CN113241155B - Method and system for acquiring mark points in skull side position slice - Google Patents
Method and system for acquiring mark points in skull side position slice Download PDFInfo
- Publication number
- CN113241155B CN113241155B CN202110345400.6A CN202110345400A CN113241155B CN 113241155 B CN113241155 B CN 113241155B CN 202110345400 A CN202110345400 A CN 202110345400A CN 113241155 B CN113241155 B CN 113241155B
- Authority
- CN
- China
- Prior art keywords
- point
- mark
- marker
- skull
- side position
- 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
- 210000003625 skull Anatomy 0.000 title claims abstract description 96
- 238000000034 method Methods 0.000 title claims abstract description 53
- 239000003550 marker Substances 0.000 claims abstract description 90
- 238000001514 detection method Methods 0.000 claims abstract description 68
- 238000012545 processing Methods 0.000 claims abstract description 32
- 238000003860 storage Methods 0.000 claims abstract description 12
- 230000006870 function Effects 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000010801 machine learning Methods 0.000 claims description 5
- 238000003062 neural network model Methods 0.000 claims description 5
- 210000002569 neuron Anatomy 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000012937 correction Methods 0.000 description 11
- 238000005259 measurement Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000003745 diagnosis Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 210000004556 brain Anatomy 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 210000004872 soft tissue Anatomy 0.000 description 2
- 238000010146 3D printing Methods 0.000 description 1
- 241000083547 Columella Species 0.000 description 1
- 206010061274 Malocclusion Diseases 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 210000001061 forehead Anatomy 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000007639 printing Methods 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Epidemiology (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention provides a method and a system for acquiring a mark point in a skull side position slice, electronic equipment and a computer storage medium, wherein the method comprises the following steps: acquiring a skull side position piece to be calibrated with a set size; inputting the skull side position piece to be calibrated into a first detection model for detection calibration, and identifying a set number of candidate mark areas in the skull side position piece to be calibrated; extracting a preset number of initial marker point sub-areas from each candidate marker area according to a preset offset algorithm; inputting the preset number of initial marker point sub-areas extracted from each candidate marker area into a second detection model for coordinate prediction processing to obtain a marker coordinate point corresponding to each candidate marker area; and inputting each marking coordinate point into a skull coordinate system to which the skull side position slice to be calibrated belongs, and setting each marking coordinate point input in the skull coordinate system as a target marking point in the skull side position slice.
Description
Technical Field
The invention belongs to the technical field of tooth correction, and particularly relates to a method and a system for acquiring a mark point in a skull side position slice, electronic equipment and a computer storage medium.
Background
The skull side position slice is one of the criteria that a doctor uses to judge whether a patient is dentate bulge or osseous bulge, and in the orthodontic treatment process, the doctor needs to mark some mark points according to the skull side position slice of the patient, and calculate some medical index information by using the mark points, so that diagnosis is performed by using the index information and a treatment scheme is formulated by orthodontic treatment.
The automation of the X-ray projection measurement analysis enables the diagnosis and treatment design of the maxillofacial deformity to be more accurate, and greatly reduces the burden of operators. However, if the accuracy of the marker point positioning is directly affected by the reliability of the measurement result, if the input of the marker point still stays at the level of the manual fixed point, a great amount of time is required, subjective factors are strong, human errors are inevitably generated, and the accuracy of the measurement result is endangered, so that the correction effect is affected.
The technical scheme for solving the technical problems is provided based on the application.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and provides a method and a system for acquiring mark points in a skull side position slice, electronic equipment and a computer storage medium, which solve the problems that the staff gauge distance of the skull side position slice in the prior art needs to be manually participated, the data has errors and the efficiency is lower.
The technical scheme provided by the invention is as follows:
a method for acquiring a mark point in a skull side position slice comprises the following steps:
acquiring a skull side position piece to be calibrated with a set size;
inputting the skull side position piece to be calibrated into a first detection model for detection calibration, and identifying a set number of candidate mark areas in the skull side position piece to be calibrated;
extracting a preset number of initial marker point sub-areas from each candidate marker area according to a preset offset algorithm;
inputting the preset number of initial marker point sub-areas extracted from each candidate marker area into a second detection model for coordinate prediction processing to obtain a marker coordinate point corresponding to each candidate marker area;
and inputting each marking coordinate point into a skull coordinate system to which the skull side position slice to be calibrated belongs, and setting each marking coordinate point input in the skull coordinate system as a target marking point in the skull side position slice.
Further preferably, extracting the initial marker sub-region includes:
calculating the center point of each candidate mark area;
and intercepting the mark areas with center points in different directions corresponding to the preset number from the candidate mark areas, and setting the mark areas as the initial mark point sub-areas.
Further preferably, the method further comprises:
acquiring distance information among the set number of candidate mark areas in the skull coordinate system; and carrying out grouping processing on the candidate mark areas according to the difference of the acquired distance information and meeting preset conditions.
Further preferably, the method comprises:
inputting the initial marker point sub-areas of which the number corresponds to a preset number after each grouping of the candidate marker areas into the second detection model, and carrying out prediction operation on each initial marker point sub-area according to a prediction algorithm embedded in the second detection model to obtain corresponding single-point information;
and carrying out average value operation processing on the single-point information with the preset number corresponding to the same candidate mark region, and setting a result obtained by the average value operation processing as the mark coordinate point.
Further preferably, the method comprises:
inputting the obtained initial marker point sub-region into a convolution layer in a second detection model to extract multidimensional features of the initial marker point sub-region;
carrying out operation processing on the extracted features and an activation function, and further extracting deep features of the initial marker point sub-region after pooling processing;
flattening the extracted deep features of the initial marker point sub-region, performing full-connection operation processing on the flattened deep features, and obtaining the corresponding marker coordinate points.
Further preferably, performing a loss function operation in the second detection model includes:
where n represents the number of neurons in the output layer, i.e., the predicted point coordinates (x, y) graphically, and Batch represents the size of the Batch data input to the neural network model, x * Representing the corresponding coordinates of the preceding marker point.
Further preferably, the method comprises:
acquiring a historical skull side position slice sample of the marked mark points;
and performing machine learning on the historical skull side slice sample, and training the first detection model through a neural network algorithm.
The first detection model includes regions of marker points.
An acquisition system capable of executing the method for acquiring the mark point in the skull side position slice, comprising:
the information obtaining module to be calibrated obtains a skull side position piece to be calibrated with a set size;
the mark region acquisition module inputs the skull side position piece to be calibrated into the first detection model for detection calibration, and the set number of candidate mark regions are marked in the skull side position piece to be calibrated;
the mark sub-region extraction module is used for extracting a preset number of initial mark point sub-regions on each candidate mark region according to a preset offset algorithm;
the sign coordinate point prediction module is used for inputting the preset number of initial sign point sub-areas extracted from each candidate sign area into a second detection model for coordinate prediction processing to obtain a sign coordinate point corresponding to each candidate sign area;
the target mark point setting module inputs each mark coordinate point into a skull coordinate system to which the skull side position slice to be calibrated belongs, and sets each mark coordinate point input in the skull coordinate system as a target mark point in the skull side position slice.
An electronic device comprises a processor and a memory, wherein the processor executes computer instructions stored in the memory, so that the electronic device executes the method for acquiring the mark point in the skull side bit slice.
A computer storage medium comprising computer instructions that, when executed on an electronic device, cause the electronic device to perform the method of acquiring a marker point in a cranial flap described above.
The invention provides a method and a system for acquiring a mark point in a skull side position slice, electronic equipment and a computer storage medium, which can bring at least one of the following beneficial effects:
in the method, the mark points of the skull side position slice can be automatically calibrated, the quality requirement on the input picture is not high, and no pretreatment or post-treatment is needed, so that the method has wider application range, higher prediction precision and higher prediction speed.
The method solves the problems that in the prior art, manual operation is adopted, or measurement is completed by virtue of experience of doctors, the digital age of the existing large environment is not met, and automatic diagnosis and treatment cannot be met.
Drawings
The above features, technical features, advantages and implementation thereof will be further described in the following detailed description of preferred embodiments with reference to the accompanying drawings in a clearly understandable manner.
FIG. 1 is a flowchart of an embodiment of a method for acquiring marker points in a cranial flap of the present invention;
FIG. 2 is another flow chart of an embodiment of a method for acquiring marker points in a cranial flap of the present invention;
FIG. 3 is a diagram of candidate marker regions in a cranial flap of the present invention;
FIG. 4 is a target mark initial mark point sub-area obtained from a cranial flap of the present invention;
FIG. 5 is a block diagram of an embodiment of a system for acquiring marker points in a cranial flap of the present invention;
fig. 6 is a structural diagram of the electronic device of the present invention.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following description will explain the specific embodiments of the present invention with reference to the accompanying drawings. It is evident that the drawings in the following description are only examples of the invention, from which other drawings and other embodiments can be obtained by a person skilled in the art without inventive effort.
The system and the method enable the diagnosis and treatment design of the maxillofacial deformity to be more accurate based on the automation of X-ray projection measurement and analysis, and greatly reduce the burden of operators. However, if the accuracy of the marker point positioning is directly affected by the reliability of the measurement result, if the input of the marker point still stays at the level of the manual fixed point, a great amount of time is required, subjective factors are strong, human errors are inevitably generated, and the accuracy of the measurement result is endangered, so that the correction effect is affected. In order to solve the problems in the prior art, the method and the device utilize the deep neural network to learn a large number of skull side position pieces marked by experts in case information, train a neural network model, and therefore achieve accurate automatic calibration of common marking points on the skull side position pieces and rapidly calculate required medical indexes. The solution has the advantages of high precision and high speed, greatly reduces the workload of doctors and improves the working efficiency.
Specific embodiments based on the above application include the following:
referring to fig. 1-6, the present application provides an embodiment of a method for acquiring a marker point in a skull lateral patch, specifically including:
step S100, acquiring a skull side position piece to be calibrated with a set size;
in the orthodontic process, skull state information of a patient needs to be displayed through the skull side position piece, and particularly, a more complete correction scheme is designed for the patient according to the skull state in the malocclusion correction process. In order to solve the problem of human participation in the prior art, the present application is directed to providing a solution for automatically marking a mark point required by a doctor on a skull side position slice, wherein the mark point comprises: the skull comprises a skull mark point, a maxillary mark point, a mandibular mark point and a soft tissue side mark point; the cranium marking points specifically comprise forehead points (G.), soft tissue nasal root points (Ns.), nasal apex (Prn.) nasal columella points (Cm.), subnasal points (Sn.) and the like, and can be referred to by textbooks. Therefore, the side position sheet with the set size is selected to complete the complete marking of the required mark point. Firstly, the obtained skull slice is arbitrated, and the specific adjudication size is that the required mark points are all included, and secondly, the requirement of computer modeling is met, and meanwhile, redundant work is saved.
Step S200, inputting the skull side position slice to be calibrated into a first detection model for detection calibration, and identifying a set number of candidate mark areas in the skull side position slice to be calibrated;
in the application, firstly, rough calibration is carried out through a first detection model, an approximate area of a point to be marked is found, rough calibration is carried out on the approximate area, a range with a set size is framed, the size of a specific size is set according to the use requirement, and the specific size is set as a candidate mark area.
The specific candidate marker region may be obtained, which is automatically identified in the present application, and a first detection model is constructed, where the first detection model includes: YOLO v3 model, and other target model reports (detection models such as Faster R-CNN, mask R-CNN, SSD, efficientDet, etc.); the YOLO v3 model is verified by a large number of experiments to be applied to the skull side position slice for region detection, the result can be achieved simply, the effect accords with the model of medical characteristic data, and the positioning of the positioned mark points is accurate, so that the time for subsequent data information processing is reserved. Referring to fig. 2, the following are specifically included:
s010 obtains a historical skull side position slice sample of the marked mark points;
and S020, performing machine learning on the historical skull side position slice sample, and training the first detection model through a neural network algorithm. And predicting the region of the mark point in the first detection model.
And when the first detection model is constructed, the feature extraction is carried out on the side position pieces marked by the doctor in a machine learning mode, and the first detection model meeting the use requirement is constructed according to a large number of learning processes. The first detection model can predict the detection area where each mark point is located, and because the input mode of the side bit slice is based on the pixel unit, the detected candidate mark area is a square detection area; the number of the candidate mark areas in the application is determined according to the current use requirement, and the method is not limited.
Step S300, extracting a preset number of initial marker point sub-areas from each candidate marker area according to a preset offset algorithm;
after the candidate mark area is positioned, further analyzing the candidate mark area to realize fine positioning and prepare for finding mark points; the candidate mark areas are sheared, the shearing mode is multidirectional, one candidate mark area can be sheared in multiple directions, the specific size is smaller than or equal to the number of the candidate mark areas, the number of the candidate mark areas can be set according to requirements, the number of the candidate mark areas can be 15, 20 or 30, 15-30 and the like, and the more the number of the interception is, the more accurate the accurate positioning is.
Step S400, inputting the preset number of initial marker point sub-areas extracted from each candidate marker area into a second detection model for coordinate prediction processing to obtain a marker coordinate point corresponding to each candidate marker area;
inputting each initial mark point sub-area into a second detection model for learning to obtain a required model; specific inputs include, if, as shown in FIG. 3, the candidate marker regions selected in the embodiments of the present application are m, set to A 1 ，A 2 ，···A m Each A 1 When n pieces are sheared, A is respectively 11 ,A 12 ···A 1n And A m1 ,A m2 ···A mn The method comprises the steps of carrying out a first treatment on the surface of the When the coordinate is input into the second detection model, the input mode is in the form of an initial mark point sub-area, the second detection model carries out operation processing and predicts to obtain the point coordinate, and the coordinate is the obtained candidate mark area A 1 ，A 2 ，···A m Corresponding mark points.
Referring to fig. 4, step S500 includes inputting each marker coordinate point into a skull coordinate system to which the skull side piece to be calibrated belongs, and setting each marker coordinate point input in the skull coordinate system as a target marker point in the skull side piece, x 1 ，x 2 ，···x n 。
In the method, a first detection model is constructed according to historical sample data information, a first model is utilized to roughly position a mark point on a skull side position piece to be marked, the second detection model is used for precisely positioning the mark point on the basis of rough positioning information, and in the detection positioning process, a convolution layer extraction feature required by the field of tooth correction appliances is built, so that automatic positioning of the mark point is achieved in a skull coordinate system.
Embodiments of a method for extracting an initial marker point sub-region in a candidate marker region are also provided herein, including:
step S310, calculating the center point of each candidate mark area;
step S320 intercepts a preset number of marker areas with center points in different orientations from the candidate marker areas, and sets the marker areas as the initial marker sub-areas.
See the figures; the position information of the center points is calculated in a candidate mark area, and the position information of each center point is appropriately shifted within a certain range, for example, a 96×96 or 128×128 area is cut after shifting 30 or 20 pixels, so that each candidate mark area is cut 20 times or 30 times, as in the previous embodiment. And the cutting is carried out 20 times or 30 times by taking the central point position information of the candidate mark area as the center, and each initial mark point sub-area cut is contained in the candidate mark area so as to take a more accurate positioning mark point as a data basis. In this application, after the coarse positioning is completed, the candidate marker areas need to be grouped before the input of the accurate positioning to the second detection model, and in a specific embodiment, the method is as follows:
acquiring distance information among the set number of candidate mark areas in the skull coordinate system; and carrying out grouping processing on the candidate mark areas according to the difference of the acquired distance information and meeting preset conditions.
Referring to the drawings, specific examples are labeled regions, A 1 ，A 2 ，···A m The m candidate mark areas are arranged in groups, and each area is better distinguished by the group arrangement, so that overlapping is avoided, and the method and the device are used for the applicationThe method can be used for grouping according to the size of the distance range between each candidate mark region, wherein the distance between the regions is 7mm, or different divisions such as 6mm are adopted for grouping, or a distance range is adopted, the distance ranges with the proportion of 4-5 mm are a group and are arranged according to the sizes of the brains, and the specific set distance information classification mode is dynamically adjusted according to the sizes of the brains of different persons, children and the like in the correction process due to the different sizes of the brains of each person. The present application is not limited to other grouping methods, and is applicable as long as the problem of different overlapping can be solved.
Preferably, the method comprises the steps of:
inputting the initial marker point sub-areas of which the number corresponds to a preset number after each grouping of the candidate marker areas into the second detection model, and carrying out prediction operation on each initial marker point sub-area according to a prediction algorithm embedded in the second detection model to obtain corresponding single-point information;
and carrying out average value operation processing on the single-point information with the preset number corresponding to the same candidate mark region, and setting a result obtained by the average value operation processing as the mark coordinate point.
The operation of inputting it into the second detection model includes:
inputting the obtained initial marker point sub-region into a convolution layer in a second detection model to extract multidimensional features of the initial marker point sub-region;
carrying out operation processing on the extracted features and an activation function, and further extracting deep features of the initial marker point sub-region after pooling processing;
flattening the extracted deep features of the initial marker point sub-region, performing full-connection operation processing on the flattened deep features, and obtaining the corresponding marker coordinate points.
When operation is carried out, the number of layers of the convolution layer is not limited, and the phenomena of over fitting and under fitting are avoided. The activation function is set according to actual use, and can be maximum pooling, average pooling and the like when deep features are acquired; further, extracting deep features of the input image from each sign initial sign point sub-area obtained in the first stage through a feature extraction module consisting of three convolution+activation functions+maximum pooling, flattening the three-dimensional feature layer to form a one-dimensional form, connecting three layers of full-connection layers and activation functions as sign point regression modules, and forming a final output layer by two neurons as sign coordinate points (x, y) of final prediction.
Performing a loss function operation in the second detection model includes:
where n represents the number of neurons in the output layer, i.e., the coordinates (x, y) of the landmark points on the skull coordinate system, and Batch represents the size of the Batch data input to the neural network model, x * Representing corresponding coordinates of a prior mark point, wherein the prior mark point is a mark point marked on a history side bit slice; the number of neurons in this application is set according to the use.
In the training of neural network models, for example, data is typically divided into individual batches, such as 128 pictures, and if one batch contains 32 pictures, there are 4 batches. When the loss function is calculated, the loss index value of each picture of 32 pictures of one batch is preset to be summed, so that the model is more stably converged to a local optimal solution when gradient descent is performed. For example, if the value of batch is 1, then as the gradient decreases, the parameters will oscillate so much that the resulting solution may be unstable. Thus making an appropriate determination.
See fig. 5; also provided in the present application is an embodiment of a system for acquiring a marker point in a cranial flap, comprising:
the information to be calibrated acquisition module 100 acquires a skull side position piece to be calibrated with a set size;
the mark region acquisition module 200 inputs the skull side position slices to be calibrated into a first detection model for detection calibration, and marks a set number of candidate mark regions in the skull side position slices to be calibrated;
the marker sub-region extraction module 300 extracts a preset number of initial marker sub-regions on each candidate marker region according to a preset offset algorithm;
the sign coordinate point conversion module 400 inputs the preset number of initial sign point sub-areas extracted from each candidate sign area into a second detection model to perform coordinate prediction processing, so as to obtain a sign coordinate point corresponding to each candidate sign area;
the target mark point setting module 500 inputs each mark coordinate point into a skull coordinate system to which the skull side position slice to be calibrated belongs, and sets each mark coordinate point input in the skull coordinate system as a target mark point in the skull side position slice.
The embodiment of the system for acquiring the mark point in the skull side position slice in the present application may be implemented by the embodiment of the method for acquiring the mark point in the skull side position slice, which is not described herein.
In the present application, a large number of marked sample data of the skull side pieces are required in the construction of the first detection model, and the commonly used marker points are marked on the side pieces by means of an expert. Approximately 3000 cranial side plates with prior mark points were acquired in the actual training. Each side position sheet is provided with 50-100 common mark points. The first stage is to obtain a positioning model for positioning, i.e. a first detection model. The scheme adopted is a YOLO v3 single-stage model. A skull side position sheet is input, the skull side position sheet outputs detection frames containing 50-100 mark points corresponding to the number, and the detection frames also give the category of the mark points contained in the frame. The method comprises the steps of constructing a plurality of convolution layers to extract features by a YOLO v3 model, outputting detection frames of different scales by combining a ResNet residual error connection idea and a feature pyramid structure, and selecting a target detection frame with the maximum probability by a non-maximum value inhibition method. The marker point positioning problem is converted into the target detection problem in the first stage, so that the target detection network can be used for detection. The corresponding automatic positioning and punctuation models are trained by using the data through a neural network.
The second stage is to input the initial marker sub-areas of the automatic positioning area into the second detection model, so that the accurate positions of the markers in the marker areas can be determined.
For M marker points, the data are divided into H groups altogether, and H models with the same structure are trained. The model mainly uses a convolution layer to extract features from an input image, then connects with a full link layer, calculates a loss function by using a mean square error loss, and performs back propagation. The input of the whole algorithm is a skull side position picture, and the output is a model containing M mark point positions and mark points.
In the application, the training network adopts a two-stage training scheme, whether the first-stage training image block contains the mark points or not is judged according to the image block containing the mark points, the area of each mark point is calculated, the second-stage image block is taken again according to the neighborhood information of each mark point, the regression network is used for respectively training the final position of each mark point, and the positioning precision of the mark point is improved.
The function achieved by the application is to automatically position and predict 50 mark points on the side of the skull. Two functions are required to be expanded in the future, one is to automatically determine another point from the scale, so that the interval between the points is exactly 1.5-3 cm, and the physical scale of a single pixel can be calculated; the other is that using the mark points, some common contour lines on the skull side position slice are automatically drawn, and contour curves required by subsequent analysis are generated according to the characteristic features on the skull; providing reliable medical data for orthodontic correction of doctors.
The present embodiment provides an electronic device, whose block diagram is shown in fig. 6, and the electronic device 000 may be a tablet computer, a notebook computer, or a desktop computer. The electronic device 000 may also be referred to by other names as a portable terminal, a laptop terminal, a desktop terminal, etc.
The electronic device 000 is built with a processor 010 and a memory 020, wherein the memory 020 is stored with a computer program, and the processor 010 realizes a method for acquiring the mark point in the skull side bit slice when running the computer program in the memory 020.
Processor 010 may include one or more processing cores, such as 4 core processors, 8 core processors, etc. The processor 010 may be implemented in at least one hardware form of DSP (Digital Signal Processing ), FPGA (Field-Programmable Gate Array, field programmable gate array), PLA (Programmable Logic Array ). The processor 010 may also include a main processor, which is a processor for processing data in an awake state, also called a CPU (Central Processing Unit ), and a coprocessor; a coprocessor is a low-power processor for processing data in a standby state.
In some embodiments, the processor 010 may be integrated with a GPU (Graphics Processing Unit, image processor) for rendering and drawing of content required to be displayed by the display screen. In some embodiments, the processor 010 may also include an AI (Artificial Intelligence ) processor for processing computing operations related to machine learning.
Memory 020 includes one or more computer-readable storage media, which may be non-transitory. Memory 020 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, a non-transitory computer readable storage medium in memory 020 is used to store at least one instruction, at least one program, a set of codes, or a set of instructions for execution by processor 010 to implement a method of obtaining a marker point in a cranial-lateral slice provided in an embodiment of the present invention.
And in the electronic device 000, a 050 peripheral interface device is further provided, which can be used for communicating external information numbers, and further includes creating an correction scheme after the contour line is constructed after the mark points are acquired, and further performing the production of the correction device, a 3D printing device for printing the dental model 040 in the production of the correction device, and a 030-mouth scanner for acquiring the digital dental model.
The modules of the embodiments may be stored in a computer readable storage medium if implemented as software functional modules and sold or used as a stand-alone product. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only memory (ROM), a random access memory (Random access memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It should be noted that the above embodiments can be freely combined as needed. The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.
Claims (9)
1. The method for acquiring the mark point in the skull side position slice is characterized by comprising the following steps of:
acquiring a skull side position piece to be calibrated with a set size;
inputting the skull side position piece to be calibrated into a first detection model for detection calibration, and identifying a set number of candidate mark areas in the skull side position piece to be calibrated;
extracting a preset number of initial marker point sub-areas from each candidate marker area according to a preset offset algorithm;
inputting a preset number of initial marker point sub-areas extracted from each candidate marker area into a second detection model for coordinate prediction processing to obtain a marker coordinate point corresponding to each candidate marker area, wherein the second detection model is a prediction model obtained by training a convolutional neural network;
inputting each marking coordinate point into a skull coordinate system to which the skull side position piece to be calibrated belongs, and setting each marking coordinate point input in the skull coordinate system as a target marking point in the skull side position piece;
wherein extracting the initial marker sub-region comprises:
calculating the center point of each candidate mark area;
and intercepting the mark areas with center points in different directions corresponding to the preset number from the candidate mark areas, and setting the mark areas as the initial mark point sub-areas.
2. The method for acquiring a marker point in a cranial flap according to claim 1, further comprising:
acquiring distance information among the set number of candidate mark areas in the skull coordinate system; and grouping the candidate mark areas according to the difference of the acquired distance information and meeting preset conditions, wherein the preset conditions comprise that all areas in the same group are not overlapped after grouping.
3. The method for acquiring a marker point in a cranial flap according to claim 2, comprising:
inputting the initial marker point sub-areas of which the number corresponds to a preset number after each grouping of the candidate marker areas into the second detection model, and carrying out prediction operation on each initial marker point sub-area according to a prediction algorithm embedded in the second detection model to obtain corresponding single-point information;
and carrying out average value operation processing on the single-point information with the preset number corresponding to the same candidate mark region, and setting a result obtained by the average value operation processing as the mark coordinate point.
4. A method of obtaining a marker point in a cranial flap according to claim 3, comprising:
inputting the obtained initial marker point sub-region into a convolution layer in a second detection model to extract multidimensional features of the initial marker point sub-region;
carrying out operation processing on the extracted features and the activation function, and further extracting deep features of the initial marker point sub-region after pooling processing;
flattening the extracted deep features of the initial marker point sub-region, performing full-connection operation processing on the flattened deep features, and obtaining the corresponding marker coordinate points.
5. The method of claim 4, wherein performing a loss function operation in the second detection model comprises:
wherein n represents the number of neurons of the output layer, x b,n Is the coordinates (x, y) of the predicted point on the graph, batch represents the size of the Batch data input to the neural network model, x * b,n Representing the corresponding coordinates of the preceding marker point.
6. The method for acquiring a marker point in a cranial flap according to claim 1, comprising:
acquiring a historical skull side position slice sample of the marked mark points;
and performing machine learning on the historical skull side slice sample, and training the first detection model through a neural network algorithm.
7. An acquisition system operable to perform the method for acquiring a marker point in a cranial flap according to any one of claims 1 to 6, comprising:
the information obtaining module to be calibrated obtains a skull side position piece to be calibrated with a set size;
the mark region acquisition module inputs the skull side position piece to be calibrated into the first detection model for detection calibration, and the set number of candidate mark regions are marked in the skull side position piece to be calibrated;
the mark sub-region extraction module is used for extracting a preset number of initial mark point sub-regions on each candidate mark region according to a preset offset algorithm;
the sign coordinate point prediction module is used for inputting a preset number of initial sign point sub-areas extracted from each candidate sign area into a second detection model for coordinate prediction processing to obtain a sign coordinate point corresponding to each candidate sign area, wherein the second detection model is a prediction model obtained by training a convolution model;
the target mark point setting module is used for inputting each mark coordinate point into a skull coordinate system to which the skull side position piece to be calibrated belongs, and setting each mark coordinate point input in the skull coordinate system as a target mark point in the skull side position piece; wherein, the mark subregion draws the module to include:
calculating the center point of each candidate mark area;
and intercepting the mark areas with center points in different directions corresponding to the preset number from the candidate mark areas, and setting the mark areas as the initial mark point sub-areas.
8. An electronic device comprising a processor and a memory, the processor executing computer instructions stored by the memory, causing the electronic device to perform the method of acquiring a landmark in a cranial-lateral slice according to any one of claims 1 to 6.
9. A computer storage medium comprising computer instructions which, when run on an electronic device, cause the electronic device to perform the method of acquiring a landmark in a cranial-lateral slice as claimed in any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110345400.6A CN113241155B (en) | 2021-03-31 | 2021-03-31 | Method and system for acquiring mark points in skull side position slice |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110345400.6A CN113241155B (en) | 2021-03-31 | 2021-03-31 | Method and system for acquiring mark points in skull side position slice |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113241155A CN113241155A (en) | 2021-08-10 |
| CN113241155B true CN113241155B (en) | 2024-02-09 |
Family
ID=77130875
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110345400.6A Active CN113241155B (en) | 2021-03-31 | 2021-03-31 | Method and system for acquiring mark points in skull side position slice |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113241155B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113948190A (en) * | 2021-09-02 | 2022-01-18 | 上海健康医学院 | Method and equipment for automatically identifying X-ray skull positive position film cephalogram measurement mark points |
| CN115797341B (en) * | 2023-01-16 | 2023-04-14 | 四川大学 | Method for automatically and immediately judging natural head position of skull side position X-ray film |
| CN117726788A (en) * | 2023-05-16 | 2024-03-19 | 荣耀终端有限公司 | Image area positioning method and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108846342A (en) * | 2018-06-05 | 2018-11-20 | 四川大学 | A kind of harelip operation mark point recognition system |
| CN109146845A (en) * | 2018-07-16 | 2019-01-04 | 中南大学 | Head image sign point detecting method based on convolutional neural networks |
| CN109544536A (en) * | 2018-11-26 | 2019-03-29 | 中国科学技术大学 | The fast automatic analysis method of hip joint x-ray image |
| CN111933253A (en) * | 2020-07-14 | 2020-11-13 | 北京邮电大学 | Neural network-based marking point marking method and device for bone structure image |
-
2021
- 2021-03-31 CN CN202110345400.6A patent/CN113241155B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108846342A (en) * | 2018-06-05 | 2018-11-20 | 四川大学 | A kind of harelip operation mark point recognition system |
| CN109146845A (en) * | 2018-07-16 | 2019-01-04 | 中南大学 | Head image sign point detecting method based on convolutional neural networks |
| CN109544536A (en) * | 2018-11-26 | 2019-03-29 | 中国科学技术大学 | The fast automatic analysis method of hip joint x-ray image |
| CN111933253A (en) * | 2020-07-14 | 2020-11-13 | 北京邮电大学 | Neural network-based marking point marking method and device for bone structure image |
Non-Patent Citations (3)
| Title |
|---|
| 复杂背景图像的文本信息提取研究;孙巧榆;《中国博士学位论文全文数据库 信息科技辑》;正文第25-30页 * |
| 平面凸轮的CAD/CAM***设计及研究;李宗成;《中国优秀硕士学位论文全文数据库 工程科技II辑》;正文第42-43页 * |
| 探索下颌数字化模型重叠的方法;王梦含等;《中国美容医学》;第第29卷卷(第第10期期);第44-48页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113241155A (en) | 2021-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113241155B (en) | Method and system for acquiring mark points in skull side position slice | |
| CN110599528B (en) | Unsupervised three-dimensional medical image registration method and system based on neural network | |
| EP4044115A1 (en) | Image processing method and apparatus based on artificial intelligence, and device and storage medium | |
| JP6607364B2 (en) | Prediction system | |
| CN111932547B (en) | Method and device for segmenting target object in image, electronic device and storage medium | |
| CN112790879B (en) | Tooth axis coordinate system construction method and system of tooth model | |
| CN115187540A (en) | Automatic segmentation method and system for human three-dimensional vertebral cancellous bone image | |
| CN115797730B (en) | Model training method and device and head shadow measurement key point positioning method and device | |
| CN113782184A (en) | Cerebral apoplexy auxiliary evaluation system based on facial key point and feature pre-learning | |
| CN113421333B (en) | Tooth local coordinate system determination method, system, equipment and computer storage medium | |
| CN110827283A (en) | Head and neck blood vessel segmentation method and device based on convolutional neural network | |
| CN113270172B (en) | Method and system for constructing contour lines in skull lateral position slice | |
| US11475568B2 (en) | Method for controlling display of abnormality in chest x-ray image, storage medium, abnormality display control apparatus, and server apparatus | |
| CN116824209A (en) | Bone window prediction method and system | |
| CN112215829B (en) | Positioning method of hip joint standard tangent plane and computer equipment | |
| CN113920128A (en) | Knee joint femur tibia segmentation method and device | |
| CN109509189B (en) | Abdominal muscle labeling method and labeling device based on multiple sub-region templates | |
| CN112734886A (en) | Biological model die-threading detection method and device, electronic equipment and storage medium | |
| CN111383212A (en) | Method for analyzing ultrasonic video image of pelvic floor | |
| CN113674261B (en) | Bone detection method, system, electronic device and storage medium | |
| CN113160186B (en) | Lung lobe segmentation method and related device | |
| CN112651959B (en) | CT intracranial hemorrhage detection system based on dynamic map loss neural network | |
| CN112700487B (en) | Method and system for acquiring measuring scale marks in skull lateral position slice | |
| CN110706222B (en) | Method and device for detecting bone region in image | |
| CN117152407A (en) | Automatic positioning method for head shadow measurement mark points |
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 |