CN110279433B - Automatic and accurate fetal head circumference measuring method based on convolutional neural network - Google Patents
Automatic and accurate fetal head circumference measuring method based on convolutional neural network Download PDFInfo
- Publication number
- CN110279433B CN110279433B CN201910384672.XA CN201910384672A CN110279433B CN 110279433 B CN110279433 B CN 110279433B CN 201910384672 A CN201910384672 A CN 201910384672A CN 110279433 B CN110279433 B CN 110279433B
- Authority
- CN
- China
- Prior art keywords
- image
- head circumference
- ultrasound
- convolutional neural
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/08—Detecting organic movements or changes, e.g. tumours, cysts, swellings
- A61B8/0866—Detecting organic movements or changes, e.g. tumours, cysts, swellings involving foetal diagnosis; pre-natal or peri-natal diagnosis of the baby
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/52—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/5215—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves involving processing of medical diagnostic data
Abstract
The invention discloses a method for automatically and accurately measuring a fetal head circumference based on a convolutional neural network, which sequentially comprises the following steps of: inputting an original fetus ultrasound thalamic section ultrasound scanning image; removing sensitive information of an input image; cutting off the desensitization ultrasonic image from multiple angles; inputting the image data to the trained DeepLabv2 to obtain a multi-angle image segmentation result; overlapping and fusing the multi-angle image segmentation results; and finishing measurement after extracting the segmentation boundary. The invention realizes accurate measurement of the head circumference of the fetal ultrasound thalamus section, effectively partitions and identifies the craniocerebral area of the fetal thalamus section by using the DeepLabV2 convolutional neural network model, accurately finds out the edge of the skull which presents strong echo in an ultrasound image, further realizes accurate measurement of the head circumference of the fetus, is suitable for prenatal ultrasound examination, has positive effects on relieving the working pressure of doctors and improving the working efficiency of prenatal ultrasound diagnosis, can relieve increasingly sharp doctor-patient contradictions and improve the effective utilization of medical resources, and has great social value and practical use value.
Description
Technical Field
The invention relates to the technical field of medical ultrasonic diagnosis, in particular to a fetal head circumference automatic accurate measurement method based on a convolutional neural network. .
Background
Accurate assessment of fetal development is critical to ensure continued health of the mother and newborn during and after pregnancy. Two-dimensional ultrasound has the advantages of strong real-time performance, low cost, wide usability, no harmful radiation and the like, so the two-dimensional ultrasound is widely used for antenatal examination. Prenatal ultrasonic examination is the most common method for observing the growth and development conditions of the fetus, and is also an important basis for preventing the birth of a defective fetus and evaluating the development degree of the fetus.
The main flow of prenatal ultrasound examination comprises the following steps: firstly, obtaining an ultrasonic standard section of a key part of a fetus; secondly, measuring biological parameters based on the ultrasonic standard section of the fetus; thirdly, using the estimation of the body weight of the fetus, the growth and development conditions of the fetus are judged. As known from the prenatal ultrasound examination process, a doctor obtains an ultrasound standard section of each part of a fetus by using two-dimensional ultrasound, and measures biological parameters based on the ultrasound standard section, such as Chen Hao, Dou Qi, Ni Dong, et al, automatic real ultrasound standard plate detection using knowledge transmitted real network [ C ]// Nassir Navab, Joachim Hornegger, Medical Image Computing and Computer-Assisted interaction, music, spring International publication, 2015: 507 and 514, in the section of the upper abdomen, the abdominal circumference is measured; measuring the circumference of the head on a thalamic section; measuring the length of the femur on a femur section; the hip length and the thickness of the transparent layer of the neck are measured on the median sagittal section.
Currently, in the actual clinical diagnosis, the measurement of the circumference of the head on the thalamic section is manually performed by the sonographer as shown in fig. 2, which causes the following five disadvantages:
1. the inspection quality of the ultrasonic image is difficult to effectively control; the parameter measurement is acquired depending on the clinical experience and professional level of an ultrasonic doctor, and if the experience level is insufficient, the quality of an ultrasonic image is difficult to guarantee;
2. the ultrasonic inspection results are not uniform; different sonographers have different abilities and experiences, so that the head circumference measurement results of the fetus are different;
3. the working efficiency of ultrasonic inspection is difficult to be effectively improved; as the complete prenatal ultrasonic examination needs to acquire the parameters of the head circumference of the fetus, the manual measurement mode causes low efficiency;
4. sonographers are prone to occupational disease; on one hand, the number of the current sonographers is lack, the working frequency is very high, and meanwhile, the sonographers need to carry out a large amount of repetitive work such as moving probes, freezing images and the like, so that the sonographers are prone to repetitive pressure injury;
5. the ultrasonic image diagnosis technology of the primary hospital needs to be improved; the diagnosis technology of ultrasonic doctors in primary hospitals needs to be further improved, and particularly for the diagnosis of old puerperae, the patients need to go to the third hospital for diagnosis, so that the patients are inconvenient to visit and ask for a doctor, the working intensity of the third hospital is increased, and medical resources are not really sunk to the primary level.
In recent years, many automatic head circumference measurement systems based on Hough transform (Hough transform), Haar-like features (Haar-like), multi-threshold (Multilevel threshold), Circular shortest path (Circular shortest path), Active contour model (Active constraining), and the like have been proposed. However, the methods are approximate measurement modes based on ellipse fitting, no scheme based on full-automatic accurate measurement of skull contour lines exists, and the method is a problem worthy of breakthrough in the field of intelligent ultrasound.
Computer Vision technology (CV) is rapidly developing, and Deep Convolutional Neural Networks (DCNN) is the most rapidly developing. The DCNN is derived from the enlightenment and variation of the structure of cerebral neurons on the basis of the traditional multilayer perceptron, and abstracts characteristics extracted from a large amount of complex high-dimensional data through the combined application of a convolution kernel and a pooling kernel, and the spatial correlation in the data is mined. DCNN has been successfully applied in the field of medical images, such as in image registration, localization, anatomy/cellular structure detection, tissue segmentation, computer-aided disease diagnosis and prognosis evaluation, etc. However, no DCNN-based fully automatic measurement method is available for head circumference measurement on a fetal ultrasound thalamus standard section.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the invention provides a fetal head circumference automatic accurate measurement method based on a convolutional neural network, which solves the problems that the efficiency is low, the influence of subjective factors of doctors is large, the accurate detection is not facilitated and the like in the clinical actual diagnosis process of fetal head circumference measurement, so that the invention provides a fully automatic accurate measurement method of fetal head circumference based on a deep convolutional neural network-deep Labv 2.
The invention is realized by the following technical scheme:
a fetal head circumference automatic accurate measurement method based on a convolutional neural network comprises the following steps:
the method comprises the following steps of (1) inputting an original fetus ultrasound thalamic section ultrasound scanning image by using a Setp 1;
setp2, multi-angle cropping of ultrasound images;
the Setp3 is input to the trained DeepLabv2 to obtain a multi-angle image segmentation result;
setp4, overlapping and fusing the multi-angle image segmentation results;
and Setp5, extracting the boundary of the superposed and fused image, and finishing measurement after acquiring the length of the boundary pixel of the image.
Further, the measuring method is to design and test the head circumference automatic measuring system under a deep learning framework TensorFlow.
Further, sensitive information to remove input images, including information about patient privacy, is also included between Setp1 and Setp 2.
Further, the Setp2 multi-angle cropping refers to selecting multiple angle directions above an image for cropping. Namely, any position above the image is selected for cutting, and the distance between the orthographic projection point of the selected cutting position and the target object on the image is more than or equal to 0.
Further, in the Setp3, the DeepLabv2 model training method comprises the following steps:
setp31, constructing a DeepLabv2 network model;
setp32, fine tuning is carried out by adopting a pre-training set to complete a pre-trained DeepLabv2 network model;
a Setp33 for determining the difference between the predicted value and the actual value to modify the deplab v2 network model;
setp34, in the training process, the original image and the marked image are input and trained in pairs, and the cross entropy of each pixel in the image to the loss function is an equivalent weight;
setp35, using a random gradient descent algorithm to optimize the deplab v2 network model parameters.
Further, the deep Labv2 model training method is specifically operated as follows: the Setp31 is used for pre-training the DeepLabv2 network model by using the ImageNet data set to complete initialization so as to achieve the purpose of transfer learning; the Setp32 finishes a pretrained DeepLabv2 network model, continues training on the fetal ultrasound image, and performs fine adjustment to adapt to the craniocerebral segmentation task of the fetal ultrasound thalamic section image; setp33, in the training process, a cross entropy cost function is used as a loss function to measure the difference between a predicted value and an actual value; setp34, in the training process, the original image and the marked image are input and trained in pairs, and the cross entropy of each pixel in the image to the loss function is an equivalent weight; setp35, using stochastic gradient descent algorithm, optimize DeepLabv2 network model parameters to achieve the optimization, the whole training process, total iteration 2 x 105Next, the initial learning rate is 2.5 × 10-4Every 1X 104The learning rate is reduced by 5 × 10-4And (4) doubling.
Further, in the Setp4, the specific operation method is as follows: and according to a multi-angle cutting mode, reversely splicing and restoring the binary segmented images into an original image, wherein the original image is in one-to-one correspondence with the input images cut at multiple angles.
Further, in the Setp5, the specific operation of extracting the boundary of the superimposed fusion image is as follows: the superimposed and fused image obtained by Setp4 was first dilated, and then the superimposed and fused image obtained by Setp4 was subtracted from the dilated image.
The invention has the following advantages and beneficial effects:
the invention realizes accurate measurement of the head circumference of the fetal ultrasound thalamus section, effectively partitions and identifies the craniocerebral area of the fetal thalamus section by using the DeepLabV2 convolutional neural network model, accurately finds out the edge of the skull which presents strong echo in an ultrasound image, further realizes accurate measurement of the head circumference of the fetus, is suitable for prenatal ultrasound examination, has positive effects on relieving the working pressure of doctors and improving the working efficiency of prenatal ultrasound diagnosis, can relieve increasingly sharp doctor-patient contradictions and improve the effective utilization of medical resources, and has great social value and practical use value.
Drawings
The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the principles of the invention. In the drawings:
FIG. 1 is a schematic view of a sectional thalamic view;
FIG. 2 is a prior art manual ellipsometry diagram for a physician;
FIG. 3 is an original image cut from multiple angles;
FIG. 4 is a multi-angle cropped image;
FIG. 5 is a flow chart of a head circumference measurement method of the present invention;
fig. 6 is a comparison diagram of a labeling diagram of the ultrasonic head circumference of the fetus.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to examples and accompanying drawings, and the exemplary embodiments and descriptions thereof are only used for explaining the present invention and are not meant to limit the present invention.
Example 1
The embodiment provides a method for automatically and accurately measuring the head circumference of a fetus based on a convolutional neural network, which sequentially comprises the following operation steps:
the method comprises the following steps of (1) inputting an original fetus ultrasound thalamic section ultrasound scanning image by using a Setp 1; the automatic accurate measurement method of the head circumference of the fetus based on the convolutional neural network is characterized in that a head circumference automatic measurement system is designed and tested under a deep learning framework TensorFlow;
setp2, removing sensitive information of the input image, wherein the sensitive information comprises privacy information such as patient name, age and the like of the image edge;
setp3, cutting off the sensitive ultrasonic image from multiple angles; as shown in fig. 3 and 4, an original image having a length W and a width H is cut at a ratio Q (0 < Q <1, Q = 0.9) from four angles of upper left, upper right, lower right, and lower left, and the cut image is input to the model at a size of 321 × 321;
the Setp4 is input to the trained DeepLabv2 to obtain a multi-angle image segmentation result, and the boundary of the confusion area of the target area and the background area of the image is segmented; the DeepLabv2 related art can be found in Chen L C, Papandrou G, Kokkinos I, et al, DeepLab: semiconductor Image Segmentation with deep illumination networks, atom illumination, and fusion Connected CRFs [ J ]. IEEE Transactions on Pattern Analysis & Machine Analysis, 2018, 40(4):834-848.
The system experiment operating hardware environment is 32-core 2.9GHz Intel Xeon E5-2670 CPU, NVIDIA1080TI GPU and 64GB memory computer. In order to avoid the over-fitting problem, the deep labv2 model used in the invention is not initialized by adopting gaussian distribution, but directly uses a migration learning mode to initialize the migration of each layer parameter learned on a natural image and carries out fine adjustment in the fetal ultrasound image training process, which is described in detail as follows.
DeepLabv2 model training procedure: 1. firstly, pretraining a DeepLabv2 network model by using an ImageNet data set to complete initialization so as to achieve the aim of transfer learning; 2. finishing a pretrained DeepLabv2 network model, continuing training on the fetal ultrasound image, and performing fine adjustment to adapt to a craniocerebral segmentation task of the fetal ultrasound thalamus section image; 3. in the training process, a cross entropy cost function is used as a loss function to measure the difference between a predicted value and an actual value; 4. in the training process, the original image and the marked image are input and trained in pairs, and the cross entropy of each pixel pair loss function in the image is equivalent to the cross entropy of the loss function; 5. optimizing DeepLabv2 network model parameters by using a Stochastic Gradient Descent (SGD) algorithm to achieve the optimal condition; 6. the whole training process, total iteration 2X 105Next, the initial learning rate is 2.5 × 10-4Every 1X 104The learning rate is reduced by 5 × 10-4And (4) doubling.
Setp5, overlapping and fusing the multi-angle segmentation results; splicing and restoring the binary segmented images into an original image in a multi-angle cutting mode in a reverse direction, wherein the original image corresponds to the input image cut in multiple angles one to one; the pixel values of the points of the division map corresponding to the original image obtained by the division operation are maintained in the range of 0 to 255 for the overlapped portion.
Setp6, extracting the boundary of the superposed fusion image and then finishing measurement; for the superposed and fused image A obtained in the step 5, firstly expanding the image A, and then subtracting the original image A from the expanded image B; and finally, obtaining the length of the boundary pixel to finish measurement.
Example 2
Based on the automatic and accurate measuring method for the head circumference of the fetus provided by the embodiment 1 and based on the convolutional neural network, the result evaluation is carried out:
1261 fetal sonothalamic data sets used in the invention, 1145 training sets and 116 testing sets.
1. Qualitative assessment
Fig. 6 is a labeling diagram of the ultrasonic head circumference of the fetus. Arranged from left to right in sequence, the first column is the clinically common semi-automatic labeling of ellipses (dotted line), the second column is the manual outline labeling (solid line), and the third column is the automatic labeling result of the algorithm (solid line). It can be obviously found that the algorithm successfully identifies and segments the fetal cranium and accurately finds out the edge of the skull with strong echo in the ultrasonic image. Meanwhile, the segmentation result of the algorithm is more accurate than the clinical common ellipse labeling and basically consistent with the manual labeling, and the feasibility of the algorithm on the problem is demonstrated.
2. Quantitative assessment
The present invention uses region-based evaluation indices: coincidence (Dice), similarity (Jaccard), to evaluate the segmentation results; the mean (d), standard deviation(s) were used to evaluate the measurements.
Suppose a is a manually marked target region and B is a target region that the algorithm automatically segments. The calculation formula of Dice and Jaccard is as follows:
where Area (, denotes an Area calculator). The evaluation results are shown in table 1:
TABLE 1 quantitative evaluation
| Dice | Jaccard | d（cm） | S（cm） | |
| Ellipse labeling | — | — | +1.5621 | 0.2557 |
| Manual marking | 0.9758 | 0.9530 | -0.3517 | 0.2816 |
As can be seen from Table 1, the coincidence rate and the similarity of the result of the invention and the manual labeling both exceed 0.95, the mean error of the measurement is-0.3517 cm (-indicating that the measurement result of the manual labeling is smaller than that of the invention), and the standard deviation is 0.2816, indicating that the invention and the manual labeling have better consistency; meanwhile, the common clinical ellipse marking measurement result is generally larger than the measurement result of the invention, the mean error is +1.5621cm, and the mean error with the manual marking measurement is +1.9128cm, which shows the effectiveness of the invention in accurately measuring the head circumference.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (8)
1. A fetal head circumference automatic accurate measurement method based on a convolutional neural network is characterized by comprising the following steps:
the method comprises the following steps of (1) inputting an original fetus ultrasound thalamic section ultrasound scanning image by using a Setp 1;
setp2, multi-angle cropping of ultrasound images;
the Setp3 is input to the trained DeepLabv2 to obtain a multi-angle image segmentation result;
setp4, overlapping and fusing the multi-angle image segmentation results;
and Setp5, extracting the boundary of the superposed and fused image, and finishing measurement after acquiring the length of the boundary pixel of the image.
2. The automatic precise measurement method for the head circumference of the fetus based on the convolutional neural network as claimed in claim 1, wherein the measurement method is to design and test an automatic head circumference measurement system under a deep learning framework Tensorflow.
3. The automatic fetal head circumference accurate measurement method based on the convolutional neural network as claimed in claim 1, further comprising sensitive information between Setp1 and Setp2 for removing input images, wherein the sensitive information comprises privacy information about patients.
4. The automatic fetal head circumference accurate measurement method based on the convolutional neural network as claimed in claim 1, wherein the Setp2 multi-angle cropping refers to selecting a plurality of angle directions above an image for cropping.
5. The automatic precise measurement method for the head circumference of the fetus based on the convolutional neural network as claimed in claim 1, wherein in the Setp3, the deep lab v2 model training method comprises the following steps:
setp31, constructing a DeepLabv2 network model;
setp32, fine tuning is carried out by adopting a pre-training set to complete a pre-trained DeepLabv2 network model;
a Setp33 for determining the difference between the predicted value and the actual value to modify the deplab v2 network model;
setp34, in the training process, the original image and the marked image are input and trained in pairs, and the cross entropy of each pixel in the image to the loss function is an equivalent weight;
setp35, using a random gradient descent algorithm to optimize the deplab v2 network model parameters.
6. The automatic precise measurement method for the head circumference of the fetus based on the convolutional neural network as claimed in claim 5, wherein the DeepLabv2 model training method is specifically operated as follows: the Setp31 is used for pre-training the DeepLabv2 network model by using the ImageNet data set to complete initialization so as to achieve the purpose of transfer learning; the Setp32 finishes a pretrained DeepLabv2 network model, continues training on the fetal ultrasound image, and performs fine adjustment to adapt to the craniocerebral segmentation task of the fetal ultrasound thalamic section image; setp33, in the training process, a cross entropy cost function is used as a loss function to measure the difference between a predicted value and an actual value; setp34, training process, input training of original image and labeled image pair, each pixel in the image is equivalent to the cross entropy of loss functionA weight; setp35, using stochastic gradient descent algorithm, optimize DeepLabv2 network model parameters to achieve the optimization, the whole training process, total iteration 2 x 105Next, the initial learning rate is 2.5 × 10-4Every 1X 104The learning rate is reduced by 5 × 10-4And (4) doubling.
7. The automatic precise measurement method for the head circumference of the fetus based on the convolutional neural network as claimed in claim 1, wherein in the Setp4, the specific operation method is as follows: and according to a multi-angle cutting mode, reversely splicing and restoring the binary segmented images into an original image, wherein the original image is in one-to-one correspondence with the input images cut at multiple angles.
8. The automatic precise measurement method for the head circumference of the fetus based on the convolutional neural network as claimed in claim 1, wherein the specific operation of extracting the boundary of the superimposed fusion image in the Setp5 is as follows: the superimposed and fused image obtained by Setp4 was first dilated, and then the superimposed and fused image obtained by Setp4 was subtracted from the dilated image.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811108648 | 2018-09-21 | ||
| CN2018111086485 | 2018-09-21 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110279433A CN110279433A (en) | 2019-09-27 |
| CN110279433B true CN110279433B (en) | 2020-03-27 |
Family
ID=68002206
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910384672.XA Active CN110279433B (en) | 2018-09-21 | 2019-05-09 | Automatic and accurate fetal head circumference measuring method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110279433B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111242929A (en) * | 2020-01-13 | 2020-06-05 | 中国科学技术大学 | Fetal skull shape parameter measuring method, system, equipment and medium |
| CN113469941B (en) * | 2021-05-27 | 2022-11-08 | 武汉楚精灵医疗科技有限公司 | Method for measuring width of bile-pancreatic duct in ultrasonic bile-pancreatic duct examination |
| CN114782407A (en) * | 2022-05-23 | 2022-07-22 | 中山大学附属第一医院 | Fetal ultrasonic key section identification method and device based on convolutional neural network |
| CN114782457A (en) * | 2022-05-23 | 2022-07-22 | 中山大学附属第一医院 | Fetal brain ultrasonic structure segmentation method and device based on three-dimensional convolutional neural network |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600595A (en) * | 2016-12-21 | 2017-04-26 | 厦门可睿特信息科技有限公司 | Human body characteristic dimension automatic measuring method based on artificial intelligence algorithm |
| CN106897573A (en) * | 2016-08-01 | 2017-06-27 | 12西格玛控股有限公司 | Use the computer-aided diagnosis system for medical image of depth convolutional neural networks |
| WO2018009405A1 (en) * | 2016-07-08 | 2018-01-11 | Avent, Inc. | System and method for automatic detection, localization, and semantic segmentation of anatomical objects |
| CN108053410A (en) * | 2017-12-11 | 2018-05-18 | 厦门美图之家科技有限公司 | Moving Object Segmentation method and device |
| CN108288271A (en) * | 2018-02-06 | 2018-07-17 | 上海交通大学 | Image detecting system and method based on three-dimensional residual error network |
-
2019
- 2019-05-09 CN CN201910384672.XA patent/CN110279433B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018009405A1 (en) * | 2016-07-08 | 2018-01-11 | Avent, Inc. | System and method for automatic detection, localization, and semantic segmentation of anatomical objects |
| CN106897573A (en) * | 2016-08-01 | 2017-06-27 | 12西格玛控股有限公司 | Use the computer-aided diagnosis system for medical image of depth convolutional neural networks |
| CN106600595A (en) * | 2016-12-21 | 2017-04-26 | 厦门可睿特信息科技有限公司 | Human body characteristic dimension automatic measuring method based on artificial intelligence algorithm |
| CN108053410A (en) * | 2017-12-11 | 2018-05-18 | 厦门美图之家科技有限公司 | Moving Object Segmentation method and device |
| CN108288271A (en) * | 2018-02-06 | 2018-07-17 | 上海交通大学 | Image detecting system and method based on three-dimensional residual error network |
Non-Patent Citations (2)
| Title |
|---|
| DeepLab:Semantic Image Segmentation with Deep Convolutional Nets,Atrous Convolution, and Fully Connected CRFs;Chen L C et al.;《IEEE Transactions on Pattern Analysis & Machine Intelligence》;20180430;第40卷(第4期);第834-848页 * |
| 胎儿腹部超声图像自动质量控制与参数测量;吴凌云;《中国优秀硕士学位论文全文数据库 医药卫生科技辑（月刊）》;20170715(第07期);第E068-48页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110279433A (en) | 2019-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110279433B (en) | Automatic and accurate fetal head circumference measuring method based on convolutional neural network | |
| Xu et al. | A hybrid method for airway segmentation and automated measurement of bronchial wall thickness on CT | |
| CN109255354B (en) | Medical CT-oriented computer image processing method and device | |
| CN110910377A (en) | Cerebral infarction MRI image identification method based on neural network | |
| El-Hariri et al. | Comparative evaluation of hand-engineered and deep-learned features for neonatal hip bone segmentation in ultrasound | |
| Liu et al. | Automated classification and measurement of fetal ultrasound images with attention feature pyramid network | |
| Shinohara et al. | Using deep learning for ultrasound images to diagnose carpal tunnel syndrome with high accuracy | |
| de Jesus Guerrero-Turrubiates et al. | Fast Parabola Detection Using Estimation of Distribution Algorithms. | |
| Zeng et al. | TUSPM-NET: A multi-task model for thyroid ultrasound standard plane recognition and detection of key anatomical structures of the thyroid | |
| US10729389B2 (en) | 3D assessment of conjugant eye deviation for the identificaiton of acute ischemic stroke | |
| Onal et al. | Assessment of a semiautomated pelvic floor measurement model for evaluating pelvic organ prolapse on MRI | |
| Lin et al. | Evaluation of pediatric hydronephrosis using deep learning quantification of fluid-to-kidney-area ratio by ultrasonography | |
| CN108836394B (en) | Automatic measuring method for descending angle of fetal head | |
| Dan et al. | DeepGA for automatically estimating fetal gestational age through ultrasound imaging | |
| Ou et al. | A sub-pixel accurate quantification of joint space narrowing progression in rheumatoid arthritis | |
| CN113393456A (en) | Automatic quality control method of early pregnancy fetus standard section based on multiple tasks | |
| Ni et al. | Automatic cystocele severity grading in transperineal ultrasound by random forest regression | |
| CN112508943A (en) | Breast tumor identification method based on ultrasonic image | |
| CN111481233B (en) | Thickness measuring method for transparent layer of fetal cervical item | |
| Oguz et al. | Automated placenta segmentation from 3D ultrasound images | |
| WO2023133929A1 (en) | Ultrasound-based human tissue symmetry detection and analysis method | |
| Marcomini et al. | Quantitative evaluation of automatic methods for lesions detection in breast ultrasound images | |
| Radhi et al. | An automatic segmentation of breast ultrasound images using U-Net model | |
| Nekooeimehr et al. | Automated tracking, segmentation and trajectory classification of pelvic organs on dynamic MRI | |
| CN116211353B (en) | Wearable ultrasonic bladder capacity measurement and multi-mode image morphology evaluation system |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Luo Hong Inventor after: Zhang Bo Inventor before: Luo Hong |
|
| CB03 | Change of inventor or designer information |