature refinement data.

4. The method for constructing a cloud platform for intravenous therapy management according to claim 1, characterized in that: Step S34 includes the following steps: Step S341: introducing a channel attention module into the convolutional neural network optimization model, thereby obtaining a convolutional neural network channel model; Step S342: using the convolutional neural network channel model to perform global average pooling on the gene image feature refinement data, thereby obtaining global average pooling layer data; Step S343: inputting the global average pooling layer data into a preset channel weight generator for channel weight learning, thereby obtaining channel weight data; Step S344: weighting the gene image feature refinement data by the treatment response contribution channel using the channel weight data, thereby obtaining contribution channel weight data; Step S345: Adjust the contribution channel weight data according to the gradient descent optimization algorithm and the convolutional neural network channel model to obtain key channel feature mapping data.

5. The method for constructing a cloud platform for intravenous therapy management according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: obtaining treatment parameters and real-time physiological parameters of the patient through the sensor network; performing interactive impact analysis on the treatment parameters and the real-time physiological parameters of the patient, thereby obtaining interactive impact analysis data; Step S42: performing physiologically relevant edge computing on the treatment parameters to obtain physiologically relevant edge computing data; performing event-driven analysis and marking on the physiologically relevant edge computing data to obtain event-driven analysis data; Step S43: performing time series analysis on the treatment parameters to obtain periodic variation data; performing time window cluster analysis on the periodic variation data to obtain time window cluster data; Step S44: performing time-driven fusion of treatment parameters according to the event-driven analysis data and the time window clustering data, thereby obtaining time-driven data of treatment parameters; Step S45: Predicting the real-time physiological parameter change trend of the treatment parameters and the patient's real-time physiological parameters according to the interactive impact analysis data and the treatment parameter time-driven data, thereby obtaining real-time physiological parameter change trend data.

Images