Remote intelligent disease diagnosis and treatment system and method based on block chain

Abstract

The invention relates to the technical field of intelligent diagnosis and treatment, and discloses a remote intelligent disease diagnosis and treatment system and method based on a block chain. The method comprises the steps that multi-modal physiological data such as real-time vital signs, historical medical record texts and medical images of a patient are collected, an encrypted data packet is generated, distributed storage is conducted through a block chain node network, and a data hash value and a timestamp are generated; and calling an intelligent contract to extract cross-modal features, generating a standardized feature vector set containing a multi-feature association identifier, and constructing a diagnosis and treatment decision tree capable of dynamically adjusting branch paths according to the standardized feature vector set. Verifying decision validity through a consensus mechanism, and generating a treatment instruction sequence containing drug configuration, instrument operation and a follow-up visit period; and monitoring feedback data in real time and updating the feedback data to the block chain, triggering an intelligent contract to generate a correction scheme when the deviation exceeds a threshold value, and writing the difference data as a new transaction record into the block chain, thereby realizing multi-modal data security management and control and dynamic precise diagnosis and treatment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent diagnosis and treatment, in particular to a remote disease intelligent diagnosis and treatment system and method based on a block chain. BACKGROUND

[0002] With the rapid development of medical informatization, remote disease diagnosis and treatment has become an important way to solve uneven distribution of medical resources and improve medical service accessibility due to its advantage of breaking through time and space limitations. In the remote diagnosis and treatment scene, the physiological data of patients presents multi-source and heterogeneous characteristics, covering real-time dynamic vital sign data, structured historical medical record text data, and non-structured medical image data and other multi-modal information. These data are the core basis for realizing accurate diagnosis and treatment.

[0003] However, the current remote diagnosis and treatment system still faces many problems to be solved. In terms of data security and management, the traditional centralized data storage mode has significant security risks. Multi-modal physiological data contains a large amount of patient privacy information, and centralized storage is vulnerable to hacker attacks, data tampering or leakage. At the same time, the data formats of different medical institutions are not unified, forming a "data island", which makes it difficult to realize cross-institutional data sharing and collaborative diagnosis and treatment. In terms of data processing and feature extraction, the time sequence dynamics, spatial correlation and semantic complexity of cross-modal data bring great challenges to feature fusion. Existing methods often fail to effectively mine the internal relations between different types of data, resulting in lack of standardization and comprehensiveness of the extracted features, which cannot provide reliable support for diagnosis and treatment decisions.

[0004] In terms of diagnosis and treatment decision mechanism, the traditional diagnosis and treatment model adopts fixed structure decision logic, which is difficult to dynamically adjust the decision path according to the changes of real-time physiological data of patients, and cannot adapt to the dynamic evolution process of the disease, which easily leads to the lag or limitation of diagnosis and treatment scheme. In addition, the effectiveness verification of diagnosis and treatment scheme lacks a unified authoritative mechanism, and the recognition degree of diagnosis and treatment scheme by different medical subjects is different, which may cause the problem of non-standard diagnosis and treatment behavior. At the same time, the feedback data of patients after executing the diagnosis and treatment scheme cannot be processed and applied in time, which cannot quickly trigger the optimization adjustment of diagnosis and treatment scheme, affecting the continuous improvement of treatment effect. These problems jointly restrict the accuracy, safety and efficiency of remote disease diagnosis and treatment, and a new technical system needs to be built to break through. SUMMARY

[0005] The purpose of the present application is to provide a remote disease intelligent diagnosis and treatment system and method based on a block chain to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a remote disease intelligent diagnosis and treatment method based on a block chain, which comprises:

[0007] Collecting multi-modal physiological data of a patient and generating an encrypted data package, the multi-modal physiological data including real-time vital sign data, historical medical record text data and medical image data;

[0008] Distributing the encrypted data package through a blockchain node network for distributed storage, and generating data hash values and time stamps at each node;

[0009] Calling a smart contract to perform cross-modal feature extraction on the encrypted data package, generating a standardized feature vector set containing the association identifiers of time series features, spatial features and semantic features;

[0010] Building a dynamic diagnosis and treatment decision tree based on the standardized feature vector set, the dynamic diagnosis and treatment decision tree dynamically adjusting the branch path according to the feature weight;

[0011] Verifying the node validity of the dynamic diagnosis and treatment decision tree through a consensus mechanism, and marking the decision path that passes the verification as an executable diagnosis and treatment scheme, generating a treatment instruction sequence containing drug configuration parameters, instrument operation steps and follow-up cycles;

[0012] Real-time monitoring of feedback data of the patient executing the treatment instruction sequence, and updating to the blockchain node network;

[0013] When the feedback data deviates from the expected efficacy by more than a threshold value, triggering the smart contract to regenerate a revised diagnosis and treatment scheme;

[0014] Writing the difference data between the revised diagnosis and treatment scheme and the original diagnosis and treatment scheme as a new transaction record to the blockchain.

[0015] Preferably, the calling of the smart contract to perform cross-modal feature extraction on the encrypted data package comprises:

[0016] Decomposing the real-time vital sign data in the encrypted data package into equally spaced time series segments, and calculating the mean, variance and peak frequency of each segment;

[0017] Performing keyword entity recognition and context relationship analysis on the historical medical record text data to generate a structured semantic graph;

[0018] Extracting the gray scale distribution matrix and edge gradient features of the medical image data to construct a three-dimensional spatial feature vector;

[0019] Aligning the time series segment features, structured semantic graph and three-dimensional spatial feature vector by time axis, and fusing them into a cross-modal association matrix.

[0020] Preferably, the building of the dynamic diagnosis and treatment decision tree based on the standardized feature vector set comprises:

[0021] According to the Euclidean distance of the eigenvectors in the cross-modal correlation matrix, an initial decision node is divided;

[0022] Calculate the mutual information entropy of each eigenvector and disease classification label, and dynamically allocate the weight coefficient of the decision branch;

[0023] When new feedback data is added, the splitting priority of the decision node is adjusted based on the weight coefficient.

[0024] Preferably, the node validity of the dynamic diagnosis and treatment decision tree is verified by a consensus mechanism, comprising:

[0025] The node hash value of the decision path is broadcast to a preset number of verification nodes;

[0026] Each verification node compares the matching degree of the locally stored eigenvector and the node hash value;

[0027] When the matching degree exceeds a preset proportion, mark the node as a valid node and generate a digital signature.

[0028] Preferably, the treatment instruction sequence is generated, comprising:

[0029] Parsing the disease staging identifier and complication risk level in the executable diagnosis and treatment scheme;

[0030] Match the treatment template in the preset clinical guideline library to generate a basic instruction set;

[0031] According to the individual difference parameters of the patient, the basic instruction set is prioritized and dosed.

[0032] Preferably, the real-time monitoring of the feedback data of the patient executing the treatment instruction sequence comprises:

[0033] Collect the concentration change curve of the metabolic product after the patient takes medicine and the physiological index fluctuation data;

[0034] Record the mechanical parameters and time intervals during the operation of the equipment through the Internet of Things device;

[0035] Pack the metabolic product concentration change curve, physiological index fluctuation data and mechanical parameters into a feedback data block.

[0036] Preferably, the smart contract is triggered to regenerate a revised diagnosis and treatment scheme, comprising:

[0037] Compare the standard deviation and deviation direction of the feedback data block and the expected efficacy data;

[0038] When the standard deviation exceeds a threshold value, the feature extraction module is called to recalculate the cross-modal correlation matrix;

[0039] Reconstruct the branch path of the diagnosis and treatment decision tree based on the updated cross-modal correlation matrix.

[0040] Preferably, the difference data between the modified diagnosis and treatment scheme and the original diagnosis and treatment scheme is written into the blockchain as a new transaction record, including:

[0041] Extract the change instruction item and parameter adjustment amount in the difference data;

[0042] Generate a transaction structure containing a change timestamp, an executing doctor digital identity, and a difference hash value;

[0043] Write the transaction structure into a new block through a proof-of-work mechanism.

[0044] Preferably, the method further comprises:

[0045] Periodically scan the diagnosis and treatment schemes of similar cases stored in the blockchain;

[0046] Cluster analysis of the feature vector distribution law of the diagnosis and treatment schemes of similar cases;

[0047] Optimize the feature weight distribution strategy in the smart contract according to the distribution law.

[0048] Preferably, the present application further comprises a remote disease intelligent diagnosis and treatment system based on blockchain, which comprises a processor and a memory, the memory stores a computer program, and the processor executes the computer program to realize the remote disease intelligent diagnosis and treatment method based on blockchain as described above.

[0049] Compared with the prior art, the present application has the following beneficial effects:

[0050] By collecting multi-modal physiological data of patients and generating encrypted data packets, comprehensive integration and secure packaging of different types of medical data are realized, covering key information such as real-time vital signs, historical medical records, and medical images, ensuring the integrity of the data foundation relied on by diagnosis and treatment decisions. The encrypted data packets are stored in a distributed manner through a blockchain node network, and data hash values and timestamps are generated. With the help of the decentralized nature of the blockchain, the security risks brought about by centralized storage are avoided, and each node retains a copy of the data, preventing data loss or information leakage caused by attacks on a single node. Through the hash value and timestamp, the data is traceable throughout its life cycle, ensuring the authenticity and non-tamperability of the data, and providing a solid technical guarantee for patient privacy protection.

[0051] The smart contract is called to perform cross-modal feature extraction and generate a standardized feature vector set, which integrates the correlation identification of time sequence features, spatial features and semantic features. The automatic execution characteristics of the smart contract ensure the efficiency and consistency of the feature extraction process, effectively breaking through the bottleneck of inconsistent formats and insufficient correlation mining in traditional cross-modal data processing, fully releasing the value of different types of data, and providing standardized and comprehensive feature support for subsequent diagnosis and treatment decisions. The dynamic diagnosis and treatment decision tree constructed based on the standardized feature vector set can dynamically adjust the branch path according to the feature weight, accurately match the individual physiological characteristics and disease change of the patient, and break through the limitations of the traditional fixed structure decision model, making the diagnosis and treatment plan more personalized and adaptive, and better coping with the differences of different patients and the dynamic evolution process of the disease.

[0052] The node effectiveness of the dynamic diagnosis and treatment decision tree is verified through the consensus mechanism, and the distributed consensus among the blockchain nodes ensures the authority and reliability of the diagnosis and treatment plan, avoiding the deviation caused by the subjective judgment of a single medical subject, and making the generated treatment instruction sequence more scientific. The treatment instruction sequence explicitly includes drug configuration parameters, instrument operation steps and follow-up cycle, etc., providing clear guidance for the standardized execution of remote diagnosis and treatment, and facilitating patients or primary medical personnel to accurately implement the diagnosis and treatment plan.

[0053] Real-time monitoring of feedback data of patients executing the treatment instruction sequence and updating to the blockchain node network forms a closed-loop management of the diagnosis and treatment process. When the deviation between the feedback data and the expected efficacy exceeds the threshold, the smart contract is triggered to regenerate the revised diagnosis and treatment plan, realizing the dynamic optimization of the diagnosis and treatment plan, and timely responding to the changes of the patient's condition, avoiding poor treatment effect caused by lagging behind. The difference data between the revised diagnosis and treatment plan and the original diagnosis and treatment plan is written into the blockchain as a new transaction record, further improving the traceability of the diagnosis and treatment process, not only providing complete data basis for subsequent medical quality evaluation and diagnosis and treatment experience summary, but also laying a data foundation for cross-institutional medical collaboration and academic exchange, promoting the development of the remote diagnosis and treatment system in a more standardized, more accurate and more efficient direction. At the same time, the decentralized characteristics of the blockchain also provide a trust basis for collaborative diagnosis and treatment between different medical institutions, breaking the "data island", promoting the efficient circulation and sharing of medical resources, and improving the quality and accessibility of the overall remote medical service. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The working principle diagram of the remote disease intelligent diagnosis and treatment method based on blockchain described in the present application;

[0055] Figure 2 The flowchart of calling the smart contract for cross-modal feature extraction;

[0056] Figure 3 Flowchart for real-time monitoring of patient feedback data. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0058] Please refer to Figure 1 The present application provides a blockchain-based remote disease intelligent diagnosis and treatment method, which comprises: through the integration of multi-modal data acquisition, blockchain storage, intelligent contract analysis and dynamic decision optimization, safe and reliable remote diagnosis and treatment is realized. The method starts with the collection of multi-modal physiological data of patients, including real-time vital sign data, historical medical record text data and medical image data. The collection equipment such as wearable sensors, electronic health record systems and medical imaging instruments work synchronously, and the original data is encrypted and packaged to generate an encrypted data packet. The encryption process uses an asymmetric encryption algorithm to ensure data privacy. The encrypted data packet is then uploaded to the blockchain node network, which is composed of multiple distributed nodes. Each node independently stores a copy of the encrypted data packet and calculates a unique data hash value, while adding an unforgeable timestamp. The smart contract is automatically called to decrypt the encrypted data packet and extract cross-modal features, outputting a standardized feature vector set containing time series features, spatial features and semantic features. Based on the standardized feature vector set, a dynamic diagnosis and treatment decision tree is constructed, and the branch path of the decision tree is dynamically adjusted according to the weight of the feature vector to reflect the changes in the patient's state. The consensus mechanism verifies the effectiveness of each node in the dynamic diagnosis and treatment decision tree, and the decision path that passes the verification is marked as an executable diagnosis and treatment plan, and a treatment instruction sequence is generated, which includes drug configuration parameters, instrument operation steps and follow-up cycle. The feedback data of the patient executing the treatment instruction sequence is monitored in real time, including physiological indicator changes and operation records. These data are updated to the blockchain node network. When the deviation of the feedback data from the expected efficacy exceeds the preset threshold, the smart contract triggers the generation of a revised diagnosis and treatment plan. The difference data between the revised diagnosis and treatment plan and the original diagnosis and treatment plan is written as a new transaction record in the blockchain, ensuring the transparency and traceability of the diagnosis and treatment process.

[0059] Embodiment 1: Please refer to Figure 2The real-time vital sign data in the encrypted data packet is decomposed into equally spaced time segments, and the division of the time segments is based on the inherent sampling frequency of the data acquisition device. For example, the millisecond-level continuous waveform data from an electrocardiogram monitor is divided into fixed-length segments for analysis. Each time segment needs to calculate its mean, variance, and peak frequency multiple statistical features. The mean reflects the baseline level of the physiological signal in that time period, the variance characterizes the signal's fluctuation range and stability, and the peak frequency identifies the dominant rhythm or periodic pattern from the frequency domain through fast Fourier transform. These features collectively constitute a quantitative description of the dynamic changes in vital signs. Historical medical record text data performs keyword entity recognition and context relationship analysis. The entity recognition process uses a pre-trained medical named entity recognition model to extract key medical terms from unstructured text, including disease names, drug names, clinical symptoms, and examination index values. Context relationship analysis establishes semantic associations between extracted entities through dependency syntax analysis techniques, such as determining the treatment relationship between a certain drug and a specific symptom, or the support relationship between a certain examination result and a diagnosis, thereby generating a structured semantic graph with entities as nodes and semantic relationships as edges. The processing of medical image data focuses on extracting its gray scale distribution matrix and edge gradient features. The gray scale distribution matrix is obtained by normalizing the intensity values of the image pixels and regridding them, reflecting the density difference information of tissues or organs. The edge gradient features are calculated by convolving the image with edge detection operators such as Sobel or Canny, outlining the contours and boundaries of anatomical structures. Based on these basic features, a three-dimensional spatial feature vector can be constructed to express the spatial relationship of the regions of interest in the image.

[0060] The time segment features, structured semantic graph, and three-dimensional spatial feature vector from different modalities need to be aligned along the time axis. The alignment operation needs to address the differences in timestamp accuracy and sampling interval between different sources of data. Linear interpolation or nearest neighbor interpolation algorithms are used to unify non-uniformly sampled data sequences to consistent time points. The aligned multi-modal data is fused into a cross-modal correlation matrix. The fusion process involves complex tensor operations, such as the outer product of time series feature vectors, semantic graph embedding vectors, and spatial feature vectors to form a high-order tensor, which is then reduced in dimension through tensor decomposition or expansion techniques to a comprehensive two-dimensional matrix. The rows of this cross-modal correlation matrix usually correspond to different feature dimensions, and the columns correspond to the unified time points. Each element in the matrix not only contains the feature information of a single modality, but also encodes the correlation strength of different modal features at the same time. The cross-modal correlation matrix constitutes the core data structure of the standardized feature vector set, which facilitates efficient processing by subsequent machine learning models.

[0061] The entire feature extraction process is coded in the logic of the smart contract, which is triggered to execute on the blockchain node, and calls the computing resources deployed on the node to complete the above complex feature engineering tasks. The feature extraction process strictly follows international standard medical data standards, such as the Health Level Seven standard and the Digital Imaging and Communications in Medicine standard, to ensure that the generated feature vectors have interoperability between different systems. The original encrypted data packet used for feature extraction is temporarily decrypted in the smart contract, and after the feature extraction task is completed, the generated intermediate data and the final feature vector are encrypted like the original data. The encrypted data packet is re-written to the blockchain distributed storage to prevent any unauthorized data access. The cross-modal correlation matrix is not a static data structure, when new patient data is added to the system through the blockchain node network, the cross-modal correlation matrix will be updated through the incremental learning algorithm. The incremental learning algorithm only calculates the feature changes based on new data and adjusts the corresponding part of the matrix, avoiding the resource consumption of recalculating the entire historical data set. This mechanism maintains the consistency of the feature representation when facing new information. The execution of the smart contract is atomic, the success or failure of the feature extraction process is recorded as a complete transaction on the blockchain, ensuring the reliability and auditability of the data processing process. The quality of the feature vector is directly related to the accuracy of the subsequent diagnosis and treatment decision, therefore the smart contract integrates a feature quality verification module, which checks the valid range of feature values, whether there are a large number of missing values, and whether the correlation between features is reasonable. Unqualified feature generation will trigger a re-computation process.

[0062] The construction of the dynamic diagnosis decision tree starts with a deep analysis of the cross-modal correlation matrix in the standardized feature vector set, which contains the complex internal relationship between the multi-modal physiological data of the patient. The Euclidean distance of the feature vector is accurately calculated, which measures the relative position relationship of different feature points in multi-dimensional space. Feature points with smaller distance values are considered to have higher physiological state similarity. The initial decision node is divided according to the preset distance threshold, which is obtained by statistical analysis of the distance between the feature clustering centers of the historical healthy population and the patient population. Each initial decision node represents a preliminary disease classification judgment or health status interval. The mutual information entropy of each feature vector and the known disease classification label is evaluated, which quantifies the contribution of a single feature or feature combination to reducing diagnostic uncertainty. Features with high mutual information entropy are considered to have stronger classification discrimination ability. The weight coefficient of the dynamic allocation decision branch is based on the calculation result of the mutual information entropy, which directly determines the influence of the feature in the decision tree branch path selection. New feedback data flows into the system from the blockchain node network in real time, which contains the physiological response information of the patient to the previous treatment instructions. The weight coefficient adjustment mechanism is activated, which dynamically updates the weight coefficient according to the degree of fit between new data and original feature vectors. If the original weight coefficient is high but inconsistent with new data, the coefficient will be adjusted downward, and vice versa. The splitting priority of the decision node is reordered based on the updated weight coefficient, which determines the order of feature selection in the decision tree growth process. High-priority branch paths represent the most likely diagnosis and treatment direction. The node hash value of the decision path is broadcast to a predetermined number of verification nodes, which is a unique digital fingerprint obtained by encrypting the feature vector data, weight coefficient and node generation timestamp contained in the node using the SHA-256 algorithm.

[0063] After receiving the broadcast node hash value, each verification node starts the local comparison process, retrieves the feature vector data set stored in the verification node, and calculates the matching degree between the local feature vector and the feature vector described in the broadcast information. The cosine similarity algorithm is used to calculate the matching degree, which focuses on measuring the difference in direction of the feature vector, effectively avoiding the interference of the absolute value of the number on the similarity judgment. When the matching degree calculation result exceeds the preset proportion threshold, for example, set to 80%, it means that most of the verification nodes agree on the effectiveness of the decision node. The decision node is marked as an effective node, which is the basic unit of the final executable diagnosis and treatment scheme. The digital signature generation process is started immediately, and the verification node uses its private key to encrypt and sign the node hash value and the timestamp of the verification, generating a unique digital signature attached to the effective node.

[0064] The binding information of the digital signature and the valid node is broadcast back to the blockchain network for other nodes to record and verify. Decision nodes that fail to reach the preset proportion threshold of matching degree are marked as invalid nodes, which will not enter the subsequent diagnosis and treatment scheme construction process and trigger the local reconstruction mechanism of the decision tree. The consensus mechanism adopts a practical Byzantine fault tolerance algorithm to coordinate the status among the verification nodes. The practical Byzantine fault tolerance algorithm can tolerate a certain number of node failures or malicious behaviors, and still ensure reliable consensus on the validity of the decision nodes. The updating and verification process of the dynamic diagnosis and treatment decision tree forms a closed loop, with new feedback data constantly input, the decision tree structure continuously fine-tuned, the node validity cyclically verified, and the entire system kept dynamic evolution. The generation process of the node hash value contains an incremental version number, which ensures that the node hash value is unique even in the case of the same feature vector data but different generation times, effectively preventing the repeated use of old decision nodes. All digital signatures generated during the verification process are stored together in a chain, and the digital signature chain clearly records the complete audit trail of each valid node from generation to verification, providing an unalterable responsibility trace basis for diagnosis and treatment decisions. Invalid nodes and their corresponding feature vector data are temporarily isolated and archived. The isolated and archived data can be used for subsequent analysis of the reasons for the divergence of the decision model, providing a data basis for optimizing feature extraction and decision algorithms. The verification status of the entire decision tree is recorded as a transaction in a new block of the blockchain, and the non-tamperability of the blockchain ensures the historical traceability of the diagnosis and treatment decision logic.

[0065] Example 3: see Figure 3The system parses the disease staging identifiers and complication risk levels in the executable treatment plan. The disease staging identifiers are machine-readable, strictly adhering to the coding system in the 11th revision of the International Classification of Diseases, dividing the patient's disease progression into discrete and clearly defined stages. The complication risk level is calculated using a multivariate assessment model, whose input parameters include patient age, genetic markers, organ function indicators, and key events in past medical history. A pre-defined clinical guideline library is queried by the system. This library is a read-only knowledge graph stored on the blockchain, containing evidence-based medicine guidelines published by authoritative medical associations. The process of matching treatment templates and generating a basic instruction set is a rule-based automated reasoning process. The system maps the parsed precise disease staging identifiers and quantified complication risk levels to diagnostic nodes and conditional paths in the knowledge graph to find the most suitable standardized treatment plan template. The basic instruction set includes the chemical drug name, administration method, initial dose range, physical therapy parameters, surgical procedure sequence, and a matrix of planned follow-up time intervals. Patient-specific parameters are integrated into the calibration algorithm. These parameters include real-time measured body surface area, creatinine clearance, drug-metabolizing enzyme genotypes, serum albumin levels, and patient-reported quality of life scores. Prioritizing and dose calibration of the basic instruction set requires the application of clinical decision support logic. Prioritization is based on a quantitative assessment of the potential benefit-risk ratio of treatment measures, while dose calibration relies on pharmacokinetic models to simulate in vivo exposure at different doses. The final generation of the treatment instruction sequence transforms abstract medical knowledge into a series of precise, executable machine instructions. The treatment instruction sequence is encoded using the HL7FHIR standard resource format to ensure seamless parsing and execution of instructions in heterogeneous hospital information systems. Drug configuration parameters specify the total amount of drug, single dose, administration rate, infusion carrier, and dosing interval. Device operation steps are decomposed into atomic operation units with strict order. The follow-up period defines the assessment time points, the types of biological samples to be collected, and the required imaging examinations.

[0066] Feedback data from the patient's execution of treatment instructions is collected via a heterogeneous Internet of Things (IoT) network. Metabolic product concentration changes after medication administration are continuously recorded by minimally invasive or non-invasive biosensors. These sensors package the detected specific metabolite concentration values ​​along with high-precision timestamps into data packets and send them to the gateway device. Fluctuations in physiological indicators, such as continuous arterial pressure, EEG spectrum, and respiratory waveforms, are recorded by bedside monitoring devices using high-frequency sampling. The built-in algorithms of these monitoring devices calculate the statistical characteristics of each indicator in real time. IoT devices synchronously record the mechanical parameters and time intervals during device operation. Mechanical parameters include the torque applied by the rehabilitation robot, the base rate and amplitude of the insulin pump, and the inspiratory-to-expiratory ratio setting of the ventilator. Time intervals precisely record the start time, duration, and stop time of each intervention. Metabolic product concentration changes, physiological indicator fluctuation data, and mechanical parameters are aggregated into a structured feedback data block. This feedback data block is encapsulated in a format conforming to the IEEE 11073 standard, balancing data exchange efficiency with semantic clarity. Before being transmitted over the network, feedback data blocks are appended with a digital signature based on an elliptic curve digital signature algorithm. This signature is generated using the private key of the data acquisition device to verify that the data has not been tampered with since acquisition and that its source is reliable. Feedback data blocks are transmitted in real-time to the blockchain node network via a dedicated medical IoT network with quality of service assurance. The transmission protocol prioritizes the Advanced Message Queuing Protocol, which ensures reliable message delivery. The monitoring process is designed for continuous operation. The execution status of the feedback data stream and the treatment instruction sequence is associated with a globally unique treatment event identifier. This treatment event identifier must be included in the feedback data block for accurate efficacy attribution analysis. An anomaly detection module is activated when data flows into the processing pipeline. This module uses a Z-score-based statistical process control chart method to dynamically calculate individualized normal ranges for each patient's physiological parameters. Any data point exceeding the control limits triggers an immediate alarm. The mapping relationship between feedback data and the treatment instruction sequence is recorded in a distributed mapping table, stored as key-value pairs on the blockchain's distributed ledger, providing irrefutable evidence of the association. Data packets remain encrypted throughout transmission. Application-layer encryption utilizes the public key of the target blockchain node to ensure data confidentiality across the transmission chain. The entire feedback data collection and transmission system is designed according to privacy principles, anonymizing data at its source to minimize the risk of patient privacy breaches. The successful reception and recording of a feedback data block on the blockchain network triggers a pre-defined state machine transition in the smart contract, updating the corresponding treatment instruction's status to "Executing - Data Returned," thus creating the necessary conditions for subsequent efficacy evaluation and treatment plan adjustments.The continuously accumulated feedback data forms a high-resolution health status trajectory curve over time. This health status trajectory curve is a valuable data asset for assessing the long-term effectiveness of treatment plans, identifying adverse drug reactions, and studying the natural history of diseases.

[0067] The drug dosage calibration process requires adjustments based on a specific pharmacokinetic model, the core formula of which is as follows:

[0068] ;

[0069] in: This represents the calculated calibrated drug dose. This represents the baseline drug dosage recommended by the baseline instruction set matched from a pre-defined clinical guideline library. This represents the target blood drug concentration that is expected to be achieved during treatment. This represents the blood drug concentration at the current dose predicted using a pharmacokinetic model based on individual patient differences. This represents the dose adjustment factor based on the risk level of complications. Its value is determined by a risk level lookup table, for example, 1.0 for low risk, 0.7 for medium risk, and 0.5 for high risk.

[0070] Example 4: The system compares the standard deviation and offset direction of the feedback data block with the expected efficacy data. The feedback data block includes metabolite concentration change curves, physiological index fluctuation data, and mechanical parameters collected from IoT devices. The expected efficacy data comes from an efficacy model library established based on large-scale clinical studies, which sets the expected physiological index change trajectory for each standard treatment regimen. The standard deviation is calculated using a sliding window method, with the sliding window covering data points within the most recent continuous monitoring period, calculating the dispersion of the difference between the feedback data and the expected data at each time point. Offset direction analysis determines the overall trend of the feedback data deviating from the expected trajectory. Offset directions are divided into positive offset, negative offset, and volatility offset. Positive offset indicates that the actual efficacy is better than expected, negative offset indicates that the efficacy is not as expected, and volatility offset indicates that the condition is unstable. When the calculated standard deviation exceeds a preset threshold, the preset threshold is dynamically adjusted according to the disease type and treatment stage, triggering the scheme correction logic in the smart contract. The smart contract calls the feature extraction module to recalculate the cross-modal correlation matrix. The feature extraction module accesses the latest encrypted data packet stored on the blockchain, decrypts it, and executes the same feature engineering process as the initial analysis. The recalculation process prioritizes feature dimensions that show the largest deviation in the standard deviation analysis. For example, if the standard deviation of blood pressure fluctuation data significantly exceeds the standard, the time-series feature extraction will focus on a more refined analysis of the blood pressure signal. The updated cross-modal correlation matrix incorporates the latest feedback information, reflecting the patient's real-time response to the initial treatment plan. Based on the updated cross-modal correlation matrix, the branch paths of the treatment decision tree are reconstructed. The reconstruction process involves re-dividing decision nodes and redistributing branch weights. Feature paths highly correlated with negative offsets are pruned, and new decision paths based on the latest data are generated.

[0071] The differences between the revised treatment plan and the original plan are extracted. These differences include changed drug names, adjusted dosages, modified device operating parameters, and updated follow-up frequencies. Change instructions and parameter adjustments are clearly identified. Change instructions refer to the specific treatment measures added, deleted, or replaced, while parameter adjustments refer to the numerical changes made from the original values. A transaction structure is generated, containing a change timestamp, the executing physician's digital identity, and a difference hash value. The change timestamp records the precise moment the plan revision occurred. The executing physician's digital identity is a CA-certified digital certificate used to identify the responsible medical individual initiating the revision. The difference hash value is a unique digest obtained by performing a SHA-256 hash on the entire difference data block. A proof-of-work mechanism is used to write the transaction structure into a new block. Nodes in the network compete for the right to write the transaction by solving a complex mathematical problem. The node that successfully finds the solution appends a new block containing the transaction structure to the end of the blockchain. The revised treatment plan is generated and stored in parallel on the blockchain with the original plan. Each plan has an independent version number and a hash pointer to its parent version, forming a complete version history chain. The difference hash value is not only used to identify transactions, but it is also incorporated into the Merkle root calculation of subsequent blocks. This ensures that any tampering with the difference data will result in a change in the Merkle root value, which will then be detected by the system. Writing a new transaction record updates the smart contract's state variable, which records the version number of the currently effective treatment plan, ensuring the system always executes the latest treatment plan. The entire correction process, from deviation detection to the on-chain update of the new plan, is automatically executed by the smart contract, minimizing delays caused by human intervention and enabling rapid response to patient conditions. The difference data written to the blockchain serves as an audit trail for medical decisions, providing an immutable data foundation for medical quality assessment and liability determination. See Table 1 for an example of the structure of the difference data transaction:

[0072] Table 1: Example of Differential Data Transaction Structure

[0073] Nodes in the blockchain network verify the received transaction structure, including the validity of the digital signature, the correctness of the differential hash value, and the standardization of the transaction format. The proof-of-work mechanism ensures that the network reaches consensus on the content of the new block in a distributed environment, preventing malicious nodes from tampering with medical records. Once the transaction is confirmed and written into the new block, the revised treatment plan officially takes effect, and the smart contract automatically sends the new treatment instruction sequence to the patient's execution device. The system continuously monitors new feedback data after the revised plan is executed, initiating a new round of evaluation-revision cycle to achieve closed-loop management and continuous optimization of the treatment process. This automatic correction mechanism based on blockchain and smart contracts enhances the adaptability and reliability of the remote medical system, providing patients with dynamic and personalized health management services.

[0074] Example 5: Treatment plans for similar cases stored in the blockchain node network are periodically scanned. Scanning tasks are automatically initiated by smart contracts according to preset time cycles, such as a global scan every 24 hours. The scanning process generates a query request based on the disease diagnosis code, the patient's basic demographics, and the core feature vectors of the treatment plan. This query request is sent to every complete node in the blockchain network. Each complete node searches for matching case data in its local index and returns the matching treatment plans for similar cases and their corresponding standardized feature vector sets to the aggregation node. The number of treatment plans for similar cases increases with the growth of the blockchain, forming a dynamically expanding clinical knowledge base. Cluster analysis identifies the distribution patterns of the feature vectors of the collected treatment plans for similar cases. The cluster analysis uses a density-based spatial clustering algorithm, which can discover clusters of arbitrary shapes and effectively handle noise points. The distribution patterns of feature vectors are reflected in the density and clustering patterns of data points in a multi-dimensional feature space. For example, in the clustering of treatment plans for patients with type 2 diabetes, feature vectors may form different clusters around several core dimensions such as "insulin sensitivity," "degree of obesity," and "number of complications." Each cluster represents a subgroup of patients with significantly shared clinical characteristics, and cases within the same cluster exhibit high similarity in treatment protocols regarding drug selection and dosage adjustment strategies. The discovery of these distribution patterns provides data-driven insights into the complex relationship between disease presentation and treatment response.

[0075] The feature weighting strategy in the smart contract is optimized based on the distribution patterns revealed by cluster analysis. The optimization process is a dynamic adjustment mechanism based on reinforcement learning. The smart contract maintains a feature weight matrix, which defines the importance score of each feature in constructing a dynamic treatment decision tree. The reinforcement learning algorithm uses the clustering results as environmental feedback. If a feature shows high discriminative power in distinguishing different clinical subgroups, its weight score in the feature weight matrix will be increased. Conversely, features with weak correlation to treatment outcomes will have their weights decreased. For example, cluster analysis might reveal that for a certain subgroup of heart disease, the "late ventricular potential" feature is far more important than the "resting heart rate" feature in predicting treatment outcomes. In this case, the smart contract will assign a higher weight to the "late ventricular potential" feature when constructing a decision tree for new patients belonging to this subgroup. The optimized feature weighting strategy is immediately updated in the smart contract's state variables, and the new strategy will be applied to the next process of generating a treatment plan for a patient. The entire optimization loop is closed and automated, requiring no manual intervention from data scanning and cluster analysis to strategy updates. The results of cluster analysis, including cluster center coordinates, the number of cases within each cluster, and other statistical information, are written back to the blockchain as a new transaction in an anonymous, aggregated form. This transaction provides auditable traceability for adjusting feature weights, allowing any participant to verify the data patterns underlying the weight changes. The frequency of periodic scans can be flexibly adjusted based on network conditions and clinical needs. During disease outbreaks or periods of data surge following the launch of new drugs, the scan frequency can be increased to once every twelve hours to more quickly capture new treatment patterns. The algorithm parameters used in cluster analysis, such as neighborhood size and minimum number of points, can also be fine-tuned according to the characteristics of the disease type to ensure the clinical significance of the clustering results. Optimizing the feature weight allocation strategy is a continuous learning process. By constantly absorbing new case data, the system's decision-making logic adapts to the evolution of medical knowledge and changes in the disease spectrum. This self-optimization capability based on real-world evidence makes the blockchain-based remote intelligent disease diagnosis and treatment system not merely an execution tool, but a learning system capable of accumulating and elevating collective clinical wisdom.

[0076] To illustrate this, consider a case study on hypertension management. A blockchain node network periodically scans all treatment protocols for cases diagnosed with "essential hypertension." Cluster analysis may reveal three main patient subgroups: one characterized by "high salt sensitivity" and "low renin activity," whose treatment regimens respond well to diuretics; a second characterized by "high sympathetic excitability" and "rapid heart rate," whose feature vector distribution shows significant effectiveness of beta-blockers; and a third characterized by "high angiotensin II levels," whose distribution indicates that angiotensin-converting enzyme inhibitors are preferred. Based on these cluster analysis results, the feature weighting strategy in the smart contract is optimized, significantly increasing the weight of features such as "salt sensitivity test results," "plasma renin activity," and "resting sympathetic tone index" in the decision tree. When a new patient's hypertension data is uploaded, the system will match its feature vector with these clusters. If its feature vector is closest to the "high sympathetic nerve excitability" cluster, then when generating a dynamic diagnosis and treatment decision tree for it, the system will give priority to treatment pathways that can effectively reduce sympathetic nerve activity and assign higher monitoring weights to relevant physiological indicators.

[0077] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0078] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A blockchain-based remote intelligent disease diagnosis and treatment method, characterized in that, include: Collect patients' multimodal physiological data and generate encrypted data packets. The multimodal physiological data includes real-time vital signs data, historical medical record text data, and medical image data. The encrypted data packets are distributed and stored through a blockchain node network, and data hash values ​​and timestamps are generated on each node. The smart contract is invoked to extract cross-modal features from the encrypted data packet and generate a standardized feature vector set, which includes the association identifiers of temporal features, spatial features and semantic features. A dynamic diagnosis and treatment decision tree is constructed based on the standardized feature vector set, and the dynamic diagnosis and treatment decision tree dynamically adjusts the branch paths according to the feature weights. The validity of the nodes in the dynamic diagnosis and treatment decision tree is verified through a consensus mechanism, and the verified decision paths are marked as executable treatment plans to generate a treatment instruction sequence, which includes drug configuration parameters, device operation steps and follow-up period. Real-time monitoring of feedback data from the patient's execution of the treatment instruction sequence, and updating to the blockchain node network; When the deviation between the feedback data and the expected therapeutic effect exceeds a threshold, the smart contract is triggered to regenerate a corrected treatment plan. The difference between the revised treatment plan and the original treatment plan is written into the blockchain as a new transaction record.

2. The blockchain-based remote intelligent disease diagnosis and treatment method according to claim 1, characterized in that, The step of invoking the smart contract to perform cross-modal feature extraction on the encrypted data packet includes: The real-time vital signs data in the encrypted data packet are decomposed into equally spaced time segments, and the mean, variance and peak frequency of each segment are calculated. Keyword entity recognition and contextual relationship parsing are performed on the historical medical record text data to generate a structured semantic graph; Extract the grayscale distribution matrix and edge gradient features of the medical image data to construct a three-dimensional spatial feature vector; The temporal segment features, structured semantic graphs, and three-dimensional spatial feature vectors are aligned along the time axis and then fused into a cross-modal correlation matrix.

3. The blockchain-based remote intelligent disease diagnosis and treatment method according to claim 2, characterized in that, The construction of a dynamic diagnosis and treatment decision tree based on the standardized feature vector set includes: The initial decision nodes are determined based on the Euclidean distance between the eigenvectors in the cross-modal correlation matrix. Calculate the mutual information entropy between each feature vector and the disease classification label, and dynamically allocate the weight coefficients of the decision branches; When new feedback data is added, the splitting priority of the decision node is adjusted based on the weight coefficient.

4. The blockchain-based remote intelligent disease diagnosis and treatment method according to claim 3, characterized in that, The verification of the node validity of the dynamic diagnosis and treatment decision tree through a consensus mechanism includes: Broadcast the node hash value of the decision path to a preset number of verification nodes; Each verification node compares the matching degree between the locally stored feature vector and the node hash value; When the matching degree exceeds the preset ratio, the node is marked as a valid node and a digital signature is generated.

5. The blockchain-based remote intelligent disease diagnosis and treatment method according to claim 4, characterized in that, The generation of the treatment instruction sequence includes: Analyze the disease staging markers and complication risk levels in the executable treatment plan; Match treatment templates from a pre-defined clinical guideline library to generate a basic instruction set; Prioritize and calibrate the basic instruction set based on individual patient differences.

6. The blockchain-based remote intelligent disease diagnosis and treatment method according to claim 5, characterized in that, The real-time monitoring of feedback data on the patient's execution of the treatment instruction sequence includes: Collect data on the changes in the concentration of metabolites and the fluctuations in physiological indicators after patients take medication. The mechanical parameters and time intervals during the operation of the instrument are recorded using IoT devices; The concentration change curves of the metabolites, the fluctuation data of physiological indicators, and the mechanical parameters are packaged into a feedback data block.

7. The blockchain-based remote intelligent disease diagnosis and treatment method according to claim 6, characterized in that, The triggering of the smart contract to regenerate the corrected treatment plan includes: Compare the standard deviation and offset direction of the feedback data block with the expected therapeutic effect data; When the standard deviation exceeds the threshold, the feature extraction module is invoked to recalculate the cross-modal correlation matrix; The branch paths of the diagnosis and treatment decision tree are reconstructed based on the updated cross-modal association matrix.

8. The blockchain-based remote intelligent disease diagnosis and treatment method according to claim 7, characterized in that, The step of writing the difference data between the revised treatment plan and the original treatment plan into the blockchain as a new transaction record includes: Extract the change instructions and parameter adjustment amounts from the difference data; Generate a transaction structure that includes a change timestamp, the executing physician's digital identity, and a differential hash value; The transaction structure is written into a new block using a proof-of-work mechanism.

9. The blockchain-based remote intelligent disease diagnosis and treatment method according to claim 8, characterized in that, Also includes: Regularly scan the treatment protocols for similar cases stored in the blockchain; Cluster analysis was performed to study the distribution patterns of feature vectors in the treatment protocols for similar cases. Optimize the feature weight allocation strategy in smart contracts based on distribution patterns.

10. A blockchain-based remote intelligent disease diagnosis and treatment system, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the blockchain-based remote intelligent disease diagnosis and treatment method according to any one of claims 1 to 9.

Images