CN109903278A - Ultrasonic tumor of breast form quantization characteristic extracting method based on shape histogram - Google Patents
Ultrasonic tumor of breast form quantization characteristic extracting method based on shape histogram Download PDFInfo
- Publication number
- CN109903278A CN109903278A CN201910138613.4A CN201910138613A CN109903278A CN 109903278 A CN109903278 A CN 109903278A CN 201910138613 A CN201910138613 A CN 201910138613A CN 109903278 A CN109903278 A CN 109903278A
- Authority
- CN
- China
- Prior art keywords
- tumor
- breast
- shape histogram
- shape
- maximum curvature
- 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
Abstract
The ultrasonic tumor of breast form quantization characteristic extracting method based on shape histogram that the invention discloses a kind of, for in computer-aided diagnosis system, the disadvantage of conventional ultrasound Breast Tumors quantization characteristic description inaccuracy, from tumor of breast profile, construct the shape histogram of tumor of breast, again based on shape histogram, design three form quantization characteristics: maximum curvature and maximum curvature peak value and maximum curvature standard deviation and, the good pernicious difference in part to Efficient Characterization ultrasound tumor of breast；And three form quantization characteristics proposed by the present invention are applied in classifier, the good pernicious discrimination of obtained ultrasonic tumor of breast is significantly increased compared to the discrimination of traditional form quantization characteristic, has practical value.The subjective factor that the present invention can effectively reduce doctor influences, and helps to improve the accuracy rate of diagnosis, is applicable to the diagnosis of other benign from malignant tumors, has a good application prospect.
Description
Technical field
The present invention relates to computer medical aided diagnosis technical fields, and in particular to a kind of ultrasound based on shape histogram
Tumor of breast form quantization characteristic extracting method.
Background technique
Instantly a kind of malignancy disease that breast cancer is often suffered from as global women, has high disease incidence and death
Rate, in worldwide every year because the lethal female patient number of breast cancer be as high as it is millions of, and throughout the year come lethality be in
Constantly soaring gesture.Ultrasonic imaging is by the advantages such as its radiationless wound, cheap, portable in real time, it has also become tumor of breast
A kind of important means of early detection and diagnosis.Sonographer by virtue of experience carries out good evil to the ultrasonic tumor of breast image of acquisition
Property differentiate, this is a subjective judgement process, diagnostic result by doctor experienced degree, know-how height and it is tired
The limitation of the conditions such as labor degree, different doctors are to same ultrasound image, and possible diagnostic result is entirely different or even same doctor exists
It is also not identical to the diagnostic result of same ultrasound image under different time, different conditions.Therefore, computer-aided diagnosis is utilized
(Computer Aided Diagnosis, CAD) system carries out good pernicious differentiation with extremely heavy to ultrasonic tumor of breast image
The researching value and application prospect wanted, the subjective factor that it can effectively reduce doctor influence, and improve disease and explain consistency, make
Diagnostic result is more objective, more acurrate.
Ultrasonic Breast Tumors detection based on CAD is to carry out feature to ultrasonic tumor of breast image using computer
The process extracted and classified, main includes pretreatment, tumor target detection and region segmentation, tumoral character extract and benign tumors are disliked
Property classification four steps.Wherein, tumor of breast feature extraction is an important link of ultrasonic Breast Tumors detection,
Whether the tumoral character of description has accurately been largely fixed the accuracy rate of Breast Tumors classification.Currently, to ultrasound
Tumor of breast properties study focuses mostly in terms of form quantization characteristic extraction, for example, Chang etc. uses level set algorithm from ultrasound
It is partitioned into tumor of breast region in picture, is then extracted like circularity, form factor, smoothness, length-width ratio, protrusion degree and area
Six kinds of morphological features of ratio are used for classifier training；Huang etc. extracts 19 morphological features from ROI, defeated after PCA dimensionality reduction
Enter SVM and carries out classification based training；Chen etc. on the basis of traditional tumour morphological feature, have also been proposed borderline tumor fluctuating number,
Five kinds of novel morphological features such as leaflet index, normalization oval perimeters, and utilize multilayer feedforward neural network training classifier.
Currently, common some traditional form quantization characteristics mainly have: edge roughness, regular shape degree, aspect ratio, ellipse
The compact degree of circle, circularity etc..However, the aufbauprinciple of these traditional ultrasonic tumor of breast form quantization characteristics is mostly from complete
Office angle describe tumor shape variation, for example, edge roughness be calculate borderline tumor to tumour mass center radical length it is poor；
Regular shape degree is the coincidence factor for analyzing tumor area and fitted ellipse area；Aspect ratio is the ratio for calculating ellipse long and short shaft；
Oval compact degree is the ratio for calculating oval perimeters and tumor's profiles perimeter；Circularity is then evaluation tumor shape and circular close
Like degree.As it can be seen that they do not account for the local form variation of tumor of breast, the Breast Tumors feature of extraction is not
Accurately.
Therefore, a kind of method for accurately and reliably judging the good evil of tumor of breast how is established, is the master for effectively reducing doctor
Sight factor influences, and helps to improve the key of the accuracy rate of diagnosis, is current urgently to be resolved.
Summary of the invention
The purpose of the present invention is solve the problems, such as traditional tumor of breast form quantization characteristic extraction process.This hair
The bright ultrasonic tumor of breast form quantization characteristic extracting method based on shape histogram, from tumor of breast profile, construction
The shape histogram of tumor of breast, then based on shape histogram, designs three form quantization characteristics: maximum curvature and most
Deep camber peak value and maximum curvature standard deviation and, the good pernicious difference in part to Efficient Characterization ultrasound tumor of breast, Ke Yiyou
The subjective factor that effect reduces doctor influences, and helps to improve the accuracy rate of diagnosis, is applicable to other benign from malignant tumors
Diagnosis, has a good application prospect.
In order to achieve the above object, the technical scheme adopted by the invention is that:
A kind of ultrasonic tumor of breast form quantization characteristic extracting method based on shape histogram, includes the following steps,
Step (A) is partitioned into tumor of breast administrative division map in ultrasonic tumor of breast original image；
Step (B) obtains the tumor of breast edge graph of the tumor of breast administrative division map；
Step (C), the ellipse for obtaining tumor target shape in the tumor of breast edge graph by oval or circle fitting are bent
Line, and obtain the shape histogram of the ultrasound tumor of breast；
Three of the tumor of breast form are depicted according to the shape histogram of obtained ultrasonic tumor of breast in step (D)
Quantization characteristic: maximum curvature and maximum curvature peak value and maximum curvature standard deviation and.
Ultrasonic tumor of breast form quantization characteristic extracting method above-mentioned based on shape histogram, step (C), by ellipse
Circle fitting obtains the elliptic curve of tumor target shape in the tumor of breast edge graph, and obtains the shape of the ultrasound tumor of breast
Histogram includes the following steps,
(C1), according to the pixel coordinate of the borderline tumor profile of tumor of breast edge graph, the tumor target being fitted
The elliptic curve of shape；
(C2), point makees a series of rays centered on the mass center of the elliptic curve, and this series of ray is with transverse
Starting point is rotated in counterclockwise direction around oval mass center, and the value range of the interval angles θ between adjacent ray is 5 °≤
θ≤6°；
(C3), there is intersection point at this series of ray and elliptic curve, tumor of breast edge respectively, calculate every ray with it is ellipse
Difference between the intersection point of circular curve, every ray and the intersection point at tumor of breast edge, obtains the shape of the ultrasound tumor of breast
Histogram.
Ultrasonic tumor of breast form quantization characteristic extracting method above-mentioned based on shape histogram, step (D), according to
The maximum curvature and quantization characteristic of the tumor of breast form is depicted in the shape histogram of the ultrasonic tumor of breast arrived, and process is such as
Under,
(D11), numerical value all in the shape histogram are connected, obtain the curve of shape histogram；
(D12), the curvature for calculating all numerical points in each section in the shape histogram, as shown in formula (1),
Wherein, h 'iIt (j) is first derivative of i-th of j-th of section numerical point on curve in shape histogram, h 'i(j)
For second dervative of i-th of j-th of section numerical point on curve, C in shape histogramijIt is then i-th in shape histogram
The curvature of j-th of section numerical point；
(D13), in the shape histogram all numerical points in each section curvature, take out the maximum curvature in each sectionWherein i=1,2 ..., N, N are the section number of shape histogram, and by the maximum curvature in each section
It is added, obtains the maximum curvature and SMC of shape histogram, as shown in formula (2),
Ultrasonic tumor of breast form quantization characteristic extracting method above-mentioned based on shape histogram, step (D), according to
The maximum curvature peak value and quantization characteristic of the tumor of breast form, mistake is depicted in the shape histogram of the ultrasonic tumor of breast arrived
Journey is as follows,
(D21), according to the maximum curvature in each sectionAs shown in formula (3), shape histogram is obtained most
Deep camber peak value and SMCP,
Wherein,For the maximum curvature in each section,It is every
The peak-peak in a section.
Ultrasonic tumor of breast form quantization characteristic extracting method above-mentioned based on shape histogram, step (D), according to
The maximum curvature standard deviation and quantization characteristic of the tumor of breast form is depicted in the shape histogram of the ultrasonic tumor of breast arrived,
Process is as follows,
(D31), according to formula (4), the peak value standard deviation in each section in shape histogram is calculated,
Wherein,For the average peak in i-th of section in shape histogram, obtaining PSD (i) then is the peak in i-th of section
It is worth standard deviation；
(D32), curvature is weighted by peak value standard deviation, obtains maximum curvature standard deviation and SMCSD, such as formula
(5) shown in,
Wherein,For the maximum curvature in each section.
Ultrasonic tumor of breast form quantization characteristic extracting method above-mentioned based on shape histogram, further includes step (E),
According to three quantization characteristics of the tumor of breast form that step (D) obtains, the good of the ultrasound tumor of breast pernicious is sentenced
Not.
Ultrasonic tumor of breast form quantization characteristic extracting method above-mentioned based on shape histogram, step (E) use
Three quantization characteristics of the tumor of breast form that SVM classifier obtains step (D) carry out the good evil of the ultrasound tumor of breast
Property differentiate.
The beneficial effects of the present invention are: the ultrasonic tumor of breast form quantization characteristic of the invention based on shape histogram mentions
Method is taken, in computer-aided diagnosis system, conventional ultrasound Breast Tumors quantization characteristic description inaccuracy is lacked
Point constructs the shape histogram of tumor of breast from tumor of breast profile, then based on shape histogram, designs three
Form quantization characteristic: maximum curvature and maximum curvature peak value and maximum curvature standard deviation and, to Efficient Characterization ultrasound mammary gland
The good pernicious difference in the part of tumour；And three form quantization characteristics proposed by the present invention are applied in classifier, what is obtained is super
The good pernicious discrimination of sound tumor of breast is significantly increased compared to the discrimination of traditional form quantization characteristic, has practical value；
Meanwhile maximum curvature and maximum curvature peak value and maximum curvature standard deviation and the form for also enriching existing ultrasonic tumor of breast
Quantization characteristic library, the subjective factor that the present invention can effectively reduce doctor influence, and help to improve the accuracy rate of diagnosis, can
Suitable for the diagnosis of other benign from malignant tumors, have a good application prospect.
Detailed description of the invention
Fig. 1 is the process of the ultrasonic tumor of breast form quantization characteristic extracting method of the invention based on shape histogram
Figure；
Fig. 2 is the flow chart of the shape histogram for obtaining the ultrasound tumor of breast of the invention；
Fig. 3 is the image of benign ultrasonic tumor of breast；
Fig. 4 is the image of pernicious ultrasonic tumor of breast；
Fig. 5 is the edge contour figure of benign ultrasonic tumor of breast；
Fig. 6 is the edge contour figure of pernicious ultrasonic tumor of breast；
Fig. 7 is the shape histogram of benign ultrasonic tumor of breast；
Fig. 8 is the shape histogram of pernicious ultrasonic tumor of breast；
Fig. 9 is the curve graph of the shape histogram of benign ultrasonic tumor of breast；
Figure 10 is the curve graph of the shape histogram of pernicious ultrasonic tumor of breast；
Figure 11 is the sample graph of ultrasonic tumor of breast.
Specific embodiment
Below in conjunction with Figure of description, the present invention is further illustrated.
As shown in Figure 1, the ultrasonic tumor of breast form quantization characteristic extracting method of the invention based on shape histogram, packet
Include following steps,
Step (A) is partitioned into tumor of breast administrative division map in ultrasonic tumor of breast original image；
Step (B) obtains the tumor of breast edge graph of the tumor of breast administrative division map；
Step (C), the ellipse for obtaining tumor target shape in the tumor of breast edge graph by oval or circle fitting are bent
Line, and the shape histogram of the ultrasound tumor of breast is obtained, as shown in Fig. 2, include the following steps,
(C1), according to the pixel coordinate of the borderline tumor profile of tumor of breast edge graph, the tumor target being fitted
The elliptic curve of shape；
(C2), point makees a series of rays centered on the mass center of the elliptic curve, and this series of ray is with transverse
Starting point is rotated in counterclockwise direction around oval mass center, and the value range of the interval angles θ between adjacent ray is 5 °≤
θ≤6 ° show that, when θ=5.7 °, effect is best through many experiments；
(C3), there is intersection point at this series of ray and elliptic curve, tumor of breast edge respectively, calculate every ray with it is ellipse
Difference between the intersection point of circular curve, every ray and the intersection point at tumor of breast edge, obtains the shape of the ultrasound tumor of breast
Histogram
Three of the tumor of breast form are depicted according to the shape histogram of obtained ultrasonic tumor of breast in step (D)
Quantization characteristic: maximum curvature and maximum curvature peak value and maximum curvature standard deviation and,
Wherein, the process for describing maximum curvature sum is as follows,
(D11), numerical value all in the shape histogram are connected, obtain the curve of shape histogram；
(D12), the curvature for calculating all numerical points in each section in the shape histogram, as shown in formula (1),
Wherein, h 'iIt (j) is first derivative of i-th of j-th of section numerical point on curve in shape histogram, h 'i(j)
For second dervative of i-th of j-th of section numerical point on curve, C in shape histogramijIt is then i-th in shape histogram
The curvature of j-th of section numerical point；
(D13), in the shape histogram all numerical points in each section curvature, take out the maximum curvature in each sectionWherein i=1,2 ..., N, N are the section number of shape histogram, and by the maximum curvature in each section
It is added, obtains the maximum curvature and SMC of shape histogram, not such as formula (2) institute,
The maximum curvature peak value and quantization characteristic of the tumor of breast form is depicted, process is as follows,
(D21), according to the maximum curvature in each sectionAs shown in formula (3), shape histogram is obtained most
Deep camber peak value and SMCP,
Wherein,For the maximum curvature in each section,It is every
The peak-peak in a section；
The maximum curvature standard deviation and quantization characteristic of the tumor of breast form is depicted, process is as follows,
(D31), according to formula (4), the peak value standard deviation in each section in shape histogram is calculated,
Wherein,For the average peak in i-th of section in shape histogram, obtaining PSD (i) then is the peak in i-th of section
It is worth standard deviation；
(D32), curvature is weighted by peak value standard deviation, obtains maximum curvature standard deviation and SMCSD, such as formula
(5) shown in,
Wherein,For the maximum curvature in each section；
Ultrasonic tumor of breast form quantization characteristic extracting method based on shape histogram of the invention, further includes step
(E), three quantization characteristics of the tumor of breast form obtained according to step (D), to the ultrasound tumor of breast it is good it is pernicious into
Row differentiates, SVM classifier can be used and carry out ultrasound cream to three quantization characteristics of the tumor of breast form that step (D) obtains
The good pernicious differentiation of adenoncus tumor.
Below according to the ultrasonic tumor of breast form quantization characteristic extracting method of the invention based on shape histogram, specifically
One implementation process is described,
(1) ultrasonic tumor of breast image is inputted, as shown in Figure 3, Figure 4, wherein Fig. 3 is benign, and Fig. 4 is pernicious；
(2) tumor target segmentation is carried out to ultrasonic tumor of breast image, borderline tumor profile diagram is obtained, such as Fig. 5 and Fig. 6 institute
Show, wherein Fig. 5 is the edge contour figure of benign breast tumor, and Fig. 6 is the edge contour figure of malignant breast tumor；
(3) according to the pixel coordinate of borderline tumor profile, the elliptic curve for the tumor target shape being fitted；
(4) point makees a series of rays centered on elliptical mass center, and the ray is using transverse as starting point, along counterclockwise
Direction surrounds oval mass center rotation, ray interval angle, θ=5.7 °；
(5) ray and elliptic curve and borderline tumor have intersection point respectively；
(6) difference between the intersection point and ray of emergent ray and elliptic curve and the intersection point of borderline tumor is calculated, is obtained
The shape histogram of tumor of breast, as shown in Figures 7 and 8, wherein Fig. 7 is the shape histogram of benign breast tumor, and Fig. 8 is
The shape histogram of malignant breast tumor.
From in Fig. 7 and Fig. 8 as can be seen that numerical fluctuations situation and shape of tumor change and are in shape histogram --- it is corresponding
, therefore corresponding quantization characteristic is designed with feasibility based on shape histogram.
Traditional form quantization characteristic is the good pernicious difference that tumor of breast is described from global angle, not for this
Foot, the present invention is based on shape histograms, and quantization characteristic is designed as unit of section, and the shape of tumor of breast is described from the angle of part
State variation, thus the good pernicious difference of more acurrate characterization tumor of breast.Here, the positive value section of shape histogram indicates tumour side
Edge is convex on the outside of ellipse, and negative value section indicates that borderline tumor is recessed on the inside of ellipse, wherein maximum curvature and maximum curvature peak value and
And maximum curvature standard deviation and the design details of three form quantization characteristics it is as follows,
(1) maximum curvature and (Sum of Maximum Curvature, SMC)
Numerical value all in shape histogram are connected, the curve graph of shape histogram is obtained, as shown in FIG. 9 and 10,
Wherein, Fig. 9 be benign breast tumor shape histogram junction curve, Figure 10) be malignant breast tumor shape histogram company
Connect curve；
(2) maximum curvature peak value and (Sum of Maximum Curvature and Peak, SMCP)
The curvature representation bending degree of the curve in each section, but the size of peak of curve cannot be embodied, therefore, counting
Calculate maximum curvature and when also need to consider the variation degree of the bent peak number value in each section in shape histogram, the present invention is further
With peak of curve weight curvature, obtain maximum curvature peak value and；
(3) maximum curvature standard deviation and (Sum of Maximum Curvature and Standard Deviation,
SMCSD)
Likewise, the peak value degree of fluctuation in each section also embodies the good pernicious of ultrasonic tumor of breast in shape histogram
Difference, by being weighted with peak value standard deviation to curvature, obtain maximum curvature standard deviation and.
Lower mask body introduction, the ultrasonic tumor of breast form quantization characteristic according to the present invention based on shape histogram extract
Method obtains three form quantization characteristics: maximum curvature and (SMC), maximum curvature peak value and (SMCP), maximum curvature standard deviation
(SMCSD) carries out the implementation process of Benign and malignant mammary tumor analysis,
Experiment test in, data source capability ultrasound tumor of breast data source in diasonograph (VINNO 70, fly according to
Promise Science and Technology Ltd., Suzhou), probe tranmitting frequency is 5MHz~14MHz, acquires 192 pictures altogether, wherein malignant tumour
Picture 71 is opened, and benign tumour picture 121 is opened, and part sample is as shown in figure 11, wherein figure (a-d) be it is benign, figure (e-h) be dislike
Property.All data all obtain subject's written consent, and meet the approval of hospital's human body ethics.
(1) good malignant breast tumor discrimination compares
Using three form quantization characteristics of the invention: maximum curvature and (SMC), maximum curvature peak value and (SMCP), most
Deep camber standard deviation and (SMCSD) and traditional form quantization characteristic: roughness, rule degree, aspect ratio, oval compact degree, circle
Shape degree carries out feature extraction to the ultrasonic tumor of breast image that actual acquisition arrives, and carries out good pernicious differentiation using SVM classifier,
Recognition result is as shown in table 1.Here, 50 malignant tumour pictures and 90 benign tumour pictures are chosen as training sample, are remained
Remaining picture as test sample,
1 present invention of table and traditional form quantization characteristic discrimination in ultrasonic tumor of breast data compare
From table 1 it follows that using three form quantization characteristics of the invention: maximum curvature and (SMC), maximum curvature
When peak value and (SMCP), maximum curvature standard deviation and (SMCSD) carry out ultrasonic Breast Tumors differentiation, it can achieve
82.69% discrimination, the discrimination compared to five traditional form quantization characteristics improve 15.38%.When conventional quantization feature
When combining with three quantization characteristics proposed by the present invention, discrimination is still 82.69%, it is seen then that most yeast proposed by the present invention
Rate and (SMC), maximum curvature peak value and (SMCP), maximum curvature standard deviation and (SMCSD) are poor in description Breast Tumors
Different time has absolute predominance.
(2) good malignant breast tumor form quantization characteristic numerical value compares
Validity in order to better illustrate the present invention calculates separately good malignant breast tumor under various feature operators
Maximum value, minimum value and mean value, as shown in table 2.
The form quantization characteristic numerical value of the good malignant breast tumor of table 2 compares
The good pernicious differentiation distance D of each form quantization characteristic, calculation formula (6) are further calculated based on 2 numerical value of table
It is as follows,
Wherein, MaxbeniMaximum value, Max for benign breast tumormaliFor the maximum value of malignant breast tumor, Minbeni
For the minimum value of benign breast tumor, MinmaliFor the minimum value of malignant breast tumor, μbeniFor the mean value of benign breast tumor,
μmaliFor the mean value of malignant breast tumor.Arranged according to formula (6) and obtain table 3, as can be seen from Table 3, by maximum curvature and
(SMC), good malignant breast tumor that maximum curvature peak value and (SMCP), maximum curvature standard deviation and (SMCSD) are calculated
Differentiate that distance D is significantly greater than traditional characteristic, therefore, further illustrates, three form quantization characteristics proposed by the present invention can be more
The good pernicious difference of accurate description tumor of breast,
The good evil part of 3 form quantization characteristic of table differentiates distance
(3) performance evaluation of the method provided by the present invention
Validity in order to further illustrate the present invention, used here as the performance of 5 index evaluation classifiers, respectively subject to
Exactness (Accu racy), sensitivity (Sensitivity), specificity (Specificity), positive predictive value (Positive
Predictive Value, PPV), negative predictive value (Negative Predictive Value, NPV), be defined as follows:
Wherein, TP is true positives number of cases, and TN is true negative number of cases, and FP is false positive number of cases, and FN is false negative number of cases.This hair
Bright and conventional quantization feature performance parameter is more as shown in table 4,
The performance parameter of 4 present invention of table and conventional method compares
As can be seen from Table 4, the present invention is compared with traditional form quantization characteristic accuracy with higher, sensitivity, special
It is 80.56% that degree, positive prediction rate and negative predictive rate, especially PPV value, which are 87.5%, NPV value, this illustrates that the present invention is examining
With brilliant precision ratio when surveying positive and negative findings, that is, it is detected as patient's certain illness or disease-free of positive or negative
Possibility is maximum.
In conclusion the ultrasonic tumor of breast form quantization characteristic extracting method of the invention based on shape histogram, needle
To in computer-aided diagnosis system, conventional ultrasound Breast Tumors quantization characteristic describes the disadvantage of inaccuracy, from mammary gland
Tumor's profiles set out, and construct the shape histogram of tumor of breast, then based on shape histogram, and it is special to design three form quantizations
Sign: maximum curvature and maximum curvature peak value and maximum curvature standard deviation and, the part to Efficient Characterization ultrasound tumor of breast
Good pernicious difference；And three form quantization characteristics proposed by the present invention are applied in classifier, obtained ultrasonic tumor of breast
Good pernicious discrimination be significantly increased compared to the discrimination of traditional form quantization characteristic, there is practical value；Meanwhile most yeast
Rate and maximum curvature peak value and maximum curvature standard deviation and the form quantization characteristic library for also enriching existing ultrasonic tumor of breast,
The subjective factor that the present invention can effectively reduce doctor influences, and helps to improve the accuracy rate of diagnosis, is applicable to other
The diagnosis of benign from malignant tumors, has a good application prospect.
Basic principles and main features and advantage of the invention have been shown and described above.The technical staff of the industry should
Understand, the present invention is not limited to the above embodiments, and the above embodiments and description only describe originals of the invention
Reason, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these changes and improvements
It all fall within the protetion scope of the claimed invention.The claimed scope of the invention is by appended claims and its equivalent circle
It is fixed.
Claims (7)
1. the ultrasonic tumor of breast form quantization characteristic extracting method based on shape histogram, it is characterised in that: including following step
Suddenly,
Step (A) is partitioned into tumor of breast administrative division map in ultrasonic tumor of breast original image；
Step (B) obtains the tumor of breast edge graph of the tumor of breast administrative division map；
Step (C) obtains the elliptic curve of tumor target shape in the tumor of breast edge graph by oval or circle fitting, and
Obtain the shape histogram of the ultrasound tumor of breast；
Three quantizations of the tumor of breast form are depicted according to the shape histogram of obtained ultrasonic tumor of breast in step (D)
Feature: maximum curvature and maximum curvature peak value and maximum curvature standard deviation and.
2. the ultrasonic tumor of breast form quantization characteristic extracting method according to claim 1 based on shape histogram,
Be characterized in that: step (C) obtains the elliptic curve of tumor target shape in the tumor of breast edge graph by ellipse fitting, and
The shape histogram of the ultrasound tumor of breast is obtained, is included the following steps,
(C1), according to the pixel coordinate of the borderline tumor profile of tumor of breast edge graph, the tumor target shape that is fitted
Elliptic curve；
(C2), point makees a series of rays centered on the mass center of the elliptic curve, and this series of ray is starting with transverse
Point is rotated in counterclockwise direction around oval mass center, the value range of the interval angles θ between adjacent ray be 5 °≤θ≤
6°；
(C3), there is intersection point at this series of ray and elliptic curve, tumor of breast edge respectively, calculate every ray and ellipse is bent
Difference between the intersection point of line, every ray and the intersection point at tumor of breast edge obtains the shape histogram of the ultrasound tumor of breast
Figure.
3. the ultrasonic tumor of breast form quantization characteristic extracting method according to claim 1 based on shape histogram,
Be characterized in that: the tumor of breast form is depicted most according to the shape histogram of obtained ultrasonic tumor of breast in step (D)
Deep camber and quantization characteristic, process is as follows,
(D11), numerical value all in the shape histogram are connected, obtain the curve of shape histogram；
(D12), the curvature for calculating all numerical points in each section in the shape histogram, as shown in formula (1),
Wherein, h 'iIt (j) is first derivative of i-th of j-th of section numerical point on curve in shape histogram, h "iIt (j) is shape
Second dervative of i-th of j-th of section numerical point on curve, C in shape histogramijIt is then i-th of section in shape histogram
The curvature of j-th of numerical point；
(D13), in the shape histogram all numerical points in each section curvature, take out the maximum curvature in each sectionWherein i=1,2 ..., N, N are the section number of shape histogram, and by the maximum curvature in each section
It is added, obtains the maximum curvature and SMC of shape histogram, as shown in formula (2),
4. the ultrasonic tumor of breast form quantization characteristic extracting method according to claim 3 based on shape histogram,
Be characterized in that: the tumor of breast form is depicted most according to the shape histogram of obtained ultrasonic tumor of breast in step (D)
Deep camber peak value and quantization characteristic, process is as follows,
(D21), according to the maximum curvature in each sectionAs shown in formula (3), the maximum curvature of shape histogram is obtained
Peak value and SMCP,
Wherein,For the maximum curvature in each section,For each section
Peak-peak.
5. the ultrasonic tumor of breast form quantization characteristic extracting method according to claim 3 based on shape histogram,
Be characterized in that: the tumor of breast form is depicted most according to the shape histogram of obtained ultrasonic tumor of breast in step (D)
Deep camber standard deviation and quantization characteristic, process is as follows,
(D31), according to formula (4), the peak value standard deviation in each section in shape histogram is calculated,
Wherein,For the average peak in i-th of section in shape histogram, obtaining PSD (i) then is the peak value standard in i-th of section
Difference；
(D32), curvature is weighted by peak value standard deviation, obtains maximum curvature standard deviation and SMCSD, such as formula (5) institute
Show,
Wherein,For the maximum curvature in each section.
6. the ultrasonic tumor of breast form quantization characteristic extracting method according to claim 1 based on shape histogram,
It is characterized in that: further including step (E), according to three quantization characteristics of the tumor of breast form that step (D) obtains, to the ultrasound
The good of tumor of breast pernicious is differentiated.
7. the ultrasonic tumor of breast form quantization characteristic extracting method according to claim 6 based on shape histogram,
Be characterized in that: step (E) is carried out using three quantization characteristics of the SVM classifier to the tumor of breast form that step (D) obtains
The good pernicious differentiation of the ultrasound tumor of breast.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910138613.4A CN109903278B (en) | 2019-02-25 | 2019-02-25 | Ultrasonic breast tumor morphological quantitative feature extraction method based on shape histogram |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910138613.4A CN109903278B (en) | 2019-02-25 | 2019-02-25 | Ultrasonic breast tumor morphological quantitative feature extraction method based on shape histogram |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109903278A true CN109903278A (en) | 2019-06-18 |
| CN109903278B CN109903278B (en) | 2020-10-27 |
Family
ID=66945495
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910138613.4A Active CN109903278B (en) | 2019-02-25 | 2019-02-25 | Ultrasonic breast tumor morphological quantitative feature extraction method based on shape histogram |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109903278B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111311578A (en) * | 2020-02-17 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Object classification method and device based on artificial intelligence and medical imaging equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2000065985A2 (en) * | 1999-04-29 | 2000-11-09 | University Of South Florida | Method and system for knowledge guided hyperintensity detection and volumetric measurement |
| CN1518719A (en) * | 2001-04-23 | 2004-08-04 | 西门子共同研究公司 | Method and system for automatically detecting lung nodules from multi-slice high resolution computed tomography (MSHR CT) images |
| CN101517614A (en) * | 2006-09-22 | 2009-08-26 | 皇家飞利浦电子股份有限公司 | Advanced computer-aided diagnosis of lung nodules |
| CN103425986A (en) * | 2013-08-31 | 2013-12-04 | 西安电子科技大学 | Breast lump image feature extraction method based on edge neighborhood weighing |
| CN109242016A (en) * | 2018-08-30 | 2019-01-18 | 天津理工大学 | A kind of similitude judgment method of space curve |
-
2019
- 2019-02-25 CN CN201910138613.4A patent/CN109903278B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2000065985A2 (en) * | 1999-04-29 | 2000-11-09 | University Of South Florida | Method and system for knowledge guided hyperintensity detection and volumetric measurement |
| CN1518719A (en) * | 2001-04-23 | 2004-08-04 | 西门子共同研究公司 | Method and system for automatically detecting lung nodules from multi-slice high resolution computed tomography (MSHR CT) images |
| CN101517614A (en) * | 2006-09-22 | 2009-08-26 | 皇家飞利浦电子股份有限公司 | Advanced computer-aided diagnosis of lung nodules |
| CN103425986A (en) * | 2013-08-31 | 2013-12-04 | 西安电子科技大学 | Breast lump image feature extraction method based on edge neighborhood weighing |
| CN109242016A (en) * | 2018-08-30 | 2019-01-18 | 天津理工大学 | A kind of similitude judgment method of space curve |
Non-Patent Citations (1)
| Title |
|---|
| 于丽主编: "《高等数学 上》", 31 July 2015, 哈尔滨工业大学出版社 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111311578A (en) * | 2020-02-17 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Object classification method and device based on artificial intelligence and medical imaging equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109903278B (en) | 2020-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pang et al. | A computer-aided diagnosis system for dynamic contrast-enhanced MR images based on level set segmentation and ReliefF feature selection. | |
| CN106971147A (en) | A kind of traditional Chinese medical science facial diagnosis system and facial diagnosis method split based on human face region | |
| Sewak et al. | SVM approach to breast cancer classification | |
| CN101714153A (en) | Visual perception based interactive mammography image searth method | |
| Zhang et al. | Automatic image quality assessment and measurement of fetal head in two-dimensional ultrasound image | |
| CN108765427A (en) | A kind of prostate image partition method | |
| Alsubaie et al. | A bottom-up approach for tumour differentiation in whole slide images of lung adenocarcinoma | |
| Zhao et al. | Automatic thyroid ultrasound image classification using feature fusion network | |
| Langer et al. | Computer-aided diagnostics in digital pathology: automated evaluation of early-phase pancreatic cancer in mice | |
| Jha et al. | Nucleus and cytoplasm–based segmentation and actor-critic neural network for acute lymphocytic leukaemia detection in single cell blood smear images | |
| Wei et al. | Automatic classification of benign and malignant breast tumors in ultrasound image with texture and morphological features | |
| CN103455821B (en) | Image analysis apparatus and method based on BI-RADS | |
| CN109065150A (en) | A kind of ultrasonic tumor of breast stage division based on multi-feature extraction and Linear SVM | |
| Ben-Cohen et al. | Automated method for detection and segmentation of liver metastatic lesions in follow-up CT examinations | |
| Feng et al. | Robust point correspondence matching and similarity measuring for 3D models by relative angle-context distributions | |
| Zhang et al. | Fully multi-target segmentation for breast ultrasound image based on fully convolutional network | |
| CN109903278A (en) | Ultrasonic tumor of breast form quantization characteristic extracting method based on shape histogram | |
| Sengodan et al. | Early detection and classification of malignant lung nodules from CT images: an optimal ensemble learning | |
| Marcomini et al. | Evaluation of a computer-aided diagnosis system in the classification of lesions in breast strain elastography imaging | |
| Zhang | Classification and diagnosis of thyroid carcinoma using reinforcement residual network with visual attention mechanisms in ultrasound images | |
| Wu et al. | An Effective Machine-Learning Based Feature Extraction/RecognitionModel for Fetal Heart Defect Detection from 2D Ultrasonic Imageries. | |
| Hsieh et al. | Stage classification in chronic kidney disease by ultrasound image | |
| Lu et al. | Automatic scoring of erythema and scaling severity in psoriasis diagnosis | |
| Wang et al. | Automatic recognition of parathyroid nodules in ultrasound images based on fused prior pathological knowledge features | |
| Li et al. | A dual attention-guided 3D convolution network for automatic segmentation of prostate and tumor |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201230 Address after: Room 108, 7 Yingcui Road, Jiangning Development Zone, Nanjing, Jiangsu 210000 Patentee after: Nanjing tianzhixin Technology Co.,Ltd. Address before: 211167 No.1, Hongjing Avenue, Jiangning Science Park, Jiangning District, Nanjing City, Jiangsu Province Patentee before: NANJING INSTITUTE OF TECHNOLOGY |
|
| TR01 | Transfer of patent right |