Medical image data storage space management method and system

Abstract

The invention relates to the field of medical image information systems, and discloses a medical image data storage space management method and system. The storage space management method comprises the following steps: actively pre-caching image data of a patient to a local storage space in a bandwidth idle period according to a patient list managed by a doctor; monitoring the utilization rate of the local storage space, the clinical state information of the patient corresponding to the image data sequence and the storage time of the image data sequence; and when the corresponding trigger condition is monitored, calculating a retention value score of the corresponding image data sequence according to a storage space cleaning mode triggered by the trigger condition, and cleaning the image data sequence of which the score is lower than a preset score threshold so as to realize local storage space management. According to the storage space management method, the I / O bottleneck problem in massive image data retrieval is effectively solved, the cache hit rate is maximized in a limited physical storage space through asynchronous event driving and multi-dimensional retention value calculation, cache jitter is avoided, and efficient scheduling of storage resources is achieved.

Description

Technical Field

[0001] This invention relates to the intersection of medical imaging information systems and computer storage technology, specifically to a method and system for managing the storage space of medical imaging data. Background Technology

[0002] With the rapid development of medical imaging technology, medical imaging information systems (such as PACS, Picture Archiving and Communication Systems) are being used more and more widely in hospitals. Doctors can easily view patients' medical imaging data, enabling diagnosis and treatment, through these systems.

[0003] However, existing medical imaging information systems face the following technical problems when viewing image data: (1) I / O bottleneck and high latency: Medical image data is a typical large binary object (BLOB), with a huge volume (hundreds of MB to several GB). In the traditional request-response mode, high-concurrency access will cause the server I / O throughput to be saturated instantly, resulting in significant network transmission delays, which cannot meet the real-time rendering needs of scenarios such as emergency. (2) Resource contention of physical storage media: The physical storage space (usually SSD) of local edge nodes is a scarce resource. Traditional general page replacement algorithms such as LRU (Least Recently Used) or FIFO (First In First Out) only separate hot and cold data based on the single dimension of "access time", and cannot perceive the "business weight" and "physical volume cost" of the data. This leads to a large number of low-value large-volume files occupying physical sectors for a long time, while high-value small-volume files are frequently swapped in and out, causing invalid wear of storage media and jitter of cache hit rate.

[0004] Existing PACS systems and caching algorithms primarily focus on "download priority" or "computation task scheduling," which are essentially monotonically increasing functions based on "time" or "frequency." However, at edge nodes with limited physical storage space, the value of data depends not only on "how important it is (revenue)" but also on "how much space it occupies (cost)." Current technologies lack consideration of the crucial dimension of "value density per unit of storage space," making the system easily filled with frequently accessed but massive amounts of "inefficient data," crowding out the space for a large amount of small, high-value data. Furthermore, passive cleanup is lagging: traditional solutions typically use fixed thresholds (e.g., 90 days) or passive replacement after space is full (LazyEviction). This mechanism cannot respond to sudden changes in clinical business status (e.g., a patient's sudden discharge), resulting in a large amount of expired "zombie data" occupying valuable SSD sectors for extended periods, causing ineffective resource locking.

[0005] Furthermore, existing cache eviction algorithms have limitations. For example, the LRU algorithm only considers the most recent access time, evictioning the least accessed data. It cannot distinguish between data from "emergency patients" and "regular outpatients," potentially evictions important data from emergency patients; it also doesn't consider data volume, potentially evictions small files while retaining large files, leading to high storage space utilization. The LFU algorithm only considers access frequency, evictioning the least frequently accessed data. Similarly, it cannot consider clinical value and storage cost, potentially evictions important data from emergency patients; it also doesn't consider data volume, failing to balance clinical value and storage cost.

[0006] How can we reduce download time through intelligent pre-caching mechanisms while intelligently retaining the most clinically valuable data and cleaning up low-value data within limited local storage space, thus balancing clinical value and storage costs and achieving optimal storage space utilization? Existing technologies primarily focus on issues such as caching efficiency, download speed, and retrieval speed, but lack intelligent cleanup and recycling mechanisms for situations with limited local storage space. They cannot proactively release storage space when insufficient, nor do they consider the crucial factor of data volume (storage cost). In particular, existing technologies lack a complete solution of "pre-caching + intelligent cleanup," failing to effectively manage local storage space while reducing download time. Summary of the Invention

[0007] In view of the problems and shortcomings of the existing technology, the purpose of this invention is to provide a method and system for managing the storage space of medical image data.

[0008] To achieve the objectives of this invention, the technical solution adopted is as follows: The first aspect of this invention provides a method for managing the storage space of medical image data, comprising the following steps: S1: Based on the patient list managed by the doctor, proactively pre-cache the patient's image data to the local storage space during periods of bandwidth idleness; S2: Monitor the local storage space usage rate, the patient clinical status information corresponding to the image data sequences in the local storage space, and the storage time of the image data sequences in the local storage space; when the local storage space usage rate exceeds the preset storage threshold, execute step S3; when the patient clinical status information corresponding to the image data sequences in the local storage space changes to discharge, execute step S4; when the storage time of the image data sequences in the local storage space exceeds the preset time threshold, execute step S5. S3: Calculate the retention value score of all image data sequences in the local storage space, and clean up image data sequences whose retention value score is lower than the preset score threshold; S4: Place the image data sequences of patients whose clinical status information has been changed to discharge into the delay queue, and set the time to live (TTL) period for the delay queue. When the delay queue reaches the TTL period, calculate the retention value score of all image data sequences of the patients in the delay queue, and clean up the image data sequences whose retention value score is lower than the preset score threshold. S5: Calculate the retention value score of all image sequences whose storage time exceeds a preset time threshold, and clean up image data sequences whose retention value score is lower than the preset score threshold.

[0009] According to the above storage space management method, preferably, the formula for calculating the retention value score is as shown in Equation 1: Score =W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume Equation 1; Among them, Score is the retention value score; S_severity is the severity score, which is obtained and quantified by acquiring the patient's clinical information through the hospital information system (such as the HIS system); S_frequency is the historical access frequency score, which is obtained based on the statistics of the number of historical accesses to the image data sequence; S_time is the data timeliness score, which is calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, which is calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weight coefficients of S_severity, S_frequency, S_time, and S_volume, respectively.

[0010] According to the above storage space management method, preferably, the formula for calculating the retention value score is as shown in Equation 2: Score = (W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume)×m Equation 2; Wherein, Score is the retention value score; S_severity is the severity score, obtained by quantifying the patient's clinical information through the hospital information system; S_frequency is the historical access frequency score, obtained based on the statistical analysis of the number of historical accesses to the image data sequence; S_time is the data timeliness score, calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weighting coefficients of S_severity, S_frequency, S_time, and S_volume, respectively; m is the correlation correction coefficient. If multiple image data sequences of the same patient exist in the local storage space, and the time interval between the generation times of the multiple image data sequences is less than a preset time value, the value of m is 1.2; if multiple image data sequences of the same patient exist in the local storage space, and the time interval between the generation times of the multiple image data sequences is greater than a preset time value, the value of m is 1; if only one image data sequence of the same patient exists in the local storage space, the value of m is 1. More preferably, the preset time value is 30 days. By introducing the correlation correction coefficient m, it can be applied to complex clinical correlation scenarios.

[0011] According to the above storage space management method, preferably, the clinical information includes clinical status information and clinical diagnosis results. The calculation formula for S_severity is: S_severity = clinical status information score + clinical diagnosis result score; wherein, the clinical status information includes emergency, ICU, surgery, general hospitalization, general outpatient, discharge / transfer, and the corresponding clinical status information scores for emergency, ICU, surgery, general hospitalization, general outpatient, discharge / transfer are 10 points, 8 points, 6 points, 4 points, 2 points, and 0 points, respectively; the clinical diagnosis results include: critical diagnosis, severe diagnosis, and general diagnosis, and the corresponding clinical diagnosis result scores for critical diagnosis, severe diagnosis, and general diagnosis are 2 points, 1 point, and 0 points, respectively.

[0012] According to the above storage space management method, preferably, the formula for calculating S_frequency is: S_frequency = min(10, N_access × k); where N_access is the number of times the image data sequence has been accessed within the most recent sliding window (e.g., the most recent 7 days); and k is the frequency sensitivity coefficient, which takes a value of 2.

[0013] According to the above storage space management method, preferably, the calculation formula for S_time is: S_time = max(0, 10 - (T_current - T_create) / T_decay); where T_current is the current system time, T_create is the image data generation time, and T_decay is the decay period constant. More preferably, T_decay is set to 3.

[0014] According to the above storage space management method, preferably, the formula for calculating S_volume is: S_volume = S v × Storage medium trimming factor; where S v = min(10, (File_Size / Max_Size)×10), where File_Size is the size of the image data sequence file, and Max_Size is the size of the largest image data sequence among all image data sequences in the local storage space; the storage media trimming factor includes the SSD storage factor and the HDD storage factor, with the SSD storage factor being 1 and the HDD storage factor being 0.5.

[0015] According to the above storage space management method, preferably, the calculation method of Max_Size is as follows: take the 95th percentile of the size of all image data sequence files in the local storage space in the past 3 months as the value of Max_Size or Max_Size = total capacity × 0.1.

[0016] According to the above storage space management method, preferably, the storage space management method further includes: after cleaning up medical image data sequences with low retention value scores, storing the remaining image data sequences in the local storage space in multiple levels according to their retention value scores.

[0017] According to the above storage space management method, preferably, the specific operation of multi-level storage of the remaining image data sequences in the local storage space based on retention value scores is as follows: The remaining image data sequences are divided into three levels—hot data, warm data, and cold data—based on their retention value scores. Hot data level image data sequences are retained in local SSD fast storage, warm data level image data sequences are migrated to network edge cache, and cold data level image data sequences are archived to cloud storage. More preferably, the retention value score of the hot data level image data sequences is greater than the hot data classification threshold, the retention value score of the warm data level image data sequences is greater than the cold data classification threshold but less than or equal to the hot data classification threshold, and the retention value score of the cold data level image data sequences is less than or equal to the cold data classification threshold. Furthermore, when a cleanup condition is triggered, it can be set to prioritize the cleanup of cold data, followed by warm data, and finally hot data (cleaning only when local storage space is severely insufficient).

[0018] According to the above storage space management method, preferably, the hot data classification threshold is set to 7 and the cold data classification threshold is set to 3. Furthermore, the hot data classification threshold and cold data classification threshold are not fixed. The system periodically reviews the retention and cleanup of image data sequences, and dynamically adjusts the values ​​of the hot data classification threshold and cold data classification threshold based on the recall rate and storage pressure index. The recall rate refers to the proportion of image data sequences that, after being identified as "cold data" and stored in the cloud, are retrieved by doctors again within a preset short-term observation window (e.g., 24h or 48h). The storage pressure index refers to the average occupancy rate of local storage space within a statistical period (e.g., the past 3-7 days).

[0019] According to the above storage space management method, preferably, the preset scoring threshold is 3; and the TTL is 7 days.

[0020] According to the above storage space management method, preferably, in step S1, the pre-caching priority is determined based on the severity of the patient's condition, examination time, and access prediction, and high-priority and medium-priority image data are cached first; among them, image data of emergency patients, image data of ICU patients, and examination data that have just been performed in the past 12-24 hours (the specific time can be adjusted according to the needs) are high priority, image data of surgical patients and ordinary inpatients are medium priority, and image data of ordinary outpatients are low priority.

[0021] According to the above storage space management method, preferably, in step S1, the bandwidth idle period includes the period when the network bandwidth utilization rate is less than 30% and the system idle time.

[0022] According to the above storage space management method, preferably, the weight coefficients W1, W2, W3, and W4 are determined based on historical data analysis and the experience of storage management experts. Historical data analysis involves statistically analyzing the storage space utilization rate, data access hit rate, proportion of emergency / ICU patient data, and average file size over the past three months. Based on this historical data, the initial weight ranges are determined as follows: W1 = 0.3–0.5, W2 = 0.2–0.4, W3 = 0.1–0.3, and W4 = 0.1–0.3. Expert experience adjustment involves adjusting the weight ranges determined by the historical data based on the experience of storage management experts. After expert adjustment, the initial weight coefficients W1, W2, W3, and W4 are as follows: W1 = 0.4, W2 = 0.3, W3 = 0.2, and W4 = 0.1. Furthermore, after determining the initial values ​​of the weight coefficients W1, W2, W3, and W4, the values ​​of the weight coefficients W1, W2, W3, and W4 can be adaptively adjusted and optimized according to the actual situation.

[0023] A second aspect of this invention provides a medical image data storage space management system, including a data pre-caching module, a storage status monitoring module, a trigger condition detection module, a data retention value assessment module, and a storage space cleanup module. The data pre-caching module is connected to a hospital information system (HIS system) to obtain a list of patients managed by doctors, monitor network bandwidth status, and proactively pre-cachise image data to local storage space during periods of bandwidth idleness, based on the list of patients managed by doctors. The storage status monitoring module monitors the usage of local storage space and the status of image data sequences stored in local storage space in real time. The trigger condition detection module receives signals from the storage status monitoring module and listens to the hospital information system (HIS system). The system (IS system) detects whether the patient's clinical status information meets the storage space cleanup trigger conditions; the data retention value assessment module receives the storage space cleanup trigger signal and calculates the retention value score for each image data sequence; the storage space cleanup module executes the image data cleanup strategy in the local storage space (cleansing image data sequences with retention value scores lower than a preset score threshold) based on the data retention value score, and releases physical storage space; wherein, the trigger conditions are: monitoring the local storage space usage rate to exceed the preset storage threshold, monitoring the patient's clinical status information corresponding to the image data sequence in the local storage space to change to discharge, or monitoring the storage time of the image data sequence in the local storage space to exceed the preset time threshold.

[0024] According to the aforementioned storage space management system, preferably, the data retention value assessment module uses the following formula to calculate the retention value score for each image data sequence: Score =W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume Equation 1 Among them, Score is the retention value score; S_severity is the severity score, which is obtained and quantified by acquiring the patient's clinical information through the hospital information system; S_frequency is the historical access frequency score, which is obtained based on the statistics of the number of historical accesses of the image data sequence; S_time is the data timeliness score, which is calculated based on the difference between the generation time of the image data sequence and the current time; S_volume is the data storage volume score, which is calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weight coefficients of S_severity, S_frequency, S_time, and S_volume, respectively.

[0025] According to the aforementioned storage space management system, preferably, the data retention value assessment module uses the following formula to calculate the retention value score for each image data sequence: Score = (W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume)×m Equation 2 Wherein, Score is the retention value score; S_severity is the severity score, obtained by quantifying the patient's clinical information through a hospital information system (such as a HIS system); S_frequency is the historical access frequency score, obtained based on the statistical analysis of the number of historical accesses to the image data sequence; S_time is the data timeliness score, calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weighting coefficients of S_severity, S_frequency, S_time, and S_volume, respectively; m is the correlation correction coefficient. If multiple image data sequences of the same patient exist in the local storage space, and the time interval between the generation times of the multiple image data sequences is less than a preset time value, the value of m is 1.2; if multiple image data sequences of the same patient exist in the local storage space, and the time interval between the generation times of the multiple image data sequences is greater than the preset time value, the value of m is 1; if only one image data sequence of the same patient exists in the local storage space, the value of m is 1. More preferably, the preset time value is 30 days.

[0026] According to the aforementioned storage space management system, preferably, the clinical information includes clinical status information and clinical diagnostic results. The formula for calculating S_severity is: S_severity = Clinical status information score + Clinical diagnostic result score; wherein, the clinical status information includes emergency, ICU, surgery, general hospitalization, general outpatient, discharge / transfer, and the corresponding clinical status information scores are 10, 8, 6, 4, 2, and 0 points, respectively; the clinical diagnostic results include: critical diagnosis, severe diagnosis, and general diagnosis, and the corresponding clinical diagnostic result scores are 2, 1, and 0 points, respectively.

[0027] According to the above storage space management system, preferably, the formula for calculating S_frequency is: S_frequency = min(10, N_access × k); where N_access is the number of times the image data sequence has been accessed within the most recent sliding window (e.g., the most recent 7 days); and k is the frequency sensitivity coefficient, which takes a value of 2.

[0028] According to the aforementioned storage space management system, preferably, the calculation formula for S_time is: S_time = max(0, 10 - (T_current - T_create) / T_decay); where T_current is the current system time, T_create is the image data generation time, and T_decay is the decay period constant. More preferably, T_decay is set to 3.

[0029] According to the aforementioned storage space management system, preferably, the formula for calculating S_volume is: S_volume = S v × Storage medium trimming factor; where S v = min(10, (File_Size / Max_Size)×10), where File_Size is the size of the image data sequence file, and Max_Size is the size of the largest image data sequence among all image data sequences in the local storage space; the storage media trimming factor includes the SSD storage factor and the HDD storage factor, with the SSD storage factor being 1 and the HDD storage factor being 0.5.

[0030] According to the above storage space management system, preferably, the calculation method of Max_Size is as follows: take the 95th percentile of the size of all image data sequence files in the local storage space in the past 3 months as the value of Max_Size or Max_Size = total capacity × 0.1.

[0031] According to the above-mentioned storage space management system, preferably, the storage space management system further includes a data hierarchical storage module. The data hierarchical storage module is used to divide the remaining image data sequences after being cleaned by the storage space cleaning module into three levels: hot data, warm data, and cold data, based on the data retention value score. The hot data level image data sequences are retained in local SSD fast storage, the warm data level image data sequences are migrated to the network edge cache, and the cold data level image data sequences are archived to cloud storage.

[0032] According to the aforementioned storage space management system, preferably, and more preferably, the retention value score of the hot data level image data sequence is greater than the hot data classification threshold, the retention value score of the warm data level image data sequence is greater than the cold data classification threshold and less than or equal to the hot data classification threshold, and the retention value score of the cold data level image data sequence is less than or equal to the cold data classification threshold. Further, when a cleanup condition is triggered, it can be set to prioritize the cleanup of cold data, followed by the cleanup of warm data, and finally the cleanup of hot data (cleaning only when local storage space is severely insufficient).

[0033] According to the aforementioned storage space management system, preferably, the hot data classification threshold is set to 7 and the cold data classification threshold is set to 3. Furthermore, the hot data classification threshold and cold data classification threshold are not fixed. The system periodically reviews the retention and cleanup of image data sequences, dynamically adjusting the values ​​of the hot data classification threshold and cold data classification threshold based on the recall rate and storage pressure index. The recall rate refers to the proportion of image data sequences that, after being identified as "cold data" and stored in the cloud, are retrieved by doctors again within a preset short-term observation window (e.g., 24h or 48h). The storage pressure index refers to the average occupancy rate of local storage space within a statistical period (e.g., the past 3-7 days).

[0034] According to the aforementioned storage space management system, preferably, the data pre-caching module determines the pre-caching priority based on the severity of the patient's condition, examination time, and access prediction, prioritizing the caching of high-priority and medium-priority image data; among them, image data of emergency patients, ICU patients, and examination data recently performed within the past 12-24 hours (the specific time can be adjusted according to needs) are of high priority, image data of surgical patients and general inpatients are of medium priority, and image data of general outpatients are of low priority.

[0035] Compared with the prior art, the technical effects achieved by the present invention are as follows: (1) Based on bandwidth load monitoring, this invention performs predictive data prefetching during network idle periods, transforming high-concurrency random I / O requests into local sequential reads; it constructs a multi-dimensional evaluation model that includes physical storage cost and business value weights, and monitors the status of local storage media; when triggering conditions such as insufficient storage space, changes in patient discharge status, or data aging are met, a data retention value score is calculated; a negative-weighted data volume score item is introduced into the retention value score to balance physical storage costs, and dynamic zeroing of discharge status and cache invalidation are achieved by listening to HIS status signals; a weighted cache replacement strategy is executed according to the score. Therefore, the storage space management method of this invention effectively solves the I / O bottleneck problem in massive image data retrieval. Through asynchronous event-driven and multi-dimensional value calculation, it maximizes the cache hit rate within limited physical storage space, avoids cache jitter, and achieves efficient scheduling of storage resources.

[0036] (2) Existing technologies (such as LRU / LFU) and conventional scoring algorithms are all based on the logic of "maximizing benefits", that is, "the one with the higher score stays". When calculating the retention value score, this invention innovatively introduces the algorithmic idea of ​​"return on investment (ROI)" and constructs a "value density optimization model" through the negative feedback penalty term "-W4×S_volume", which breaks through the "value density blind spot" of traditional cache replacement algorithms. Moreover, experiments show that compared with algorithms that only consider access frequency, this scheme can accommodate 40%-60% more effective files (Cache Entries) under the same storage capacity, which significantly reduces I / O thrashing caused by "frequent swapping in and out of large files".

[0037] (3) This invention abandons the traditional "polling" or "time-to-live (TTL)" mechanism and establishes a "cross-system cache consistency protocol". By subscribing to the ADT (Admission / Discharge / Transfer) signaling of the HIS system, the system can capture the "cache invalidation interruption" of "patient discharge" within milliseconds. This is not only a mapping of business rules, but also solves the typical "data state synchronization" problem in distributed systems. It ensures that the data lifecycle in the storage system is strictly synchronized with the clinical business cycle, eliminating the invalid occupation of physical resources by "zombie data".

[0038] (4) By performing predictive prefetching during periods of bandwidth idle time (triggered by traffic monitoring), this invention transforms random read I / O requests, which were originally concentrated during peak treatment periods, into sequential write I / O operations during off-peak times. This not only reduces the peak bandwidth pressure on the core network, but also reduces the image loading latency from seconds to milliseconds through the fast random read characteristics of the local SSD.

[0039] (5) Based on the scoring results, the data is mapped to storage media of different performance levels (local SSD, edge cache, cloud object storage), achieving Pareto Optimality of storage cost and access performance. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of multi-level storage. Figure 2 A schematic diagram of the storage space management method triggered when the local storage space utilization rate exceeds a preset storage threshold; Figure 3A schematic diagram of the storage space management method triggered by changes in patient clinical status information corresponding to image data sequences in local storage space; Figure 4 This is a schematic diagram of the medical image data storage space management system of the present invention. Detailed Implementation

[0041] To enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below with reference to specific embodiments.

[0042] Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0043] Terminology Explanation: In the embodiments of this invention, unless otherwise specified, the "Clean / Eviction" operation specifically refers to releasing physical storage space on edge nodes or local servers (e.g., executing the unlink command of a file system or the TRIM command of an SSD) to free up cache resources. This does not imply permanent data destruction. In the full lifecycle management of medical image data, all original copies of image data are archived long-term or permanently retained in cloud object storage or cold storage systems in hospital central computer rooms. Therefore, the "cleaning" in this invention is actually "cache replacement" or "local copy removal," ensuring data security and traceability.

[0044] Example 1: A method for managing the storage space of medical image data, comprising the following steps: S1: Based on the patient list managed by the doctor, proactively pre-cache patient image data to local storage during periods of bandwidth idleness; these periods of bandwidth idleness include times when network bandwidth utilization is below 30% and system idle time. S2: Monitor the local storage space usage rate, the patient clinical status information corresponding to the image data sequences in the local storage space, and the storage time of the image data sequences in the local storage space; when the local storage space usage rate exceeds the preset storage threshold, execute step S3; when the patient clinical status information corresponding to the image data sequences in the local storage space changes to discharge, execute step S4; when the storage time of the image data sequences in the local storage space exceeds the preset time threshold, execute step S5. S3: Calculate the retention value score of all image data sequences in the local storage space, sort the image data sequences in order of retention value score from low to high, and clean up image data sequences with retention value scores lower than the preset score threshold. S4: Place the image data sequences of patients whose clinical status information has been changed to discharge into the delay queue, and set the TTL period of the delay queue. When the delay queue reaches the TTL period, calculate the retention value score of all image data sequences of the patient in the delay queue, and clean up the image data sequences whose retention value score is lower than the preset score threshold. S5: Calculate the retention value score of all image sequences whose storage time exceeds a preset time threshold, sort the image data sequences in order of retention value score from low to high, and clean up image data sequences whose retention value score is lower than the preset score threshold.

[0045] In step S1, the pre-caching priority is determined based on the severity of the patient's condition, examination time, and access prediction. High-priority and medium-priority image data are cached first. Among them, image data of emergency patients, ICU patients, and examination data that have just been performed within the past 12-24 hours (the specific time can be adjusted according to the needs) are high-priority, image data of surgical patients and general inpatients are medium-priority, and image data of general outpatients are low-priority.

[0046] In steps S3, S4, and S5, the calculation formula for the retention value score is the same, and the specific calculation formula is as follows: Score =W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume; Among them, Score is the retention value score; S_severity is the severity score, which is obtained and quantified by acquiring the patient's clinical information through a hospital information system (such as a HIS system), and the clinical information includes clinical status information and clinical diagnosis results; S_frequency is the historical access frequency score, which is obtained based on the statistics of the number of historical accesses to the image data sequence; S_time is the data timeliness score, which is calculated based on the difference between the generation time of the image data sequence and the current time; S_volume is the data storage volume score, which is calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weight coefficients of S_severity, S_frequency, S_time, and S_volume, respectively.

[0047] The formula for calculating S_severity is: S_severity = Clinical Status Information Score + Clinical Diagnosis Result Score; where clinical status information includes emergency room, ICU, surgery, general inpatient, general outpatient, and discharge / transfer, with scores of 10, 8, 6, 4, 2, and 0 respectively; clinical diagnosis results include critical, severe, and general diagnoses, with scores of 2, 1, and 0 respectively. Patient clinical status and diagnosis results are obtained through the API interface of a hospital information system (such as a HIS system), establishing a patient clinical information synchronization mechanism. When the patient's status changes in the HIS system, the information is updated synchronously. The synchronization cycle is either immediate synchronization (synchronization immediately upon change of patient status) or scheduled synchronization (scheduled synchronization can be set to every 5 minutes). When the patient's status changes, the severity score is updated synchronously. Specifically, when a patient is discharged, although the severity score S_severity drops to 0, if the doctor continues to frequently review the patient's imaging data for reasons such as research review, discussion of difficult cases, or re-examination, the review frequency score S_frequency in the scoring formula will remain at a high level, thus keeping the total score within a certain range and preventing the data from being deleted immediately. This reflects the system's intelligent adaptation to actual clinical use scenarios.

[0048] The formula for calculating S_frequency is: S_frequency = min(10, N_access × k); where N_access is the number of times the image data sequence has been accessed within the most recent sliding window (e.g., the last 7 days), obtained from access log statistics; k is the frequency sensitivity coefficient, with a default value of 2. This means that if the sequence is accessed 5 times or more in the last 7 days, it receives a perfect score (10 points), reflecting a high sensitivity to hot data. Calculation example: A CT image data sequence was viewed by a doctor 3 times in the last 7 days: S_frequency = min(10, 3×2) = 6 points; An MRI sequence used for teaching was viewed 20 times in the last 7 days: S_frequency = min(10, 20×2) = 10 points.

[0049] The formula for calculating S_time is: S_time = max(0, 10 - (T_current - T_create) / T_decay); where T_current is the current system time, T_create is the image data generation time, and T_decay is the decay period constant, usually set to 3 by default (i.e., 1 point is deducted every 3 days). Calculation example: Data generated just 1 day ago: (T_current - T_create) / T_decay) = (1 / 3) ≈ 0.33, 10 - (T_current - T_create) / T_decay) = 9.67, S_time = max(0, 10 - (T_current - T_create) / T_decay) = 9.67 points. Data for 15 days was generated: (T_current - T_create) / T_decay) = 15 / 3 = 5, 10 - (T_current - T_create) / T_decay) = 5, S_time = max(0, 10 - (T_current - T_create) / T_decay) = 5 minutes. Data for 30 days was generated: (T_current - T_create) / T_decay) = 30 / 3 = 10, 10 - (T_current - T_create) / T_decay) = 0, S_time = max(0, 10 - (T_current - T_create) / T_decay) = 0 minutes.

[0050] The formula for calculating S_volume is: S_volume = S v × Storage medium trimming factor; where S v = min(10, (File_Size / Max_Size)×10), where File_Size is the size of the image data sequence file, and Max_Size is the size of the largest image data sequence among all image data sequences in the local storage space; the storage media adjustment factor includes SSD storage factor and HDD storage factor, with SSD storage factor being 1 and HDD storage factor being 0.5. Preferably, Max_Size is calculated as follows: take the 95th percentile of the size of all image data sequence files in the local storage space over the past 3 months as the value of Max_Size, or Max_Size = total capacity × 0.1.

[0051] The weighting coefficients W1, W2, W3, and W4 were determined based on historical data analysis and the experience of storage management experts. Historical data analysis involved statistically analyzing storage space utilization, data access hit rate, proportion of emergency / ICU patient data, and average file size over the past three months. Based on this historical data, the initial weighting ranges were determined as follows: W1 = 0.3-0.5, W2 = 0.2-0.4, W3 = 0.1-0.3, and W4 = 0.1-0.3. Expert adjustment incorporated the experience of storage management experts to adjust the weighting ranges determined by the historical data. After expert adjustment, the initial weighting coefficients W1, W2, W3, and W4 were as follows: W1 = 0.4, W2 = 0.3, W3 = 0.2, and W4 = 0.1. Furthermore, after determining the initial values ​​of the weight coefficients W1, W2, W3, and W4, their values ​​can be adaptively adjusted and optimized according to the actual situation. The adjustment and optimization method is as follows: statistically analyze the storage space utilization rate, data access hit rate, frequency of re-downloading after cleaning, and data retention rate of emergency / ICU patients over the past 30 days; if the storage space utilization rate is <70%, increase the weight of W4 (considering the need to increase storage costs), with an adjustment range not exceeding 0.05; if the data access hit rate is <75%, increase the weight of W2 (to increase access frequency). Considering the following: If the re-download frequency is >20%, increase the W1 weight (to increase the severity of the condition), with an adjustment range not exceeding 0.05; if the data retention rate for emergency / ICU patients is <95%, increase the W1 weight (to increase the severity of the condition), with an adjustment range not exceeding 0.05; Adjustment cycle: every 30 days, or immediately when indicators such as storage space utilization and data access hit rate are abnormal; Optimization goal: to keep storage space utilization within a reasonable range while ensuring that the data retention rate for emergency / ICU patients is ≥95%. Additionally, doctors in different departments can manually adjust the values ​​of weight coefficients W1, W2, W3, and W4 according to their actual work needs.

[0052] Example 2: The content of Example 2 is basically the same as that of Example 1, except that: In steps S3, S4, and S5, the calculation formula for the retention value score is the same, and the specific calculation formula is as follows: Score = (W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume)×m; Among them, Score is the retention value score; S_severity is the severity score, obtained and quantified by acquiring patient clinical information through a hospital information system (such as a HIS system), which includes clinical status information and clinical diagnosis results; S_frequency is the historical access frequency score, obtained based on the number of historical accesses to the image data sequence; S_time is the data timeliness score, calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are S_severity, S_frequency, S_time, and S_volume, respectively. S_volume is the weighting coefficient; m is the correlation correction coefficient. If there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation time of the multiple image data sequences is less than a preset threshold (the preset threshold can be set to 30 days), the value of m is 1.2; if there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation time of the multiple image data sequences is greater than the preset threshold (the preset threshold can be set to 30 days), the value of m is 1; if there is only one image data sequence of the same patient in the local storage space, the value of m is 1.

[0053] Example 3: The content of Example 3 is basically the same as that of Example 1, except that the storage space management method further includes: after cleaning up the image data sequences whose retention value scores are lower than the preset score threshold, storing the remaining image data sequences in the local storage space in multiple levels according to the retention value scores.

[0054] The specific operation of storing the remaining image data sequences in the local storage space in multiple levels according to the retention value score is as follows (e.g.) Figure 1 As shown): Based on the retention value score, the remaining image data sequences are divided into three levels: hot data, warm data, and cold data. Hot data sequences are stored locally on a fast SSD, warm data sequences are migrated to a network edge cache, and cold data sequences are archived to cloud storage. Specifically, the retention value score of a hot data sequence is greater than the hot data classification threshold; the retention value score of a warm data sequence is greater than the cold data classification threshold but less than or equal to the hot data classification threshold; and the retention value score of a cold data sequence is less than or equal to the cold data classification threshold. Furthermore, subsequent settings can be configured to prioritize the cleanup of cold data, followed by warm data, and finally hot data when image data sequence cleanup conditions are triggered.

[0055] Preferably, the hot data classification threshold is set to 7, and the cold data classification threshold is set to 3. Furthermore, the hot data classification threshold and cold data classification threshold are not fixed. The system periodically reviews the retention and cleaning of image data sequences and dynamically adjusts the values ​​of the hot data classification threshold and cold data classification threshold based on the recall rate and storage pressure index. The recall rate refers to the proportion of image data sequences that, after being identified as "cold data" and stored in the cloud, are retrieved by doctors within a preset short-term observation window (e.g., 24h or 48h). If the recall rate exceeds a preset warning value (e.g., 5%), it indicates that the current cold data judgment standard is too aggressive, leading to erroneous cleaning. The system should automatically lower the cold data classification threshold to increase the amount of data retained. The storage pressure index refers to the average occupancy rate of local storage space within a statistical period (e.g., the past 3 to 7 days). If the average occupancy rate is consistently higher than the preset safety level (e.g., 90%), it indicates that local high-speed storage resources are strained. The system should automatically increase the hot data classification threshold to reduce the number of image data sequences residing in local storage space and free up physical space.

[0056] Example 4: Example 4 is basically the same as Example 3, except that: In steps S3, S4, and S5, the calculation formula for the retention value score is the same, and the specific calculation formula is as follows: Score = (W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume)×m; Among them, Score is the retention value score; S_severity is the severity score, obtained and quantified by acquiring patient clinical information through a hospital information system (such as a HIS system), which includes clinical status information and clinical diagnosis results; S_frequency is the historical access frequency score, obtained based on the number of historical accesses to the image data sequence; S_time is the data timeliness score, calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are S_severity, S_frequency, S_time, and S_volume, respectively. S_volume is the weighting coefficient; m is the correlation correction number. If there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation time of the multiple image data sequences is less than a preset threshold (the preset threshold can be set to 30 days), the value of m is 1.2; if there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation time of the multiple image data sequences is greater than the preset threshold (the preset threshold can be set to 30 days), the value of m is 1; if there is only one image data sequence of the same patient in the local storage space, the value of m is 1.

[0057] Example 5: A method for proactively pre-caching patient image data to local storage during periods of bandwidth idleness. This embodiment provides an implementable operational example of step S1 in the medical image data storage space management method described in Embodiment 1 or Embodiment 2 above. The specific steps are as follows: S11: Building the Prefetch Queue: The system scans the task list of the doctor's workstation through the interface (Example: Dr. Zhang currently manages 40 patients, including 5 emergency patients, 10 ICU patients, 15 general inpatients, and 10 general outpatients). S12: Priority Calculation (1) High priority (P1): A. Imaging data of emergency patients (5 people) and ICU patients (10 people); these two types of patients have rapidly changing conditions and have the highest requirements for real-time image access, so they are high priority.

[0058] B. New image data generated in the past 12 hours: This is real-time hotspot data, and therefore, it is also of high priority.

[0059] (2) Medium priority (P2): Imaging data of ordinary inpatients (15 people) (because the condition of ordinary inpatients is relatively stable, the time for reviewing data during ward rounds is relatively fixed).

[0060] (3) Low priority (P3): Imaging data of general outpatients (10 people) (general outpatients have high mobility and a relatively low probability of re-examination).

[0061] S13: Perform Asynchronous Prefetching: When the network traffic monitoring module detects that the core switch bandwidth utilization is <30% (e.g., 02:00-05:00), the system starts a background transmission thread to write image data chunks to local high-speed storage (SSD) in the order of P1>P2>P3. For example, in P1 stage: data from 5 emergency patients and 10 ICU patients (approximately 10.5GB in total) are written first, along with recent examination hotspot data. In P2 stage: data from 15 general inpatients are then written.

[0062] Through the operations described in steps S11 to S13, the "high-concurrency random network reads" that might occur during peak consultation periods are transformed into "off-peak sequential network transmission + local SSD random reads". The performance benefits obtained through steps S11 to S13 are: 1) Reduced I / O latency: When doctors click to view images during the morning peak the next day, the data is read directly from the local NVMe SSD, reducing the response time from an average of 1800ms (network loading) to 20ms (local I / O); 2) Bandwidth protection: Avoiding intra-hospital network congestion caused by a large number of concurrent downloads during the morning rounds peak.

[0063] Example 6: Storage space management triggered when local storage space utilization exceeds a preset storage threshold This embodiment mainly provides an implementable example of the storage space management method triggered in step S2 of the medical image data storage space management method described in Embodiment 1 or Embodiment 3 above, where the local storage space usage rate exceeds a preset storage threshold. Specific steps (such as...) Figure 2 As shown below: S31: Storage Status Monitoring: The total local storage space is 500GB. Currently, it stores image data sequences of 500 patients. The storage space occupied by the image data sequences of 500 patients is 425GB. The system monitors that the local storage space utilization rate is 85%, which exceeds the preset threshold of 80%. S32: Calculate the data retention value score: Calculate the retention value score of each of the 500 image data sequences according to the retention value score calculation formula; wherein, the retention value score calculation formula is: Score = W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume.

[0064] Taking patient A's CT image data sequence as an example, the retention value score is calculated as follows: 1) Severity of condition: emergency state, S_severity = 10 points; 2) Viewing frequency: 5 times in the last 7 days, S_frequency = 8 points; 3) Time since: 3 days ago, S_time = 9 points; 4) Data volume: file size 200MB, S_volume = 4 points; 5) Weighting coefficients: W1=0.4, W2=0.3, W3=0.2, W4=0.1; Therefore, the retention value score of patient A's CT image data sequence is: Score = 0.4×10 + 0.3×8 + 0.2×9 - 0.1×4 = 7.8 points.

[0065] S33: Select cleaning targets and perform cleaning: According to the retention value score calculated in step S32, sort the 500 image data sequences in order of retention value score from low to high, select 100 image data sequences with retention value lower than the preset score threshold for cleaning, release 80GB of storage space, and reduce the storage space utilization rate to 69%.

[0066] By performing the steps S31-S33 described above, storage space is effectively freed up while retaining high-clinical-value data.

[0067] Example 7: Storage space management triggered by changes in patient clinical status information corresponding to image data sequences in local storage space. This embodiment mainly provides an implementable example of the storage space management method for medical image data described in Embodiment 1 or Embodiment 3 above, specifically the storage space management method triggered by step S2 when monitoring the change of the patient's clinical status information corresponding to the image data sequence in the local storage space to discharge. Specific steps (such as...) Figure 3 As shown below: S41: Event Capture: The trigger condition detection module subscribes to the HIS system's ADT (admission / discharge / transfer) events via WebHook or MQ (message queue) and captures patient B's Discharge_Event (discharge event) with a timestamp of T_discharge.

[0068] S42: Delay Queue Processing: To avoid temporary backtracking during discharge procedures, cleanup tasks are placed in a delay queue with a TTL of 7 days.

[0069] S43: When the delayed queue reaches its TTL (i.e., 7 days later), the retention value score calculation is triggered. The retention value score for all image data sequences in the delayed queue is calculated according to the retention value score calculation formula: Score = W1 × S_severity + W2 × S_frequency + W3 × S_time - W4 × S_volume. Because Patient B's status changes to "discharged," the S_severity factor in the retention value score calculation formula for Patient B's image data sequences is forcibly set to zero. Furthermore, because S_severity is zero, the retention value score of Patient B's image data sequences will decrease significantly.

[0070] S44: Select cleaning targets and perform cleaning: Based on the retention value score calculated in step S43, select image data sequences with retention value scores lower than the preset score threshold (preferably, the preset score threshold is 3 points) for cleaning to release storage space.

[0071] Special Scenario Description (High-Frequency Access Protection): Assume patient B has been discharged, but due to the special nature of the case, the doctor conducted 10 teaching reviews within 3 days of discharge. Dynamic Correction: At this point, S_frequency (access frequency) spikes to 10 points. Furthermore, patient B has only one image data sequence for the same patient in the local storage space, with a correlation coefficient m of 1. The retention value score is calculated as: Score = (0.4 × 0 + 0.3 × 10 + 0.2 × 9 - 0.1 × 3) × 1 = 4.5 points. Since the retention value score of 4.5 points is greater than the preset score threshold for cleaning (preferably 3 points), the system identifies this as "atypical hot data," intelligently intercepts the cleaning command, and retains or downgrades it to warm storage, rather than deleting it directly. This demonstrates the algorithm's ability to accurately capture the "true business popularity."

[0072] By performing the above steps S41-S44, storage space is effectively freed up while retaining high-clinical-value data.

[0073] Example 8: Storage space management triggered when the storage time of image data sequences in local storage exceeds a preset time threshold. This embodiment mainly provides an implementable example of the storage space management method triggered in step S2 of the medical image data storage space management method described in Embodiment 1 or Embodiment 3 above, where the storage time of the image data sequence in the local storage space exceeds a preset time threshold. The specific steps are as follows: S51: It was detected that the storage time of 50 image data sequences in the local storage space has exceeded the preset time threshold (90 days). S52: Calculate the data retention value score: Calculate the retention value score of the 50 image data sequences whose storage time exceeds the preset time threshold according to the retention value score calculation formula; wherein, the retention value score calculation formula is: Score = W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume.

[0074] S53: Select cleaning targets and perform cleaning: Based on the retention value score calculated in step S52, sort the 50 image data sequences in order of retention value score from low to high, select 30 image data sequences with retention value lower than the preset score threshold for cleaning, and release approximately 60GB of storage space.

[0075] By performing the steps S51-S53 described above, storage space is effectively freed up while retaining high-clinical-value data.

[0076] Example 9: Adaptive adjustment method for weight coefficients W1, W2, W3, and W4 of S_severity, S_frequency, S_time, and S_volume This embodiment mainly provides an implementable example of the adaptive adjustment method for the weight coefficients W1, W2, W3, and W4 of S_severity, S_frequency, S_time, and S_volume in the retention value score calculation formula of the medical image data storage space management method described in Embodiment 1 or Embodiment 3 above. The specific steps are as follows: 1. Historical data statistics: 1) Storage space utilization rate over the past 30 days: average 75%; 2) Access hit rate of statistical data: average 85%; 3) Frequency of re-downloading after cleaning: average 10%; 4) Data retention rate of emergency / ICU patients: average 98%.

[0077] 2. Weight Adjustment: The initial values ​​of the weighting coefficients W1, W2, W3, and W4 are as follows: W1 = 0.4, W2 = 0.3, W3 = 0.2, and W4 = 0.1.

[0078] If storage space utilization is below 80%, increase the weight of w4: adjust w4 from 0.1 to 0.15 (adjustment amount 0.05); If the data access hit rate is higher than 80%, keep the w2 weight unchanged; Emergency / ICU patient data retention rate is above 95%, and the w1 weight remains unchanged; Adjusted weights: W1=0.4, W2=0.3, W3=0.15, W4=0.15.

[0079] The effects of adaptive adjustment: considering increased storage costs, large-volume data is prioritized for cleaning; storage space utilization is increased to 82%; and the retention rate of emergency / ICU patient data remains above 98%.

[0080] Example 10: Medical Imaging Data Storage Management System A medical image data storage space management system, such as Figure 4 As shown, the system includes a data pre-caching module, a storage status monitoring module, a trigger condition detection module, a data retention value assessment module, and a storage space cleanup module. Once operational, this medical image data storage space management system can implement the medical image data storage space management method described in Example 1.

[0081] The system includes several modules: a data pre-caching module connected to the Hospital Information System (HIS) to obtain the patient list managed by doctors, monitor network bandwidth status, and proactively pre-cachise image data to local storage during periods of bandwidth idleness; a storage status monitoring module to monitor the usage of local storage space and the status of image data sequences stored therein in real time; a trigger condition detection module receiving signals from the storage status monitoring module and listening to the clinical status information of patients in the HIS to detect whether the storage space cleanup trigger conditions are met; a data retention value assessment module receiving storage space cleanup trigger signals and calculating the retention value score for each image data sequence; and a storage space cleanup module executing a local storage data cleanup strategy (cleaning up image data sequences with retention value scores below a preset threshold) based on the data retention value scores, thus releasing physical storage space. The trigger conditions are: monitoring that the local storage space usage exceeds a preset storage threshold; monitoring that the clinical status information of the patient corresponding to the image data sequence in the local storage space changes to discharge; or monitoring that the storage time of the image data sequence in the local storage space exceeds a preset time threshold.

[0082] The data retention value assessment module uses the following formula to calculate the retention value score for each image data sequence: Score =W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume; Among them, Score is the retention value score; S_severity is the severity score, which is obtained and quantified by acquiring the patient's clinical information through the hospital information system (such as the HIS system); S_frequency is the historical access frequency score, which is obtained based on the statistics of the number of historical accesses to the image data sequence; S_time is the data timeliness score, which is calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, which is calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weight coefficients of S_severity, S_frequency, S_time, and S_volume, respectively.

[0083] The formula for calculating S_severity is: S_severity = Clinical Status Information Score + Clinical Diagnosis Result Score; where clinical status information includes emergency room, ICU, surgery, general inpatient, general outpatient, and discharge / transfer, with scores of 10, 8, 6, 4, 2, and 0 respectively; clinical diagnosis results include critical, severe, and general diagnoses, with scores of 2, 1, and 0 respectively. Patient clinical status and diagnosis results are obtained through the API interface of a hospital information system (such as a HIS system), establishing a patient clinical information synchronization mechanism. When the patient's status changes in the HIS system, the information is updated synchronously. The synchronization cycle is either immediate synchronization (synchronization immediately upon change of patient status) or scheduled synchronization (scheduled synchronization can be set to every 5 minutes). When the patient's status changes, the severity score is updated synchronously. Specifically, when a patient is discharged, although the severity score S_severity drops to 0, if the doctor continues to frequently review the patient's imaging data for reasons such as research review, discussion of difficult cases, or re-examination, the review frequency score S_frequency in the scoring formula will remain at a high level, thus keeping the total score within a certain range and preventing the data from being deleted immediately. This reflects the system's intelligent adaptation to actual clinical use scenarios.

[0084] The formula for calculating S_frequency is: S_frequency = min(10, N_access × k); where N_access is the number of times the image data sequence has been accessed within the most recent sliding window (e.g., the last 7 days), obtained from access log statistics; k is the frequency sensitivity coefficient, with a default value of 2. This means that as long as it has been accessed 5 times or more in the last 7 days, this item will receive full marks (10 points), reflecting a high sensitivity to hot data.

[0085] The formula for calculating S_time is: S_time = max(0, 10 - (T_current - T_create) / T_decay); where T_current is the current system time, T_create is the image data generation time, and T_decay is the decay period constant, which is usually set to 3 by default (i.e., 1 point is deducted every 3 days).

[0086] The formula for calculating S_volume is: S_volume = S v × Storage medium trimming factor; where S v = min(10, (File_Size / Max_Size)×10), where File_Size is the size of the image data sequence file, and Max_Size is the size of the largest image data sequence among all image data sequences in the local storage space; the storage media adjustment factor includes SSD storage factor and HDD storage factor, with SSD storage factor being 1 and HDD storage factor being 0.5. Preferably, Max_Size is calculated as follows: take the 95th percentile of the size of all image data sequence files in the local storage space over the past 3 months as the value of Max_Size, or Max_Size = total capacity × 0.1.

[0087] The weighting coefficients W1, W2, W3, and W4 were determined based on historical data analysis and the experience of storage management experts. Historical data analysis involved statistically analyzing storage space utilization, data access hit rate, proportion of emergency / ICU patient data, and average file size over the past three months. Based on this historical data, the initial weighting ranges were determined as follows: W1 = 0.3-0.5, W2 = 0.2-0.4, W3 = 0.1-0.3, and W4 = 0.1-0.3. Expert adjustment incorporated the experience of storage management experts to adjust the weighting ranges determined by the historical data. After expert adjustment, the initial weighting coefficients W1, W2, W3, and W4 were as follows: W1 = 0.4, W2 = 0.3, W3 = 0.2, and W4 = 0.1. Furthermore, after determining the initial values ​​of the weight coefficients W1, W2, W3, and W4, the values ​​of the weight coefficients W1, W2, W3, and W4 can be adaptively adjusted and optimized according to the actual situation. The specific adjustment and optimization method is the same as the method described in Example 1.

[0088] Example 11: Medical Image Data Storage Management System Example 11 is basically the same as Example 10, except that the calculation formula for the data retention value assessment module to calculate the retention value score of each image data sequence is as follows: Score = (W1× S_severity + W2× S_frequency + W3× S_time - W4× S_volume)×m; Among them, Score is the retention value score; S_severity is the severity score, obtained by quantifying the patient's clinical information through the hospital information system; S_frequency is the historical access frequency score, obtained based on the statistical analysis of the number of historical accesses to the image data sequence; S_time is the data timeliness score, calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weighting coefficients of S_severity, S_frequency, S_time, and S_volume, respectively; m is the correlation correction coefficient. If there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation times of the multiple image data sequences is less than a preset time value, the value of m is 1.2; if there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation times of the multiple image data sequences is greater than a preset time value, the value of m is 1; if there is only one image data sequence of the same patient in the local storage space, the value of m is 1.

[0089] Example 12: Medical Image Data Storage Management System Example 12 is essentially the same as Example 10, except that the medical image data storage management system further includes a data tiered storage module. This module categorizes the remaining image data sequences after storage space cleanup into three levels: hot data, warm data, and cold data, based on their retention value score. Hot data sequences are stored locally on a fast SSD, warm data sequences are migrated to a network edge cache, and cold data sequences are archived to cloud storage. Furthermore, subsequent settings can be configured to prioritize cleaning cold data, followed by warm data, and finally hot data when cleanup conditions are triggered.

[0090] Specifically, the retention value score of hot data level image data sequences is greater than the hot data classification threshold; the retention value score of warm data level image data sequences is greater than the cold data classification threshold but less than or equal to the hot data classification threshold; and the retention value score of cold data level image data sequences is less than or equal to the cold data classification threshold. Preferably, the hot data classification threshold is 7 and the cold data classification threshold is 3. Furthermore, the hot data classification threshold and cold data classification threshold are not fixed. The system periodically reviews the retention and cleaning of image data sequences and dynamically adjusts the values ​​of the hot data classification threshold and cold data classification threshold based on the recall rate and storage pressure index. The recall rate refers to the proportion of image data sequences that, after being identified as "cold data" and stored in the cloud, are retrieved by doctors again within a preset short-term observation window (e.g., 24h or 48h). If the recall rate exceeds a preset warning value (e.g., 5%), it indicates that the current cold data judgment standard is too aggressive, leading to erroneous cleaning, and the cold data threshold should be lowered (e.g., the system automatically lowers the cold data classification threshold by 0.5 to increase the amount of image data retained). The storage pressure index refers to the average occupancy rate of local storage space within a statistical period (such as the past 3 to 7 days). If the average occupancy rate is consistently higher than the preset safety level (such as 90%), it indicates that local high-speed storage resources are strained. The hot data classification threshold should be automatically increased (for example, the system should automatically increase the hot data classification threshold by 0.5) to reduce the number of image data sequences residing in local storage space and free up physical space.

[0091] Through this "closed-loop feedback" mechanism that dynamically adjusts the threshold values ​​for hot data grading and cold data grading, the system can evolve itself, finding a dynamic balance between "access performance" and "storage cost" to always maintain optimal operating status.

[0092] (I) Comparative Experiment of the Medical Image Data Storage Space Management Method of the Present Invention with the Existing LRU Algorithm Experimental environment: 500GB local storage space, 500 patient image data sequences, and an experimental period of 30 days. Under this experimental environment, the storage space management method for medical image data of this invention and the existing LRU algorithm were used to manage the local storage space. The storage space utilization rate, data access hit rate, frequency of re-downloading after cleanup, retention rate of emergency / ICU patient data, average file size, image viewing waiting time, and pre-cache hit rate were evaluated and compared. Specific comparison results are shown in Table 1.

[0093] Table 1 shows the impact of cache hit rate and I / O optimization: The LRU algorithm, based on the principle of "temporal locality," is prone to cache pollution during hospital-wide data scanning (such as research statistics), leading to the evicting of critical data such as emergency room data. This invention introduces S_severity as a weight protection mechanism, ensuring the caching of critical business data and achieving a 100% local hit rate for emergency / ICU data, eliminating network I / O latency. Storage media efficiency: The LRU algorithm is insensitive to file size and may evict a 1GB file to retain a recently accessed 1KB file (or conversely, frequent swapping in and out of large files leads to cache thrashing). This invention introduces a physical cost constraint through -W4×S_volume, tending to retain smaller data volumes with equal value or prioritizing the eviction of large files with low value density, thereby reducing the write amplification factor of the SSD, extending media lifespan, and optimizing storage space utilization from 85% to 40%, achieving the goal of "carrying more business value per unit space." System response speed: By combining idle prefetching, the system reduces the average loading latency perceived by users by 99% (from minutes to milliseconds), essentially shifting the user's waiting time to the background processing during system idle hours.

[0094] (II) Comparative Experiment of the Medical Image Data Storage Space Management Method of the Present Invention with the Existing LFU Algorithm Experimental environment: 500GB local storage space, 500 patient image data sequences, and an experimental period of 30 days. Under this experimental environment, the storage space management method of this invention and the existing LFU algorithm were used to manage the local storage space. The storage space utilization rate, data access hit rate, frequency of re-downloading after cleaning, retention rate of emergency / ICU patient data, average file size, image viewing waiting time, and pre-cache hit rate were evaluated and compared. Specific comparison results are shown in Table 2.

[0095] As shown in Table 2, the LFU algorithm only considers access frequency and cannot distinguish between emergency room patients and general outpatients, potentially erasing important data from emergency room patients. This invention, considering the urgency of the condition, prioritizes the retention of data from emergency room and ICU patients, achieving a 100% data retention rate for these patients. Furthermore, the LFU algorithm does not consider data volume, failing to balance clinical value with storage costs. This invention, by considering data volume, balances clinical value with storage costs, reducing storage space utilization to 40% while maintaining a 100% data retention rate for emergency room and ICU patients. Moreover, the LFU algorithm lacks a pre-caching mechanism, requiring doctors to wait 3-5 minutes for images to download. This invention, through an intelligent pre-caching mechanism, achieves an 85% pre-caching hit rate, reducing waiting time for doctors to view images to 2-5 seconds, a reduction of 99%.

[0096] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may use the above technical content as inspiration to make changes or modifications. These are equivalent embodiments with variations. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the technical concept of the present invention still fall within the protection scope of the claims of the present invention.

Claims

1. A method for managing the storage space of medical image data, characterized in that, Includes the following steps: S1: Based on the patient list managed by the doctor, proactively pre-cache the patient's image data to the local storage space during periods of bandwidth idleness; S2: Monitor local storage space usage, patient clinical status information corresponding to image data sequences in local storage space, and storage time of image data sequences in local storage space; When the local storage space usage exceeds the preset storage threshold, step S3 is executed; When the monitoring detects that the patient's clinical status information corresponding to the image data sequence in the local storage space has changed to discharge, step S4 is executed; when the monitoring detects that the storage time of the image data sequence in the local storage space exceeds the preset time threshold, step S5 is executed. S3: Calculate the retention value score of all image data sequences in the local storage space, and clean up image data sequences whose retention value score is lower than the preset score threshold; S4: Place all image data sequences of patients whose clinical status information has been changed to discharge into a delay queue, and set a timed cleanup period for the delay queue. When the timed cleanup period for the delay queue is reached, calculate the retention value score of all image data sequences of the patients in the delay queue, and clean up image data sequences whose retention value score is lower than the preset score threshold. S5: Calculate the retention value score of all image sequences whose storage time exceeds a preset time threshold, and clean up image data sequences whose retention value score is lower than the preset score threshold.

2. The storage space management method according to claim 1, characterized in that, The formula for calculating the retention value score is shown in Equation 1: Score =W1× S_severity + W2 × S_frequency + W3× S_time - W4 × S_volume Formula 1 Among them, Score is the retention value score; S_severity is the severity score, which is obtained and quantified by acquiring the patient's clinical information through the hospital information system; S_frequency is the historical access frequency score, which is obtained based on the statistics of the number of historical accesses of the image data sequence; S_time is the data timeliness score, which is calculated based on the difference between the generation time of the image data sequence and the current time; S_volume is the data storage volume score, which is calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weight coefficients of S_severity, S_frequency, S_time, and S_volume, respectively.

3. The storage space management method according to claim 1, characterized in that, The formula for calculating the retention value score is shown in Equation 2: Score = (W1× S_severity + W2 × S_frequency + W3× S_time - W4 × S_volume) × m Equation 2; Among them, Score is the retention value score; S_severity is the severity score, obtained by quantifying the patient's clinical information through the hospital information system; S_frequency is the historical access frequency score, obtained based on the statistical analysis of the number of historical accesses to the image data sequence; S_time is the data timeliness score, calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weighting coefficients of S_severity, S_frequency, S_time, and S_volume, respectively; m is the correlation correction coefficient. If there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation times of the multiple image data sequences is less than a preset time value, the value of m is 1.2; if there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation times of the multiple image data sequences is greater than a preset time value, the value of m is 1; if there is only one image data sequence of the same patient in the local storage space, the value of m is 1.

4. The storage space management method according to claim 2 or 3, characterized in that, The clinical information includes clinical status information and clinical diagnosis results. The formula for calculating S_severity is: S_severity = Clinical status information score + Clinical diagnosis result score. Clinical status information includes emergency room, ICU, surgery, general inpatient, general outpatient, and discharge / transfer, with scores of 10, 8, 6, 4, 2, and 0 respectively. Clinical diagnosis results include critical diagnosis, severe diagnosis, and general diagnosis, with scores of 2, 1, and 0 respectively.

5. The storage space management method according to claim 2 or 3, characterized in that, The formula for calculating S_frequency is: S_frequency = min(10, N_access × k); where N_access is the number of times the image data sequence has been accessed within the most recent sliding window; k is the frequency sensitivity coefficient, with a value of 2; the formula for calculating S_time is: S_time = max(0, 10 - (T_current - T_create) / T_decay); where T_current is the current system time, T_create is the image data generation time; and T_decay is the decay period constant.

6. The storage space management method according to claim 2 or 3, characterized in that, The formula for calculating S_volume is: S_volume = S v × Storage medium trimming factor; where S v = min(10, (File_Size / Max_Size)×10), where File_Size is the size of the image data sequence file, and Max_Size is the size of the largest image data sequence among all image data sequences in the local storage space; the storage media trimming factor includes the SSD storage factor and the HDD storage factor, with the SSD storage factor being 1 and the HDD storage factor being 0.

5.

7. The storage space management method according to any one of claims 1-6, characterized in that, The storage space management method further includes: cleaning up image data sequences with retention value scores lower than a preset score threshold, and storing the remaining image data sequences in the local storage space in multiple levels according to their retention value scores; the specific operation of storing the remaining image data sequences in the local storage space in multiple levels according to their retention value scores is as follows: the remaining image data sequences are divided into three levels: hot data, warm data, and cold data according to their retention value scores, wherein the retention value score of the hot data level image data sequences is greater than the hot data classification threshold, the retention value score of the warm data level image data sequences is greater than the cold data classification threshold and less than or equal to the hot data classification threshold, and the retention value score of the cold data level image data sequences is less than or equal to the cold data classification threshold; the hot data level image data sequences are stored locally on a fast SSD, the warm data level image data sequences are migrated to a network edge cache, and the cold data level image data sequences are archived to cloud storage.

8. The storage space management method according to claim 7, characterized in that, In step S1, the pre-caching priority is determined based on the severity of the patient's condition, examination time, and access prediction, prioritizing the caching of high-priority and medium-priority image data. Among these, image data of emergency patients, ICU patients, and examination data performed within the past 12-24 hours are of high priority; image data of surgical patients and general inpatients are of medium priority; and image data of general outpatients are of low priority. Bandwidth idle periods include times when network bandwidth utilization is below 30% and system idle time.

9. A storage space management system for medical image data, characterized in that, The system includes a data pre-caching module, a storage status monitoring module, a trigger condition detection module, a data retention value assessment module, and a storage space cleanup module. The data pre-caching module connects to the hospital information system to obtain the patient list managed by doctors, monitor network bandwidth status, and proactively pre-caches image data to local storage during periods of bandwidth idleness based on the patient list managed by doctors. The storage status monitoring module monitors the usage of local storage space and the status of image data sequences stored in local storage space in real time. The trigger condition detection module receives signals from the storage status monitoring module and listens to the clinical status information of patients in the hospital information system to detect whether the storage space cleanup trigger conditions are met. The data retention value assessment module receives storage space cleanup trigger signals and calculates the retention value score for each image data sequence. The storage space cleanup module executes image data cleanup strategies in local storage space based on the data retention value scores, releasing physical storage space. The trigger conditions are: monitoring that the local storage space usage rate exceeds a preset storage threshold; monitoring that the clinical status information of the patient corresponding to the image data sequence in local storage space changes to discharge; or monitoring that the storage time of the image data sequence in local storage space exceeds a preset time threshold.

10. The storage space management system according to claim 9, characterized in that, The data retention value assessment module uses the following formula to calculate the retention value score for each image data sequence: Equation 1 or Equation 2. Score =W1 × S_severity + W2 × S_frequency + W3 × S_time - W4 × S_volume Formula 1 Score =(W1× S_severity + W2 × S_frequency + W3 × S_time - W4 × S_volume)×m Equation 2; Among them, Score is the retention value score; S_severity is the severity score, obtained by quantifying the patient's clinical information through the hospital information system; S_frequency is the historical access frequency score, obtained based on the statistical analysis of the number of historical accesses to the image data sequence; S_time is the data timeliness score, calculated based on the difference between the image data sequence generation time and the current time; S_volume is the data storage volume score, calculated based on the storage space occupied by the image data sequence; W1, W2, W3, and W4 are the weight coefficients of S_severity, S_frequency, S_time, and S_volume, respectively; m is the correlation correction coefficient. If there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation times of the multiple image data sequences is less than a preset threshold, the value of m is 1.2; if there are multiple image data sequences of the same patient in the local storage space, and the time interval between the generation times of the multiple image data sequences is greater than a preset threshold, the value of m is 1; if there is only one image data sequence of the same patient in the local storage space, the value of m is 1.

Images