CN112652394A - Multi-focus target detection-based retinopathy of prematurity diagnosis system - Google Patents
Multi-focus target detection-based retinopathy of prematurity diagnosis system Download PDFInfo
- Publication number
- CN112652394A CN112652394A CN202110045855.6A CN202110045855A CN112652394A CN 112652394 A CN112652394 A CN 112652394A CN 202110045855 A CN202110045855 A CN 202110045855A CN 112652394 A CN112652394 A CN 112652394A
- Authority
- CN
- China
- Prior art keywords
- module
- target detection
- fundus
- diagnosis
- picture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 95
- 238000003745 diagnosis Methods 0.000 title claims abstract description 66
- 206010038933 Retinopathy of prematurity Diseases 0.000 title claims description 27
- 238000012545 processing Methods 0.000 claims abstract description 46
- 210000001525 retina Anatomy 0.000 claims abstract description 32
- 230000002028 premature Effects 0.000 claims abstract description 27
- 230000000007 visual effect Effects 0.000 claims abstract description 14
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 8
- 230000001575 pathological effect Effects 0.000 claims description 50
- 238000012549 training Methods 0.000 claims description 27
- 238000000034 method Methods 0.000 claims description 26
- 230000008569 process Effects 0.000 claims description 13
- 230000000877 morphologic effect Effects 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 10
- 238000013135 deep learning Methods 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 5
- 230000002207 retinal effect Effects 0.000 claims description 5
- 101000878595 Arabidopsis thaliana Squalene synthase 1 Proteins 0.000 claims description 3
- 101000642811 Oryza sativa subsp. indica Soluble starch synthase 1, chloroplastic/amyloplastic Proteins 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000005352 clarification Methods 0.000 claims description 3
- 208000017442 Retinal disease Diseases 0.000 abstract description 5
- 206010038923 Retinopathy Diseases 0.000 abstract description 5
- 230000009286 beneficial effect Effects 0.000 abstract description 4
- 238000011158 quantitative evaluation Methods 0.000 description 3
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 208000030533 eye disease Diseases 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 229910052760 oxygen Inorganic materials 0.000 description 2
- 239000001301 oxygen Substances 0.000 description 2
- 230000004083 survival effect Effects 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 206010012689 Diabetic retinopathy Diseases 0.000 description 1
- 208000010412 Glaucoma Diseases 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000035935 pregnancy Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06T5/00—Image enhancement or restoration
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Abstract
The invention discloses a premature infant retinopathy diagnosis system based on multi-focus target detection, which belongs to the technical field of image processing and comprises an eyeground picture storage module, an eyeground picture acquisition module, an eyeground picture processing module, an eyeground picture target detection module, a target detection sub-model generation module, an integrated model generation module, a retina analysis diagnosis module, a diagnosis report formation module and a visual display module; the system comprises a fundus picture storage module, a fundus picture acquisition module, a fundus picture processing module, a fundus picture target detection module, a target detection sub-model generation module, an integrated model generation module, a retina analysis and diagnosis module, a diagnosis report formation module and a visual display module, wherein the fundus picture storage module, the fundus picture acquisition module, the fundus picture processing module, the fundus picture target detection module, the target detection sub-model generation module, the integrated model generation module, the retina analysis and; the invention constructs a plurality of target detection submodels and an integrated model, thereby being beneficial to reducing the overall prediction deviation of the system and improving the interpretability of a diagnosis result.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a retinopathy of prematurity diagnosis system based on multi-focus target detection.
Background
Through retrieval, Chinese patent No. CN108717869A discloses a diabetic retinal complication diagnosis auxiliary system based on a convolutional neural network, and the invention adopts a single deep neural network model, although the diagnosis process is optimized, the overall prediction deviation is larger, and the interpretability of the result is lower; the premature infant is an infant born in less than 37 weeks during pregnancy, high-concentration oxygen inhalation is an important technology for ensuring the survival rate of the premature infant, but excessive oxygen inhalation can cause retinopathy of the premature infant, the survival rate of the premature infant is remarkably increased along with the continuous improvement of the medical level in developing China represented by China, but the incidence rate of retinopathy of the premature infant is on the rise, and the retinopathy of the premature infant becomes the main cause of blindness of children all over the world; fundus photos are an important means for diagnosing and screening retinopathy of prematurity, and are effective methods for discovering occult eye diseases in early stage, for example, occult eye diseases such as glaucoma and diabetic retinopathy have no vision change symptom in early stage, the fundus photos have important significance for guiding diagnosis and treatment of fundus diseases and evaluating the health condition of the whole body, with the popularization of deep learning technology in recent years, deep neural models with strong fitting capability and feature extraction capability are designed, and the models are used for target detection and classification prediction of various medical images and achieve the level close to the special class of human; therefore, it becomes important to invent a retinopathy of prematurity diagnosis system based on multi-focal target detection.
Most of the existing retinopathy diagnosis systems adopt a single deep neural network model, the integral prediction deviation is large, the interpretability of the result is low, and quantitative evaluation of various pathological structures cannot be output; to this end, we propose a system for diagnosing retinopathy of prematurity based on multi-focal target detection.
Disclosure of Invention
The invention aims to solve the defects in the prior art and provides a retinopathy of prematurity diagnosis system based on multi-focus target detection.
In order to achieve the purpose, the invention adopts the following technical scheme:
the system for diagnosing the retinopathy of prematurity based on multi-focus target detection comprises an eyeground picture storage module, an eyeground picture acquisition module, an eyeground picture processing module, an eyeground picture target detection module, a target detection sub-model generation module, an integrated model generation module, a retina analysis and diagnosis module, a diagnosis report formation module and a visual display module;
the eye fundus picture storage module, the eye fundus picture acquisition module, the eye fundus picture processing module, the eye fundus picture target detection module, the target detection sub-model generation module, the integrated model generation module, the retina analysis and diagnosis module, the diagnosis report formation module and the visual display module are integrated in the computer; the visual display module is specifically a computer display screen.
Preferably, the fundus picture storage module is used for storing a large number of fundus pictures of the retina of the premature infant; the fundus picture acquisition module is used for taking fundus pictures of a large number of retinas of premature infants in the fundus picture storage module and sending the fundus pictures to the fundus picture processing module.
Preferably, the fundus picture processing module is configured to receive a large number of fundus pictures of the premature infant retina sent by the fundus picture acquisition module, and perform optimization processing on image parameters by using a spatial histogram enhancement method and a frequency domain homomorphic filtering enhancement method; the image parameters include sharpness, hue, gray scale, size, and noise.
Preferably, the fundus picture target detection module is used for performing target detection on fundus pictures of a large number of premature infant retinas subjected to optimization processing by the fundus picture processing module, identifying and analyzing a target area of the fundus pictures, and obtaining various pathological morphological characteristics in retinopathy of prematurity; the target detection submodel generation module is used for training various pathological morphological characteristics in retinopathy of prematurity by utilizing a deep learning algorithm to obtain a plurality of target detection submodels; the specific generation process of the target detection submodel generation module is as follows:
s1: extracting a large number of fundus pictures of the retina of the premature infant by using a fundus picture acquisition module, and processing the fundus pictures by using a fundus picture processing module and a fundus picture target detection module to obtain various pathomorphology feature sets;
s2: randomly distributing various pathological morphological feature sets to form a training set 1, a training set 2, … … and a training set n;
s3: constructing a plurality of classifiers, and respectively putting a training set 1, a training set 2, a training set … … and a training set n into the plurality of classifiers to obtain a target detection submodel 1, a target detection submodel 2, a target detection submodel … … and a target detection submodel n;
the deep learning algorithm is specifically a deep neural network.
Preferably, the integrated model generation module is used for extracting quantitative features of various pathological structures of the multiple target detection submodels and constructing the integrated model according to the quantitative features of the various pathological structures; the integrated model generation module specifically generates the following processes:
SS 1: for each fundus picture, respectively predicting by a plurality of target detection submodels, and obtaining quantitative characteristics 1 of a pathological structure, 2 and … … of the pathological structure and n of the pathological structure according to the size, the range, the number and the like of focuses;
SS 2: constructing an integrated classifier, taking the quantitative feature 1 of the pathological structure, the quantitative features 2 and … … of the pathological structure and the quantitative feature n of the pathological structure as input data of the integrated classifier, marking a diagnosis result by a human expert, and training through an integrated learning strategy to obtain an integrated model.
Preferably, the retinal analysis and diagnosis module is configured to receive the quantitative characteristics of the new fundus image, and input the quantitative characteristics into the integrated model to obtain a diagnosis result, and the retinal analysis and diagnosis module specifically performs an analysis and diagnosis process as follows:
SSS 1: aiming at the new fundus picture, respectively predicting by each target detection sub-model to obtain quantitative characteristics of various pathological structures;
SSS 2: and inputting the extracted quantitative characteristics of various pathological structures into the integrated model to obtain a diagnosis result.
Preferably, the diagnostic report forming module is configured to perform a sharpening process on the diagnostic result by using graph marking software to form a diagnostic report, where the graph type of the graph marking software includes a column graph, a line graph, a pie graph, a bar graph, and a three-line graph; the visual display module is used for visually displaying the diagnosis report and providing the diagnosis assistance for the medical staff.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention is provided with a fundus picture processing module and a fundus picture target detection module; the fundus picture processing module receives a large number of fundus pictures of the retina of the premature infant sent by the fundus picture acquisition module, and performs optimization processing on the aspect of image parameters by utilizing a space domain histogram enhancement method and a frequency domain homomorphic filtering enhancement method; the fundus picture target detection module performs target detection on a large number of fundus pictures of the premature infant retina after optimization processing by the fundus picture processing module, identifies and analyzes a target area of the fundus pictures, and obtains various pathological morphological characteristics in retinopathy of prematurity; the method is beneficial to improving various parameters of the fundus photo and improving the accuracy of a plurality of subsequent target detection submodels and integrated models; thereby being beneficial to improving the diagnosis reliability of the whole system;
2. the system is provided with a target detection submodel generation module and an integrated model generation module, wherein the target detection submodel generation module can also train various pathomorphism characteristics in retinopathy of prematurity by utilizing a deep learning algorithm to obtain a plurality of target detection submodels; the integrated model generation module extracts the quantitative features of various pathological structures of the multiple target detection submodels and constructs an integrated model according to the quantitative features of various pathological structures; the method is favorable for reducing the overall prediction deviation, improves the interpretability of a diagnosis result compared with the current mainstream single DNN black box model, gives a general conclusion, and can output quantitative evaluation of various pathological structures; and the complexity of algorithm diagnosis and iterative optimization is reduced; because each extracted sub-feature is transparent, local sub-models with poor performance can be conveniently positioned.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention.
Fig. 1 is a schematic diagram of the overall structure of a retinopathy of prematurity diagnosis system based on multi-focal target detection according to the present invention;
fig. 2 is a schematic view of the overall structure of a fundus picture according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
In the description of the present invention, it is to be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", and the like, indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be construed as limiting the present invention.
Referring to fig. 1-2, the system for diagnosing retinopathy of prematurity based on multi-focus target detection comprises an eyeground picture storage module, an eyeground picture acquisition module, an eyeground picture processing module, an eyeground picture target detection module, a target detection sub-model generation module, an integrated model generation module, a retina analysis and diagnosis module, a diagnosis report formation module and a visual display module;
the system comprises a fundus picture storage module, a fundus picture acquisition module, a fundus picture processing module, a fundus picture target detection module, a target detection sub-model generation module, an integrated model generation module, a retina analysis and diagnosis module, a diagnosis report formation module and a visual display module, wherein the fundus picture storage module, the fundus picture acquisition module, the fundus picture processing module, the fundus picture target detection module, the target detection sub-model generation module, the integrated model generation module, the retina analysis and; the visual display module is specifically a computer display screen.
The fundus picture storage module is used for storing a large number of fundus pictures of the retina of the premature infant; the fundus picture acquisition module is used for taking fundus pictures of a large number of retinas of premature infants in the fundus picture storage module and sending the fundus pictures to the fundus picture processing module.
The fundus picture processing module is used for receiving a large number of fundus pictures of the premature infant retina sent by the fundus picture acquisition module and carrying out optimization processing on the aspect of image parameters by utilizing a spatial domain histogram enhancement method and a frequency domain homomorphic filtering enhancement method; image parameters include sharpness, hue, gray scale, size, and noise.
The fundus picture target detection module is used for carrying out target detection on fundus pictures of a large number of premature infant retinas which are subjected to optimization processing by the fundus picture processing module, identifying and analyzing a target area of the fundus pictures, and obtaining various pathological morphological characteristics in retinopathy of prematurity; the target detection submodel generation module is used for training various pathological morphological characteristics in retinopathy of prematurity by utilizing a deep learning algorithm to obtain a plurality of target detection submodels; the specific generation process of the target detection sub-model generation module is as follows:
s1: extracting a large number of fundus pictures of the retina of the premature infant by using a fundus picture acquisition module, and processing the fundus pictures by using a fundus picture processing module and a fundus picture target detection module to obtain various pathomorphology feature sets;
s2: randomly distributing various pathological morphological feature sets to form a training set 1, a training set 2, … … and a training set n;
s3: constructing a plurality of classifiers, and respectively putting a training set 1, a training set 2, a training set … … and a training set n into the plurality of classifiers to obtain a target detection submodel 1, a target detection submodel 2, a target detection submodel … … and a target detection submodel n;
the deep learning algorithm is specifically a deep neural network.
The integrated model generation module is used for extracting quantitative characteristics of various pathological structures of the multiple target detection submodels and constructing an integrated model according to the quantitative characteristics of the various pathological structures; the integrated model generation module specifically generates the following processes:
SS 1: for each fundus picture, respectively predicting by a plurality of target detection submodels, and obtaining quantitative characteristics 1 of a pathological structure, 2 and … … of the pathological structure and n of the pathological structure according to the size, the range, the number and the like of focuses;
SS 2: constructing an integrated classifier, taking the quantitative feature 1 of the pathological structure, the quantitative features 2 and … … of the pathological structure and the quantitative feature n of the pathological structure as input data of the integrated classifier, marking a diagnosis result by a human expert, and training through an integrated learning strategy to obtain an integrated model.
The retina analysis and diagnosis module is used for receiving the quantitative characteristics of the new fundus picture and inputting the quantitative characteristics into the integrated model to obtain a diagnosis result, and the retina analysis and diagnosis module specifically analyzes and diagnoses the process as follows:
SSS 1: aiming at the new fundus picture, respectively predicting by each target detection sub-model to obtain quantitative characteristics of various pathological structures;
SSS 2: and inputting the extracted quantitative characteristics of various pathological structures into the integrated model to obtain a diagnosis result.
The diagnosis report forming module is used for carrying out clarification processing on the diagnosis result by using chart marking software to form a diagnosis report, and the chart type of the chart marking software comprises a column chart, a broken line chart, a pie chart, a bar chart and a three-line chart; the visual display module is used for visually displaying the diagnosis report and providing the diagnosis assistance for the medical staff.
The working principle and the using process of the invention are as follows: before the system is used, a plurality of target detection submodels and an integrated model need to be constructed; the specific construction process is as follows: firstly, a fundus picture acquisition module is used for taking a large number of fundus pictures of the retina of the premature infant in a fundus picture storage module and sending the fundus pictures to a fundus picture processing module; then the fundus picture processing module receives a large number of fundus pictures of the retina of the premature infant sent by the fundus picture acquisition module, and performs optimization processing on the aspect of image parameters by utilizing a space domain histogram enhancement method and a frequency domain homomorphic filtering enhancement method; then, the fundus picture target detection module performs target detection on a large number of fundus pictures of the premature infant retina after optimization processing by the fundus picture processing module, identifies and analyzes a target area of the fundus pictures, and obtains various pathological morphological characteristics in retinopathy of prematurity; then, the target detection submodel generation module trains various pathological morphological characteristics in retinopathy of prematurity by using a deep learning algorithm to obtain a plurality of target detection submodels; then, the integrated model generation module extracts the quantitative features of various pathological structures of the multiple target detection submodels and constructs an integrated model according to the quantitative features of the various pathological structures; after the construction of the plurality of target detection submodels and the integrated model is completed, the system can be used normally, and for a new fundus photo, the quantitative characteristics of various pathological structures are obtained by respectively predicting each target detection submodel; then inputting the extracted quantitative characteristics of various pathological structures into the integrated model to obtain a diagnosis result; the diagnosis report forming module can utilize chart marking software to carry out clarification processing on the diagnosis result to form a diagnosis report, and finally the visualization display module can carry out visualization display on the diagnosis report to provide diagnosis assistance for medical staff; the invention reduces the overall prediction deviation through the integrated learning strategy of a plurality of target detection submodels, improves the interpretability of the diagnosis result compared with the current mainstream single DNN black box model, and can output the quantitative evaluation of various pathological structures while giving a general conclusion.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (7)
1. The system for diagnosing retinopathy of prematurity based on multi-focus target detection is characterized by comprising an eyeground picture storage module, an eyeground picture acquisition module, an eyeground picture processing module, an eyeground picture target detection module, a target detection sub-model generation module, an integrated model generation module, a retina analysis and diagnosis module, a diagnosis report formation module and a visual display module;
the eye fundus picture storage module, the eye fundus picture acquisition module, the eye fundus picture processing module, the eye fundus picture target detection module, the target detection sub-model generation module, the integrated model generation module, the retina analysis and diagnosis module, the diagnosis report formation module and the visual display module are integrated in the computer; the visual display module is specifically a computer display screen.
2. The system for diagnosing retinopathy of prematurity based on multi-focal target detection of claim 1, wherein the fundus picture storage module is configured to store fundus pictures of a large number of retinas of premature infants; the fundus picture acquisition module is used for taking fundus pictures of a large number of retinas of premature infants in the fundus picture storage module and sending the fundus pictures to the fundus picture processing module.
3. The system for diagnosing retinopathy of prematurity based on multi-lesion target detection as claimed in claim 1, wherein the fundus picture processing module is used for receiving fundus pictures of a large number of retinas of premature infants sent by the fundus picture collecting module and performing optimization processing on image parameters by using a spatial histogram enhancement method and a frequency domain homomorphic filtering enhancement method; the image parameters include sharpness, hue, gray scale, size, and noise.
4. The system for diagnosing retinopathy of prematurity based on multi-lesion target detection as claimed in claim 1, wherein the fundus picture target detection module is used for performing target detection on fundus pictures of a large number of retinas of premature infants after optimized processing by the fundus picture processing module, identifying and analyzing target areas of the fundus pictures, and obtaining various types of pathomorphological characteristics in retinopathy of prematurity; the target detection submodel generation module is used for training various pathological morphological characteristics in retinopathy of prematurity by utilizing a deep learning algorithm to obtain a plurality of target detection submodels; the specific generation process of the target detection submodel generation module is as follows:
s1: extracting a large number of fundus pictures of the retina of the premature infant by using a fundus picture acquisition module, and processing the fundus pictures by using a fundus picture processing module and a fundus picture target detection module to obtain various pathomorphology feature sets;
s2: randomly distributing various pathological morphological feature sets to form a training set 1, a training set 2, … … and a training set n;
s3: constructing a plurality of classifiers, and respectively putting a training set 1, a training set 2, a training set … … and a training set n into the plurality of classifiers to obtain a target detection submodel 1, a target detection submodel 2, a target detection submodel … … and a target detection submodel n;
the deep learning algorithm is specifically a deep neural network.
5. The system for diagnosing retinopathy of prematurity based on multi-focal target detection of claim 1, wherein the integrated model generation module is configured to extract quantitative features of various pathological structures of a plurality of target detection submodels and construct an integrated model according to the quantitative features of various pathological structures; the integrated model generation module specifically generates the following processes:
SS 1: for each fundus picture, respectively predicting by a plurality of target detection submodels, and obtaining quantitative characteristics 1 of a pathological structure, 2 and … … of the pathological structure and n of the pathological structure according to the size, the range, the number and the like of focuses;
SS 2: constructing an integrated classifier, taking the quantitative feature 1 of the pathological structure, the quantitative features 2 and … … of the pathological structure and the quantitative feature n of the pathological structure as input data of the integrated classifier, marking a diagnosis result by a human expert, and training through an integrated learning strategy to obtain an integrated model.
6. The system for diagnosing retinopathy of prematurity based on multi-focal target detection of claim 1, wherein the retinal analysis and diagnosis module is used for receiving the quantified characteristics of the new fundus picture and inputting the quantified characteristics into the integrated model to obtain the diagnosis result, and the retinal analysis and diagnosis module specifically analyzes and diagnoses the process as follows:
SSS 1: aiming at the new fundus picture, respectively predicting by each target detection sub-model to obtain quantitative characteristics of various pathological structures;
SSS 2: and inputting the extracted quantitative characteristics of various pathological structures into the integrated model to obtain a diagnosis result.
7. The system for diagnosing retinopathy of prematurity based on multi-lesion target detection of claim 1, wherein the diagnosis report forming module is used for performing clarification processing on the diagnosis result by using chart labeling software to form a diagnosis report, and the chart types of the chart labeling software comprise a column chart, a broken line chart, a pie chart, a bar chart and a three-line chart;
the visual display module is used for visually displaying the diagnosis report and providing the diagnosis assistance for the medical staff.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110045855.6A CN112652394A (en) | 2021-01-14 | 2021-01-14 | Multi-focus target detection-based retinopathy of prematurity diagnosis system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110045855.6A CN112652394A (en) | 2021-01-14 | 2021-01-14 | Multi-focus target detection-based retinopathy of prematurity diagnosis system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112652394A true CN112652394A (en) | 2021-04-13 |
Family
ID=75368325
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110045855.6A Pending CN112652394A (en) | 2021-01-14 | 2021-01-14 | Multi-focus target detection-based retinopathy of prematurity diagnosis system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112652394A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107680684A (en) * | 2017-10-12 | 2018-02-09 | 百度在线网络技术（北京）有限公司 | For obtaining the method and device of information |
| CN109671049A (en) * | 2018-11-07 | 2019-04-23 | 哈尔滨工业大学（深圳） | A kind of medical image processing method, system, equipment, storage medium |
| CN109949302A (en) * | 2019-03-27 | 2019-06-28 | 天津工业大学 | Retinal feature Structural Techniques based on pixel |
| CN110010219A (en) * | 2019-03-13 | 2019-07-12 | 杭州电子科技大学 | Optical coherence tomography image retinopathy intelligent checking system and detection method |
| CN110163844A (en) * | 2019-04-17 | 2019-08-23 | 平安科技（深圳）有限公司 | Eyeground lesion detection method, device, computer equipment and storage medium |
| CN110648344A (en) * | 2019-09-12 | 2020-01-03 | 电子科技大学 | Diabetic retinopathy classification device based on local focus characteristics |
| CN111127425A (en) * | 2019-12-23 | 2020-05-08 | 北京至真互联网技术有限公司 | Target detection positioning method and device based on retina fundus image |
| CN111507932A (en) * | 2019-01-31 | 2020-08-07 | 福州依影健康科技有限公司 | High-specificity diabetic retinopathy characteristic detection method and storage equipment |
-
2021
- 2021-01-14 CN CN202110045855.6A patent/CN112652394A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107680684A (en) * | 2017-10-12 | 2018-02-09 | 百度在线网络技术（北京）有限公司 | For obtaining the method and device of information |
| CN109671049A (en) * | 2018-11-07 | 2019-04-23 | 哈尔滨工业大学（深圳） | A kind of medical image processing method, system, equipment, storage medium |
| CN111507932A (en) * | 2019-01-31 | 2020-08-07 | 福州依影健康科技有限公司 | High-specificity diabetic retinopathy characteristic detection method and storage equipment |
| CN110010219A (en) * | 2019-03-13 | 2019-07-12 | 杭州电子科技大学 | Optical coherence tomography image retinopathy intelligent checking system and detection method |
| CN109949302A (en) * | 2019-03-27 | 2019-06-28 | 天津工业大学 | Retinal feature Structural Techniques based on pixel |
| CN110163844A (en) * | 2019-04-17 | 2019-08-23 | 平安科技（深圳）有限公司 | Eyeground lesion detection method, device, computer equipment and storage medium |
| CN110648344A (en) * | 2019-09-12 | 2020-01-03 | 电子科技大学 | Diabetic retinopathy classification device based on local focus characteristics |
| CN111127425A (en) * | 2019-12-23 | 2020-05-08 | 北京至真互联网技术有限公司 | Target detection positioning method and device based on retina fundus image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108876775B (en) | Method for rapidly detecting diabetic retinopathy | |
| CN109472781B (en) | Diabetic retinopathy detection system based on serial structure segmentation | |
| CN109635862B (en) | Sorting method for retinopathy of prematurity plus lesion | |
| CN109858540B (en) | Medical image recognition system and method based on multi-mode fusion | |
| CN107609503A (en) | Intelligent cancerous tumor cell identifying system and method, cloud platform, server, computer | |
| CN111046835A (en) | Eyeground illumination multiple disease detection system based on regional feature set neural network | |
| CN107209933A (en) | For assessing retinal images and the method and system of information being obtained from retinal images | |
| CN110335266B (en) | Intelligent traditional Chinese medicine visual inspection image processing method and device | |
| US20210391056A1 (en) | Health big data service method and system based on remote fundus screening | |
| CN112233087A (en) | Artificial intelligence-based ophthalmic ultrasonic disease diagnosis method and system | |
| CN102567734B (en) | Specific value based retina thin blood vessel segmentation method | |
| CN109919938A (en) | The optic disk of glaucoma divides map acquisition methods | |
| CN110613483A (en) | Method and system for detecting fetal craniocerebral abnormality based on machine learning | |
| Hassan et al. | CDC-Net: Cascaded decoupled convolutional network for lesion-assisted detection and grading of retinopathy using optical coherence tomography (OCT) scans | |
| CN114343563A (en) | Method, device and system for assisting dry eye diagnosis and typing through multi-modal fusion | |
| CN114334124A (en) | Pathological myopia detection system based on deep neural network | |
| CN114332910A (en) | Human body part segmentation method for similar feature calculation of far infrared image | |
| CN108665474B (en) | B-COSFIRE-based retinal vessel segmentation method for fundus image | |
| Yadav et al. | Automatic Cataract Severity Detection and Grading Using Deep Learning | |
| CN116758038A (en) | Infant retina disease information identification method and system based on training network | |
| Triyadi et al. | Deep learning in image classification using vgg-19 and residual networks for cataract detection | |
| CN112652394A (en) | Multi-focus target detection-based retinopathy of prematurity diagnosis system | |
| CN115661101A (en) | Premature infant retinopathy detection system based on random sampling and deep learning | |
| Nasir et al. | A faster RCNN based diabetic retinopathy detection method using fused features from retina images | |
| CN112927808A (en) | Thyroid ultrasound image-based nodule grading system and method |
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 |