CN112184684A - Improved YOLO-v3 algorithm and application thereof in lung nodule detection - Google Patents
Improved YOLO-v3 algorithm and application thereof in lung nodule detection Download PDFInfo
- Publication number
- CN112184684A CN112184684A CN202011074796.7A CN202011074796A CN112184684A CN 112184684 A CN112184684 A CN 112184684A CN 202011074796 A CN202011074796 A CN 202011074796A CN 112184684 A CN112184684 A CN 112184684A
- Authority
- CN
- China
- Prior art keywords
- image
- algorithm
- data set
- yolo
- detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 44
- 206010056342 Pulmonary mass Diseases 0.000 title claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 11
- 230000032823 cell division Effects 0.000 claims abstract description 8
- 238000012545 processing Methods 0.000 claims description 7
- 230000002685 pulmonary effect Effects 0.000 claims description 7
- 238000000034 method Methods 0.000 claims description 6
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 2
- 230000011218 segmentation Effects 0.000 claims 2
- 230000006870 function Effects 0.000 description 10
- 210000004072 lung Anatomy 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 230000035945 sensitivity Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000002591 computed tomography Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 208000019693 Lung disease Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 210000000038 chest Anatomy 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 210000001519 tissue Anatomy 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23211—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with adaptive number of clusters
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4084—Transform-based scaling, e.g. FFT domain scaling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/32—Normalisation of the pattern dimensions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
- G06T2207/30064—Lung nodule
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The invention discloses an improved YOLO-v3 algorithm and application thereof in lung nodule detection, wherein a detection frame loss function of the YOLO-v3 algorithm is optimized firstly, and an anchor frame is optimized according to average intersection and comparison; then, carrying out image scaling on the acquired data set, and carrying out cell division and feature extraction on the input image to complete improvement on the YOLO-v3 algorithm; then acquiring a Luna16 data set, and preprocessing images in the Luna16 data set, wherein the preprocessing comprises converting CT image gray scale values into HU values, then generating a mask, and finally normalizing and unifying the size; and then, respectively turning the preprocessed image by 90 degrees, 180 degrees and 270 degrees, renaming according to the naming rule of PASCAL VOC, converting all data into a data set in a VOC format, inputting a training set divided from the data set into an improved YOLO-v3 algorithm for feature extraction, generating a nodule prediction frame and a prediction confidence probability, and improving the detection precision.
Description
Technical Field
The invention relates to the technical field of medical image processing, in particular to an improved YOLO-v3 algorithm and application thereof in lung nodule detection.
Background
Medical ct (computed tomography) is one of the most common and effective imaging examinations, and it uses an X-ray collimation system to obtain clear cross-sectional images, accurate layer thickness, high density resolution, and no interference data of out-of-layer structures. With the widespread use of medical CT, the lung CT image data is "explosively" growing, and usually a whole lung CT contains 100-500 lung sectional images. In addition, the early pulmonary nodules generally have the characteristics of small volume, blurred edge and difficulty in distinguishing by naked eyes, so that the early pulmonary nodules and the positions thereof are identified from the CT images, the workload of doctors is greatly increased, very high requirements are also provided for professional level and experience of doctors, the judgment results are different from person to person, and different doctor judgment results may be different, so that a method for processing CT image data is needed, and the method can assist doctors in quickly and accurately identifying the pulmonary nodules and the positions thereof.
Disclosure of Invention
The invention aims to provide an improved YOLO-v3 algorithm and application thereof in lung nodule detection, and detection accuracy is improved.
To achieve the above object, in a first aspect, the present invention provides an improved YOLO-v3 algorithm, comprising the steps of:
optimizing a loss function of the detection frame, and optimizing the anchor frame according to average intersection and comparison;
and carrying out image scaling on the acquired data set, and carrying out cell division and feature extraction on the input image.
Wherein, carry out image scaling to the data set who obtains to carry out cell division and feature extraction with input image, include:
the method comprises the steps of scaling the size of an image in an acquired data set, dividing the image into a plurality of cells after the size of the image is unified, and carrying out target identification detection on each cell.
Wherein, carry out image scaling to the data set who obtains to carry out cell division and feature extraction with input image, still include:
and performing up-sampling on the cell which completes target identification detection, combining the up-sampling with the output characteristics of adjacent residual blocks, and controlling the output of characteristic extraction by using two layers of ResNet.
In a second aspect, the present invention provides an application of a modified YOLO-v3 algorithm in lung nodule detection, the modified YOLO-v3 algorithm being applied to lung nodule detection, comprising the following steps:
acquiring a Luna16 data set, and preprocessing images in the Luna16 data set;
performing data enhancement on the preprocessed image, and converting the image into a VOC format according to a naming rule;
inputting a training set into the improved YOLO-v3 algorithm for feature extraction, and generating a nodule prediction box and a prediction confidence probability.
Acquiring a Luna16 data set, and preprocessing images in the Luna16 data set, wherein the preprocessing comprises:
and converting the CT expected gray value in the acquired Luna16 data set into an HU value according to the linear attenuation coefficients of air, water and X-rays, filtering the HU value according to a set threshold value, and performing expansion processing to obtain a corresponding mask map.
The method comprises the following steps of performing data enhancement on the preprocessed image, converting the image into a VOC format according to a naming rule, and comprising the following steps:
and respectively turning the preprocessed image by 90 degrees, 180 degrees and 270 degrees, renaming according to the naming rule of PASCALVOC, converting all data into a data set in a VOC format, and dividing the data set.
The invention relates to an improved YOLO-v3 algorithm and application thereof in lung nodule detection, which comprises the steps of firstly optimizing a detection frame loss function of the YOLO-v3 algorithm, and optimizing an anchor frame according to average intersection and comparison; then, carrying out image scaling on the acquired data set, and carrying out cell division and feature extraction on the input image to complete improvement on the YOLO-v3 algorithm; then acquiring a Luna16 data set, and preprocessing images in the Luna16 data set, wherein the preprocessing comprises converting CT image gray values into HU values, then generating a mask, removing interference parts to enable nodule characteristics to be obvious, and finally normalizing and unifying sizes; and then, respectively turning the preprocessed image by 90 degrees, 180 degrees and 270 degrees, renaming according to the naming rule of PASCAL VOC, converting all data into a data set in a VOC format, inputting a training set divided from the data set into an improved YOLO-v3 algorithm for feature extraction, generating a nodule prediction frame and a prediction confidence probability, and improving the detection precision.
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 of 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 schematic diagram of the steps of an improved YOLO-v3 algorithm provided by the present invention.
Fig. 2 is a schematic diagram of the steps of the application of the improved YOLO-v3 algorithm in lung nodule detection provided by the invention.
Fig. 3 is a flow chart of preprocessing an image according to the present invention.
FIG. 4 is a schematic structural diagram of improved YOLO-v3 provided by the present invention.
Fig. 5 is a schematic diagram of the detection of lung nodules provided by the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
Referring to FIG. 1, the present invention provides an improved YOLO-v3 algorithm, comprising the following steps:
s101, optimizing a loss function of the detection frame, and optimizing the anchor frame according to average intersection and comparison.
Specifically, the detection box evaluation index in the YOLO-v3 network is the intersection ratio (IoU), and the index has some problems: since IoU calculates the cross-over ratio, the quality of the regression box can be better reflected, but the cross-over ratio has no strong correlation with the common loss; if the two objects do not overlap, the value of loU will be zero, while in optimizing the loss function, the gradient is zero, meaning that optimization cannot be performed; and the distance between the two shapes will not be reflected, nor will the alignment of the two objects be correctly distinguished. The formula of GIOU as the loss function of the regression target box is as follows:
wherein, A and B are two rectangular frames with arbitrary size, C is the smallest rectangular frame which can contain A, B at the same time, and the area of C \ C (Acu B) is the area of C minus the area of Acu B. GIOU is similar to IoU, and has excellent properties as a metric in addition to scale invariance; including nonnegativity, identity, symmetry, and the nature of the triangle inequality. When A and B are not well aligned, the area of C is increased, so that the value of GIOU is reduced, and when the two rectangular frames are not overlapped, the GIOU can be calculated, so that the problem that IoU is not suitable for being used as a loss function is solved to a certain extent.
YOLO-v3 employs the anchor frame as the initial candidate frame, so the selection of the anchor frame is also crucial. The anchor frame is different from the object boundary frame and is a frame which is supposed by people. The size and shape of the anchor frame are set, and then a rectangular frame is drawn by taking a certain point on the image as a center. The mean cross-over ratio (avg IoU) is used as an index for evaluating the target clustering analysis in YOLO-v3, and the target function F can be expressed as:
wherein A represents a real sample frame, B represents a clustering center, n is the number of samples, k is the number of clustering centers, n is the number of the clustering centerskNumber of samples representing the k-th clustering center, IIoU(A, B) represents the intersection ratio of the two.
Calculating the values of the objective function F of the number k of the cluster centers from 1 to 9 in turn, it can be found that the number k is gradually reduced when k is 3, so that the cluster centers are three, and the improved cluster results are (6, 6), (11, 10), (21, 20), and the cluster results can be adjusted according to the actual situation of the data set.
And S102, carrying out image scaling on the acquired data set, and carrying out cell division and feature extraction on the input image.
Specifically, when the pictures in the acquired dataset are transmitted into a YOLO-v3 network, the input image is scaled first, the image size most suitable for processing by the YOLO-v3 algorithm is found, and finally the input image is uniformly adjusted to 416 × 416.
The YOLO-v3 algorithm then divides the input image into a number of M × M cells, each cell responsible for detecting the lower right corner region, and if the center point of the object falls in this region, the position of the object is determined by the grid point. As shown in fig. 4: the feature map output by the third layer of residual block is up-sampled, and then the up-sampled feature map is spliced with the output of the second layer of residual block, so that the front feature layer can be mapped into the rear feature layer, and training and feature extraction are facilitated.
In addition, in order to overcome the problems of gradient explosion and gradient disappearance caused by the deepening of the network, a two-layer ResNet structure is added into the network, and the ResNet can effectively classify images and identify objects, so that the improved whole network has higher detection sensitivity to small nodes.
Referring to fig. 2, the present invention provides an application of an improved YOLO-v3 algorithm in lung nodule detection, and the improved YOLO-v3 algorithm is applied in lung nodule detection, including the following steps:
s201, acquiring a Luna16 data set, and preprocessing images in the Luna16 data set.
Specifically, the original LIDC-IDRI dataset is subjected to CT image data removal after the slice thickness is larger than 3mm and the lung nodule is smaller than 3mm, so as to obtain a Luna16 dataset, wherein the Luna16 dataset is a Luna16 dataset; the original image is a three-dimensional image. Each image contains a series of axial slices of the thorax. The three-dimensional images are composed of a different number of two-dimensional images, wherein the mhd file contains the basic information of the CT image and the raw file stores the specific data of the CT image.
The CT images for each patient (each patient's image information contains a. mhd file and a.raw file) are then pre-processed. The preprocessing comprises the steps of converting the gray value of an original CT image into an HU value, then generating a mask, removing an interference part to enable the nodule characteristics to be obvious, and finally normalizing, unifying the sizes and the like. As shown in fig. 3:
1. and converting the CT image gray value into an HU value. The HU value refers to the transmittance of a human tissue and organ to radiation, and is calculated as follows:
where μ is a linear attenuation coefficient, related to the X-ray intensity. Since the HU values of the lungs are around-500, regions with values within-1000, +400 are retained (from air to bone) and regions outside this range are considered to be discarded regardless of lung disease detection.
2. A mask is generated from the HU values. The thresholding operation is used here and the threshold T is set to a gray value at HU-600 so that water, air is substantially filtered and the remaining portion is expanded to fill the small holes in the lungs and create the mask.
3. And (6) normalization processing. Firstly, the pixel value is cut off to be [ -1200, 600], the value smaller than-1200 is set as-1200, the value larger than 600 is set as 600, then the scaling is carried out to be 0-255, and finally the picture is added with the mask image and is stored as the png format.
4. And (5) adjusting the size. The Luna16 dataset was sliced to a size of 512X 512, which was redefined here to a size of 416X 416, scaled to the size best suited for network training.
S202, performing data enhancement on the preprocessed image, and converting the image into a VOC format according to a naming rule.
Specifically, considering that the Luna16 data set is provided with nodule data, which may cause the samples to have singularity, for the problem of unbalance of the positive and negative samples, the positive sample is subjected to data enhancement, and the partial data is respectively turned over by 90 degrees, 180 degrees and 270 degrees from three dimensions. Noise may be introduced in the detection of actual lung nodules due to the mode of operation and other factors, disturbing the image, and thus gaussian blurring of portions of the data. Then, renaming all the data according to the naming rule of PASCAL VOC, and converting the data into a data set in a VOC format, wherein the training set accounts for 90 percent, and the test set accounts for 10 percent.
S203, inputting the training set into the improved YOLO-v3 algorithm for feature extraction, and generating a nodule prediction box and a prediction confidence probability.
Specifically, inputting training set data into an improved YOLO-v3 network for feature extraction, specifically: when the pictures in the data set with the VOC format are transmitted into a YOLO-v3 network, firstly, the input image is zoomed, the image size which is most suitable for the processing of a YOLO-v3 algorithm is found, finally, the input image is uniformly adjusted to be 416 multiplied by 416, and in order to prevent image distortion, gray bars are added to the edges of the image during training.
The YOLO-v3 algorithm then divides the input image into a number of M × M cells, each cell responsible for detecting the lower right corner region, and if the center point of the object falls in this region, the position of the object is determined by the grid point. For the present invention, lung nodules are the targets that need to be detected or identified finally, and lung nodules mostly exist in the form of small targets in the lung CT image, and in order to improve the detection effect of lung nodules in the CT image, the network must acquire the feature information of more small targets. Then, by performing feature extraction through upsampling and the like in S102 in the first embodiment, the improved YOLO-v3 algorithm increases a feature fusion network for small targets, so that there is a high sensitivity for small nodules, and two layers of ResNet are added before output, so that not only is the detection sensitivity and accuracy of the improved whole network for small nodules higher, but also the positions of lung nodules in the CT image can be quickly identified, thereby reducing the dependence on human labor, and avoiding the reduction of detection accuracy caused by human identification. In addition, in order to overcome the problems of gradient explosion and gradient disappearance caused by the deepening of the network, a two-layer ResNet structure is added into the network, and the ResNet can effectively classify images and identify objects, so that the improved whole network has higher detection sensitivity to small nodes.
Finally, a prediction box and a prediction confidence probability of the nodule are generated by using the network model. And the lung nodule detection is realized by adopting an end-to-end YOLO-v3 target detection deep learning network. The method comprises the steps of firstly generating a series of candidate regions on a picture according to a certain rule, then labeling the candidate regions according to the position relation between the candidate regions and an object real frame on the picture, finally extracting picture characteristics by using a convolutional neural network and predicting the positions and the types of the candidate regions, and has a good detection effect on pulmonary nodules.
Fig. 5 shows the detection results of the CT scan image trained for 10000 times (the detection effect of randomly selecting 3 samples). For example, as shown in FIG. 5(a), "nodule: 0.91" and "nodule: 1.0" mean: in this case, two lung nodules were detected, one of which had a probability of 91% and the other of which had a probability of 100%. The algorithm can help a doctor to quickly locate and identify the position of the pulmonary nodule of the patient, and greatly reduces the workload of the doctor while providing convenience for subsequent diagnosis.
The invention relates to an improved YOLO-v3 algorithm and application thereof in lung nodule detection, which comprises the steps of firstly optimizing a detection frame loss function of the YOLO-v3 algorithm, and optimizing an anchor frame according to average intersection and comparison; then, carrying out image scaling on the acquired data set, and carrying out cell division and feature extraction on the input image to complete improvement on the YOLO-v3 algorithm; then acquiring a Luna16 data set, and preprocessing images in the Luna16 data set, wherein the preprocessing comprises converting CT image gray values into HU values, then generating a mask, removing interference parts to enable nodule characteristics to be obvious, and finally normalizing and unifying sizes; and then, respectively turning the preprocessed image by 90 degrees, 180 degrees and 270 degrees, renaming according to the naming rule of PASCAL VOC, converting all data into a data set in a VOC format, inputting a training set divided from the data set into an improved YOLO-v3 algorithm for feature extraction, generating a nodule prediction frame and a prediction confidence probability, and improving the detection precision.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (6)
1. An improved YOLO-v3 algorithm, comprising the steps of:
optimizing a loss function of the detection frame, and optimizing the anchor frame according to average intersection and comparison;
and carrying out image scaling on the acquired data set, and carrying out cell division and feature extraction on the input image.
2. The improved YOLO-v3 algorithm of claim 1, wherein image scaling the acquired dataset and cell segmentation and feature extraction of the input image comprises:
the method comprises the steps of scaling the size of an image in an acquired data set, dividing the image into a plurality of cells after the size of the image is unified, and carrying out target identification detection on each cell.
3. The improved YOLO-v3 algorithm of claim 2, wherein image scaling the acquired dataset and cell segmentation and feature extraction of the input image further comprises:
and performing up-sampling on the cell which completes target identification detection, combining the up-sampling with the output characteristics of adjacent residual blocks, and controlling the output of characteristic extraction by using two layers of ResNet.
4. The application of a modified YOLO-v3 algorithm in pulmonary nodule detection, wherein the modified YOLO-v3 algorithm is applied to pulmonary nodule detection, and is characterized by comprising the following steps:
acquiring a Luna16 data set, and preprocessing images in the Luna16 data set;
performing data enhancement on the preprocessed image, and converting the image into a VOC format according to a naming rule;
inputting a training set into the improved YOLO-v3 algorithm for feature extraction, and generating a nodule prediction box and a prediction confidence probability.
5. The use of the improved YOLO-v3 algorithm in lung nodule detection according to claim 4, wherein acquiring a Luna16 dataset and preprocessing images in the Luna16 dataset comprises:
and converting the CT expected gray value in the acquired Luna16 data set into an HU value according to the linear attenuation coefficients of air, water and X-rays, filtering the HU value according to a set threshold value, and performing expansion processing to obtain a corresponding mask map.
6. The use of the modified YOLO-v3 algorithm in lung nodule detection according to claim 5, wherein the pre-processed image is data enhanced and converted to VOC format according to naming rules, comprising:
and respectively turning the preprocessed image by 90 degrees, 180 degrees and 270 degrees, renaming according to the naming rule of PASCALVOC, converting all data into a data set in a VOC format, and dividing the data set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011074796.7A CN112184684A (en) | 2020-10-09 | 2020-10-09 | Improved YOLO-v3 algorithm and application thereof in lung nodule detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011074796.7A CN112184684A (en) | 2020-10-09 | 2020-10-09 | Improved YOLO-v3 algorithm and application thereof in lung nodule detection |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112184684A true CN112184684A (en) | 2021-01-05 |
Family
ID=73948295
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011074796.7A Pending CN112184684A (en) | 2020-10-09 | 2020-10-09 | Improved YOLO-v3 algorithm and application thereof in lung nodule detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112184684A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112633066A (en) * | 2020-11-20 | 2021-04-09 | 苏州浪潮智能科技有限公司 | Aerial small target detection method, device, equipment and storage medium |
CN113299363A (en) * | 2021-04-27 | 2021-08-24 | 西安理工大学 | Yoov 5-based dermatology over-the-counter medicine selling method |
CN113537211A (en) * | 2021-06-21 | 2021-10-22 | 电子科技大学 | Deep learning license plate frame positioning method based on asymmetric IOU |
CN114387520A (en) * | 2022-01-14 | 2022-04-22 | 华南农业大学 | Precision detection method and system for intensive plums picked by robot |
CN115439686A (en) * | 2022-08-30 | 2022-12-06 | 一选(浙江)医疗科技有限公司 | Method and system for detecting attention object based on scanned image |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110942446A (en) * | 2019-10-17 | 2020-03-31 | 付冲 | Pulmonary nodule automatic detection method based on CT image |
CN110969623A (en) * | 2020-02-28 | 2020-04-07 | 北京深睿博联科技有限责任公司 | Lung CT multi-symptom automatic detection method, system, terminal and storage medium |
CN111127438A (en) * | 2019-12-25 | 2020-05-08 | 昆明理工大学 | Pulmonary CT image junction detection method based on improved YOLO algorithm |
-
2020
- 2020-10-09 CN CN202011074796.7A patent/CN112184684A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110942446A (en) * | 2019-10-17 | 2020-03-31 | 付冲 | Pulmonary nodule automatic detection method based on CT image |
CN111127438A (en) * | 2019-12-25 | 2020-05-08 | 昆明理工大学 | Pulmonary CT image junction detection method based on improved YOLO algorithm |
CN110969623A (en) * | 2020-02-28 | 2020-04-07 | 北京深睿博联科技有限责任公司 | Lung CT multi-symptom automatic detection method, system, terminal and storage medium |
Non-Patent Citations (3)
Title |
---|
KEJIA XU等: "A New Object Detection Algorithm Based on YOLOv3 for Lung Nodules", 《ICCAI "20:PROCEEDINGS OF THE 2020 6TH INTERNATIONAL CONFERENCE ON COMPUTING AND ARTIFICIAL INTELLIGENCE》 * |
刘学平等: "嵌入SENet结构的改进YOLOV3目标识别算法", 《计算机工程》 * |
孙佳等: "基于改进的YOLOv3网络的实时目标检测", 《激光与光电子学进展》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112633066A (en) * | 2020-11-20 | 2021-04-09 | 苏州浪潮智能科技有限公司 | Aerial small target detection method, device, equipment and storage medium |
CN113299363A (en) * | 2021-04-27 | 2021-08-24 | 西安理工大学 | Yoov 5-based dermatology over-the-counter medicine selling method |
CN113537211A (en) * | 2021-06-21 | 2021-10-22 | 电子科技大学 | Deep learning license plate frame positioning method based on asymmetric IOU |
CN113537211B (en) * | 2021-06-21 | 2023-06-02 | 电子科技大学 | Asymmetric IOU-based deep learning license plate frame positioning method |
CN114387520A (en) * | 2022-01-14 | 2022-04-22 | 华南农业大学 | Precision detection method and system for intensive plums picked by robot |
CN115439686A (en) * | 2022-08-30 | 2022-12-06 | 一选(浙江)医疗科技有限公司 | Method and system for detecting attention object based on scanned image |
CN115439686B (en) * | 2022-08-30 | 2024-01-09 | 一选(浙江)医疗科技有限公司 | Method and system for detecting object of interest based on scanned image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111539930B (en) | Dynamic ultrasonic breast nodule real-time segmentation and identification method based on deep learning | |
CN109461495B (en) | Medical image recognition method, model training method and server | |
CN112184684A (en) | Improved YOLO-v3 algorithm and application thereof in lung nodule detection | |
CN108898595B (en) | Construction method and application of positioning model of focus region in chest image | |
CN105957063B (en) | CT image liver segmentation method and system based on multiple dimensioned weighting similarity measure | |
CN109363698B (en) | Method and device for identifying mammary gland image signs | |
CN112132166B (en) | Intelligent analysis method, system and device for digital cell pathology image | |
CN109363699B (en) | Method and device for identifying focus of breast image | |
CN111340827A (en) | Lung CT image data processing and analyzing method and system | |
CN112365973B (en) | Pulmonary nodule auxiliary diagnosis system based on countermeasure network and fast R-CNN | |
EP1789920A1 (en) | Feature weighted medical object contouring using distance coordinates | |
CN110992377B (en) | Image segmentation method, device, computer-readable storage medium and equipment | |
CN109363697B (en) | Method and device for identifying focus of breast image | |
Farag et al. | Automatic detection and recognition of lung abnormalities in helical CT images using deformable templates | |
Costaridou | Medical image analysis methods | |
CN112581458B (en) | Image processing method and device | |
CN109919254B (en) | Breast density classification method, system, readable storage medium and computer device | |
US8983154B2 (en) | Method for identifying colon polyps in a computed tomography colonography dataset | |
CN113744183A (en) | Pulmonary nodule detection method and system | |
CN111932495B (en) | Medical image detection method, device and storage medium | |
CN113850328A (en) | Non-small cell lung cancer subtype classification system based on multi-view deep learning | |
JP4651370B2 (en) | Radiation image processing method | |
CN114581698A (en) | Target classification method based on space cross attention mechanism feature fusion | |
CN113012127A (en) | Cardiothoracic ratio measuring method based on chest medical image | |
CN112801992A (en) | Pulmonary nodule image classification method based on 3D residual error network |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210105 |