CN114926439B - Bone growth point delineating method and device, storage medium and processor - Google Patents
Bone growth point delineating method and device, storage medium and processor Download PDFInfo
- Publication number
- CN114926439B CN114926439B CN202210571515.1A CN202210571515A CN114926439B CN 114926439 B CN114926439 B CN 114926439B CN 202210571515 A CN202210571515 A CN 202210571515A CN 114926439 B CN114926439 B CN 114926439B
- Authority
- CN
- China
- Prior art keywords
- image
- target
- layer
- initial
- result
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000008468 bone growth Effects 0.000 title claims abstract description 192
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000012545 processing Methods 0.000 claims abstract description 107
- 230000012010 growth Effects 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims description 57
- 230000005284 excitation Effects 0.000 claims description 18
- 238000001125 extrusion Methods 0.000 claims description 18
- 230000021332 multicellular organism growth Effects 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 abstract description 6
- 230000006870 function Effects 0.000 description 34
- 230000008569 process Effects 0.000 description 17
- 238000004590 computer program Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 8
- 238000001959 radiotherapy Methods 0.000 description 8
- 210000000988 bone and bone Anatomy 0.000 description 5
- 238000011161 development Methods 0.000 description 5
- 230000018109 developmental process Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 210000003128 head Anatomy 0.000 description 4
- 210000002745 epiphysis Anatomy 0.000 description 3
- 210000002758 humerus Anatomy 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 206010028980 Neoplasm Diseases 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 208000012191 childhood neoplasm Diseases 0.000 description 2
- 210000000527 greater trochanter Anatomy 0.000 description 2
- 210000004095 humeral head Anatomy 0.000 description 2
- 210000000528 lesser trochanter Anatomy 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 210000000492 nasalseptum Anatomy 0.000 description 2
- 210000003049 pelvic bone Anatomy 0.000 description 2
- 210000004061 pubic symphysis Anatomy 0.000 description 2
- 210000001991 scapula Anatomy 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 230000009645 skeletal growth Effects 0.000 description 2
- 206010067484 Adverse reaction Diseases 0.000 description 1
- 208000032170 Congenital Abnormalities Diseases 0.000 description 1
- 206010061619 Deformity Diseases 0.000 description 1
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 1
- 206010058314 Dysplasia Diseases 0.000 description 1
- 206010023509 Kyphosis Diseases 0.000 description 1
- 208000031481 Pathologic Constriction Diseases 0.000 description 1
- 206010034156 Pathological fracture Diseases 0.000 description 1
- 208000005250 Spontaneous Fractures Diseases 0.000 description 1
- 210000002659 acromion Anatomy 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000006838 adverse reaction Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000002512 chemotherapy Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000004821 effect on bone Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 210000002414 leg Anatomy 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000035764 nutrition Effects 0.000 description 1
- 235000016709 nutrition Nutrition 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 206010039722 scoliosis Diseases 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000036262 stenosis Effects 0.000 description 1
- 208000037804 stenosis Diseases 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 210000000689 upper leg Anatomy 0.000 description 1
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
- G06T2207/10072—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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30008—Bone
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The application discloses a method and a device for outlining a bone growth point, a storage medium and a processor. The method comprises the following steps: obtaining a target image, wherein the target image comprises: a target layer CT image to be sketched, wherein a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image; and inputting the CT images of the target layer to be sketched, namely the CT image of the upper layer of the CT images of the target layer and the CT image of the lower layer of the CT images of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of skeleton growth points in the CT images of the target layer. The application solves the problem of lower accuracy of automatic sketching of the bone growth points caused by complex structure of the bone growth points and difficult identification on CT images in the related technology.
Description
Technical Field
The application relates to the technical field of medical image processing, in particular to a method and a device for delineating bone growth points, a storage medium and a processor.
Background
In recent years, with the comprehensive application of various treatment modes such as surgery, chemotherapy, radiotherapy and the like, the survival time of the children tumor patients is obviously prolonged. But the quality of life of many survivors of pediatric tumor patients is affected by treatment-related adverse effects, resulting in treatment-related disability or life-threatening long-term complications. Radiotherapy is an important treatment means for children tumors, bones of children are in the development process, and compared with adult patients, the radiotherapy has more serious inhibition effect on bones of children patients in the growth and development stage. The radiotherapy-related skeletal long-term adverse reactions are related to patient exposure dose and age, including facial dysplasia, orbital defects, collarbone stenosis, arm length differences, leg length differences, asymmetry, deformity, pathological fracture, scoliosis, kyphosis, and the like. Therefore, there is a need to protect the growth and development of organs such as bones of children without affecting the target dose for pediatric tumor patients receiving radiation therapy.
The automatic sketching model of the bone growth points of the children based on deep learning is not available at present, in the planning process of the radiotherapy of the children, important bone growth centers in the radiotherapy process can only be sketched manually, the number of the bone growth centers of the children is large, the size is small and the bone growth centers are distributed on the whole body, so that the manual sketching of the bone growth centers of the children in the radiotherapy plan is very time-consuming (the total sketching takes about 40 minutes), meanwhile, because the bone growth points are complex in structure and difficult to identify on CT images, anatomical information of upper and lower horizontal layers of the CT images needs to be consulted when the bone growth points are sketched, and the sketching difference among doctors can be caused. Because the growth and development of the bone of the children are a continuous and complex stage, the bone of the growth and development is influenced by factors such as age, nutrition condition, puberty condition, existence of basic diseases and the like, and bone growth points of children patients at different or same age stage can show different shapes and volumes on CT images, and the difficulty is brought to the accuracy of automatic segmentation of the bone growth points of the children.
Aiming at the problem that the accuracy of automatic sketching of bone growth points is lower because of complex structure of the bone growth points and difficult identification on CT images in the related art, no effective solution is proposed at present.
Disclosure of Invention
The application mainly aims to provide a method and a device for outlining a bone growth point, a storage medium and a processor, which are used for solving the problem that the accuracy of automatic outlining of the bone growth point is lower because the bone growth point is complex in structure and difficult to identify on a CT image in the related technology.
In order to achieve the above object, according to one aspect of the present application, there is provided a method of delineating a bone growth point. The method comprises the following steps: obtaining a target image, wherein the target image comprises: a target layer CT image to be sketched, wherein a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image; and inputting the CT images of the target layer to be sketched, namely the CT image of the upper layer of the CT images of the target layer and the CT image of the lower layer of the CT images of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of skeleton growth points in the CT images of the target layer.
Further, inputting the CT images of the target layer to be sketched, which are the CT images of the upper layer and the CT images of the lower layer of the CT images of the target layer, into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of a bone growth point in the CT images of the target layer includes: processing the CT image of the upper layer through the first bypass branch network to obtain first target characteristic information of a bone growth point in the CT image of the upper layer; processing the next layer CT image through the second bypass branch network to obtain second target characteristic information of bone growth points in the next layer CT image; encoding the target layer CT image through a main encoder in the main branch network to obtain third target characteristic information of a bone growth point in the target layer CT image; combining the first target feature information, the second target feature information and the third target feature information to obtain first total target feature information; and decoding the first total target characteristic information through a main decoder in the main branch network to obtain a sketching result of skeleton growth points in the target layer CT image.
Further, processing the previous layer CT image through the first bypass branch network, and obtaining first target feature information in the previous layer CT image includes: encoding the upper layer CT image through an encoder in the first bypass branch network to obtain first initial characteristic information of a bone growth point in the upper layer CT image; and carrying out semantic enhancement on the initial characteristic information through an extrusion and excitation module in the first bypass branch network to obtain the first target characteristic information.
Further, before inputting the CT image of the target layer to be sketched and the CT image of the upper layer of the CT image of the target layer and the CT image of the lower layer of the CT image of the target layer into the main branch network, the first bypass branch network and the second bypass branch network in the target network model for processing, to obtain the sketching result of the bone growth point in the CT image of the target layer, the method further includes: acquiring a plurality of training sample sets, wherein the plurality of training sample sets at least comprise a plurality of CT image sets and standard sketching results of bone growth points in the plurality of CT image sets, one CT image set comprises an upper CT image, a middle CT image and a lower CT image, the upper CT image is a CT image of the upper layer of the middle CT image, and the lower CT image is a CT image of the lower layer of the middle CT image; inputting the plurality of training sample groups into an initial network model to obtain a target prediction sketch result of the training samples; calculating the target prediction sketching result and a loss function of the target network model according to the standard sketching result to obtain a target loss value; and optimizing parameters of the initial network model according to the target loss value to obtain the target network model.
Further, inputting the plurality of training samples into an initial network model to obtain a target prediction sketch result of the training samples, including: processing the upper CT image through an initial first bypass branch network in the initial network model to obtain a first prediction sketch result; processing the middle-layer CT image through an initial main branch network in the initial network model to obtain a second prediction sketch result; processing the lower CT image through an initial second bypass branch network in the initial network model to obtain a third prediction sketch result; and taking the first prediction sketch result, the second prediction sketch result and the third prediction sketch result as the target prediction sketch result.
Further, according to the standard sketching result, the target prediction sketching result and the loss function of the target network model are calculated, and obtaining a target loss value includes: calculating according to the first prediction sketching result, the standard sketching result of the bone growth points in the upper CT image and the loss function to obtain a first loss value; calculating according to the second prediction sketch result, the standard sketch result of the bone growth point in the middle-layer CT image and the loss function to obtain a second loss value; calculating according to the third prediction sketch result, the standard sketch result of the bone growth point in the lower CT image and the loss function to obtain a third loss value; and taking the first loss value, the second loss value and the third loss value as the target loss values.
Further, processing the upper layer CT image through an initial first bypass branch network in the initial network model, and obtaining a first prediction sketch result includes: encoding the upper CT image through an encoder in the initial first bypass branch network to obtain second initial characteristic information of bone growth points in the upper CT image; and decoding the second initial characteristic information through a decoder in the initial first bypass branch network to obtain the first prediction sketch result.
Further, processing the middle layer CT image through the initial main branch network to obtain a second prediction sketch result includes: semantic enhancement is carried out on the second initial characteristic information through a first extrusion and excitation module in the initial first bypass branch network, so that fourth target characteristic information is obtained; acquiring fifth target characteristic information of bone growth points in the lower CT image; encoding the middle-layer CT image through a main encoder of the initial main branch network to obtain sixth target characteristic information of bone growth points in the middle-layer CT image; combining the fourth target characteristic information, the fifth target characteristic information and the sixth target characteristic information to obtain second total characteristic information; and decoding the second total characteristic information through a main decoder of the initial main branch network to obtain the second prediction sketch result.
Further, obtaining a plurality of training sample sets includes: creating an initial sketching rule of bone growth points according to the human body growth rule; optimizing the initial sketching rule according to the characteristics of the bone growth points to obtain a target sketching rule; acquiring the plurality of CT image groups, and adopting the target sketching rule to sketch bone growth points of the plurality of CT image groups to obtain the standard sketching result; and constructing the CT image group and the standard sketching result into a plurality of training sample groups.
In order to achieve the above object, according to another aspect of the present application, there is provided a delineating device of a bone growth point. The device comprises: the first acquisition unit is used for acquiring a target image, wherein the target image comprises: a target layer CT image to be sketched, wherein a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image; the first processing unit is used for inputting the CT images of the target layer to be sketched, wherein the CT images of the upper layer of the CT images of the target layer and the CT images of the lower layer of the CT images of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of skeleton growing points in the CT images of the target layer.
Further, the first processing unit includes: the first processing subunit is used for processing the CT image of the upper layer through the first bypass branch network to obtain first target characteristic information of a bone growth point in the CT image of the upper layer; the second processing subunit is used for processing the next layer of CT image through the second bypass branch network to obtain second target characteristic information of bone growth points in the next layer of CT image; the encoding subunit is used for encoding the target layer CT image through a main encoder in the main branch network to obtain third target characteristic information of a bone growth point in the target layer CT image; the merging subunit is used for merging the first target feature information, the second target feature information and the third target feature information to obtain first total target feature information; and the decoding subunit is used for decoding the first total target characteristic information through a main decoder in the main branch network to obtain a sketching result of skeleton growth points in the target layer CT image.
Further, the first processing subunit includes: the first coding module is used for coding the CT image of the upper layer through an encoder in the first bypass branch network to obtain first initial characteristic information of a bone growth point in the CT image of the upper layer; the first processing module is used for carrying out semantic enhancement on the initial characteristic information through the extrusion and excitation module in the first bypass branch network to obtain the first target characteristic information.
Further, the apparatus further comprises: the second obtaining unit is configured to obtain a plurality of training sample sets before inputting a CT image of a target layer to be sketched and a CT image of an upper layer of the CT image of the target layer and a CT image of a lower layer of the CT image of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model to process, so as to obtain a sketching result of a bone growth point in the CT image of the target layer, where the plurality of training sample sets at least include a plurality of CT image sets and a standard sketching result of the bone growth point in the plurality of CT image sets, and one CT image set is composed of an upper layer CT image, a middle layer CT image and a lower layer CT image, and the upper layer CT image is a CT image of the upper layer of the middle layer CT image and the lower layer CT image is a CT image of the lower layer of the middle layer CT image; the second processing unit is used for inputting the training sample groups into an initial network model to obtain a target prediction sketch result of the training samples; the calculation unit is used for calculating the target prediction sketching result and the loss function of the target network model according to the standard sketching result to obtain a target loss value; and the optimizing unit is used for optimizing the parameters of the initial network model according to the target loss value to obtain the target network model.
Further, the second processing unit includes: the third processing subunit is used for processing the upper CT image through an initial first bypass branch network in the initial network model to obtain a first prediction sketch result; a fourth processing subunit, configured to process the middle layer CT image through an initial main branch network in the initial network model, to obtain a second prediction sketch result; a fifth processing subunit, configured to process the lower layer CT image through an initial second bypass branch network in the initial network model, to obtain a third prediction sketch result; and the first determination subunit is used for taking the first prediction sketching result, the second prediction sketching result and the third prediction sketching result as the target prediction sketching result.
Further, the computing unit includes: the first calculating subunit is used for calculating according to the first prediction sketching result, the standard sketching result of the bone growth point in the upper CT image and the loss function to obtain a first loss value; the second calculating subunit is used for calculating according to the second prediction sketching result, the standard sketching result of the bone growth points in the middle-layer CT image and the loss function to obtain a second loss value; the third calculation subunit is used for calculating according to the third prediction sketching result, the standard sketching result of the bone growth points in the lower CT image and the loss function to obtain a third loss value; and the second determining subunit is used for taking the first loss value, the second loss value and the third loss value as the target loss value.
Further, the third processing subunit includes: the second coding module is used for coding the upper CT image through an encoder in the initial first bypass branch network to obtain second initial characteristic information of bone growth points in the upper CT image; and the decoding module is used for decoding the second initial characteristic information through a decoder in the initial first bypass branch network to obtain the first prediction sketch result.
Further, the fourth processing subunit includes: the second processing module is used for carrying out semantic enhancement on the second initial characteristic information through a first extrusion and excitation module in the initial first bypass branch network to obtain fourth target characteristic information; the acquisition module is used for acquiring fifth target characteristic information of bone growth points in the lower CT image; the third coding module is used for coding the middle-layer CT image through a main encoder of the initial main branch network to obtain sixth target characteristic information of bone growth points in the middle-layer CT image; the merging module is used for merging the fourth target characteristic information, the fifth target characteristic information and the sixth target characteristic information to obtain second total characteristic information; and the determining module is used for decoding the second total characteristic information through a main decoder of the initial main branch network to obtain the second prediction sketch result.
Further, the second acquisition unit includes: the creation subunit is used for creating an initial sketching rule of skeleton growing points according to the human body growing rule; the optimizing subunit is used for optimizing the initial sketching rule according to the characteristics of the bone growth points to obtain a target sketching rule; the acquisition subunit is used for acquiring the plurality of CT image groups, and carrying out bone growth point sketching on the plurality of CT image groups by adopting the target sketching rule to obtain the standard sketching result; and the constructing subunit is used for constructing the CT image group and the standard sketching result into the training sample groups.
In order to achieve the above object, according to another aspect of the present application, there is also provided a computer-readable storage medium storing a program, wherein the program, when run, controls a device in which the storage medium is located to perform the method of delineating bone growth points according to any one of the above.
To achieve the above object, according to another aspect of the present application, there is also provided a processor, wherein the processor is configured to execute a program, wherein the program executes the method for delineating bone growth points according to any one of the above.
According to the application, the following steps are adopted: obtaining a target image, wherein the target image comprises: a target layer CT image to be sketched, a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image; and inputting a CT image of a target layer to be sketched and a CT image of an upper layer of the CT image of the target layer and a CT image of a lower layer of the CT image of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing to obtain a sketching result of skeleton growing points in the CT image of the target layer, thereby solving the problem that the accuracy of automatic sketching of the skeleton growing points is lower due to the complex structure of the skeleton growing points and difficult identification on the CT image in the related technology. Through inputting three-layer adjacent CT images into the target network model, the sketching of the bone growth points in the middle CT images is realized through integrating the characteristic information of the adjacent CT images, so that the target network model can more accurately and automatically divide important bone growth centers of children, and the effect of improving the accuracy of the automatic sketching of the bone growth points is further achieved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:
FIG. 1 is a flow chart of a method of delineating bone growth points provided in accordance with an embodiment of the present application;
FIG. 2 is a schematic diagram of a target network model provided in accordance with an embodiment of the present application;
fig. 3 is a schematic view of a bone growth point delineating device according to an embodiment of the present application.
Detailed Description
It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. The application will be described in detail below with reference to the drawings in connection with embodiments.
In order that those skilled in the art will better understand the present application, a technical solution in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, shall fall within the scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate in order to describe the embodiments of the application herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The application will be described with reference to preferred embodiments, and FIG. 1 is a flowchart of a method for delineating bone growth points according to an embodiment of the application, as shown in FIG. 1, comprising the steps of:
Step S101, obtaining a target image, where the target image includes: the target layer CT image to be sketched, a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image.
Specifically, CT images of adjacent three layers are obtained from image data of a patient, and delineating bone growth points in the intermediate CT image (i.e., the target layer CT image described above) is achieved by the three layers of CT images.
Step S102, inputting the CT images of the target layer to be sketched, namely the CT image of the upper layer of the CT images of the target layer and the CT image of the lower layer of the CT images of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of bone growth points in the CT images of the target layer.
Specifically, the target network model is composed of a main branch network, a first bypass branch network and a second bypass branch network, and spatial information of adjacent CT images is integrated through the target network model, so that accurate sketching of bone growth points in the CT images of the target layer is realized.
In summary, three layers of adjacent CT images are input into the target network model, and the sketching of the bone growth points in the middle CT image is realized by integrating the characteristic information of the adjacent CT images, so that the target network model can more accurately and automatically sketch important bone growth centers of children, and the accuracy of the automatic sketching of the bone growth points is improved.
How to accurately sketch the CT image through the target network model is important, and in the sketching method of the bone growth points provided by the embodiment of the application, the sketching of the bone growth points of the CT image through the target network model is realized in the following way: processing the CT image of the upper layer through a first bypass branch network to obtain first target characteristic information of bone growth points in the CT image of the upper layer; processing the next layer of CT image through a second bypass branch network to obtain second target characteristic information of bone growth points in the next layer of CT image; encoding the target layer CT image through a main encoder in a main branch network to obtain third target characteristic information of a bone growth point in the target layer CT image; combining the first target feature information, the second target feature information and the third target feature information to obtain first total target feature information; and decoding the first total target characteristic information through a main decoder in the main branch network to obtain a sketching result of skeleton growth points in the CT image of the target layer.
Specifically, as shown in fig. 2, the target network model is composed of a main branch network, a first bypass branch network, and a second bypass branch network, each having an encoder and a decoder. Both the encoder and decoder are stacked by four residual blocks. The encoder of the bypass branch network is connected with a squeezing and exciting Module (SE Module) to realize that the characteristic information of the adjacent CT images is transmitted into the main branch network. During the process of delineating the target image, only two bypass-branched encoders and main-branched networks are maintained, as shown by the solid line in fig. 2.
The method comprises the steps of firstly, processing a CT image of an upper layer through a first bypass branch network to obtain first target characteristic information of a bone growth point of the CT image of the upper layer, processing a CT image of a lower layer through a second bypass branch network to obtain second target characteristic information of the bone growth point of the CT image of the lower layer, and encoding the CT image of the target layer through a main encoder of a main branch network to obtain third target characteristic information of the bone growth point in the CT image of the target layer. And combining the first target characteristic information, the second target characteristic information and the third target characteristic information to obtain total characteristic information (namely the first total characteristic information). And finally, decoding the first total characteristic information through a main decoder of the main branch network, so as to obtain a sketching result of skeleton growth points in the CT image of the target layer.
Through the steps, when the CT image of the target layer is sketched, the characteristic information of the CT images of the adjacent layers is fully combined, so that the sketching of the bone growth points in the CT image of the target layer can be accurately realized.
The target characteristic information of the adjacent CT images is accurately obtained through the following steps: encoding the CT image of the upper layer by an encoder in the first bypass branch network to obtain first initial characteristic information of a bone growth point in the CT image of the upper layer; and carrying out semantic enhancement on the initial characteristic information through an extrusion and excitation module in the first bypass branch network to obtain first target characteristic information.
Specifically, an encoder of a first bypass branch network is utilized to encode a CT image of a previous layer to obtain initial characteristic information of bone growth points in the CT image of the previous layer, and then the initial characteristic information is subjected to semantic enhancement through an extrusion and excitation module to obtain first target characteristic information. The same procedure is performed on the next layer CT image through the second bypass branch network. And encoding the CT image of the next layer by using an encoder of the second bypass branch network to obtain initial characteristic information of skeleton growth points in the CT image of the next layer, and then carrying out semantic enhancement on the initial characteristic information by using the extrusion and excitation module to obtain second target characteristic information.
By carrying out semantic enhancement on the initial characteristic information, the characteristic information of the CT images of the adjacent layers can be captured more accurately, and further, the accuracy of the bone growth point sketching of the CT images of the target layers is improved.
It is also important how to train to obtain a target network model, and in the method for delineating bone growth points provided by the embodiment of the application, the following method is adopted to obtain the target network model: acquiring a plurality of training sample sets, wherein the plurality of training sample sets at least comprise a plurality of CT image sets and standard sketching results of bone growth points in the CT image sets, one CT image set comprises an upper CT image, a middle CT image and a lower CT image, the upper CT image is a CT image of the upper layer of the middle CT image, and the lower CT image is a CT image of the lower layer of the middle CT image; inputting a plurality of training sample groups into an initial network model to obtain a target prediction sketch result of the training samples; calculating a target prediction sketch result and a loss function of a target network model according to the standard sketch result to obtain a target loss value; and optimizing parameters of the initial network model according to the target loss value to obtain a target network model.
Specifically, during training, the bypass branch network outputs a sketching result of the CT image of the adjacent layer, and the corresponding loss value (i.e., the target loss value) is obtained through the sketching result output by the three layers of the two bypass branch networks and the main branch network, and the initial network model is trained by using the loss value, so as to finally obtain the target network model.
A plurality of training sample sets are obtained through the CT image of a patient, for example, one patient shoots CT, 80 layers of CT images can be obtained, and then the adjacent three layers of CT images can form one training sample set. And meanwhile, the training sample set comprises standard sketching results of each CT image. And calculating to obtain the target loss value by using the standard sketching result and the sketching result (namely the target prediction sketching result) output by the initial network model.
In an alternative embodiment, the dice losses and the profile constraint losses are combined to obtain a loss function, the overall loss function beingWherein p is a sketching result obtained through an initial network model,For a binarized training sample, alpha is a weight factor, default is set to 1.0,/>For Dice Loss, CL (p) is the contour constraint Loss.
Through the steps, convergence stability, accuracy and smoothness of the target network model are guaranteed.
In an alternative embodiment, the target prediction delineation result and the target loss value of the training sample are obtained by the following method: processing the upper CT image through an initial first bypass branch network in an initial network model to obtain a first prediction sketch result; processing the middle-layer CT image through an initial main branch network in an initial network model to obtain a second prediction sketch result; processing the lower CT image through an initial second bypass branch network in the initial network model to obtain a third prediction sketch result; and taking the first prediction sketching result, the second prediction sketching result and the third prediction sketching result as target prediction sketching results. Calculating according to the first prediction sketching result, the standard sketching result of the bone growth points in the upper CT image and the loss function to obtain a first loss value; calculating according to the second prediction sketching result, the standard sketching result of the bone growth points in the middle-layer CT image and the loss function to obtain a second loss value; calculating according to the third prediction sketching result, the standard sketching result of the bone growth points in the lower CT image and the loss function to obtain a third loss value; the first loss value, the second loss value and the third loss value are set as target loss values.
Specifically, an encoder in an initial first bypass branch network is used for encoding the upper CT image to obtain second initial characteristic information of a bone growth point in the upper CT image; and decoding the second initial characteristic information through a decoder in the initial first bypass branch network to obtain a first prediction sketch result of the upper CT image.
Processing the middle-layer CT image through an initial main branch network to obtain a second prediction sketch result of the middle-layer CT image; and processing the lower CT image through the initial second bypass branch network to obtain a third prediction sketch result. Calculating to obtain a first loss value through a first prediction sketch result and a standard sketch result of an upper CT image; and calculating a second loss value through a second prediction sketch result and a standard sketch result of the middle-layer CT image, calculating a third loss value through a third prediction sketch result and a standard sketch result of the lower-layer CT image, and taking the first loss value, the second loss value and the third loss value as target loss values.
It should be noted that, when the middle layer CT image is sketched, the feature information of the upper layer CT image and the lower layer CT image are combined, and the specific steps are as follows: semantic enhancement is carried out on the second initial characteristic information through a first extrusion and excitation module in the initial first bypass branch network, so that fourth target characteristic information is obtained; acquiring fifth target characteristic information of a bone growth point in the lower CT image; encoding the middle layer CT image through a main encoder of an initial main branch network to obtain sixth target characteristic information of bone growth points in the middle layer CT image; combining the fourth target characteristic information, the fifth target characteristic information and the sixth target characteristic information to obtain second total characteristic information; and decoding the second total characteristic information through a main decoder of the initial main branch network to obtain a second prediction sketch result.
Specifically, after the initial characteristic information of the upper layer CT image is obtained through the encoder of the initial first bypass branch network, the first extrusion and excitation module of the initial first bypass branch network is utilized to carry out semantic enhancement on the initial characteristic information, so as to obtain fourth target characteristic information. After initial characteristic information of a lower layer CT image is obtained by using an encoder of an initial second bypass branch network, semantic enhancement is carried out on the initial characteristic information by using a second extrusion and excitation module of the initial second bypass branch network to obtain fifth target characteristic information, the middle layer CT image is encoded by using a main encoder of an initial main branch network to obtain sixth target characteristic information of a bone growth point in the middle layer CT image, and then fourth target characteristic information, the fifth target characteristic information and the sixth target characteristic information are combined to obtain second total characteristic information, and a main decoder of the initial main branch network is used for decoding the second total characteristic information to finally obtain a second prediction sketch result.
Through the training process, the accuracy of the target network model on the sketching of the bone growth points is further improved.
In the prior art, a set of strict and standard sketching rules for bone growth points does not exist, so in order to ensure that standard sketching results of a training sample set are accurately obtained, in the sketching method for bone growth points provided by the embodiment of the application, the following steps are adopted to obtain the standard sketching results of the training sample set: creating an initial sketching rule of bone growth points according to the human body growth rule; optimizing the initial sketching rule according to the characteristics of the bone growth points to obtain a target sketching rule; acquiring a plurality of CT image groups, and adopting a target sketching rule to sketch bone growth points of the CT image groups to obtain a standard sketching result; the CT image set and the standard sketching result are constructed into a plurality of training sample sets.
In an alternative embodiment, an initial sketching of skeletal growth points is created based on human growth rules, i.e., the important child skeletal growth centers are divided into three groups: the craniofacial, shoulder and pelvic bone growth centers total 22 structures.
(1) The craniofacial bone growth center consists of frontal suture (FRS), frontal Zygomatic Suture (FZS), zygomatic suture (ZMS), frontal jaw suture (FMS), nasal septum (NAS), condyle (COP), and coronary process (CRP).
(2) Shoulder bone growth centers include Scapula (SCA), acromion (APS), coracoid Process (CPS), collarbone (CLA), humeral head, humeral tuberosity, and greater humerus tuberosity.
(3) The pelvic bone growth center includes sacroiliac coccyx region (SIR), iliac crest (ILC), pubic symphysis (PUS), ischial tuberosity (IST), greater trochanter (GRT), lesser trochanter (LRT), epiphysis of femur, femoral head.
In order to achieve a more accurate segmentation effect, anatomically adjacent growth centers (i.e., the above-described features according to bone growth points, the initial delineation rules are optimized to obtain the target delineation rules) are merged. For example, when sketched, the merger humerus head, the lesser humeral tuberosity and the greater humerus tuberosity are Humeral Head Tuberosities (HHT), and the merger femoral head and femoral epiphysis are named Femoral Head Epiphysis (FHE). 19 structures of each patient were manually depicted on the horizontal, coronal and sagittal CT images of the bone window by the above-described target delineation rules and examined by decreakers and radiologists to obtain multiple training sample sets.
In an alternative embodiment, DICE SIMILARITY Coefficient (DSC) and Hausdorff Distance th-percentile (HD 95) may be used to evaluate the target network model in order to quantitatively evaluate the performance of the target network model. Through experimental evaluation, compared with U-Net and V-Net network models in the prior art, the target network model can more accurately and automatically sketch skeleton growth points in the CT influence diagram.
The method for outlining the bone growth points provided by the embodiment of the application comprises the steps of obtaining a target image, wherein the target image comprises the following steps: a target layer CT image to be sketched, a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image; and inputting a CT image of a target layer to be sketched and a CT image of an upper layer of the CT image of the target layer and a CT image of a lower layer of the CT image of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing to obtain a sketching result of skeleton growing points in the CT image of the target layer, thereby solving the problem that the accuracy of automatic sketching of the skeleton growing points is lower due to the complex structure of the skeleton growing points and difficult identification on the CT image in the related technology. Through inputting three-layer adjacent CT images into the target network model, the sketching of the bone growth points in the middle CT images is realized through integrating the characteristic information of the adjacent CT images, so that the target network model can more accurately and automatically divide important bone growth centers of children, and the effect of improving the accuracy of the automatic sketching of the bone growth points is further achieved.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.
The embodiment of the application also provides a device for outlining the bone growth points, and the device for outlining the bone growth points can be used for executing the method for outlining the bone growth points. The following describes a bone growth point delineating device provided by the embodiment of the application.
Fig. 3 is a schematic view of a bone growth point delineating device according to an embodiment of the present application. As shown in fig. 3, the apparatus includes: a first acquisition unit 301 and a first processing unit 302.
The first obtaining unit 301 is configured to obtain a target image, where the target image includes: a target layer CT image to be sketched, a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image;
The first processing unit 302 is configured to input a target layer CT image to be sketched, a CT image of a previous layer of the target layer CT image, and a CT image of a next layer of the target layer CT image into a main branch network, a first bypass branch network, and a second bypass branch network in the target network model for processing, so as to obtain a sketching result of a bone growth point in the target layer CT image.
According to the bone growth point delineation device provided by the embodiment of the application, the first acquisition unit 501 acquires the target image, wherein the target image comprises: a target layer CT image to be sketched, a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image; the first processing unit 502 inputs the CT images of the target layer to be sketched, the CT image of the upper layer of the CT images of the target layer and the CT image of the lower layer of the CT images of the target layer into the main branch network, the first bypass branch network and the second bypass branch network in the target network model for processing, so as to obtain the sketching result of the bone growth points in the CT images of the target layer, and solve the problem that the accuracy of automatic sketching of the bone growth points is lower due to the complex structure of the bone growth points and the difficult recognition on the CT images in the related art. Through inputting three-layer adjacent CT images into the target network model, the sketching of the bone growth points in the middle CT images is realized through integrating the characteristic information of the adjacent CT images, so that the target network model can more accurately and automatically divide important bone growth centers of children, and the effect of improving the accuracy of the automatic sketching of the bone growth points is further achieved.
Optionally, in the bone growth point delineation device provided in the embodiment of the present application, the first processing unit 302 includes: the first processing subunit is used for processing the CT image of the upper layer through a first bypass branch network to obtain first target characteristic information of a bone growth point in the CT image of the upper layer; the second processing subunit is used for processing the next layer of CT image through a second bypass branch network to obtain second target characteristic information of a bone growth point in the next layer of CT image; the encoding subunit is used for encoding the target layer CT image through a main encoder in the main branch network to obtain third target characteristic information of a bone growth point in the target layer CT image; the merging subunit is used for merging the first target feature information, the second target feature information and the third target feature information to obtain first total target feature information; and the decoding subunit is used for decoding the first total target characteristic information through a main decoder in the main branch network to obtain a sketching result of skeleton growth points in the target layer CT image.
Optionally, in the bone growth point delineation device provided by the embodiment of the present application, the first processing subunit includes: the first coding module is used for coding the CT image of the upper layer through an encoder in the first bypass branch network to obtain first initial characteristic information of a bone growth point in the CT image of the upper layer; the first processing module is used for carrying out semantic enhancement on the initial characteristic information through the extrusion and excitation module in the first bypass branch network to obtain first target characteristic information.
Optionally, in the device for delineating a bone growth point provided by the embodiment of the present application, the device further includes: the second acquisition unit is used for acquiring a plurality of training sample sets before inputting a target layer CT image to be sketched, a CT image of an upper layer of the target layer CT image and a CT image of a lower layer of the target layer CT image into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing to obtain a sketching result of a bone growth point in the target layer CT image, wherein the plurality of training sample sets at least comprise a plurality of CT image sets and standard sketching results of the bone growth point in the plurality of CT image sets, one CT image set comprises an upper layer CT image, a middle layer CT image and a lower layer CT image, the upper layer CT image is a CT image of the upper layer of the middle layer CT image, and the lower layer CT image is a CT image of the lower layer of the middle layer CT image; the second processing unit is used for inputting a plurality of training sample groups into the initial network model to obtain a target prediction sketch result of the training samples; the calculation unit is used for calculating the target prediction sketching result and the loss function of the target network model according to the standard sketching result to obtain a target loss value; and the optimizing unit is used for optimizing the parameters of the initial network model according to the target loss value to obtain the target network model.
Optionally, in the bone growth point delineation device provided by the embodiment of the present application, the second processing unit includes: the third processing subunit is used for processing the upper CT image through an initial first bypass branch network in the initial network model to obtain a first prediction sketch result; a fourth processing subunit, configured to process the middle layer CT image through an initial main branch network in the initial network model, to obtain a second prediction sketch result; the fifth processing subunit is used for processing the lower layer CT image through an initial second bypass branch network in the initial network model to obtain a third prediction sketch result; and the first determination subunit is used for taking the first prediction sketching result, the second prediction sketching result and the third prediction sketching result as target prediction sketching results.
Optionally, in the bone growth point delineating device provided by the embodiment of the present application, the computing unit includes: the first calculating subunit is used for calculating according to the first prediction sketching result, the standard sketching result of the bone growth point in the upper layer CT image and the loss function to obtain a first loss value; the second calculating subunit is used for calculating according to the second prediction sketching result, the standard sketching result of the bone growth point in the middle-layer CT image and the loss function to obtain a second loss value; the third calculation subunit is used for calculating according to a third prediction sketch result, a standard sketch result of a bone growth point in the lower-layer CT image and a loss function to obtain a third loss value; and the second determining subunit is used for taking the first loss value, the second loss value and the third loss value as target loss values.
Optionally, in the bone growth point delineation device provided in the embodiment of the present application, the third processing subunit includes: the second coding module is used for coding the upper CT image through an encoder in the initial first bypass branch network to obtain second initial characteristic information of bone growth points in the upper CT image; and the decoding module is used for decoding the second initial characteristic information through a decoder in the initial first bypass branch network to obtain a first prediction sketch result.
Optionally, in the device for delineating a bone growth point provided in the embodiment of the present application, the fourth processing subunit includes: the second processing module is used for carrying out semantic enhancement on the second initial characteristic information through a first extrusion and excitation module in the initial first bypass branch network to obtain fourth target characteristic information; the acquisition module is used for acquiring fifth target characteristic information of bone growth points in the lower CT image; the third coding module is used for coding the middle layer CT image through a main coder of the initial main branch network to obtain sixth target characteristic information of a bone growth point in the middle layer CT image; the merging module is used for merging the fourth target characteristic information, the fifth target characteristic information and the sixth target characteristic information to obtain second total characteristic information; and the determining module is used for decoding the second total characteristic information through a main decoder of the initial main branch network to obtain a second prediction sketch result.
Optionally, in the bone growth point delineating device provided in the embodiment of the present application, the second obtaining unit includes: the creation subunit is used for creating an initial sketching rule of skeleton growing points according to the human body growing rule; the optimizing subunit is used for optimizing the initial sketching rule according to the characteristics of the bone growth points to obtain a target sketching rule; the acquisition subunit is used for acquiring a plurality of CT image groups, and carrying out bone growth point sketching on the CT image groups by adopting a target sketching rule to obtain a standard sketching result; and the constructing subunit is used for constructing the CT image group and the standard sketching result into a plurality of training sample groups.
The bone growth point delineation device comprises a processor and a memory, wherein the first acquisition unit 301, the first processing unit 302 and the like are stored in the memory as program units, and the processor executes the program units stored in the memory to realize corresponding functions.
The processor includes a kernel, and the kernel fetches the corresponding program unit from the memory. The kernel can be provided with one or more than one kernel, and the delineation of the bone growth points is realized by adjusting kernel parameters.
The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), among other forms in computer readable media, the memory including at least one memory chip.
Embodiments of the present invention provide a computer-readable storage medium having stored thereon a program that, when executed by a processor, implements a method of delineating bone growth points.
The embodiment of the invention provides a processor, which is used for running a program, wherein the program runs to execute a sketching method of a bone growth point. The processor when executing the program implements the following steps: obtaining a target image, wherein the target image comprises: a target layer CT image to be sketched, a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image; and inputting the CT images of the target layer to be sketched, namely the CT image of the upper layer of the CT images of the target layer and the CT image of the lower layer of the CT images of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of skeleton growth points in the CT images of the target layer.
Optionally, inputting the target layer CT image to be sketched, a CT image of a previous layer of the target layer CT image and a CT image of a next layer of the target layer CT image into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of a bone growth point in the target layer CT image includes: processing the CT image of the upper layer through a first bypass branch network to obtain first target characteristic information of bone growth points in the CT image of the upper layer; processing the next layer of CT image through a second bypass branch network to obtain second target characteristic information of bone growth points in the next layer of CT image; encoding the target layer CT image through a main encoder in a main branch network to obtain third target characteristic information of a bone growth point in the target layer CT image; combining the first target feature information, the second target feature information and the third target feature information to obtain first total target feature information; and decoding the first total target characteristic information through a main decoder in the main branch network to obtain a sketching result of skeleton growth points in the CT image of the target layer.
Optionally, processing the CT image of the previous layer through the first bypass branch network, and obtaining the first target feature information in the CT image of the previous layer includes: encoding the CT image of the upper layer by an encoder in the first bypass branch network to obtain first initial characteristic information of a bone growth point in the CT image of the upper layer; and carrying out semantic enhancement on the initial characteristic information through an extrusion and excitation module in the first bypass branch network to obtain first target characteristic information.
Optionally, before inputting the target layer CT image to be sketched, the CT image of the upper layer of the target layer CT image, and the CT image of the lower layer of the target layer CT image into the main branch network, the first bypass branch network, and the second bypass branch network in the target network model for processing, to obtain the sketching result of the bone growth point in the target layer CT image, the method further includes: acquiring a plurality of training sample sets, wherein the plurality of training sample sets at least comprise a plurality of CT image sets and standard sketching results of bone growth points in the CT image sets, one CT image set comprises an upper CT image, a middle CT image and a lower CT image, the upper CT image is a CT image of the upper layer of the middle CT image, and the lower CT image is a CT image of the lower layer of the middle CT image; inputting a plurality of training sample groups into an initial network model to obtain a target prediction sketch result of the training samples; calculating a target prediction sketch result and a loss function of a target network model according to the standard sketch result to obtain a target loss value; and optimizing parameters of the initial network model according to the target loss value to obtain a target network model.
Optionally, inputting a plurality of training samples into the initial network model to obtain a target prediction sketch result of the training samples, including: processing the upper CT image through an initial first bypass branch network in an initial network model to obtain a first prediction sketch result; processing the middle-layer CT image through an initial main branch network in an initial network model to obtain a second prediction sketch result; processing the lower CT image through an initial second bypass branch network in the initial network model to obtain a third prediction sketch result; and taking the first prediction sketching result, the second prediction sketching result and the third prediction sketching result as target prediction sketching results.
Optionally, calculating the target prediction sketching result and the loss function of the target network model according to the standard sketching result, and obtaining the target loss value includes: calculating according to the first prediction sketching result, the standard sketching result of the bone growth points in the upper CT image and the loss function to obtain a first loss value; calculating according to the second prediction sketching result, the standard sketching result of the bone growth points in the middle-layer CT image and the loss function to obtain a second loss value; calculating according to the third prediction sketching result, the standard sketching result of the bone growth points in the lower CT image and the loss function to obtain a third loss value; the first loss value, the second loss value and the third loss value are set as target loss values.
Optionally, processing the upper layer CT image through an initial first bypass branch network in the initial network model, and obtaining a first prediction sketch result includes: encoding the upper CT image through an encoder in the initial first bypass branch network to obtain second initial characteristic information of bone growth points in the upper CT image; and decoding the second initial characteristic information through a decoder in the initial first bypass branch network to obtain a first prediction sketch result.
Optionally, processing the middle layer CT image through the initial main branch network to obtain a second prediction sketch result includes: semantic enhancement is carried out on the second initial characteristic information through a first extrusion and excitation module in the initial first bypass branch network, so that fourth target characteristic information is obtained; acquiring fifth target characteristic information of a bone growth point in the lower CT image; encoding the middle layer CT image through a main encoder of an initial main branch network to obtain sixth target characteristic information of bone growth points in the middle layer CT image; combining the fourth target characteristic information, the fifth target characteristic information and the sixth target characteristic information to obtain second total characteristic information; and decoding the second total characteristic information through a main decoder of the initial main branch network to obtain a second prediction sketch result.
Optionally, acquiring the plurality of training sample sets includes: creating an initial sketching rule of bone growth points according to the human body growth rule; optimizing the initial sketching rule according to the characteristics of the bone growth points to obtain a target sketching rule; acquiring a plurality of CT image groups, and adopting a target sketching rule to sketch bone growth points of the CT image groups to obtain a standard sketching result; the CT image set and the standard sketching result are constructed into a plurality of training sample sets. The device herein may be a server, PC, PAD, cell phone, etc.
The application also provides a computer program product adapted to perform, when executed on a data processing device, a program initialized with the method steps of: obtaining a target image, wherein the target image comprises: a target layer CT image to be sketched, a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image; and inputting the CT images of the target layer to be sketched, namely the CT image of the upper layer of the CT images of the target layer and the CT image of the lower layer of the CT images of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of skeleton growth points in the CT images of the target layer.
Optionally, inputting the target layer CT image to be sketched, a CT image of a previous layer of the target layer CT image and a CT image of a next layer of the target layer CT image into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of a bone growth point in the target layer CT image includes: processing the CT image of the upper layer through a first bypass branch network to obtain first target characteristic information of bone growth points in the CT image of the upper layer; processing the next layer of CT image through a second bypass branch network to obtain second target characteristic information of bone growth points in the next layer of CT image; encoding the target layer CT image through a main encoder in a main branch network to obtain third target characteristic information of a bone growth point in the target layer CT image; combining the first target feature information, the second target feature information and the third target feature information to obtain first total target feature information; and decoding the first total target characteristic information through a main decoder in the main branch network to obtain a sketching result of skeleton growth points in the CT image of the target layer.
Optionally, processing the CT image of the previous layer through the first bypass branch network, and obtaining the first target feature information in the CT image of the previous layer includes: encoding the CT image of the upper layer by an encoder in the first bypass branch network to obtain first initial characteristic information of a bone growth point in the CT image of the upper layer; and carrying out semantic enhancement on the initial characteristic information through an extrusion and excitation module in the first bypass branch network to obtain first target characteristic information.
Optionally, before inputting the target layer CT image to be sketched, the CT image of the upper layer of the target layer CT image, and the CT image of the lower layer of the target layer CT image into the main branch network, the first bypass branch network, and the second bypass branch network in the target network model for processing, to obtain the sketching result of the bone growth point in the target layer CT image, the method further includes: acquiring a plurality of training sample sets, wherein the plurality of training sample sets at least comprise a plurality of CT image sets and standard sketching results of bone growth points in the CT image sets, one CT image set comprises an upper CT image, a middle CT image and a lower CT image, the upper CT image is a CT image of the upper layer of the middle CT image, and the lower CT image is a CT image of the lower layer of the middle CT image; inputting a plurality of training sample groups into an initial network model to obtain a target prediction sketch result of the training samples; calculating a target prediction sketch result and a loss function of a target network model according to the standard sketch result to obtain a target loss value; and optimizing parameters of the initial network model according to the target loss value to obtain a target network model.
Optionally, inputting a plurality of training samples into the initial network model to obtain a target prediction sketch result of the training samples, including: processing the upper CT image through an initial first bypass branch network in an initial network model to obtain a first prediction sketch result; processing the middle-layer CT image through an initial main branch network in an initial network model to obtain a second prediction sketch result; processing the lower CT image through an initial second bypass branch network in the initial network model to obtain a third prediction sketch result; and taking the first prediction sketching result, the second prediction sketching result and the third prediction sketching result as target prediction sketching results.
Optionally, calculating the target prediction sketching result and the loss function of the target network model according to the standard sketching result, and obtaining the target loss value includes: calculating according to the first prediction sketching result, the standard sketching result of the bone growth points in the upper CT image and the loss function to obtain a first loss value; calculating according to the second prediction sketching result, the standard sketching result of the bone growth points in the middle-layer CT image and the loss function to obtain a second loss value; calculating according to the third prediction sketching result, the standard sketching result of the bone growth points in the lower CT image and the loss function to obtain a third loss value; the first loss value, the second loss value and the third loss value are set as target loss values.
Optionally, processing the upper layer CT image through an initial first bypass branch network in the initial network model, and obtaining a first prediction sketch result includes: encoding the upper CT image through an encoder in the initial first bypass branch network to obtain second initial characteristic information of bone growth points in the upper CT image; and decoding the second initial characteristic information through a decoder in the initial first bypass branch network to obtain a first prediction sketch result.
Optionally, processing the middle layer CT image through the initial main branch network to obtain a second prediction sketch result includes: semantic enhancement is carried out on the second initial characteristic information through a first extrusion and excitation module in the initial first bypass branch network, so that fourth target characteristic information is obtained; acquiring fifth target characteristic information of a bone growth point in the lower CT image; encoding the middle layer CT image through a main encoder of an initial main branch network to obtain sixth target characteristic information of bone growth points in the middle layer CT image; combining the fourth target characteristic information, the fifth target characteristic information and the sixth target characteristic information to obtain second total characteristic information; and decoding the second total characteristic information through a main decoder of the initial main branch network to obtain a second prediction sketch result.
Optionally, acquiring the plurality of training sample sets includes: creating an initial sketching rule of bone growth points according to the human body growth rule; optimizing the initial sketching rule according to the characteristics of the bone growth points to obtain a target sketching rule; acquiring a plurality of CT image groups, and adopting a target sketching rule to sketch bone growth points of the CT image groups to obtain a standard sketching result; the CT image set and the standard sketching result are constructed into a plurality of training sample sets.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises an element.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and variations of the present application will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the application are to be included in the scope of the claims of the present application.
Claims (11)
1. A method of delineating a bone growth point, comprising:
Obtaining a target image, wherein the target image comprises: a target layer CT image to be sketched, wherein a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image;
Inputting a CT image of a target layer to be sketched and a CT image of a previous layer of the CT image of the target layer and a CT image of a next layer of the CT image of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing to obtain a sketching result of skeleton growth points in the CT image of the target layer;
Inputting the CT images of the target layer to be sketched, namely the CT image of the upper layer of the CT images of the target layer and the CT image of the lower layer of the CT images of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of skeleton growth points in the CT images of the target layer comprises the following steps:
Processing the CT image of the upper layer through the first bypass branch network to obtain first target characteristic information of a bone growth point in the CT image of the upper layer;
processing the next layer CT image through the second bypass branch network to obtain second target characteristic information of bone growth points in the next layer CT image;
Encoding the target layer CT image through a main encoder in the main branch network to obtain third target characteristic information of a bone growth point in the target layer CT image;
Combining the first target feature information, the second target feature information and the third target feature information to obtain first total target feature information;
and decoding the first total target characteristic information through a main decoder in the main branch network to obtain a sketching result of skeleton growth points in the target layer CT image.
2. The method of claim 1, wherein processing the previous layer CT image through the first bypass branch network to obtain first target feature information in the previous layer CT image comprises:
Encoding the upper layer CT image through an encoder in the first bypass branch network to obtain first initial characteristic information of a bone growth point in the upper layer CT image;
and carrying out semantic enhancement on the initial characteristic information through an extrusion and excitation module in the first bypass branch network to obtain the first target characteristic information.
3. The method of claim 1, wherein before inputting the target layer CT image to be sketched, a CT image of a layer above the target layer CT image and a CT image of a layer below the target layer CT image into a main branch network, a first bypass branch network, and a second bypass branch network in a target network model for processing, the method further comprises:
Acquiring a plurality of training sample sets, wherein the plurality of training sample sets at least comprise a plurality of CT image sets and standard sketching results of bone growth points in the plurality of CT image sets, one CT image set comprises an upper CT image, a middle CT image and a lower CT image, the upper CT image is a CT image of the upper layer of the middle CT image, and the lower CT image is a CT image of the lower layer of the middle CT image;
Inputting the plurality of training sample groups into an initial network model to obtain a target prediction sketch result of the training samples;
calculating the target prediction sketching result and a loss function of the target network model according to the standard sketching result to obtain a target loss value;
And optimizing parameters of the initial network model according to the target loss value to obtain the target network model.
4. The method of claim 3, wherein inputting the plurality of training samples into an initial network model to obtain a target prediction delineation result for the training samples comprises:
processing the upper CT image through an initial first bypass branch network in the initial network model to obtain a first prediction sketch result;
Processing the middle-layer CT image through an initial main branch network in the initial network model to obtain a second prediction sketch result;
Processing the lower CT image through an initial second bypass branch network in the initial network model to obtain a third prediction sketch result;
And taking the first prediction sketch result, the second prediction sketch result and the third prediction sketch result as the target prediction sketch result.
5. The method of claim 4, wherein calculating the target prediction sketch result and the loss function of the target network model based on the standard sketch result, the target loss value comprises:
calculating according to the first prediction sketching result, the standard sketching result of the bone growth points in the upper CT image and the loss function to obtain a first loss value;
calculating according to the second prediction sketch result, the standard sketch result of the bone growth point in the middle-layer CT image and the loss function to obtain a second loss value;
calculating according to the third prediction sketch result, the standard sketch result of the bone growth point in the lower CT image and the loss function to obtain a third loss value;
and taking the first loss value, the second loss value and the third loss value as the target loss values.
6. The method of claim 4, wherein processing the upper layer CT image through an initial first bypass branch network in the initial network model to obtain a first predictive delineation result comprises:
encoding the upper CT image through an encoder in the initial first bypass branch network to obtain second initial characteristic information of bone growth points in the upper CT image;
And decoding the second initial characteristic information through a decoder in the initial first bypass branch network to obtain the first prediction sketch result.
7. The method of claim 6, wherein processing the mid-layer CT image through the initial main branch network to obtain a second predicted delineation result comprises:
Semantic enhancement is carried out on the second initial characteristic information through a first extrusion and excitation module in the initial first bypass branch network, so that fourth target characteristic information is obtained;
Acquiring fifth target characteristic information of bone growth points in the lower CT image;
Encoding the middle-layer CT image through a main encoder of the initial main branch network to obtain sixth target characteristic information of bone growth points in the middle-layer CT image;
combining the fourth target characteristic information, the fifth target characteristic information and the sixth target characteristic information to obtain second total characteristic information;
And decoding the second total characteristic information through a main decoder of the initial main branch network to obtain the second prediction sketch result.
8. The method of claim 3, wherein obtaining a plurality of training sample sets comprises:
Creating an initial sketching rule of bone growth points according to the human body growth rule;
Optimizing the initial sketching rule according to the characteristics of the bone growth points to obtain a target sketching rule;
acquiring the plurality of CT image groups, and adopting the target sketching rule to sketch bone growth points of the plurality of CT image groups to obtain the standard sketching result;
and constructing the CT image group and the standard sketching result into a plurality of training sample groups.
9. A bone growth point delineation device for performing the bone growth point delineation method of claim 1, comprising:
The first acquisition unit is used for acquiring a target image, wherein the target image comprises: a target layer CT image to be sketched, wherein a CT image of the upper layer of the target layer CT image and a CT image of the lower layer of the target layer CT image;
The first processing unit is used for inputting the CT images of the target layer to be sketched, wherein the CT images of the upper layer of the CT images of the target layer and the CT images of the lower layer of the CT images of the target layer into a main branch network, a first bypass branch network and a second bypass branch network in a target network model for processing, and obtaining a sketching result of skeleton growing points in the CT images of the target layer.
10. A computer-readable storage medium, characterized in that the storage medium stores a program, wherein the program performs the delineating method of bone growth points according to any one of claims 1 to 8.
11. A processor for running a program, wherein the program runs on performing the method of delineating bone growth points according to any one of claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210571515.1A CN114926439B (en) | 2022-05-24 | 2022-05-24 | Bone growth point delineating method and device, storage medium and processor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210571515.1A CN114926439B (en) | 2022-05-24 | 2022-05-24 | Bone growth point delineating method and device, storage medium and processor |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114926439A CN114926439A (en) | 2022-08-19 |
CN114926439B true CN114926439B (en) | 2024-06-11 |
Family
ID=82811435
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210571515.1A Active CN114926439B (en) | 2022-05-24 | 2022-05-24 | Bone growth point delineating method and device, storage medium and processor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114926439B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110517275A (en) * | 2019-09-03 | 2019-11-29 | 上海联影医疗科技有限公司 | A kind of exchange method, device, equipment and storage medium |
CN112614144A (en) * | 2020-12-30 | 2021-04-06 | 深圳市联影高端医疗装备创新研究院 | Image segmentation method, device, equipment and storage medium |
CN112767411A (en) * | 2020-12-31 | 2021-05-07 | 罗雄彪 | Synchronous segmentation method for multiple organs of lung |
CN113538359A (en) * | 2021-07-12 | 2021-10-22 | 北京曙光易通技术有限公司 | System and method for finger vein image segmentation |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111583165B (en) * | 2019-02-19 | 2023-08-08 | 京东方科技集团股份有限公司 | Image processing method, device, equipment and storage medium |
-
2022
- 2022-05-24 CN CN202210571515.1A patent/CN114926439B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110517275A (en) * | 2019-09-03 | 2019-11-29 | 上海联影医疗科技有限公司 | A kind of exchange method, device, equipment and storage medium |
CN112614144A (en) * | 2020-12-30 | 2021-04-06 | 深圳市联影高端医疗装备创新研究院 | Image segmentation method, device, equipment and storage medium |
CN112767411A (en) * | 2020-12-31 | 2021-05-07 | 罗雄彪 | Synchronous segmentation method for multiple organs of lung |
CN113538359A (en) * | 2021-07-12 | 2021-10-22 | 北京曙光易通技术有限公司 | System and method for finger vein image segmentation |
Also Published As
Publication number | Publication date |
---|---|
CN114926439A (en) | 2022-08-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11944463B2 (en) | Pseudo-CT generation from MR data using a feature regression model | |
CN108596882B (en) | The recognition methods of pathological picture and device | |
US11386557B2 (en) | Systems and methods for segmentation of intra-patient medical images | |
Lu et al. | Deep spine: automated lumbar vertebral segmentation, disc-level designation, and spinal stenosis grading using deep learning | |
CN109308495B (en) | Apparatus and system for automatically predicting physiological condition from medical image of patient | |
CN113012155B (en) | Bone segmentation method in hip joint image, electronic equipment and storage medium | |
WO2017066317A1 (en) | Pseudo-ct generation from mr data using tissue parameter estimation | |
Cocilovo et al. | Effects of artificial deformation on cranial morphogenesis in the south central Andes | |
EP3457356A1 (en) | Method and apparatus for fully automated segmentation of joint using patient-specific optimal thresholding and watershed algorithm | |
CN111369574B (en) | Thoracic organ segmentation method and device | |
CN111798976A (en) | DDH artificial intelligence auxiliary diagnosis method and device | |
CN114170193A (en) | Automatic nasopharyngeal carcinoma target area delineating method and system based on deep learning | |
CN115375897A (en) | Image processing method, apparatus, device and medium | |
Bayat et al. | Cranial implant prediction using low-resolution 3D shape completion and high-resolution 2D refinement | |
CN114926439B (en) | Bone growth point delineating method and device, storage medium and processor | |
Wang et al. | Cranial implant design using a deep learning method with anatomical regularization | |
CN112529849B (en) | CT rib automatic counting method and device | |
CN116167990B (en) | Target recognition and neural network model processing method based on image | |
CN113129327B (en) | Method and system for generating internal general target area based on neural network model | |
CN115861464A (en) | Pseudo CT (computed tomography) synthesis method based on multimode MRI (magnetic resonance imaging) synchronous generation | |
Sample et al. | Development of a CT-based auto-segmentation model for prostate-specific membrane antigen (PSMA) positron emission tomography-delineated tubarial glands | |
Zhao et al. | [Retracted] Application of Artificial Intelligence in Radiotherapy of Nasopharyngeal Carcinoma with Magnetic Resonance Imaging | |
CN114882135A (en) | CT image synthesis method, device, equipment and medium based on MR image | |
Saraswathi et al. | Ab-mtedeep classifier trained with aagan for the identification and classification of alopecia areata | |
US11995830B2 (en) | Deep learning architecture for analyzing medical images for body region recognition and delineation |
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 |