Big data fast indexing and query system based on diagnosis and treatment characteristics of general surgery department

Abstract

The application relates to the technical field of electric digital data processing, and discloses a big data fast indexing and querying system based on general surgery diagnosis and treatment characteristics, which comprises a data stream acquisition module, a topological sequence reorganization module, a physical space mapping module, an index key synthesis module and a preloading execution engine; the topological sequence reorganization module analyzes heterogeneous diagnosis and treatment data streams and reorganizes the data streams into logically continuous patient whole-course data aggregates; the physical space mapping module sequentially maps the aggregates to a continuous physical address space of a bottom-layer storage; the index key synthesis module generates composite index items according to physical address offsets among data blocks; and the preloading execution engine drives the bottom-layer storage to burst read, the application converts cross-modal diagnosis and treatment correlation queries into one-way sequential reading of the storage, reduces the track delay and access overhead in the searching process, and guarantees the searching response delay certainty under the condition that the data volume continuously grows.

Description

Technical Field

[0001] This invention belongs to the field of electronic digital data processing technology, and in particular relates to a big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics. Background Technology

[0002] Current distributed database systems typically employ a storage-compute separation architecture, with their underlying physical storage following a columnar storage layout. This layout assumes statistical independence between data dimensions, sacrificing cross-dimensional correlation for write throughput. It exhibits good scalability when processing general internet data. However, when processing general surgical diagnostic data with strong temporal dependencies and causal coupling, the existing layout reveals physical layer addressing bottlenecks due to heterogeneous payloads such as preoperative vital sign sequences, intraoperative endoscopic video frames, and postoperative pathological texts. When the system responds to cross-modal relational queries, the retrieval engine needs to extract heterogeneous data scattered across different physical sectors and perform high-density relational operations in memory. This strong coupling at the logical level and the high degree of dispersion at the physical storage level result in a large number of random disk seek requests when the system processes full-process data backtracking.

[0003] Common linear improvement approaches focus on increasing hardware computing power or storage bandwidth. However, given the high-dimensional features of massive amounts of clinical data, simply increasing hardware resources often fails due to the non-linear increase in the number of underlying seek operations. This inherent contradiction between logical relationships and addressing actions cannot be fundamentally resolved through conventional software layer patches. Existing technologies mainly suffer from the following shortcomings: 1. Mismatch between the topological relationship of the clinical process and the underlying storage layout; 2. High dispersion of heterogeneous data at the physical layer causes random addressing pressure; 3. General caching mechanisms cannot utilize the spatial locality features in the clinical path, for example, [the following is an example from the patent announcement number CN11556]. Chinese invention patent 3127B discloses a query method and system based on a big data medical general retrieval index. It automatically builds an Elasticsearch index through metadata configuration, which lowers the development threshold of the medical retrieval system at the logical level. However, it is still a general logical mapping scheme. The scheme fails to penetrate the underlying objective attributes of strong temporal causality and cross-modal coupling in the general surgery diagnosis and treatment path. The heterogeneous diagnosis and treatment data blocks are distributed in a discrete and hashed manner on the underlying physical disk. The logical layer is ordered and the physical layer is discrete, resulting in a misalignment of the architecture. When faced with cross-stage diagnosis and treatment feature association queries, it cannot trigger local preloading of the storage medium space, bears the seek loss, and cannot guarantee the deterministic nature of the retrieval response delay.

[0004] Therefore, how to establish a physical layout mechanism at the storage layer that matches the topology of the diagnosis and treatment process, and eliminate the random seek loss of heterogeneous data during retrieval, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics, comprising: The data stream acquisition module is used to acquire heterogeneous diagnostic and treatment data streams consisting of preoperative vital signs parameters, intraoperative imaging features, and postoperative follow-up text. The topology reorganization module is used to parse the metadata of heterogeneous diagnosis and treatment data streams to identify patient entity identifiers and nodes of the directed acyclic graph of the diagnosis and treatment path. It reorganizes the cross-time domain diagnosis and treatment data fragments of the same patient into a patient's full disease course data aggregate according to the topological order of the directed acyclic graph of the diagnosis and treatment path. The physical space mapping module is used to sequentially map the patient's entire course data aggregate as atomic storage units to the continuous physical address space of the underlying memory, so that the physical end address of the preceding diagnosis and treatment data block in the patient's entire course data aggregate is adjacent to the physical start address of the subsequent diagnosis and treatment data block, thereby solidifying the temporal dependency relationship defined by the directed acyclic graph of the diagnosis and treatment path into the storage distribution pattern of the underlying memory. The index key synthesis module is used to calculate the physical address offset of the subsequent diagnosis and treatment data block relative to the preceding diagnosis and treatment data block. By performing bit fusion operation on the physical address offset and the topological coordinates in the directed acyclic graph of the diagnosis and treatment path, a composite index item containing the content hash value, topological coordinates and physical address offset is generated. The preloaded execution engine is used to retrieve the physical address of the starting point of the target topology path in the composite index entry according to the query request, and drive the underlying memory to perform burst read operations based on the physical address offset recorded in the composite index entry.

[0006] Preferably, when reconstructing the patient's full-course data aggregate, the topology order reconstructing module allocates a virtual continuous address segment to the patient's full-course data aggregate and maps the structured diagnosis and treatment text, time-series physiological parameters, and unstructured medical video frames to each sub-partition of the virtual continuous address segment; the topology order reconstructing module calculates the physical offset step size of each sub-partition based on the coupling relationship determined by the directed acyclic graph of the diagnosis and treatment path.

[0007] Preferably, before performing the write operation, the physical space mapping module constructs a physical image of the patient's entire disease course data aggregate in the memory buffer and performs data alignment processing according to the physical block boundary of the underlying memory, so that the physical start address of the atomic storage unit is anchored to the logical block start bit of the underlying memory.

[0008] Preferably, when generating a composite index entry, the index key synthesis module extracts the feature bitmap summary embedded in the preceding medical data block and stores it in the metadata area of ​​the composite index entry; the feature bitmap summary is used to logically intercept non-target feature query requests at the index addressing layer to reduce the frequency of invalid addressing for large-scale heterogeneous medical data.

[0009] Preferably, the preloading execution engine interacts with the underlying memory through the direct memory access mechanism. While retrieving the preceding diagnostic data blocks, it preloads the associated subsequent diagnostic data blocks into the cache space based on the physical address offset recorded in the composite index entry, thereby achieving hardware-level spatial locality prefetching.

[0010] Preferably, the system also includes a bandwidth control module; the bandwidth control module is used to monitor the real-time bitrate changes of unstructured medical video frames and dynamically adjust the storage interval of adjacent video frames in the continuous physical address space to match the bus transmission bandwidth of the preloaded execution engine in burst read mode.

[0011] Preferably, when calculating the physical address offset, the index key synthesis module introduces a weighting factor to compensate for the addressing overhead of the text modality and the video modality; the physical address offset in the composite index item is used to convert the logical jump instruction of the query layer into the hardware prefetch instruction of the underlying storage controller.

[0012] Preferably, the preloading execution engine determines the dynamic prefetch window size based on the topological span of the patient's full-course data aggregate; the byte length L of the dynamic prefetch window is calculated using the following formula: Where n is the number of nodes covered by the query path. Δ represents the length of the data block corresponding to the i-th node in the physical address space, and Δ is the redundancy compensation amount of 4KB to 64KB based on storage latency.

[0013] Preferably, the system also includes a feature mapping module; the feature mapping module is used to map the preoperative sign sequence, intraoperative video features and postoperative pathology report to a unified feature vector space, and use the dimension-reduced feature code as the guiding identifier of the composite index item, so as to realize the normalized representation of cross-modal diagnosis and treatment features in the index dimension.

[0014] Preferably, the index key synthesis module embeds diagnostic feature masks in the index entries, converting composite query requests involving cross-diagnosis and treatment stage related features into unidirectional sequential scan operations targeting a continuous physical address space, thereby avoiding disk random seek latency in the process of retrieving heterogeneous medical big data.

[0015] Compared with existing technologies, the big data rapid indexing and query system based on general surgery diagnosis and treatment characteristics of this invention has the following advantages: 1. In fast indexing and querying of big data, by converting the topological logic of the diagnosis and treatment path into the physical arrangement of the underlying storage space during the index building stage, the distribution of heterogeneous data on the storage medium is changed, and the cross-modal data association that relies on multiple random seeks is converted into unidirectional sequential reading along the topological coordinate axis. This reduces the disk mechanical latency and memory random access overhead in the data retrieval process, and ensures the determinism of retrieval response latency under the condition of continuous growth in data volume.

[0016] 2. By leveraging the synergy between physical clustering writes and hard offset pointers, a driving relationship is established between logical indexes and hardware prefetching mechanisms. This enables query requests targeting specific diagnostic features to trigger burst transmission modes in the underlying storage medium. While retrieving preceding diagnostic records, the spatial locality principle of hardware is used to preload potentially related subsequent data payloads into the cache space, avoiding repeated physical seek operations in complex relational queries and improving the utilization efficiency of computing resources and storage bandwidth.

[0017] 3. By embedding feature bitmap digests in the metadata area of ​​the routing pointer, the verification logic of high-dimensional features is moved from the data record layer to the index addressing layer, thus constructing a logical interception barrier against invalid query requests. When the system faces sparse or non-existent target feature combinations, it directly blocks the subsequent physical extraction process based on the bit operation results, intercepts the movement of invalid data blocks from the storage medium to the memory space, and maintains the stability of system bandwidth under extreme load conditions. Attached Figure Description

[0018] Figure 1 This is a diagram of the heterogeneous diagnostic and treatment data topology reconstruction and continuous physical space mapping architecture of the present invention; Figure 2 This is a flowchart of the composite index item structure parsing and feature bitmap logic interception retrieval process of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0020] A rapid indexing and querying system for big data based on general surgical diagnosis and treatment characteristics, comprising: The data stream acquisition module is used to acquire heterogeneous diagnostic and treatment data streams consisting of preoperative vital signs parameters, intraoperative imaging features, and postoperative follow-up text. The topology reorganization module is used to parse the metadata of heterogeneous diagnosis and treatment data streams to identify patient entity identifiers and nodes of the directed acyclic graph of the diagnosis and treatment path. It reorganizes the cross-time domain diagnosis and treatment data fragments of the same patient into a patient's full disease course data aggregate according to the topological order of the directed acyclic graph of the diagnosis and treatment path. The physical space mapping module is used to sequentially map the patient's entire course data aggregate as atomic storage units to the continuous physical address space of the underlying memory, so that the physical end address of the preceding diagnosis and treatment data block in the patient's entire course data aggregate is adjacent to the physical start address of the subsequent diagnosis and treatment data block, thereby solidifying the temporal dependency relationship defined by the directed acyclic graph of the diagnosis and treatment path into the storage distribution pattern of the underlying memory. The index key synthesis module is used to calculate the physical address offset of the subsequent diagnosis and treatment data block relative to the preceding diagnosis and treatment data block. By performing bit fusion operation on the physical address offset and the topological coordinates in the directed acyclic graph of the diagnosis and treatment path, a composite index item containing the content hash value, topological coordinates and physical address offset is generated. The preloaded execution engine is used to retrieve the physical address of the starting point of the target topology path in the composite index entry according to the query request, and drive the underlying memory to perform burst read operations based on the physical address offset recorded in the composite index entry.

[0021] Preferably, when reconstructing the patient's full-course data aggregate, the topology order reconstructing module allocates a virtual continuous address segment to the patient's full-course data aggregate and maps the structured diagnosis and treatment text, time-series physiological parameters, and unstructured medical video frames to each sub-partition of the virtual continuous address segment; the topology order reconstructing module calculates the physical offset step size of each sub-partition based on the coupling relationship determined by the directed acyclic graph of the diagnosis and treatment path.

[0022] Preferably, before performing the write operation, the physical space mapping module constructs a physical image of the patient's entire disease course data aggregate in the memory buffer and performs data alignment processing according to the physical block boundary of the underlying memory, so that the physical start address of the atomic storage unit is anchored to the logical block start bit of the underlying memory.

[0023] Preferably, when generating a composite index entry, the index key synthesis module extracts the feature bitmap summary embedded in the preceding medical data block and stores it in the metadata area of ​​the composite index entry; the feature bitmap summary is used to logically intercept non-target feature query requests at the index addressing layer to reduce the frequency of invalid addressing for large-scale heterogeneous medical data.

[0024] Preferably, the preloading execution engine interacts with the underlying memory through the direct memory access mechanism. While retrieving the preceding diagnostic data blocks, it preloads the associated subsequent diagnostic data blocks into the cache space based on the physical address offset recorded in the composite index entry, thereby achieving hardware-level spatial locality prefetching.

[0025] Preferably, the system also includes a bandwidth control module; the bandwidth control module is used to monitor the real-time bitrate changes of unstructured medical video frames and dynamically adjust the storage interval of adjacent video frames in the continuous physical address space to match the bus transmission bandwidth of the preloaded execution engine in burst read mode.

[0026] Preferably, when calculating the physical address offset, the index key synthesis module introduces a weighting factor to compensate for the addressing overhead of the text modality and the video modality; the physical address offset in the composite index item is used to convert the logical jump instruction of the query layer into the hardware prefetch instruction of the underlying storage controller.

[0027] Preferably, the preloading execution engine determines the dynamic prefetch window size based on the topological span of the patient's full-course data aggregate; the byte length L of the dynamic prefetch window is calculated using the following formula: Where n is the number of nodes covered by the query path. Δ represents the length of the data block corresponding to the i-th node in the physical address space, and Δ represents the redundancy compensation amount of 4KB to 64KB based on storage latency.

[0028] Preferably, the system also includes a feature mapping module; the feature mapping module is used to map the preoperative sign sequence, intraoperative video features and postoperative pathology report to a unified feature vector space, and use the dimension-reduced feature code as the guiding identifier of the composite index item, so as to realize the normalized representation of cross-modal diagnosis and treatment features in the index dimension.

[0029] Preferably, the index key synthesis module embeds diagnostic feature masks in the index entries, converting composite query requests involving cross-diagnosis and treatment stage related features into unidirectional sequential scan operations targeting a continuous physical address space, thereby avoiding disk random seek latency in the process of retrieving heterogeneous medical big data.

[0030] Example 1: In a specific scenario of massive cross-modal diagnosis and treatment data retrospection in general surgery, the underlying database continuously receives heterogeneous diagnosis and treatment data streams consisting of preoperative vital signs parameters, intraoperative imaging features, and postoperative follow-up texts. Traditional general inverted index architectures strip data from different modalities and store it in discrete physical sectors. This causes the query engine to perform Cartesian product operations in memory and initiate massive random seek requests to the underlying storage medium when processing composite search intentions that are related to specific preoperative medical history and specific postoperative reactions under specific intraoperative operations. In such cases, the big data fast indexing and query system based on general surgery diagnosis and treatment features transforms the dynamic association set matching logic during the query period into a static physical space topology alignment layout during the data writing period, reducing the underlying physical architecture conflict between the discrete distribution pattern of general columnar storage and the temporal coupling characteristics of general surgery diagnosis and treatment events.

[0031] After the data stream acquisition module acquires heterogeneous medical data streams, the topology reorganization module parses the metadata of these streams to identify patient entity identifiers and nodes in the directed acyclic graph (DAG) of the treatment pathway. It then reorganizes the cross-temporal medical data fragments of the same patient into a complete patient disease course data aggregate according to the topological order of the DAG. When constructing the DAG, the topology reorganization module extracts globally unique patient identifiers from the heterogeneous medical data streams as the aggregation benchmark. It retrieves the timestamps of medical actions from the clinical information system and the hardware clock signal from the video equipment, performing spatiotemporal alignment within a ±500 millisecond synchronization window. It merges intraoperative endoscopic image frames and synchronized physiological parameter data for specific time periods into the same topology node. The directed edges between nodes are determined by the predefined treatment time sequence logic of the general surgery clinical pathway, ensuring that the source heterogeneous medical data fragments are logically aligned in the disease course. The physical space mapping module then maps the patient's complete disease course data. The process data aggregate is sequentially mapped to the continuous physical address space of the underlying memory as an atomic storage unit, so that the physical end address of the preceding diagnosis and treatment data block in the aggregate is strictly adjacent to the physical start address of the subsequent diagnosis and treatment data block. In this way, the temporal dependency relationship defined by the directed acyclic graph of the diagnosis and treatment path is solidified into the storage distribution law of the underlying memory. On this basis, the index key synthesis module calculates the physical address offset of the subsequent diagnosis and treatment data block relative to the preceding diagnosis and treatment data block. By performing bit fusion operation on the physical address offset and the topological coordinates in the directed acyclic graph of the diagnosis and treatment path, a composite index item containing the content hash value, topological coordinates and physical address offset is generated. The continuous physical address layout established by the physical space mapping module provides the physical premise for the index key synthesis module to extract the hardware-level offset. This composite index item enables the subsequent data extraction operation to have the spatial locality prefetching execution condition based on the physical offset.

[0032] When the system receives a query request involving cross-time-domain feature association, the preloading execution engine retrieves the physical address of the starting point of the target topology path from the composite index entry based on the query request, and determines the size of the dynamic prefetch window based on the physical address offset recorded in the composite index entry. The byte length L of the dynamic prefetch window is determined according to the formula... The result is obtained through calculation, where n is the number of nodes covered by the query path. Let be the length of the data block corresponding to the i-th node in the physical address space. To set a redundancy compensation amount in the range of 4KB to 64KB, the preload execution engine drives the underlying memory to perform burst read operations according to this byte length. Through the direct memory access mechanism, while extracting the preceding diagnosis and treatment data blocks, the associated subsequent diagnosis and treatment data blocks are preloaded into the cache space. At the same time, the feature bitmap summary and additional filter feature words embedded in the preceding diagnosis and treatment data blocks are extracted and bitwise ANDed to implement logical interception of non-target features at the addressing layer. This transforms the cross-modal data association action that originally relied on multiple discrete disk random seeks into a unidirectional sequential scan operation along the continuous physical address space. The system extracts data records that fall within the target coordinate range and summarizes them as a query result set, while maintaining the determinism of the retrieval response latency.

[0033] Example 2: In the cross-modal association query response test scenario of general surgery, a distributed storage test bench configured with an enterprise-grade solid-state drive array was set as the data acquisition source. The test bench was injected with a simulated multimodal aggregated dataset of general surgery with a total scale of 10TB. Random read actions with a peak of 10,000 times / second were superimposed throughout the test to simulate background seek noise interference during the peak operation of the hospital information system. In order to determine the redundancy compensation amount of the preload execution engine in the burst read phase, a calibration logic based on the fragmentation degree of the underlying physical sector was constructed. The value of the redundancy compensation amount is considered to balance the cache hit rate of prefetched data and the bandwidth utilization of the underlying storage bus. The decision rule is to dynamically adjust the compensation scale based on the local physical address fragmentation degree caused by the heterogeneous data blocks in the merge write phase and the single direct memory access load threshold of the data bus. When the physical address fragmentation degree increases, the redundancy compensation amount tends to the upper limit of the value range. Based on this calibration logic, the baseline redundancy compensation amount of the test group was established as 32KB under the current hardware load.

[0034] To observe the operating parameters of the above test scenario, a control group and an experimental group were set up. The control group adopted a general inverted index architecture without a physical space mapping module. Under a load condition with a concurrent query volume of 1000 queries / second, the heterogeneous data distribution caused random seek command accumulation due to background noise interference. The extracted underlying addressing indicators showed that the average number of disk seeks triggered by a single query in the control group was 312, resulting in an average query response latency of 1250.5ms and a cache miss rate of 85.6%. In the experimental group using the complete technical solution, the physical space mapping module sequentially mapped the patient's entire disease course data aggregate to a continuous address space. The index key synthesis module extracted the physical address offset based on the physical end address of the preceding data block. The preload execution engine was then loaded according to the formula... Calculate the dynamic prefetch window size, where n is the number of nodes covered by the query path. Let Δ be the length of the data block corresponding to the i-th node in the physical address space, in bytes, and let Δ be the redundancy compensation amount, in bytes. When the baseline redundancy compensation amount is set to 32KB, the unidirectional sequential scan operation of the continuous physical address suppresses random read interference, the number of disk seeks triggered by a single query is sharply reduced to 1, the average query response latency of the experimental group is reduced to 42.5ms and the cache miss rate converges to 1.1%. To verify the boundary of the redundancy compensation amount, an out-of-range control group is set up. The bandwidth control module monitors the real-time throughput of the underlying storage bus and the cache water level, calculates the deviation value E between the available bandwidth and the target read rate, and introduces a feedback adjustment loop to adjust the size of the data block between adjacent video frames in the continuous physical address space. The length D of the data block is determined according to the formula D=K⋅E, where D is the length of the data block, K is the proportional adjustment coefficient, the value of which is set according to the page alignment step of the underlying memory, and E is the difference between the remaining bus bandwidth and the bandwidth required by the preloaded task.

[0035] By adjusting the storage distribution density through inserting padding data blocks during the physical write phase and matching the hardware throughput limit of the burst read mode of the preload execution engine, when the redundancy compensation amount of the first out-of-range control group is 2KB, the prefetch window is unable to cover the physical alignment gap, the cache miss rate rebounds to 15.2%, and a second seek operation is triggered. When the redundancy compensation amount of the second out-of-range control group is 128KB, the system operation exhibits non-linear degradation characteristics. At this time, the cache hit rate changes slightly to 0.9%, but the unidirectional sequential scan operation extracts a large amount of adjacent useless sector data, causing the data bus load rate to reach saturation, and the average query response latency... The latency was reduced to 85.6ms due to bus queuing effects. The physical addressing parameters and response latency indicators extracted by the test demonstrated the technical mechanism of the system architecture. The continuous physical address layout established by the physical space mapping module provides a prerequisite for the index key synthesis module to extract hardware-level offsets. The composite index item drives the preloading execution engine to perform burst read operations, which transforms the discrete logical set matching process into a continuous stream extraction action at the physical address layer. The determined 4KB to 64KB redundancy compensation working window resolves the technical contradiction between prefetch integrity and bus bandwidth occupation, and maintains the determinism of the retrieval response latency of the system under background read and write noise interference.

[0036] Example 3: In a scenario of real-time data entry and precise traceability of high-concurrency heterogeneous data in general surgery, the data stream acquisition module acquires a heterogeneous diagnostic and treatment data stream consisting of preoperative vital signs, intraoperative imaging features, and postoperative follow-up text. General databases face constraints such as index node combination explosion and physical hash fragmentation when processing this type of multimodal data with both temporal span and causal coupling. The topology order reorganization module of this system parses the metadata of this heterogeneous diagnostic and treatment data stream to identify patient entity identifiers and nodes in the directed acyclic graph (DAG) of the treatment path. It reorganizes the cross-temporal-domain diagnostic and treatment data of the same patient into a complete patient disease course data aggregate according to the topological order of the DAG of the treatment path. To quantify the temporal dependencies between nodes, the topology order reorganization module divides each node in the DAG of the treatment path into... A topological coordinate is assigned, which is composed of a time depth value representing the chronological order of the disease stages and a modal branch identifier code used to distinguish concurrent diagnosis and treatment operations. At the same time, when reconstructing the patient's full course data aggregate, the topological order reconstruction module allocates a virtual continuous address segment to the patient's full course data aggregate and maps the structured diagnosis and treatment text, time-series physiological parameters, and unstructured medical video frames to each sub-partition of the virtual continuous address segment. The topological order reconstruction module calculates the physical offset step size of each sub-partition based on the coupling relationship determined by the directed acyclic graph of the diagnosis and treatment path. The value of the physical offset step size is determined by the product of the calibration sampling rate of the single modal data stream and the duration span of a single diagnosis and treatment event, thereby ensuring that the allocated virtual continuous address segment can accommodate heterogeneous data increments.

[0037] After establishing the virtual address boundaries and topological coordinates, the physical space mapping module constructs a physical image of the patient's entire disease course data aggregate in the memory buffer before performing the write operation. It then performs data alignment processing based on the physical block boundaries of the underlying memory, anchoring the physical start address of the atomic storage unit to the logical block start position of the underlying memory. The physical space mapping module sequentially maps the patient's entire disease course data aggregate as atomic storage units to the continuous physical address space of the underlying memory, ensuring that the physical end address of the preceding treatment data block in the patient's entire disease course data aggregate is adjacent to the physical start address of the subsequent treatment data block. This solidifies the temporal dependencies defined by the directed acyclic graph of the treatment path into the storage distribution pattern of the underlying memory. Based on this physical layout, the index key synthesis module calculates the physical address offset of the subsequent treatment data block relative to the preceding treatment data block. The index key synthesis module then calculates the offset based on the formula... Construct an index truncation and verification benchmark, where The total bit length of the composite index entry is defined as 352 bits. The bit length of the content hash value is defined as 256 bits. The bit length for topological coordinates is defined as 64 bits. The physical address offset is defined as 32 bits in length. When generating a composite index entry, the index key synthesis module performs bit fusion operations on the physical address offset and the topological coordinates in the directed acyclic graph of the treatment path. The module stores a 256-bit content hash value generated by a secure hash algorithm in the high-order space, concatenates the time depth value and modality identifier to form a 64-bit topological coordinate displacement embedded in the middle segment, and retains the 32-bit physical address offset in the low-order space, forming a unique binary sequence of 352 bits. When the physical address offset reaches the upper limit threshold of a 32-bit unsigned integer, the address extension logic is triggered. An overflow flag is written to the composite index entry's metadata area and associated with the secondary offset record table. This maintains the global uniqueness of the address pointer in single treatment scenarios where the data volume exceeds four gigabytes. The topological coordinates are shifted to the set high-order interval and a logical OR operation is performed with the physical address offset residing in the low-order interval to generate a composite index entry containing the content hash value, topological coordinates, and physical address offset. At the same time, when generating the composite index entry, the index key synthesis module extracts the feature bitmap summary embedded in the preceding diagnosis and treatment data block and stores it in the metadata area of ​​the composite index entry. The preload execution engine retrieves the physical starting address of the target topological path in the composite index entry according to the query request, and drives the underlying memory to perform a burst read operation based on the physical address offset recorded in the composite index entry. This bit-level address hardening encoding mechanism transforms cross-modal association operations into sequential byte stream truncation operations on the storage bus, maintaining the determinism of retrieval response latency while avoiding the hash collision risk of conventional query engines.

[0038] Example 4: In the deployment of a new cluster across hospital campuses, before the topology reordering module takes over the heterogeneous diagnostic and treatment data stream from the new source, the baseline calibration unit at the system front end extracts a sampling benchmark dataset containing preoperative vital signs parameters, intraoperative imaging features, and postoperative follow-up text. The timestamps of various diagnostic and treatment events in the sampling benchmark dataset are shifted toward the relative time zero point. Based on the time interval difference after shifting, a one-dimensional time depth mapping matrix is ​​constructed in memory. This time depth mapping matrix discretizes the continuous time axis into a quantized scale with a fixed step size and outputs the corresponding time depth value. The topology reordering module concatenates this time depth value with the modal branch identifier returned by the hardware interface protocol layer of the access device to establish a topology coordinate generation benchmark for the new data source, thus solving the coordinate drift and misalignment caused by the difference in time record formats across hospital campuses.

[0039] To adapt to the requirements of underlying memory physical block boundary alignment and preloading execution engine parameter settings, the physical space mapping module continuously sends multiple sets of probe data packets with increasing byte lengths to the deployed underlying storage medium. It extracts the page write latency returned by the media controller and the full load flag of the direct memory access bus status register. Based on the step point presented on the write latency curve, it locates the actual capacity boundary of the underlying physical page and establishes it as the benchmark alignment unit when constructing a physical mirror of the patient's entire disease course data aggregate in the memory buffer. The preloading execution engine extracts the upper limit of the burst read bytes when the bus full load flag is triggered. The difference between this upper limit of bytes and the sum of the data lengths of the nodes covered by a single query path is set as the redundancy compensation amount in the current hardware environment. The specific calibration steps for the redundancy compensation amount Δ within the safe range are as follows: During the initial stress test phase of system deployment, monitor the physical address hole rate of the underlying memory. When the hole rate is below 5%, fix the initial value of Δ to 8KB. If the hole rate increases in increments of 1%, then Δ is corrected upwards in increments of 4KB until the 64KB bus load threshold is reached. This step-by-step compensation strategy based on the degree of fragmentation ensures that when the ratio of hard disk seek time to data transfer time fluctuates between 2:1 and 5:1, the preloading engine can still cover all physical alignment gaps through a single burst read, avoiding secondary seek delays caused by cross-sector access. This is combined with the formula for determining the statistical variance of local address fragmentation caused by the sampling benchmark dataset. The initial values ​​of the parameters are used in this pre-debugging procedure to anchor the physical space mapping rules and dynamic prefetch window size to the physical response characteristics of the access device.

[0040] Example 5: In a dynamic retrieval environment that continuously receives unstructured general surgery records and multidimensional physiological indicators, heterogeneous data written into a continuous physical address space, coupled with semantic features that are difficult to adapt to the physical limitations of direct addressing of the underlying logic gate array, leads the data stream acquisition module to establish feature mapping standards based on baseline construction procedures after extracting various modal features. The specific operating procedures are as follows: For preoperative vital signs, ECG signals are acquired using a 100Hz sampling frequency, and the mean and standard deviation of the signal within a 10s rolling window are extracted and quantized into 64-bit feature codes; for intraoperative video features, motion vectors are extracted at a resolution of 1920 x 1080 pixels, using 16 x 16 pixels as macroblock units, and then hashed and dimensionality reduced to 128-bit feature codes; for postoperative follow-up text, the top-ranked words by frequency are extracted. The system uses medical keywords (5) and performs a 64-bit consistent hash operation. Finally, it concatenates the above-mentioned physical signs, video, and text features into a 256-bit global feature vector in a fixed bit-width order, realizing a physical normalized representation of cross-modal data in the index dimension. This module uses a consistent hash algorithm to map variable-length text roots and numerical ranges to a fixed-length binary data structure to generate a feature bitmap summary. Its bit width is set to 256 bits to match the single instruction multiple data stream operation register of the underlying processor. The system calculates the hash value of a specific clinical entity in the hash space and sets the corresponding physical bitmap coordinates to a logic high level state. This construction process forms a reference model of the underlying electrical signal sequence for subsequent state judgment, resolving the architectural conflict between high-dimensional semantic information and physical bus transmission mechanism.

[0041] When the preloading execution engine drives the underlying memory to perform a burst read operation based on the additional filter features in the query request, the system calls the aforementioned benchmark reference model to initiate a false positive compensation procedure. The preloading execution engine converts the additional filter features into a 256-bit query bitmap array, extracts the feature bitmap summary embedded in the preceding diagnostic data block, and performs a bitwise AND operation with the additional filter features to implement logical interception of non-target features at the addressing layer, according to the formula... Output the result sequence representing the matching state, where For the result sequence, For feature bitmap summaries, To query a bitmap array, when the result sequence is equivalent to the binary sequence of the queried bitmap array, the preload execution engine allows the corresponding data record to be released into the cache space. In response to the physical phenomenon of spatial overlap caused by hash mapping, the system divides the cache space into byte-level verification isolation areas. The original continuous byte sequence is extracted from the data record initially screened by bitwise AND operation. The original continuous byte sequence is compared with the plaintext features. This procedure filters out redundant heterogeneous data and reduces the invalid load on the bus, and cuts off the false positive data backflow path to converge the physical boundary of the cross-modal retrieval result set.

[0042] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A rapid indexing and querying system for big data based on general surgical diagnostic and treatment characteristics, characterized in that, The system includes: The data stream acquisition module is used to acquire heterogeneous diagnostic and treatment data streams consisting of preoperative vital signs parameters, intraoperative imaging features, and postoperative follow-up text. The topology reorganization module is used to parse the metadata of heterogeneous diagnosis and treatment data streams to identify patient entity identifiers and nodes of the directed acyclic graph of the diagnosis and treatment path. It reorganizes the cross-time domain diagnosis and treatment data fragments of the same patient into a patient's full disease course data aggregate according to the topological order of the directed acyclic graph of the diagnosis and treatment path. The physical space mapping module is used to sequentially map the patient's entire course data aggregate as atomic storage units to the continuous physical address space of the underlying memory, so that the physical end address of the preceding diagnosis and treatment data block in the patient's entire course data aggregate is adjacent to the physical start address of the subsequent diagnosis and treatment data block, thereby solidifying the temporal dependency relationship defined by the directed acyclic graph of the diagnosis and treatment path into the storage distribution pattern of the underlying memory. The index key synthesis module is used to calculate the physical address offset of the subsequent diagnosis and treatment data block relative to the preceding diagnosis and treatment data block. By performing bit fusion operation on the physical address offset and the topological coordinates in the directed acyclic graph of the diagnosis and treatment path, a composite index item containing the content hash value, topological coordinates and physical address offset is generated. The preloaded execution engine is used to retrieve the physical address of the starting point of the target topology path in the composite index entry according to the query request, and drive the underlying memory to perform burst read operations based on the physical address offset recorded in the composite index entry.

2. The big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, When reconstructing the patient's full-course data aggregate, the topology order reconstructing module allocates a virtual continuous address segment to the patient's full-course data aggregate and maps the structured diagnosis and treatment text, time-series physiological parameters, and unstructured medical video frames to each sub-partition of the virtual continuous address segment; the topology order reconstructing module calculates the physical offset step size of each sub-partition based on the coupling relationship determined by the directed acyclic graph of the diagnosis and treatment path.

3. The big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, Before performing the write operation, the physical space mapping module constructs a physical image of the patient's entire disease course data aggregate in the memory buffer and performs data alignment processing according to the physical block boundaries of the underlying memory, so that the physical start address of the atomic storage unit is anchored to the logical block start bit of the underlying memory.

4. The big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, When generating composite index entries, the index key synthesis module extracts the feature bitmap summary embedded in the preceding medical data block and stores it in the metadata area of ​​the composite index entry. The feature bitmap summary is used to logically intercept non-target feature query requests at the index addressing layer to reduce the frequency of invalid addressing for large-scale heterogeneous medical data.

5. A big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, The preloading execution engine interacts with the underlying memory through the direct memory access mechanism. While retrieving the preceding diagnostic data blocks, it preloads the associated subsequent diagnostic data blocks into the cache space based on the physical address offset recorded in the composite index entry, thereby achieving hardware-level spatial locality prefetching.

6. A big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, The system also includes a bandwidth control module; the bandwidth control module is used to monitor the real-time bitrate changes of unstructured medical video frames and dynamically adjust the storage interval of adjacent video frames in the continuous physical address space to match the bus transmission bandwidth of the preloaded execution engine in burst read mode.

7. A big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, When calculating the physical address offset, the index key synthesis module introduces a weighting factor to compensate for the addressing overhead of the text modality and the video modality; the physical address offset in the composite index item is used to convert the logical jump instruction of the query layer into the hardware prefetch instruction of the underlying storage controller.

8. A big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, The preloading execution engine determines the dynamic prefetch window size based on the topological span of the patient's full-course data aggregate; the byte length L of the dynamic prefetch window is calculated using the following formula: Where n is the number of nodes covered by the query path. Δ represents the length of the data block corresponding to the i-th node in the physical address space, and Δ represents the redundancy compensation amount of 4KB to 64KB based on storage latency.

9. A big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, The system also includes a feature mapping module; The feature mapping module is used to map preoperative physical signs sequence, intraoperative video features, and postoperative pathology reports to a unified feature vector space, and uses the dimensionality-reduced feature code as a guiding identifier for composite index items to achieve a normalized representation of cross-modal diagnosis and treatment features in the index dimension.

10. A big data rapid indexing and query system based on general surgical diagnosis and treatment characteristics according to claim 1, characterized in that, The index key synthesis module transforms complex query requests involving cross-stage diagnostic features into unidirectional sequential scan operations targeting a continuous physical address space by embedding diagnostic feature masks in the index entries, thus avoiding disk random seek latency during heterogeneous medical big data retrieval.

Images