CN109978861A - Marrow grey matter detection method, device, equipment and computer readable storage medium - Google Patents
Marrow grey matter detection method, device, equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN109978861A CN109978861A CN201910237817.3A CN201910237817A CN109978861A CN 109978861 A CN109978861 A CN 109978861A CN 201910237817 A CN201910237817 A CN 201910237817A CN 109978861 A CN109978861 A CN 109978861A
- Authority
- CN
- China
- Prior art keywords
- marrow
- image
- grey matter
- area
- mri
- 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
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30008—Bone
- G06T2207/30012—Spine; Backbone
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Image Processing (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The present invention discloses a kind of marrow grey matter detection method, device, equipment and computer readable storage medium, this method comprises: each MRI image of vertebra is sequentially inputted to carry out the segmentation of marrow grey matter in preparatory trained U-Net model, at least one initial marrow grey matter image is obtained；Belong to a determining pixel in the corresponding initial marrow grey matter image of the MRI image in the middle part of vertebra from what is selected, and acquires the location information of pixel；Seed point is selected respectively in each MRI image to be measured according to location information, and extracts the segmented image comprising seed point respectively from each MRI image to be measured using region growing algorithm；When the area of the segmented image of extraction is less than or equal to default error with the error of the area of corresponding initial marrow grey matter image, determine that there are marrow grey matter regions for the corresponding MRI image to be measured of segmented image.The present invention is suitable for the detection to vertebra marrow grey matter regions, further increases the speed of processing detection data, and automation and intelligence degree are high.
Description
Technical field
The present invention relates to grey matter detection technique field, in particular to a kind of marrow grey matter detection method, device, equipment and meter
Calculation machine readable storage medium storing program for executing.
Background technique
Medical image can be used in describing the form of human body different tissues organ, and the health condition of reflection human body, have
Help auxiliary diagnosis and treatment.In the conventional technology, most of for the detection of marrow grey matter by the way of artificial detection, from
And lead to low efficiency and error is big, even if the technology for thering is machine to detect marrow grey matter automatically, while facing mass data
It is not high that there is also intelligence degrees when detection, causes the low problem for causing detection effect bad of at high cost and detection accuracy.
Summary of the invention
It is an object of the invention to for the deficiency in traditional technology, provide a kind of marrow grey matter detection method, device, set
Standby and computer readable storage medium.
A kind of marrow grey matter detection method of the embodiment of the present invention, comprising the following steps:
Each MRI image of vertebra is sequentially inputted to carry out the segmentation of marrow grey matter in preparatory trained U-Net model, is obtained
Take at least one initial marrow grey matter image；
Belong to a determining pixel in the corresponding initial marrow grey matter image of the MRI image in the middle part of vertebra from what is selected,
And acquire the location information of pixel；
Seed point is selected respectively in each MRI image to be measured according to location information, and using region growing algorithm from each
The segmented image comprising seed point is extracted in MRI image to be measured respectively, wherein each MRI image to be measured is each initial marrow
The corresponding MRI image of grey matter image；
When the area of the segmented image of extraction is less than or equal to the error of the area of corresponding initial marrow grey matter image
When default error, determine that there are marrow grey matter regions for the corresponding MRI image to be measured of segmented image.
In one of the embodiments, " when area and the corresponding initial marrow grey matter image of the segmented image of extraction
When the error of area is less than or equal to default error, determine that there are marrow grey matter regions for the corresponding MRI image to be measured of segmented image "
Later, further includes:
There are the areas of the corresponding initial marrow grey matter image of the MRI image of marrow grey matter regions as the first face using each
Product, and using the area of corresponding segmented image as second area；
Using the first area with the average value of corresponding second area as there are the final of the MRI image of marrow grey matter regions
Marrow grey matter regions area.
In one of the embodiments, " using the first area with the average value of corresponding second area as there are marrow ashes
After the final marrow grey matter regions area of the MRI image in matter region ", further includes:
Obtain that each there are the corresponding thickness of the MRI image of marrow grey matter regions；
Calculate the product of each thickness with corresponding final marrow grey matter regions area；
The summation of each product is calculated, and using summation as the marrow grey matter total volume of vertebra.
The training process of trained U-net model includes: in one of the embodiments,
The MRI sample image for the preset quantity that will acquire is input in preset initial U-Net model, obtains initial U-
After Net model divides MRI sample image, each marrow grey matter training image of output；
Calculate the deviation of the marrow grey matter standard picture of marrow grey matter training image and corresponding MRI sample image；
The weight coefficient of initial U-Net model is updated using back-propagation algorithm based on deviation, until updated initial
U-Net model meets preset condition, and using the initial U-Net model of final updated as trained U-Net model.
In one of the embodiments, " seed point is selected in each MRI image to be measured according to location information respectively, and
The segmented image comprising seed point is extracted respectively from each MRI image to be measured using region growing algorithm ", comprising:
Initial pixel point identical with location information is chosen from MRI image to be measured, and using initial pixel point as seed
Point；
Using seed point as growth starting point, and gradually the pixel of the default growing strategy of adjacent and satisfaction is merged into and is worked as
In preceding image-region；
It extracts and completes combined present image area as segmented image.
On the other hand, the embodiment of the invention also provides a kind of marrow ash quality detecting devices, comprising:
Initial pictures obtain module, for each MRI image of vertebra to be sequentially inputted to preparatory trained U-Net model
Middle progress marrow grey matter segmentation obtains at least one initial marrow grey matter image；
Module is chosen, for belonging in the corresponding initial marrow grey matter image of the MRI image in the middle part of vertebra from what is selected
It determines a pixel, and acquires the location information of pixel；
Region growing module for selecting seed point respectively in each MRI image to be measured according to location information, and utilizes
Region growing algorithm extracts the segmented image comprising seed point respectively from each MRI image to be measured, wherein each MRI figure to be measured
As being the corresponding MRI image of each initial marrow grey matter image；
Marrow grey matter determining module, for area and the corresponding initial marrow grey matter image when the segmented image extracted
When the error of area is less than or equal to default error, determine that there are marrow grey matter regions for the corresponding MRI image to be measured of segmented image.
In one of the embodiments, further include:
Area determining module, for there are the corresponding initial marrow grey matter figures of the MRI image of marrow grey matter regions by each
The area of picture is as the first area, and using the area of corresponding segmented image as second area；
Computing module, for using the first area with the average value of corresponding second area as there are marrow grey matter regions
The final marrow grey matter regions area of MRI image.
In one of the embodiments, further include:
Thickness data obtaining module, for obtaining, each there are the corresponding thickness of the MRI image of marrow grey matter regions；
Layer volume calculation module, for calculating the product of each thickness with corresponding final marrow grey matter regions area；
Grey matter volume calculation module, for calculating the summation of each product, and summation is overall as the marrow grey matter of vertebra
Product.
On the other hand, the embodiment of the invention also provides a kind of marrow grey matter detection device, including memory and processor,
Memory is stored with computer program, and processor realizes marrow grey matter detection method when executing computer program.
On the other hand, the embodiment of the invention also provides a kind of computer readable storage medium, it is stored with computer program,
Marrow grey matter detection method is realized when computer program is executed by processor.
A technical solution in above-mentioned technical proposal is had the following advantages and beneficial effects:
Marrow grey matter detection method, device, equipment and computer readable storage medium of the invention can be based on training
U-Net model be partitioned into the initial marrow grey matter image in the MRI image data of vertebra.Since there are bones in the middle part of vertebra
Therefore marrow facilitates in initial marrow grey matter image to select corresponding image in the middle part of backbone.In order to improve detection vertebrae
It the precision of marrow grey matter regions and prevents from judging by accident, a pixel is determined in the initial marrow grey matter image selected and acquires its position
Confidence ceases, and the foundation in this, as selected seed point is simultaneously proposed in each MRI image to be measured corresponding based on algorithm of region growing
Segmented image.By the comparison of the segmented image and initial marrow grey matter image, if area error is being preset in error range,
Determine that there are marrow grey matter regions for corresponding MRI image to be measured.Various embodiments of the present invention are increased by U-Net models coupling region
Algorithm, and according to the marrow grey matter characteristic distributions of vertebra, double check is carried out to marrow grey matter regions, can quickly navigated to
While marrow grey matter regions, the generation of erroneous detection erroneous judgement problem can also be prevented.Present invention is particularly suitable for vertebra marrow ash
The detection in matter region, further improves the speed of processing detection data, and automation and intelligence degree are high.
Detailed description of the invention
Fig. 1 is the flow diagram that one embodiment of marrow grey matter detection method of the present invention provides；
Fig. 2 is that the process that the marrow grey matter volume that one embodiment of marrow grey matter detection method of the present invention provides calculates is shown
It is intended to；
Fig. 3 is the flow diagram that the region that one embodiment of marrow grey matter detection method of the present invention provides increases；
Fig. 4 is the structural schematic diagram of marrow ash quality detecting device of the present invention；
Fig. 5 is the structural schematic diagram of marrow grey matter detection device of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, referring to the drawings in conjunction with the embodiments, right
The present invention is further described.
Referring to Fig. 1, in one embodiment, the present invention provides a kind of marrow grey matter detection methods, comprising the following steps:
Step S110: each MRI image of vertebra is sequentially inputted to carry out marrow ash in preparatory trained U-Net model
Matter segmentation, and obtain at least one initial marrow grey matter image.
U-Net model is the convolutional neural networks of Biomedical Image segmentation, is adapted to do the segmentation of medical image.This
Preparatory trained U-Net model is the MRI (Magnetic that initial U-Net model passes through many cases vertebra in inventive embodiments
Resonance Imaging, magnetic resonance imaging) sample image training obtains, the marrow that can be partitioned into MRI sample image
Grey matter regions.Therefore, each MRI image of vertebra is input in trained U-Net model, trained U-Net model energy
Enough be partitioned into think include marrow grey matter initial marrow grey matter image.Wherein, the vertebra MRI image of every detected personnel
It include multipage.
The marrow grey matter detection method of the embodiment of the present invention, can be based on preparatory trained U-Net model automatic identification
Marrow grey matter in each MRI image of vertebra out, so that Fast Segmentation goes out initial marrow grey matter image.
Step S120: it is corresponded to determining one in initial marrow grey matter image from the MRI image in the middle part of vertebra that belongs to selected
Pixel, and acquire the location information of pixel.
Since marrow grey matter is present in the middle part of vertebra, and when carrying out MRI scan to vertebra, be from vertebra it is upper down
Scanning or under up scan, until scan through entire vertebra position.Therefore, the initial marrow obtained based on above-mentioned steps S110
Grey matter image can therefrom select that corresponding MRI image in the middle part of vertebra, the reason is that the MRI image in the middle part of vertebra is determining
There is marrow grey matter.Wherein, orderly number can be marked on every page of MRI image to represent figure sequence, the sequence of serial number is that basis is swept
Retouch vertebra sequence and determine.Therefore, the corresponding initial marrow grey matter figure of MRI image in the middle part of vertebra can be selected according to serial number
Picture, for example, that one page corresponding among serial number is the MRI image in the middle part of vertebra.Such as, the vertebra for scanning a tester is total
The MRI image of page totally 200, then or so page 100 intermediate, such as page 100 is MRI image corresponding in the middle part of vertebra, thus further
To the corresponding initial marrow grey matter image of the MRI image.Wherein, the mode of selection can be chosen to be artificial, or calculate
Machine is chosen automatically.
Specifically, the pixel determined in the corresponding initial marrow grey matter image of MRI image in the middle part of vertebra is bone
Any point in marrow grey matter regions can be more preferably the center of mass point of marrow grey matter regions.MRI image pair in the middle part of vertebra
A pixel is determined in the initial marrow grey matter image answered, and acquires the location information of the pixel, wherein the location information can
Pixel coordinate is thought, the reason is that marrow grey matter is distributed in the middle part of vertebra, institute based on marrow grey matter in the distribution situation of vertebra
It is that determination has marrow grey matter, and then the pixel position is to indicate that marrow grey matter is deposited with the MRI image in the middle part of vertebra
One of position.
And trained U-Net model carries out marrow grey matter regions to the vertebra MRI image of detected personnel page by page in advance
Segmentation, but can exist erroneous judgement the case where because not every MRI image can all have marrow grey matter.Thus train
U-Net model there are the feelings that can will be accidentally partitioned into without the MRI image of marrow grey matter an initial marrow grey matter image originally
Condition.Whether above-mentioned location information can determine the reference frame of marrow grey matter as each MRI image to be measured of detection.As a result,
It can be using the location information of the pixel as examination criteria, in the corresponding MRI image of each initial marrow grey matter image further
Marrow grey matter regions are detected, help to ensure that the accuracy of marrow ash quality detection.
Step S130: seed point is selected according to location information respectively in each MRI image to be measured, and is increased using region
Algorithm extracts the segmented image comprising seed point respectively from each MRI image to be measured, wherein each MRI image to be measured is each
The initially corresponding MRI image of marrow grey matter image.
Region growing algorithm can specify seed point as growth starting point in the picture, then will lead around the growth starting point
The pixel in domain and the growth starting point compare, and the point with like attribute is combined continuation to outgrowth, until not having
Until thering is the pixel of the condition of satisfaction to be included.Specifically, according to the location information acquired in above-mentioned steps S120, each
The position being closer to quickly is navigated in MRI image to be measured, for example, in the pixel coordinate of the position and above-mentioned location information
Pixel coordinate distance be less than or equal to pre-determined distance, more preferably, the position navigated to is identical as above-mentioned location information, i.e.,
The pixel coordinate of the two is identical.Thus using the corresponding pixel in the position navigated to as seed point, to be based on this kind
Son point carries out region and increases to obtain the segmented image comprising seed point.
The marrow grey matter detection method of the embodiment of the present invention, the location information that can be determined based on above-mentioned steps S120 are effective
Ground navigates to corresponding position in each MRI image to be measured, it is possible thereby to which accurately finding out out marrow grey matter should existing position
To select seed point, selection is with clearly defined objective, reduces the redundancy of program.Area can be utilized in each MRI image to be measured simultaneously
Domain growth algorithm merges the pixel with seed point like attribute, and then can further detect whether MRI image to be measured determines
There are marrow grey matters, to avoid erroneous judgement caused by single detection mode.
Step S140: when the area and the error of the area of corresponding initial marrow grey matter image of the segmented image of extraction are small
When default error, determine that there are marrow grey matter regions for the corresponding MRI image to be measured of segmented image.
If there is the pixel with seed point like attribute on MRI image to be measured, can be closed by region growing algorithm
And segmented image that is providing certain area area and including seed point.Further, if there are marrow ashes for MRI image to be measured
Matter image, then its obtained segmented image area, the area error of initial marrow grey matter image corresponding with MRI image to be measured compared with
It is small；If marrow grey matter image, segmented image initial marrow corresponding with MRI image to be measured is not present in MRI image to be measured originally
The error of grey matter image is larger, can give up the MRI image to be measured of the no marrow grey matter regions of the determination.More preferably, it presets and misses
Difference may be set to 20%.
Marrow grey matter detection method of the invention can be partitioned into the MRI image of vertebra based on trained U-Net model
Initial marrow grey matter image in data.Due in the middle part of vertebra there are marrow, facilitate in initial marrow grey matter image
In select corresponding image in the middle part of backbone.In order to improve the precision of detection vertebra marrow grey matter regions and prevent from judging by accident,
A pixel is determined in the initial marrow grey matter image selected and acquires its location information, in this, as the foundation of selected seed point
And corresponding segmented image is proposed in each MRI image to be measured based on algorithm of region growing.Pass through the segmented image and initial bone
The comparison of marrow grey matter image, if area error is in default error range, it is determined that there are marrow ashes for corresponding MRI image to be measured
Matter region.The embodiment of the present invention is distributed spy by U-Net models coupling region growing algorithm, and according to the marrow grey matter of vertebra
Point, carrying out double check to marrow grey matter regions can also prevent from missing while can quickly navigate to marrow grey matter regions
Examine the generation of erroneous judgement problem.Present invention is particularly suitable for the detections to vertebra marrow grey matter regions, further improve processing inspection
The speed of measured data, automation and intelligence degree are high.
In a specific embodiment, " when area and the corresponding initial marrow grey matter image of the segmented image of extraction
The error of area when being less than or equal to default error, determine that there are marrow grey matter areas for the corresponding MRI image to be measured of segmented image
After domain ", further includes:
Step S2: there are the area of the corresponding initial marrow grey matter image of the MRI image of marrow grey matter regions works by each
For the first area, and using the area of corresponding segmented image as second area.
Step S4: using the first area with the average value of corresponding second area as there are the MRI of marrow grey matter regions figures
The final marrow grey matter regions area of picture.
The marrow grey matter detection method of the embodiment of the present invention is detecting that MRI image to be measured, can be comprehensive there are after marrow grey matter
It closes and determines that there are the first area of the corresponding initial marrow grey matter image of the MRI image of marrow grey matter regions and corresponding points
Cut the mean value of the second area of image.Thereby, it is possible to further improve the detection accuracy of marrow grey matter, erroneous judgement erroneous detection is prevented.
Referring to fig. 2, in a specific embodiment, " using the first area and the average value of corresponding second area as
There are the final marrow grey matter regions areas of the MRI image of marrow grey matter regions " after, further includes:
Step S210: acquisition is each, and there are the corresponding thickness of the MRI image of marrow grey matter regions.
The corresponding thickness of MRI image refers to the thickness for the layer that is excited when carrying out MRI scan.
Step S220: the product of each thickness with corresponding final marrow grey matter regions area is calculated.
Corresponding this layer of marrow grey matter of the layer that is excited based on this step, where can get final marrow grey matter regions area
Volume helps to improve the resolving power calculated marrow grey matter total volume.
Step S230: the summation of each product is calculated, and using summation as the marrow grey matter total volume of vertebra.
The marrow grey matter detection method of the embodiment of the present invention, by using each, there are the MRI images of marrow grey matter regions
Corresponding thickness, first calculate vertebra marrow grey matter be respectively excited layer volume to obtaining the total volume of marrow grey matter.This hair
Bright embodiment automation and intelligence degree are high, effectively can accurately calculate marrow grey matter volume and resolving power is high, especially suitable
For the diagnosis to vertebra.
In a specific embodiment, the training process of trained U-Net model includes:
Step S8: the MRI sample image for the preset quantity that will acquire is input in preset initial U-Net model, is obtained
After dividing MRI sample image to initial U-Net model, each marrow grey matter training image of output.
MRI sample image is the MRI image for including marrow grey matter, for training U-Net model, can be identified simultaneously
The marrow grey matter being partitioned into MRI sample image can be more preferably the MRI image comprising vertebra marrow grey matter.
Step S10: calculating marrow grey matter training image is inclined with the marrow grey matter standard picture of corresponding MRI sample image
Difference.
Image is presented by marrow grey matter in the marrow grey matter standard picture of MRI sample image in MRI image, for judging
Marrow grey matter training image whether be marrow grey matter MRI image.Deviation for indicate marrow grey matter training image with it is corresponding
The similarity of marrow grey matter standard picture, the deviation of the two are smaller, then it represents that the two is more similar, and U-Net model divides MRI image
In marrow grey matter precision it is higher, closer to training objective.
Step S12: updating the weight coefficient of initial U-Net model based on deviation using back-propagation algorithm, until updating
Initial U-Net model afterwards meets preset condition, and using the initial U-Net model of final updated as trained U-Net mould
Type.
Initial U-Net model includes input layer, hidden layer and output layer, and initial U-Net model is in not up to default item
Before part, by the marrow grey matter training image of output and the deviation of corresponding marrow grey matter standard picture, backpropagation is utilized
Algorithm from output layer, hidden layer to input layer sequential update and adjust the weight coefficient of each node, until initial U-Net mould
Type reaches preset condition, and then using the initial U-Net model for meeting preset condition of final updated as trained U-Net mould
Type.Wherein, preset condition is that deviation is less than or equal to predetermined deviation, i.e., updated initial U-Net model divides marrow grey matter
The accuracy rate of image reaches preset value, and more preferably, accuracy rate can be set as 99%.
The marrow grey matter detection method of the embodiment of the present invention, can by way of artificial intelligence training can high-precision from
The U-Net model of marrow grey matter image is partitioned into MRI image, intelligence degree height is, it can be achieved that automatic detection marrow grey matter
Diagnosis efficiency can be improved in region.
Ginseng Fig. 3 " selects kind in each MRI image to be measured according to location information in a specific embodiment respectively
It is sub-, and the segmented image comprising seed point is extracted respectively from each MRI image to be measured using region growing algorithm ", comprising:
Step S310: choosing identical with location information initial pixel point from MRI image to be measured, and by initial pixel point
As seed point.
Step S320: using seed point as growth starting point, gradually by the pixel of the default growing strategy of adjacent and satisfaction
It is merged into present image area.
Seed point is the growth starting point that region increases, while being also initial present image area, is carrying out region growth
When adjacent pixel is traversed centered on the seed point.For example, if the gray value of neighbor pixel and the gray value of seed point
When difference is less than or equal to default gray scale difference value, then neighbor pixel and seed point are merged as new present image area
Domain then traverses the upper one adjacent pixel of pixel for being merged into present image area, if the pixel traversed with it is upper
When the difference of the gray value of one combined pixel is less than or equal to default gray scale difference value, then the pixel traversed is merged
Into present image area, and continue to traverse the adjacent pixel of the upper one new pixel for being merged into present image area, with this
The mode of iteration gradually merges adjacent, like attribute pixel into present image area, until without meeting default life
Until the pixel of long rule.Wherein, preset growing strategy be not merged into the pixel of present image area with it is adjacent and
The gray value difference for belonging to the pixel of present image area is less than or equal to default gray scale difference value.
Step S330: it extracts and completes combined present image area as segmented image.
The marrow grey matter detection method of the embodiment of the present invention, can be by the attributive character of seed point, in MRI image to be measured
Connected region with similar features is identified, meets the marrow grey matter regions of condition convenient for extracting, it is high-efficient and facilitate
Complete image is obtained, guarantees the continuity of image.The embodiment of the present invention especially improves the essence of the detection to vertebra marrow grey matter
Degree.
Referring to fig. 4, in one embodiment, the present invention also provides a kind of marrow ash quality detecting devices, comprising:
Initial pictures obtain module 410, for each MRI image of vertebra to be sequentially inputted to preparatory trained U-Net
The segmentation of marrow grey matter is carried out in model, obtains at least one initial marrow grey matter image.
Module 420 is chosen, for from the corresponding initial marrow grey matter image of the MRI image belonged in the middle part of vertebra selected
Middle one pixel of determination, and acquire the location information of pixel.
Region growing module 430, for selecting seed point respectively in each MRI image to be measured according to location information, and
The segmented image comprising seed point is extracted respectively from each MRI image to be measured using region growing algorithm, wherein each to be measured
MRI image is the corresponding MRI image of each initial marrow grey matter image.
Marrow grey matter determining module 440, for area and the corresponding initial marrow grey matter figure when the segmented image extracted
When the error of the area of picture is less than or equal to default error, determine that there are marrow grey matters for the corresponding MRI image to be measured of segmented image
Region.
The marrow ash quality detecting device of the embodiment of the present invention can be partitioned into vertebra based on trained U-Net model
Initial marrow grey matter image in MRI image data.Due in the middle part of vertebra there are marrow, facilitate in initial marrow
Corresponding image in the middle part of backbone is selected in grey matter image.In order to improve the precision of detection vertebra marrow grey matter regions and prevent
Erroneous judgement determines a pixel in the initial marrow grey matter image selected and acquires its location information, in this, as selected seed
The foundation of point simultaneously proposes corresponding segmented image in each MRI image to be measured based on algorithm of region growing.Pass through the segmented image
With the comparison of initial marrow grey matter image, if area error is in default error range, it is determined that corresponding MRI image to be measured is deposited
In marrow grey matter regions.The embodiment of the present invention passes through U-Net models coupling region growing algorithm, and according to the marrow grey matter of vertebra
Characteristic distributions carry out double check to marrow grey matter regions, can be with while can quickly navigate to marrow grey matter regions
Prevent the generation of erroneous detection erroneous judgement problem.Present invention is particularly suitable for the detections to vertebra marrow grey matter regions, further improve
The speed of processing detection data, automation and intelligence degree are high.
In a specific embodiment, further includes:
Area determining module, for there are the corresponding initial marrow grey matter figures of the MRI image of marrow grey matter regions by each
The area of picture is as the first area, and using the area of corresponding segmented image as second area.
Computing module, for using the first area with the average value of corresponding second area as there are marrow grey matter regions
The final marrow grey matter regions area of MRI image.
The marrow ash quality detecting device of the embodiment of the present invention is detecting that MRI image to be measured, can be comprehensive there are after marrow grey matter
It closes and determines that there are the first area of the corresponding initial marrow grey matter image of the MRI image of marrow grey matter regions and corresponding points
Cut the mean value of the second area of image.Thereby, it is possible to further improve the detection accuracy of marrow grey matter, erroneous judgement erroneous detection is prevented.
In a specific embodiment, further includes:
Thickness data obtaining module, for obtaining, each there are the corresponding thickness of the MRI image of marrow grey matter regions.
Layer volume calculation module, for calculating the product of each thickness with corresponding final marrow grey matter regions area.
Grey matter volume calculation module, for calculating the summation of each product, and summation is overall as the marrow grey matter of vertebra
Product.
The marrow ash quality detecting device of the embodiment of the present invention, by using each, there are the MRI images of marrow grey matter regions
Corresponding thickness calculates the volume of each layer of vertebra marrow grey matter first to obtain the total volume of marrow grey matter.The present invention is implemented
Example automation and intelligence degree are high, effectively can accurately calculate marrow grey matter volume and resolving power is high, are particularly suitable for pair
The diagnosis of vertebra.
In a specific embodiment, further includes:
Training image obtains module, and the MRI sample image of the preset quantity for will acquire is input to preset initial
In U-Net model, after obtaining initial U-Net model segmentation MRI sample image, each marrow grey matter training image of output.
Deviation computing module, for calculating marrow grey matter mark of the marrow grey matter training image with corresponding MRI sample image
The deviation of quasi- image.
Study module, for the weight coefficient of initial U-Net model to be updated using back-propagation algorithm based on deviation, until
Updated initial U-Net model meets preset condition, and using the initial U-Net model of final updated as trained U-
Net model.
The marrow ash quality detecting device of the embodiment of the present invention, can by way of artificial intelligence training can high-precision from
The U-Net model of marrow grey matter image is partitioned into MRI image, intelligence degree height is, it can be achieved that automatic detection marrow grey matter
Diagnosis efficiency can be improved in region.
In a specific embodiment, growth module in region includes:
Seed point selection unit, for choosing initial pixel point identical with location information from MRI image to be measured, and will
Initial pixel point is as seed point.
Combining unit, for using seed point as growth starting point, adjacent and satisfaction gradually to be preset to the picture of growing strategy
Vegetarian refreshments is merged into present image area.
Extraction unit completes combined present image area as segmented image for extracting.
The marrow ash quality detecting device of the embodiment of the present invention, can be by the attributive character of seed point, in MRI image to be measured
Connected region with similar features is identified, meets the marrow grey matter regions of condition convenient for extracting, it is high-efficient and facilitate
Complete image is obtained, guarantees the continuity of image.The embodiment of the present invention especially improves the essence of the detection to vertebra marrow grey matter
Degree.
Referring to Fig. 5, in one embodiment, a kind of marrow grey matter detection device, including memory and processor are also provided,
Memory is stored with computer program, and processor realizes marrow grey matter detection method when executing computer program.
The marrow grey matter detection device can be terminal, and internal structure chart can be as shown in Figure 5.The marrow ash quality detection
Equipment includes processor, memory, network interface, display screen and the input unit connected by system bus.Wherein, the marrow
The processor of grey matter detection device is for providing calculating and control ability.The memory of the marrow grey matter detection device includes non-easy
The property lost storage medium, built-in storage.The non-volatile memory medium is stored with operating system and computer program.The built-in storage
Operation for operating system and computer program in non-volatile memory medium provides environment.The marrow grey matter detection device
Network interface is used to communicate with external terminal by network connection.To realize one kind when the computer program is executed by processor
Marrow grey matter detection method.The display screen of the marrow grey matter detection device can be liquid crystal display or electric ink is shown
Screen, the input unit of the marrow grey matter detection device can be the touch layer covered on display screen, be also possible to marrow ash quality inspection
Key, trace ball or the Trackpad being arranged on measurement equipment shell can also be external keyboard, Trackpad or mouse etc..
In one embodiment, a kind of computer readable storage medium is additionally provided, computer program is stored thereon with, it should
Any one marrow grey matter detection method in as above each embodiment is realized when program is executed by processor.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
To any reference of memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
In several embodiments provided herein, it should be understood that disclosed device and method can also pass through
Other modes are realized.The apparatus embodiments described above are merely exemplary, for example, flow chart and structure in attached drawing
Figure shows the system frame in the cards of the device of multiple embodiments according to the present invention, method and computer program product
Structure, function and operation.In this regard, each box in flowchart or block diagram can represent a module, section or code
A part, a part of the module, section or code includes one or more for implementing the specified logical function
Executable instruction.It should also be noted that function marked in the box can also be to be different from the implementation as replacement
The sequence marked in attached drawing occurs.For example, two continuous boxes can actually be basically executed in parallel, they are sometimes
It can execute in the opposite order, this depends on the function involved.It is also noted that in structure chart and/or flow chart
The combination of each box and the box in structure chart and/or flow chart, can function or movement as defined in executing it is dedicated
Hardware based system realize, or can realize using a combination of dedicated hardware and computer instructions.
In addition, each functional module or unit in each embodiment of the present invention can integrate one independence of formation together
Part, be also possible to modules individualism, an independent part can also be integrated to form with two or more modules.
It, can be with if the function is realized and when sold or used as an independent product in the form of software function module
It is stored in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be intelligence
Can mobile phone, personal computer, server or network equipment etc.) execute each embodiment the method for the present invention whole or
Part steps.And storage medium above-mentioned include: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory),
Random access memory (RAM, Random Access Memory), magnetic or disk etc. be various to can store program code
Medium.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent substitution, improvement and etc. done be should be included within the scope of the present invention.
Claims (10)
1. a kind of marrow grey matter detection method, which comprises the following steps:
Each MRI image of vertebra is sequentially inputted to carry out the segmentation of marrow grey matter in preparatory trained U-Net model, is obtained extremely
A few initial marrow grey matter image；
Belong to a determining pixel in the corresponding initial marrow grey matter image of the MRI image in the middle part of the vertebra from what is selected,
And acquire the location information of the pixel；
Select seed point respectively in each MRI image to be measured according to the positional information, and using region growing algorithm from each
The segmented image comprising the seed point is extracted in MRI image to be measured respectively, wherein each MRI image to be measured is each
The corresponding MRI image of the initial marrow grey matter image；
When the area of the segmented image of extraction is less than or equal to the error of the area of corresponding initial marrow grey matter image
When default error, determine that there are marrow grey matter regions for the corresponding MRI image to be measured of the segmented image.
2. marrow grey matter detection method according to claim 1, which is characterized in that described " when the segmentation figure of extraction
When the area of picture is less than or equal to default error with the error of the area of corresponding initial marrow grey matter image, the segmentation is determined
There are marrow grey matter regions for the corresponding MRI image to be measured of image " after, further includes:
There are the areas of the corresponding initial marrow grey matter image of the MRI image of marrow grey matter regions as the first area using each,
And using the area of the corresponding segmented image as second area；
Using first area and the average value of the corresponding second area as described, there are the MRI of marrow grey matter regions figures
The final marrow grey matter regions area of picture.
3. marrow grey matter detection method according to claim 2, which is characterized in that it is described " by first area with it is right
There are the final marrow grey matter regions faces of the MRI image of marrow grey matter regions as described for the average value for the second area answered
After product ", further includes:
Obtain that each there are the corresponding thickness of the MRI image of marrow grey matter regions；
Calculate the product of each thickness with the corresponding final marrow grey matter regions area；
The summation of each product is calculated, and using the summation as the marrow grey matter total volume of the vertebra.
4. marrow grey matter detection method according to claim 1, which is characterized in that the trained U-net model
Training process includes:
The MRI sample image for the preset quantity that will acquire is input in preset initial U-Net model, obtains the initial U-
After Net model divides the MRI sample image, each marrow grey matter training image of output；
Calculate the deviation of the marrow grey matter standard picture of the marrow grey matter training image and the corresponding MRI sample image；
The weight coefficient for updating the initial U-Net model using back-propagation algorithm based on the deviation, until updated
The initial U-Net model meets preset condition, and using the initial U-Net model of final updated as described trained
U-Net model.
5. marrow grey matter detection method according to claim 1, which is characterized in that described " to exist according to the positional information
Seed point is selected in each MRI image to be measured respectively, and extracts packet respectively from each MRI image to be measured using region growing algorithm
Segmented image containing the seed point ", comprising:
Choose identical with location information initial pixel point from the MRI image to be measured, and by the initial pixel point
As seed point；
Using the seed point as growth starting point, and gradually the pixel of the default growing strategy of adjacent and satisfaction is merged into and is worked as
In preceding image-region；
It extracts and completes combined present image area as the segmented image.
6. a kind of marrow ash quality detecting device characterized by comprising
Initial pictures obtain module, for by each MRI image of vertebra be sequentially inputted in preparatory trained U-Net model into
The segmentation of row marrow grey matter, obtains at least one initial marrow grey matter image；
Module is chosen, for belonging in the corresponding initial marrow grey matter image of the MRI image in the middle part of the vertebra from what is selected
It determines a pixel, and acquires the location information of the pixel；
Region growing module for selecting seed point respectively in each MRI image to be measured according to the positional information, and utilizes
Region growing algorithm extracted respectively from each MRI image to be measured include the seed point segmented image, wherein it is each it is described to
Surveying MRI image is the corresponding MRI image of each initial marrow grey matter image；
Marrow grey matter determining module, for area and the corresponding initial marrow grey matter image when the segmented image extracted
When the error of area is less than or equal to default error, determine that there are marrow for the corresponding MRI image to be measured of the segmented image
Grey matter regions.
7. marrow ash quality detecting device according to claim 6, which is characterized in that further include:
Area determining module, for by each there are the corresponding initial marrow grey matter image of the MRI image of marrow grey matter regions
Area is as the first area, and using the area of the corresponding segmented image as second area；
Computing module, for there are marrow ashes using first area and the average value of the corresponding second area as described
The final marrow grey matter regions area of the MRI image in matter region.
8. marrow ash quality detecting device according to claim 7, which is characterized in that further include:
Thickness data obtaining module, for obtaining, each there are the corresponding thickness of the MRI image of marrow grey matter regions；
Layer volume calculation module, for calculating multiplying for each thickness and the corresponding final marrow grey matter regions area
Product；
Grey matter volume calculation module, for calculating the summation of each product, and using the summation as the marrow grey matter of the vertebra
Total volume.
9. a kind of marrow grey matter detection device, which is characterized in that including memory and processor, the memory is stored with calculating
Machine program, the processor realize marrow ash quality inspection described in any one of claims 1 to 5 when executing the computer program
Survey method.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer program, the computer program is located
Reason device realizes marrow grey matter detection method described in any one of claims 1 to 5 when executing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910237817.3A CN109978861B (en) | 2019-03-27 | 2019-03-27 | Polio detection method, apparatus, device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910237817.3A CN109978861B (en) | 2019-03-27 | 2019-03-27 | Polio detection method, apparatus, device and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109978861A true CN109978861A (en) | 2019-07-05 |
| CN109978861B CN109978861B (en) | 2021-03-26 |
Family
ID=67080929
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910237817.3A Active CN109978861B (en) | 2019-03-27 | 2019-03-27 | Polio detection method, apparatus, device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109978861B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110599505A (en) * | 2019-09-17 | 2019-12-20 | 上海微创医疗器械（集团）有限公司 | Organ image segmentation method and device, electronic equipment and storage medium |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102006041619A1 (en) * | 2005-09-07 | 2007-04-19 | Siemens Medical Solutions Usa, Inc. | Systems and methods for computer-assisted detection of spinal curvature using images and angle measurements |
| CN101756710A (en) * | 2008-12-19 | 2010-06-30 | 曹淑兰 | Measuring method of volume of intracranial hematoma |
| CN102727200A (en) * | 2011-03-31 | 2012-10-17 | 深圳迈瑞生物医疗电子股份有限公司 | Method and device for dividing spine centrum and intervertebral disk, and magnetic resonance imaging system |
| CN104346799A (en) * | 2013-08-01 | 2015-02-11 | 上海联影医疗科技有限公司 | Method for extracting spinal marrow in CT (Computed Tomography) image |
| CN104751178A (en) * | 2015-03-31 | 2015-07-01 | 上海理工大学 | Pulmonary nodule detection device and method based on shape template matching and combining classifier |
| CN105719276A (en) * | 2016-01-07 | 2016-06-29 | 于翠妮 | Liver parenchymal segmentation method based on CT images |
| CN107977971A (en) * | 2017-11-09 | 2018-05-01 | 哈尔滨理工大学 | The method of vertebra positioning based on convolutional neural networks |
| CN108040496A (en) * | 2015-06-01 | 2018-05-15 | 尤尼伐控股有限公司 | The computer implemented method of distance of the detection object away from imaging sensor |
| CN108197526A (en) * | 2017-11-23 | 2018-06-22 | 西安艾润物联网技术服务有限责任公司 | Detection method, system and computer readable storage medium |
| CN108537784A (en) * | 2018-03-30 | 2018-09-14 | 四川元匠科技有限公司 | A kind of CT figure pulmonary nodule detection methods based on deep learning |
| CN109146885A (en) * | 2018-08-17 | 2019-01-04 | 深圳蓝胖子机器人有限公司 | Image partition method, equipment and computer readable storage medium |
| CN109409503A (en) * | 2018-09-27 | 2019-03-01 | 深圳市铱硙医疗科技有限公司 | Training method, image conversion method, device, equipment and the medium of neural network |
| CN109447969A (en) * | 2018-10-29 | 2019-03-08 | 北京青燕祥云科技有限公司 | Hepatic space occupying lesion recognition methods, device and realization device |
-
2019
- 2019-03-27 CN CN201910237817.3A patent/CN109978861B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102006041619A1 (en) * | 2005-09-07 | 2007-04-19 | Siemens Medical Solutions Usa, Inc. | Systems and methods for computer-assisted detection of spinal curvature using images and angle measurements |
| CN101756710A (en) * | 2008-12-19 | 2010-06-30 | 曹淑兰 | Measuring method of volume of intracranial hematoma |
| CN102727200A (en) * | 2011-03-31 | 2012-10-17 | 深圳迈瑞生物医疗电子股份有限公司 | Method and device for dividing spine centrum and intervertebral disk, and magnetic resonance imaging system |
| CN104346799A (en) * | 2013-08-01 | 2015-02-11 | 上海联影医疗科技有限公司 | Method for extracting spinal marrow in CT (Computed Tomography) image |
| CN104751178A (en) * | 2015-03-31 | 2015-07-01 | 上海理工大学 | Pulmonary nodule detection device and method based on shape template matching and combining classifier |
| CN108040496A (en) * | 2015-06-01 | 2018-05-15 | 尤尼伐控股有限公司 | The computer implemented method of distance of the detection object away from imaging sensor |
| CN105719276A (en) * | 2016-01-07 | 2016-06-29 | 于翠妮 | Liver parenchymal segmentation method based on CT images |
| CN107977971A (en) * | 2017-11-09 | 2018-05-01 | 哈尔滨理工大学 | The method of vertebra positioning based on convolutional neural networks |
| CN108197526A (en) * | 2017-11-23 | 2018-06-22 | 西安艾润物联网技术服务有限责任公司 | Detection method, system and computer readable storage medium |
| CN108537784A (en) * | 2018-03-30 | 2018-09-14 | 四川元匠科技有限公司 | A kind of CT figure pulmonary nodule detection methods based on deep learning |
| CN109146885A (en) * | 2018-08-17 | 2019-01-04 | 深圳蓝胖子机器人有限公司 | Image partition method, equipment and computer readable storage medium |
| CN109409503A (en) * | 2018-09-27 | 2019-03-01 | 深圳市铱硙医疗科技有限公司 | Training method, image conversion method, device, equipment and the medium of neural network |
| CN109447969A (en) * | 2018-10-29 | 2019-03-08 | 北京青燕祥云科技有限公司 | Hepatic space occupying lesion recognition methods, device and realization device |
Non-Patent Citations (2)
| Title |
|---|
| SZU-HAO HUANG等: "Learning-Based Vertebra Detection and Iterative Normalized-Cut Segmentation for Spinal MRI", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
| 解威: "脊髓瘢痕分割算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110599505A (en) * | 2019-09-17 | 2019-12-20 | 上海微创医疗器械（集团）有限公司 | Organ image segmentation method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109978861B (en) | 2021-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11049014B2 (en) | Learning apparatus, detecting apparatus, learning method, and detecting method | |
| CN110599503B (en) | Detection model training method and device, computer equipment and storage medium | |
| US10997466B2 (en) | Method and system for image segmentation and identification | |
| CN111862044B (en) | Ultrasonic image processing method, ultrasonic image processing device, computer equipment and storage medium | |
| CN111080584B (en) | Quality control method for medical image, computer device and readable storage medium | |
| CN109948684A (en) | Quality detecting method, device and its relevant device of point cloud data mark quality | |
| US5813984A (en) | Forensic skull and soft tissue database and on-line facial reconstruction of victims and age progession portrait rendering of missing children through utilization of advance diagnostic radiologic modalities | |
| US20130070997A1 (en) | Systems, methods, and media for on-line boosting of a classifier | |
| CN111340800A (en) | Image detection method, computer device, and storage medium | |
| CN108492298A (en) | Based on the multispectral image change detecting method for generating confrontation network | |
| CN109978867A (en) | Toy appearance quality determining method and its relevant device | |
| CN110473171A (en) | Brain age detection method, computer equipment and storage medium | |
| CN114387201A (en) | Cytopathic image auxiliary diagnosis system based on deep learning and reinforcement learning | |
| CN111128348B (en) | Medical image processing method, medical image processing device, storage medium and computer equipment | |
| CN110570425B (en) | Pulmonary nodule analysis method and device based on deep reinforcement learning algorithm | |
| CN111091539A (en) | Network model training method, medical image processing method, device, medium and equipment | |
| CN109978861A (en) | Marrow grey matter detection method, device, equipment and computer readable storage medium | |
| CN109978868A (en) | Toy appearance quality determining method and its relevant device | |
| AU2020223750B2 (en) | Method and System for Image Annotation | |
| KR20220094791A (en) | Method and system for data augmentation | |
| CN117745704A (en) | Vertebral region segmentation system for osteoporosis recognition | |
| CN111598144B (en) | Training method and device for image recognition model | |
| CN114372970A (en) | Operation reference information generation method and device | |
| CN113643263A (en) | Identification method and system for upper limb bone positioning and forearm bone fusion deformity | |
| CN111627554A (en) | Fracture image automatic classification system based on deep convolutional neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |