CN109009110A - Axillary lymphatic metastasis forecasting system based on MRI image - Google Patents
Axillary lymphatic metastasis forecasting system based on MRI image Download PDFInfo
- Publication number
- CN109009110A CN109009110A CN201810666312.4A CN201810666312A CN109009110A CN 109009110 A CN109009110 A CN 109009110A CN 201810666312 A CN201810666312 A CN 201810666312A CN 109009110 A CN109009110 A CN 109009110A
- Authority
- CN
- China
- Prior art keywords
- module
- image
- lump
- region
- feature
- 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.)
- Pending
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves
- A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7271—Specific aspects of physiological measurement analysis
- A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
Abstract
The present invention provides a kind of axillary lymphatic metastasis forecasting system based on MRI image, is related to computer-aided diagnosis technical field.The system includes: input module reception user's input to Diagnosis of Breast DCE-MR image sequence；Region of interesting extraction module extracts the area-of-interest in mammary gland DCE-MR image sequence；Lump segmentation module splits the lump in area-of-interest；Point visualization model carries out visualization display to every image of segmentation and extracts lesion edge；Characteristic extracting module is according to lump information extraction associated eigenvalue and is transmitted to Feature Dimension Reduction module；Feature Dimension Reduction module carries out Feature Dimension Reduction to the feature set after extraction；Each lump characteristic value is inputted classifier by classification diagnosis module, is carried out computer automatic sorting identification, is determined whether lymph node shifts；Output module shows branch prediction result and transition probability.The accurate segmentation of breast lesion may be implemented in the present invention, effectively assists the Accurate Diagnosis of mammary gland axillary lymphatic metastasis.
Description
Technical field
The present invention relates to computer-aided diagnosis technical field more particularly to a kind of axillary glands based on MRI image
Branch prediction system.
Background technique
Breast cancer is the most common malignant tumour of women in world wide, and causes female cancer dead after lung cancer
The second largest reason.When lump diameter is greater than 2cm, breast cancer cell is possible to shift.Breast cancer cell often first turns
Ipsilateral axillary gland is moved on to, the lymph node numbers shifted are more, and the survival rate of patient is lower.Patient with breast cancer's armpit
Lymph node whether occur shifting to breast cancer by stages, treatment and prognosis is significant and the weight of postoperation radiotherapy and chemotherapy
Want one of reference index.
The goldstandard whether current diagnosis lymph node shifts still is pathological examination, frequently with axillary lymph node dissection with before
Whistle lymph node biopsy.Axillary lymph node dissection can provide most complete, accurate information, but its wound is big, and complication is more, than
Such as lymph node oedema, armpit pain, ipsilateral shoulder mobility are limited, and it is feminine gender that part, which cleans result, not to lymph node
The patient for occurring shifting brings unnecessary over-treatment.Therefore, Accurate Prediction Status of axillary lymph node touches axillary gland
Negative patient with breast cancer is examined, can avoid unnecessary axillary lymph node dissection, reduces pain and expense.Sentinel lymph node
Biopsy is compared with axillary lymph node dissection few intercurrent disease, but it is incomplete to the assessment of Status of axillary lymph node, increases operating time,
And if lymph node Skip metastasis occurs or early diagnosis mistake occurs and false negative case is caused to occur, and patient will carry out two
Secondary operation.If the state of axillary gland can be assessed accurately in the preoperative, sentinel lymph node biopsy can be skipped and direct
It carries out or without axillary lymph node dissection.Therefore, clinical that a noninvasive accurate forecasting system assessment is needed to be diagnosed as in the recent period
Patient's Status of axillary lymph node of breast cancer, to determine which patient needs directly to carry out axillary lymph node dissection, and armpit drenches
Fawn on the over-treatment that the patient not shifted avoids axillary lymph node dissection.
Currently, commonly the method for assessment axillary lymph knot includes that clinical physical examination, mammary X-ray photography, CT, ultrasound and magnetic are total
Vibration imaging.Clinical physical examination is usually the inspection that patient receives at first, and due to being influenced by rib cage and chest muscle, physical examination can only be touched
And the shallow lymph node compared with table, and physical examination is difficult to discriminate between inflammatory and metastatic enlarged lymph node.Therefore physical examination is for armpit
The diagnostic value of lymph node is extremely limited.Mammary X-ray, which is photographed, is widely used in the screening of mammary gland disease.Sometimes in mammary X-ray
Above it can be observed that part axillary gland, the increase of lymph node asymmetry volume, increase in density, lymph door disappearance etc. prompt
Transfer may.But the lymph node of mammary X-ray photography display is usually imperfect, and structure is discontinuous, so that mammary X-ray photography is for turning
There are larger deficiencies for the diagnosis of shifting property lymph node, and the lymph node for that can show, the specificity of mammary X-ray photography diagnosis is also
It is lower.CT has higher spatial resolution, and image taking speed is fast, and scanning range is also relatively complete, can clearly show that the shape of lymph node
State feature, enhancing scanning can provide partial function information.But CT has radiation, is not suitable as the method checked repeatedly.It is super
Acoustic inspection is simple and easy, can carry out the imaging of arbitrary tangent, and leaching that can be pernicious for strong suspicion to lymph node in real time
Fawn on carry out Puncture Biopsy under Ultrasound Guidance.Due to having the structures such as pectoralis major, musculus pectoralis minor around the area II lymph node, there is lung in the area III
The interference of sharp gas, leading to display and diagnosis of the ultrasound to this two groups of lymph nodes, there are certain difficulties.Magnetic resonance imaging is a kind of
Non-invasive, radiationless inspection method, soft tissue resolving power is high, has been widely used in the diagnosis of mammary gland disease, diagnoses
Value also has been widely recognized, but due to the limitation of its FOV, some higher lymph nodes in position are difficult to show, and cream
The spatial resolution that gland Coil on Axillary Lymph Nodes is shown is not good enough.
Summary of the invention
It is a kind of based on MRI image the technical problem to be solved by the present invention is in view of the above shortcomings of the prior art, provide
Using computer-aided diagnosis technology the accurate segmentation of breast lesion may be implemented, effectively in axillary lymphatic metastasis forecasting system
Assist the Accurate Diagnosis of mammary gland axillary lymphatic metastasis.
In order to solve the above technical problems, the technical solution used in the present invention is:
A kind of axillary lymphatic metastasis forecasting system based on MRI image, including input module, region of interesting extraction mould
Block, lump segmentation module, visualization model, characteristic extracting module, Feature Dimension Reduction module, classification diagnosis module and output module；
Input module, for receiving the mammary gland DCE-MR image sequence to be diagnosed of user's input；Region of interesting extraction
Module, the area-of-interest in mammary gland DCE-MR image sequence for extracting input, and the area-of-interest figure that will be extracted
As sequence sends lump segmentation module to；Lump divides module, the region of interest for extracting region of interesting extraction module
Lump in domain is split, and obtains segmented image sequence, and send visualization model and characteristic extracting module to；Visualize mould
Block, for carrying out visualization display to every image in segmented image sequence and extracting the side of lesion in the image after visualization
Edge, to clearly see lesion present position；Characteristic extracting module extracts correlation for lump information based on the received
Characteristic value, and send Feature Dimension Reduction module to；Feature Dimension Reduction module, for carrying out Feature Dimension Reduction, shape to the feature set after extraction
The character subset of Cheng Xin；Classification diagnosis module, for the new feature set obtained after dimensionality reduction to be input in classifier, to lump
Characteristic value carries out computer automatic sorting identification, determines whether lymph node shifts, and send result to output module；Export mould
Block is for showing prediction result and transition probability that mammary cancer armpit lymph gland shifts.
The region of interesting extraction module is realized especially by following steps:
Step 2.1: for the mammary gland DCE-MR image sequence to be diagnosed of input, showing lump middle layer figure first
Picture；
Step 2.2: in display image, interception includes the rectangular area of lump, and lump is made to be located at rectangle frame as far as possible
Center；
Step 2.3: one variable of setting automatically saves the top left co-ordinate of rectangle frame and the length and width dimensions of rectangle；
Step 2.4: by first layer lump image, successively showing each tomographic image；
Step 2.5: the variable saved in step 2.3 being acted in each tomographic image, every layer interested is automatically extracted
Region obtains image of interest sequence.
The lump segmentation module is realized using 3D region growth segmentation lump especially by following steps:
Step 3.1: the gray level of every image in image of interest sequence being compressed, compression image sequence is obtained
Column；
Step 3.2: the middle layer in display compression image sequence；
Step 3.3: being selected in intermediate tomographic image a little as seed point, if its pixel is (x0, y0)；
Step 3.4: in compression image sequence, centered on (x0, y0), consider 6 neighborhood territory pixels (x, y) of (x0, y0),
If neighborhood territory pixel (x, y) and sub-pixel (x0, y0) gray value phase absolute value of the difference are less than some thresholding T, then the pixel packet
Containing into the region where sub-pixel, forming region；
Step 3.5: centered on region, continuing with 6 neighborhood territory pixels, the point for meeting growth conditions is incorporated to, again shape
The region of Cheng Xin；The growth conditions is when the segmented good region all pixels point of the sum of the grayscale values of pixel to be added
When the absolute value of the difference of average gray value is less than or equal to thresholding T, which is added to the region being divided into；
Step 3.6: step 3.5 is repeated, until region growing terminates when meeting this growth conditions there is no pixel；
Step 3.7: segmentation result being expanded, cavity is supplemented, deletes three Morphological scale-spaces of small area object, is obtained
Ideal segmentation result.
The visualization model is realized especially by following steps:
Step 4.1: the edge of lump is detected in lump cutting procedure, and its pixel value is denoted as 1, rest of pixels value note
It is 0；
Step 4.2: visualization display lump segmentation as a result, and marginal information is shown on the image, so as to clearly
See lesion form.
The characteristic extracting module extracts four kinds of features, including morphological feature, gray level co-occurrence matrixes feature, intensity histogram
Figure feature and Tamura feature.
The Feature Dimension Reduction module is especially by LASSO, in pact of the sum of the absolute value of regression coefficient less than a constant
Under the conditions of beam, residual sum of squares (RSS) is minimized, generates certain regression coefficients exactly equal to 0, achieve the purpose that dimensionality reduction.
The classification diagnosis module uses support vector machines (i.e. SVM) classifier, realizes especially by following steps:
Step 7.1: the new data set formed after dimensionality reduction being divided into 5 parts at random using 5 folding cross-validation methods；
Step 7.2: every time will wherein 4 parts be used as training set, be left 1 part and be used as test set, carry out 5 experiments, guarantee every
Part data all did test set；
Step 7.3: the classification results for providing each sample in test set are 1 or 0, while also exporting in the model SVM points
The accuracy of class device；The accuracy result that 5 times are tested saves, and seeks mean value as final classification results.
The beneficial effects of adopting the technical scheme are that the armpit leaching provided by the invention based on MRI image
Branch prediction system is fawned on, mammary gland DCE-MR image is based on, inputs a mammary gland DCE-MR image sequence to be diagnosed to system first
Column obtain a series of associated eigenvalues by region of interesting extraction, lump segmentation and feature extraction and dimensionality reduction, then will be special
Value indicative is input to classifier and whether shifts carry out classification diagnosis to lymph node.In short, the present invention is based on mammary gland DCE- by a kind of
The axillary lymphatic metastasis forecasting system based on MRI image of MR image automatically detects breast lump, thus to armpit
The transfer condition of nest lymph node judges, and improves radiologist to a certain extent and examines lymph no transfer
Disconnected accuracy and efficiency, selection (such as whether needing to carry out axillary lymph node dissection etc.) to mammary cancer surgery mode and subsequent
The selection of clinical treatment is significant.
Detailed description of the invention
Fig. 1 is the structural block diagram of the axillary lymphatic metastasis forecasting system provided in an embodiment of the present invention based on MRI image；
Fig. 2 is the flow chart of the axillary lymphatic metastasis prediction technique provided in an embodiment of the present invention based on MRI image；
Fig. 3 is mammary gland DCE-MR image sequence schematic diagram to be processed provided in an embodiment of the present invention；
Fig. 4 is the region of interest of middle layer image interception in image sequence in Fig. 3；
Fig. 5 is the region of interest of image sequence interception in Fig. 3；
Fig. 6 is the relevant treatment image schematic diagram of intermediate tomographic image in image sequence in Fig. 3；Wherein, a is in image sequence
The original sequence figure of intermediate tomographic image；B is the area-of-interest schematic diagram of intermediate tomographic image in image sequence；C is image
The lump of intermediate tomographic image divides schematic diagram in sequence；D is the Morphological scale-space schematic diagram of intermediate tomographic image in image sequence；e
For the segmentation contour schematic diagram of tomographic image intermediate in image sequence；F is the final segmentation result of intermediate tomographic image in image sequence
Schematic diagram.
Specific embodiment
With reference to the accompanying drawings and examples, specific embodiments of the present invention will be described in further detail.Implement below
Example is not intended to limit the scope of the invention for illustrating the present invention.
Accurate Prediction Status of axillary lymph node is significant to the patient with breast cancer of axillary gland feminine gender, can avoid not
Necessary axillary lymph node dissection reduces pain and expense.As shown in Figure 1, provided in this embodiment based on MRI image
Axillary lymphatic metastasis forecasting system, including input module, region of interesting extraction module, lump segmentation module, visualization mould
Block, characteristic extracting module, Feature Dimension Reduction module, classification diagnosis module and output module.Base is carried out using the system of the present embodiment
In the method flow that the axillary lymphatic metastasis of MRI image is predicted as shown in Fig. 2, mainly comprising the steps that
(1) a mammary gland DCE-MR image sequence to be diagnosed is inputted；
(2) to the DCE-MR image preprocessing of input, area-of-interest is extracted；
(3) lump region is partitioned into area-of-interest；
(4) visualization display segmentation result；
(5) characteristic value for having divided lump is extracted；
(6) Feature Dimension Reduction is carried out to the feature set after extraction, forms new character subset；
(7) new character subset is input to classifier, carries out the modeling and analysis of axillary lymphatic metastasis.
Each module concrete analysis is as follows in the system of the present embodiment.
Input module is used to receive the mammary gland DCE-MR image sequence to be diagnosed of user's input.In the present embodiment, input
Mammary gland DCE-MR image sequence to be diagnosed it is as shown in Figure 3.
Region of interesting extraction module is used to extract the area-of-interest in the mammary gland DCE-MR image sequence of input, and will
The area-of-interest image sequence extracted sends lump segmentation module to.
The lump segmentation of mammary gland DCE-MR image sequence is the base of mammary cancer armpit lymph gland transfer computer-aided diagnosis
The key link of plinth and image detection and diagnosis.Entire image is analyzed, not there is only bulk redundancy information, is also held
It is easily introduced mistake.In order to improve the speed and accuracy of processing, needs to deal with objects from entire image and narrow down to several cells
Domain, i.e. area-of-interest.Region of interesting extraction module is realized especially by following steps:
Step 2.1: for the mammary gland DCE-MR image sequence to be diagnosed of input, showing lump middle layer figure first
Picture；
Step 2.2: in display image, interception includes the rectangular area of lump, and lump is made to be located at rectangle frame as far as possible
Center；In the present embodiment, as shown in figure 4, to the image sequence middle layer image interception region of interest in Fig. 3；
Step 2.3: one variable of setting automatically saves the top left co-ordinate of rectangle frame and the length and width dimensions of rectangle；
Step 2.4: by first layer lump image, successively showing each tomographic image；
Step 2.5: the variable saved in step 2.3 being acted in each tomographic image, every layer interested is automatically extracted
Region obtains image of interest sequence.As shown in figure 5, for the region of interest intercepted in the present embodiment to image sequence in Fig. 3.
Lump segmentation module is used to split the lump in area-of-interest that region of interesting extraction module is extracted,
Segmented image sequence is obtained, and sends visualization model and characteristic extracting module to.
In view of lump has high brightness, approximately round and pair that generates due to the gray difference with surrounding tissue
Than features such as degree, the present embodiment is using 3D region growth segmentation lump.
The basic thought of region growing is that the pixel set with similitude is got up to constitute region, first to each needs
Starting point of the sub-pixel as growth is found out in the region of segmentation, then will be had in sub-pixel surrounding neighbors with seed identical
Or the pixel (being determined according to pre-determined growth or similarity criterion) of similar quality is merged into the region where sub-pixel
In, and new pixel continues to grow around as seed, and the pixel until not meeting condition again may include coming in, and one
Region, which is just grown, to be formed.There are 3 crucial problems during this: growing the determination of seed point, the condition of region growing, area
The condition that domain growth stops.
Region growing realizes lump segmentation, and specific step is as follows:
Step 3.1: the gray level of every image in image of interest sequence being compressed, compression image sequence is obtained
Column；
Step 3.2: the middle layer in display compression image sequence；In the present embodiment, intermediate tomographic image in image sequence
Shown in original sequence figure such as Fig. 6 (a), in image sequence shown in area-of-interest such as Fig. 6 (b) of intermediate tomographic image；
Step 3.3: being selected in intermediate tomographic image a little as seed point, if its pixel is (x0, y0)；
Step 3.4: in compression image sequence, centered on (x0, y0), consider 6 neighborhood territory pixels (x, y) of (x0, y0),
If neighborhood territory pixel (x, y) and sub-pixel (x0, y0) gray value phase absolute value of the difference are less than some thresholding T, then the pixel packet
Containing into the region where sub-pixel, forming region；
Step 3.5: centered on region, continuing with 6 neighborhood territory pixels, the point for meeting growth conditions is incorporated to, again shape
The region of Cheng Xin；The growth conditions is when the segmented good region all pixels point of the sum of the grayscale values of pixel to be added
When the absolute value of the difference of average gray value is less than or equal to thresholding T, which is added to the region being divided into；
Step 3.6: step 3.5 is repeated, until region growing terminates when meeting this growth conditions there is no pixel；This reality
It applies in example, in image sequence shown in lump segmentation schematic diagram such as Fig. 6 (c) of intermediate tomographic image；
Step 3.7: segmentation result being expanded, cavity is supplemented, deletes three Morphological scale-spaces of small area object, is obtained
Ideal segmentation result；In the present embodiment, in image sequence shown in Morphological scale-space result such as Fig. 6 (d) of intermediate tomographic image.
Visualization model, after carrying out visualization display to every image in segmented image sequence and extracting visualization
Image in the edge of lesion realized to clearly see lesion present position especially by following steps:
Step 4.1: the edge of lump is detected in lump cutting procedure, and its pixel value is denoted as 1, rest of pixels value note
It is 0；
Step 4.2: visualization display lump segmentation as a result, and marginal information is shown on the image, so as to clearly
See lesion form.
In the present embodiment, in image sequence the segmentation contour of intermediate tomographic image and final segmentation result respectively such as Fig. 6 (e) and
Shown in 6 (f).
Characteristic extracting module extracts associated eigenvalue, and send Feature Dimension Reduction to for lump information based on the received
Module.
Feature extraction is the important component and image information of mammary nodes transfer computer-aided diagnosis system
The important information of quantization and expression basis.Correctly extracting and select effective characteristics of image is to judge that lymph no shifts
Important link, and improve the important prerequisite of mammary cancer armpit lymph gland transfer precision degree.
The characteristic value of selective extraction should follow following feature:
1. identifiability: the characteristic value of different class objects has notable difference；
2. reliability: homogeneous object applies similar characteristic value；
3. independence: should not have strong correlation between characteristic value；
Morphological feature is the variable for describing lump shape, edge and geometrical characteristic.
Gray level co-occurrence matrixes are a kind of texture characteristic extracting methods, mainly by four energy, entropy, contrast, correlation spies
The mean value and variance for levying parameter are constituted.Energy is the quadratic sum of gray level co-occurrence matrixes element value, reflects image grayscale distribution
Uniformity coefficient and texture fineness degree.Contrast reflects the clarity of image and the degree of the texture rill depth.Texture rill is got over
Deep, contrast is bigger, and visual effect is more clear；Conversely, contrast is small, then rill is shallow, and effect is fuzzy.Relativity measurement space
Gray level co-occurrence matrixes element be expert at or column direction on similarity degree, therefore, correlation size reflects local gray level in image
Correlation.Entropy illustrates the non-uniform degree or complexity of texture in image.
Grey level histogram is the function about grey level distribution, is the statistics to grey level distribution in image, it indicates figure
The number of pixel as in certain gray level reflects the frequency that certain gray scale occurs in image.
Psychologic research based on the mankind to the visual perception of texture, Tamura et al. propose the table of textural characteristics
It reaches.Six components of Tamura textural characteristics correspond to Psychological Angle on textural characteristics six attribute, be respectively roughness,
Contrast, direction degree, line picture degree, regularity and rough degree.Wherein, first three component is even more important for image retrieval.
In the present embodiment, four kinds of features, including morphological feature, gray level co-occurrence matrixes feature, intensity histogram are extracted altogether
Figure feature and Tamura feature.Each lump extracts 33 dimensional features altogether, including area, perimeter, like circularity, length-width ratio, rectangle, matter
The heart, degree of eccentricity etc. 16 ties up morphological feature；Energy, entropy, 8 Wei Huidugongshengjuzhente of contrast, the mean value of correlation and standard deviation
Sign；The 3 dimension textural characteristics extracted by Tamura, i.e. contrast, roughness and direction degree；6 dimension grey level histogram features:
Value, mean square deviation, smoothness, third moment, consistency, entropy.
Feature Dimension Reduction module is used to carry out Feature Dimension Reduction to the feature set after extraction, forms new character subset.
Feature Dimension Reduction plays an important role in terms of training classifier, reduction computation complexity, raising.
Minimum absolute retract and selection operator (LASSO) method are suitable for the recurrence of high dimensional data, select for concentrating from initial data
The most useful predicted characteristics.The basic thought of LASSO is in constraint item of the sum of the absolute value of regression coefficient less than a constant
Under part, residual sum of squares (RSS) is minimized, so as to generate certain regression coefficients exactly equal to 0, reduces redundancy and uncorrelated
Feature, and then achieve the purpose that dimensionality reduction.In the present embodiment, it is concentrated using LASSO from primitive character and selects 14 optimal dimensions
Feature set forms new feature set.
In the new feature set input classifier that classification diagnosis module is used to obtain after dimensionality reduction, lump characteristic value is carried out
Computer automatic sorting identification, determines whether lymph node shifts, and send result to output module.
In axillary lymphatic metastasis forecasting system of the breast cancer based on MRI image, it is necessary to lesion segmentation, feature will be passed through
The characteristic parameter obtained after the processing such as extraction and selection, which is input to classifier, which carries out classification analysis, could obtain final lymph node
Branch prediction is as a result, realize complete lesion diagnosis.
In the present embodiment, using support vector machines (SVM) classifier.SVM method be by a Nonlinear Mapping,
Sample space is mapped in a higher-dimension or even infinite dimensional feature space, so that the Nonlinear separability in original sample space
The problem of the problem of being converted into the linear separability in feature space.Briefly, peacekeeping linearisation is exactly risen.SVM needs to instruct
A segmentation hyperplane is practised, which is exactly the decision boundary classified, and being divided on plane both sides is exactly two classes.In addition, in order to
Over-fitting is reduced to a certain extent, and obtains effective information from limited data as much as possible, uses 5 times of cross validations
To determine the robustness of classifier.
Specific step is as follows for the classification shifted to mammary cancer armpit lymph gland:
Step 7.1: the new data set formed after dimensionality reduction being divided into 5 parts at random using 5 folding cross-validation methods；
Step 7.2: every time will wherein 4 parts be used as training set, be left 1 part and be used as test set, carry out 5 experiments, guarantee every
Part data all did test set；
Step 7.3: the classification results for providing each sample in test set are 1 or 0, while also exporting in the model SVM points
The accuracy of class device；The accuracy result that 5 times are tested saves, and seeks mean value as final classification results.
In the present embodiment, the axillary lymphatic metastasis based on mammary gland DCE-MR image finally obtained diagnoses average accuracy
Up to 89.14%.
Output module is for showing prediction result and transition probability that mammary cancer armpit lymph gland shifts.
Axillary lymphatic metastasis based on MRI image provided in this embodiment based on mammary gland DCE-MR image predicts system
System after carrying out region of interesting extraction to breast MRI image sequence early period, using Region growing segmentation method, carries out mastosis
The accurate segmentation of stove, and lump region after segmentation is clearly indicated out, it extracts 35 features such as form, texture, gray scale and uses
In svm classifier and modeling, the diagnosis that can effectively assist mammary cancer armpit lymph gland to shift.The above is of the invention specific
Embodiment, mammary gland DCE-MR image can while detecting breast cancer thoroughly evaluating axillary gland, be preoperative evaluation armpit
Effective inspection method of lymph node status.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify to technical solution documented by previous embodiment, or some or all of the technical features are equal
Replacement；And these are modified or replaceed, model defined by the claims in the present invention that it does not separate the essence of the corresponding technical solution
It encloses.
Claims (7)
1. a kind of axillary lymphatic metastasis forecasting system based on MRI image, it is characterised in that: including input module, interested
Region extraction module, lump segmentation module, visualization model, characteristic extracting module, Feature Dimension Reduction module, classification diagnosis module and
Output module；
Input module, for receiving the mammary gland DCE-MR image sequence to be diagnosed of user's input；Region of interesting extraction module,
The area-of-interest in mammary gland DCE-MR image sequence for extracting input, and the area-of-interest image sequence that will be extracted
Send lump segmentation module to；Lump divides module, in the area-of-interest for extracting region of interesting extraction module
Lump is split, and obtains segmented image sequence, and send visualization model and characteristic extracting module to；Visualization model is used
In in segmented image sequence every image carry out visualization display and extract visualization after image in lesion edge, with
Just lesion present position is clearly seen；Characteristic extracting module extracts correlated characteristic for lump information based on the received
Value, and send Feature Dimension Reduction module to；Feature Dimension Reduction module is formed new for carrying out Feature Dimension Reduction to the feature set after extraction
Character subset；Classification diagnosis module, for inputting the new feature set obtained after dimensionality reduction in classifier, to lump characteristic value
Computer automatic sorting identification is carried out, determines whether lymph node shifts, and send result to output module；Output module is used for
The prediction result of mammary cancer armpit lymph gland transfer and transition probability are shown.
2. the axillary lymphatic metastasis forecasting system according to claim 1 based on MRI image, it is characterised in that: described
Region of interesting extraction module is realized especially by following steps:
Step 2.1: for the mammary gland DCE-MR image sequence to be diagnosed of input, showing tomographic image among lump first；
Step 2.2: in display image, interception includes the rectangular area of lump, and lump is made to be located at the center of rectangle frame as far as possible
Position；
Step 2.3: one variable of setting automatically saves the top left co-ordinate of rectangle frame and the length and width dimensions of rectangle；
Step 2.4: by first layer lump image, successively showing each tomographic image；
Step 2.5: the variable saved in step 2.3 is acted in each tomographic image, every layer of area-of-interest is automatically extracted,
Obtain image of interest sequence.
3. the axillary lymphatic metastasis forecasting system according to claim 1 based on MRI image, it is characterised in that: described
Lump divides module using 3D region growth segmentation lump, realizes especially by following steps:
Step 3.1: the gray level of every image in image of interest sequence being compressed, compression image sequence is obtained；
Step 3.2: the middle layer in display compression image sequence；
Step 3.3: being selected in intermediate tomographic image a little as seed point, if its pixel is (x0, y0)；
Step 3.4: in compression image sequence, centered on (x0, y0), consider 6 neighborhood territory pixels (x, y) of (x0, y0), if
Neighborhood territory pixel (x, y) and sub-pixel (x0, y0) gray value phase absolute value of the difference are less than some thresholding T, then the pixel include into
Region where sub-pixel, forming region；
Step 3.5: centered on region, continuing with 6 neighborhood territory pixels, the point for meeting growth conditions is incorporated to, formed again new
Region；The growth conditions is being averaged when the segmented good region all pixels point of the sum of the grayscale values of pixel to be added
When the absolute value of the difference of gray value is less than or equal to thresholding T, which is added to the region being divided into；
Step 3.6: step 3.5 is repeated, until region growing terminates when meeting this growth conditions there is no pixel；
Step 3.7: segmentation result being expanded, cavity is supplemented, deletes three Morphological scale-spaces of small area object, obtains ideal
Segmentation result.
4. the axillary lymphatic metastasis forecasting system according to claim 1 based on MRI image, it is characterised in that: described
Visualization model is realized especially by following steps:
Step 4.1: detecting the edge of lump in lump cutting procedure, and its pixel value is denoted as 1, rest of pixels value is denoted as 0；
Step 4.2: visualization display lump segmentation as a result, and marginal information is shown on the image, to can be clearly seen that
Lesion form.
5. the axillary lymphatic metastasis forecasting system according to claim 1 based on MRI image, it is characterised in that: described
Characteristic extracting module extract four kinds of features, including morphological feature, gray level co-occurrence matrixes feature, grey level histogram feature and
Tamura feature.
6. the axillary lymphatic metastasis forecasting system according to claim 1 based on MRI image, it is characterised in that: described
Feature Dimension Reduction module makes residual especially by LASSO in the sum of absolute value of regression coefficient less than under the constraint condition of a constant
Poor quadratic sum minimizes, and generates certain regression coefficients exactly equal to 0, achievees the purpose that dimensionality reduction.
7. the axillary lymphatic metastasis forecasting system according to claim 1 based on MRI image, it is characterised in that: described
Classification diagnosis module uses support vector machines (i.e. SVM) classifier, realizes especially by following steps:
Step 7.1: the new data set formed after dimensionality reduction being divided into 5 parts at random using 5 folding cross-validation methods；
Step 7.2: every time will wherein 4 parts be used as training set, be left 1 part be used as test set, carry out 5 times experiment, guarantee every number
According to all doing test set；
Step 7.3: the classification results for providing each sample in test set are 1 or 0, while also exporting SVM classifier in the model
Accuracy；The accuracy result that 5 times are tested saves, and seeks mean value as final classification results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810666312.4A CN109009110A (en) | 2018-06-26 | 2018-06-26 | Axillary lymphatic metastasis forecasting system based on MRI image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810666312.4A CN109009110A (en) | 2018-06-26 | 2018-06-26 | Axillary lymphatic metastasis forecasting system based on MRI image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109009110A true CN109009110A (en) | 2018-12-18 |
Family
ID=64610840
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810666312.4A Pending CN109009110A (en) | 2018-06-26 | 2018-06-26 | Axillary lymphatic metastasis forecasting system based on MRI image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109009110A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109740608A (en) * | 2018-12-28 | 2019-05-10 | 北京工业大学 | A kind of image partition method based on deep learning |
| CN109785310A (en) * | 2019-01-11 | 2019-05-21 | 南京信息工程大学 | A kind of automatic Staging System calculated based on mammary nodes panoramic picture |
| CN110472629A (en) * | 2019-08-14 | 2019-11-19 | 青岛大学附属医院 | A kind of pathological image automatic recognition system and its training method based on deep learning |
| CN110596117A (en) * | 2019-08-15 | 2019-12-20 | 山东科技大学 | Hyperspectral imaging-based rapid nondestructive detection method for apple surface damage |
| CN111178449A (en) * | 2019-12-31 | 2020-05-19 | 浙江大学 | Liver cancer image classification method and device combining computer vision characteristics and imaging omics characteristics |
| CN111353998A (en) * | 2020-05-13 | 2020-06-30 | 温州医科大学附属第一医院 | Tumor diagnosis and treatment prediction model and device based on artificial intelligence |
| CN111354470A (en) * | 2020-02-11 | 2020-06-30 | 中国医学科学院肿瘤医院 | Data analysis method and device based on lymph node multiphase phase |
| WO2020135792A1 (en) * | 2018-12-28 | 2020-07-02 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for generating image metric |
| CN112132917A (en) * | 2020-08-27 | 2020-12-25 | 盐城工学院 | Intelligent diagnosis method for rectal cancer lymph node metastasis |
| CN112233058A (en) * | 2019-07-15 | 2021-01-15 | 上海交通大学医学院附属第九人民医院 | Method for detecting lymph nodes in head and neck CT image |
| CN112545478A (en) * | 2019-09-26 | 2021-03-26 | 北京赛迈特锐医疗科技有限公司 | System and method for automatically segmenting, measuring and locating breast masses on MRI |
| CN112669929A (en) * | 2020-12-30 | 2021-04-16 | 深圳大学 | Crohn's disease infliximab drug effect prediction method and terminal equipment |
| CN112734723A (en) * | 2021-01-08 | 2021-04-30 | 温州医科大学 | Multi-source data-oriented breast tumor image classification prediction method and device |
| CN115272253A (en) * | 2022-08-02 | 2022-11-01 | 中国医学科学院北京协和医院 | Sentinel lymph node image analysis method and device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104794426A (en) * | 2015-01-13 | 2015-07-22 | 宁夏医科大学 | Method for improving prostate tumor MRI (Magnetic Resonance Imaging) image identification rate based on CAD (Computer-Aided Diagnosis) system |
| CN106778046A (en) * | 2017-02-28 | 2017-05-31 | 思派（北京）网络科技有限公司 | The system and method that a kind of feature based value carries out prostate cancer recurrence prediction |
| CN107358267A (en) * | 2017-07-20 | 2017-11-17 | 东北大学 | A kind of breast ultrasound image multivariate classification system and method based on cross-correlation feature |
| CN107392928A (en) * | 2017-07-20 | 2017-11-24 | 东北大学 | Breast cancer diagnosis system and method based on MRF Snake segmentations |
| CN107945181A (en) * | 2017-12-30 | 2018-04-20 | 北京羽医甘蓝信息技术有限公司 | Treating method and apparatus for breast cancer Lymph Node Metastasis pathological image |
-
2018
- 2018-06-26 CN CN201810666312.4A patent/CN109009110A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104794426A (en) * | 2015-01-13 | 2015-07-22 | 宁夏医科大学 | Method for improving prostate tumor MRI (Magnetic Resonance Imaging) image identification rate based on CAD (Computer-Aided Diagnosis) system |
| CN106778046A (en) * | 2017-02-28 | 2017-05-31 | 思派（北京）网络科技有限公司 | The system and method that a kind of feature based value carries out prostate cancer recurrence prediction |
| CN107358267A (en) * | 2017-07-20 | 2017-11-17 | 东北大学 | A kind of breast ultrasound image multivariate classification system and method based on cross-correlation feature |
| CN107392928A (en) * | 2017-07-20 | 2017-11-24 | 东北大学 | Breast cancer diagnosis system and method based on MRF Snake segmentations |
| CN107945181A (en) * | 2017-12-30 | 2018-04-20 | 北京羽医甘蓝信息技术有限公司 | Treating method and apparatus for breast cancer Lymph Node Metastasis pathological image |
Non-Patent Citations (3)
| Title |
|---|
| SCHACHT等: "Using quantitative image analysis to classify axillary lymph nodes onbreast MRI: A new application for the Z 0011 Era", 《EUROPEAN JOURNAL OF RADIOLOGY》 * |
| 季云飞: "磁共振图像中半月板的自动提取和重建问题研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 陈日晖: "面向肺部CAD的三维ROI分割、特征提取与分类方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11436720B2 (en) | 2018-12-28 | 2022-09-06 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for generating image metric |
| CN109740608A (en) * | 2018-12-28 | 2019-05-10 | 北京工业大学 | A kind of image partition method based on deep learning |
| WO2020135792A1 (en) * | 2018-12-28 | 2020-07-02 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for generating image metric |
| CN109785310A (en) * | 2019-01-11 | 2019-05-21 | 南京信息工程大学 | A kind of automatic Staging System calculated based on mammary nodes panoramic picture |
| CN112233058A (en) * | 2019-07-15 | 2021-01-15 | 上海交通大学医学院附属第九人民医院 | Method for detecting lymph nodes in head and neck CT image |
| CN110472629A (en) * | 2019-08-14 | 2019-11-19 | 青岛大学附属医院 | A kind of pathological image automatic recognition system and its training method based on deep learning |
| CN110472629B (en) * | 2019-08-14 | 2022-03-25 | 青岛大学附属医院 | Pathological image automatic identification system based on deep learning and training method thereof |
| CN110596117A (en) * | 2019-08-15 | 2019-12-20 | 山东科技大学 | Hyperspectral imaging-based rapid nondestructive detection method for apple surface damage |
| CN112545478A (en) * | 2019-09-26 | 2021-03-26 | 北京赛迈特锐医疗科技有限公司 | System and method for automatically segmenting, measuring and locating breast masses on MRI |
| CN112545478B (en) * | 2019-09-26 | 2022-07-15 | 北京赛迈特锐医疗科技有限公司 | System and method for automatically segmenting, measuring and locating breast masses on MRI |
| CN111178449B (en) * | 2019-12-31 | 2021-11-05 | 浙江大学 | Liver cancer image classification method combining computer vision characteristics and imaging omics characteristics |
| CN111178449A (en) * | 2019-12-31 | 2020-05-19 | 浙江大学 | Liver cancer image classification method and device combining computer vision characteristics and imaging omics characteristics |
| CN111354470A (en) * | 2020-02-11 | 2020-06-30 | 中国医学科学院肿瘤医院 | Data analysis method and device based on lymph node multiphase phase |
| CN111353998A (en) * | 2020-05-13 | 2020-06-30 | 温州医科大学附属第一医院 | Tumor diagnosis and treatment prediction model and device based on artificial intelligence |
| CN112132917A (en) * | 2020-08-27 | 2020-12-25 | 盐城工学院 | Intelligent diagnosis method for rectal cancer lymph node metastasis |
| CN112669929A (en) * | 2020-12-30 | 2021-04-16 | 深圳大学 | Crohn's disease infliximab drug effect prediction method and terminal equipment |
| CN112669929B (en) * | 2020-12-30 | 2021-08-03 | 深圳大学 | Crohn's disease infliximab drug effect prediction method and terminal equipment |
| CN112734723A (en) * | 2021-01-08 | 2021-04-30 | 温州医科大学 | Multi-source data-oriented breast tumor image classification prediction method and device |
| CN112734723B (en) * | 2021-01-08 | 2023-06-30 | 温州医科大学 | Multi-source data-oriented breast tumor image classification prediction method and device |
| CN115272253A (en) * | 2022-08-02 | 2022-11-01 | 中国医学科学院北京协和医院 | Sentinel lymph node image analysis method and device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109009110A (en) | Axillary lymphatic metastasis forecasting system based on MRI image | |
| Zhou et al. | A radiomics approach with CNN for shear-wave elastography breast tumor classification | |
| Rajan et al. | Brain tumor detection and segmentation by intensity adjustment | |
| Zhang et al. | Photoacoustic image classification and segmentation of breast cancer: a feasibility study | |
| US10593035B2 (en) | Image-based automated measurement model to predict pelvic organ prolapse | |
| Maitra et al. | Identification of abnormal masses in digital mammography images | |
| Wang et al. | Automatic breast tumor detection in ABVS images based on convolutional neural network and superpixel patterns | |
| Qiu et al. | Lung Nodules Detection in CT Images Using Gestalt‐Based Algorithm | |
| Zhang et al. | Comparison of multiple feature extractors on Faster RCNN for breast tumor detection | |
| CN109065150A (en) | A kind of ultrasonic tumor of breast stage division based on multi-feature extraction and Linear SVM | |
| JP2000126181A (en) | Method for extracting and treating tumor | |
| Razzaq et al. | Brain tumor detection from mri images using bag of features and deep neural network | |
| Femil et al. | An Efficient Hybrid Optimization for Skin Cancer Detection Using PNN Classifier. | |
| Bi et al. | Classification of low-grade and high-grade glioma using multiparametric radiomics model | |
| Poernama et al. | Feature extraction and feature selection methods in classification of brain MRI images: a review | |
| Kumar et al. | Survey on recent CAD system for liver disease diagnosis | |
| Chaiyakhan et al. | Feature selection techniques for breast cancer image classification with support vector machine | |
| Choubey et al. | A fully automatic approach to detect brain cancer using random walk algorithm | |
| Youssef et al. | Contour detection of mammogram masses using ChanVese model and B-spline approximation | |
| Bandyopadhyay et al. | Identification of abnormal masses in digital mammography images | |
| Cao et al. | Lesion classification by model-based feature extraction: A differential affine invariant model of soft tissue elasticity | |
| Sreekumari et al. | Breast lesion detection and characterization with 3D features | |
| Dixit et al. | Proposing a Framework to Analyze Breast Cancer in Mammogram Images Using Global Thresholding, Gray Level Co‐Occurrence Matrix, and Convolutional Neural Network (CNN) | |
| CN108960313A (en) | Ultrasonic breast lump hierarchical detection method based on Shearlet feature and level binary tree SVM classifier | |
| CN116415649B (en) | Breast micro cancer analysis method based on multi-mode ultrasonic image self-supervision learning |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181218 |
|
| WD01 | Invention patent application deemed withdrawn after publication |