CN106157230A - Magnanimity brain tissue 3 d image data fast call method - Google Patents
Magnanimity brain tissue 3 d image data fast call method Download PDFInfo
- Publication number
- CN106157230A CN106157230A CN201610351789.4A CN201610351789A CN106157230A CN 106157230 A CN106157230 A CN 106157230A CN 201610351789 A CN201610351789 A CN 201610351789A CN 106157230 A CN106157230 A CN 106157230A
- Authority
- CN
- China
- Prior art keywords
- data block
- image data
- magnanimity
- buffer area
- brain tissue
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/0007—Image acquisition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/60—Memory management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/04—Indexing scheme for image data processing or generation, in general involving 3D image data
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of magnanimity brain tissue 3 d image data fast call method, the method includes: in the magnanimity brain tissue 3 d image data using partitioned organization to store, tendency by matching nerve fibre, predict, along the natural bearing of trend of nerve fibre, the 3 d image data block that will use, and be loaded onto in advance in Computer Cache district；If the 3 d image data block that user's actual request is called has been located in the middle of buffer area, then read from buffer area, it is achieved quick calling.The inventive method by the tendency of nerve fibre is analyzed, targetedly preloading data to caching in, make in process of reconstruction read data time shorten, reduce period of reservation of number, improve the efficiency of neuron morphology structural remodeling.
Description
Technical field
The present invention relates to the process of mass data, in particular to a kind of magnanimity brain tissue based on nerve fibre forward prediction
Three-dimensional data caches call method, belongs to Biomedical Image process field.
Background technology
Neuron morphology structural remodeling refers to extract the shape of neuron from the image comprising neuron morphology structure
View data is converted into the vector data of neuron morphology by state.Neuron morphology structural remodeling is the basis of brain science research
Method, rebuilds neuron morphology structure accurately and plays help greatly to cognitive cerebral function.The form of neuron has local
, also there is long-range to project, the neuron of long-range projection discloses Nao Neinao district or the annexation of core group, so nerve fibre
Long-range segmentation be significant.
For long-range projection nerve fibre for, owing to nerve fibre span is longer, required from a large scale, high-resolution
3-D view in be partitioned into nerve fibre, the Treatment Analysis of data needs the data in the face of magnanimity, hundreds of GB at least, at most closely
Hundred TB.Owing to overall data volume is very big, mass image data typically can be according to some rule with the form compared with small data block
Piecemeal preserves, and only reads the data block being currently needed for processing, then use and carry out automatic or manual cutting operation when needing, complete
After becoming current data block, split followed by reading other data blocks comprising nerve fibre, so repeatedly, until completing whole
The reconstruction of individual nerve fibre.But, during segmentation is rebuild, read every time stage of data when all can consume relatively more
Between, such as several minutes, have a strong impact on the continuity of neuron process of reconstruction, total operating efficiency is substantially reduced.
Content of the invention
Present invention aim to overcome that prior art, in the deficiency solving in the problems referred to above, provides a kind of magnanimity brain tissue three
Dimensional data image fast call method, when the nerve fibre projecting long-range interacts formula segmentation, by nerve
The tendency of fiber is analyzed, in advance the data in order to use, and is loaded in advance in caching.This method makes process of reconstruction
The average time of middle reading data is greatly shortened, and reduces the stand-by period of user or calculation procedure, improves neuron morphology knot
The efficiency that structure is rebuild.
Realize that the object of the invention the technical scheme is that a kind of magnanimity brain tissue 3 d image data quick calling side
Method, the method includes:
In the magnanimity brain tissue 3 d image data using partitioned organization to store, by walking of matching nerve fibre
Gesture, predicts, along the natural bearing of trend of nerve fibre, the 3 d image data block that will use, and it is slow to be loaded onto computer in advance
Deposit in district；If the 3 d image data block that user's actual request is called has been located in the middle of buffer area, then read from buffer area, it is achieved
Quick calling.
Described buffer area, in the storage device with high-speed read-write performance, can be used for depositing 3 d image data block
Storage region.
The 3 d image data block that described actual request is called such as is present in buffer area, then read from buffer area, otherwise
Read from the low speed storage device storing magnanimity brain tissue 3 d image data.
The tendency of described matching nerve fibre includes:
The neurofibril split take from nearest one section of current data block location to be split, to this section of known fibre
The path of dimension is fitted, fit line extended line enters in current data block to be split, it was predicted that nerve fibre is currently being treated point
Cut the extension path in data block.
The method of the 3 d image data block that described prediction will use is:
Described fit line extended line enters current data block to be split, and a face passes from which, chooses and this face phase
Adjacent data block is as the preferential data block reading；Continue to choose intersection point when fit line passes data block to be split, determine and work as
The data block adjacent with this face is read by another face closest with this intersection point in front data block to be split as second priority
Data block.
When setting up buffer area, setting up two groups of dynamic labels of equal number, it is empty that one of them preserves each data block simultaneously
Between idle, another preserves the distance between each data block and current data block to be split；Utilize the letter in described dynamic labels
Data block in buffer area is ranked up by breath, deletes the portion in buffer area according to not using at most with maximum distance two indices
Divided data block, and update described dynamic labels.
Traditional segmentation neuron mode, for first reading a data block, then calculates again or processes, and this data block completes it
After, the direction further according to the nerve fibre being partitioned into is gone to read next data block.Due to read block stand-by period too
Long, affect operating efficiency.Therefore, compared with traditional segmentation neuron mode, the inventive method has the advantage that
The inventive method is carrying out neuron to the magnanimity brain tissue 3 d image data using partitioned organization to store
During the segmentation of fiber long-range, while splitting certain data block, i.e. when also not processed current data block, by obtaining
Take the position splitting nerve fibre, calculate nerve fibre tendency to prefetch data block to caching in, anticipation goes out next number
According to the probable ranges of block, carry out data loading in advance so that the time that data block loads is greatly shortened, improve operating efficiency.
Brief description
Fig. 1 is the flow chart of the present invention a kind of magnanimity brain tissue 3 d image data fast call method.
Fig. 2 a is the schematic diagram that data block is split by nerve fibre, and Fig. 2 b is the schematic diagram that DCC'D' face is divided into 4 parts,
Fig. 2 c is to be calculated the schematic diagram needing to read 2 data blocks into buffer area.
Fig. 3 a does not uses the result schematic diagram of sequence at most for using, and Fig. 3 b does not uses the result of sequence to show for using at most
Being intended to, Fig. 3 c does not uses the result schematic diagram of sequence at most for using.
Detailed description of the invention
The present invention is described in further detail with specific embodiment below in conjunction with the accompanying drawings.
In the present embodiment, the storage device of low speed read-write selects hard disk, and the storage device of high-speed read-write selects internal memory.
The targeted three-dimensional data to be split of the present invention uses partitioned organization to store, and is stored on hard disk.This enforcement
The 3 d image data that the partitioned organization using in example stores is by the data chunk of a lot of 512 × 512 × 512 sizes
Become.The size of each data block is 128MB.Refering to Fig. 1, the method that present invention caching calls comprises the steps:
S100, set up buffer area
Setting up the buffer area for depositing some data block sizes in internal memory, the quantity depositing data block can be according to meter
The configuration of calculation machine is changed, if internal memory is sufficiently large, can set up many, if big not, then that can set up lacks.Additionally build
Stand same amount of two groups of labels, one preserve each data block how long be not loaded (i.e. rest in buffer area when
Between), another preserves the distance of the data block central point from current segmentation for this data block central point.
S200, matching nerve fibre tendency, calculate required reading data
When reading data to buffer area, the face, adjacent 6 of current data block to be split is all read entrance caching
District, fetch data so next time block when always have a data block in buffer area, but do not read 6 numbers due to sliced time
Time according to block is long, and reads 6 data blocks every time and also do not have too big necessity, so the present embodiment uses selective reading
2 data blocks.
In nerve fibre cutting procedure, owing to the tendency of nerve fibre generally will not occur very violent suddenly
Change, within the specific limits, the general direction of tendency of this fiber is constant.So by segmented node of nerve fiber
Really, the tendency of also undivided part can be predicted.
This process is refering to shown in Fig. 2, and in Fig. 2 a, data block 1_1_1 is the current data block reading dividing for nerve fibre
Cutting, PQ is segmented nerve fibre, near a number of point of side draw of Q on PQ, such as 5-10, uses certain
Fit approach, such as least square method, be fitted, matching can obtain one pass through current data block a fit line
QR, fit line QR are straight line or curve (according to approximating method), and this fit line QR intersects at S with the DCC'D' face of data block
Point.
Taking the center of current data block as initial point, data block is divided into 8 parts by three normal surfaces, the line QR simulating with
Data block to be split can intersect at ABCD face and Liang Ge face, DCC'D' face, and this line pierces into from ABCD face, passes from DCC'D' face.Intend
Zygonema QR and the DCC'D' face passing and 1 S intersecting.DCC'D' face is divided into 4 regions, and as shown in Figure 2 b, S point falls
In the JOKD' of region, calculate the distance of the four edges at place, DCC'D' face that S point distance passes, it is obvious that S point and limit DC and
The distance of limit CC' ratio is remote with the distance on DD' limit and C'D' limit, therefore, only need to by distance m of S point and limit DD' and limit C'D' and
N simultaneously compares, and the distance obtaining n is short, so to read the data block into buffer area is data block 2_1_1 adjacent with DCC'D',
And data block 1_1_2 adjacent with n vertical plane A'B'C'D', as shown in Figure 2 c.The adjacent data block of preferential reading DCC'D'
2_1_1, second priority reads the adjacent data blocks being perpendicular to S point and limit C'D' place line segment.
S300, data load and read
It after being calculated data block to be read, is loaded into data block in buffer area according to precedence.Load
User is waited to read data after one-tenth, if user needs the data reading in buffer area, then direct from buffer area reading, explanation
The results contrast of matching is accurate, and data load effectively；If data are not in buffer area, then read from hard disk, matching is described
Result does not corresponds with actual, is not loaded into the data needing.
If data are in loading procedure, user's requests data reading, then stop the loading to buffer area for the data, read number
According to, if buffer area has, then load from buffer area, if buffer area does not has, then direct from hard disk loading.
S400, flush buffers district
When loading data entrance buffer area, owing to the size of buffer area is limited, so there is the full situation of buffer area, need
Want flush buffers district, some data blocks are removed from buffer area, be i.e. the refreshing of buffer area.
The refreshing of buffer area uses not use at most and is weighted sequence with maximum distance two indices, deletes part data
Block.Two groups of labels in buffer area how long are not loaded each data block in buffer area and data block is from current data
The distance of block has carried out record.When each flush buffers district, all data blocks in buffer area are entered according to the two label
Row sequence, by the two sequence according to weights be 0.5 again weighting sequence, obtain one weighting sequence, according to this ranking replacement delay
Deposit the data in district.This process can be refering to shown in Fig. 3: employing does not uses the result of sequence as shown in Figure 3 a at most, is counted
Come to be eliminated first according to block A, use shown in the result Fig. 3 b of maximum distance sequence, obtain data block C and come and wait to eliminate
First, be 0.5 weighting sequence again to data block A and data block C according to weights, obtain a weighting sequence, ranking results
As shown in Figure 3 c, data block C is eliminated, and i.e. refreshes data block C.
Claims (6)
1. a magnanimity brain tissue 3 d image data fast call method, it is characterised in that:
In the magnanimity brain tissue 3 d image data using partitioned organization to store, by the tendency of matching nerve fibre,
Predict, along the natural bearing of trend of nerve fibre, the 3 d image data block that will use, and be loaded onto Computer Cache district in advance
In；
If the 3 d image data block that user's actual request is called has been located in the middle of buffer area, then read from buffer area, it is achieved fast
Velocity modulation is used.
2. magnanimity brain tissue 3 d image data fast call method according to claim 1, it is characterised in that: described caching
District, in the storage device with high-speed read-write performance, is used for depositing the storage region of 3 d image data block.
3. magnanimity brain tissue 3 d image data fast call method according to claim 1, it is characterised in that: described reality
The 3 d image data block of request call is such as present in buffer area, then read from buffer area, otherwise from storing magnanimity brain group
Knit and read in the low speed storage device of 3 d image data.
4. magnanimity brain tissue 3 d image data fast call method according to claim 1, it is characterised in that described matching
The method of nerve fibre tendency is:
The neurofibril split take from nearest one section of current data block location to be split, to this section of known fiber
Path is fitted, and fit line extended line enters in current data block to be split, it was predicted that go out nerve fibre at current number to be split
According to the extension path in block.
5. magnanimity brain tissue 3 d image data fast call method according to claim 4, it is characterised in that described prediction
The method of the 3 d image data block that will use is:
Described fit line extended line enters current data block to be split, and a face passes from which, chooses adjacent with this face
Data block is as the preferential data block reading；
Continue to choose intersection point when fit line passes data block to be split, determine in current data block to be split with this intersection point distance
Another nearest face, the data block that the data block adjacent with this face is read as second priority.
6. magnanimity brain tissue 3 d image data fast call method according to any one of Claims 1 to 5, its feature exists
In:
When setting up buffer area, set up two groups of dynamic labels of equal number simultaneously, preserve the free time of each data block respectively,
And the distance between each data block and current data block to be split；
Utilizing the information in described dynamic labels to be ranked up the data block in buffer area, foundation does not uses and long distance at most
Delete the part data block in buffer area from two indices, and update described dynamic labels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610351789.4A CN106157230B (en) | 2016-05-25 | 2016-05-25 | Magnanimity brain tissue 3 d image data fast call method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610351789.4A CN106157230B (en) | 2016-05-25 | 2016-05-25 | Magnanimity brain tissue 3 d image data fast call method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106157230A true CN106157230A (en) | 2016-11-23 |
| CN106157230B CN106157230B (en) | 2019-08-16 |
Family
ID=57353203
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610351789.4A Active CN106157230B (en) | 2016-05-25 | 2016-05-25 | Magnanimity brain tissue 3 d image data fast call method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106157230B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101036393A (en) * | 2004-07-16 | 2007-09-12 | 索尼株式会社 | Information processing system, information processing method, and computer program |
| US20080175509A1 (en) * | 2007-01-24 | 2008-07-24 | General Electric Company | System and method for reconstructing restored facial images from video |
| CN102567944A (en) * | 2012-03-09 | 2012-07-11 | 中国人民解放军信息工程大学 | Computed tomography (CT) image reconstruction hardware accelerating method based on field programmable gate array (FPGA) |
-
2016
- 2016-05-25 CN CN201610351789.4A patent/CN106157230B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101036393A (en) * | 2004-07-16 | 2007-09-12 | 索尼株式会社 | Information processing system, information processing method, and computer program |
| US20080175509A1 (en) * | 2007-01-24 | 2008-07-24 | General Electric Company | System and method for reconstructing restored facial images from video |
| CN102567944A (en) * | 2012-03-09 | 2012-07-11 | 中国人民解放军信息工程大学 | Computed tomography (CT) image reconstruction hardware accelerating method based on field programmable gate array (FPGA) |
Non-Patent Citations (1)
| Title |
|---|
| 潘君: "基于统计模型的DTI神经纤维追踪算法研究", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106157230B (en) | 2019-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021120789A1 (en) | Data writing method and apparatus, and storage server and computer-readable storage medium | |
| CN103885728B (en) | A kind of disk buffering system based on solid-state disk | |
| CN101730885B (en) | Cache for a multi thread and multi core system and methods thereof | |
| CN104407989A (en) | Cache memory having unified tag and sliced data | |
| US20140351519A1 (en) | System and method for providing cache-aware lightweight producer consumer queues | |
| CN107247675B (en) | A kind of caching selection method and system based on classification prediction | |
| JP6020448B2 (en) | Database performance prediction apparatus and database prediction method | |
| CN110297787B (en) | Method, device and equipment for accessing memory by I/O equipment | |
| CN110532200A (en) | A kind of memory system based on mixing memory architecture | |
| CN107122126A (en) | The moving method of data, device and system | |
| CN105320608A (en) | Memory controller and method for controlling a memory device to process access requests | |
| US10489296B2 (en) | Quality of cache management in a computer | |
| CN113157605B (en) | Resource allocation method, system, storage medium and computing device for two-level cache | |
| CN113268457B (en) | Self-adaptive learning index method and system supporting efficient writing | |
| CN116560585B (en) | Data hierarchical storage method and system | |
| Lu et al. | TridentKV: A read-optimized LSM-tree based KV store via adaptive indexing and space-efficient partitioning | |
| US20200226461A1 (en) | Asynchronous early stopping in hyperparameter metaoptimization for a neural network | |
| US20240192880A1 (en) | Data processing method, apparatus, and system | |
| CN118414598A (en) | Method for managing near memory processing commands and non-near memory processing commands in a memory controller | |
| JPH08147218A (en) | Cache controller | |
| CN103903295B (en) | Method for real-time scheduling of multiple high-precision three-dimensional scene data on basis of InfiniBand | |
| CN107562806A (en) | Mix the adaptive perception accelerated method and system of memory file system | |
| CN105045894A (en) | Cache method and system oriented to distributed sequence list | |
| DE102013100169A1 (en) | Computer-implemented method for selection of a processor, which is incorporated in multiple processors to receive work, which relates to an arithmetic problem | |
| CN105955971B (en) | A kind of implementation method and device of key assignments caching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |