Ultrasonic mammary gland image classification method and system based on ultrasonic fuzzy entropy analysis

Abstract

The invention discloses an ultrasonic mammary gland image classification method and system based on ultrasonic fuzzy entropy analysis. The method comprises the following steps: S1, imaging by taking a mammary gland as an imaging target; s2, performing pretreatment; s3, selecting a target area image completely containing the whole focus; s4, sliding the window; s5, calculating the fuzzy entropy of the signal in the window, and taking the fuzzy entropy as the pixel value of the center coordinate of the window; s6, calculating fuzzy entropies of all pixel points to obtain an entropy image; s7, performing logarithmic compression on the envelope data to obtain a B ultrasonic image; s8, calculating an average value of entropy values in the data blocks as a focus entropy value; s9, compounding the images; and S10, classifying the ultrasonic mammary gland images. According to the method, the breast tumor is imaged and classified by using the fuzzy entropy, the ultrasonic imaging quality can be improved, so that the breast tumor is diagnosed more accurately, and the benign and malignant breast tumor is preliminarily classified and diagnosed under the condition that puncture and pathological experiments do not need to be performed on a patient.

Description

Technical Field

[0001] The present invention relates to the technical field of ultrasonic detection and ultrasonic imaging, and in particular to an ultrasonic breast image classification method and system based on ultrasonic fuzzy entropy analysis. Background Art

[0002] Cancer is one of the leading causes of death worldwide, with nearly 20 million new cancer cases and 9.7 million cancer deaths in 2022, according to the World Health Organization1 [1] Among them, breast cancer is the most common cancer in women, so the diagnosis and classification of breast diseases are particularly important.

[0003] The morphology of breast parenchyma, which is composed of adipose tissue and fibroglandular tissue, has been shown to be associated with the risk of breast cancer. Medical ultrasound imaging is one of the important means of diagnosing breast diseases. For healthy women, breast tissue includes mammary lobes and mammary duct tissue. Normal breast tissue appears as uniform medium-intensity spots on B-ultrasound images, and the ducts are circular or elliptical low-echo areas of similar size but irregular arrangement. Benign lesions are mostly composed of glandular tissue and local fibrous or calcified components. Local fibrosis or calcification can cause changes in the scattering characteristics of ultrasonic signals, causing scatterers within the tumor to show higher variability in the scattering cross section. [2] Invasive mammary ductal carcinoma (IDC) is the most common malignant lesion. Cancer cells can spread to other parts of the body through the lymphatic system and blood. Malignant tumors may have different structures and calcification patterns. [3-5] Extracting tissue information from ultrasound signals and distinguishing benign and malignant tumors through quantitative ultrasound technology is an important development direction of medical ultrasound imaging.

[0004] Ultrasound entropy imaging is an emerging ultrasound quantitative technology. Information entropy has a high research potential in ultrasound signal analysis. Hughes [6-8] Shannon entropy was first applied to ultrasound, proving that Shannon entropy can reflect the complexity and changes of the internal structure of the scatterer. The disordered characteristics of ultrasound backscattered signals in tissue lesions are the basis of ultrasound entropy imaging. In clinical practice, normal soft tissue parenchyma, such as the liver [9] and breast

[10] , can be regarded as a homogeneous medium with a considerable number of randomly distributed scatterers. The pathological change of soft tissue from normal to abnormal state can be explained by the change of scatterer arrangement from uniform to non-uniform, that is, expressed by entropy

[11] However, during thermal ablation, the temperature of the target area in the tissue increases, causing irreversible coagulative necrosis, in which collagen septa and cell nuclei are the sources of ultrasound backscatter.

[12] Thermocoagulation necrosis leads to protein denaturation, cytoplasm collapse, destruction of nuclei and cell membranes, and the appearance of bubbles. These changes in tissue microstructure can be described by entropy.

[13] Therefore, entropy can be used to characterize tissues, whether they are caused by physiological damage or treatment-induced tissue degeneration, and to identify and classify whether biological tissues are normal.

[0005] Entropy is a parameter that measures the degree of information disorder and has strong adaptability. Hughes et al. [14,15] first used entropy for ultrasound signal analysis and proved that information entropy can quantitatively describe changes in tissue structure. Entropy imaging is an imaging method that uses information entropy as a medium to extract tissue scattering information from ultrasound backscatter signals, thereby enhancing tissue features and improving contrast.

[16] It has been shown that information entropy imaging is superior to Nakagami imaging and B-ultrasound imaging in the assessment of hepatic steatosis. Entropy has also been applied to the diagnosis of prostate cancer.

[17] and lesion detectability

[18] . Tsui et al. [11,19] proposed small window entropy to measure breast tumors, and proved that entropy imaging achieved better detectability than Nakagami imaging. Both traditional Shannon entropy (TSE) imaging and Nakagami imaging can outline the morphology of lesions, and the TSE value of malignant tumors is slightly lower than that of benign lesions, which can be used as one of the bases for lesion classification. In addition, entropy has been applied to the evaluation of ultrasound contrast agent velocity for angiogenesis imaging of prostate cancer

[20] . Behnam et al. [21,22] explored the changes in radiofrequency signals during high-intensity focused ultrasound (HIFU) ablation and demonstrated the detectability of lesion entropy after HIFU treatment. Weighted Shannon entropy (WSE) is an improved entropy that simultaneously considers the probability distribution and amplitude of the backscattered signal [23,24]. Tsui et al.

[25] It was demonstrated that WSE is more sensitive to tissue information in ultrasound signals than TSE. WSE has been widely used in image segmentation

[26] , neural network construction

[27] , tissue characterization

[28] and other fields. Li et al. [29,30] The horizontally normalized Shannon entropy (hNSE) for monitoring microwave ablation thermal damage is proposed. It takes into account the influence of the horizontal axis limit of the probability density histogram on the shape of the data distribution and emphasizes the influence of the distribution of local data relative to the global data on the imaging results. hNSE enhances the contrast of parametric images by increasing the difference in data distribution.

[0006] The above entropies are all based on probability density function (PDF), which has limitations in ultrasonic signal analysis. Chan et al.

[31] It is proposed to use sample entropy imaging to evaluate hepatic steatosis and fibrosis. Compared with PDF-based entropy, sample entropy can more comprehensively describe the interference in different microstructures by monitoring the similarity of signals, thereby better evaluating the uncertainty in ultrasound time series.

[32] Fuzzy entropy is also a parameter to measure the probability of a new pattern in a time series. Fuzzy entropy has similar characteristics to sample entropy and was first proposed for the analysis of muscle electrical signals.

[33] , which is also widely used in the field of image processing. Ismail et al.

[34] Fuzzy entropy is used as a threshold in the level set algorithm to segment medical images. Afterwards, fuzzy entropy is used for ultrasound image enhancement to improve the contrast of liver ultrasound images.

[35] Based on the characteristics of fuzzy entropy for signal analysis, fuzzy entropy imaging has also been applied to the diagnosis and classification of thyroid lesions, improving the diagnostic accuracy of thyroid lesions.

[36] Compared with sample entropy, fuzzy entropy uses a fuzzy function based on an exponential function, which makes the fuzzy entropy change continuously with the parameters. Applying fuzzy entropy to ultrasound parameter imaging and tissue characterization can enhance the robustness of the signal and provide more biological tissue information.

[0007] References:

[0008] [1]BRAY F,LAVERSANNE M,SUNG H,et al.Global cancer statistics 2022:GLOBOCAN estimates of incidence and mortality worldwide for 36cancers in185countries[J].CA:A Cancer Journal for Clinicians,2024,74(3):229-263.

[0009] [2]TSUIP-H,YEH CK,LIAO Y,et al.Ultrasonic nakagamiimaging: a strategy to visualize the scatterer properties of benign and malignant breast tumors[J].Ultrasound in medicine&biology,2009,36:209-217.

[0011] [3]BüSING C M,KEPPLER U,MENGES V.Differences in microcalcification inbreast tumors[J].Virchows Archiv A,1981,393(3):307-313.

[0013] [4]OLSON S L,FAM B W,WINTER P F,et al.Breast calcifications:analysisof imaging properties[J].Radiology,1988,169 2:329-332.

[0014] [5]LIANG S,RANGAYYAN R M,DESAUTELS J E L.Application of shapeanalysis to mammographic calcifications[J].IEEE Transactions on MedicalImaging,1994,13(2):263-274.

[0016] [6]HUGHES M S.Analysis of ultrasonic waveforms using Shannon entropy;proceedings of the IEEE 1992Ultrasonics Symposium Proceedings,Tucson,AZ,USA,1992[C].Tucson,AZ,USA:IEEE,1992.

[0018] [7]HUGHES M.Analysis of digitized waveforms using Shannon entropy[J].The Journal of the Acoustical Society of America,1993,93:892-906.

[0019] [8]HUGHES M S,MCCARTHY J E,MARSH J N,et al.Joint entropy ofcontinuously differentiable ultrasonic waveforms[J].The Journal of theAcoustical Society of America,2013,133(1):283-300.

[0021] [9]YAMADAH,EBARAM,YAMAGUCHI T,et al.Apilot approach for quantitativeassessment of liver fibrosis using ultrasound:preliminary results in 79cases[J].Journal of Hepatology,2006,44(1):68-75.

[0023]

[10] SHANKAR P M,REID J M,ORTEGAH,et al.Use of non-Rayleigh statisticsfor the identification of tumors in ultrasonic B-scans of the breast[J].IEEETransactions on Medical Imaging,1993,12(4):687-692.

[0024]

[11] TSUI P H,CHEN C K,KUO W H,et al.Small-window parametric imagingbased on information entropy for ultrasound tissue characterization[J].Scientific Reports,2017,7:41004.

[0026]

[12] HUNT J W,WORTHINGTON AE,XUAN A,et al.Amodel based upon pseudoregular spacing of cells combined with the randomisation of the nuclei canexplain the significant changes in high-frequency ultrasound signals duringapoptosis[J].Ultrasound in Medicine&Biology,2002,28(2):217-226.

[0028]

[13] LI X,JIAX,SHEN T,et al.Ultrasound entropy imaging for detectionand monitoring of thermal lesion during microwave ablation of liver[J].IEEEJournal of Biomedical and Health Informatics,2022,26(8):4056-4066.

[0029]

[14] HUGHES M S.Analysis of digitized waveforms using Shannon entropy[J].Journal of the Acoustical Society of America,1993,93:892-906.

[0030]

[15] HUGHES M S,MCCARTHY J E,MARSH J N,et al.Joint entropy ofcontinuously differentiable ultrasonic waveforms[J].J Acoust Soc Am,2013,133(1):283-300.

[0031]

[16] ZHOU Z,TAI D-I,WAN Y-L,et al.Hepatic Steatosis Assessment withUltrasound Small-Window Entropy Imaging[J].Ultrasound in Medicine&Biology,2018,44(7):1327-1340.

[0033]

[17] VAN SLOUN R J G,DEMI L,POSTEMAAW,et al.Entropy of ultrasound-contrast-agent velocity fields for angiogenesis imaging in prostate cancer[J].IEEE Transactions on Medical Imaging,2017,36(3):826-837.

[0034]

[18] GHASEMIFARD H,BEHNAM H,TAVAKKOLI J.High-Intensity FocusedUltrasound Lesion Detection Using Adaptive Compressive Sensing Based onEmpirical Mode Decomposition[J].J Med Signals Sens,2019,9(1):24-32.

[0035]

[19] YAN D,LI Q,LIN C W,et al.Clinical Evaluation of Duchenne MuscularDystrophy Severity Using Ultrasound Small-Window Entropy Imaging[J].Entropy(Basel),2020,22(7).

[0036]

[20] VAN SLOUN R J G,DEMI L,POSTEMAAW,et al.Entropy of Ultrasound-Contrast-Agent Velocity Fields for Angiogenesis Imaging in Prostate Cancer[J].IEEE Trans Med Imaging,2017,36(3):826-837.

[0038]

[21] MOBASHERI S,BEHNAM H,RANGRAZ P,et al.Radio frequency ultrasoundtime series signal analysis to evaluate high-intensity focused ultrasoundlesion formation status in tissue[J].Journal of Medical Signals&Sensors,2016,6(2):91-98.

[0040]

[22] MONFARED M M,BEHNAM H,RANGRAZ P,et al.High-intensity focusedultrasound thermal lesion detection using entropy imaging of ultrasound radiofrequency signal time series[J].Journal of Medical Ultrasonics,2018,26(1):24-30.

[0042]

[23] S.Weighted entropy[J].Reports on Mathematical Physics,1971,2(3):165-179.

[0043]

[24] GUIASU S.Grouping data by using the weighted entropy[J].JournalofStatistical Planning and Inference,1986,15:63-69.

[0044]

[25] TSUI P-H.Ultrasound detection of scatterer concentration byweighted entropy[J].Entropy,2015,17:6598-6616.

[0046]

[26] HE Y,LIU T,HOU Y,et al.Afast iris image quality evaluation methodbased on weighted entropy;proceedings of the International Symposium onPhotoelectronic Detection and Imaging2007,Beijing,China,2008[C].Beijing,China:SPIE,2008.

[0048]

[27] PARK E,AHN J,YOO S.Weighted-entropy-based quantization for deepneural networks;

[0049] proceedings of the 2017 IEEE Conference on Computer Vision andPattern Recognition(CVPR),Honolulu,HI,USA,2017[C].New York:IEEE,2017.

[0051]

[28] ZHANG M-M,GAO F,TU J,et al.Classification of benign and malignantbreast masses using entropy from nonlinear ultrasound radiofrequency signal[J].Acta Physica Sinica,2021,70:084302.

[0053]

[29] LI X,JIAX,SHEN T,et al.Ultrasound Entropy Imaging for Detectionand Monitoring of Thermal Lesion During Microwave Ablation of Liver[J].IEEE JBiomed Health Inform,2022,26(8):4056-4066.

[0055]

[30] YANG G,LIU J,YANG B,et al.Multiple ultrasonic parametric imagingfor the detection and monitoring of high-intensity focused ultrasoundablation[J].Ultrasonics,2024,139:107274.

[0056]

[31] CHAN H J,ZHOU Z,FANG J,et al.Ultrasound sample entropy imaging:anew approach for evaluating hepatic steatosis and fibrosis[J].IEEE Journal ofTranslational Engineering in Health and Medicine,2021,9:1800612.

[0057]

[32] RICHMAN J,MOORMAN J.Physiological time-series analysis usingapproximate entropy and sample entropy[J].American Journal of PhysiologyHeart and Circulatory Physiology,2000,278:H2039-2049.

[0058]

[33] CHEN W,WANG Z,XIE H,et al.Characterization of surface EMG signalbased on fuzzy entropy[J].IEEE Transactions on Neural Systems andRehabilitation Engineering,2007,15(2):266-272.

[0059]

[34] MAOLOOD I Y,AL-SALHI Y E A,LU S.Thresholding for medical imagesegmentation for cancer using fuzzy entropy with level set algorithm[J].OpenMedicine 2018,13:374-383.

[0060]

[35] ZHANG X,JIAO Y,ZHANG D,et al.Pixel-based small-window parametricultrasound imaging for liver tumor characterization[J].Biomedical EngineeringLetters,2024,14(5):1011-1021.

[0061]

[36] ZHANG X,HUANG W,LI X,et al.Ultrasound fuzzy entropy imaging basedon time-series signal for tissue characterization[J].Applied Acoustics,2024,224:110158. Summary of the Invention

[0062] The technical problem to be solved by the present invention is to provide an ultrasonic breast image classification method and system based on ultrasonic fuzzy entropy analysis in view of the deficiencies in the above-mentioned prior art. The present invention provides a non-invasive breast tumor benign and malignant classification method, which can improve the quality of ultrasonic imaging so as to more accurately diagnose breast tumors, and make a preliminary diagnosis of the benign and malignant nature of breast tumors without the need for puncture and pathological experiments on patients. Analyzing ultrasonic signals through information entropy does not require ultrasonic signals to conform to a specific statistical distribution model, and can be used as a flexible parameter for signal analysis, and the type of lesion can be preliminarily judged by the size of the fuzzy entropy value.

[0063] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows: In a first aspect of the present invention, a method for classifying ultrasonic breast images based on ultrasonic fuzzy entropy analysis is provided, comprising the following steps:

[0064] S1. Using an ultrasonic signal acquisition module to perform imaging with the breast as an imaging target, and obtaining radio frequency data of the original echo signal;

[0065] S2, performing beamforming and Hilbert transform on the original RF data, demodulating it into envelope data, and then normalizing it to obtain a preprocessed image;

[0066] S3, selecting a target area image that completely includes the entire lesion from the preprocessed image;

[0067] S4, setting the size of the sliding window in the target area image to M×N, and traversing the entire target area image with a step size of one pixel;

[0068] S5, rearrange the two-dimensional signal in the window into a one-dimensional signal, and calculate the fuzzy entropy of the signal in the window as the pixel value of the center coordinate of the window;

[0069] S6, calculating the fuzzy entropy of all pixels to obtain an entropy image;

[0070] S7, logarithmically compressing the envelope data to obtain a B-ultrasound image;

[0071] S8, selecting an entropy value data block of appropriate size in the lesion area in the entropy image, and calculating the average value of the entropy value in the data block as the lesion entropy value, wherein the size of the data block is determined according to the lesion area, and the area is as large as possible but does not exceed the lesion boundary;

[0072] S9, compounding the B-ultrasound image and the entropy image to obtain a composite image, thereby improving image contrast and lesion boundary clarity, which is beneficial to improving diagnostic accuracy;

[0073] S10. Ultrasonic breast image classification is performed by comparing the lesion entropy value with the tumor classification standard value.

[0074] Preferably, the classification method in step S9 is: if the lesion entropy value is less than a standard value, the ultrasound breast image is classified as a malignant tumor image, otherwise it is classified as a benign tumor image.

[0075] Preferably, the method for calculating the fuzzy entropy in step S5 comprises the following steps:

[0076] S5-1. Sequence segmentation: For a time signal sequence X = {x(1), x(2), ..., x(n)} consisting of N data, a vector sequence of dimension m is formed in sequence in 1≤i≤N-m+1; express The mean value of

[0077] S5-2. Define distance: Define The distance between is the Chebyshev distance, where 1≤j≤Nm,j≠i, that is, the maximum absolute value of the numerical difference of each element:

[0078]

[0079] S5-3. Defining Similarity: Introducing Fuzzy Membership n Measurement Similarity, for 1≤j≤Nm,j≠i, introduce fuzzy membership function

[0080]

[0081] Among them, r is the similarity tolerance threshold;

[0082] S5-4. Find the average value: Find the average value φ of the fuzzy membership function for each i m,n,r :

[0083]

[0084] S5-5. Increase the dimension to m+1 to find φ m+1,n,r ,calculate The distance between Right now

[0085]

[0086] For 1≤j≤Nm,j≠i, calculate the similarity of m+1 dimension and mean φ m+1,n,r :

[0087]

[0088] Among them, r is the similarity tolerance threshold, n determines the gradient of the similarity tolerance boundary, and the larger n is, the larger the gradient is;

[0089] S5-6. Calculate the fuzzy entropy F(m,r,n) based on the average fuzzy membership function values ​​of m and m+1 dimensions:

[0090] F(m,r,n)=lnφ m,r,n -lnφ m+1,r,n .

[0091] Preferably, the sliding window size is set to 3 to 10 times the wavelength of the ultrasonic signal, and the number of rows M of the signal in the window is:

[0092]

[0093] Where k is the set wavelength multiple, f s is the signal sampling frequency, λ is the wavelength of the ultrasonic signal emitted by the ultrasonic signal acquisition module, c is the speed of sound, f is c is the ultrasonic frequency.

[0094] Preferably, the number of columns N of the signal in the window is:

[0095]

[0096] Where L is the total number of ultrasonic echo signal columns, and W is the width of the ultrasonic image.

[0097] Preferably, both M and N are rounded integers.

[0098] Preferably, when the signal sequence length is 100 to 1000, m=2 is selected; and when the signal length is 1000 to 30000, m=3 is selected.

[0099] Preferably, the value of r is usually between 0.1 and 0.25 times the standard deviation of the signal.

[0100] Preferably, n is between 0.5 and 3.

[0101] A second aspect of the present invention provides an ultrasound breast image classification system based on ultrasound fuzzy entropy analysis, which uses the above-mentioned method to classify ultrasound breast images. The system comprises:

[0102] An ultrasonic signal acquisition module, which is used to image the imaging target using the method of step S1 and obtain radio frequency data of the original echo signal;

[0103] A data preprocessing module, which is used to preprocess the original radio frequency data using the method of step S2;

[0104] An entropy image generation module, which uses the method of steps S3-S6 to obtain an entropy image;

[0105] A B-ultrasound image generation module, which is used to logarithmically compress the envelope data using the method of step S7 to obtain a B-ultrasound image;

[0106] An entropy value calculation module, which is used to calculate the lesion entropy value using the method of step S8;

[0107] and a classification module, which is used to classify the ultrasonic breast image using the method of step S9.

[0108] The beneficial effects of the present invention are:

[0109] The present invention utilizes fuzzy entropy to image and classify breast tumors, which can improve image contrast, improve the quality of ultrasonic imaging, and thus more accurately diagnose breast tumors, and preliminarily classify and diagnose the benign and malignant nature of breast tumors without the need for puncture and pathological experiments on patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0110] Figure 1 It is a flow chart of the ultrasound breast image classification method based on ultrasound fuzzy entropy analysis;

[0111] Figure 2 Diagnosis and imaging results of breast fibroadenoma (benign lesions);

[0112] Figure 3 Diagnosis and imaging results of invasive breast cancer (malignant lesions). DETAILED DESCRIPTION

[0113] The present invention is further described in detail below in conjunction with embodiments so that those skilled in the art can implement the invention with reference to the description.

[0114] It should be understood that the terms such as “having”, “including” and “comprising” used herein do not exclude the existence or addition of one or more other elements or combinations thereof.

[0115] The test methods used in the following examples are conventional methods unless otherwise specified. The materials and reagents used in the following examples are all commercially available unless otherwise specified. In the following examples, if no specific conditions are specified, the experiments were carried out under conventional conditions or conditions recommended by the manufacturer. The reagents or instruments used, if the manufacturer is not specified, are all conventional products that can be purchased commercially.

[0116] Example 1

[0117] A method for classifying ultrasonic breast images based on ultrasonic fuzzy entropy analysis comprises the following steps:

[0118] S1. Using an ultrasonic signal acquisition module to perform imaging with the breast as an imaging target, and obtaining radio frequency data of the original echo signal;

[0119] S2, beamforming and Hilbert transform the original RF data, demodulating it into envelope data, and then normalizing it to reduce system impact to obtain a preprocessed image;

[0120] S3, selecting a target area image that completely includes the entire lesion from the preprocessed image;

[0121] S4, setting the size of the sliding window in the target area image to M×N, and traversing the entire target area image with a step size of one pixel;

[0122] S5, rearrange the two-dimensional signal in the window into a one-dimensional signal, and calculate the fuzzy entropy of the signal in the window as the pixel value of the center coordinate of the window;

[0123] The specific calculation method of fuzzy entropy is:

[0124] S5-1. Sequence segmentation: A vector sequence with dimension m is formed in the order of 1 to N-m+1 (1 to N-m+1). in 1≤i≤N-m+1; express The mean value of

[0125] S5-2. Define distance: Define The distance between is the Chebyshev distance, where 1≤j≤Nm,j≠i, that is, the maximum absolute value of the numerical difference of each element:

[0126]

[0127] S5-3. Defining Similarity: Introducing Fuzzy Membership n Measurement Similarity, for 1≤j≤Nm,j≠i, introduce fuzzy membership function

[0128]

[0129] Among them, r is the similarity tolerance threshold;

[0130] S5-4. Find the average value: Find the average value φ of the fuzzy membership function for each i m,n,r :

[0131]

[0132] S5-5. Increase the dimension to m+1 to find φ m+1,n,r,calculate The distance between Right now

[0133]

[0134] For 1≤j≤Nm,j≠i, calculate the similarity of m+1 dimension and mean φ m+1,n,r :

[0135]

[0136] Among them, r is the similarity tolerance threshold, n determines the gradient of the similarity tolerance boundary, and the larger n is, the larger the gradient is; it plays an important weight role in the calculation process of the similarity between fuzzy entropy vectors;

[0137] S5-6. Calculate the fuzzy entropy F(m,r,n) based on the average fuzzy membership function values ​​of m and m+1 dimensions:

[0138] F(m,r,n)=lnφ m,r,n -lnφ m+1,r,n .

[0139] S6, calculating the fuzzy entropy of all pixels to obtain an entropy image;

[0140] S7, logarithmically compressing the envelope data to obtain a B-ultrasound image; the entropy image can be compositely superimposed with the B-ultrasound image to highlight the characteristics of the lesion;

[0141] S8, selecting an entropy value data block of appropriate size in the lesion area in the entropy image, and calculating the average value of the entropy value in the data block as the lesion entropy value, wherein the size of the data block is determined according to the lesion area, and the area is as large as possible but does not exceed the lesion boundary;

[0142] S9, compounding the B-ultrasound image and the entropy image to obtain a composite image, thereby improving image contrast and lesion boundary clarity, which is beneficial to improving diagnostic accuracy;

[0143] S10. Ultrasound breast image classification is performed by comparing the lesion entropy value with the tumor classification standard value:

[0144] If the lesion entropy value is less than the standard value, the ultrasound breast image is classified as a malignant tumor image, otherwise it is classified as a benign tumor image.

[0145] In this embodiment, the size of the sliding window is set to 3 to 10 times the wavelength of the ultrasonic signal, and the number of rows M of the signal in the window is:

[0146]

[0147] Where k is the set wavelength multiple, f sis the signal sampling frequency, λ is the wavelength of the ultrasonic signal emitted by the ultrasonic signal acquisition module, c is the speed of sound, f is c is the ultrasonic frequency (specifically, the center frequency of the ultrasonic transducer in the ultrasonic signal acquisition module).

[0148] In this embodiment, the number of columns N of the signal in the window is:

[0149]

[0150] Where L is the total number of ultrasonic echo signal columns, and W is the width of the ultrasonic image.

[0151] In this embodiment, M and N are both rounded integers.

[0152] In this embodiment, when the signal sequence length is 100 to 1000, m=2 is selected; when the signal length is 1000 to 30000, m=3 is selected.

[0153] In this embodiment, the value of r is usually between 0.1 and 0.25 times of the signal standard deviation. When r is large, more information will be lost, and when r is too small, the statistical characteristics of the system cannot be estimated ideally.

[0154] In this embodiment, n is between 0.5 and 3. The larger n is, the larger the gradient is, and the greater the weight on the signal distance is.

[0155] Example 2

[0156] An ultrasonic breast image classification system based on ultrasonic fuzzy entropy analysis, which uses the method of embodiment 1 to classify ultrasonic breast images, and the system includes:

[0157] An ultrasonic signal acquisition module, which is used to image the imaging target using the method of step S1 and obtain radio frequency data of the original echo signal;

[0158] A data preprocessing module, which is used to preprocess the original radio frequency data using the method of step S2;

[0159] An entropy image generation module, which uses the method of steps S3-S6 to obtain an entropy image;

[0160] A B-ultrasound image generation module, which is used to logarithmically compress the envelope data using the method of step S7 to obtain a B-ultrasound image;

[0161] An entropy value calculation module, which is used to calculate the lesion entropy value using the method of step S8;

[0162] and a classification module, which is used to classify the ultrasonic breast image using the method of step S9.

[0163] Application effect verification test

[0164] See also Figure 2 , which is the diagnosis and imaging results of breast fibroadenoma (benign lesions):

[0165] Using the method of Example 1, the present invention's verification experiment uses the breast as the target tissue for imaging. The ultrasound clinician diagnoses and collects signals on the volunteers, and obtains 47 groups of breast lesion data. Then the radio frequency data is saved to the host computer for processing and imaging. The resulting images will be used for the evaluation of the entropy imaging effect. The clinical data is collected by the Resona 9 ultrasound system (Shenzhen Mindray Bio-Medical Electronics Co., Ltd., Shenzhen, China), and the ultrasound probe uses an L14-3 linear transducer with a center frequency of 12MHz. After completing the data acquisition, the radio frequency data is stored in the computer, and the signal processing and entropy imaging are performed by the MATLAB program (Mathworks Inc., Natick, MA, USA). The ultrasound physician performs ultrasound examination diagnosis and marks the B-ultrasound image. The marked B-ultrasound image is used as a reference for evaluating the performance of the entropy image and calculating the effective range of the average entropy value. Based on these marked areas, the corresponding fuzzy entropy image is generated, and then superimposed with the B-ultrasound image to further enhance the imaging effect. Patients diagnosed by ultrasound undergo puncture biopsy and pathological experiments, and the lesion type is divided into benign and malignant according to the pathological results. Fifteen benign lesions (fibroadenomas) and 19 malignant lesions (invasive carcinomas) were selected to verify the classification and characterization performance of each imaging method. After calculating the entropy value and imaging, the average entropy value of the lesion area in each imaging method was calculated. The entropy values ​​of all benign and malignant tumors were averaged. Finally, an unpaired t-test was used to verify whether there was a significant difference in the entropy values ​​of benign and malignant lesions in each method, and to determine whether the entropy imaging method could directly characterize the lesions.

[0166] Figure 2 It is the diagnosis and imaging result of breast fibroadenoma (benign lesion). Breast fibroadenoma is a benign tumor composed of two components, glandular epithelium and fibrous tissue, and it is more common in young women. The degree and composition of the fibrous components and glandular epithelial hyperplasia in this benign tumor are different, and it can be divided into three categories: fibroadenoma, adenofibroma and adenoma. When the tumor is mainly composed of glandular epithelial hyperplasia and less fibrous components, it is called fibroadenoma; when the tumor has more fibrous tissue and fewer glandular duct components, it is called adenofibroma; and when the tumor is composed of a large number of glandular duct components, it is called adenoma. Figure 2 As shown in the ultrasound image of (a), fibroadenomas are mostly round or oval low-echo masses with no blood flow signals and capsules. The echo behind the mass is normal or slightly enhanced, and lateral acoustic shadows can be seen. In this case, ultrasound suggests BI-RADS 4a, but the lesions in the image are blurred due to the spots. Figure 2The fuzzy entropy image in (c) can significantly improve the image contrast and more accurately depict the lesion contour.

[0167] See also Figure 3 , which is the diagnosis and imaging results of invasive breast cancer (malignant lesions):

[0168] Invasive ductal breast carcinoma Figure 3 In the B-ultrasound image (a), the ultrasound indicates BI-RADS 4b, which is characterized by low-density echoes, rough and unclear lesion boundaries, and increased echo density in the surrounding tissues. The biggest difference between breast invasive ductal carcinoma and fibroadenoma in images is whether the boundaries are clear and have capsules. In B-ultrasound images, the lesion outline is often blurred due to the influence of spots, which affects the doctor's judgment of the lesion. Figure 3 The fuzzy entropy imaging result of (c) clearly shows the infiltrated tissue with a scattering intensity between normal tissue and solid lesions, and shows the rough and unclear outline of the lesion.

[0169] See Table 1 for the fuzzy entropy values ​​of benign and malignant breast tumors:

[0170]

[0171] Patients diagnosed by ultrasound underwent puncture biopsy and pathological experiments, and the lesion types were divided into benign and malignant according to the pathological results. After calculating the entropy value and imaging, the average entropy value of the lesion area in each imaging method was calculated. The entropy values ​​of all benign and malignant tumors were averaged respectively. Finally, the unpaired t-test was used to verify whether there was a significant difference in the entropy values ​​of benign and malignant lesions in each method, and to determine whether the entropy imaging method could directly characterize the lesions. The p value is the unpaired t-test result, and p<0.05 indicates that the entropy value of the malignant tumor lesion area is significantly different from that of the benign tumor. In the experimental results, p=0.26, and the fuzzy entropy value of the benign tumor is significantly higher than that of the malignant tumor. Therefore, by calculating the entropy value of the lesion area, benign and malignant tumors can be characterized and classified. This is because fuzzy entropy has a high sensitivity to tissue scattering information in ultrasound signals. Therefore, for low-echo lesions such as breast tumors, even with low amplitude, tissue scattering information can be effectively extracted from the signal.

[0172] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to specific details.

Claims

1. A method for classifying ultrasonic breast images based on ultrasonic fuzzy entropy analysis, characterized in that: The following steps are involved: S1. Using an ultrasonic signal acquisition module to perform imaging with the breast as an imaging target, and obtaining radio frequency data of the original echo signal; S2, performing beamforming and Hilbert transform on the original RF data, demodulating it into envelope data, and then normalizing it to obtain a preprocessed image; S3, selecting a target area image that completely includes the entire lesion from the preprocessed image; S4, setting the size of the sliding window in the target area image to M×N, and traversing the entire target area image with a step size of one pixel; S5, rearrange the two-dimensional signal in the window into a one-dimensional signal, and calculate the fuzzy entropy of the signal in the window as the pixel value of the center coordinate of the window; S6, calculating the fuzzy entropy of all pixels to obtain an entropy image; S7, logarithmically compressing the envelope data to obtain a B-ultrasound image; S8, selecting an entropy value data block of appropriate size in the lesion area in the entropy image, and calculating the average value of the entropy value in the data block as the lesion entropy value, wherein the size of the data block is determined according to the lesion area, and the area does not exceed the lesion boundary; S9, compounding the B-ultrasound image and the entropy image to obtain a composite image, thereby improving the image contrast and the clarity of the lesion boundary; S10. Ultrasonic breast image classification is performed by comparing the lesion entropy value with the tumor classification standard value.

2. The ultrasonic breast image classification method based on ultrasonic fuzzy entropy analysis according to claim 1 is characterized in that: The classification method in step S9 is: if the lesion entropy value is less than the standard value, the ultrasound breast image is classified as a malignant tumor image, otherwise it is classified as a benign tumor image.

3. The ultrasonic breast image classification method based on ultrasonic fuzzy entropy analysis according to claim 2 is characterized in that: The method for calculating the fuzzy entropy in step S5 comprises the following steps: S5-1. Sequence segmentation: For a time signal sequence X = {x(1), x(2), ..., x(n)} consisting of N data, a vector sequence of dimension m is formed in sequence in 1≤i≤N-m+1; express The mean value of S5-2. Define distance: Define and The distance between is the Chebyshev distance, 1≤j≤Nm,j≠i, that is, the maximum absolute value of the numerical difference of each element: S5-3. Defining Similarity: Introducing Fuzzy Membership n Measurement and Similarity, for 1≤j≤Nm,j≠i, introduce fuzzy membership function Among them, r is the similarity tolerance threshold; S5-4. Find the average value: Find the average value φ of the fuzzy membership function for each i m,n,r : S5-5. Increase the dimension to m+1 to find φ m+1,n,r ,calculate and The distance between Right now For 1≤j≤Nm,j≠i, calculate the similarity of m+1 dimension and mean φ m+1,n,r : Among them, r is the similarity tolerance threshold, n determines the gradient of the similarity tolerance boundary, and the larger n is, the larger the gradient is; S5-6. Calculate the fuzzy entropy F(m,r,n) based on the average fuzzy membership function values ​​of m and m+1 dimensions: F(m,r,n)=lnφ m,r,n -lnφ m+1,r,n 。 4. The ultrasonic breast image classification method based on ultrasonic fuzzy entropy analysis according to claim 3 is characterized in that: The sliding window size is set between 3 and 10 times the wavelength of the ultrasonic signal, and the number of rows M of the signal in the window is: Where k is the set wavelength multiple, f s is the signal sampling frequency, λ is the wavelength of the ultrasonic signal emitted by the ultrasonic signal acquisition module, c is the speed of sound, f is c is the ultrasonic frequency.

5. The ultrasonic breast image classification method based on ultrasonic fuzzy entropy analysis according to claim 4 is characterized in that: The number of columns N of the signal in the window is: Where L is the total number of ultrasonic echo signal columns, and W is the width of the ultrasonic image.

6. The ultrasonic breast image classification method based on ultrasonic fuzzy entropy analysis according to claim 5, characterized in that: Both M and N are rounded integers.

7. The ultrasonic breast image classification method based on ultrasonic fuzzy entropy analysis according to claim 6 is characterized in that: When the signal sequence length is 100 to 1000, m=2 is selected; when the signal length is 1000 to 30000, m=3 is selected.

8. The ultrasonic breast image classification method based on ultrasonic fuzzy entropy analysis according to claim 7 is characterized in that: The value of r is usually between 0.1 and 0.25 times the standard deviation of the signal.

9. The ultrasonic breast image classification method based on ultrasonic fuzzy entropy analysis according to claim 8, characterized in that: The value of n is between 0.5 and 3.

10. An ultrasonic breast image classification system based on ultrasonic fuzzy entropy analysis, characterized in that: The method according to any one of claims 1 to 9 is used to classify ultrasonic breast images, and the system comprises: An ultrasonic signal acquisition module, which is used to image the imaging target using the method of step S1 and obtain radio frequency data of the original echo signal; A data preprocessing module, which is used to preprocess the original radio frequency data using the method of step S2; An entropy image generation module, which uses the method of steps S3-S6 to obtain an entropy image; A B-ultrasound image generation module, which is used to logarithmically compress the envelope data using the method of step S7 to obtain a B-ultrasound image; An entropy value calculation module, which is used to calculate the lesion entropy value using the method of step S8; and a classification module, which is used to classify the ultrasonic breast image using the method of step S9.

Images