Search

CN-122018803-A - Storage space management method and system for medical image data

CN122018803ACN 122018803 ACN122018803 ACN 122018803ACN-122018803-A

Abstract

The invention relates to the field of medical image information systems, and discloses a storage space management method and a storage space management system for medical image data. The storage space management method comprises the steps of 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 use rate of the local storage space, clinical state information of the patient corresponding to an image data sequence and the storage time of the image data sequence, and calculating a reserve value score of the corresponding image data sequence according to a storage space cleaning mode triggered by a triggering condition when the corresponding triggering condition is monitored, and cleaning the image data sequence with the score lower than a preset scoring threshold value to realize the local storage space management. The storage space management method effectively solves the problem of I/O bottleneck in the retrieval of massive image data, maximizes the cache hit rate in a limited physical storage space through asynchronous event driving and multidimensional reserve value calculation, avoids cache jitter, and realizes efficient scheduling of storage resources.

Inventors

  • LIANG JUNXIN
  • SUN HONGWEI
  • TAN JIAN
  • ZHAO JINGLING

Assignees

  • 郑州大学第一附属医院

Dates

Publication Date
20260512
Application Date
20260130

Claims (10)

  1. 1. A method for managing storage space of medical image data, comprising the steps of: s1, 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; S2, monitoring the utilization rate of a local storage space, the clinical state information of a patient corresponding to an image data sequence in the local storage space and the storage time of the image data sequence in the local storage space, executing a step S3 when the utilization rate of the local storage space is monitored to exceed a preset storage threshold value, executing a step S4 when the clinical state information of the patient corresponding to the image data sequence in the local storage space is monitored to be changed to be discharged, and executing a step S5 when the storage time of the image data sequence in the local storage space is monitored to exceed the preset time threshold value; s3, calculating reserve value scores of all image data sequences in a local storage space, and cleaning the image data sequences with reserve value scores lower than a preset score threshold; S4, changing clinical state information into all image data sequences of the discharged patient, putting the image data sequences into a delay queue, setting a timing cleaning period of the delay queue, calculating reserve value scores of all the image data sequences of the patient in the delay queue after the delay queue is monitored to reach the timing cleaning period, and cleaning the image data sequences with reserve value scores lower than a preset score threshold; and S5, calculating the reserve value scores of all the image sequences with the storage time exceeding a preset time threshold, and cleaning the image data sequences with the reserve value scores lower than the preset score threshold.
  2. 2. The storage space management method according to claim 1, wherein the calculation formula of the reserve value score is as shown in formula 1: Score =W 1 × S_severity + W 2 × S_frequency + W 3 × S_time - W 4 × S_volume 1 (1) The Score is a reserved value Score, the S_quality is a disease criticality Score, clinical information of patients is obtained through a hospital information system and quantified, the S_frequency is a historical access frequency Score, the S_frequency is obtained through statistics based on historical access times of image data sequences, the S_time is a data timeliness Score and is obtained through calculation based on the difference value from the time of image data sequence generation to the current time, the S_volume is a data storage volume Score and is obtained through calculation based on the size of storage space occupied by the image data sequences, and W 1 、W 2 、W 3 、W 4 is the weight coefficient of the S_quality, the S_frequency, the S_time and the S_volume respectively.
  3. 3. The storage space management method according to claim 1, wherein the calculation formula of the reserve value score is as shown in formula 2: Score =（W 1 × S_severity + W 2 × S_frequency + W 3 × S_time - W 4 × S_volume）×m Formula 2; the method comprises the steps of obtaining clinical information of a patient through a hospital information system, obtaining and quantifying the clinical information of the patient, obtaining S_priority which is a disease state criticality Score through a hospital information system, obtaining S_frequency which is a historical access frequency Score based on historical access times of image data sequences, obtaining S_time which is a data timeliness Score through calculation based on a difference value from time generation time to current time of the image data sequences, obtaining S_volume which is a data storage volume Score through calculation based on the size of storage space occupied by the image data sequences, obtaining W 1 、W 2 、W 3 、W 4 which is a weight coefficient of the S_priority, the S_frequency, the S_time and the S_volume respectively, obtaining m which is an associated correction coefficient, obtaining 1.2 value if a plurality of image data sequences exist in a local storage space of the same patient, obtaining 1 value of the m value if the time interval of the generation time of the plurality of image data sequences of the same patient exists in the local storage space is larger than the preset time value, and obtaining 1 value of m value if the time interval of the generation time of the image data sequences of the same patient in the local storage space is only.
  4. 4. The storage space management method according to claim 2 or 3, wherein the clinical information comprises clinical state information and clinical diagnosis results, the calculation formula of s_quality is s_quality=clinical state information score+clinical diagnosis result score, the clinical state information comprises emergency, ICU, operation, general hospitalization, general outpatient service, discharge/transfer, the clinical state information score corresponding to emergency, ICU, operation, general hospitalization, general outpatient service, discharge/transfer is 10, 8, 6, 4, 2 and 0 respectively, and the clinical diagnosis result comprises critical diagnosis, severe diagnosis, general diagnosis, critical diagnosis, severe diagnosis and general diagnosis is 2,1 and 0 respectively.
  5. 5. The storage space management method according to claim 2 or 3, wherein the calculation formula of the s_frequency is s_frequency=min (10, n_access×k), wherein n_access is the number of times the image data sequence is accessed in the latest sliding window, k is a frequency sensitivity coefficient, the value is 2, and the calculation formula of the s_time is s_time=max (0, 10- (t_current-t_create)/t_decay), wherein t_current is the current system time, t_create is the image data generation time, and t_decay is the decay period constant.
  6. 6. The storage space management method according to claim 2 or 3, wherein the calculation formula of s_volume is s_volume=s v ×storage medium trimming coefficient, wherein S v =min (10, (file_size/max_size) ×10), file_size is the Size of an image data sequence File, max_size is the Size of the largest image data sequence among all image data sequences in the local storage space, and the storage medium trimming coefficient includes an SSD storage coefficient and an HDD storage coefficient, the SSD storage coefficient is 1, and the HDD storage coefficient is 0.5.
  7. 7. The method for managing storage space according to any one of claims 1 to 6, further comprising the steps of cleaning image data sequences with reserve values lower than a preset score threshold, storing the remaining image data sequences in the local storage space in multiple stages according to the reserve values, dividing the remaining image data sequences into three levels of hot data, warm data and cold data according to the reserve values, wherein the reserve values of the image data sequences of the hot data levels are greater than the hot data score, the reserve values of the image data sequences of the warm data levels are greater than the cold data score and less than or equal to the hot data score, the reserve values of the image data sequences of the cold data levels are less than or equal to the cold data score, storing the image data sequences of the hot data levels in the local SSD quickly, migrating the image data sequences of the warm data levels to a cloud edge cache, and archiving the image data sequences of the cold data levels to storage.
  8. 8. The method according to claim 7, wherein in step S1, the pre-cache priority is determined according to patient illness state criticality, examination time and access prediction, and image data with high priority and medium priority are cached preferentially, wherein the image data of emergency patients, the image data of ICU patients and the examination data just performed in the past 12-24 hours are high priority, the image data of operation patients and general inpatients are medium priority, the image data of general outpatients are low priority, and the bandwidth idle period comprises a period of network bandwidth utilization less than 30% and system idle time.
  9. 9. A storage space management system of medical image data is characterized by comprising a data pre-caching module, a storage state monitoring module, a trigger condition detection module, a data retention value evaluation module and a storage space cleaning module, wherein the data pre-caching module is connected with a hospital information system and is used for acquiring a patient list managed by a doctor, monitoring a network bandwidth state and actively pre-caching image data to a local storage space in a bandwidth idle period according to the patient list managed by the doctor, the storage state monitoring module is used for monitoring the use condition of the local storage space and the state of an image data sequence stored in the local storage space in real time, the trigger condition detection module is used for receiving signals of the storage state monitoring module and monitoring clinical state information of a patient in the hospital information system and detecting whether a storage space cleaning trigger condition is met, the data retention value evaluation module is used for receiving a storage space cleaning trigger signal and calculating a retention value score of each image data sequence, and the storage space cleaning module is used for executing an image data cleaning strategy in the local storage space according to the data retention value and releasing the physical storage space, wherein the trigger condition is that the use ratio of the local storage space is monitored to exceed a preset storage threshold or the use ratio of the local storage space to the local storage space or the use condition of the image data corresponding to the local storage space is monitored to the preset storage space, the state of the local storage space exceeds the preset storage threshold or the time.
  10. 10. The storage space management system according to claim 9, wherein the data retention value evaluation module is configured to calculate a retention value score for each image data sequence according to the following formula 1 or formula 2 Score =W 1 × S_severity + W 2 × S_frequency + W 3 × S_time - W 4 × S_volume 1 (1) Score =(W 1 × S_severity + W 2 × S_frequency + W 3 × S_time - W 4 × S_volume)×m Formula 2; The method comprises the steps of obtaining clinical information of a patient through a hospital information system, obtaining and quantifying the clinical information, wherein the Score is a reserved value Score, the S_quality is a disease criticality Score, the S_frequency is a historical access frequency Score, the historical access times of image data sequences are counted and obtained based on the historical access times of the image data sequences, the S_time is a data timeliness Score, the difference value from the generation time of the image data sequences to the current time is calculated, the S_volume is a data storage volume Score and is calculated based on the size of a storage space occupied by the image data sequences, W 1 、W 2 、W 3 、W 4 is a weight coefficient of the S_quality, the S_frequency, the S_time and the S_volume respectively, m is an associated correction coefficient, if a plurality of image data sequences exist in a local storage space of the same patient, the time interval of the generation time of the image data sequences is smaller than a preset threshold, the m value is 1.2, if the time interval of the generation time of the image data sequences of the same patient in the local storage space of the same patient is larger than the preset threshold, and the m value is 1, and the m value of the image data sequences of the same patient in the local storage space is only takes 1.

Description

Storage space management method and system for medical image data Technical Field The invention relates to the crossing field of medical image information systems and computer storage technologies, in particular to a storage space management method and a storage space management system for medical image data. Background With the rapid development of medical imaging technology, medical image information systems (such as PACS systems, picture ARCHIVING AND Communication System, image archiving and communication systems) are increasingly used in hospitals. A doctor can conveniently check the medical image data of a patient through the medical image information system to perform diagnosis and treatment. However, the existing medical image information system is faced with the technical problems of (1) I/O bottleneck and high delay when viewing image data, namely medical image data belongs to a typical large binary object (BLOB), and has huge volume (hundreds of MB to several GB). In the traditional Request-Response (Request-Response) mode, high concurrency retrieval can cause instantaneous saturation of server I/O throughput, cause significant network transmission delay, and cannot meet real-time rendering requirements in emergency and other scenes. (2) The problem of resource competition for physical storage media is that the physical storage space (typically SSD) of the local edge node is a scarce resource. Conventional LRU (least recently used) or FIFO (first in first out) general page replacement algorithms perform cold and hot data separation based on only a single dimension of "access time", and cannot perceive "business weight" and "physical volume cost" of data. This results in a large number of low value high volume files occupying physical sectors for a long period of time, while high value low volume files are swapped in and out frequently (THRASHING), causing ineffective wear of the storage medium and jitter in cache hit rates. Existing PACS systems or caching algorithms focus on "download priority" or "computational task scheduling", essentially a monotonically increasing function based on "time" or "frequency". However, at an edge node where physical storage space is limited, the value of data depends not only on "how much it is (revenue)", but also on "how much it is (cost)". The prior art lacks the consideration of the key dimension of Value Density (Value Density) of unit storage space, so that the system is very easy to be filled with low-efficiency data which is accessed by high frequency but has huge volume, and the living space of a large amount of small-volume high-Value data is occupied. Additionally, the hysteresis of passive cleaning, conventional schemes typically employ a fixed threshold (e.g., 90 days) or passive replacement after space is filled (LazyEviction). This mechanism fails to respond to abrupt changes in clinical traffic status (e.g., sudden discharge of a patient), resulting in a large amount of failed o "zombie data" occupying valuable SSD sectors for a long period of time, resulting in an ineffective lock on resources. In addition, the limitations of existing cache cleaning algorithms. For example, the LRU algorithm cleans up the data that was not accessed the longest considering only the last access time. The data of the emergency patient and the data of the common outpatient can not be distinguished, important data of the emergency patient can be cleaned, small files can be cleaned to reserve large files without considering the data volume, and the utilization rate of the storage space is high. And the LFU algorithm only considers the access frequency and cleans up the data with the lowest access frequency. The clinical value and the storage cost cannot be considered, important data of emergency patients can be cleaned, and the clinical value and the storage cost cannot be balanced without considering the data volume. How to reduce the time consumption of downloading through an intelligent pre-caching mechanism, and simultaneously, intelligently reserve the data with the most clinical value in a limited local storage space, clear low-value data, balance the clinical value and the storage cost and realize the optimal utilization of the storage space. The prior art mainly focuses on the problems of cache efficiency, downloading speed, retrieval speed and the like, but lacks an intelligent cleaning and recycling mechanism aiming at the condition of limited local storage space, cannot actively release the storage space when the space is insufficient, and cannot consider the key factor of data volume (storage cost). In particular, the prior art lacks a complete solution of "pre-caching+intelligent cleaning", and cannot effectively manage the local storage space while reducing the time consumed for downloading. Disclosure of Invention Aiming at the problems and the defects existing in the prior art, the invention aims to provide a storage space management method and a storage