CN114913169B - Neonatal necrotizing enterocolitis screening system - Google Patents
Neonatal necrotizing enterocolitis screening system Download PDFInfo
- Publication number
- CN114913169B CN114913169B CN202210655628.XA CN202210655628A CN114913169B CN 114913169 B CN114913169 B CN 114913169B CN 202210655628 A CN202210655628 A CN 202210655628A CN 114913169 B CN114913169 B CN 114913169B
- Authority
- CN
- China
- Prior art keywords
- necrotizing enterocolitis
- screening
- model
- convolution
- neonatal
- 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
- 206010051606 Necrotising colitis Diseases 0.000 title claims abstract description 56
- 208000004995 necrotizing enterocolitis Diseases 0.000 title claims abstract description 56
- 201000006195 perinatal necrotizing enterocolitis Diseases 0.000 title claims abstract description 56
- 238000012216 screening Methods 0.000 title claims abstract description 41
- 230000003187 abdominal effect Effects 0.000 claims abstract description 15
- 230000004913 activation Effects 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 10
- 238000004590 computer program Methods 0.000 claims abstract description 7
- 230000007246 mechanism Effects 0.000 claims abstract description 6
- 238000003062 neural network model Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 26
- 230000006870 function Effects 0.000 claims description 15
- 238000000034 method Methods 0.000 claims description 14
- 210000001015 abdomen Anatomy 0.000 claims description 11
- 238000012795 verification Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 230000005855 radiation Effects 0.000 claims description 6
- 230000007704 transition Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 210000003127 knee Anatomy 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 210000003689 pubic bone Anatomy 0.000 claims description 3
- 210000004061 pubic symphysis Anatomy 0.000 claims description 3
- 210000002417 xiphoid bone Anatomy 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 238000010200 validation analysis Methods 0.000 claims description 2
- 238000003745 diagnosis Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 210000004185 liver Anatomy 0.000 description 3
- 208000004930 Fatty Liver Diseases 0.000 description 2
- 206010019708 Hepatic steatosis Diseases 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 239000008280 blood Substances 0.000 description 2
- 210000004369 blood Anatomy 0.000 description 2
- 230000034994 death Effects 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 208000010706 fatty liver disease Diseases 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 231100000240 steatosis hepatitis Toxicity 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 241000426499 Chilo Species 0.000 description 1
- 206010017982 Gastrointestinal necrosis Diseases 0.000 description 1
- 206010061218 Inflammation Diseases 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000000090 biomarker Substances 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000004054 inflammatory process Effects 0.000 description 1
- 208000028774 intestinal disease Diseases 0.000 description 1
- 230000000968 intestinal effect Effects 0.000 description 1
- 208000037817 intestinal injury Diseases 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 208000018773 low birth weight Diseases 0.000 description 1
- 231100000533 low birth weight Toxicity 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 208000037890 multiple organ injury Diseases 0.000 description 1
- 230000017074 necrotic cell death Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 230000007170 pathology Effects 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 238000002604 ultrasonography 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
- A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
-
- 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
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2503/00—Evaluating a particular growth phase or type of persons or animals
- A61B2503/04—Babies, e.g. for SIDS detection
- A61B2503/045—Newborns, e.g. premature baby monitoring
-
- 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
- G06T2207/30028—Colon; Small intestine
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Radiology & Medical Imaging (AREA)
- Evolutionary Computation (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Animal Behavior & Ethology (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Optics & Photonics (AREA)
- Pathology (AREA)
- General Engineering & Computer Science (AREA)
- Heart & Thoracic Surgery (AREA)
- Surgery (AREA)
- High Energy & Nuclear Physics (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention discloses a neonatal necrotizing enterocolitis screening system, which comprises a computer memory, a computer processor and a computer program stored in the computer memory and executable on the computer processor, wherein a trained neonatal necrotizing enterocolitis screening model is stored in the computer memory; the screening model of the neonatal necrotizing enterocolitis is based on an improved DenseNet neural network model, and a PReLU activation function and an ECA attention mechanism are introduced on the basis of an original DenseNet framework; the computer processor, when executing the computer program, performs the steps of: preprocessing abdominal plain data to be screened, and inputting the trained neonatal necrotizing enterocolitis screening model to obtain a screening result of whether necrotizing enterocolitis exists. The invention can improve the diagnosis accuracy of necrotizing enterocolitis of newborn, thereby providing high-quality medical assistance to the sick children as soon as possible.
Description
Technical Field
The invention belongs to the field of medical artificial intelligence, and particularly relates to a neonatal necrotizing enterocolitis screening system.
Background
Necrotizing Enterocolitis (NEC) is a common, potentially catastrophic, intestinal disease in very low birth weight premature infants. NEC affects up to 15% of newborns with birth weights below 1500 grams, and causes sudden episodes of progressive intestinal inflammation and necrosis, which can lead to severe intestinal necrosis, multiple organ injury, or death. The unified mechanism of treatment of NEC has not been established, nor has any reliable biomarkers indicated the risk of an individual patient to suffer from the disease. Due to the inability to predict NEC in advance, current medical strategies include close clinical monitoring to treat NEC infants as soon as possible before unrecoverable intestinal injury occurs. The current clinical examination includes five types of biochemistry, blood routine, blood and qi, abdominal plain film, B ultrasonic and the like. The abdominal plain film is the detection mode with the highest accuracy.
Multiple studies show that the convolutional neural network is widely applied to medical science, particularly to classification of medical images, and obtains a good classification result.
For example, chinese patent publication No. CN109620152A discloses an electrocardiosignal classifying method based on MutiFacolLoss-densenert, because the time duration of the electrocardiosignal measured by each record is different or too long, it cannot be directly classified, it is necessary to divide each record at equal time intervals, then normalize the divided electrocardiosignal segments, and finally input the processed electrocardiosignal into a convolutional neural network for classification. The backbone network of the method mainly adopts the thought of a DenseNet structure, and the thought has the advantages of high classification recognition rate, parameter quantity and the like. The input of the backbone network is segmented electrocardiosignal segments, the output is the number of electrocardiosignal categories, and each output of the network is the probability of the category to which the output belongs, so the method is an end-to-end electrocardiosignal classification method.
For example, chinese patent publication No. CN114067092A discloses a method for classifying B-mode ultrasonic images of fatty liver based on DenseNet and lightGBM, which relates to the field of medical image processing, and obtains a medical image data set of liver, and preprocesses the obtained liver image data set; constructing a DenseNet201 network, inputting a preprocessed image data set, extracting an image characteristic matrix, dividing the extracted characteristic matrix into a training set, a verification set and a test set according to a set proportion; training a lightGBM model by using a training set, evaluating the trained model by using a verification set, and adjusting model parameters to improve the accuracy of model classification; and inputting the test set into the adjusted lightGBM model to obtain a liver classification result and test the generalization capability of the model. The histogram normalization and USM sharpening enhancement algorithm are adopted to optimize the classification characteristics of the fatty liver B-mode ultrasound image, and the performance is more excellent
Many scholars apply convolutional neural networks to image-based disease classification. However, aiming at the research field of the invention, namely the primary screening of necrotizing enterocolitis of neonate based on abdominal plain, no relevant research and report exists, and the existing network model is not suitable for the screening of necrotizing enterocolitis of neonate.
Disclosure of Invention
The invention provides a neonatal necrotizing enterocolitis screening system which can improve the diagnosis accuracy of neonatal necrotizing enterocolitis, thereby providing high-quality medical assistance to a sick child as soon as possible, and having important significance for relieving symptoms of the sick child and reducing the death rate.
A neonatal necrotizing enterocolitis screening system comprising computer memory, a computer processor, and a computer program stored in and executable on the computer memory, the computer memory having a trained neonatal necrotizing enterocolitis screening model stored therein;
the neonatal necrotizing enterocolitis screening model is based on an improved Densenet neural network model, and a PReLU activation function and an ECA attention mechanism are introduced on the basis of an original Densenet framework;
the computer processor, when executing the computer program, performs the steps of:
preprocessing abdominal plain data to be screened, and inputting the trained neonatal necrotizing enterocolitis screening model to obtain a screening result of whether necrotizing enterocolitis exists.
Further, the specific structure of the neonatal necrotizing enterocolitis screening model is as follows:
in the framework of the original DenseNet, each feature extraction layer DenseLayer is modified, and the modified DenseLayer includes: batch Norm layer, PReLU activation function, 1x1 convolution, 3x3 convolution and ECA attention module, using the construction form of bottleeck; the 1x1 convolution is used for reducing the dimension and inputting the feature map, the 3x3 convolution is used for extracting the features, and the ECA attention module is used for obtaining an enhanced feature map;
the original input feature map is sequentially subjected to a Batch Norm layer, a PReLU activation function, a 1x1 convolution, a Batch Norm layer, a PReLU activation function and a 3x3 convolution, the output feature map enters an ECA attention module, attention weights distributed to each channel are calculated according to the input feature map, the attention weights are used for performing point multiplication on the feature map to obtain feature weight calibration, and finally the original input feature map and the reinforced feature map are combined and spliced in channel dimensions to obtain a final feature map.
The framework of the original DenseNet is formed by cross combination of a DenseBlock module with a feature extraction function and a Transition module with a down-sampling function; the DenseBlock module comprises a plurality of DenseLayer modules connected in series, and the Transition module is composed of an average pooling layer and a convolution layer.
The neonatal necrotizing enterocolitis screening model training process is as follows:
(1) Collecting data of abdominal plain tablets of children suffering from necrotizing enterocolitis of the newborn;
(2) Preprocessing the abdominal plain film data, including image enhancement, format conversion and size unification;
(3) Dividing the preprocessed data into a training set and a verification set according to a proportion;
(4) And (3) sending the training set into the constructed neonatal necrotizing enterocolitis screening model for training, evaluating the performance of the classification model by using the verification set, adjusting the hyper-parameters of the model according to the evaluation effect, and finally finishing model training.
In the step (1), collected data of the abdomen plain film of the infant with necrotizing enterocolitis neonatorum meet the following standards:
the infant lies on the back on the photographic table, lifts the arms of the infant and clamps the head and the hands of the infant; the head, shoulders and knees of the infant patient are tightly attached to the photographic table, and the middle surface of the body is aligned with the central line of the table top, so that the body position is ensured not to be skewed; the central line is aligned with the midpoint of the joint connecting line of the xiphoid process and the pubis and vertically shot into the detector; the field of projection is minimized, with the upper margin including the diaphragm surface and the lower margin including the pubic symphysis, to reduce scattered radiation and radiation dose to the neonate.
In the step (2), the specific process of preprocessing the abdomen plain film data is as follows:
firstly, carrying out image enhancement on original abdomen plain film data by using a CLAHE algorithm so as to highlight key image characteristics of the abdomen plain film; the image size is unified into 224 × 224 size, and the image format is converted into jpg format by DICOM.
In the step (3), the preprocessed data are divided into a training set and a verification set according to the proportion of 8.
In the step (4), in the training process, 64 image data columns are arranged as a group for model training, and random left-right or horizontal rotation of 25 degrees is adopted to achieve the effect of sample size expansion.
Compared with the prior art, the invention has the following beneficial effects:
the system provided by the invention is used for innovatively constructing an improved DenseNet neural network algorithm to construct a newborn necrotizing enterocolitis prescreening model, the model introduces a PReLU activation function and uses an ECA attention mechanism on the basis of the original framework of the DenseNet neural network, and the prescreening judgment can be carried out on whether the newborn suffers from necrotizing enterocolitis or not only through an abdominal plain film, so that the diagnosis accuracy and efficiency of primary doctors are greatly improved.
Drawings
FIG. 1 is a flow chart of the overall structure of a neonatal necrotizing enterocolitis screening system according to the present invention;
fig. 2 is a thermodynamic diagram for lesion visualization in an embodiment of the invention.
Detailed Description
The invention will be described in further detail below with reference to the drawings and examples, which are intended to facilitate the understanding of the invention without limiting it in any way.
A neonatal necrotizing enterocolitis screening system includes a computer memory having a trained neonatal necrotizing enterocolitis screening model stored therein, a computer processor, and a computer program stored in the computer memory and executable on the computer processor.
The screening model for necrotizing enterocolitis neonatorum constructed by the invention is based on an improved DenseNet neural network model, and introduces a PReLU activation function and an ECA attention mechanism on the basis of an original DenseNet framework, so that the classification performance of the model is effectively improved by the improved model.
As shown in FIG. 1, the DenseNet algorithm model is mainly formed by the cross combination of a DenseBlock module with a feature extraction function and a Transition module with a down-sampling function, wherein the DenseBlock module comprises a plurality of DenseLayers connected in series, and the Transition module is formed by an average pooling layer and a convolution layer.
Neonatal necrotizing enterocolitis screening model each DenseLayer (feature extraction layer) was modified in the original DenseNet, the modified DenseLayer consisting of the following components, including: the method comprises the following steps of Batch Norm layer, PReLU activation function, 1x1 convolution, 3x3 convolution and ECA module, and a structured form of bottleeck is used, wherein the 1x1 convolution is located before the 3x3 convolution, the 1x1 convolution mainly plays a role in reducing dimension input feature maps, and meanwhile, the 3x3 convolution mainly completes a feature extraction task. The ECA attention module is introduced to obtain an enhanced characteristic diagram, and the specific implementation mode is as follows: the input data of the ECA attention module is a feature map of the final layer 3x3 convolution output, the attention weight distributed to each channel can be calculated according to the input feature map, the attention weight is used for carrying out point multiplication on the feature map to obtain feature weight calibration, and finally the original input feature map and the reinforced feature map are combined and spliced in the channel dimension to obtain a final feature map.
The training process of the neonatal necrotizing enterocolitis screening model is as follows:
(1) Abdominal plain data were collected from neonates with necrotizing enterocolitis. The collection of the image data of the abdominal plain film meets the following standards:
the infant is supine on the photographic table, a parent is instructed to lift the two arms of the infant, and the head and the hands of the infant are clamped tightly; the head, shoulders and knees of the infant patient are tightly attached to the photographic table, and the middle surface of the body is aligned with the central line of the table top, so that the body position is ensured not to be skewed; the central line is aligned with the midpoint of the joint connecting line of the xiphoid process and the pubis and vertically shot into the detector; the projection field is minimized, the upper edge includes the diaphragm surface, the lower edge includes the pubic symphysis, thereby reducing scattered radiation and radiation dose to the newborn.
(2) And preprocessing the abdominal plain film data, including image enhancement, format conversion and size unification.
The collected image data is subjected to data preprocessing, firstly, the original abdomen plain film data is subjected to image preprocessing by using a CLAHE algorithm to highlight key image features of the abdomen plain film, and the purpose of the step is that external conditions cause the abdomen plain film data to have blurred edges, contrast, uneven brightness or be too low and too high, such as: the activity when the neonate shoots the image, the sampling precision of equipment, the light intensity, the pathological characteristics of the patient, the level of shooting personnel and the like. The quality of the collected data directly influences the learning efficiency of the model, and meanwhile, in order to register images, the sizes of the images are unified into 224 × 224 sizes, and the image format is converted into a jpg format from DICOM.
(3) The preprocessed data were randomly divided into training and validation sets in a ratio of 8.
(4) And sending the training set into the constructed neonatal necrotizing enterocolitis screening model for training, evaluating the performance of the classification model by using the verification set, adjusting the hyper-parameters of the model according to the evaluation effect, and finally finishing model training.
During training, 64 image data are listed as a group for model training, and random left-right or horizontal rotation of 25 degrees is adopted to achieve the effect of enlarging the sample size.
In order to verify the effect of the invention, multiple indexes are adopted to verify the advancement of the model, including the specificity of 0.92, the sensitivity of 0.8, the accuracy of 0.85, the recall rate of 0.87, the F1 score of 0.98, the AUC value of 91.6 and the ROC curve of the test data set which are frequently appeared in the image classification task. Because NEC diagnosis is difficult, the experimental results show that the system reaches the diagnosis level of middle-level doctors.
Furthermore, a plurality of algorithm models are adopted for performance comparison, including VGGNet, alexNet and RestNet, and the model has advancement in specificity, sensitivity, accuracy, recall rate, F1 score and AUC value.
The performance is visualized, and whether feature selection is concentrated on the position on clinical pathology or not in the neonatal abdomen plain screening process is evaluated through feature thermodynamic diagram mapping of a Grad-CAM model so as to enhance the interpretability of the algorithm model, the deeper the color of the thermodynamic diagram is, the larger the contribution value is represented, and the heat diagram is well matched with the focus part through the verification of doctors with the radiology work experience of a Hospital department at Chilo level of 8 years, as shown in the focus visualization thermodynamic diagram of fig. 2.
The embodiments described above are intended to illustrate the technical solutions and advantages of the present invention, and it should be understood that the above-mentioned embodiments are only specific embodiments of the present invention, and are not intended to limit the present invention, and any modifications, additions and equivalents made within the scope of the principles of the present invention should be included in the scope of the present invention.
Claims (7)
1. A neonatal necrotizing enterocolitis screening system comprising computer memory, a computer processor, and a computer program stored in the computer memory and executable on the computer processor, characterized by: the trained neonatal necrotizing enterocolitis screening model is stored in the computer memory;
the screening model of the neonatal necrotizing enterocolitis is based on an improved DenseNet neural network model, and a PReLU activation function and an ECA attention mechanism are introduced on the basis of an original DenseNet framework; the specific structure of the neonatal necrotizing enterocolitis screening model is as follows:
in the framework of the original DenseNet, each feature extraction layer DenseLayer is modified, and the modified DenseLayer includes: batch Norm layer, PReLU activation function, 1x1 convolution, 3x3 convolution and ECA attention module, using the construction form of bottleeck; the 1x1 convolution is used for reducing the dimension and inputting the feature map, the 3x3 convolution is used for extracting the features, and the ECA attention module is used for obtaining an enhanced feature map;
the method comprises the steps that an original input feature map is subjected to Batch Norm layer convolution, PReLU activation function convolution, 1x1 convolution, batch Norm layer convolution, PReLU activation function convolution and 3x3 convolution in sequence, an output feature map enters an ECA attention module, attention weights distributed to all channels are calculated according to the input feature map, the attention weights are used for carrying out point multiplication on the feature map to obtain feature weight calibration, and finally the original input feature map and an enhanced feature map are combined and spliced in channel dimensions to obtain a final feature map;
the computer processor, when executing the computer program, performs the steps of:
preprocessing abdominal plain data to be screened, and inputting the trained neonatal necrotizing enterocolitis screening model to obtain a screening result of whether necrotizing enterocolitis exists.
2. The neonatal necrotizing enterocolitis screening system of claim 1, wherein the framework of the original DenseNet is formed by cross-combining a DenseBlock module with feature extraction and a Transition module with down-sampling; the DenseBlock module comprises a plurality of DenseLayers connected in series, and the Transition module is composed of an average pooling layer and a convolution layer.
3. The system for screening of neonatal necrotizing enterocolitis according to claim 1, wherein the training process of the neonatal necrotizing enterocolitis screening model is as follows:
(1) Collecting data of abdominal plain tablets of children suffering from necrotizing enterocolitis of the newborn;
(2) Preprocessing the abdomen plain film data, including image enhancement, format conversion and size unification;
(3) Dividing the preprocessed data into a training set and a verification set according to a proportion;
(4) And sending the training set into the constructed neonatal necrotizing enterocolitis screening model for training, evaluating the performance of the classification model by using the verification set, adjusting the hyper-parameters of the model according to the evaluation effect, and finally finishing model training.
4. The system for screening neonatal necrotizing enterocolitis according to claim 3, wherein in step (1), the collected data of the neonatal abdominal plain for necrotizing enterocolitis shall meet the following criteria:
the infant lies on the back on the photographic table, lifts the arms of the infant and clamps the head and the hands of the infant; the head, shoulders and knees of the infant patient are tightly attached to the photographic table, and the middle surface of the body is aligned with the central line of the table top, so that the body position is ensured not to be skewed; the central line is aligned with the midpoint of the joint connecting line of the xiphoid process and the pubis and vertically shot into the detector; the field of projection is minimized, with the upper margin including the diaphragm surface and the lower margin including the pubic symphysis, to reduce scattered radiation and radiation dose to the neonate.
5. The neonatal necrotizing enterocolitis screening system of claim 3, wherein in step (2), the specific process of preprocessing the abdominal surview data is as follows:
firstly, carrying out image enhancement on original abdomen plain film data by using a CLAHE algorithm so as to highlight key image characteristics of the abdomen plain film; the image size is unified into 224 × 224 size, and the image format is converted into jpg format by DICOM.
6. The neonatal necrotizing enterocolitis screening system of claim 3, wherein in step (3), the preprocessed data are divided into training and validation sets in a ratio of 8.
7. The neonatal necrotizing enterocolitis screening system of claim 3, wherein in the step (4), 64 image data columns are grouped for model training during training, and random left-right or horizontal rotation of 25 degrees is adopted to enlarge the sample size.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210655628.XA CN114913169B (en) | 2022-06-10 | 2022-06-10 | Neonatal necrotizing enterocolitis screening system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210655628.XA CN114913169B (en) | 2022-06-10 | 2022-06-10 | Neonatal necrotizing enterocolitis screening system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114913169A CN114913169A (en) | 2022-08-16 |
| CN114913169B true CN114913169B (en) | 2023-03-24 |
Family
ID=82770889
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210655628.XA Active CN114913169B (en) | 2022-06-10 | 2022-06-10 | Neonatal necrotizing enterocolitis screening system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114913169B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113674252A (en) * | 2021-08-25 | 2021-11-19 | 上海鹏冠生物医药科技有限公司 | Histopathology image diagnosis system based on graph neural network |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013138431A1 (en) * | 2012-03-15 | 2013-09-19 | The Penn State Research Foundation, The Pennsylvania State University | Methods for predicting and treating necrotizing enterocolitis in neonates |
| CN111340828A (en) * | 2020-01-10 | 2020-06-26 | 南京航空航天大学 | Brain glioma segmentation based on cascaded convolutional neural networks |
| CN111401480B (en) * | 2020-04-27 | 2023-07-25 | 上海市同济医院 | Novel mammary gland MRI automatic auxiliary diagnosis method based on fusion attention mechanism |
| AU2021342157A1 (en) * | 2020-09-11 | 2023-04-20 | Inseer, Inc. | Methods and apapratus for machine learning to analyze musculo-skeletal rehabilitation from images |
| CN114037686B (en) * | 2021-11-09 | 2022-05-17 | 浙江大学 | Children intussusception automatic check out system based on degree of depth learning |
| CN114188021B (en) * | 2021-12-13 | 2022-06-10 | 浙江大学 | Intelligent analysis system for children intussusception diagnosis based on multi-mode fusion |
| CN114565786A (en) * | 2022-01-30 | 2022-05-31 | 天津大学 | Tomography image classification device and method based on channel attention mechanism |
-
2022
- 2022-06-10 CN CN202210655628.XA patent/CN114913169B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113674252A (en) * | 2021-08-25 | 2021-11-19 | 上海鹏冠生物医药科技有限公司 | Histopathology image diagnosis system based on graph neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114913169A (en) | 2022-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gozes et al. | Coronavirus detection and analysis on chest ct with deep learning | |
| Dong et al. | A generic quality control framework for fetal ultrasound cardiac four-chamber planes | |
| CN111986177B (en) | Chest rib fracture detection method based on attention convolution neural network | |
| Wang et al. | Using CFW-net deep learning models for X-ray images to detect COVID-19 patients | |
| CN112784856A (en) | Channel attention feature extraction method and identification method of chest X-ray image | |
| CN109255354B (en) | Medical CT-oriented computer image processing method and device | |
| Pasyar et al. | Hybrid classification of diffuse liver diseases in ultrasound images using deep convolutional neural networks | |
| CN111104961A (en) | Method for classifying breast cancer based on improved MobileNet network | |
| WO2024016575A1 (en) | Cbam mechanism-based residual network medical image auxiliary detection method | |
| CN111260639A (en) | Multi-view information-collaborative breast benign and malignant tumor classification method | |
| Gao et al. | A deep Siamese-based plantar fasciitis classification method using shear wave elastography | |
| Yu et al. | Intelligent diagnosis algorithm for thyroid nodules based on deep learning and statistical features | |
| Ivanescu et al. | Evolutionary computation paradigm to determine deep neural networks architectures | |
| CN114913169B (en) | Neonatal necrotizing enterocolitis screening system | |
| CN116363438B (en) | Child encephalitis classification system based on multiple modes | |
| CN112508943A (en) | Breast tumor identification method based on ultrasonic image | |
| CN114821176B (en) | Viral encephalitis classification system for MR (magnetic resonance) images of children brain | |
| Mao et al. | Efficient and accurate diagnosis of otomycosis using an ensemble deep-learning model | |
| Wang et al. | Triplanar convolutional neural network for automatic liver and tumor image segmentation | |
| Kazemi et al. | A new parallel deep learning algorithm for breast cancer classification | |
| CN115330600A (en) | Lung CT image super-resolution method based on improved SRGAN | |
| Mathina Kani et al. | Classification of skin lesion images using modified Inception V3 model with transfer learning and augmentation techniques | |
| Estermann et al. | Pwml detection in 3d cranial ultrasound volumes using over-segmentation and multimodal classification with deep learning | |
| Cao et al. | EFAG-CNN: Effectively fused attention guided convolutional neural network for WCE image classification | |
| Feng | PCXRNet: Condense attention block and Multiconvolution spatial attention block for Pneumonia Chest X-Ray detection |
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 |