Blockchain-based embryo culture image visualization management method and system

Abstract

This invention relates to a blockchain-based method and system for visual management of embryo culture images, specifically in the fields of medical image processing and blockchain evidence storage. The method involves acquiring time-series images at a preset sampling frequency and extracting the digital digest hash value of the second frame. Boundary constraint processing is applied to the two frames to obtain target region feature images. Morphological variation feature values ​​are obtained through similarity analysis. If the mutation condition is not met, the on-chain rights confirmation request is blocked, and the digital digest is concatenated with the historical root hash value from the local cache pool to generate the target cumulative root hash value. If the mutation condition is met, the target cumulative root hash value is generated by packaging it with the digital digest and submitting it to the blockchain node for on-chain consensus rights confirmation. The sampling frequency and mutation condition are updated based on feedback and optimized parameters. This method addresses the systemic technical defects of existing solutions, such as on-chain resource depletion and severe delays in keyframe anchoring.

Description

Technical Field

[0001] This application relates to the fields of medical image processing and blockchain evidence storage technology, specifically a blockchain-based method and system for visual management of embryo culture images. Background Technology

[0002] In the field of assisted reproduction, time-lapse incubators provide crucial data for assessing embryonic developmental potential through continuous image acquisition. With the advancement of digital healthcare and the increasing demand for medical evidence, achieving reliable storage and traceability of these time-series images has become a technical challenge. Blockchain technology, due to its decentralized and tamper-proof characteristics, has been introduced. Existing conventional solutions extract digital summaries of images at a fixed frequency and store them on the blockchain, implicitly assuming that each frame of the image has equal evidentiary value.

[0003] However, this assumption is seriously contrary to the true biological laws of embryonic development. Embryonic development exhibits a nonlinear, skipping characteristic of long-term homeostasis and transient mutations: during the intercleavage period, which lasts for tens of hours, the morphology remains almost unchanged; dramatic remodeling only occurs during brief cleavage events. Existing isofrequency evidence storage strategies completely ignore this homeostatic delay characteristic, leading to two major systemic defects:

[0004] First, resource depletion: Hundreds of on-chain transactions occur throughout the entire embryonic development cycle, with over 95% being background noise frames during the steady-state period, causing on-chain storage bloat and wasted gas fees. Second, keyframe delay: Indiscriminate high-frequency transactions congest the memory pool, delaying the confirmation of truly valuable critical mutation frames, thus compromising the immediacy and time-anchoring accuracy required for medical evidence. Therefore, a novel management method is urgently needed that can dynamically sense the embryonic development status and adaptively adjust the frequency of evidence storage. Summary of the Invention

[0005] The purpose of this application is to provide a blockchain-based method and system for visual management of embryo culture images, aiming to solve the systemic technical defects caused by the mismatch between fixed-frequency indiscriminate hashing on the chain and the nonlinear steady-state characteristics of embryo development, resulting in the depletion of on-chain resources and severe delays in keyframe anchoring.

[0006] The objective of this application can be achieved through the following technical solution: Firstly, a blockchain-based method for visual management of embryo culture images, comprising the following steps:

[0007] At a preset sampling frequency, the first time-series image and the second time-series image of the target object are acquired sequentially, and the digital digest hash value of the second time-series image is extracted.

[0008] Boundary constraint processing based on a preset mask matrix is ​​performed on the first time series image and the second time series image to obtain the first target region feature image and the second target region feature image;

[0009] A similarity analysis is performed on the first target region feature image and the second target region feature image to obtain morphological variation feature values ​​that characterize the structural evolution state of the target object, and it is determined whether the morphological variation feature values ​​meet the preset mutation conditions.

[0010] When the morphological variation feature value does not meet the preset mutation condition, the on-chain confirmation request sent to the blockchain node is blocked, and the digital digest hash value and the historical cumulative root hash value in the local cache pool are concatenated by one-way hashing to generate the target cumulative root hash value.

[0011] When the morphological variation feature value meets the preset mutation condition, the target cumulative root hash value is extracted from the local cache pool, and it is packaged with the digital digest hash value and submitted to the blockchain node to perform on-chain consensus and rights confirmation.

[0012] After executing on-chain consensus and confirming rights, feedback optimization parameters representing the image evidence preservation effect are obtained, and the preset sampling frequency and preset mutation conditions are updated accordingly.

[0013] Secondly, the blockchain-based embryo culture image visualization management system includes the following modules:

[0014] The image acquisition module is used to sequentially acquire a first time-series image and a second time-series image of the target object at a preset sampling frequency, and extract the digital digest hash value of the second time-series image;

[0015] The image processing module is used to perform boundary constraint processing based on a preset mask matrix on the first time-series image and the second time-series image to obtain a first target region feature image and a second target region feature image;

[0016] The image analysis module is used to perform similarity analysis on the first target area feature image and the second target area feature image to obtain morphological variation feature values ​​that characterize the structural evolution state of the target object, and to determine whether the morphological variation feature values ​​meet the preset mutation conditions.

[0017] The off-chain caching module is used to block the on-chain confirmation request sent to the blockchain node when the morphological variation feature value does not meet the preset mutation condition, and to perform one-way hash concatenation of the digital digest hash value and the historical cumulative root hash value in the local cache pool to generate the target cumulative root hash value.

[0018] The on-chain evidence storage module is used to extract the target cumulative root hash value from the local cache pool when the morphological variation feature value meets the preset mutation condition, and then package it with the digital digest hash value and submit it to the blockchain node to perform on-chain consensus and rights confirmation.

[0019] The feedback optimization module is used to obtain feedback optimization parameters that characterize the image evidence preservation effect after executing on-chain consensus and confirmation, and update the preset sampling frequency and preset mutation conditions according to them.

[0020] Thirdly, a computer storage medium stores computer-executable instructions, which, when executed, implement the blockchain-based embryo culture image visualization management method described in the first aspect.

[0021] Compared with the prior art, the beneficial effects of this application are:

[0022] This application deeply couples the biological morphological mutation rate with the blockchain block generation logic, breaking down the cross-layer isolation between the image processing layer and the consensus layer, and achieving non-linear routing where data content determines the evidence storage channel. It can reduce the number of on-chain transactions from hundreds to just over ten within the cultivation period, significantly reducing gas consumption and network congestion. Through a folding and compression mechanism of off-chain incrementally accumulated root hash values, it provides unforgeable cryptographic endorsement for massive amounts of steady-state frames, while ensuring zero-latency on-chain confirmation of key mutation frames. Combining forced triggering based on elapsed time differences with multi-parameter adaptive feedback optimization to construct a closed-loop control system, it can maintain stable and reliable evidence storage performance even under extreme conditions. Attached Figure Description

[0023] Figure 1 This is a schematic diagram illustrating the steps of the blockchain-based embryo culture image visualization management method of this application;

[0024] Figure 2 This is a schematic diagram of the modules of the blockchain-based embryo culture image visualization management system of this application. Detailed Implementation

[0025] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but only to illustrate selected embodiments of this application.

[0026] Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item has been defined in one figure, it does not need to be further defined and explained in subsequent figures. The terms "first", "second", etc. are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0027] The development of assisted reproductive technology has made in-vitro fertilization (IVF) and embryo transfer an important clinical method for treating infertility. Time-lapse incubators, as core equipment in modern assisted reproductive laboratories, use built-in microscopic optical cameras to continuously acquire images of the entire embryonic development process, providing embryologists with undisturbed dynamic observation data and becoming a key tool for assessing embryonic developmental potential. However, with the deepening of digital healthcare and the increasing demand for evidence in medical disputes, how to achieve tamper-proof preservation and reliable traceability of these clinically valuable continuous time-series image data has gradually become an urgent technical problem to be solved.

[0028] Blockchain technology, with its decentralized, immutable, and fully traceable characteristics, is considered an ideal solution to the problem of medical image evidence preservation. The conventional approach of existing technologies is to extract the current frame image at a fixed frequency (e.g., every 10 minutes), generate its digital digest hash value, and then call a smart contract deployed on the blockchain to write the hash value as evidence data onto the chain. This approach is based on the default assumption that each frame image has equal biological evolutionary value and tamper-proof anchoring requirements, thus employing an equal-frequency, undifferentiated evidence preservation strategy.

[0029] However, the aforementioned baseline assumptions suffer from a serious mean-averaging fallacy, contradicting the true biological laws of embryonic development. Under actual embryo culture conditions, the developmental process exhibits typical non-linear, jump-like characteristics: during the intercleavage phase (i.e., the homeostatic phase), which lasts for tens of hours, cell morphology shows almost no substantial change, and the differences between adjacent time frames are extremely small; while during extremely brief cleavage events (such as pronuclear disappearance, first cleavage, blastocyst formation, and other key biological milestones) (i.e., the mutation phase), cell structure undergoes drastic morphological remodeling within minutes or even less. This biological homeostatic delay characteristic of long-term homeostasis followed by brief mutations is a unique physical evolutionary law of embryonic development, constituting the core cognitive blind spot of current technology.

[0030] Existing equal-frequency evidence storage strategies completely ignore the aforementioned biological steady-state latency characteristics, leading to two systemic technical defects. First, resource depletion: During a typical 5-6 day culture cycle, if data is continuously uploaded to the blockchain at a frequency of one frame every 10 minutes, a single device will inject over 720 on-chain transactions into the blockchain network. Of these, over 95% of the images are in a steady-state phase. These frames, whose morphology remains almost unchanged, are written to the chain in large quantities as background noise data, causing a rapid expansion of on-chain storage capacity and unnecessary depletion of smart contract gas fees. Second, keyframe latency: Due to indiscriminate high-frequency concurrent transactions, the transaction mempool in the blockchain network will experience severe congestion. This causes significant delays for truly valuable medical evidence-based mutation frames (such as cleavage event frames) while they wait in the queue for confirmation, completely undermining the immediate validity and time-anchoring accuracy required for medical evidence.

[0031] In summary, the existing fixed-frequency indiscriminate hashing strategy for blockchain is fundamentally structurally mismatched with the nonlinear steady-state characteristics of embryonic development. This is a long-standing common bottleneck in the field, and there is an urgent need to provide a new management method that can dynamically sense the embryonic development status and adaptively regulate the frequency of on-chain evidence storage.

[0032] Therefore, such as Figure 1 As shown, this application provides a blockchain-based method for visual management of embryo culture images, including the following steps:

[0033] At a preset sampling frequency, the first time-series image and the second time-series image of the target object are acquired sequentially, and the digital digest hash value of the second time-series image is extracted.

[0034] Boundary constraint processing based on a preset mask matrix is ​​performed on the first time series image and the second time series image to obtain the first target region feature image and the second target region feature image;

[0035] A similarity analysis is performed on the first target region feature image and the second target region feature image to obtain morphological variation feature values ​​that characterize the structural evolution state of the target object, and it is determined whether the morphological variation feature values ​​meet the preset mutation conditions.

[0036] When the morphological variation feature value does not meet the preset mutation condition, the on-chain confirmation request sent to the blockchain node is blocked, and the digital digest hash value and the historical cumulative root hash value in the local cache pool are concatenated by one-way hashing to generate the target cumulative root hash value.

[0037] When the morphological variation feature value meets the preset mutation condition, the target cumulative root hash value is extracted from the local cache pool, and it is packaged with the digital digest hash value and submitted to the blockchain node to perform on-chain consensus and rights confirmation.

[0038] After executing on-chain consensus and confirming rights, feedback optimization parameters representing the image evidence preservation effect are obtained, and the preset sampling frequency and preset mutation conditions are updated accordingly.

[0039] In another implementation, such as Figure 2 As shown, this application also provides a blockchain-based embryo culture image visualization management system, including the following modules:

[0040] The image acquisition module is used to sequentially acquire a first time-series image and a second time-series image of the target object at a preset sampling frequency, and extract the digital digest hash value of the second time-series image;

[0041] The image processing module is used to perform boundary constraint processing based on a preset mask matrix on the first time-series image and the second time-series image to obtain a first target region feature image and a second target region feature image;

[0042] The image analysis module is used to perform similarity analysis on the first target area feature image and the second target area feature image to obtain morphological variation feature values ​​that characterize the structural evolution state of the target object, and to determine whether the morphological variation feature values ​​meet the preset mutation conditions.

[0043] The off-chain caching module is used to block the on-chain confirmation request sent to the blockchain node when the morphological variation feature value does not meet the preset mutation condition, and to perform one-way hash concatenation of the digital digest hash value and the historical cumulative root hash value in the local cache pool to generate the target cumulative root hash value.

[0044] The on-chain evidence storage module is used to extract the target cumulative root hash value from the local cache pool when the morphological variation feature value meets the preset mutation condition, and then package it with the digital digest hash value and submit it to the blockchain node to perform on-chain consensus and rights confirmation.

[0045] The feedback optimization module is used to obtain feedback optimization parameters that characterize the image evidence preservation effect after executing on-chain consensus and confirmation, and update the preset sampling frequency and preset mutation conditions according to them.

[0046] Specifically, the image acquisition module is responsible for connecting to the data output interface of the time-difference incubator's microscopic optical camera and sampling at a preset frequency. The system periodically receives raw high-resolution time-series images, sequentially labeling the image from the previous time step as the first time-series image and the image from the current time step as the second time-series image. Simultaneously, it performs a SHA-256 hash operation on the second time-series image, extracts and caches the digital digest hash value. .

[0047] The image processing module receives the image pairs output by the image acquisition module and performs a series of operations such as grayscale processing, threshold segmentation, hole boundary contour extraction, pixel distance reduction to generate target area contour, construction of a binary preset mask matrix, and pixel-level multiplication. It outputs the first target area feature image and the second target area feature image respectively, and completes the filtering of edge optical noise.

[0048] The image analysis module receives two target region feature images, calculates the pixel mean, pixel variance, and pixel covariance of each image, and then calculates the structural similarity index using the formula. Thus, the morphological variation characteristic values ​​are obtained. Before executing the mutation condition judgment, the image analysis module also needs to call the system clock to query the elapsed time difference and compare it with the preset fault tolerance time threshold. If it times out, the state is directly set to meet the mutation condition, and the on-chain process is triggered first; if it does not time out, it further... Compared with the preset mutation critical threshold The system performs a comparison, outputs a routing status identifier, and distributes it to the corresponding downstream module.

[0049] Upon receiving the steady-state folding routing status flag, the off-chain caching module performs the following operations: blocks API requests to the blockchain node; queries the local cache pool status, and if it is empty, then... If not empty, then follow the incremental folding formula. Update the target cumulative root hash value and overwrite the storage; move the original image. The data is asynchronously uploaded to an off-chain cloud object storage database for archiving and future reference, without any on-chain interaction.

[0050] Upon receiving the mutation direct path status flag, the on-chain evidence storage module retrieves the target cumulative root hash value from the local cache pool. Together with the current timestamp information and the digital digest hash value Each independent data field is encapsulated together in the preset transaction payload, and a digital signature operation is performed using an asymmetric private key to generate a transaction data packet. The data packet is submitted to the blockchain node network with high priority, and the on-chain consensus confirmation receipt is listened to and received. After the transaction is confirmed, the local cache pool is cleared and the elapsed time counter is reset.

[0051] At the end of each preset time window, the feedback optimization module retrieves historical consensus logs from the blockchain nodes, analyzes the average block latency, and normalizes and calculates the network congestion index. Simultaneously, the proportion of forced triggering is statistically analyzed. Then, calculations were performed according to the three update formulas. , and The updated parameters are then sent to the image acquisition module, image analysis module, and other relevant modules, so that they can run with the new parameters in the next time window.

[0052] I. Image acquisition and digital digest hash value extraction;

[0053] At a preset sampling frequency, the first time-series image and the second time-series image of the target object are acquired sequentially, and the digital digest hash value of the second time-series image is extracted.

[0054] Specifically, the built-in microscope in the time-difference incubator operates at a preset sampling frequency (the reference sampling frequency under initial conditions). It can be set to capture one frame every 10 minutes (approximately 0.1 frames / minute) and continuously output a raw high-resolution image stream. At the current sampling time... The image received by the edge computing device is the second time-series image. At the immediately preceding sampling time The received image is the first time-series image. Together, they form a temporally continuous image pair, which is used for morphological difference quantification analysis between subsequent adjacent frames.

[0055] For the second time series image Edge computing devices synchronously invoke digital digest hash algorithms (such as SHA-256) to perform one-way hash operations on the complete byte stream of the original image, extracting and storing its digital digest hash value. ,satisfy The hash value of this digital digest has a fixed length (e.g., 256 bits) and is unidirectionally irreversible and highly collision-resistant. Any tampering with the original image will result in a complete change of the hash value, forming the basis for tamper-proof evidence for subsequent on-chain storage.

[0056] II. Boundary constraint processing based on a preset mask matrix;

[0057] Boundary constraint processing based on a preset mask matrix is ​​performed on the first time series image and the second time series image to obtain the first target region feature image and the second target region feature image.

[0058] In the actual operating conditions of a time-varying incubator, optical artifacts often occur at the edges of the wells in the culture dish due to minute oscillations in the culture medium surface, reflections from the inner walls of the wells, or refraction from condensation droplets. If these edge optical noises are not filtered out, they will seriously interfere with the accurate quantification of subsequent morphological variation features, leading to false triggering of on-chain requests even when no biological deformation has occurred. Therefore, it is necessary to perform precise target region constraint extraction on the input image.

[0059] The specific process of handling the above boundary constraints is as follows:

[0060] First, for the first time series image The original pixel matrix is ​​converted to grayscale, transforming the color image into a single-channel grayscale image to reduce the computational complexity of subsequent processing and eliminate interference between color channels. Subsequently, a thresholding algorithm (e.g., adaptive thresholding) is applied to the grayscale image to identify the clear physical circular boundaries of the petri dish pores, extracting the pore boundary contour that characterizes the physical boundary of the target object. This pore boundary contour is a closed geometric curve that precisely delineates the boundary between the inner wall of the pore and the external environment.

[0061] Next, the outline of the cavity boundary is shrunk by a preset pixel distance (e.g., 50 pixels) towards its geometric center (i.e., the center coordinates of the cavity's circle) to generate the target area outline. The purpose of this shrinkage operation is to actively remove the edge region of the cavity's inner wall—the interference zone where the refraction of condensed water droplets and the concentration of reflection from the inner wall—from the effective analysis area, thereby ensuring that subsequent feature extraction is performed only within the effective internal region containing the core morphological information of the embryo.

[0062] Based on this, a preset mask matrix is ​​constructed. Its size is exactly the same as the pixel matrix of the input image. Based on the target region contour, pixels inside the target region contour in the preset mask matrix are all assigned a first preset constant (value of 1), while pixels outside the contour are all assigned a second preset constant (value of 0), and the first preset constant is greater than the second preset constant. Therefore, the preset mask matrix... This constitutes a binary spatial selection operator, with the inner region set to 1 (reserved) and the outer region set to 0 (masked).

[0063] Finally, the first time series images are respectively... The original pixel matrix and the second time-series image The original pixel matrix and the preset mask matrix Pixel-level multiplication is performed, which involves multiplying the pixel values ​​at corresponding positions in the two matrices point by point. Pixel values ​​located inside the target region outline are multiplied by 1 and retained as is, while pixel values ​​located outside are multiplied by 0 and removed to a black background (pixel value of 0). After the above calculation, the first target region feature image, filtered to remove edge optical noise, is output. Second target region feature image Both only retain the effective regions containing core embryonic biological morphological information.

[0064] III. Quantitative extraction of morphological variation feature values ​​and judgment of preset mutation conditions;

[0065] A similarity analysis is performed on the first target region feature image and the second target region feature image to obtain morphological variation feature values ​​that characterize the structural evolution state of the target object, and it is determined whether the morphological variation feature values ​​meet the preset mutation conditions.

[0066] Compared to the simple pixel mean square error method, the structural similarity index algorithm has a stronger causal sensitivity to structural texture changes caused by embryonic cell division, and at the same time has a good anti-interference ability to the overall brightness drift caused by light source aging. Therefore, this invention adopts the morphological difference quantification method based on structural similarity as the core feature extraction means.

[0067] The specific process is as follows: Extract the feature images of the first target region respectively. and the second target region feature image The pixel mean, pixel variance, and pixel covariance between them. and These represent the pixel mean and pixel variance of the feature image of the second target region, respectively. and These are the pixel mean and pixel variance of the feature image of the first target region, respectively. Let be the pixel covariance of the first target region feature image and the second target region feature image. Based on this, calculate the structural similarity index, which characterizes the comprehensive differences in brightness, contrast, and spatial structure between the two images. The calculation formula is as follows:

[0068] ;

[0069] in, and The first and second preset constants are used to prevent the calculated value from becoming unstable when the denominator is close to zero; they are usually set to... , , The dynamic range of pixel values. , A preset constant that is much less than 1 (e.g.) , Structural similarity index The range of values ​​is When the two images are exactly the same The greater the difference, the better. The smaller.

[0070] Based on structural similarity index Define morphological variation feature values ​​that characterize the structural evolution state of the target object. ;

[0071] The physical meaning of the above morphological variation feature values ​​is as follows: When the embryo is in the biological homeostasis of the intercleavage stage, the feature images of the target area in two adjacent frames are highly similar. If it approaches 1, then A value close to 0 indicates that the embryo is in a morphologically stable, mutation-free state; when the embryo undergoes morphological mutation events such as cytokinesis, the structural difference between two adjacent frames increases significantly. A significant drop, then The significant increase indicates the occurrence of biological cleavage mutations.

[0072] In obtaining morphological variation characteristic values Then, compare it with the preset mutation critical threshold. Perform a comparison and determine the routing status: when Less than the preset mutation critical threshold When the morphological variation feature value does not meet the preset mutation condition, the embryo is currently in a biological homeostasis, and the system triggers the chain-down folding routing mechanism; when Greater than or equal to the preset mutation critical threshold When the morphological variation feature value meets the preset mutation condition, indicating that the embryo has undergone a morphological mutation event of core medical value, the system triggers the on-chain direct pathway routing mechanism. The preset mutation threshold... Normal distribution statistical analysis can be performed on historical massive cultivation cycle morphological change data to pre-calibrate the data.

[0073] IV. Off-chain incremental folding during steady state – generation of the target cumulative root hash value;

[0074] When the morphological variation feature value does not meet the preset mutation condition, the system blocks the on-chain confirmation request sent to the blockchain node, and performs one-way hash concatenation of the digital digest hash value and the historical cumulative root hash value in the local cache pool to generate the corresponding target cumulative root hash value.

[0075] This step is one of the core innovative mechanisms of this invention. In essence, without consuming any blockchain resources, it utilizes the one-way irreversibility and collision resistance of cryptographic hash functions to weave the hash value of each frame of the image during the steady state period into an evolving cumulative root hash value through incremental folding, thereby constructing a cryptographically verifiable proof of existence for a large number of steady-state frames that are not yet on the blockchain.

[0076] The specific process for generating the target cumulative root hash value is as follows:

[0077] First, the storage allocation status of the local cache pool is read to determine if it is empty. When the local cache pool is empty (i.e., the current frame is the first steady-state frame at the start of the cultivation cycle, or the first steady-state frame after the previous on-chain consensus confirmation), the digital digest hash value of the second time-series image is... It is directly used as the target cumulative root hash value and written to the specified memory address of the local cache pool.

[0078] When the local cache pool is not empty, an incremental folding operation is performed: the target cumulative root hash value recorded in the specified memory address is read out and used as the historical cumulative root hash value. Subsequently, the historical cumulative root hash value is used as prefix data, and the digital digest hash value of the second time-series image is used as the prefix data. As suffix data, ordered one-way hash concatenation is performed at the byte stream level, and the concatenated data is then linked together. ( (This involves byte-level concatenation) to perform a hash operation to generate a new target cumulative root hash value. ;

[0079] ;

[0080] Finally, the newly generated target cumulative root hash value is... The historical accumulated root hash value written to the specified memory address is overwritten, completing the state update of the local cache pool. At this point, the processing of the current steady-state frame is entirely completed locally, without incurring any on-chain gas fees or sending any requests to the blockchain network. However, the hash value of this frame has been cryptographically folded and anchored to the continuously evolving target accumulated root hash value. Among them.

[0081] The aforementioned folding mechanism exhibits significant space compression in mathematics: regardless of the number of frames accumulated during the steady-state period, the corresponding target cumulative root hash value remains a fixed-length string, rather than a data set that grows linearly with the number of frames. This is the underlying mathematical foundation for this scheme to completely solve the on-chain storage explosion problem.

[0082] V. On-chain joint packaging and rights confirmation during the mutation period;

[0083] When the morphological variation feature value meets the preset mutation condition, the target cumulative root hash value is extracted from the local cache pool, and then packaged with the digital digest hash value and submitted to the blockchain node to perform on-chain consensus and rights confirmation.

[0084] This stage indicates that the embryo has undergone a morphological change event with core medical assessment and traceability value. The system needs to immediately initiate an on-chain strong ownership confirmation process to achieve zero-latency timestamp anchoring. The specific process of extracting the target cumulative root hash value and packaging it with the digital digest hash value for submission is as follows:

[0085] First, obtain the second time-series image (i.e., the current mutation frame). The data is collected using the precise timestamp information (e.g., a UTC millisecond-level timestamp). Subsequently, a preset transaction payload is constructed, combining the timestamp information and the digital digest hash value. and the target cumulative root hash value extracted from the specified memory address in the local cache pool. Each of the three fields is written into a separate, independent data field within the predefined transaction payload. The three fields have different functional semantics: the timestamp field is used to solidify the absolute time anchor point of the mutation event on the chain; the digital digest hash value field is used to tamper-proof and confirm the ownership of the current mutation frame itself; and the target cumulative root hash value field serves as a cryptographic endorsement proxy for all preceding steady-state frames, indirectly completing the notarization of massive steady-state frames through a single on-chain action.

[0086] Secondly, a digital signature operation is performed on the preset transaction payload based on the pre-allocated asymmetric private key (e.g., a private key generated based on the Elliptic Curve Digital Signature Algorithm ECDSA) to generate a transaction data packet carrying a valid signature. This data packet is then submitted to the blockchain node in a high-priority broadcast manner, triggering the on-chain consensus node to verify the validity of the digital signature carried in the transaction data packet. After the verification is successful, a distributed consensus algorithm is run to package the target ownership confirmation transaction data and add it to the target block, generating a blockchain transaction receipt with an absolute timestamp (containing the on-chain transaction hash value and block height information).

[0087] Finally, after receiving the consensus confirmation receipt from the blockchain node, confirming that the transaction has been successfully recorded on the chain, the edge computing device immediately triggers the execution logic to clear the local cache pool, completely destroying the historical accumulated root hash value of the local cache, and preparing to welcome the steady-state frame accumulation of the next embryo development computing cycle.

[0088] VI. Engineering Fault Tolerance Guarantee – Forced Trigger Mechanism Driven by Elapsed Time Difference;

[0089] Before determining whether the morphological variation feature value meets the preset mutation condition, the following engineering fault tolerance steps are also included: obtaining the time interval between the current time and the last time the on-chain consensus confirmation was completed, marking it as the elapsed time difference, and comparing the elapsed time difference with the preset fault tolerance time threshold.

[0090] In practical engineering applications, extreme operating conditions can sometimes cause a system to remain trapped in a steady-state folding cycle for an extended period, preventing the natural triggering of mutations on the chain. Examples include an embryo ceasing deformation due to developmental failure, or an image sensor hardware malfunction leading to a pseudo-steady state where the same image is output for an extended period. In these situations, without a fault-tolerant backup mechanism, the hash values ​​of a large number of steady-state frames will remain in the local cache pool for a long time, posing an engineering risk of complete data loss due to unexpected power outages.

[0091] Therefore, the system maintains an independent real-time clock to continuously record the elapsed time difference since the last successful execution of on-chain consensus and rights confirmation. When the elapsed time difference exceeds a preset fault tolerance time threshold, the system blocks the normal judgment process of the morphological mutation feature value, directly sets its state to meet the preset mutation condition, and immediately triggers the step of extracting the target accumulated root hash value to execute the on-chain consensus and rights confirmation. This forced triggering mechanism has the highest priority, and its execution is not constrained by the size of the morphological mutation feature value.

[0092] Through the above fault tolerance mechanism, the worst data storage failure period under any physical extreme conditions is strictly limited to within the preset fault tolerance time threshold, forming a closed-loop storage guarantee with extremely high engineering robustness.

[0093] VII. Feedback Optimization Mechanism – Multi-parameter Adaptive Control Based on Network State Awareness;

[0094] After executing on-chain consensus and confirming rights, feedback optimization parameters characterizing the image evidence preservation effect are obtained, and the preset sampling frequency and preset mutation conditions are updated accordingly. This invention designs a multi-parameter adaptive control mechanism based on the feedback of the real state of the blockchain network, realizing the coordinated dynamic updating of the preset sampling frequency, preset mutation critical threshold, and preset fault tolerance time threshold, so that the entire system can maintain optimal resource utilization efficiency and evidence preservation quality under different network congestion states and different embryo development activity levels.

[0095] First, extract the historical consensus logs returned by the blockchain nodes, parse out the time consumed for each block confirmation within a preset time window (e.g., the most recent complete training monitoring round or the most recent 24 hours), calculate the average block latency, and normalize it (e.g., divide the average block latency by the system's preset standard block latency benchmark value to map it to...). (range), to obtain the network congestion index representing the real-time state of the underlying network. . The closer a value is to 1, the more congested the current blockchain network is, and the higher the cost of getting things on the chain. The closer a value is to 0, the smoother the network connection.

[0096] Secondly, within the corresponding preset time window, the number of times on-chain consensus confirmation is triggered due to the elapsed time difference exceeding the preset fault tolerance time threshold is counted. The ratio of this number of triggers to the total number of on-chain consensus confirmations executed within the corresponding preset time window is obtained and marked as the forced trigger ratio. ,satisfy . The larger the value, the longer the embryo is in an inactive state (its morphology hardly changes) within that time window, and the lower the proportion of naturally triggered on-chain activity. The smaller the value, the more active the embryonic development, and the more naturally occurring events triggered by mutations on the blockchain. Network congestion index. With forced trigger ratio Together, they constitute the feedback optimization parameters.

[0097] According to the network congestion index Preset reference sampling frequency Perform negative weight decay adjustment to obtain the updated preset sampling frequency. It is then applied to the next preset time window, calculated using the following formula:

[0098] ;

[0099] in, This is a preset frequency adjustment coefficient, and its value range is usually [value range missing]. The causal logic of this formula is as follows: when the network congestion index... A larger value indicates high on-chain transaction costs, and the system reduces the sampling frequency accordingly. This reduces the amount of hash generation and potential on-chain triggering by edge computing devices at the source, thus achieving proactive load reduction; when When it approaches 0, near The system operates normally at a frequency close to the reference frequency.

[0100] The update of the preset mutation conditions includes the synchronous update of the preset mutation critical threshold and the preset fault tolerance time threshold.

[0101] When the network congestion index When the current preset congestion threshold is greater than or equal to the preset congestion threshold, the current preset sudden change critical threshold is... Perform positive compensation updates to obtain the updated preset mutation critical threshold. The calculation formula is as follows:

[0102] ;

[0103] in, This is a preset mutation adjustment coefficient. The causal logic of this formula is as follows: when the network is severely congested, the mutation threshold is increased. This increases the system's tolerance for minor morphological variations, allowing on-chain events to be triggered only when extremely violent splitting events occur. This further reduces unnecessary on-chain events, enabling more rigorous content filtering in exchange for precise utilization of limited on-chain resources.

[0104] When the forced trigger ratio When the current preset fault tolerance time threshold is greater than or equal to the preset trigger threshold, Perform dynamic extension updates to obtain the updated preset fault tolerance time threshold. The calculation formula is as follows:

[0105] ;

[0106] in, This is the preset fault tolerance adjustment coefficient. The causal logic of this formula is as follows: if and only if the network congestion index... High and forced trigger ratio When the values ​​are large, the product of these two factors will trigger a significant extension of the fault tolerance time threshold, lengthening the waiting time for timed forced on-chain events. This greatly reduces the frequency of invalid on-chain events during periods when the embryo has not undergone substantial development, maximizing gas cost savings. This dual-factor triggering design effectively prevents erroneous operations that might result from a single-factor judgment.

[0107] Through the aforementioned multi-parameter linkage adaptive control mechanism, the system of this invention can perform refined dynamic control of on-chain resource consumption while ensuring the integrity and timeliness of evidence storage, thus fully addressing the dual uncertainties of network state fluctuations and individual embryo development differences in assisted reproductive clinical scenarios.

[0108] This invention can achieve the following technical effects: by deeply coupling the unique biological morphological variation rate feature of the business application layer with the block generation logic of the underlying blockchain state machine, it breaks the long-standing cross-layer isolation barrier between the image processing layer and the blockchain consensus layer, and realizes a non-linear dedicated routing mechanism in which the data content itself determines the topology of the data storage channel.

[0109] In terms of quantitative results, during a typical embryo culture cycle of 5 to 6 days, approximately 120 hours (calculated at one frame every 10 minutes), the traditional equal-frequency on-chain solution requires approximately 720 on-chain transactions. However, this invention, through a dynamic routing mechanism driven by morphological variation feature values, triggers on-chain consensus and rights confirmation only at approximately 10 to 15 key biological nodes with core clinical value, such as the appearance of the fertilized pronucleus, various stages of cleavage, and blastocyst formation. This reduces the number of on-chain transactions by approximately 98%, significantly reduces on-chain gas consumption and network bandwidth usage, and completely eliminates the congestion of the Mempool caused by invalid background noise frames.

[0110] From the perspective of evidence preservation quality, this invention utilizes the folding and compression mechanism of off-chain incremental cumulative root hash values, leveraging the unidirectional irreversibility of cryptographic hashes, to provide unforgeable cryptographic endorsement for massive steady-state frames with a fixed-length hash value. This ensures zero-latency on-chain confirmation of rights for mutation frames with clinical medical evidentiary value while guaranteeing 100% frame-level traceability. From the perspective of system robustness, this invention combines an engineering fault-tolerant forced triggering mechanism based on elapsed time difference with a multi-parameter adaptive feedback optimization mechanism based on network congestion index and forced triggering ratio to construct a closed-loop adaptive control system. This allows the system to maintain stable and reliable evidence preservation even under extreme conditions.

[0111] In another embodiment, this application also provides a computer storage medium storing computer-executable instructions, which, when executed, implement the blockchain-based embryo culture image visualization management method.

[0112] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A blockchain-based method for visual management of embryo culture images, characterized in that, Includes the following steps: At a preset sampling frequency, the first time-series image and the second time-series image of the target object are acquired sequentially, and the digital digest hash value of the second time-series image is extracted. Boundary constraint processing based on a preset mask matrix is ​​performed on the first time series image and the second time series image to obtain the first target region feature image and the second target region feature image; A similarity analysis is performed on the first target region feature image and the second target region feature image to obtain morphological variation feature values ​​that characterize the structural evolution state of the target object, and it is determined whether the morphological variation feature values ​​meet the preset mutation conditions. When the morphological variation feature value does not meet the preset mutation condition, the on-chain confirmation request sent to the blockchain node is blocked, and the digital digest hash value and the historical cumulative root hash value in the local cache pool are concatenated by one-way hashing to generate the target cumulative root hash value. When the morphological variation feature value meets the preset mutation condition, the target cumulative root hash value is extracted from the local cache pool, and it is packaged with the digital digest hash value and submitted to the blockchain node to perform on-chain consensus and rights confirmation. After executing on-chain consensus and confirming rights, feedback optimization parameters representing the image evidence preservation effect are obtained, and the preset sampling frequency and preset mutation conditions are updated accordingly.

2. The method for visual management of embryo culture images based on blockchain according to claim 1, characterized in that, The process of obtaining the feature images of the first and second target regions includes: After performing grayscale processing on the original pixel matrix of the first time series image, a threshold segmentation algorithm is used to extract the hole boundary contour that represents the physical boundary of the target object, and the hole boundary contour is shrunk towards its geometric center by a preset pixel distance to generate the target area contour. In the preset mask matrix, all pixels located inside the target area outline are assigned a first preset constant value, while all pixels located outside are assigned a second preset constant value, wherein the first preset constant value is greater than the second preset constant value. The original pixel matrix of the first time-series image and the original pixel matrix of the second time-series image are multiplied by the preset mask matrix at the pixel level to output the first target area feature image and the second target area feature image after filtering out edge optical noise.

3. The method for visual management of embryo culture images based on blockchain according to claim 2, characterized in that, The process of obtaining morphological variation feature values ​​and determining whether they meet preset mutation conditions includes: The pixel mean, pixel variance, and pixel covariance between the first and second target region feature images are extracted respectively, and a structural similarity index representing the difference between the first and second target region feature images is obtained. ; ； The morphological variation characteristic value , and These represent the pixel mean and pixel variance of the feature image of the second target region, respectively. and These are the pixel mean and pixel variance of the feature image of the first target region, respectively. and For the first preset constant and the second preset constant, The pixel covariance of the first target region feature image and the second target region feature image; The morphological variation feature value is compared with a preset mutation critical threshold. When the morphological variation feature value is less than the preset mutation critical threshold, it is determined that it does not meet the preset mutation condition. When the morphological variation feature value is greater than or equal to the preset mutation critical threshold, it is determined that it meets the preset mutation condition.

4. The method for visual management of embryo culture images based on blockchain according to claim 1, characterized in that, The process of generating the target cumulative root hash value includes: Read the storage allocation status of the local cache pool. When it is empty, use the digital digest hash value of the second time series image as the target cumulative root hash value and write it to the specified memory address of the local cache pool. When it is not empty, the target cumulative root hash value recorded in the specified memory address is used as the historical cumulative root hash value. The historical cumulative root hash value is used as the prefix data and the digital digest hash value is used as the suffix data to perform one-way hash concatenation. The concatenated data is hashed to generate the target cumulative root hash value and overwrites the historical cumulative root hash value in the specified memory address.

5. The blockchain-based embryo culture image visualization management method according to claim 4, characterized in that, The process of extracting the target cumulative root hash value and packaging it with the digital digest hash value for submission includes: Obtain the timestamp information corresponding to the time when the second time series image is acquired, and write the timestamp information, the digital digest hash value, and the target cumulative root hash value into different data fields of the preset transaction payload respectively; A digital signature operation is performed on a preset transaction payload based on a pre-allocated asymmetric private key to generate a transaction data packet and submit it to the blockchain node. After on-chain consensus and confirmation of rights are completed, the local cache pool is cleared.

6. The blockchain-based embryo culture image visualization management method according to claim 3, characterized in that, Before determining whether the morphological variation characteristic value meets the preset mutation conditions, the following steps are also included: Obtain the time interval between the current moment and the last time the on-chain consensus confirmation was completed, and mark it as the elapsed time difference. Compare the elapsed time difference with a preset fault tolerance time threshold. When the elapsed time difference is greater than the preset fault tolerance time threshold, block the judgment process of the morphological mutation feature value, directly set it as meeting the preset mutation condition, and trigger the step of extracting the target cumulative root hash value to execute the on-chain consensus confirmation.

7. The blockchain-based embryo culture image visualization management method according to claim 6, characterized in that, The process of updating the preset sampling frequency includes: Extract the historical consensus logs returned by the blockchain nodes, parse out the average block production latency within a preset time window, and normalize it to obtain a network congestion index that characterizes the underlying network state. ; Within a corresponding preset time window, the number of times on-chain consensus confirmation is triggered due to the elapsed time difference exceeding a preset fault tolerance time threshold is counted. The ratio of this number of triggers to the total number of on-chain consensus confirmations executed within the corresponding preset time window is obtained and marked as the forced trigger ratio. The network congestion index and the forced trigger ratio are both feedback optimization parameters; Based on the network congestion index, a preset benchmark sampling frequency is used. Adjustments are made to obtain the updated preset sampling frequency. And apply it to the next preset time window. This is the preset frequency adjustment coefficient.

8. The blockchain-based embryo culture image visualization management method according to claim 7, characterized in that, The process of updating the preset mutation conditions includes: The update of the preset mutation conditions includes updating the preset mutation critical threshold and the preset fault tolerance time threshold. When the network congestion index is greater than or equal to the preset congestion threshold, the current preset mutation critical threshold is updated. Perform an update to obtain the updated preset mutation critical threshold. ; When the forced trigger ratio is greater than or equal to the preset trigger threshold, the current preset fault tolerance time threshold is... Perform an update to obtain the updated preset fault tolerance time threshold. And apply both to the next preset time window. and These are the preset mutation adjustment coefficient and fault tolerance adjustment coefficient.

9. A blockchain-based embryo culture image visualization management system, characterized in that, Includes the following modules: The image acquisition module is used to sequentially acquire a first time-series image and a second time-series image of the target object at a preset sampling frequency, and extract the digital digest hash value of the second time-series image; The image processing module is used to perform boundary constraint processing based on a preset mask matrix on the first time-series image and the second time-series image to obtain a first target region feature image and a second target region feature image; The image analysis module is used to perform similarity analysis on the first target area feature image and the second target area feature image to obtain morphological variation feature values ​​that characterize the structural evolution state of the target object, and to determine whether the morphological variation feature values ​​meet the preset mutation conditions. The off-chain caching module is used to block the on-chain confirmation request sent to the blockchain node when the morphological variation feature value does not meet the preset mutation condition, and to perform one-way hash concatenation of the digital digest hash value and the historical cumulative root hash value in the local cache pool to generate the target cumulative root hash value. The on-chain evidence storage module is used to extract the target cumulative root hash value from the local cache pool when the morphological variation feature value meets the preset mutation condition, and then package it with the digital digest hash value and submit it to the blockchain node to perform on-chain consensus and rights confirmation. The feedback optimization module is used to obtain feedback optimization parameters that characterize the image evidence preservation effect after executing on-chain consensus and confirmation, and update the preset sampling frequency and preset mutation conditions according to them.

10. A computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, they implement the blockchain-based embryo culture image visualization management method as described in any one of claims 1-8.

Images