CN110969613A - Intelligent pulmonary tuberculosis identification method and system with image sign interpretation - Google Patents
Intelligent pulmonary tuberculosis identification method and system with image sign interpretation Download PDFInfo
- Publication number
- CN110969613A CN110969613A CN201911238180.6A CN201911238180A CN110969613A CN 110969613 A CN110969613 A CN 110969613A CN 201911238180 A CN201911238180 A CN 201911238180A CN 110969613 A CN110969613 A CN 110969613A
- Authority
- CN
- China
- Prior art keywords
- suspected
- image
- focus
- symptom
- tuberculosis
- 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.)
- Granted
Links
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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10116—X-ray image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
-
- 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
- G06T2207/30096—Tumor; Lesion
Abstract
The embodiment of the invention provides an intelligent identification method and system for tuberculosis with image symptom explanation, wherein the method comprises the following steps: preprocessing an X ray chest film to convert the X ray chest film into a vector diagram; identifying abnormal regions to obtain classification results of whether suspected lesions exist; judging whether a suspected focus is identified in the abnormal area; carrying out classification correction processing to obtain the conditional probability of the suspected focus; judging whether a suspected lesion caused by the pulmonary tuberculosis exists in the abnormal area; processing the suspected region corresponding to the suspected focus on an original image to obtain a sub-image vector diagram; interpreting the abnormal area to obtain image symptom, and judging whether the abnormal area has the tuberculosis or not by obtaining the symptom description; and (4) judging the activity of the pulmonary tuberculosis based on the symptom description. The embodiment of the invention can effectively obtain the internal relation between the image characteristic features in the chest radiography, and compared with the judgment made only based on the image, the embodiment of the invention can better accord with the judgment logic of the iconography, and can also greatly improve the identification precision and efficiency.
Description
Technical Field
The invention relates to the technical field of medical image processing, in particular to an intelligent pulmonary tuberculosis identification method and system with image sign interpretation.
Background
Tuberculosis has attracted attention as an infectious disease that poses a great threat to human health. According to the investigation of the world health organization (WTO), Tuberculosis (TB) caused about one hundred and thirty million deaths in 2017, with the majority of patients located in developing countries. Aiming at the abuse of TB, WTO provides a combined prevention and treatment means, wherein an important link is to carry out X-ray chest radiography screening of the TB. Although the development of chest film screening for TB is of great significance, it is still difficult and serious to perform widespread TB chest film screening due to the fact that chest film images themselves and the general lack of experienced physicians in developing countries. In underdeveloped areas, missing detection, false detection and the like are common in TB chest radiography screening, which is a huge barrier for eliminating the threat of TB to human health in the areas.
In recent years, with the rapid development of artificial intelligence technology, methods represented by deep learning are generally introduced in the field of computer-aided diagnosis to improve the diagnosis accuracy. The development of these techniques has made it possible to design a rapid, accurate screening of chest radiographs for tuberculosis. However, chest film screening for TB has some specificity compared to other areas of image recognition technology. Firstly, the evolution process of TB disease evolution is analyzed, and the evolution of image representation for identifying TB on a chest film has high correlation; secondly, reasonably explaining that the identification result of the chest radiograph after identifying TB meets medical logic; therefore, identifying whether a chest film has TB based on image characterization is an important priority in determining whether a patient is contagious.
At present, a general TB identification method based on artificial intelligence usually adopts a chest film of a patient to make a judgment, such a judgment is usually based on abnormal identification or judgment made by a picture and a model, but ignores the clinical meaning of the abnormal region morphology itself, and these identification methods do not consider the internal relation of lesion evolution contained between image representations in the chest film. The above drawbacks make it difficult for generic methods to give acceptable reasons for identification or determination, and these methods have difficulty in giving acceptable judgment of TB activity because of neglecting the inherent relationship between image characterizations.
Disclosure of Invention
Aiming at the problems in the prior art, the embodiment of the invention provides an intelligent pulmonary tuberculosis identification method and system with image symptom explanation.
In a first aspect, an embodiment of the present invention provides an intelligent tuberculosis identification method with image symptom interpretation, including:
preprocessing an X ray chest film, and converting the X ray chest film into a vector diagram;
carrying out abnormal region identification according to the vector diagram to obtain a classification result of whether the identification result comprises a suspected lesion or not;
judging whether a suspected focus is identified in the abnormal region or not according to the identification result;
carrying out classification correction processing according to the suspected area and the suspected point corresponding to the suspected focus to obtain the conditional probability of the suspected focus;
judging whether a suspected lesion caused by the pulmonary tuberculosis exists in the abnormal region or not according to the conditional probability;
processing the suspected region corresponding to the suspected focus on an original image to obtain a subgraph vector diagram of the suspected focus;
and performing abnormal region interpretation according to the subgraph vector diagram of the suspected focus to obtain a symptom description, judging whether the suspected focus has tuberculosis, and completing intelligent identification of the whole chest radiograph.
Further, the pretreatment process of the X ray chest radiography is specifically as follows:
scaling an X ray chest picture to a picture of 2048X 2048 pixel size by a downsampling mode;
normalizing the scaled picture to a set of 2-dimensional vector maps on [0,1 ];
the abnormal area identification process specifically comprises the following steps: inputting the vector diagram by taking a RetinaNet network as a reference network, and identifying abnormal regions to obtain an identification result of the abnormal regions;
the identification result judgment process specifically comprises the following steps: judging according to the identification result, judging whether the probability of the classification result is greater than a set threshold value, if so, identifying a suspected focus in an abnormal area, and performing classification correction processing; if not, the suspected focus is not identified, and the whole process is terminated.
Further, the specific procedure of the classification correction process is as follows:
constructing a syndrome model, wherein the syndrome model comprises two analysis layers, and the analysis layers comprise a convolution layer, an activation layer and an example standard layer;
coding correction information, and selecting a surrounding area and an area formed in a symmetrical position set radius according to the set radius by taking the suspected point as a center to form a 2-channel tensor;
importing the correction information into the correction submodel to obtain an output probability P (C/O); and randomly taking points within a set distance around the suspected point, repeatedly encoding correction information by taking the points as centers, and importing the information into the syndrome model to obtain the output probability of the syndrome model
And performing conditional probability correction calculation according to the output probability, as shown in the following formula:
wherein, OiIndicating a suspected i-th lesion in the chest radiograph, CiRepresenting the influence of the area around the suspected area and the symmetric area; p (O | C)*) The probability is given, namely the judgment is made on the model under the ideal condition; p (O) is the judgment made by the model on the basis of no conditional reference, namely the probability given by the original RetinaNet network;
the probability of an inverse event is 1-P (O).
Further, the conditional probability determination process specifically includes: judging whether the conditional probability is greater than a preset threshold value or not according to the conditional probability, if so, indicating that the suspected focus does not belong to the tuberculosis focus, and terminating the whole process; if the number is larger than the preset value, the suspected lesion caused by the pulmonary tuberculosis exists, and original image processing is carried out;
the specific process of processing the original image comprises the following steps:
expanding 20 pixels outwards on the original image according to the suspected area corresponding to the suspected focus, and intercepting a subgraph corresponding to the suspected focus;
scaling the subgraph to a 512x 512 pixel picture;
and normalizing the zoomed picture to obtain a sub-picture vector diagram distributed in [0,1 ].
Further, the specific process of the abnormal area interpretation includes the following steps:
constructing a relevance mining model according to image symptom data of the chest radiography to obtain a correction tensor;
according to the sub-image vector diagram, constructing an abnormal region interpretation model for image feature extraction to obtain an image feature analysis tensor;
performing point multiplication on the image feature analysis tensor and the correction tensor to obtain a result tensor;
obtaining the symptom description of the sub-image vector diagram according to the result tensor;
judging the symptom description, if the symptom contains the relevant symptom of the pulmonary tuberculosis, considering that the subgraph vector diagram contains the focus of the pulmonary tuberculosis, and giving out the corresponding symptom;
repeating the above process for other suspected areas in the chest film to obtain all the symptom descriptions in the chest film, and confirming whether the chest film has the tuberculosis.
Further, the specific process of constructing the relevance mining model is as follows:
setting a label for describing image signs according to the image sign characteristics of the chest radiograph, and labeling the suspected focus by adopting the label;
constructing an One-hot matrix and an adjacent matrix reflecting the initial relation of the labels according to the labels;
and constructing a two-layer graph convolution neural network model as a relevance mining model by adopting a graph convolution neural network method, and inputting the one-hot matrix and the adjacent matrix to obtain a correction tensor.
The method for constructing the abnormal region interpretation model specifically comprises the following steps: and constructing an abnormal region interpretation model by adopting the first 15 layers of the ResNet101 network, inputting the sub-image vector diagram to obtain a characteristic tensor, and carrying out a Global Max boosting process on the characteristic tensor to obtain an image characteristic analysis tensor.
Further, an activity judgment model is constructed by adopting a gradient lifting iterative decision tree method, a symptom vector formed by the symptom description is input, the value of each element is 0 or 1, 1 represents that the symptom exists, and if 0 is not the case, the activity judgment model is not the case; the probability of whether the indication is active is obtained.
In a second aspect, an embodiment of the present invention provides an intelligent tuberculosis identification system with image symptom interpretation, including:
chest radiography preprocessing module: the X ray chest film preprocessing device is used for preprocessing the X ray chest film and converting the X ray chest film into a vector diagram;
an abnormal area identification module: the abnormal region recognition is carried out according to the vector diagram to obtain a classification result of whether the recognition result comprises a suspected lesion or not;
and an identification result judgment module: the system is used for judging whether a suspected focus is identified in the abnormal area according to the identification result;
a classification correction processing module: the system is used for carrying out classification correction processing according to the suspected area and the suspected point corresponding to the suspected focus to obtain the conditional probability of the suspected focus;
a conditional probability judging module: the system is used for judging whether a suspected focus caused by the pulmonary tuberculosis exists in the abnormal area or not according to the conditional probability;
original image processing module: the suspected region corresponding to the suspected focus is processed on an original image to obtain a subgraph vector diagram of the suspected focus;
an abnormal region interpretation module: the system is used for performing abnormal area explanation according to the sub-image vector diagram of the suspected focus to obtain the symptom description of the abnormal area explanation, judging whether the pulmonary tuberculosis exists or not, and completing intelligent identification of the whole chest radiograph;
an activity judgment module: and the method is used for constructing an activity judgment model according to the symptom description, judging the activity of the pulmonary tuberculosis symptom and obtaining the activity probability of the pulmonary tuberculosis symptom.
In a third aspect, an embodiment of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method provided in the first aspect when executing the program.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the steps of the method as provided in the first aspect.
According to the intelligent pulmonary tuberculosis identification method and system with the image symptom explanation, the suspected area corresponding to the suspected focus is obtained by identifying the abnormal area in the chest radiograph, the suspected area is further explained based on the image symptom information, the symptom description of the suspected area corresponding to the suspected focus in the chest radiograph can be obtained, whether the suspected focus in the chest radiograph belongs to TB can be confirmed through the symptom description, and reliable activity judgment can be further carried out on the suspected focus. Therefore, the embodiment of the invention can effectively obtain the internal relation between the image characteristic features in the chest film, the constructed abnormal region interpretation model is driven by data, and can better accord with logic compared with judgment made only based on the image, the identification precision and efficiency can be greatly improved, and the effectiveness and accuracy are ensured.
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, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a flowchart of an intelligent tuberculosis identification method according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a classification correction process in the method according to the embodiment of the present invention;
FIG. 3 is a schematic diagram of conditional probability calculation in the method according to the embodiment of the present invention;
fig. 4 is a flowchart illustrating an original image processing procedure in the method according to the embodiment of the present invention;
fig. 5 is a flowchart of an abnormal area interpretation process in the method provided by the embodiment of the present invention;
FIG. 6 is a flowchart of constructing a relevance mining model in the method according to the embodiment of the present invention;
fig. 7 is a schematic diagram of an abnormal area interpretation model constructed in the method provided by the embodiment of the invention;
FIG. 8 is a schematic diagram of an intelligent tuberculosis identification system according to an embodiment of the present invention;
fig. 9 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a flowchart of an intelligent tuberculosis identification method with image symptom interpretation according to an embodiment of the present invention, and as shown in fig. 1, the method includes:
step S1: the X ray chest film is preprocessed and converted into a vector diagram. Specifically, the X ray chest radiograph is converted into a set of 2-dimensional vector images distributed on [0,1 ].
In step S1 of the embodiment of the present invention, the preprocessing process of the X ray chest film is specifically as follows:
the X ray chest slices are scaled to a 2048X 2048 pixel size picture by a downsampling manner.
The scaled pictures are normalized to a set of 2-dimensional vector maps on [0,1 ].
In the above, since the input of the network is limited by hardware conditions, and cannot support the input of pictures with any size, there must be a limit to the size of the image, and too large or too small will adversely affect the abnormal region (i.e. too small will result in too small focus, and too large will result in too large hardware overhead), and the effect of scaling the chest radiography to 2048 × 2048 size is the best by comprehensively considering all the factors, i.e. the effect is the best under the condition of general hardware support. And the zoomed picture is normalized, so that the network learning is easier and the interference can be effectively eliminated.
Step S2: and according to the vector diagram of the chest picture, identifying abnormal regions to obtain an identification result.
In step S2 of the embodiment of the present invention, the abnormal area identification process specifically includes:
and inputting the vector diagram by taking a RetinaNet network as a reference network, and identifying an abnormal region to obtain an identification result of the abnormal region, wherein the identification result comprises a classification result of whether the suspected lesion is in the abnormal region and a position coordinate matched with the classification result.
In step S2 of the embodiment of the present invention, a learning training is performed using an existing RetinaNet network (reference Lin, Tsung-Yi, et. "local loss for noise object detection." Proceedings of the international conference on computer vision.2017 ") as a reference network, and only the vector map is input, so that a recognition result for an abnormal region can be obtained. The RetinaNet network can effectively utilize image information with different scales and layers, and has a simple structure and easy operation. Two branches in the RetinaNet network provide the identification result, and the two branches generate corresponding results based on the same characteristic information, namely a batch of abnormal regions can be detected for subsequent screening, and the abnormal information can be caused by other diseases, so that further judgment and identification are needed.
Step S3: judging according to the identification result, judging whether the probability of the classification result is greater than a set threshold value, if so, identifying a suspected focus in an abnormal area, and carrying out the next step; if not, the suspected focus is not identified, and the whole process is terminated.
In step S3, a simple preliminary screening is performed on the recognition result, and due to individual differences, lung textures are unlikely to be similar, and for general individuals, there should be a certain low degree of abnormality, and the abnormality is reflected by individual differences rather than actual lesions, so a low threshold is used for filtering, interference caused by individual differences is eliminated as much as possible, and at the same time, the data size of subsequent steps is reduced, and the accuracy of recognition is improved.
In the embodiment of the present invention, if no suspected lesion is identified in the abnormal region through the above process, it indicates that there is no lesion region caused by TB in the chest radiograph, that is, no TB is identified, and the overall process may be terminated. If the suspected lesion is identified in the abnormal area, the suspected area needs to be further corrected by combining the environmental information (including the corresponding position and the surrounding information) of the suspected area with respect to the suspected area corresponding to the suspected lesion. Therefore, the intelligent tuberculosis identification method provided by the embodiment of the invention further comprises the following steps:
step S4: and carrying out classification correction processing according to the suspected area and the suspected point corresponding to the suspected focus to obtain the conditional probability of the suspected focus.
In step S4 of the embodiment of the present invention, referring to fig. 2, the specific process of the classification correction process is as follows:
step S41: a correction submodel is constructed, the model comprising two resolution layers, the resolution layers comprising a convolutional layer (Conv), an activation layer (ReLu) and an instance specification layer (IN). The number of the analysis layers can be increased or decreased according to actual needs, and two layers are preferred in the embodiment of the invention, so that the correction effect is optimal.
Step S42: and coding correction information, namely, forming a 2-channel tensor by using the area formed by setting the radius around the pseudo point and the area formed by setting the radius at the symmetrical position of the pseudo point as the center. Further, the set radius is preferably 10 pixels, which can collect enough surrounding information without increasing excessive hardware consumption.
Step S43: importing the correction information into the correction submodel to obtain an output probability P (C/O); and randomly taking points within a set distance around the suspected point, and repeating the process by taking the points as a center to obtain the output probability P (C/O) of the syndrome model. Further, the set distance is preferably 5pixel radii around the suspected point.
Step S44: and performing a conditional probability correction calculation according to the output probability, and referring to fig. 3, obtaining the conditional probability as shown in the following formula:
wherein, OiIndicating a suspected i-th lesion in the chest radiograph, CiRepresenting the influence of the area around the suspected area and the symmetric area; p (O | C)*) The probability is given, namely the judgment is made on the model under the ideal condition; p (O) is the judgment made by the model on the basis of no conditional reference, namely the probability given by the original RetinaNet network;
probability of an inverse event, namely 1-P (O); p (C | O) is the output probability obtained by encoding the above conditions (surrounding area, symmetric area) on the basis of the probability given by the RetinaNet network;
the output probability of the submodel is corrected in the event of the inverse event, and the result of replacing the adjacent area of the suspected area is used here.
In step S4 of the embodiment of the present invention, when determining whether a suspected region is a TB lesion region, the surrounding tissue morphology and the tissue morphology at a symmetric position of the region are referred to, for example, if the suspected lesion is found in the left lung, the tissue morphology of the corresponding region around the left lung and the right lung needs to be observed to make a comprehensive determination, and this process is mathematically described as a conditional probability, so that for a lesion, the final optimized conditional probability function is as shown in the above formula. By comprehensively judging by means of the surrounding area and the symmetrical area of the suspected focus, the false positive can be effectively reduced, and the probability of the real focus is improved and more prominent.
Step S5: judging whether the conditional probability is greater than a preset threshold value or not according to the conditional probability, if so, indicating that the suspected focus does not belong to the TB focus, and terminating the whole process; if the abnormal area is larger than the preset abnormal area, the suspected focus caused by the TB exists in the abnormal area, and the next step is carried out.
In steps S4-S5 according to the embodiment of the present invention, the suspected lesion identified in the abnormal region is further classified and corrected, and conditional judgment is performed based on environmental information that includes the surrounding position and the symmetric position of the suspected region to obtain conditional probability, so as to further determine whether a lesion region caused by TB exists in the abnormal region. If the abnormal area is eliminated, which indicates that no TB is identified, the whole process can be terminated; if a suspected lesion is still identified, it is necessary to further confirm whether the suspected lesion belongs to TB. Therefore, the intelligent tuberculosis identification method provided by the embodiment of the invention further comprises the following steps:
step S6: and processing the suspected region corresponding to the suspected focus on an original drawing to obtain a sub-picture vector diagram of the suspected focus.
In step S6 of the embodiment of the present invention, referring to fig. 4, the specific process of processing the original image includes the following steps:
step S61: and expanding 20 pixels outwards on the original image according to the suspected area corresponding to the suspected focus, and intercepting a subgraph corresponding to the suspected focus.
Step S62: the subgraph is scaled to a 512x 512 pixel picture.
Step S63: and normalizing the zoomed picture to obtain a sub-picture vector diagram distributed in [0,1 ].
In step S6, in order to ensure that the inherent deficiency of the neural network convolution calculation can be effectively avoided, the suspected region is expanded outward, so that more additional information can be introduced, and the suspected lesion is ensured to be located at the center of the sub-image, thereby avoiding the influence of poor effect of the neural network on processing the information at the edge of the image. Expand 20 pixels and scale to 512x 512 pixels, which works best.
Step S7: according to the sub-image vector diagram of the suspected focus, the abnormal area is explained to obtain a sign description to confirm whether the suspected focus has the tuberculosis focus or not, and the intelligent tuberculosis identification of the whole chest film is completed.
In step S7 of the embodiment of the present invention, referring to fig. 5, a specific process of the abnormal area interpretation includes the following steps:
step S71: and constructing a relevance mining model according to the image symptom data of the chest radiography to obtain a correction tensor.
In step S71 in the embodiment of the present invention, referring to fig. 6, the building of the relevance mining model specifically includes the following steps:
step S711: and setting a label for describing the image symptoms according to the image symptom characteristics of the chest radiography, and labeling the suspected focus by adopting the label.
In step S711 of the embodiment of the present invention, 17 types of the tags are set, respectively: fibrosis, calcification, exudation, foci proliferation, coalescent bulbs, cavities, millet marks, pleural thickening, pleural effusion, tree bud marks, ground glass, interstitial lesions, secondary bronchiectasis, secondary cavities, micro-nodules, caseous lesions or others. The suspected lesion is located and marked with one or more of the 17 labels according to its image characteristics.
Step S712: and constructing an One-hot matrix and an adjacent matrix reflecting the initial relation of the labels according to the labels.
In step S712 of the embodiment of the present invention, the One-hot matrix has 17 × 17 dimensions, and rows and columns represent the values of the labels, for example, the 0 th row and the 0 th column represent fibrosis symptoms, the 1 st row and the 1 st column represent calcification symptoms, and so on. The adjacency matrix is also 17 × 17 dimensions, and the element at each position represents the normalized conditional probability of each symptom, for example, the row 1, the column 0 indicate the normalized probability that a calcification label exists in the image labeled with the calcification label in the training data, and the row 0, the column 1 indicate the normalized probability that the calcification label exists in the image labeled with the calcification label in the training data.
Step S713: and constructing a two-layer graph convolution neural network model as a relevance mining model by adopting a graph convolution neural network method, and inputting the one-hot matrix and the adjacent matrix to obtain a correction tensor.
In step S713, the obtained correction tensors are subjected to learning correction by a reverse derivation method of a neural network, and the set of correction tensors is used to correct the output tensors of the subsequent image feature extraction model. The number of layers of the graph convolution neural network model can be increased or decreased according to actual needs, and two layers are optimized according to the embodiment of the invention, so that the effect is optimal.
Step S72: and constructing an abnormal region interpretation model according to the sub-image vector diagram obtained in the step S6 to extract image features, so as to obtain an image feature analysis tensor.
The method for constructing the abnormal region interpretation model in step S72 includes: referring to fig. 7, an abnormal region interpretation model is constructed by using the first 15 layers of the ResNet101 network (refer to Deep reactive Learning for Image Recognition (He, Kaiming, et al. "Deep reactive Learning for Image Recognition." Proceedings of the ieee reference on computer vision and pattern Recognition. 2016)), the vector diagram is input to obtain a feature tensor, and the feature tensor is subjected to a Global map boosting process to obtain an Image feature analysis tensor.
Step S73: and performing point multiplication on the image feature analysis tensor and the correction tensor to obtain a result tensor.
In step S73 of the embodiment of the present invention, the result tensor form is n x 17, n is the number of input images, 17 represents the scores given to the 17 kinds of symptom labels designed above, and the threshold used here is 0, that is, if a score of a certain item in the 17 values is greater than 0, it indicates that the symptom represented by the item exists in the sub-image vector diagram.
Step S74: and obtaining the symptom description of the sub-image vector diagram according to the result tensor.
Step S75: judging the symptom description, if the symptom is other, considering that the subgraph vector diagram is a similar case instead of Tuberculosis (TB), and terminating the whole process; and if the symptoms have TB related symptoms, the sub-image vector diagram is considered to have the TB focus, and corresponding symptoms are given.
Step S76: and repeating the process aiming at other suspected areas in the chest picture to obtain all the symptom descriptions in the chest picture, and finishing the identification of the whole chest picture.
In steps S75-S76 of the embodiment of the invention, TB focus points and position coordinates corresponding to the TB focus points can be obtained through judging and determining whether the TB focus exists, so that the intelligent identification of TB is completed, and the symptom descriptions of other suspected areas are sequentially obtained, namely the intelligent identification of the whole chest radiograph is completed.
In the above, the intrinsic information between the types of the lesions is obtained based on the graph convolution neural network, and an image-to-text label process is performed based on the image recognition method to obtain the label describing the lesion area, so that the reason that whether the model recognizes the lesions is explained, and the false positive elimination and the result checking are facilitated. In addition, the method of using the graph convolution neural network can effectively use the characters except the image to describe the information, and compared with the method of simply generating the character label from the image angle, the method is reliable in accuracy and logic.
According to the intelligent identification method provided by the embodiment of the invention, the suspected area corresponding to the suspected focus is obtained by identifying the abnormal area, the suspected area is further explained based on the image symptom information, the symptom description of the suspected area corresponding to the suspected focus in the chest radiograph can be obtained, and whether the suspected focus in the chest radiograph belongs to TB or not can be confirmed through the symptom description.
In the embodiment of the invention, because the active tuberculosis is infectious and needs to be treated differently, the active diagnosis of TB is the key point of TB control. Therefore, the intelligent tuberculosis identification method provided by the embodiment of the invention further comprises the following steps:
step S8: and constructing an activity judgment model according to the symptom description, and judging the activity of the TB symptom to obtain the activity probability of the TB symptom.
The method specifically comprises the following steps: constructing all symptom descriptions into a 1x 17 symptom vector, constructing an activity judgment model by adopting a gradient lifting iterative decision tree method (GBDT), inputting the symptom vector, wherein the value of each element is 0 or 1, 1 represents that the symptom exists, and if 0 is not the case, the activity judgment model is constructed; the probability of whether the indication is active is obtained.
In step S8, the activity probability of tuberculosis in chest radiography identification is determined by constructing an activity determination model based on the symptom description of the lesion features, and the accuracy and generalization ability thereof are more reliable and accurate than the determination of the activity of tuberculosis directly obtained from the image.
In the intelligent identification method provided by the embodiment of the invention, two processes are totally adopted to judge whether TB exists in one X ray chest film, namely if no suspected focus is identified in the identification process of the abnormal area, the abnormal area explanation is not carried out at the moment, and the chest film is considered to have no TB; if the abnormal area identification process identifies a suspected lesion, but the symptom description obtained in the abnormal area interpretation process is other than the others, it is also assumed that the chest radiograph should be of a similar disease species rather than TB. Therefore, only if a suspected lesion is identified in the abnormal region identification process and the symptom obtained in the abnormal region interpretation process is described as TB, the suspected lesion existing in the chest film is determined as TB, and the accuracy of intelligent identification for TB identification in the chest film can be ensured.
Based on any of the above embodiments, fig. 8 is a schematic structural diagram of an intelligent tuberculosis identification system with image symptom interpretation according to an embodiment of the present invention, as shown in fig. 8, the system includes:
chest radiography preprocessing module: used for preprocessing the X ray chest film to be converted into a vector diagram.
An abnormal area identification module: and the abnormal region identification is carried out according to the vector diagram to obtain a classification result of whether the identification result comprises the suspected lesion or not.
And an identification result judgment module: and the method is used for judging whether the suspected focus is identified in the abnormal area according to the identification result.
A classification correction processing module: and the method is used for carrying out classification correction processing according to the suspected area and the suspected point corresponding to the suspected focus to obtain the conditional probability of the suspected focus.
A conditional probability judging module: and the method is used for judging whether a suspected lesion caused by the pulmonary tuberculosis exists in the abnormal region according to the conditional probability.
Original image processing module: and the method is used for processing the suspected region corresponding to the suspected focus on an original drawing to obtain a sub-image vector diagram of the suspected focus.
An abnormal region interpretation module: and the system is used for performing abnormal region interpretation to obtain the symptom description according to the sub-image vector diagram of the suspected focus, confirming whether the suspected focus has the tuberculosis focus or not and finishing the intelligent identification of the whole chest radiograph.
An activity judgment module: and the activity judgment module is used for constructing an activity judgment model according to the symptom description, judging the activity of the TB symptom and obtaining the activity probability of the TB symptom.
Specifically, the system provided in the embodiment of the present invention is specifically configured to execute the method embodiment described above, and details of the method embodiment of the present invention are not described again. According to the system provided by the embodiment of the invention, the suspected area corresponding to the suspected focus is obtained by identifying the abnormal area, and the suspected area is further explained based on the image symptom information, so that the symptom description of the suspected area corresponding to the suspected focus in the chest radiograph can be obtained, whether the suspected focus in the chest radiograph belongs to TB or not can be confirmed through the symptom description, and the system is simple, reliable and more logical.
In summary, the method and system provided by the embodiments of the present invention can effectively obtain the internal relationship between the image characteristic features in the chest radiograph based on the information of the image characteristic features, the constructed abnormal region interpretation model is driven by data, and compared with the judgment made only based on the image itself, the method and system can better meet the logic, can also greatly improve the identification precision and efficiency, and can ensure the effectiveness and accuracy of identification. In addition, the embodiment of the invention can also carry out reliable activity judgment according to the symptom description, and further ensures the reliability and the accuracy of the identification of the tuberculosis in the chest radiograph.
Fig. 9 is a schematic entity structure diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 9, the electronic device may include: a processor (processor)301, a communication Interface (communication Interface)302, a memory (memory)303 and a communication bus 304, wherein the processor 301, the communication Interface 302 and the memory 303 complete communication with each other through the communication bus 304. The processor 301 may invoke a computer program stored on the memory 303 and executable on the processor 301 to perform the methods provided by the various embodiments described above, including, for example:
preprocessing an X ray chest film, and converting the X ray chest film into a vector diagram;
carrying out abnormal region identification according to the vector diagram to obtain a classification result of whether the identification result comprises a suspected lesion or not;
judging whether a suspected focus is identified in the abnormal region or not according to the identification result;
carrying out classification correction processing according to the suspected area and the suspected point corresponding to the suspected focus to obtain the conditional probability of the suspected focus;
judging whether a suspected lesion caused by the pulmonary tuberculosis exists in the abnormal region or not according to the conditional probability;
processing the suspected region corresponding to the suspected focus on an original image to obtain a subgraph vector diagram of the suspected focus;
and performing abnormal region interpretation according to the subgraph vector diagram of the suspected focus to obtain a symptom description, judging whether the suspected focus has tuberculosis, and completing intelligent identification of the whole chest radiograph.
In addition, the logic instructions in the memory 303 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solutions of the embodiments of the present invention may be essentially implemented or make a contribution to the prior art, or may be implemented in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Embodiments of the present invention further provide a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented to perform the method provided in the foregoing embodiments when executed by a processor, and the method includes:
preprocessing an X ray chest film, and converting the X ray chest film into a vector diagram;
carrying out abnormal region identification according to the vector diagram to obtain a classification result of whether the identification result comprises a suspected lesion or not;
judging whether a suspected focus is identified in the abnormal region or not according to the identification result;
carrying out classification correction processing according to the suspected area and the suspected point corresponding to the suspected focus to obtain the conditional probability of the suspected focus;
judging whether a suspected lesion caused by the pulmonary tuberculosis exists in the abnormal region or not according to the conditional probability;
processing the suspected region corresponding to the suspected focus on an original image to obtain a subgraph vector diagram of the suspected focus;
and performing abnormal region interpretation according to the subgraph vector diagram of the suspected focus to obtain a symptom description, judging whether the suspected focus has tuberculosis, and completing intelligent identification of the whole chest radiograph.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. An intelligent identification method for tuberculosis with image symptom explanation is characterized by comprising the following steps:
preprocessing an X ray chest film, and converting the X ray chest film into a vector diagram;
carrying out abnormal region identification according to the vector diagram to obtain a classification result of whether the identification result comprises a suspected lesion or not;
judging whether a suspected focus is identified in the abnormal region or not according to the identification result;
carrying out classification correction processing according to the suspected area and the suspected point corresponding to the suspected focus to obtain the conditional probability of the suspected focus;
judging whether a suspected lesion caused by the pulmonary tuberculosis exists in the abnormal region or not according to the conditional probability;
processing the suspected region corresponding to the suspected focus on an original image to obtain a subgraph vector diagram of the suspected focus;
and performing abnormal region interpretation according to the subgraph vector diagram of the suspected focus to obtain a symptom description, judging whether the suspected focus has tuberculosis, and completing intelligent identification of the whole chest radiograph.
2. The intelligent tuberculosis identification method according to claim 1, wherein the preprocessing process of the X ray chest radiograph is as follows:
scaling an X ray chest picture to a picture of 2048X 2048 pixel size by a downsampling mode;
normalizing the scaled picture to a set of 2-dimensional vector maps on [0,1 ];
the abnormal area identification process specifically comprises the following steps: inputting the vector diagram by taking a RetinaNet network as a reference network, and identifying abnormal regions to obtain an identification result of the abnormal regions;
the identification result judgment process specifically comprises the following steps: judging according to the identification result, judging whether the probability of the classification result is greater than a set threshold value, if so, identifying a suspected focus in an abnormal area, and performing classification correction processing; if not, the suspected focus is not identified, and the whole process is terminated.
3. The intelligent tuberculosis identification method according to claim 2, wherein the classification correction process comprises the following specific procedures:
constructing a syndrome model, wherein the syndrome model comprises two analysis layers, and the analysis layers comprise a convolution layer, an activation layer and an example standard layer;
coding correction information, and selecting a surrounding area and an area formed in a symmetrical position set radius according to the set radius by taking the suspected point as a center to form a 2-channel tensor;
importing the correction information into the correction submodel to obtain an output probability P (C/O); and randomly taking points within a set distance around the suspected point, repeatedly encoding correction information by taking the points as centers, and importing the information into the syndrome model to obtain the output probability of the syndrome model
And performing conditional probability correction calculation according to the output probability, as shown in the following formula:
wherein, OiIndicating a suspected i-th lesion in the chest radiograph, CiRepresenting the influence of the area around the suspected area and the symmetric area; p (O | C)*) The probability is given, namely the judgment is made on the model under the ideal condition; p (O) is the judgment made by the model on the basis of no conditional reference, namely the probability given by the original RetinaNet network;
the probability of an inverse event is 1-P (O).
4. The intelligent tuberculosis identification method according to claim 3, wherein the conditional probability judgment process specifically comprises: judging whether the conditional probability is greater than a preset threshold value or not according to the conditional probability, if so, indicating that the suspected focus does not belong to the tuberculosis focus, and terminating the whole process; if the number is larger than the preset value, the suspected lesion caused by the pulmonary tuberculosis exists, and original image processing is carried out;
the specific process of processing the original image comprises the following steps:
expanding 20 pixels outwards on the original image according to the suspected area corresponding to the suspected focus, and intercepting a subgraph corresponding to the suspected focus;
scaling the subgraph to a 512x 512 pixel picture;
and normalizing the zoomed picture to obtain a sub-picture vector diagram distributed in [0,1 ].
5. The intelligent tuberculosis identification method according to claim 4, wherein the specific process of interpreting the abnormal area comprises the following steps:
constructing a relevance mining model according to image symptom data of the chest radiography to obtain a correction tensor;
according to the sub-image vector diagram, constructing an abnormal region interpretation model for image feature extraction to obtain an image feature analysis tensor;
performing point multiplication on the image feature analysis tensor and the correction tensor to obtain a result tensor;
obtaining the symptom description of the sub-image vector diagram according to the result tensor;
judging the symptom description, if the symptom contains the relevant symptom of the pulmonary tuberculosis, considering that the subgraph vector diagram contains the focus of the pulmonary tuberculosis, and giving out the corresponding symptom;
repeating the above process for other suspected areas in the chest film to obtain all the symptom descriptions in the chest film, and confirming whether the chest film has the tuberculosis.
6. The intelligent tuberculosis identification method according to claim 5, further comprising: the specific process for constructing the relevance mining model is as follows:
setting a label for describing image signs according to the image sign characteristics of the chest radiograph, and labeling the suspected focus by adopting the label;
constructing an One-hot matrix and an adjacent matrix reflecting the initial relation of the labels according to the labels;
and constructing a two-layer graph convolution neural network model as a relevance mining model by adopting a graph convolution neural network method, and inputting the one-hot matrix and the adjacent matrix to obtain a correction tensor.
The method for constructing the abnormal region interpretation model specifically comprises the following steps: and constructing an abnormal region interpretation model by adopting the first 15 layers of the ResNet101 network, inputting the sub-image vector diagram to obtain a characteristic tensor, and carrying out a Global Max Pooling process on the characteristic tensor to obtain an image characteristic analysis tensor.
7. The intelligent tuberculosis identification method according to claim 6, further comprising: constructing an activity judgment model by adopting a gradient lifting iterative decision tree method, inputting a symptom vector formed by the symptom description, wherein the value of each element is 0 or 1, 1 represents that the symptom exists, and if 0 is not the case, the other element is not the case; the probability of whether the indication is active is obtained.
8. An intelligent tuberculosis identification system with image symptom interpretation, comprising:
chest radiography preprocessing module: the X ray chest film preprocessing device is used for preprocessing the X ray chest film and converting the X ray chest film into a vector diagram;
an abnormal area identification module: the abnormal region recognition is carried out according to the vector diagram to obtain a classification result of whether the recognition result comprises a suspected lesion or not;
and an identification result judgment module: the system is used for judging whether a suspected focus is identified in the abnormal area according to the identification result;
a classification correction processing module: the system is used for carrying out classification correction processing according to the suspected area and the suspected point corresponding to the suspected focus to obtain the conditional probability of the suspected focus;
a conditional probability judging module: the system is used for judging whether a suspected focus caused by the pulmonary tuberculosis exists in the abnormal area or not according to the conditional probability;
original image processing module: the suspected region corresponding to the suspected focus is processed on an original image to obtain a subgraph vector diagram of the suspected focus;
an abnormal region interpretation module: the system is used for performing abnormal area explanation according to the sub-image vector diagram of the suspected focus to obtain the symptom description of the abnormal area explanation, judging whether the pulmonary tuberculosis exists or not, and completing intelligent identification of the whole chest radiograph;
an activity judgment module: and the method is used for constructing an activity judgment model according to the symptom description, judging the activity of the pulmonary tuberculosis symptom and obtaining the activity probability of the pulmonary tuberculosis symptom.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method according to any of claims 1 to 7 are implemented when the processor executes the program.
10. A non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911238180.6A CN110969613B (en) | 2019-12-06 | 2019-12-06 | Intelligent pulmonary tuberculosis identification method and system with image symptom explanation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911238180.6A CN110969613B (en) | 2019-12-06 | 2019-12-06 | Intelligent pulmonary tuberculosis identification method and system with image symptom explanation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110969613A true CN110969613A (en) | 2020-04-07 |
| CN110969613B CN110969613B (en) | 2023-04-14 |
Family
ID=70033134
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911238180.6A Active CN110969613B (en) | 2019-12-06 | 2019-12-06 | Intelligent pulmonary tuberculosis identification method and system with image symptom explanation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110969613B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507979A (en) * | 2020-05-08 | 2020-08-07 | 延安大学 | Computer-aided analysis method for medical image |
| CN111754485A (en) * | 2020-06-24 | 2020-10-09 | 成都市温江区人民医院 | Artificial intelligence ultrasonic auxiliary system for liver |
| CN111767952A (en) * | 2020-06-30 | 2020-10-13 | 重庆大学 | Interpretable classification method for benign and malignant pulmonary nodules |
| CN112634226A (en) * | 2020-12-18 | 2021-04-09 | 平安科技（深圳）有限公司 | Head CT image detection device, method, electronic device, and storage medium |
| CN113257412A (en) * | 2021-06-16 | 2021-08-13 | 腾讯科技（深圳）有限公司 | Information processing method, information processing device, computer equipment and storage medium |
| CN112634226B (en) * | 2020-12-18 | 2024-05-14 | 平安科技（深圳）有限公司 | Head CT image detection device, method, electronic device and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800805A (en) * | 2019-01-14 | 2019-05-24 | 上海联影智能医疗科技有限公司 | Image processing system and computer equipment based on artificial intelligence |
| CN110175993A (en) * | 2019-05-27 | 2019-08-27 | 西安交通大学医学院第一附属医院 | A kind of Faster R-CNN pulmonary tuberculosis sign detection system and method based on FPN |
| JP2019141478A (en) * | 2018-02-23 | 2019-08-29 | 株式会社日立製作所 | Image processing apparatus, method and program |
-
2019
- 2019-12-06 CN CN201911238180.6A patent/CN110969613B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2019141478A (en) * | 2018-02-23 | 2019-08-29 | 株式会社日立製作所 | Image processing apparatus, method and program |
| CN109800805A (en) * | 2019-01-14 | 2019-05-24 | 上海联影智能医疗科技有限公司 | Image processing system and computer equipment based on artificial intelligence |
| CN110175993A (en) * | 2019-05-27 | 2019-08-27 | 西安交通大学医学院第一附属医院 | A kind of Faster R-CNN pulmonary tuberculosis sign detection system and method based on FPN |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507979A (en) * | 2020-05-08 | 2020-08-07 | 延安大学 | Computer-aided analysis method for medical image |
| CN111754485A (en) * | 2020-06-24 | 2020-10-09 | 成都市温江区人民医院 | Artificial intelligence ultrasonic auxiliary system for liver |
| CN111767952A (en) * | 2020-06-30 | 2020-10-13 | 重庆大学 | Interpretable classification method for benign and malignant pulmonary nodules |
| CN111767952B (en) * | 2020-06-30 | 2024-03-29 | 重庆大学 | Interpretable lung nodule benign and malignant classification method |
| CN112634226A (en) * | 2020-12-18 | 2021-04-09 | 平安科技（深圳）有限公司 | Head CT image detection device, method, electronic device, and storage medium |
| CN112634226B (en) * | 2020-12-18 | 2024-05-14 | 平安科技（深圳）有限公司 | Head CT image detection device, method, electronic device and storage medium |
| CN113257412A (en) * | 2021-06-16 | 2021-08-13 | 腾讯科技（深圳）有限公司 | Information processing method, information processing device, computer equipment and storage medium |
| CN113257412B (en) * | 2021-06-16 | 2022-02-11 | 腾讯科技（深圳）有限公司 | Information processing method, information processing device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110969613B (en) | 2023-04-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110969613B (en) | Intelligent pulmonary tuberculosis identification method and system with image symptom explanation | |
| CN108648172B (en) | CT (computed tomography) map pulmonary nodule detection system based on 3D-Unet | |
| US10482603B1 (en) | Medical image segmentation using an integrated edge guidance module and object segmentation network | |
| US20190191988A1 (en) | Screening method for automated detection of vision-degenerative diseases from color fundus images | |
| Xu et al. | Dual-channel residual network for hyperspectral image classification with noisy labels | |
| CN111144486B (en) | Heart nuclear magnetic resonance image key point detection method based on convolutional neural network | |
| CN111754453A (en) | Pulmonary tuberculosis detection method and system based on chest radiography image and storage medium | |
| US11935213B2 (en) | Laparoscopic image smoke removal method based on generative adversarial network | |
| Wu et al. | BA-GCA net: boundary-aware grid contextual attention net in osteosarcoma MRI image segmentation | |
| Balasubramaniam et al. | Optimal Ensemble learning model for COVID-19 detection using chest X-ray images | |
| CN111863232B (en) | Remote disease intelligent diagnosis system based on block chain and medical image | |
| CN114926892A (en) | Fundus image matching method and system based on deep learning and readable medium | |
| Aria et al. | ADA-COVID: adversarial deep domain adaptation-based diagnosis of COVID-19 from lung CT scans using triplet embeddings | |
| CN111260670B (en) | Tubular structure segmentation graph fracture repairing method and system of three-dimensional image based on deep learning network | |
| CN113658165B (en) | Cup/disc ratio determining method, device, equipment and storage medium | |
| CN114998615A (en) | Deep learning-based collaborative significance detection method | |
| Singh et al. | Optimized convolutional neural network for glaucoma detection with improved optic-cup segmentation | |
| Li et al. | Developing an image-based deep learning framework for automatic scoring of the pentagon drawing test | |
| CN117218453A (en) | Incomplete multi-mode medical image learning method | |
| CN113379716B (en) | Method, device, equipment and storage medium for predicting color spots | |
| Soni et al. | Convolutional neural network based CT scan classification method for COVID-19 test validation | |
| CN113033305B (en) | Living body detection method, living body detection device, terminal equipment and storage medium | |
| Zaeri | AI Modeling to Combat COVID-19 Using CT Scan Imaging Algorithms and Simulations: A Study | |
| CN111178453A (en) | SVM-based medical image intelligent matching method and storage medium | |
| Shen et al. | Unsupervised human skeleton extraction from Kinect depth images |
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 |