CN114972322A - Mammary gland micro-calcification cluster detection method combining FFDM image and DBT image - Google Patents

Mammary gland micro-calcification cluster detection method combining FFDM image and DBT image Download PDF

Info

Publication number
CN114972322A
CN114972322A CN202210731641.9A CN202210731641A CN114972322A CN 114972322 A CN114972322 A CN 114972322A CN 202210731641 A CN202210731641 A CN 202210731641A CN 114972322 A CN114972322 A CN 114972322A
Authority
CN
China
Prior art keywords
image
dbt
ffdm
segmentation result
segmentation
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
Application number
CN202210731641.9A
Other languages
Chinese (zh)
Inventor
刘半藤
叶赞挺
王柯
陈友荣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang Shuren University
Original Assignee
Zhejiang Shuren University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang Shuren University filed Critical Zhejiang Shuren University
Priority to CN202210731641.9A priority Critical patent/CN114972322A/en
Publication of CN114972322A publication Critical patent/CN114972322A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30068—Mammography; Breast
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30096—Tumor; Lesion

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

The application discloses a breast micro-calcification cluster detection method combining an FFDM image and a DBT image, belonging to the technical field of medical diagnosis and comprising the following steps: screening out a suspicious region from the full-digital mammography FFDM image based on a Yolov4 calcification target region detection algorithm; preprocessing the digital mammary tomography DBT image according to the suspicious region; constructing and training an improved U-net network model; and outputting a DBT image containing the microcalcification cluster focus through the improved U-net network model, and finishing final judgment of the focus by utilizing a mammary tomography data fusion strategy. The FFDM image and the DBT image are combined together, whether the focus exists in a patient is detected, the accuracy and the detection efficiency are considered, and the micro-calcified cluster focus is efficiently and accurately detected.

Description

Mammary gland micro-calcification cluster detection method combining FFDM image and DBT image
Technical Field
The application belongs to the technical field of medical diagnosis, and particularly relates to a breast micro-calcification cluster detection method combining an FFDM image and a DBT image.
Background
Deep learning has a good application prospect in the aspect of medical image detection, breast cancer is used as a primary disease threatening female health, and in recent years, many scholars at home and abroad research computer-aided diagnosis methods of breast cancer. For microcalcification clusters in Breast cancer pathological features, two types of data, namely FFDM (Full Field Digital Mammography) and DBT (Digital Breast tomography), are often used for detection. The method based on deep learning usually uses image enhancement, convolutional neural network and other modes to complete the detection of the focus area by extracting and classifying the image features, and is efficient and easy.
However, the existing deep learning method often only detects single data in FFDM and DBT. The detection is performed only for FFDM data, the data volume is small, the phenomenon of tissue overlapping exists in imaging, and the accuracy rate is often insufficient when the deep learning technology is used for detection. And only DBT data is detected, the DBT number is not cut and screened, the data scale is large, and the model detection efficiency is insufficient. The existing detection technology is difficult to meet the requirement of high-efficiency and accurate detection of the microcalcification cluster focus, and the accuracy and the detection efficiency cannot be considered at the same time.
Disclosure of Invention
The embodiment of the application aims to provide a method for detecting micro-calcified clusters of mammary gland by combining an FFDM image and a DBT image, and the technical problems that the existing detection technology in the prior art is difficult to meet the requirement of high-efficiency accurate detection of micro-calcified cluster lesions, and the accuracy and the detection efficiency cannot be considered at the same time can be solved.
In order to solve the technical problem, the present application is implemented as follows:
the embodiment of the application provides a breast micro-calcification cluster detection method combining an FFDM image and a DBT image, which comprises the following steps:
s101: screening out a suspicious region from the full-digital mammography FFDM image based on a Yolov4 calcification target region detection algorithm;
s102: preprocessing the digital mammary tomography DBT image according to the suspicious region;
s103: constructing and training an improved U-net network model;
s104: and outputting a DBT image containing the microcalcification cluster focus through the improved U-net network model, and finishing final judgment of the focus by utilizing a mammary tomography data fusion strategy.
In the embodiment of the application, firstly, a suspicious region is screened out from a fully digitized mammography FFDM image based on a calcification target region detection algorithm of Yolov4, a digital mammography DBT image is preprocessed according to the suspicious region, the DBT image containing a microcalcification cluster focus is output through an improved U-net network model, and a final judgment on the focus is completed by utilizing a data fusion strategy between mammary glands. The FFDM image and the DBT image are combined together, whether the focus exists in a patient is detected, the accuracy and the detection efficiency are considered, and the micro-calcified cluster focus is efficiently and accurately detected.
Drawings
Fig. 1 is a schematic flowchart of a breast microcalcification cluster detection method combining an FFDM image and a DBT image according to an embodiment of the present disclosure;
fig. 2 is a flowchart of a method for screening suspicious regions based on Yolov4 calcification target region detection algorithm according to an embodiment of the present application;
fig. 3 is a flowchart of a DBT image preprocessing provided by an embodiment of the present application;
FIG. 4 is a flowchart of training an improved U-net network model according to an embodiment of the present disclosure;
fig. 5 is a flowchart of a DTB image segmentation and data result fusion strategy thereof according to an embodiment of the present application.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings in combination with embodiments.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The breast microcalcification cluster detection method provided by the embodiment of the present application and combining the FFDM image and the DBT image is described in detail below with reference to the accompanying drawings.
Example one
Referring to fig. 1, a schematic flow chart of a breast microcalcification cluster detection method combining an FFDM image and a DBT image provided in an embodiment of the present application is shown.
The method for detecting the micro-calcification clusters of the mammary gland by combining the FFDM image and the DBT image, provided by the embodiment of the application, comprises the following steps:
s101: a calcified target area detection algorithm based on Yolov4(You Only Look one) screens out suspicious areas from full digital mammography FFDM images.
S102: and preprocessing the digital mammary tomography DBT image according to the suspicious region.
S103: and constructing and training an improved U-net network model.
Specifically, the network structure of the improved U-net network model is composed of two symmetrical up-sampling and down-sampling paths. The down-sampling process obtains high-level semantic feature information through convolution operation, and the up-sampling process obtains target positioning information through convolution and deconvolution operation. And after the semantic features in the two paths are cut, the semantic features of the image are fused with the position features by connecting Skip-conditioner with the positioning information. Compared with the original U-net network, the improved network introduces residual connection on the basis of convolution operation, and replaces pooling operation with same-scale convolution operation to improve the generalization performance of the model.
S104: and outputting a DBT image containing the microcalcification cluster focus through the improved U-net network model, and finishing final judgment of the focus by utilizing a mammary tomography data fusion strategy.
It should be noted that the U-net network outputs the detection result of each DBT image, and then, the final determination of the lesion is completed by integrating the detection results of each image in each volume of DBT image by using a data fusion strategy.
Optionally, marking of the microcalcification cluster lesion area is done on the FFDM image according to the final judgment result. The FFDM image and the DBT image are combined together, whether the focus exists in a patient is detected, the accuracy and the detection efficiency are considered, and the micro-calcified cluster focus is efficiently and accurately detected.
Referring to fig. 2, a flowchart of a method for screening suspicious regions based on Yolov4 calcification target region detection algorithm provided by the embodiment of the present application is shown.
In a possible implementation, S101 specifically includes sub-steps S1011 to S1019:
s1011: and (5) performing contour clipping on the FFDM image and extracting a mammary gland region.
S1012: and performing 2-by-2 geometric segmentation on the cut FFDM image.
S1014: and constructing a region candidate network model based on Yolov 4.
S1013: inputting the preprocessed FFDM image and the corresponding label sample into the regional candidate network model so as to train the regional candidate network model.
S1015: and saving the optimal model parameters obtained by training.
S1016: and sequentially inputting the FFDM images to be detected after segmentation pretreatment.
S1017: and detecting the FFDM image to be detected by using the optimal model stored after the training is finished to obtain a detection frame.
S1018: the confidence of each detected frame is calculated, and if the probability p is greater than 0.1, the detected frame is displayed in the output result. If the probability p is less than 0.1, the detection frame is deleted.
S1019: and outputting the suspicious region according to the detection frame.
Referring to fig. 3, a flowchart of a DBT image preprocessing provided by an embodiment of the present application is shown.
In one possible implementation, S102 comprises sub-steps S1021 to S1025:
s1021, selecting FFDM image and DBT image of the same body position of a single patient.
S1022: with the center of the suspicious region in the FFDM image as the clipping center of the DBT image in the same position, a 320 × 320 square region is clipped on the DBT image.
S1023: and judging whether the square area exceeds the image boundary, and if so, filling the black background in the boundary area.
S1024: and (3) performing image enhancement on the square region obtained by cutting by applying gamma wavelet transform shown in formula 1.
s＝cr γ Equation 1
Where r is an input value of the grayscale image, s is an output value after gamma conversion, c is a grayscale scaling coefficient, γ is a gamma factor, and γ is used to control the scaling degree.
S1025: and saving the image and finishing the preprocessing of the DBT image.
Referring to fig. 4, a flowchart for training an improved U-net network model according to an embodiment of the present application is shown.
In a possible embodiment, S103 comprises sub-steps S1031 to S1037:
s1031: and constructing an improved U-net network model.
S1032: and taking the preprocessed DBT image and the corresponding label thereof as sample input, and training the improved U-net network model.
S1033: and obtaining the focus segmentation result of the DBT image at the end of each batch of training.
S1034: and calculating a loss function in the task according to the output focus segmentation result and a formula 2.
Figure BDA0003711834690000051
Wherein L is unify For the output of the loss function, N is the total number of pixels of the image, T i For the real label of the ith pixel, aiming at the calcification detection task T i 0 or 1, 0 represents a background region not including calcification, and 1 represents a calcified region, p' i Marking the probability of calcification for the current pixel point, and beta is weightAnd (4) the coefficient.
S1035: and calculating the Dice coefficient between each segmentation result and the real label according to a formula 3.
Figure BDA0003711834690000052
Wherein, the true positive TP is the number of samples containing microcalcification clusters correctly judged. False negative FN indicates the number of samples containing microcalcification clusters judged to be normal. False positive FP is the number of samples judged to contain microcalcification clusters without microcalcification clusters.
S1036: and judging whether the model loss is reduced along with the iteration times, if the iteration results are not reduced any more, finishing the training and carrying out S1037, otherwise, carrying out S1032 after updating the model parameters.
S1037: and (5) saving the parameters of the network model and finishing the training.
Referring to fig. 5, a flowchart of a DTB image segmentation and data result fusion strategy thereof provided in an embodiment of the present application is shown.
In one possible implementation, S104 includes sub-steps S1041 to S1046:
s1041: and inputting the preprocessed DBT image to be detected.
S1042: and loading the trained improved U-net network model.
S1043: and outputting the DBT image focus segmentation result in each volume.
S1044: and judging whether the segmentation result set is empty or not by utilizing a mammary gland inter-fault data fusion strategy.
S1045: and if the segmentation result set is not empty, marking the lesion area on the FFDM image. If the set of segmentation results is empty, the volume is determined to contain no microcalcification clusters.
S1046: and (6) finishing detection.
In one possible implementation, the sub-step S1044 further includes sub-steps S1044A to S1044D:
S1044A: in the segmentation result, when the number of pixels in the segmentation area exceeds 10 × 10, the segmentation is defined to be effective.
S1044B: and when the effective segmentation result of the same volume in the DBT image is larger than 1, calculating the Euclidean distance between the centers of the segmentation results, and determining the same segmentation result when the Euclidean distance between the two is smaller than 10 pixel points, and dividing the segmentation result into the same segmentation result set.
S1044C: the split result set in which only a single detection result exists in the volume is removed.
S1044D: and traversing the rest of the segmentation result sets, and if the segmentation result sets are not empty, outputting the starting layer number and the ending layer number of the effective segmentation results in the segmentation result sets.
In the embodiment of the application, firstly, a suspicious region is screened out from a fully digitized mammography FFDM image based on a calcification target region detection algorithm of Yolov4, a digital mammography DBT image is preprocessed according to the suspicious region, the DBT image containing a microcalcification cluster focus is output through an improved U-net network model, and a final judgment on the focus is completed by utilizing a data fusion strategy between mammary glands. The FFDM image and the DBT image are combined together, whether the focus exists in a patient is detected, the accuracy and the detection efficiency are considered, and the micro-calcified cluster focus is efficiently and accurately detected.
In summary, the invention provides a breast micro-calcification cluster detection method combining FFDM and DBT data, which locates suspicious regions in FFDM by using a regional candidate network based on Yolov4, and cuts DBT data to input into an improved U-Net network by using suspicious region information, thereby realizing efficient and accurate segmentation of DBT data. The method utilizes suspicious regions generated by FFDM to perform region cutting on DBT data, and solves the problem of overlarge input scale in the conventional DBT data detection method; and the DBT data is utilized to re-screen the FFDM detection result, so that the problem of high false positive rate in the conventional micro-calcification cluster detection task aiming at the FFDM is solved. The method integrates the information of the two, judges whether the patient contains the microcalcification cluster focus or not, can effectively improve the true positive rate of detection and reduce the false positive rate, and has better practical application value.
The above description is only an example of the present invention, and is not intended to limit the present invention. Various modifications and alterations to this invention will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims (6)

1. A breast microcalcification cluster detection method combining an FFDM image and a DBT image is characterized by comprising the following steps:
s101: screening out a suspicious region from the full-digital mammography FFDM image based on a Yolov4 calcification target region detection algorithm;
s102: preprocessing the digital mammary tomography DBT image according to the suspicious region;
s103: constructing and training an improved U-net network model;
s104: and outputting a DBT image containing the microcalcification cluster focus through the improved U-net network model, and finishing final judgment of the focus by utilizing a mammary tomography data fusion strategy.
2. The detection method according to claim 1, wherein the S101 specifically includes:
s1011: performing contour clipping on the FFDM image to extract a mammary gland region;
s1012: 2 x 2 geometric segmentation is carried out on the cut FFDM image;
s1014: constructing a regional candidate network model based on Yolov 4;
s1013: inputting the preprocessed FFDM image and a corresponding label sample into the regional candidate network model to train the regional candidate network model;
s1015: saving the optimal model parameters obtained by training;
s1016: sequentially inputting to-be-detected FFDM images subjected to segmentation pretreatment;
s1017: detecting the FFDM image to be detected by using the optimal model stored after training is finished to obtain a detection frame;
s1018: calculating the confidence of each detection frame, and if the probability p is greater than 0.1, displaying the detection frame in an output result; if the probability p is less than 0.1, deleting the detection frame;
s1019: and outputting the suspicious region according to the detection frame.
3. The detection method according to claim 1, wherein the S102 specifically includes:
s1021, selecting an FFDM image and a DBT image of a single patient in the same body position;
s1022: cutting out a 320 x 320 square area on the DBT image by taking the center of the suspicious area in the FFDM image as the cutting center of the DBT image in the same position;
s1023: judging whether the square area exceeds the image boundary, and if so, filling a black background in the boundary area;
s1024: performing image enhancement on the square area obtained by cutting by applying gamma wavelet transform shown in formula 1;
s＝cr γ equation 1
Wherein r is an input value of the gray image, s is an output value after gamma conversion, c is a gray scaling coefficient, gamma is a gamma factor, and gamma is used for controlling the scaling degree;
s1025: and saving the image and finishing the preprocessing of the DBT image.
4. The detection method according to claim 1, wherein the S103 specifically includes:
s1031: constructing the improved U-net network model;
s1032: inputting the preprocessed DBT image and a label corresponding to the DBT image as a sample, and training the improved U-net network model;
s1033: obtaining a focus segmentation result of the DBT image when each batch of training is finished;
s1034: calculating a loss function in a task according to the output focus segmentation result and a formula 2;
Figure FDA0003711834680000021
wherein L is unify For the output of the loss function, N is the total number of pixels of the image, T i For the real label of the ith pixel, aiming at the calcification detection task T i 0 or 1, 0 represents a background region not including calcification, and 1 represents a calcified region, p' i Marking the probability of calcification of the current pixel point, wherein beta is a weight coefficient;
s1035: calculating a Dice coefficient between each segmentation result and the real label according to a formula 3;
Figure FDA0003711834680000022
wherein, the positive TP is the number of samples containing microcalcification clusters correctly judged; false negative FN is the number of samples containing microcalcification cluster judged to be normal; the false positive FP is the number of samples which do not contain the microcalcification cluster and are judged as containing the microcalcification cluster;
s1036: judging whether the model loss is reduced along with the iteration times, if the iteration results are not reduced any more, finishing the training and carrying out S1037, otherwise, carrying out S1032 after updating the model parameters;
s1037: and (5) saving the parameters of the network model, and finishing training.
5. The detection method according to claim 1, wherein the S104 specifically includes:
s1041: inputting a preprocessed DBT image to be detected;
s1042: loading the trained improved U-net network model;
s1043: outputting the DBT image focus segmentation result in each volume;
s1044: judging whether the segmentation result set is empty or not by utilizing a mammary gland inter-fault data fusion strategy;
s1045: if the segmentation result set is not empty, marking a focus area on the FFDM image; if the segmentation result set is empty, judging the volume as not containing the micro calcification clusters;
s1046: and (6) finishing detection.
6. The detection method according to claim 5, wherein the S1044 specifically includes:
S1044A: in the segmentation result, when the pixel of the segmentation area exceeds 10 x 10, the segmentation is defined as effective segmentation;
S1044B: when the effective segmentation result of the same volume in the DBT image is larger than 1, calculating the Euclidean distance between the centers of all the segmentation results, and when the Euclidean distance between the two Euclidean distances is smaller than 10 pixel points, determining the same segmentation result and dividing the same segmentation result set;
S1044C: removing a segmentation result set with only a single detection result in the volume;
S1044D: and traversing the rest of the segmentation result sets, and if the segmentation result sets are not empty, outputting the starting layer number and the ending layer number of the effective segmentation results in the segmentation result sets.
CN202210731641.9A 2022-06-24 2022-06-24 Mammary gland micro-calcification cluster detection method combining FFDM image and DBT image Pending CN114972322A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210731641.9A CN114972322A (en) 2022-06-24 2022-06-24 Mammary gland micro-calcification cluster detection method combining FFDM image and DBT image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210731641.9A CN114972322A (en) 2022-06-24 2022-06-24 Mammary gland micro-calcification cluster detection method combining FFDM image and DBT image

Publications (1)

Publication Number Publication Date
CN114972322A true CN114972322A (en) 2022-08-30

Family

ID=82964793

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210731641.9A Pending CN114972322A (en) 2022-06-24 2022-06-24 Mammary gland micro-calcification cluster detection method combining FFDM image and DBT image

Country Status (1)

Country Link
CN (1) CN114972322A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115619641A (en) * 2022-10-24 2023-01-17 中山大学附属第五医院 Mammary gland image processing method, system, terminal and medium based on FFDM
CN116363155A (en) * 2023-05-25 2023-06-30 南方医科大学南方医院 Intelligent pectoral large muscle region segmentation method, device and storage medium
CN117115515A (en) * 2023-08-07 2023-11-24 南方医科大学南方医院 Digital breast three-dimensional tomography structure distortion focus image processing method

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115619641A (en) * 2022-10-24 2023-01-17 中山大学附属第五医院 Mammary gland image processing method, system, terminal and medium based on FFDM
CN116363155A (en) * 2023-05-25 2023-06-30 南方医科大学南方医院 Intelligent pectoral large muscle region segmentation method, device and storage medium
CN116363155B (en) * 2023-05-25 2023-08-15 南方医科大学南方医院 Intelligent pectoral large muscle region segmentation method, device and storage medium
CN117115515A (en) * 2023-08-07 2023-11-24 南方医科大学南方医院 Digital breast three-dimensional tomography structure distortion focus image processing method

Similar Documents

Publication Publication Date Title
CN114972322A (en) Mammary gland micro-calcification cluster detection method combining FFDM image and DBT image
CN110930416B (en) MRI image prostate segmentation method based on U-shaped network
CN109685768B (en) Pulmonary nodule automatic detection method and system based on pulmonary CT sequence
CN111563902A (en) Lung lobe segmentation method and system based on three-dimensional convolutional neural network
CN112132959B (en) Digital rock core image processing method and device, computer equipment and storage medium
CN107274402A (en) A kind of Lung neoplasm automatic testing method and system based on chest CT image
CN110689525B (en) Method and device for identifying lymph nodes based on neural network
CN113808146B (en) Multi-organ segmentation method and system for medical image
CN111402254B (en) CT image lung nodule high-performance automatic detection method and device
CN112862808A (en) Deep learning-based interpretability identification method of breast cancer ultrasonic image
CN113223005B (en) Thyroid nodule automatic segmentation and grading intelligent system
CN114758137B (en) Ultrasonic image segmentation method and device and computer readable storage medium
CN111444844A (en) Liquid-based cell artificial intelligence detection method based on variational self-encoder
CN113421240B (en) Mammary gland classification method and device based on ultrasonic automatic mammary gland full-volume imaging
CN112651929B (en) Medical image organ segmentation method and system based on three-dimensional full-convolution neural network and region growing
CN112053325A (en) Breast mass image processing and classifying system
CN112508884A (en) Comprehensive detection device and method for cancerous region
CN116758336A (en) Medical image intelligent analysis system based on artificial intelligence
CN113344933B (en) Glandular cell segmentation method based on multi-level feature fusion network
CN114565601A (en) Improved liver CT image segmentation algorithm based on DeepLabV3+
CN111062909A (en) Method and equipment for judging benign and malignant breast tumor
CN115760875A (en) Full-field medical picture region segmentation method based on self-supervision learning
CN115661029A (en) Pulmonary nodule detection and identification system based on YOLOv5
CN114140830A (en) Repeated identification inhibition method based on circulating tumor cell image
CN112529911A (en) Training method of pancreas image segmentation model, image segmentation method and device

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