Method and device for assisting positioning of dental implant site, storage medium and electronic device

Abstract

The application provides a method and device for assisting in positioning a tooth implant site based on a deep learning method, a computer readable storage medium, an electronic device and a surgical robot system. The method for assisting in positioning the tooth implant site comprises: performing dimension reduction processing on input image data; scaling the image after the dimension reduction processing to obtain a standardized two-dimensional image; using a preset tooth implant site positioning information extraction model to process the input standardized two-dimensional image file to obtain all tooth sites and tooth loss condition information of each tooth site. Thus, the tooth loss tooth site in all tooth sites can be identified, and all tooth sites with teeth, tooth sites with possible tooth loss or tooth sites with possible wisdom tooth growth are comprehensively predicted, thereby solving the problem that multiple tooth sites may be confused with each other in the traditional method and improving the accuracy. Because the input size can be arbitrarily adjusted, the demand for hardware devices is greatly reduced and the prediction speed is improved.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, specifically to a technique for accurately locating dental implant sites using an AI-assisted robot, particularly to a method and apparatus for locating dental implant sites using deep learning-assisted methods, and also to a computer-readable storage medium, electronic device, and surgical robot system. Background Technology

[0002] As people's pace of life accelerates, more and more health details are being neglected by the public, especially oral hygiene, leading to a large number of oral health problems among young people. As a result, the demand for dental implants is increasing. However, the industry faces the problem of not having enough experienced doctors to perform implant procedures, which leads to longer waiting times for patients, causing them to suffer additional pain and incurring significant costs.

[0003] The technical challenge to be addressed is how to increase the accuracy of surgical robots in locating implant sites, assist doctors with limited clinical experience, increase the implantation speed of experienced doctors, and enable skilled but inexperienced doctors to start working as soon as possible, so that dental patients can receive treatment more quickly.

[0004] Prior to this, although numerous academic methods existed for finding dental implant sites, a very difficult problem always remained. It is well known that the normal number of human teeth is 28 (although...). Figure 2 The notation shown also includes the occasional case of 32 beads, such as... Figure 2 As shown, in medical terms, these 28 teeth are divided into four regions based on the FDI tooth position representation, corresponding to quadrants 1, 2, 3, and 4 of a mathematical rectangular coordinate system. Each quadrant contains teeth numbered 1-7. For ease of communication with patients or non-dental professionals, directions are used to describe them. For example, the upper left tooth number 6 refers to the sixth molar (#26) in the second quadrant. The left-right dividing point is the upper and lower incisors, which are numbered 1, and so on inwards. Tooth number 8, outside of these, is the wisdom tooth position; even if extracted, it does not require implantation (this is why previous methods for finding dental implant sites used a 28-tooth representation instead of 32). However, even with previous techniques, finding implantable teeth is graphically very difficult because, according to the basic rules of deep learning, the number of inputs and outputs of a neural network is fixed, but tooth loss varies from person to person. Here are three examples of patients:

[0005] Patient 1 is missing one upper left seventh molar (#27), which is the most basic pathological phenomenon and one that can be resolved by most methods at present.

[0006] Patient 2 is missing the left upper 1st incisor (#21) and the right upper 1st incisor (#11), and this patient has the left upper 8th wisdom tooth (#28), which makes it difficult to use the current mainstream dental neural network based on the 28-tooth representation method for reasoning;

[0007] Patient 3 is missing the left upper 3rd canine (#23) and the right upper 1st incisor (#11), and is also missing the right lower 6th bicuspid (#46), which makes it basically impossible to reason.

[0008] As mentioned above, the input and output of a neural network are determined. In the world coordinates, a point needs 3 coordinates to be determined, i.e., the positions of the x, y, and z axes in the coordinates need to be determined, which requires 3 outputs. However, if a patient is missing multiple teeth, such as patient 3, it is not possible to complete the entire plan at once, and only one tooth at a time can be positioned, which is very inefficient. In addition, for patient 2, because the patient is missing 2 incisors and is adjacent, and has a wisdom tooth on the left upper side, the recognition effect has a high probability that: the tooth position is empty and has only 1 tooth, and the left upper side has 7 teeth from 2-8, so the left upper 2-8 teeth are recognized as 1-7 teeth, and the recognition result is that the left upper side has no missing teeth and the right upper side is missing one incisor, so the output result is incorrect. That is, although the patient is missing one incisor, because there is a wisdom tooth in the quadrant where the incisor is located, the side with the wisdom tooth still has 7 teeth, and in the past positioning method, it is possible to mistake the side with the wisdom tooth as normal and not need to implant a tooth, and of course the output result is not correct. For example, the case of patient 2, although two incisors are missing, only one gap is detected, so when recognizing, although the gap is wide, it is believed that there are 6 teeth in one quadrant and 7 teeth in the other quadrant, and it is possible to recognize that only 1 tooth is missing (see Figure 8 ).

[0009] Therefore, in order to avoid this situation, it is necessary to accurately position the tooth implant site before or during the operation. SUMMARY

[0010] To solve the above problems, the purpose of the present application is to provide a method and device for assisting in positioning the tooth implant site based on a deep learning method, which can effectively improve the positioning speed and accuracy of dental positioning implant surgery, and at the same time use more basic technology as much as possible to reduce the hardware requirements.

[0011] According to an aspect of the present application, a method for assisting in positioning a dental implant site based on a deep learning method is provided, which comprises: step S1: performing dimension reduction processing on input image data; step S2: scaling the image after dimension reduction processing to obtain a standardized two-dimensional image; and step S3: using a preset dental implant site positioning information extraction model to process the input standardized two-dimensional image file to obtain all tooth sites and tooth loss status information of each tooth site.

[0012] Preferably, the dimension reduction processing in step S1 comprises: on the basis of obtaining a complete set of dental CT image information, adding data of image pixels corresponding one by one in the Z-axis direction to reduce three-dimensional XYZ-axis data to two-dimensional XY-axis data, thereby completing dimension reduction.

[0013] Preferably, step S2 comprises: scaling the image after dimension reduction processing to a 512*512 two-dimensional image.

[0014] Preferably, after the addition of data, an equal reduction processing is performed to complete the numerical compression of image pixels.

[0015] Preferably, the specific structure of the neural network of the preset dental implant site positioning information extraction model is: after three continuous processes I of the following processing are performed on the input image file, 2 times of 3*3 convolution and 1 time of pooling are sequentially performed, then feature map flattening is performed, and then 2 times of full connection calculation are performed to reset the data to a 4*8*4 matrix, thereby obtaining the final result containing all tooth sites and tooth loss status of each tooth site, wherein the process I is that three branches of a first branch to a third branch are used in the processing of the input data, and then the three branches are spliced into one branch using concat splicing, and then 2 times of 3*3 convolution and 1 time of pooling are performed, wherein the first branch is that the input data is respectively convolved using a 3*3 convolution kernel, a 5*5 convolution kernel, and a 7*7 convolution kernel, and then spliced using add, the second branch is that the input data is respectively convolved using a 3*3 convolution kernel, a 5*5 convolution kernel, a 7*7 convolution kernel, and a 1*1 convolution kernel, and then spliced using add, and the third branch is that only 1 time of 3*3 convolution is performed.

[0016] Preferably, the output form of the image file processed in the step S3 is a three-dimensional matrix of 4*8*4, the first dimension of the matrix is the four quadrants of 1, 2, 3 and 4; the second dimension refers to each tooth in the corresponding quadrant area, and the order is tooth positions 1-7 and wisdom tooth position 8; in the third dimension, the first three values respectively represent the positions of the x, y and z three quadrant axes in the world coordinate, and the fourth value is a floating point number between 0 and 1, greater than 0.5 indicating that the tooth position at this position is vacant and needs to be implanted, and less than 0.5 indicating that the tooth position at this position is not vacant and the tooth is normal.

[0017] According to another aspect of the present application, there is provided a device for assisting in positioning tooth implant sites, which uses any of the above methods to assist in positioning tooth implant sites, comprising: a dimension reduction unit configured to perform dimension reduction processing on input image data; an image scaling unit configured to scale the image after dimension reduction processing to obtain a standardized two-dimensional image; and a tooth site and tooth loss status determination unit configured to use a preset tooth implant site positioning information extraction model to process the input standardized two-dimensional image file to obtain all tooth sites and tooth loss status information for each tooth site.

[0018] According to still another aspect of the present application, there is provided a computer readable storage medium storing a computer program, wherein the computer program is configured to cause a computer to implement any of the above methods by executing the computer program.

[0019] According to yet another aspect of the present application, there is provided an electronic device comprising a processor and a memory for storing instructions executable by the processor, wherein the processor is configured to read and execute the instructions from the memory to implement any of the above methods.

[0020] According to still another aspect of the present application, there is provided a surgical robot system using any of the above methods to assist a surgical robot in using deep learning to position dental implant sites.

[0021] According to the present application, the prediction of multiple tooth implant sites can be completed at the same time, and the number of missing adjacent teeth can also be accurately identified. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 A flowchart of a method for assisting in positioning tooth implant sites based on a deep learning method according to an embodiment of the present application is schematically shown.

[0023] Figure 2 Tooth sites based on FDI tooth site representation are schematically shown, wherein the left side of the figure corresponds to the right side of the patient.

[0024] Figure 3The principle of data addition of the input picture along the Z direction is schematically shown.

[0025] Figure 4 A flow chart of a tooth-net network infrastructure according to an embodiment of the present application is shown.

[0026] Figure 5 is Figure 4 A structural diagram of one process module and its input in the process module.

[0027] Figure 6 A structural diagram of an apparatus provided according to an embodiment of the present application is shown.

[0028] Figure 7 is a structure of an electronic device provided according to an embodiment of the present application.

[0029] Figure 8 A schematic diagram showing that two adjacent front teeth are missing, which is easily misidentified as missing only one tooth. DETAILED DESCRIPTION

[0030] Exemplary embodiments of the present application are described in detail below with reference to the accompanying drawings. The exemplary embodiments described below and illustrated in the drawings are intended to teach the principles of the present application, with the provision that the present application can be employed in various environments and for several different applications. Therefore, the protection scope of the present application is defined by the appended claims rather than the exemplary embodiments described below, and the exemplary embodiments are not intended to be, and should not be taken to be, a restrictive description of the scope of the present application. In addition, unless otherwise specified, the relative arrangement, numerical expressions, and numerical values of the components and steps set forth in these embodiments are not intended to limit the scope of the present application. Those skilled in the art can understand that the terms "first", "second", "Sn", and the like in the embodiments of the present application are only used to distinguish different steps, devices, or modules, and do not represent any specific technical meaning, nor do they represent a necessary logical sequence between them. For any component, data, or structure mentioned in the embodiments of the present application, it can be understood as one or more, unless specifically limited or given a contrary implication by the context or the prior art. The description of the various embodiments of the present application emphasizes the differences between the various embodiments, and the same or similar parts can be referred to each other, and for the sake of brevity, will not be repeated. For the technologies, methods, and devices known to those skilled in the relevant art, they can not be discussed in detail, but in appropriate cases, the technologies, methods, and devices should be considered as part of the specification.

[0031] The embodiments of the present application can be applied to terminal devices, computer systems, servers, and the like electronic devices, which can operate with many other general-purpose or special-purpose computing system environments or configurations. Examples of well-known terminal devices, computing systems, environments, and / or configurations suitable for use with terminal devices, computer systems, servers, and the like electronic devices include, but are not limited to, personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, minicomputer systems, mainframe computer systems, and distributed cloud computing technology environments comprising any of the above systems, and the like. Terminal devices, computer systems, servers, and the like electronic devices can be described in the general context of computer system-executable instructions, such as program modules, being executed by the computer system. Generally, program modules can include routines, programs, objects, components, logic, data structures, and the like, which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, in which tasks are performed by remote processing devices that are linked through a communications network. In a distributed cloud computing environment, program modules can be located in local or remote computer system storage media, including storage devices.

[0032] According to the present application, a method for positioning dental implant sites is provided, in which a brand new neural network is used to find the implant sites.

[0033] According to the present application, the above two defects are corrected, i.e., multiple dental implant sites can be predicted at the same time, and the number of missing adjacent teeth can be accurately identified. This is because the present application has broken out of the misunderstanding of the previous method and has not focused on the missing teeth, but has enlarged the field of view to include all the determined and possible dental sites in the entire oral cavity in the output. A novel neural network (which can be named tooth-net) is used in the present application, which is specially designed in the input and output, as described below.

[0034] Reference Figure 1 A method for positioning dental implant sites is provided, and the specific method steps include:

[0035] Step S1: Dimensionality reduction processing is performed on the input image data.

[0036] Thus, when the image is input, the entire set of dental CT images of the patient can be first subjected to dimensionality reduction processing.

[0037] Preferably, the dimension reduction processing comprises reducing the three-dimensional data of XYZ axis to two-dimensional data of XY axis, also known as compressing the Z axis to 1, based on the whole set of image information, so as to greatly reduce the input size, which means greatly reducing the operation cost.

[0038] The dimension reduction idea is close to the Low Variance Filter. Figure 3 The principle of this method is to directly reduce the data with similar variance in the same dimension to one data, so that the specific operation method is very simple. In this way, the original z*y*x data is compressed to 1, and the data type is still a three-dimensional matrix. Only the Z axis dimension is compressed to 1, so it can be discarded in the operation and directly changed into a two-dimensional matrix. Of course, the three-dimensional matrix can still be used directly.

[0039] In other words, each CT is an image, and the data format of the image input into the computer is a two-dimensional matrix composed of one pixel. Then, a three-dimensional matrix is formed by inputting a large number of pictures. As shown in the figure, Figure 3 Each picture in the three-dimensional matrix is added one by one to become a picture corresponding to a two-dimensional matrix.

[0040] Preferably, only the image pixels corresponding to the Z axis direction in the whole CT are added, and the data reduction effect is very significant. Because the density of the skeleton in the human body is very high, and the density of the teeth in the skeleton is very high; at the same time, the dental CT layer thickness is 0.5mm, and the length of a tooth plus the root is basically around 2cm, at which time there will be continuous multiple layers of super high density images in the image, and the dental image is from the mandible position to the position below the eyeball. This position does not contain the whole brain and skull image, so the only tissue that can appear in this space is the tooth, and the density of the upper and lower jaw bones is not as high as that of the tooth, so there is no need to calculate the variance of the data. After direct addition, simple geometric scaling can achieve efficient dimension reduction. At the same time, the purpose of geometric scaling is to make the data smaller in order to facilitate calculation.

[0041] Preferably, the scale of the equal ratio reduction is 30,000 times. Because the Hu value range of CT is -1024-3071, just like the normal optical image is 0-255 color gamut, so for the above-mentioned continuous multi-layer super high density image, a whole set of dozens or hundreds of images will get a very large value. For example, assuming the density value of the tooth is about 700 Hu, and each CT is about 0.5 mm thick, and the tooth is 20 mm long, the total Hu value is 700*20 / 0.5, plus the upper and lower dental CT images also contain other tissues, and the Hu value will be as high as tens of thousands. At this time, reduce this value by a scale, preferably 10,000 or more 30,000 or less binary whole ten numbers, such as 524,288, or select a smaller one such as 212,144 as needed, so as to be able to determine to 10.

[0042] For example, one patient has good teeth with a density value of more than 900, and another patient has poor teeth with a density value of only 500, and reducing by 30,000 times can clearly observe the density change. For example, when the test step is 200, 32,600 to 32,800 times can be preferred to obtain good observation effect.

[0043] As described above, the color gamut of the CT image is -1024-3071, and in the traditional processing method, in order to view the CT image, the CT image is compressed to 0-255 color gamut by using the image viewing software, so as to facilitate the naked eye observation. For example, open a window between -1024 and 3071 to view, the center point is the window level, and the included color gamut is the window width, for example, the window level is 200, and the window width is 400, that is, open a 400-width window at the position of 200 to view, then the view is from (200-400 / 2) to (200+400 / 2) that is 0-400, and then the values in this range are compressed to 0-255, and the values exceeding 400 are uniformly set to 255, and the values less than 0 are uniformly set to 0. Unlike this, according to the present application, the traditional processing method is no longer needed because observation is no longer needed, only the data file needs to be input into the neural network for training and prediction, and all the layers are added together, even if the window width and window level are set, but the value after addition is still very large, and the naked eye cannot observe, so the traditional processing method is not used, but the dimension reduction processing containing equal ratio reduction is used.

[0044] As a comparison example, there is also a method of reducing the density value of each image to 0-10 or 0-1 by the same proportion, but in the case of reducing the density value of each image to 0-1 by the same proportion, for example, if the highest value of the density value of a patient is 500, it is reduced to 1, and if the highest value of the density value of another patient is 900, it is also reduced to 1, a problem occurs in that two sets of data have different standards, and the "1" in one set of data represents 900, and the "1" in the other set of data represents 500, and the characteristics of the data are confused.

[0045] Step S2: standardizing and scaling the image after dimension reduction, for example, into a two-dimensional image of 512*512, inputting the two-dimensional image into a preset tooth implant site positioning information extraction model (tooth-net) as shown in the drawings. Figure 4 、 5 tooth-net).

[0046] The scaling in this step is to make the image more standard, and the initial data of most CT images of a single section is also 512*512, so as to try to completely retain the original information.

[0047] Step S3: using a preset tooth implant site positioning information extraction model to process the input standardized two-dimensional image file to obtain all tooth sites and tooth loss condition information of each tooth site.

[0048] The output form of the information is a 4*8*4 three-dimensional matrix, the first dimension of the matrix is 1, 2, 3, and 4 quadrants, and the information of 28 teeth is not directly output but output in 4 regions, which on the one hand can make it more convenient for doctors to read the results, and on the other hand can reduce the amount of information in the same dimension, thereby increasing the accuracy; the second dimension refers to each tooth in the region, and the order is tooth sites 1-7 and tooth site 8; in the third dimension, the first three values respectively represent the positions of the x, y, and z three-dimensional coordinate axes in the world coordinate system, and the fourth value is a 0-1 floating point number, greater than 0.5 indicating that the tooth site is vacant and needs to be implanted, and less than 0.5 indicating that the tooth site is not vacant and the tooth is normal.

[0049] For example, for the aforementioned patient 2, the following example of a three-dimensional matrix can be output, wherein as described above, in the third dimension (i.e., in each group of four parallel data), the first three values represent the predicted coordinates of the implant point in the case of needing to implant teeth, and the fourth value represents how much probability there is of needing to implant teeth, greater than 0.5 indicating the need, and less than indicating the need not. The following only shows an output example of a 4*8*4 three-dimensional matrix: [

[0051] 1st quadrant, upper right (four regions have the same tooth positions, so the tooth positions are omitted in other quadrants) [

[0053] central incisor [211.46, 83.11, 46.23, 0.6273],

[0054] lateral incisor [213.53, 83.93, 46.57, 0.0236],

[0055] canine [215.03, 85.02, 46.79, 0.1479],

[0056] premolar 1 [216.99, 87.50, 47.13, 0.0359],

[0057] premolar 2 [218.86, 88.97, 47.55, 0.1006],

[0058] molar 1 [220.08, 90.13, 47.97, 0.0004],

[0059] molar 2 [221.97, 91.93, 48.22, 0.0768],

[0060] wisdom tooth [223.04, 92.36, 48.85, 0.0000],

[0061] ],

[0062] 2nd quadrant, upper left [

[0064] [211.46, 83.11, 46.23, 0.8263],

[0065] [209.93, 83.93, 46.49, 0.0036],

[0066] [208.03, 85.02, 46.67, 0.1719],

[0067] [206.99, 87.50, 47.23, 0.0331],

[0068] [205.42, 88.97, 47.45, 0.0017],

[0069] [203.95, 90.13, 47.96, 0.0005],

[0070] [201.97, 91.88, 48.25, 0.0178],

[0071] [199.73, 92.59, 48.97, 0.0000],

[0072] ],

[0073] 3 Quadrant, Lower Left [

[0075] [211.46, 83.11, 40.23, 0.0273],

[0076] [213.53, 83.93, 39.57, 0.0336],

[0077] [215.03, 85.02, 39.09, 0.1489],

[0078] [216.99, 87.50, 38.13, 0.0354],

[0079] [218.86, 88.97, 38.55, 0.1015],

[0080] [220.08, 90.13, 37.97, 0.0069],

[0081] [221.97, 91.93, 37.22, 0.0448],

[0082] [223.17, 93.31, 36.52, 0.0000],

[0083] ],

[0084] 4 Quadrant, Lower Right [

[0086] [211.46, 83.11, 40.23, 0.6273],

[0087] [209.93, 83.93, 39.57, 0.0236],

[0088] [208.03, 85.02, 39.09, 0.1479],

[0089] [206.99, 87.50, 38.13, 0.0359],

[0090] [205.42, 88.97, 38.55, 0.1006],

[0091] [203.95, 90.13, 37.37, 0.0004],

[0092] [201.97, 91.93, 36.72, 0.0768],

[0093] [199.02, 93.15, 34.82, 0.0000],

[0094] ],

[0095] ].

[0096] In this way, the following neural network can be used to process the dimension-reduced image file when locating the tooth implant site, so as to obtain all tooth positions (including the possible position of the 8th wisdom tooth, if present) and the tooth loss status of each tooth position.

[0097] Figure 5 Show Figure 4 The process I in the tooth-net network shown is specifically illustrated as part of a pre-defined tooth implant location information extraction model, along with the input intended to represent the current module structure.

[0098] The specific structure of course I is as follows: For the data processing of the image file input to the current module structure, three branches (branch1 to branch3) are used. First, the input image file is convolved with a 3*3 kernel (conv3), a 5*5 kernel (conv5), and a 7*7 kernel (conv7), and then concatenated with an add function, forming the first branch (branch1). Second, the first branch is repeated, but an additional 1*1 convolution (conv1) is performed, followed by concatenation with an add function, forming the second branch (branch2). Third, the input image file is convolved with a 3*3 convolution (conv3) only once, forming the third branch (branch3). Then, the three branches are concatenated into one using concat, followed by two 3*3 convolutions and one pooling operation (pool 2), forming course I.

[0099] Figure 4A tooth-net network overall structure diagram is shown. The convolution channel number in the above process I starts from 4 (of course, it is not limited to this, and a binary integer 2^n can also be set), and the channel number is doubled every time process I is performed once. After the above standardized two-dimensional image file data is sequentially input into process I (course) three times, the image is changed from the original 512*512*1 shape to a 128*128*16 feature map. After the feature map is subjected to 2 times of 3*3 convolution (conv3) and 1 time of pooling (pool), it becomes a 64*64*16 feature map. At this time, the feature map is flattened into a 1*262144 vector, and then 2 times of full connection calculation are performed. After the data is reset to a 4*8*4 matrix, the final result (output) is obtained. In the training process, the classic structure of deep learning is adopted. This structure has higher stability after a large number of practical applications. The activation function adopts a Relu activation function (h=max(0, x)), and the loss function in the convolution structure adopts an intersection over union: IOUloss ( ), and the average absolute error L1loss (L1loss=main(abs(y-h))) is used in the full connection layer, wherein abs represents the absolute value, and main represents the average number. The final output is divided into two paths. The first path outputs the coordinate point, does not use the activation function, and directly outputs the numerical value. The output value is 4*8*3. The second path outputs whether a tooth needs to be implanted. The sigmoid activation function ( ) is used. The output of this activation function is between 0 and 1. If the result is greater than 0.5, a tooth needs to be implanted. If the result is less than 0.5, a tooth does not need to be implanted (in the previous formula, h is the prediction result, x is the input value, and y is the true result). At this time, all tooth positions and tooth loss conditions of the tooth positions can be predicted or determined. As an example of the specific final result, it has been described above and will not be repeated here. At this time, the training data 500 sets of CT and the test data 50 sets of CT of the neural network can achieve an ideal accuracy, and the Euclidean distance of the label error is within 1 millimeter. Because it is a three-dimensional coordinate, the Euclidean distance is used for measurement.

[0100] Regarding flattening the feature map, for example, a 5*5*5 matrix is tiled to become a 1*1*125 matrix, that is, a length 125 vector, so as to facilitate the subsequent full connection operation (fully). Since a convolution operation is used in image processing, a convolution kernel is initialized, and the weight update during training is also updated for the convolution kernel. The full connection is to initialize a plurality of neurons, and the neurons are updated during training.

[0101] According to the present application, the above-mentioned problems of the second patient and the third patient can be properly solved, because the eighth wisdom tooth information that may exist is contained in the output data of the present application, and through the deep learning in the present application, the missing tooth positions in all tooth positions can be easily and accurately identified from the output data file, and through the all-round prediction of all tooth positions, tooth positions that may fall out, or tooth positions that may grow wisdom teeth, the problem that multiple tooth positions may be confused with each other in the traditional method is solved, the accuracy is improved, at the same time, because the input size can be arbitrarily adjusted, the demand for hardware devices is greatly reduced, and the prediction speed is improved.

[0102] The device for assisting in positioning tooth implant positions provided by the embodiment of the present application, as shown in Figure 6 The device for assisting in positioning tooth implant positions provided by the embodiment of the present application, as shown in

[0103] The dimension reduction unit is configured to perform dimension reduction processing on the input image data;

[0104] The image scaling unit is configured to scale the image after the dimension reduction processing to obtain a standardized two-dimensional image;

[0105] The all-tooth-position-and-tooth-missing-condition determination unit is configured to use a preset tooth implant position positioning information extraction model to process the input two-dimensional image file to obtain all-tooth-position-and-tooth-missing-condition information.

[0106] As shown in Figure 7 The electronic device 50 provided by the embodiment of the present application includes one or more processors 51 and a memory 52. The electronic device can be any one or both of the first device and the second device, or a single device independent of them, which can communicate with the first device and the second device to receive the collected input signals therefrom.

[0107] The processor 51 can be a central processing unit (CPU) or other form of processing unit that has data processing and / or instruction executing capability, and can control other components in the electronic device to perform desired functions. The memory 52 can include one or more computer program products that can include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory, for example, can include random access memory (RAM), cache memory, and / or the like. The non-volatile memory, for example, can include read-only memory (ROM), hard disk, flash memory, and / or the like. One or more computer program instructions can be stored on the computer-readable storage media, which the processor 51 can execute to implement the method of positioning a dental implant site according to various embodiments of the present disclosure described above and / or other desired functions. In one example, the electronic device can further include an input device 53 and an output device 54, which are interconnected through a bus system and / or other form of connection mechanism (not shown). In addition, the input device 53 can include, for example, a keyboard, a mouse, and / or the like. The output device 54 can output various information to the outside. The output device 54 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and / or the like. Of course, in order to simplify the description, Figure 7 Only some of the components related to the present disclosure in the electronic device are shown, and components such as buses, input / output interfaces, and the like are omitted. In addition to this, the electronic device can further include any other appropriate components according to the specific application.

[0108] In addition to the above-described method and device, embodiments of the present disclosure can be a computer program product including computer program instructions that, when executed by a processor, cause the processor to perform steps of the method of positioning a dental implant site according to various embodiments of the present disclosure described in the above "Exemplary Method" section of the specification.

[0109] The computer program product can be written in any combination of one or more programming languages, including an object-oriented programming language such as Java, C++, and / or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device or entirely on the remote computing device or server.

[0110] In addition, an embodiment of the present disclosure can also be a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, cause the processor to perform the steps of the method of positioning a dental implant site according to various embodiments of the present disclosure described in the above “Exemplary Method” section of the specification.

[0111] The computer readable storage medium can take the form of one or more combinations of any type of computer readable medium. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can include, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include the following: an electrical connection having one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0112] The above describes the basic principles of the present disclosure in conjunction with specific embodiments, but it should be noted that the advantages, benefits, effects, etc. mentioned in the present disclosure are only examples and are not limiting, and these advantages, benefits, effects, etc. cannot be considered as mandatory for each embodiment of the present disclosure. In addition, the above specific details are only for the purpose of example and understanding, and are not limiting, and the above details do not limit the present disclosure to the above specific details.

[0113] The various embodiments described herein are described with respect to progressive embodiments, each embodiment highlighting differences from other embodiments, and the same or similar parts between embodiments are cross-referenced. For system embodiments, since they substantially correspond to method embodiments, they are described more simply, and the relevant parts are cross-referenced to the parts of the method embodiments. The block diagrams of devices, apparatuses, equipment, systems referred to in this disclosure are only illustrative examples and are not intended to require or imply that the connections, arrangements, configurations must be as shown in the block diagrams. As those skilled in the art will appreciate, these devices, apparatuses, equipment, systems can be connected, arranged, configured in any manner. Words such as "comprise", "include", "have", and the like are open-ended words, mean "including but not limited to", and can be used interchangeably. The words "or" and "and" used herein mean the word "and / or", and can be used interchangeably, unless the context clearly indicates otherwise. The word "such as" used herein means the phrase "such as but not limited to", and can be used interchangeably. The methods and devices of the present disclosure can be implemented in many ways. For example, the methods and devices of the present disclosure can be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is only for illustration, and the steps of the method of the present disclosure are not limited to the above specifically described order, unless otherwise specifically described. In addition, in some embodiments, the present disclosure can also be implemented as programs recorded in recording media, which include machine-readable instructions for implementing the method according to the present disclosure. Thus, the present disclosure also covers the recording media storing the programs for executing the method according to the present disclosure.

[0114] In summary, according to the present disclosure, the following technical solutions are provided.

[0115] (1) A method for assisting in positioning a dental implant site based on a deep learning method, comprising:

[0116] Step S1: performing dimensionality reduction processing on the input image data;

[0117] Step S2: scaling the image after dimensionality reduction to obtain a standardized two-dimensional image;

[0118] Step S3: using a pre-set dental implant site positioning information extraction model to process the input standardized two-dimensional image file to obtain all tooth sites and tooth loss status information of each tooth site.

[0119] (2) Preferably, the dimensionality reduction processing in step S1 comprises: on the basis of obtaining a complete set of dental CT image information, by adding data of image pixels corresponding one by one in the Z-axis direction, three-dimensional data of XYZ axes is reduced to two-dimensional data of XY axes, thereby completing the dimensionality reduction.

[0120] (3) Preferably, the step S2 comprises: scaling the image after dimension reduction processing into a 512*512 two-dimensional image.

[0121] (4) Preferably, after the addition of the above data, an isometric reduction processing is performed to complete the numerical compression of the image pixels.

[0122] (5) Preferably, the specific structure of the neural network of the preset tooth implant site positioning information extraction model is: after the following process I is performed on the input image file for three times in succession, 2 times of 3*3 convolution and 1 time of pooling are performed in turn, then feature map flattening is performed, and then 2 times of full connection calculation are performed to reset the data into a 4*8*4 matrix, so as to obtain the final result containing all tooth sites and tooth loss conditions of each tooth site, wherein the process I is to use three branches of the first branch to the third branch in the input data processing, and then use concat splicing to splice the three branches into one branch, and then perform 2 times of 3*3 convolution and 1 time of pooling, wherein the first branch is to use 3*3 convolution kernel, 5*5 convolution kernel and 7*7 convolution kernel to perform convolution on the input data respectively, and then use add to splice, the second branch is to use 3*3 convolution kernel, 5*5 convolution kernel, 7*7 convolution kernel and 1*1 convolution kernel to perform convolution on the input data respectively, and then use add to splice, and the third branch is to perform 3*3 convolution only once.

[0123] (6) Preferably, the output form of the image file after processing in the above step S3 is a 4*8*4 three-dimensional matrix, the first dimension of the matrix is four quadrants of 1, 2, 3 and 4; the second dimension refers to each tooth in the corresponding quadrant area, and the order is tooth sites of 1-7 and tooth site of 8 wisdom tooth; in the third dimension, the first three values respectively represent the positions of x, y and z three coordinate axes in the world coordinate, and the fourth value is a floating point number of 0-1, greater than 0.5 indicates that the tooth site at this position is vacant and needs to be implanted, and less than 0.5 indicates that the tooth site at this position is not vacant and the tooth is normal.

[0124] (7) Preferably, a device for assisting in positioning tooth implant sites is provided, which uses the method for assisting in positioning tooth implant sites according to any one of the above to assist in positioning tooth implant sites, comprising:

[0125] a dimension reduction unit configured to perform dimension reduction processing on the input image data;

[0126] an image scaling unit configured to scale the image after dimension reduction processing to obtain a standardized two-dimensional image;

[0127] The tooth position and tooth loss condition determination unit is configured to use preset tooth implant position positioning information extraction model to process the input standardized two-dimensional image file to obtain all tooth positions and tooth loss condition information of each tooth position.

[0128] (8) A computer-readable storage medium storing a computer program for causing a computer to implement the above method by executing the computer program.

[0129] (9) An electronic device comprising a processor and a memory for storing instructions executable by the processor, the processor being configured to read and execute the instructions from the memory to implement the above method.

[0130] (10) A surgical robot system using the above method to assist a surgical robot in using deep learning for dental implant point positioning.

[0131] In the devices, apparatuses, and methods of the present disclosure, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions of the present disclosure. The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other aspects without departing from the scope of the present disclosure. Thus, the present disclosure is not intended to be limited to the aspects shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein. While the application has been described with reference to various specific embodiments, it should be understood that modifications can be made without departing from the spirit and scope of the inventive concepts described. Accordingly, it is intended that the application not be limited to the described embodiments, but will have the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assisting in positioning a dental implant site based on a deep learning method, comprising: Step S1: performing dimension reduction processing on input image data; Step S2: scaling the image after dimension reduction processing to obtain a standardized two-dimensional image; Step S3: using a preset dental implant site positioning information extraction model to process the input standardized two-dimensional image file to obtain all tooth sites and tooth loss status information for each tooth site, The specific structure of the neural network of the preset dental implant site positioning information extraction model is as follows: after three consecutive processes I for the input image file, 2 times of 3*3 convolution and 1 time of pooling are performed in sequence, then the feature map is flattened, and then 2 times of full connection calculation are performed to reset the data to a 4*8*4 matrix, thereby obtaining the final result containing all tooth sites and tooth loss status information for each tooth site, wherein the process I is to use three branches, i.e., a first branch, a second branch, and a third branch, to process the input data, and then use concat to splice the three branches into one branch, and then perform 2 times of 3*3 convolution and 1 time of pooling, wherein the first branch is to use 3*3, 5*5, and 7*7 convolution kernels to convolve the input data respectively, and then use add to splice, the second branch is to use 3*3, 5*5, 7*7, and 1*1 convolution kernels to convolve the input data respectively, and then use add to splice, and the third branch is to perform only one time of 3*3 convolution.

2. The method of claim 1, wherein the dimension reduction processing in step S1 includes: on the basis of obtaining a complete set of dental CT image information, adding data of image pixels corresponding to the Z-axis direction one by one to reduce three-dimensional XYZ-axis data to two-dimensional XY-axis data, thereby completing dimension reduction.

3. The method of claim 1, wherein the step S2 includes: scaling the image after dimension reduction processing to a 512*512 two-dimensional image.

4. The method of claim 2, wherein after the addition of the data, an equal reduction processing is performed to complete the numerical compression of the image pixels.

5. The method of claim 1, wherein the output form after processing the image file in step S3 is a 4*8*4 three-dimensional matrix, the first dimension of the matrix is four quadrants of 1, 2, 3, and 4; the second dimension refers to each tooth in the corresponding quadrant area, and the order is tooth sites 1-7 and 8 wisdom tooth site; in the third dimension, the first three values respectively represent the positions of the x, y, and z three coordinate axes in the world coordinate system, and the fourth value is a floating point number between 0 and 1, greater than 0.5 indicating that the tooth site at this position is vacant and needs to be implanted, and less than 0.5 indicating that the tooth site at this position is not vacant and the tooth is normal. Using the method of any one of claims 1-5 to assist in positioning a dental implant site, comprising: ​ ​ ​ 6. A device for assisting in positioning a tooth implant, characterized in that, ​ a dimension reduction unit configured to perform dimension reduction on the input image data; an image scaling unit configured to scale the image after the dimension reduction to obtain a standardized two-dimensional image; a tooth site and tooth loss condition determination unit configured to use preset tooth implant site positioning information extraction model to process the input standardized two-dimensional image file to obtain all tooth site and tooth loss condition information.

7. A computer readable storage medium storing a computer program, characterized in that, The computer program is used to enable a computer to implement the method of any one of claims 1-5 by executing the computer program.

8. An electronic device comprising a processor and a memory for storing instructions executable by the processor, the electronic device characterized by: The processor is configured to read and execute the instructions from the memory to implement the method of any one of claims 1-5.

Images