CN109166108B - Automatic identification method for abnormal lung tissue of CT (computed tomography) image - Google Patents
Automatic identification method for abnormal lung tissue of CT (computed tomography) image Download PDFInfo
- Publication number
- CN109166108B CN109166108B CN201810922391.0A CN201810922391A CN109166108B CN 109166108 B CN109166108 B CN 109166108B CN 201810922391 A CN201810922391 A CN 201810922391A CN 109166108 B CN109166108 B CN 109166108B
- Authority
- CN
- China
- Prior art keywords
- template
- image
- lung
- block
- group
- 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.)
- Active
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
- G06T7/0014—Biomedical image inspection using an image reference approach
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- 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
- G06T2207/30061—Lung
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Geometry (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention relates to an automatic film reading technology realized by a computer in medical treatment, in particular to an automatic identification method of abnormal lung tissues of CT images, which is realized by adopting a template differential comparison technology. The method mainly uses a group of parameter vectors extracted from normal standard CT lung images as a template group, compares the template vectors with real-time dynamic CT lung images, and quickly finds abnormal tissues and automatically marks the abnormal tissues.
Description
Technical Field
The invention relates to an automatic film reading technology realized by a computer in medical treatment, in particular to an automatic identification method of abnormal lung tissues of CT images, which is realized by adopting a template difference comparison technology.
Background
The widespread use of a large number of CT images in medicine has caused two problems: on one hand, the information amount of the image generated by one-time examination of each patient is larger and larger, the image information which needs to be read by a doctor every day is too much, and the working intensity is larger and larger; on the other hand, the level of diagnosis is very dependent on the experience and ability of the diagnostician, and misdiagnosis occurs. With the development of the DICOM standard in the medical image format and the computer-aided diagnosis technology, it is increasingly important to realize the automatic identification of the abnormal tissue of the CT image and provide the accuracy of the film reading diagnosis.
The main application targets of the method for automatically identifying the abnormal lung tissue of the CT image are shown in two aspects: on the basis of processing a large amount of CT lung image data of a patient, suspicious lesions are found out as accurately as possible, are marked in a key way, are provided for doctors to serve as a reference for further diagnosis, reduce the workload of reading films by the doctors and provide the accuracy of diagnosis. Secondly, modeling is realized on the basis of a large number of medical imaging processing methods and a large number of analyses on some special diseases, normal tissues and focus information which are easy to be confused in imaging are distinguished through selection of model parameters, and real focuses are screened out.
Disclosure of Invention
Aiming at the technical defects in the prior art, the invention relates to an automatic film reading technology realized by a computer in medical treatment, in particular to an automatic identification method of CT image lung abnormal tissue realized by adopting a template difference comparison technology. The method mainly uses a group of parameter vectors extracted from normal standard CT lung images as a template group, compares the template vectors with real-time dynamic CT lung images, and quickly finds abnormal tissues and automatically marks the abnormal tissues.
Firstly, screening 50 healthy people lung CT images as standard templates according to gender, age group and obesity grade of men and women, wherein the total number of the lung CT images is 50, the number of the templates is the number of the male from M1 and M2 ….. M25, and the number of the female from F1 and F2 ….. F25, and then generating template data according to steps.
The invention relates to a CT standard image template characteristic constructed by describing each detail point characteristic by a plurality of groups of data such as two-dimensional polar coordinates, node types, curvatures, directivities, distances among effective associated points, expansion factors and the like, and storing the comparison templates in a database strictly according to grouping numbers.
In the actual CT lung image comparison process, we first determine the age and sex of the patient, and then perform the same analysis processing steps as the template image until the characteristic data is finally obtained. Then, the corresponding grouping feature template data are found and are respectively compared one by one.
In the template alignment process, three main processing steps are required:
(1) we respectively obtain the difference blocks S of 5 lung image template data in the same groupn(optionally none) and area class betan(0-10 levels are divided according to the area of the block);
(2) at SnOf more than 2 templates'(n)Image stripping is performed according to the effectiveness of block closure, respectively. From s 'if image closure is invalid'(n)Deleting the list;
(3) calculate each s'(n)The maximum edge distance is used as the radius r, and finally, an effective abnormal block is marked on the original CT image.
Drawings
FIG. 1: is a schematic diagram of CT lung template feature generation of the invention;
FIG. 2 is a drawing: is a schematic diagram of the CT lung image comparison operation process.
Detailed Description
The invention is further illustrated with reference to the following figures and examples:
firstly, screening 50 healthy people's lung CT images as standard templates according to gender, age group and obesity grade of men and women, wherein the total number of the CT images is 50, the template numbers are M1 and M2 ….. M25 for men and F1 and F2 ….. F25 for women, and then generating template data according to steps, as shown in figure 1.
The CT lung template feature generation in FIG. 1 is based on gender classification of male and female, and is established by groups of childhood (0-6 years), teenager (7-17 years), teenager (18-40 years), middle aged (41-65 years) and elderly (66 years later). Each template has its own ID number, which is 3 bits: the first is gender (1: male, 2: female); the second bit represents the group number (1-5); the third bit represents the serial number in the same group.
The CT standard image template feature minutiae points are characterized by each minutiae point through data such as two-dimensional polar coordinates, node types, curvatures, directivities, distances between effective associated points, expansion factors and the like, and the comparison templates are stored in a database strictly according to grouping numbers.
In the template comparison process, firstly, for any input CT lung image, the same preprocessing as the template processing step is carried out to obtain the feature data of the minutiae points. Then selecting corresponding template groups to compare one by one. Finally, respectively obtaining the difference blocks S of the 5 lung image template data in the same groupn(optionally none) and area class betan(0 to 10 stages are divided by block area), and then S is obtainednMinimum rectangle of block and save four coordinate points, for betanThe value of (c) is also saved. Allowing for multiple difference blocks, SnE.g. Diff _ Area, the data structure is as follows:
struct Diff_Area {
short Block _ Number// difference Block Number
Struct BLOCK _ INDEX [ Block _ Number ]/difference BLOCK coordinate
Short Scale [ Block _ Number ],/Difference Block area level
};
The difference block coordinate data structure is as follows:
Struct BLOCK_INDEX {
int Top _ X// X coordinate on block
Int Top _ Y// Y coordinate on block
Int Down X// X coordinate under block
Int Down Y// block Down Y coordinate
Int Left X// block Left X coordinate
Int Left Y coordinate// block
Int Right _ X// block Right X coordinate
Int Right _ Y// block Right Y coordinate
};
In the comparison algorithm, the second step is that the same operation is carried out between the difference blocks obtained by comparing the five templates. At SnOf more than 2 templates'(n)Are considered to be effective.
And respectively carrying out image closeness operation on the effective blocks. If the threshold is exceeded, then deemed closed invalid, then from s'(n)And deleting the list.
The last step is to calculate each s'(n)The maximum edge distance is used as the radius r, and finally, an effective abnormal block is inscribed on the original CT image, and the comparison algorithm flow is shown in FIG. 2.
The establishment of the standard characteristic template of the CT lung image requires an expert doctor to screen out soldiers from the existing CT file bank to complete modeling work. For the CT lung image to be checked, all working links such as comparison, labeling and the like are automatically read in and completed through a computer without manual intervention. The following example is performed with the standard feature template of the CT lung image and the comparison process.
1) Data structure for establishing standard characteristic template of CT lung image
The template data is characterized by two-dimensional polar coordinates, node types, curvatures, directivities, distances between effective association points, scaling factors and the like, and the comparison templates are stored in a database strictly according to grouping numbers. Its main data structure is thus defined as follows
struct CT_ BellowsForm {
char ID [4],/template numbering
int group ID// group numbering
char Sex// Sex: m: male, F: woman
date Birthday// birth date
int Block _ ToT _ Number// Total Number of blocks
int Block _ Serial _ No// Block Serial number
struct Block _ Area × PArea; // Block pointer
struct Block _ Direct _ PDir// Block pattern pointer
structure Block _ Curvolume PCur// Block direction Curvature
struct Block _ KeyPoint PKeyPoint// key node coordinates
struct Block _ KeyPoint _ Distance PDist/key point vector table
struct Block _ StretchFactor _ PFact; // key point scaling factor
int PScale// class of block area size (1 ~ 10)
}
struct Block_Area {
Short Block _ Number// Block Number
Struct BLOCK _ INDEX PNo// BLOCK coordinates
Short Scale [ Block _ Number ],/Block area level
} *PArea;
The difference block coordinate data structure is as follows:
Struct BLOCK_INDEX {
int Top _ X// X coordinate on block
Int Top _ Y// Y coordinate on block
Int Down X// X coordinate under block
Int Down Y// block Down Y coordinate
Int Left X// block Left X coordinate
Int Left Y coordinate// block
Int Right _ X// block Right X coordinate
Int Right _ Y// block Right Y coordinate
} *PNo;
2) CT lung image preprocessing
Baseline wander processing
l image
l image segmentation
l block division
l … ...
3) Implementation of the comparison Algorithm
First, we need to determine the preset matching amounts of the reject rate (FAR), the False Recognition Rate (FRR) and the Similarity (Similarity) as three groups:
0.1%、 0.02%、0.75%;
0.05%、0.06%、0.85%;
0.03%、0.08%、0.93%;
meanwhile, it is also necessary to set up a reasonable comparison threshold value for each important index in the images, blocks and detail points through a large amount of measurement and calculation, which is not particularly described herein.
ECGWaveVerify (WaveData, WaveTemplate, VerifyParameter) for the alignment algorithm;
4) CT image template reconstruction
The template reconstruction function name is:
ECGWaveTemplateRefactoring(OldWaveTemplate, NewWaveTemplate ,Parameter);
5) CT image template optimization
The template optimization function name is:
ECGWaveTemplateOptimize (WaveTemplate, Parameter);
the method is designed from the angle of the CT lung image template, can improve the precision and reliability of the CT image template to a certain extent, and improves the accuracy of automatic pre-diagnosis.
Claims (3)
1. A method for automatically identifying abnormal lung tissues of CT images by adopting a template differential comparison technology is characterized in that parameter vectors of CT lung images in a group of normal standard forms are extracted to serve as a template group, and the parameter vectors serve as template characteristic data;
the comparison between the template set and the CT lung image collected in real time is a method for finding out abnormal blocks by adopting a difference block segmentation technology;
wherein, the step of extracting the parameter vector of the CT lung image based on a group of normal standard forms as a template group comprises:
screening lung CT images of a set number of healthy people respectively according to sex, age group and obesity level of men and women as standard templates, numbering the standard templates, and extracting parameter vectors of the standard templates as a template group according to the steps;
the method for finding the abnormal block further comprises the following steps:
s1, obtaining difference blocks Sn and area levels betan of the lung image template data with the same set number respectively;
s2, in Sn, image stripping is carried out on the overlapped areas S ' (n) with more than 2 templates according to the validity of block closure, and if the image closure is invalid, the overlapped areas S ' (n) are deleted from the S ' (n) list;
s3, the center point of each S' (n) is calculated, the maximum edge distance is used as the radius r, and finally, the valid abnormal region is circled on the original CT image.
2. The method of claim 1, characterized by a vector adaptive feature of CT lung image size.
3. The method of claim 1, wherein the extracting the vector thereof as template feature data comprises: and (3) using a plurality of groups of two-dimensional polar coordinates, node types, curvatures, directional diagrams, distances among effective association points and expansion factors obtained after CT image preprocessing as parameter vectors of lung surrounding tissue template features.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810922391.0A CN109166108B (en) | 2018-08-14 | 2018-08-14 | Automatic identification method for abnormal lung tissue of CT (computed tomography) image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810922391.0A CN109166108B (en) | 2018-08-14 | 2018-08-14 | Automatic identification method for abnormal lung tissue of CT (computed tomography) image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109166108A CN109166108A (en) | 2019-01-08 |
CN109166108B true CN109166108B (en) | 2022-04-08 |
Family
ID=64895564
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810922391.0A Active CN109166108B (en) | 2018-08-14 | 2018-08-14 | Automatic identification method for abnormal lung tissue of CT (computed tomography) image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109166108B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113450899A (en) * | 2021-06-22 | 2021-09-28 | 上海市第一人民医院 | Intelligent diagnosis guiding method based on artificial intelligence cardiopulmonary examination images |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006114003A1 (en) * | 2005-04-27 | 2006-11-02 | The Governors Of The University Of Alberta | A method and system for automatic detection and segmentation of tumors and associated edema (swelling) in magnetic resonance (mri) images |
WO2008107809A2 (en) * | 2007-03-06 | 2008-09-12 | Koninklijke Philips Electronics, N.V. | Automated diagnosis and alignment supplemented with pet/mr flow estimation |
CN106846317A (en) * | 2017-02-27 | 2017-06-13 | 北京连心医疗科技有限公司 | A kind of feature based extracts the method for retrieving medicine image with Similarity matching |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1934589A (en) * | 2004-03-23 | 2007-03-21 | 美国西门子医疗解决公司 | Systems and methods providing automated decision support for medical imaging |
CN101727666B (en) * | 2008-11-03 | 2013-07-10 | 深圳迈瑞生物医疗电子股份有限公司 | Image segmentation method and device, and method for judging image inversion |
JP5962237B2 (en) * | 2012-06-11 | 2016-08-03 | コニカミノルタ株式会社 | Chest diagnosis support information generation method |
-
2018
- 2018-08-14 CN CN201810922391.0A patent/CN109166108B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006114003A1 (en) * | 2005-04-27 | 2006-11-02 | The Governors Of The University Of Alberta | A method and system for automatic detection and segmentation of tumors and associated edema (swelling) in magnetic resonance (mri) images |
WO2008107809A2 (en) * | 2007-03-06 | 2008-09-12 | Koninklijke Philips Electronics, N.V. | Automated diagnosis and alignment supplemented with pet/mr flow estimation |
CN106846317A (en) * | 2017-02-27 | 2017-06-13 | 北京连心医疗科技有限公司 | A kind of feature based extracts the method for retrieving medicine image with Similarity matching |
Non-Patent Citations (2)
Title |
---|
"Anatomical landmarks for registration of experimental image data to volumetric rodent brain atlasing templates";Marina Sergejeva等;《Journal of Neuroscience Methods》;20150130;第240卷;第161-169页 * |
"基于图形分析方法的函数型数据异常值检验实证研究";米子川;《统计与信息论坛》;20140630;第29卷(第6期);第18-24页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109166108A (en) | 2019-01-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Vijayarajeswari et al. | Classification of mammogram for early detection of breast cancer using SVM classifier and Hough transform | |
Tilve et al. | Pneumonia detection using deep learning approaches | |
CN108898595B (en) | Construction method and application of positioning model of focus region in chest image | |
US9355450B2 (en) | Apparatus and method for utilizing mammogram images for verification | |
US20120283574A1 (en) | Diagnosis Support System Providing Guidance to a User by Automated Retrieval of Similar Cancer Images with User Feedback | |
CN109389584A (en) | Multiple dimensioned rhinopharyngeal neoplasm dividing method based on CNN | |
Vezzetti et al. | 3D Landmarking in multiexpression face analysis: a preliminary study on eyebrows and mouth | |
Chen et al. | Anatomy-aware siamese network: Exploiting semantic asymmetry for accurate pelvic fracture detection in x-ray images | |
CN101103924A (en) | Galactophore cancer computer auxiliary diagnosis method based on galactophore X-ray radiography and system thereof | |
Sapate et al. | Breast cancer diagnosis using abnormalities on ipsilateral views of digital mammograms | |
CN101714153A (en) | Visual perception based interactive mammography image searth method | |
Senan et al. | Automated diagnosis of chest X-ray for early detection of COVID-19 disease | |
Aswathy et al. | Performance analysis of segmentation algorithms for the detection of breast cancer | |
CN109166108B (en) | Automatic identification method for abnormal lung tissue of CT (computed tomography) image | |
CN112529900B (en) | Method, device, terminal and storage medium for matching ROI in mammary gland image | |
Bessa et al. | Normal breast identification in screening mammography: a study on 18 000 images | |
CN111768845B (en) | Pulmonary nodule auxiliary detection method based on optimal multi-scale perception | |
CN114998203A (en) | System and method for accurately diagnosing occupational pneumoconiosis based on artificial intelligence | |
Mousavi Moghaddam et al. | Lung parenchyma segmentation from CT images with a fully automatic method | |
Nandanwar et al. | Analysis of pixel intensity variation by performing morphological operations for image segmentation on cervical cancer pap smear image | |
Sridevi et al. | A combined deep CNN-lasso regression feature fusion and classification of MLO and CC view mammogram image | |
KR102620046B1 (en) | Method and system for breast ultrasonic image diagnosis using weakly-supervised deep learning artificial intelligence | |
Jadhav et al. | Comparative Study of breast cancer detection methods | |
CN117476219B (en) | Auxiliary method and auxiliary system for positioning CT (computed tomography) tomographic image based on big data analysis | |
Alimudin et al. | Similarity measurement on digital mammogram classification |
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 |