CN115100180A - Pneumonia feature identification method and device based on neural network model and electronic equipment - Google Patents
Pneumonia feature identification method and device based on neural network model and electronic equipment Download PDFInfo
- Publication number
- CN115100180A CN115100180A CN202210844283.2A CN202210844283A CN115100180A CN 115100180 A CN115100180 A CN 115100180A CN 202210844283 A CN202210844283 A CN 202210844283A CN 115100180 A CN115100180 A CN 115100180A
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- pneumonia
- image data
- feature extraction
- 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
- 238000003062 neural network model Methods 0.000 title claims abstract description 73
- 206010035664 Pneumonia Diseases 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000002601 radiography Methods 0.000 claims abstract description 9
- 238000002372 labelling Methods 0.000 claims abstract description 3
- 238000000605 extraction Methods 0.000 claims description 62
- 238000004590 computer program Methods 0.000 claims description 12
- 230000004927 fusion Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 abstract description 3
- 238000004364 calculation method Methods 0.000 description 8
- 238000003745 diagnosis Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000000717 retained effect Effects 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 238000007639 printing Methods 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 210000004072 lung Anatomy 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000002834 transmittance Methods 0.000 description 1
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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06T2207/10116—X-ray 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
- G06T2207/30061—Lung
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention relates to the technical field of deep learning, in particular to a pneumonia feature identification method and device based on a neural network model and electronic equipment, wherein the method comprises the following steps: acquiring chest radiography image data to be detected; inputting chest radiograph image data to be detected into a neural network model, wherein the neural network model is obtained by training a plurality of groups of data sets, and each group of data in the plurality of groups of data training sets comprises an image with pneumonia characteristics, characteristic position labeling information and a classification name; calculating a neural network model, copying the image data of the chest radiograph to be detected, and marking the recognition result of the neural network model on the copied image data, wherein the recognition result comprises a pneumonia feature position mark and a classification name mark; and outputting the image data of the chest film to be detected and the copied image data marked with the identification result. The invention improves the identification efficiency and avoids the influence of the identification result on the original film by simultaneously outputting the original film of the chest film to be detected and the copy image data of the marked identification result.
Description
Technical Field
The invention relates to the technical field of deep learning, in particular to a pneumonia feature identification method and device based on a neural network model and electronic equipment.
Background
The pneumonia image is also called as a chest film or a lung X-ray film, and is an image of a detected part of a human body obtained by processing data obtained by measurement through an electronic computer according to different absorption and transmittance of different tissues of the human body to X-rays; however, the efficiency of manually reading pneumonia features and then judging the state of illness is too low, so that the suspected people cannot be quickly identified;
in the related art, in order to improve the efficiency of identifying pneumonia features, a medical image processing technology is used, for example, a convolutional neural network is utilized, data training is performed on images with pneumonia features, the types of the pneumonia features are automatically learned, then, the types and positions of the pneumonia features are quickly labeled through deep learning of a data model, and reliable data support is provided for subsequent diagnosis;
however, the existing neural network model has the defects that on one hand, the calculation amount is large, the operation speed is relatively slow, and on the other hand, when the existing network model outputs the result, the result is directly marked on the original image, so that the mark shields the original image, and further observation of the doctor on the chest film is influenced.
The information disclosed in this background section is only for enhancement of understanding of the general background of the invention and should not be taken as an acknowledgement or any form of suggestion that this information forms the prior art that is already known to a person skilled in the art.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: a pneumonia feature identification method and device based on a neural network model and electronic equipment are provided to reduce the influence of chest radiography marks on doctor observation.
In order to achieve the purpose, the invention adopts the technical scheme that:
in a first aspect, the invention provides a pneumonia feature identification method based on a neural network model, which comprises the following steps:
acquiring chest radiography image data to be detected;
inputting the chest radiography image data to be detected into a neural network model, wherein the neural network model is obtained by training a plurality of groups of data sets, and each group of data in the plurality of groups of data training sets comprises an image with pneumonia characteristics and characteristic position marking information;
calculating the neural network model, copying the chest image data to be detected, and marking the recognition result of the neural network model on the copied image data, wherein the recognition result comprises a pneumonia characteristic position mark;
and outputting the image data of the chest film to be detected and the copied image data marked with the identification result.
Further, the neural network model comprises a trunk feature extraction network, an enhanced feature extraction network and a prediction network;
the main feature extraction network is used for extracting preliminary features and obtaining three effective feature layers;
the reinforced feature extraction network is connected with the trunk feature extraction network and is used for fusing the three effective feature layers;
and the prediction network is connected with the enhanced feature extraction network and is used for predicting and outputting the result of the fused effective feature layer.
Further, the construction method of the neural network model comprises the following steps:
replacing a backbone feature extraction network of YOLOv4 with a backbone network structure of Mobilenetv3 and extracting features to obtain three effective feature layers;
grafting a backbone network structure of Mobilenetv3 with an enhanced feature extraction network and a prediction network of YOLOv4 to realize fusion and prediction of the three effective feature layers.
Further, when the backbone network structure of Mobilenetv3 is grafted with the enhanced feature extraction network and the prediction network of YOLOv4, the SPP and PANet structures of YOLOv4 are retained to improve the feature extraction capability.
In a second aspect, the present invention provides a pneumonia feature recognition apparatus based on a neural network model, including:
the acquisition unit is used for acquiring the image data of the chest radiograph to be detected;
the input unit is used for inputting the chest radiography image data to be detected into a neural network model, the neural network model is obtained through training of a plurality of groups of data sets, and each group of data in the plurality of groups of data training sets comprises an image with pneumonia characteristics and characteristic position marking information;
the operation unit is used for operating the neural network model, copying the chest image data to be detected and marking the identification result of the neural network model on the copied image data, wherein the identification result comprises a pneumonia characteristic position mark;
and the output unit is used for outputting the image data of the chest piece to be detected and the copied image data marked with the identification result.
Further, in the input unit, the neural network model includes a trunk feature extraction network, an enhanced feature extraction network and a prediction network;
the main feature extraction network is used for extracting primary features and obtaining three effective feature layers;
the reinforced feature extraction network is connected with the trunk feature extraction network and is used for fusing the three effective feature layers;
and the prediction network is connected with the enhanced feature extraction network and is used for predicting and outputting the result of the fused effective feature layer.
Further, in the input unit, the method for constructing the neural network model includes the following steps:
replacing a backbone feature extraction network of YOLOv4 with a backbone network structure of Mobilenetv3 and extracting features to obtain three effective feature layers;
grafting a backbone network structure of Mobilenetv3 with an enhanced feature extraction network and a prediction network of YOLOv4 to realize fusion and prediction of the three effective feature layers.
Further, in the input unit, when the backbone network structure of Mobilenetv3 is grafted with the enhanced feature extraction network and the prediction network of YOLOv4, the SPP and PANet structures of YOLOv4 are retained to improve the feature extraction capability.
In a third aspect, the present invention provides an electronic device comprising:
a memory for storing a computer program;
a processor for implementing the method for pneumonia feature recognition based on neural network model according to any one of the first aspect when the computer program is executed.
In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the storage medium stores a computer program for executing the method for identifying characteristics of pneumonia based on a neural network model according to any one of the first aspect.
The invention has the beneficial effects that: the invention identifies the image data of the chest film to be detected through the trained neural network model, detects suspicious pneumonia characteristics, and simultaneously outputs the copy image data forms of the original film of the chest film to be detected and the mark identification result through copying the chest film and marking the position and the classification label on the copied chest film, thereby facilitating the doctor to simultaneously observe the original film and the copy film with the identification result, improving the analysis efficiency and simultaneously avoiding the influence of the identification result on the original film.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flowchart illustrating steps of a method for identifying pneumonia features based on a neural network model according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an output image data of a neural network model-based pneumonia feature identification method according to an embodiment of the present invention;
FIG. 3 is a network structure diagram of a YOLOv4 network structure model according to an embodiment of the present invention;
FIG. 4 is a network structure diagram of a YOLOv4-Tiny network structure model in an embodiment of the present invention;
FIG. 5 is a network structure diagram of an improved neural network model according to an embodiment of the present invention;
FIG. 6 is a flowchart illustrating steps of a neural network model construction method according to an embodiment of the present invention;
fig. 7 is a schematic diagram of a pneumonia feature recognition device based on a neural network in an embodiment of 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.
It will be understood that when an element is referred to as being "secured to" another element, it can be directly on the other element or intervening elements may also be present. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or intervening elements may also be present. The terms "vertical," "horizontal," "left," "right," and the like as used herein are for illustrative purposes only and do not represent the only embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
It should be noted here that the inventive concept of the present invention is to improve the identification efficiency of suspicious pneumonia features on a chest film, reduce the influence of the marks on the pneumonia chest film on diagnosis and treatment, and improve the operation speed of network nerves by improving the marks and the output mode and improving the network nerve model, so as to provide data support for the diagnosis of a doctor for subsequent diagnosis and treatment, and not to contain the description of a disease diagnosis method directly, and the output result is only limited for the doctor to see the chest film more efficiently, and whether the disease is affected or not needs to be acquired by depending on the professional knowledge and skills of the doctor, and the present invention will be described in detail in the following by way of embodiments;
the pneumonia feature identification method based on the neural network model as shown in fig. 1 comprises the following steps:
s10: acquiring chest radiography image data to be detected; the chest image data to be detected refers to an image which is shot and acquired by a suspected patient from an X-ray film instrument, and can be an electronic image before printing or a film image after printing;
s20: inputting chest radiograph image data to be detected into a neural network model, wherein the neural network model is obtained through training of a plurality of groups of data sets, and each group of data in the plurality of groups of data training sets comprises an image with pneumonia characteristics and characteristic position labeling information; the neural network model may have various forms, such as a YOLOv4 algorithm model, or a YOLOv4-Tiny algorithm model with a faster speed, which will be described in detail below;
s30: calculating a neural network model, copying the chest image data to be detected, and marking the recognition result of the neural network model on the copied image data, wherein the recognition result comprises a pneumonia characteristic position mark; as shown in fig. 2, the present invention is improved in that the original is not directly marked, but the original is copied and then the marking of the position is algorithmically recognized on the copied image; it should be noted here that the position mark has many forms, and may be marked by a rectangular box in fig. 2, or may be implemented by a form of a cross cursor, for example;
s40: and outputting the image data of the chest piece to be detected and the copied image data marked with the identification result. Referring to fig. 2, in the embodiment of the present invention, two image data are finally output at the same time, and the original image and the marked copy image are output at the same time, so that the diagnostician can directly observe and position the image; of course, it should be noted here that the specific output form may be that the original image is in the form of a film, and the duplicated image is in the form of paper, so as to save cost;
in the embodiment, the image data of the chest film to be detected is identified through the trained neural network model, suspected pneumonia features are detected, the chest film is copied, the position and classification label is marked on the copied chest film, and the copied image data of the original chest film to be detected and the marked recognition result are output simultaneously, so that a doctor can observe the original film and the copied film with the recognition result simultaneously, the analysis efficiency is improved, and the influence of the recognition result on the original film is avoided.
On the basis of the above embodiments, the following describes a neural network model used before the improvement of the embodiments of the present invention, as shown in fig. 3, taking a YOLOv4 network neural structure model as an example, the neural network model includes a trunk feature extraction network, an enhanced feature extraction network, and a prediction network;
the main feature extraction network is used for extracting the primary features and obtaining three effective feature layers; CSPDarknet53 in a backbone network part of YOLOv4 uses a Mish activation function to carry out high-precision calculation, so that the feature extraction is more accurate, but the calculation cost is higher; DarknetConv2D _ BN _ Mish is a common convolution which adopts BatchNorm2D data normalization processing and a Mish activation function, Resblock _ body is a residual error network structure, and two effective layers are output through the residual error network structure;
after convolving the last feature layer of CSPDarknet53 3 times in the feature pyramid part by using SPP, the processing is carried out by adopting maximum pooling of 4 different scales.
The reinforced feature extraction network is connected with the trunk feature extraction network and is used for fusing the three effective feature layers; in the PANET structure, the input tensors are spliced through Contact operation, so that the receptive field is effectively enhanced, and the characteristic separation of the context is more remarkable;
the prediction network is connected with the reinforced characteristic extraction network and used for predicting and outputting the result of the fused effective characteristic layer. In fig. 3, YOLO Head is a prediction network part, and a final output result is obtained by performing score sorting and non-maximum inhibition screening on prediction results;
however, the inventor finds that when the YOLOv4 network structure model is used, the operation cost is high, the operation time is long, the result waiting time is increased, and the efficiency is not improved; then, a more portable YOLOv4-Tiny network structure model is used, as shown in fig. 4, YOLOv4-Tiny is a lightweight version of YOLOv4, the operation speed of which is nearly 10 times of that of YOLOv4 under the same conditions, the network layer number is firstly shortened, the original SPP and PANet structures are secondly cancelled, instead, an FPN structure is adopted, the FPN convolves the effective feature layer of the last scale, performs upsampling, and then performs stacking and convolution with the effective feature layer of the last scale, but the precision of YOLOv4-Tiny is about forty percent, which is thirty percent lower than YOLOv4, and more detection leakage situations exist; therefore, the inventor achieves the precision requirement between the two and improves the running speed by improving the structure of the YOLOv4 neural network; therefore, the inventor introduces a Mobilenet model, which is a lighter deep neural network and developed by Google aiming at embedded platforms such as mobile phones, and discloses three versions at present, wherein the Mobilenetv3 introduces a deep separable convolution and a reverse residual error structure with linear bottleneck, and the calculation cost ratio of the reverse residual error structure is between YOLov4-Tiny and YOLov4, and based on the three versions, the inventor fuses Mobilenetv3 and YOLov4 to construct a network structure model which improves the operation speed on the premise of no great loss of calculation accuracy; as shown in fig. 5, a specific bneck structure in Mobilenetv3 is utilized in the trunk feature extraction network, which can reduce the parameter amount of the network on the premise of maximally retaining the target feature, thereby improving the running speed, and the activating function adopts HS and RE, so that the precision loss can be avoided as much as possible while the calculation cost is reduced; in addition, in the improved network structure model, please continue to refer to fig. 5, in the embodiment of the present invention, the SPP and PANet structures in YOLOv4 are retained, so that it can extract the semantic information of a deeper level; through the improvement on the network structure model, the calculation speed is higher than that of YOLOv4 by an empirical calculation by about nineteen percent, and the precision is improved by about twenty percent compared with that of YOLOv4-Tiny, so that an ideal effect is achieved;
specifically, in the embodiment of the present invention, as shown in fig. 6, the method for constructing an improved neural network model includes the following steps, and since the specific principle is described in detail above, the description is not repeated here:
s21: replacing a backbone feature extraction network of YOLOv4 with a backbone network structure of Mobilenetv3 and extracting features to obtain three effective feature layers;
s22: the trunk network structure of the Mobilenetv3 is grafted with the enhanced feature extraction network and the prediction network of the YOLOv4, so that fusion and prediction of three effective feature layers are realized.
In the embodiment of the invention, when the backbone network structure of the Mobilenetv3 is grafted with the enhanced feature extraction network and the prediction network of the YOLOv4, the SPP and PANET structures of the YOLOv4 are reserved, so that the feature extraction capability is improved. This part has already been described in detail above and will not be described again;
as will be understood by those skilled in the art, embodiments of the present invention may be provided as a method, an apparatus, a storage medium or a computer program product, so that embodiments of the present invention may fully adopt hardware embodiments, hardware and software combined embodiments or pure software embodiments, and based on the above content and the same inventive concept, the following describes a pneumonia feature recognition apparatus based on a neural network model in embodiments of the present invention, where the following embodiments of the apparatus correspond to the above embodiments of the method, and those skilled in the art can understand the following implementation processes based on the above description, and will not describe here in detail, as shown in fig. 7, and another aspect of the embodiments of the present invention also provides a pneumonia feature recognition apparatus based on a neural network model, including:
an obtaining unit 100, configured to obtain chest radiography image data to be detected;
the input unit 200 is used for inputting the chest radiograph image data to be tested into a neural network model, the neural network model is obtained by training a plurality of groups of data sets, and each group of data in the plurality of groups of data training sets comprises an image with pneumonia characteristics and characteristic position marking information;
the operation unit 300 is used for operating the neural network model, copying the chest image data to be detected, and marking the recognition result of the neural network model on the copied image data, wherein the recognition result comprises a pneumonia characteristic position mark;
and an output unit 400 for outputting the image data of the chest piece to be measured and the copied image data marked with the identification result.
In the embodiment of the present invention, in the input unit 200, the neural network model includes a trunk feature extraction network, an enhanced feature extraction network, and a prediction network;
the main feature extraction network is used for extracting the primary features and obtaining three effective feature layers;
the reinforced feature extraction network is connected with the trunk feature extraction network and is used for fusing the three effective feature layers;
the prediction network is connected with the reinforced feature extraction network and used for predicting and outputting the result of the fused effective feature layer.
In the embodiment of the present invention, in the input unit 200, the method for constructing the neural network model includes the following steps:
replacing a backbone feature extraction network of YOLOv4 with a backbone network structure of Mobilenetv3 and extracting features to obtain three effective feature layers;
the trunk network structure of the Mobilenetv3 is grafted with the enhanced feature extraction network and the prediction network of the YOLOv4, so that fusion and prediction of three effective feature layers are realized.
In the embodiment of the present invention, in the input unit 200, when the backbone network structure of Mobilenetv3 is grafted with the enhanced feature extraction network and the prediction network of YOLOv4, the SPP and PANet structures of YOLOv4 are retained to improve the feature extraction capability.
In another aspect of the embodiments of the present invention, there is also provided an electronic device, including:
a memory for storing a computer program;
a processor, configured to implement any one of the above methods for pneumonia feature identification based on a neural network model when executing a computer program.
In another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, in which a computer program is stored, the computer program being configured to execute any one of the above pneumonia feature identification method based on a neural network model.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (10)
1. A pneumonia feature identification method based on a neural network model is characterized by comprising the following steps:
acquiring chest radiography image data to be detected;
inputting the chest radiograph image data to be detected into a neural network model, wherein the neural network model is obtained by training a plurality of groups of data sets, and each group of data in the plurality of groups of data training sets comprises an image with pneumonia characteristics and characteristic position labeling information;
calculating the neural network model, copying the chest image data to be detected, and marking the recognition result of the neural network model on the copied image data, wherein the recognition result comprises a pneumonia characteristic position mark;
and outputting the image data of the chest film to be detected and the copied image data marked with the identification result.
2. The pneumonia feature identification method based on the neural network model is characterized in that the neural network model comprises a trunk feature extraction network, an enhanced feature extraction network and a prediction network;
the main feature extraction network is used for extracting primary features and obtaining three effective feature layers;
the reinforced feature extraction network is connected with the trunk feature extraction network and is used for fusing the three effective feature layers;
and the prediction network is connected with the enhanced feature extraction network and is used for predicting and outputting the result of the fused effective feature layer.
3. The pneumonia feature recognition method based on the neural network model is characterized in that the construction method of the neural network model comprises the following steps:
replacing a backbone feature extraction network of YOLOv4 with a backbone network structure of Mobilenetv3 and extracting features to obtain three effective feature layers;
grafting a backbone network structure of Mobilenetv3 with an enhanced feature extraction network and a prediction network of YOLOv4 to realize fusion and prediction of the three effective feature layers.
4. The pneumonia feature recognition method based on neural network model according to claim 3, characterized in that when grafting the backbone network structure of Mobilenetv3 with the enhanced feature extraction network and the prediction network of YOLov4, the SPP and PANET structures of YOLov4 are preserved to improve the feature extraction capability.
5. A pneumonia feature recognition device based on a neural network model is characterized by comprising:
the acquisition unit is used for acquiring the image data of the chest radiograph to be detected;
the input unit is used for inputting the chest radiography image data to be detected into a neural network model, the neural network model is obtained through training of a plurality of groups of data sets, and each group of data in the plurality of groups of data training sets comprises an image with pneumonia characteristics and characteristic position marking information;
the operation unit is used for operating the neural network model, copying the chest image data to be detected and marking the identification result of the neural network model on the copied image data, wherein the identification result comprises a pneumonia characteristic position mark;
and the output unit is used for outputting the image data of the chest piece to be detected and the copied image data marked with the identification result.
6. The pneumonia feature recognition device based on the neural network model according to claim 5, wherein in the input unit, the neural network model comprises a trunk feature extraction network, an enhanced feature extraction network and a prediction network;
the main feature extraction network is used for extracting preliminary features and obtaining three effective feature layers;
the reinforced feature extraction network is connected with the trunk feature extraction network and is used for fusing the three effective feature layers;
and the prediction network is connected with the enhanced feature extraction network and is used for predicting and outputting the result of the fused effective feature layer.
7. The pneumonia feature recognition device based on the neural network model according to claim 6, wherein in the input unit, the method for constructing the neural network model comprises the following steps:
replacing a backbone feature extraction network of YOLOv4 with a backbone network structure of Mobilenetv3 and extracting features to obtain three effective feature layers;
grafting a backbone network structure of Mobilenetv3 with an enhanced feature extraction network and a prediction network of YOLOv4 to realize fusion and prediction of the three effective feature layers.
8. The pneumonia feature recognition device based on neural network model according to claim 7, characterized in that in the input unit, when the trunk network structure of Mobilenetv3 is grafted with the enhanced feature extraction network and the prediction network of YOLOv4, the SPP and PANet structures of YOLOv4 are preserved to improve the feature extraction capability.
9. An electronic device, comprising:
a memory for storing a computer program;
a processor for implementing the neural network model-based pneumonia feature identification method according to any one of claims 1 to 4 when executing the computer program.
10. A computer-readable storage medium, wherein a computer program is stored in the storage medium, and the computer program is used for executing the method for identifying characteristics of pneumonia based on neural network model according to any one of claims 1 to 4.
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210844283.2A CN115100180A (en) | 2022-07-18 | 2022-07-18 | Pneumonia feature identification method and device based on neural network model and electronic equipment |
| PCT/CN2023/100944 WO2023186190A1 (en) | 2022-07-18 | 2023-06-19 | Neural network model-based pneumonia feature recognition method and apparatus, and electronic device |
| ZA2023/09547A ZA202309547B (en) | 2022-07-18 | 2023-10-12 | Neural network model-based pneumonia feature recognition method and apparatus, and electronic device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210844283.2A CN115100180A (en) | 2022-07-18 | 2022-07-18 | Pneumonia feature identification method and device based on neural network model and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115100180A true CN115100180A (en) | 2022-09-23 |
Family
ID=83298753
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210844283.2A Pending CN115100180A (en) | 2022-07-18 | 2022-07-18 | Pneumonia feature identification method and device based on neural network model and electronic equipment |
Country Status (3)
| Country | Link |
|---|---|
| CN (1) | CN115100180A (en) |
| WO (1) | WO2023186190A1 (en) |
| ZA (1) | ZA202309547B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023186190A1 (en) * | 2022-07-18 | 2023-10-05 | 华北理工大学 | Neural network model-based pneumonia feature recognition method and apparatus, and electronic device |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118366156B (en) * | 2024-06-20 | 2024-09-24 | 中国人民解放军国防科技大学 | Automatic focusing method, device and computer equipment for sparse sample microscopic imaging |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113222984A (en) * | 2021-06-03 | 2021-08-06 | 河海大学 | X-ray film new coronary pneumonia identification and detection method based on fine-grained image identification and target detection algorithm |
| CN113516655A (en) * | 2021-09-13 | 2021-10-19 | 常州市宏发纵横新材料科技股份有限公司 | Cloth cover defect detection method, device and system based on YOLOv4-Tiny neural network |
| CN113537244A (en) * | 2021-07-23 | 2021-10-22 | 深圳职业技术学院 | Livestock image target detection method and device based on light-weight YOLOv4 |
| CN113762265A (en) * | 2021-08-27 | 2021-12-07 | 慧影医疗科技（北京）有限公司 | Pneumonia classification and segmentation method and system |
| CN114120019A (en) * | 2021-11-08 | 2022-03-01 | 贵州大学 | Lightweight target detection method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10936912B2 (en) * | 2018-11-01 | 2021-03-02 | International Business Machines Corporation | Image classification using a mask image and neural networks |
| CN115100180A (en) * | 2022-07-18 | 2022-09-23 | 华北理工大学 | Pneumonia feature identification method and device based on neural network model and electronic equipment |
-
2022
- 2022-07-18 CN CN202210844283.2A patent/CN115100180A/en active Pending
-
2023
- 2023-06-19 WO PCT/CN2023/100944 patent/WO2023186190A1/en unknown
- 2023-10-12 ZA ZA2023/09547A patent/ZA202309547B/en unknown
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113222984A (en) * | 2021-06-03 | 2021-08-06 | 河海大学 | X-ray film new coronary pneumonia identification and detection method based on fine-grained image identification and target detection algorithm |
| CN113537244A (en) * | 2021-07-23 | 2021-10-22 | 深圳职业技术学院 | Livestock image target detection method and device based on light-weight YOLOv4 |
| CN113762265A (en) * | 2021-08-27 | 2021-12-07 | 慧影医疗科技（北京）有限公司 | Pneumonia classification and segmentation method and system |
| CN113516655A (en) * | 2021-09-13 | 2021-10-19 | 常州市宏发纵横新材料科技股份有限公司 | Cloth cover defect detection method, device and system based on YOLOv4-Tiny neural network |
| CN114120019A (en) * | 2021-11-08 | 2022-03-01 | 贵州大学 | Lightweight target detection method |
Non-Patent Citations (1)
| Title |
|---|
| 曹盼等: "特征金字塔网络在胸部X线摄影图像上筛检肺结核的价值", 《中国防痨杂志》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023186190A1 (en) * | 2022-07-18 | 2023-10-05 | 华北理工大学 | Neural network model-based pneumonia feature recognition method and apparatus, and electronic device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2023186190A1 (en) | 2023-10-05 |
| ZA202309547B (en) | 2023-11-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111259982B (en) | Attention mechanism-based premature infant retina image classification method and device | |
| CN115100180A (en) | Pneumonia feature identification method and device based on neural network model and electronic equipment | |
| CN106056595A (en) | Method for automatically identifying whether thyroid nodule is benign or malignant based on deep convolutional neural network | |
| CN109544518B (en) | Method and system applied to bone maturity assessment | |
| CN109902584A (en) | A kind of recognition methods, device, equipment and the storage medium of mask defect | |
| CN111754453A (en) | Pulmonary tuberculosis detection method and system based on chest radiography image and storage medium | |
| CN113793301B (en) | Training method of fundus image analysis model based on dense convolution network model | |
| CN113012163A (en) | Retina blood vessel segmentation method, equipment and storage medium based on multi-scale attention network | |
| CN109685765A (en) | A kind of X-ray pneumonia prediction of result device based on convolutional neural networks | |
| CN116188879B (en) | Image classification and image classification model training method, device, equipment and medium | |
| CN113205524A (en) | Blood vessel image segmentation method, device and equipment based on U-Net | |
| CN110610488A (en) | Classification training and detecting method and device | |
| CN113237881A (en) | Method and device for detecting specific cells and pathological section detection system | |
| CN113554640A (en) | AI model training method, use method, computer device and storage medium | |
| CN108985302A (en) | A kind of skin lens image processing method, device and equipment | |
| CN111047582A (en) | Crohn's disease auxiliary diagnosis system under enteroscope based on degree of depth learning | |
| CN114092427B (en) | Crohn's disease and intestinal tuberculosis classification method based on multi-sequence MRI image | |
| CN116091446A (en) | Method, system, medium and equipment for detecting abnormality of esophageal endoscope image | |
| WO2022227193A1 (en) | Liver region segmentation method and apparatus, and electronic device and storage medium | |
| CN114649092A (en) | Auxiliary diagnosis method and device based on semi-supervised learning and multi-scale feature fusion | |
| CN114359133A (en) | Hand bone image analysis method based on feature extraction and related equipment | |
| CN110570417B (en) | Pulmonary nodule classification device and image processing equipment | |
| Ogundokun et al. | Application of pre-trained CNN methods to identify COVID-19 pneumonia from Chest X-Ray | |
| CN112991266A (en) | Semantic segmentation method and system for small sample medical image | |
| CN115359060B (en) | Focus example segmentation method and device for pneumonia CT image |
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 |