CN114418989A

CN114418989A - Dental segmentation method, device, equipment and storage medium for oral medical image

Info

Publication number: CN114418989A
Application number: CN202210055028.XA
Authority: CN
Inventors: 任春霞; 田庆; 韩月乔; 温静; 倪自强
Original assignee: Beijing Ruiyibo Technology Co ltd
Current assignee: Beijing Ruiyibo Technology Co ltd
Priority date: 2022-01-18
Filing date: 2022-01-18
Publication date: 2022-04-29

Abstract

The embodiment of the application provides a tooth segmentation method, a tooth segmentation device, tooth segmentation equipment and a storage medium for oral medical images. Wherein the method comprises the following steps: determining a first training set and a second training set according to the acquired oral medical image sample; generating a first pipeline fingerprint for training the tooth classification prediction module according to the first data fingerprint of the first training set, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint; generating a second pipeline fingerprint for training the tooth partition prediction module according to the second data fingerprint of the second training set, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint; respectively predicting the classification and the partition of teeth in the target oral medical image; and determining a tooth segmentation result according to the tooth classification prediction result and the tooth partition prediction result. The scheme can effectively improve the tooth segmentation efficiency and the tooth segmentation accuracy in the oral medical image.

Description

Dental segmentation method, device, equipment and storage medium for oral medical image

Technical Field

The embodiment of the application relates to the field of artificial intelligence, in particular to a tooth segmentation method and device for an oral medical image, electronic equipment and a computer readable medium.

Background

In the field of oral surgery robots, tooth segmentation is an important step in which an oral surgery robot can accurately implant teeth. In the process of tooth implantation, the oral surgery robot needs to be positioned to a vacant tooth position and to place an implant, and meanwhile, the specific position, depth and angle of the implant need to be confirmed according to the adjacent teeth of the vacant tooth position, so that the oral surgery robot is required to be capable of segmenting the teeth in the oral cavity as accurately as possible so as to identify the position of each tooth in the oral cavity.

In the prior art, tooth segmentation methods are mainly classified into two major categories, the first category is based on traditional algorithms such as threshold segmentation, region growing, etc., and the second category is deep learning methods such as PointNet model. The first method requires a priori knowledge to be provided manually, and when the medical images of the mouth become more and more complex and diverse, it cannot adapt well to all the medical images of the mouth, thereby reducing the accuracy of dental segmentation of the medical images of the mouth. Compared with the first method, the second method can learn the optimal model by processing a large amount of oral medical images, can segment the teeth end to end, and effectively improves the accuracy and generalization capability. However, before the model training, training parameters need to be manually prepared in advance for the model to be trained. Different training parameters will affect the accuracy of the model's segmentation of the teeth in the oral medical image. In addition, the efficiency of the model in segmenting teeth in oral medical images is also affected.

Therefore, how to effectively improve the tooth segmentation efficiency and the tooth segmentation accuracy in the oral medical image becomes a technical problem to be solved at present.

Disclosure of Invention

The present application aims to provide a tooth segmentation method, a tooth segmentation device, an electronic device, and a computer-readable medium for dental medical images, which are used to solve the technical problem in the prior art of how to effectively improve the tooth segmentation efficiency and the tooth segmentation accuracy in the dental medical images.

According to a first aspect of the embodiments of the present application, a tooth segmentation method of an oral medical image is provided. The method comprises the following steps: according to an oral medical image sample obtained by scanning an oral cavity by an oral cavity scanning device, determining a first training set for training a tooth classification prediction module in a tooth classification prediction model and a second training set for training a tooth partition prediction module in a tooth partition prediction model; adaptively generating a first pipeline fingerprint used for training the tooth classification prediction module according to a first data fingerprint of the first training set through a first adaptive module in the tooth classification prediction model, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint; adaptively generating a second pipeline fingerprint used for training the tooth partition prediction module according to a second data fingerprint of the second training set through a second adaptive module in the tooth partition prediction model, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint; predicting the classification of teeth in the target oral medical image through the trained tooth classification prediction module, and predicting the subareas of the teeth in the target oral medical image through the trained tooth subarea prediction module; and determining the tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image.

According to a second aspect of the embodiments of the present application, there is provided a dental segmentation apparatus for an oral medical image. The device comprises: the first determination module is used for determining a first training set used for training a tooth classification prediction module in a tooth classification prediction model and a second training set used for training a tooth partition prediction module in a tooth partition prediction model according to an oral medical image sample obtained by scanning an oral cavity by an oral cavity scanning device; the first training module is used for adaptively generating a first pipeline fingerprint used for training the tooth classification prediction module according to the first data fingerprint of the first training set through a first self-adaptive module in the tooth classification prediction model, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint; the second training module is used for adaptively generating a second pipeline fingerprint used for training the tooth partition prediction module according to a second data fingerprint of the second training set through a second self-adaptive module in the tooth partition prediction model, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint; the prediction module is used for predicting the classification of the teeth in the target oral medical image through the trained tooth classification prediction module and predicting the subareas of the teeth in the target oral medical image through the trained tooth subarea prediction module; and the second determination module is used for determining the tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image.

According to a third aspect of embodiments of the present application, there is provided an electronic apparatus, including: one or more processors; a storage configured to store one or more programs; when executed by the one or more processors, cause the one or more processors to implement a method for dental segmentation of an oral medical image as described in the first aspect of the embodiments of the present application.

According to a fourth aspect of embodiments of the present application, there is provided a computer-readable storage medium having a computer program stored thereon, which when executed by a processor, implements a tooth segmentation method for an oral medical image according to the first aspect of embodiments of the present application.

By the tooth segmentation scheme of the oral medical image provided by the embodiment of the application, the pipeline fingerprint does not need to be prepared for the tooth classification prediction module to be trained and the pipeline fingerprint does not need to be prepared for the tooth partition prediction module to be trained in advance, but the pipeline fingerprint is prepared by the first self-adaptive module in the tooth classification prediction model, adaptively generating a first pipeline fingerprint for training the tooth classification prediction module according to the first data fingerprint of the first training set, and passing through a second adaptive module in the tooth partition prediction model, according to the second data fingerprints of the second training set, the second pipeline fingerprints used for training the tooth partition prediction module are generated in a self-adaptive mode, the trouble of manually preparing the pipeline fingerprints is avoided, the interference of factors is reduced, and then the tooth segmentation efficiency in the oral medical image is effectively improved. In addition, the diversity of the segmentation results of the teeth in the oral medical images caused by manual preparation of pipeline fingerprints is effectively avoided, and the segmentation accuracy of the teeth in the oral medical images is effectively improved.

Furthermore, the classification of the teeth in the target oral medical image is predicted through the trained tooth classification prediction module, the partition of the teeth in the target oral medical image is predicted through the trained tooth partition prediction module, and the tooth segmentation result of the target oral medical image is determined according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image, so that the trained tooth classification prediction module and the trained tooth partition prediction module can more fully utilize the limited data of the target oral medical image, and the tooth segmentation accuracy of the teeth in the target oral medical image is effectively improved.

Drawings

Some specific embodiments of the present application will be described in detail hereinafter by way of illustration and not limitation with reference to the accompanying drawings. The same reference numbers in the drawings identify the same or similar elements or components. Those skilled in the art will appreciate that the drawings are not necessarily drawn to scale. In the drawings:

fig. 1 is a flowchart illustrating steps of a tooth segmentation method for medical images of the oral cavity according to an embodiment of the present application;

fig. 2 is a schematic structural diagram of a tooth segmentation apparatus for dental medical images according to a second embodiment of the present application;

fig. 3 is a schematic structural diagram of an electronic device in a third embodiment of the present application;

fig. 4 is a hardware structure of an electronic device according to a fourth embodiment of the present application.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, but not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application shall fall within the scope of the protection of the embodiments in the present application.

Referring to fig. 1, a flowchart illustrating steps of a tooth segmentation method for medical images of the oral cavity according to a first embodiment is shown.

Specifically, the tooth segmentation method for medical images of the oral cavity provided by the embodiment includes the following steps:

in step S101, a first training set for training a tooth classification prediction module in a tooth classification prediction model and a second training set for training a tooth partition prediction module in a tooth partition prediction model are determined according to an oral medical image sample obtained by scanning an oral cavity by an oral cavity scanning device.

In the present embodiment, the oral cavity scanning device may be understood as a device that scans the oral cavity of a patient to obtain an oral medical image, and the oral cavity scanning device may be a Cone beam computed tomography (Cone beam CT, CBCT) or an oral cavity scanner. The medical image of the oral cavity may be understood as a three-dimensional image obtained by imaging the oral cavity. The oral medical image sample may be understood as an oral medical image used as a training sample of a model. The tooth classification prediction model may include a first adaptation module and a tooth classification prediction module coupled to the first adaptation module. The tooth classification prediction model may be nnU-Net (Self-Adapting frame for U-Net-Based Medical Image Segmentation, an adaptive Framework for U-Net-Based Medical Image Segmentation). The method is a deep learning network model in the medical segmentation field, and is called nnU-Net for short. The improved U-Net model is improved on the basis of the traditional U-Net, the traditional U-Net is composed of an encoding part and a decoding part, the appearance is like U, therefore, the U-Net and nnU-Net do not change the model structure of the U-Net, but add an adaptive module, analyze data preprocessing, model scale and parameters, post-processing part in detail and select suitable parameters automatically. The tooth classification prediction module may be U-Net (Convolutional network for Biomedical Image Segmentation). The tooth partition prediction model may include a second adaptation module and a tooth partition prediction module connected to the second adaptation module. The tooth zone prediction model may be nnU-Net. The tooth zone prediction module can be U-Net. The second training set comprises an oral medical image sample and tooth partition labels thereof for training the tooth partition prediction module. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In one specific example, the oral medical image may be loaded from the CBCT to obtain a data set of the oral medical image sample. The data format of the data set is then converted to the image format required for training the model. For example, the dicom format of the dataset is converted to the ni.gz format required for training the model, while the dataset is scaled by 8: and 2, dividing the model into a training set and a test set in proportion, wherein the training set covers various oral medical image samples as much as possible, the model training is facilitated, different features are learned to complete the segmentation of the oral medical images, and the test set is used for evaluating the model obtained by the training. For the tooth classification task, the first training set comprises the oral medical image samples and tooth classification labels thereof used for training the tooth classification prediction module. For the tooth partition task, the second training set includes the oral medical image samples and tooth partition labels thereof for training the tooth partition prediction module. The dental medical image samples in the first training set are the same as the dental medical image samples in the second training set, but the tooth classification labels are different from the tooth partition labels. And the nnU-NET is adopted to respectively finish tooth classification and tooth partition. The tooth classification labels include 7 classifications, 1, 2, 3, 4, 5, 6, 7, respectively. The tooth zone label includes 4 categories, 1, 2, 3, 4 respectively. Specifically, 28 teeth are labeled 1-4, the upper left 7 teeth are labeled 1, the lower left 7 teeth are labeled 2, the upper right 7 teeth are labeled 3, and the lower right 7 teeth are labeled 4. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In step S102, a first pipeline fingerprint for training the tooth classification prediction module is adaptively generated according to the first data fingerprint of the first training set by a first adaptive module in the tooth classification prediction model, and the tooth classification prediction module is trained according to the first training set and the first pipeline fingerprint.

In this embodiment, the first adaptation module may be understood as an adaptively selected program module. The first data fingerprint of the first training set may be understood as attributes of the first training set, including attributes of the oral medical image sample in the first training set and attributes of a training task associated with the first training set. The first data fingerprint of the first training set may include a size of the oral medical image sample in the first training set, a number of voxels of different spatial dimensions of the oral medical image sample in the first training set, a voxel spacing of the oral medical image sample in the first training set, gray scale information of the oral medical image sample in the first training set, and a task category of a training task associated with the first training set. Wherein the task class of the training task associated with the first training set may be a dental classification. The first pipeline fingerprint may be understood as a training parameter related to the training of the tooth classification prediction module. When the tooth classification prediction module is U-Net, the U-Net comprises three networks, namely 2D-Unet, 3D-Unet and 3D cascade Unet. Therefore, the first pipeline fingerprint comprises identification information of the network to be trained in the tooth classification prediction module and network hyper-parameters of the network to be trained. The identification information may be understood as information for indicating a network type, which may indicate that the network type is 2D-Unet, may also indicate that the network type is 3D-Unet, and may also indicate that the network type is 3D cascaded Unet. The network hyper-parameters of the network to be trained may include a learning rate of the network to be trained, a loss function of the network to be trained, an optimizer of the network to be trained, a training batch size of the network to be trained, an iteration number of the network to be trained, and the like.

The learning rate of the network to be trained can be dynamically adjusted in the training process of the network to be trained:

wherein lr represents a current learning rate, baselr represents an initial learning rate, which may be 0.01, epoch represents the iteration number of the network to be trained, maxepoch represents the maximum iteration number, and power (taking 0.99) controls a curve shape.

The loss function of the network to be trained can be the sum of the Dice loss and the cross entropy loss. The optimizer employs a Stochastic Gradient Descent (SGD) algorithm. The number of iterations of the network to be trained may be 1000 per round. The training batch size of the network to be trained may be 250 samples per iteration. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, when the identification information of the network to be trained in the tooth classification prediction module is adaptively generated, the identification information of the network to be trained which is fixedly set may be determined as the identification information of the adaptively generated network to be trained by the first adaptive module. When the learning rate of the network to be trained in the tooth classification prediction module is adaptively generated, the learning rate of the network to be trained which is fixedly set can be determined as the learning rate of the adaptively generated network to be trained through the first adaptive module. When the loss function of the network to be trained in the tooth classification prediction module is adaptively generated, the loss function of the fixedly set network to be trained can be determined as the loss function of the adaptively generated network to be trained through the first adaptive module. When the optimizer of the network to be trained in the tooth classification prediction module is adaptively generated, the optimizer of the network to be trained which is fixedly arranged can be determined as the optimizer of the adaptively generated network to be trained through the first adaptive module. When the training batch size of the network to be trained in the tooth classification prediction module is generated in a self-adaptive manner, the training batch size of the network to be trained in the tooth classification prediction module can be generated in a self-adaptive manner through the first self-adaptive module according to the size of the oral medical image sample in the first training set and the size of the GPU memory. When the iteration number of the network to be trained in the tooth classification prediction module is generated in an adaptive manner, the iteration number of the network to be trained which is fixedly set can be determined as the iteration number of the network to be trained which is generated in an adaptive manner through the first adaptive module. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, the method further comprises: adaptively selecting a preprocessing mode for the first training set according to the first data fingerprint of the first training set through the first adaptive module; according to the preprocessing mode, preprocessing the first training set to obtain the preprocessed first training set, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint, including: and training the tooth classification prediction module according to the preprocessed first training set and the preprocessed first pipeline fingerprint. Therefore, the preprocessing mode is selected for the first training set in a self-adaptive mode according to the first data fingerprint of the first training set through the first self-adaptive module without manually relying on prior knowledge, and the preprocessing mode is preprocessed for the first training set according to the preprocessing mode, so that the efficiency of preprocessing the first training set is effectively improved, and the accuracy of preprocessing the first training set is also effectively improved. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the preprocessing includes clipping, resampling, and normalization. When adaptively selecting a clipping mode for the first training set, the first adaptive module may determine the clipping mode of the fixedly set oral medical image samples in the first training set as the adaptively selected clipping mode. When the resampling method is adaptively selected for the first training set, the resampling method may be adaptively selected for the first training set by the first adaptive module according to a voxel pitch of the oral medical image samples in the first training set. When the normalization method is adaptively selected for the first training set, the normalization method may be adaptively selected for the first training set by the first adaptive module according to the voxel size of the oral medical image sample in the first training set. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the preprocessing manner includes clipping, resampling and normalization, and the preprocessing the first training set according to the preprocessing manner to obtain the preprocessed first training set includes: cutting the oral medical image samples in the first training set to obtain the cut oral medical image samples; resampling the cut oral medical image samples according to the voxel spacing of the cut oral medical image samples to obtain the resampled oral medical image samples; normalizing the resampled oral medical image sample according to the voxel size of the resampled oral medical image sample to obtain the normalized oral medical image sample, and obtaining the preprocessed first training set according to the normalized oral medical image sample. And the cutting operation is used for cutting zero-value areas of all oral medical image samples in the first training set to reserve non-zero-value areas, reducing the size of the image and reducing the calculated amount of the model. Different oral medical image samples may have different voxel spacings due to different oral scanning devices or device acquisition protocols, and in order to ensure that the model can learn correct spatial semantics, the voxel spacings of the oral medical image samples are resampled to the median voxel spacings. The oral medical image sample after resampling is normalized according to the voxel size of the oral medical image sample after resampling, so that adverse effects caused by singular data can be avoided, and model convergence can be accelerated. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, when the clipped oral medical image sample is resampled according to the voxel space of the clipped oral medical image sample, calculating the ratio of the maximum voxel axial space of the clipped oral medical image sample to the minimum voxel axial space of the clipped oral medical image sample; if the ratio is less than or equal to 3, adopting cubic linear interpolation to resample the cut oral medical image sample; and if the ratio is larger than 3, resampling the cut oral medical image sample in the plane by adopting cubic linear interpolation, and resampling the cut oral medical image sample out of the plane by adopting a nearest neighbor algorithm. Calculating the average value and the standard deviation of the voxel size of the oral medical image sample after resampling when the oral medical image sample after resampling is normalized according to the voxel size of the oral medical image sample after resampling; and normalizing the resampled oral medical image sample according to the average value and the standard deviation of the voxel size of the sampled oral medical image sample to obtain the normalized oral medical image sample. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, the method further comprises: adaptively selecting a data enhancement mode for the first training set through the first adaptive module; according to the data enhancement mode, performing data enhancement on the first training set to obtain the first training set after data enhancement, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint, including: and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint after data enhancement. Therefore, the data enhancement mode is selected for the first training set in a self-adaptive mode through the first self-adaptive module without manually relying on prior knowledge, and the data enhancement mode is selected for the first training set in a self-adaptive mode according to the data enhancement mode, so that the efficiency of data enhancement on the first training set is effectively improved, and the accuracy of data enhancement on the first training set is also effectively improved. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the data enhancement mode includes rotation scaling, gaussian plus noise, gaussian blur, brightness processing, or contrast processing. When adaptively selecting a rotation scaling manner for the first training set, the fixedly set rotation scaling manner of the first training set may be determined as the adaptively generated rotation scaling manner of the first training set by the first adaptive module. When the gaussian noise adding mode is adaptively selected for the first training set, the fixedly set gaussian noise adding mode of the first training set may be determined as the adaptively generated gaussian noise adding mode of the first training set by the first adaptive module. When the gaussian fuzzy manner is adaptively selected for the first training set, the fixedly set gaussian fuzzy manner of the first training set may be determined as the adaptively generated gaussian fuzzy manner of the first training set by the first adaptive module. When adaptively selecting the brightness processing mode for the first training set, the first adaptive module may determine the brightness processing mode of the first training set that is fixedly set as the brightness processing mode of the adaptively generated first training set. When the contrast processing mode is adaptively selected for the first training set, the first adaptive module may determine the contrast processing mode of the first training set that is fixedly set as the contrast processing mode of the adaptively generated first training set. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the data enhancement mode includes rotation scaling, gaussian plus noise, gaussian blur, brightness processing, or contrast processing. When the first training set is subjected to data enhancement according to the data enhancement mode, rotationally scaling the oral medical image samples in the first training set to obtain the rotationally scaled oral medical image samples, and obtaining the data-enhanced first training set according to the rotationally scaled oral medical image samples; or, performing gaussian noise on the oral medical image samples in the first training set to obtain gaussian noise-added oral medical image samples, and obtaining data-enhanced first training set according to the gaussian noise-added oral medical image samples; or, performing gaussian blurring on the oral medical image samples in the first training set to obtain the oral medical image samples after gaussian blurring, and obtaining the first training set after data enhancement according to the oral medical image samples after gaussian blurring; or, performing brightness processing on the oral medical image samples in the first training set to obtain brightness-processed oral medical image samples, and obtaining the first training set with enhanced data according to the brightness-processed oral medical image samples; or, performing contrast processing on the oral medical image samples in the first training set to obtain the oral medical image samples after the contrast processing, and obtaining the first training set after data enhancement according to the oral medical image samples after the contrast processing. The gaussian noise adding can be understood as adding noise into the oral medical image sample, the probability density function of the noise follows gaussian distribution, the gaussian blur can be understood as actually performing filtering processing on the oral medical image sample, the filtering mode is gaussian filtering, the filtering mode is image blur, important information of the oral medical image sample is extracted, the brightness processing can be understood as changing the brightness of the oral medical image sample, and the contrast processing can be understood as changing the contrast of the oral medical image sample. The more and more the oral medical image samples in the first training set, the more beneficial the model training and the improvement of the generalization capability of the model. In order to train the model as good as possible on the limited first training set, the oral medical image samples in the first training set are expanded in a data enhancement mode, the quality of the training samples is improved, noise data are increased, and the robustness of the model is improved. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, the first pipeline fingerprint comprises identification information of a network to be trained in the tooth classification prediction module and a network hyper-parameter of the network to be trained. When the tooth classification prediction module is trained according to the first training set and the first pipeline fingerprint, determining the network to be trained according to the identification information, and setting the network to be trained according to the network hyper-parameter so as to obtain the set network to be trained; predicting the classification of teeth in the oral medical image samples in the first training set through the set network to be trained so as to obtain tooth classification prediction results of the oral medical image samples; and training the set network to be trained according to the tooth classification prediction result of the oral medical image sample and the tooth classification labeling result of the oral medical image sample. Therefore, the trained network can accurately classify the teeth of the target oral medical image. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the tooth classification prediction result of the oral medical image sample can be understood as prediction probability data of a tooth classification label to which a voxel in the oral medical image sample belongs, and the tooth classification labeling result of the oral medical image sample can be understood as labeling probability data of the tooth classification label to which the voxel in the oral medical image sample belongs. When the set network to be trained is trained according to the tooth classification prediction result of the oral medical image sample and the tooth classification labeling result of the oral medical image sample, determining the difference value between the prediction probability data and the labeling probability data through a target loss function; and adjusting the set parameters of the network to be trained based on the difference value. Wherein the objective loss function may be a sum of a Dice loss and a cross entropy loss. When the parameters of the set network to be trained are adjusted, a back propagation algorithm or a random gradient descent algorithm can be adopted to adjust the parameters of the set network to be trained. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In a specific example, the currently obtained prediction probability data is evaluated by determining a difference value between the prediction probability data and the labeled probability data, so as to be used as a basis for subsequently training the set network to be trained. Specifically, the difference value may be reversely transmitted to the set network to be trained, so as to iteratively train the set network to be trained. The training of the network to be trained after the setting is an iterative process, and this embodiment describes only one training process, but it should be understood by those skilled in the art that this training mode may be adopted for each training of the network to be trained after the setting until the training of the network to be trained after the setting is completed. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S103, a second pipeline fingerprint for training the tooth partition prediction module is adaptively generated according to a second data fingerprint of the second training set by a second adaptive module in the tooth partition prediction model, and the tooth partition prediction module is trained according to the second training set and the second pipeline fingerprint.

In this embodiment, the second adaptation module may be understood as an adaptively selected program module. The second data fingerprint of the second training set may be understood as attributes of the second training set, including attributes of the oral medical image sample in the second training set and attributes of the training task associated with the second training set. The second data fingerprint of the second training set may include a size of the oral medical image sample in the second training set, a number of voxels of different spatial dimensions of the oral medical image sample in the second training set, a voxel spacing of the oral medical image sample in the second training set, gray scale information of the oral medical image sample in the second training set, and a task category of a training task associated with the second training set. Wherein the task category of the training task associated with the second training set may be a dental zone. The second conduit fingerprint may be understood as a training parameter related to the training of the dental segment prediction module. When the tooth partition prediction module is U-Net, the U-Net comprises three networks, namely 2D-Unet, 3D-Unet and 3D cascade Unet. Thus, the second conduit fingerprint includes identification information of the network to be trained in the tooth zone prediction module and network hyper-parameters of the network to be trained. The identification information may be understood as information for indicating a network type, which may indicate that the network type is 2D-Unet, may also indicate that the network type is 3D-Unet, and may also indicate that the network type is 3D cascaded Unet. The network hyper-parameters of the network to be trained may include a learning rate of the network to be trained, a loss function of the network to be trained, an optimizer of the network to be trained, a training batch size of the network to be trained, an iteration number of the network to be trained, and the like.

In a specific example, when the identification information of the network to be trained in the tooth partition prediction module is adaptively generated, the identification information of the network to be trained which is fixedly set may be determined as the identification information of the adaptively generated network to be trained by the second adaptation module. When the learning rate of the network to be trained in the tooth partition prediction module is adaptively generated, the learning rate of the network to be trained which is fixedly set can be determined as the learning rate of the adaptively generated network to be trained through the second adaptive module. When the loss function of the network to be trained in the tooth partition prediction module is adaptively generated, the loss function of the fixedly set network to be trained can be determined as the loss function of the adaptively generated network to be trained through the second adaptive module. When the optimizer of the network to be trained in the tooth partition prediction module is adaptively generated, the optimizer of the network to be trained which is fixedly arranged can be determined as the optimizer of the adaptively generated network to be trained through the second adaptive module. When the training batch size of the network to be trained in the tooth partition prediction module is generated in a self-adaptive manner, the training batch size of the network to be trained in the tooth partition prediction module can be generated in a self-adaptive manner through the second self-adaptive module according to the size of the oral medical image sample in the second training set and the size of the GPU memory. When the iteration number of the network to be trained in the tooth partition prediction module is generated in an adaptive manner, the iteration number of the network to be trained which is fixedly set can be determined as the iteration number of the network to be trained which is generated in an adaptive manner through the second adaptive module. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, the method further comprises: adaptively selecting a preprocessing mode for the second training set according to the second data fingerprint of the second training set through the second adaptive module; according to the preprocessing mode, preprocessing the second training set to obtain the preprocessed second training set, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint, including: and training the tooth partition prediction module according to the preprocessed second training set and the second pipeline fingerprint. Therefore, the preprocessing mode is selected for the second training set in a self-adaptive mode according to the second data fingerprint of the second training set through the second self-adaptive module without manually relying on prior knowledge, and the preprocessing mode is preprocessed for the second training set according to the preprocessing mode, so that the efficiency of preprocessing the second training set is effectively improved, and the accuracy of preprocessing the second training set is also effectively improved. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the preprocessing includes clipping, resampling, and normalization. When the clipping mode is adaptively selected for the second training set, the clipping mode of the oral medical image samples in the second training set that are fixedly set can be determined as the adaptively selected clipping mode by the second adaptive module. When the resampling method is adaptively selected for the second training set, the resampling method may be adaptively selected for the second training set by the second adaptive module according to the voxel pitch of the oral medical image samples in the second training set. When the normalization mode is adaptively selected for the second training set, the normalization mode may be adaptively selected for the second training set by the second adaptive module according to the voxel size of the oral medical image sample in the second training set. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the preprocessing manner includes clipping, resampling and normalization, and the preprocessing the second training set according to the preprocessing manner to obtain the preprocessed second training set includes: clipping the oral medical image samples in the second training set to obtain the clipped oral medical image samples; resampling the cut oral medical image samples according to the voxel spacing of the cut oral medical image samples to obtain the resampled oral medical image samples; normalizing the resampled oral medical image sample according to the voxel size of the resampled oral medical image sample to obtain the normalized oral medical image sample, and obtaining the preprocessed second training set according to the normalized oral medical image sample. And the cutting operation is used for cutting zero-value areas of all oral medical image samples in the second training set to reserve non-zero-value areas, reducing the size of the image and reducing the calculated amount of the model. Different oral medical image samples may have different voxel spacings due to different oral scanning devices or device acquisition protocols, and in order to ensure that the model can learn correct spatial semantics, the voxel spacings of the oral medical image samples are resampled to the median voxel spacings. The oral medical image sample after resampling is normalized according to the voxel size of the oral medical image sample after resampling, so that adverse effects caused by singular data can be avoided, and model convergence can be accelerated. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, the method further comprises: adaptively selecting a data enhancement mode for the second training set through the second adaptive module; according to the data enhancement mode, performing data enhancement on the second training set to obtain the second training set after data enhancement, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint, including: and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint after data enhancement. Therefore, the data enhancement mode is selected for the second training set in a self-adaptive mode through the second self-adaptive module without manually relying on prior knowledge, and data enhancement is performed on the second training set according to the data enhancement mode, so that the efficiency of performing data enhancement on the second training set is effectively improved, and the accuracy of performing data enhancement on the second training set is also effectively improved. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the data enhancement mode includes rotation scaling, gaussian plus noise, gaussian blur, brightness processing, or contrast processing. When the rotation scaling manner is adaptively selected for the second training set, the fixedly set rotation scaling manner of the second training set may be determined as the adaptively generated rotation scaling manner of the second training set by the second adaptive module. When the gaussian noise adding mode is adaptively selected for the second training set, the fixedly set gaussian noise adding mode of the second training set may be determined as the adaptively generated gaussian noise adding mode of the second training set by the second adaptive module. When the gaussian fuzzy manner is adaptively selected for the second training set, the fixedly set gaussian fuzzy manner of the second training set may be determined as the adaptively generated gaussian fuzzy manner of the second training set by the second adaptive module. When adaptively selecting the brightness processing mode for the second training set, the brightness processing mode of the second training set that is fixedly set may be determined as the brightness processing mode of the adaptively generated second training set by the second adaptive module. When the contrast processing mode is adaptively selected for the second training set, the second adaptive module may determine the contrast processing mode of the second training set that is fixedly set as the contrast processing mode of the adaptively generated second training set. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the data enhancement mode includes rotation scaling, gaussian plus noise, gaussian blur, brightness processing, or contrast processing. When the second training set is subjected to data enhancement according to the data enhancement mode, rotationally scaling the oral medical image samples in the second training set to obtain the rotationally scaled oral medical image samples, and obtaining the data-enhanced second training set according to the rotationally scaled oral medical image samples; or, performing gaussian noise on the oral medical image samples in the second training set to obtain gaussian noise-added oral medical image samples, and obtaining data-enhanced second training set according to the gaussian noise-added oral medical image samples; or, performing gaussian blurring on the oral medical image samples in the second training set to obtain the oral medical image samples after gaussian blurring, and obtaining the second training set after data enhancement according to the oral medical image samples after gaussian blurring; or, performing brightness processing on the oral medical image samples in the second training set to obtain the oral medical image samples after the brightness processing, and obtaining the second training set after data enhancement according to the oral medical image samples after the brightness processing; or, performing contrast processing on the oral medical image samples in the second training set to obtain the oral medical image samples after the contrast processing, and obtaining the second training set after data enhancement according to the oral medical image samples after the contrast processing. The gaussian noise adding can be understood as adding noise into the oral medical image sample, the probability density function of the noise follows gaussian distribution, the gaussian blur can be understood as actually performing filtering processing on the oral medical image sample, the filtering mode is gaussian filtering, the filtering mode is image blur, important information of the oral medical image sample is extracted, the brightness processing can be understood as changing the brightness of the oral medical image sample, and the contrast processing can be understood as changing the contrast of the oral medical image sample. The more and more the oral medical image samples in the second training set, the more beneficial the model training and the improvement of the generalization capability of the model. In order to train the model as good as possible on the limited second training set, the oral medical image samples in the second training set are expanded in a data enhancement mode, the quality of the training samples is improved, noise data is increased, and the robustness of the model is improved. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, the second pipeline fingerprint comprises identification information of a network to be trained in the tooth segment prediction module and a network hyper-parameter of the network to be trained. When the tooth partition prediction module is trained according to the second training set and the second pipeline fingerprint, determining the network to be trained according to the identification information, and setting the network to be trained according to the network hyper-parameter so as to obtain the set network to be trained; predicting the tooth partition in the oral medical image sample in the second training set through the set network to be trained so as to obtain the tooth partition prediction result of the oral medical image sample; and training the set network to be trained according to the tooth partition prediction result of the oral medical image sample and the tooth partition marking result of the oral medical image sample. Therefore, the trained network can accurately perform tooth partition on the target oral medical image. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the tooth-zone prediction result of the oral medical image sample may be understood as prediction probability data of a tooth zone label to which a voxel in the oral medical image sample belongs, and the tooth-zone labeling result of the oral medical image sample may be understood as labeling probability data of a tooth zone label to which a voxel in the oral medical image sample belongs. When the set network to be trained is trained according to the tooth partition prediction result of the oral medical image sample and the tooth partition marking result of the oral medical image sample, determining the difference value of the prediction probability data and the marking probability data through a target loss function; and adjusting the set parameters of the network to be trained based on the difference value. Wherein the objective loss function may be a sum of a Dice loss and a cross entropy loss. When the parameters of the set network to be trained are adjusted, a back propagation algorithm or a random gradient descent algorithm can be adopted to adjust the parameters of the set network to be trained. It should be understood that the above description is only exemplary, and the embodiments of the present application are not limited in this respect.

In step S104, the trained tooth classification prediction module predicts the classification of the teeth in the target oral medical image, and the trained tooth partition prediction module predicts the partitions of the teeth in the target oral medical image.

In some optional embodiments, the method further comprises: adaptively selecting an inference mode for the trained tooth classification prediction module through the first adaptive module, and predicting the classification of teeth in the target oral medical image through the trained tooth classification prediction module, including: and predicting the classification of the teeth in the target oral medical image according to the reasoning mode through the trained tooth classification prediction module so as to obtain a tooth classification prediction result of the target oral medical image. Therefore, an inference mode does not need to be selected for the trained tooth classification prediction module by artificially relying on prior knowledge, and the classification of the teeth in the target oral medical image is predicted by the trained tooth classification prediction module according to the inference mode, so that the inference efficiency of the trained tooth classification prediction module is effectively improved, and the inference accuracy of the trained tooth classification prediction module is also effectively improved. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, when the tooth classification prediction module is U-Net, the U-Net comprises three networks, namely 2D-Unet, 3D-Unet and 3D cascaded Unet. The best/certain network of the three networks can be selected for reasoning, and the average value of the results of the 3 networks can also be selected for reasoning; the best network is chosen by default for reasoning. When the first self-adaptive module is used for self-adaptively selecting an inference mode for the trained tooth classification prediction module, the fixedly set inference mode is determined as the inference mode of the trained tooth classification prediction module through the first self-adaptive module. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, the method further comprises: adaptively selecting an inference mode for the trained tooth partition prediction module through the second adaptive module, and predicting the partition of the tooth in the target oral medical image through the trained tooth partition prediction module, including: and predicting the dental subareas in the target oral medical image according to the reasoning mode by the trained dental subarea prediction module to obtain a dental subarea prediction result of the target oral medical image. Therefore, an inference mode does not need to be selected for the trained tooth partition prediction module by artificially relying on prior knowledge, and the tooth partition prediction module predicts the partition of the tooth in the target oral medical image according to the inference mode, so that the inference efficiency of the trained tooth partition prediction module is effectively improved, and the inference accuracy of the trained tooth partition prediction module is also effectively improved. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, when the tooth zone prediction module is U-Net, the U-Net comprises three networks, namely 2D-Unet, 3D-Unet and 3D cascaded Unet. The best/certain network of the three networks can be selected for reasoning, and the average value of the results of the 3 networks can also be selected for reasoning; the best network is chosen by default for reasoning. And when the second self-adaptive module is used for self-adaptively selecting an inference mode for the trained tooth partition prediction module, determining a fixedly set inference mode as the inference mode of the trained tooth partition prediction module through the second self-adaptive module. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, when the trained tooth classification prediction module is used for predicting the classification of the tooth in the target oral medical image, the trained tooth classification prediction module is used for performing down-sampling on the target oral medical image through a plurality of cascaded down-sampling units in the trained tooth classification prediction module to obtain a plurality of first feature maps with different scales of the target oral medical image; and performing upsampling on the last first feature map obtained by downsampling the target oral medical image according to a plurality of first feature maps with different scales of the target oral medical image through a plurality of cascaded upsampling units in the trained tooth classification prediction module to obtain a tooth classification prediction result of the target oral medical image. The down-sampling unit comprises a cascaded convolution layer and a cascaded pooling layer, and the up-sampling unit comprises a cascaded up-sampling layer and a cascaded deconvolution layer. Therefore, the tooth classification prediction result of the target oral medical image can be accurately obtained. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In some optional embodiments, when the dental subarea prediction module after training predicts the subareas of the teeth in the target oral medical image, the target oral medical image is downsampled by a plurality of cascaded downsampling units in the dental subarea prediction module after training to obtain a plurality of second feature maps with different scales of the target oral medical image; and performing upsampling on the last second feature map obtained by downsampling the target oral medical image according to a plurality of second feature maps with different scales of the target oral medical image through a plurality of cascaded upsampling units in the trained tooth partition prediction module to obtain a tooth partition prediction result of the target oral medical image. The down-sampling unit comprises a cascaded convolution layer and a cascaded pooling layer, and the up-sampling unit comprises a cascaded up-sampling layer and a cascaded deconvolution layer. Therefore, the tooth partition prediction result of the target oral medical image can be accurately obtained. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In step S105, a tooth segmentation result of the target oral medical image is determined according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image.

In some optional embodiments, the method further comprises: adaptively selecting a first post-processing mode for the tooth classification prediction result through the first adaptive module; adaptively selecting a second post-processing mode for the tooth partition prediction result through the second adaptive module, and determining the tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image, including: carrying out post-processing on the tooth classification prediction result of the target oral medical image through the first post-processing mode to obtain a post-processed tooth classification prediction result; performing post-processing on the tooth partition prediction result of the target oral medical image through the second post-processing mode to obtain a post-processed tooth partition prediction result; and fusing the post-processed tooth classification prediction result and the post-processed tooth partition prediction result to obtain a tooth segmentation result of the target oral medical image. Thereby, the tooth segmentation result of the target oral medical image can be accurately obtained. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the first post-processing manner includes connected component analysis, the tooth classification prediction result includes probability data of a tooth classification label to which a voxel in the target oral medical image belongs, and the tooth classification prediction result of post-processing includes a tooth classification label to which a tooth region of the target oral medical image belongs. And when the tooth classification prediction result of the target oral medical image is subjected to post-processing through the first post-processing mode, performing connected domain analysis on probability data of the tooth classification label to which the voxel in the target oral medical image belongs to obtain the tooth classification label to which the tooth area of the target oral medical image belongs. Thereby, the tooth classification label to which the tooth region of the target oral medical image belongs can be accurately obtained. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the second post-processing manner includes connected component analysis, the tooth-partition prediction result includes probability data of a tooth-partition label to which a voxel in the target oral medical image belongs, and the tooth-partition prediction result of post-processing includes a tooth-partition label to which a tooth region of the target oral medical image belongs. And when the tooth partition prediction result of the target oral medical image is subjected to post-processing through the second post-processing mode, performing connected domain analysis on probability data of the tooth partition label to which the voxel in the target oral medical image belongs to obtain the tooth partition label to which the tooth area of the target oral medical image belongs. Therefore, the tooth partition label of the tooth area of the target oral medical image can be accurately obtained. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, the post-processed tooth classification prediction result includes a tooth classification label to which a tooth region of the target oral medical image belongs, and the post-processed tooth partition prediction result includes a tooth partition label to which a tooth region of the target oral medical image belongs. When the post-processed tooth classification prediction result and the post-processed tooth partition prediction result are fused, fusing a tooth classification label to which a tooth region of the target oral medical image belongs and a tooth partition label to which the tooth region of the target oral medical image belongs to obtain a tooth segmentation label to which the tooth region of the target oral medical image belongs; and segmenting the target oral medical image according to the tooth segmentation label of the tooth area of the target oral medical image so as to obtain the tooth segmentation result of the target oral medical image. Therefore, the target oral medical image is segmented through the tooth segmentation label of the tooth area of the target oral medical image, and the tooth segmentation result of the target oral medical image can be obtained more accurately. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific example, a tooth classification label belonging to a tooth region of the target oral medical image and a tooth partition label belonging to the tooth region of the target oral medical image are fused, the tooth classification label is 7 classifications, the labels are 1-7, the tooth partition label is 4 classifications, the labels are 1-4, and a tooth segmentation label obtained after fusion is 28 classifications, the labels are 11-17, 21-27, 31-37, and 41-47. Specifically, the labels of the upper left 7 teeth are 11-17, the labels of the lower left 7 teeth are 21-27, the labels of the upper right 7 teeth are 31-37, and the labels of the lower right 7 teeth are 41-47. The tooth segmentation is divided into two tasks of 7 classes of tooth classification and 4 classes of tooth partition, so that data are more fully utilized than direct 28 classification, the 7 classes of tooth classification tasks can extract 4 groups of teeth from one image, the 28 classes of teeth can only extract one group of teeth from one image, the data volume of the 7 classes of tooth segmentation is far more than 28 classes, the deep model learning and training are facilitated, and the classification and partition of all teeth is more accurate than the direct segmentation. Tooth segmentation is carried out through the mode of fusing tooth classification label and tooth subregion label, can make the more abundant limited data of utilization of degree of depth model, and then promotes the degree of accuracy of model. And converting the fused data format from the ni.gz format into a data format dicom format which can be read by software, and displaying the tooth segmentation result in the software. It should be understood that the above description is only exemplary, and the present embodiment is not limited thereto.

In a specific practical application, the tooth segmentation method for the oral medical image provided by the embodiment can automatically select a proper preprocessing mode and a proper post-processing mode for the oral medical image by using the nnU-NET depth model, and can select a proper network to train an oral medical image sample, so that the condition that the tooth segmentation algorithm depends on priori knowledge or needs manual intervention like a traditional tooth segmentation algorithm or other depth learning algorithms is avoided, the tooth segmentation efficiency and accuracy are improved, the oral surgery robot can automatically segment and display the teeth in the medical image, and the robot is helped to accurately position the teeth in subsequent tasks and implement high-precision dental implantation operation.

According to the tooth segmentation method of the oral medical image, pipeline fingerprints for a tooth classification prediction module to be trained and pipeline fingerprints for a tooth partition prediction module to be trained do not need to be manually prepared in advance, the first pipeline fingerprints for training the tooth classification prediction module are generated in a self-adaptive mode through a first self-adaptive module in the tooth classification prediction model according to first data fingerprints of a first training set, and second pipeline fingerprints for training the tooth partition prediction module are generated in a self-adaptive mode through a second self-adaptive module in the tooth partition prediction model according to second data fingerprints of a second training set, so that the trouble of manually preparing the pipeline fingerprints is avoided, the interference of thought factors is reduced, and the tooth segmentation efficiency in the oral medical image is effectively improved. In addition, the diversity of the segmentation results of the teeth in the oral medical images caused by manual preparation of pipeline fingerprints is effectively avoided, and the segmentation accuracy of the teeth in the oral medical images is effectively improved.

The dental segmentation method of the oral medical image provided by the present embodiment may be performed by any suitable device with data processing capability, including but not limited to: a camera, a terminal, a mobile terminal, a PC, a server, an in-vehicle device, an entertainment device, an advertising device, a Personal Digital Assistant (PDA), a tablet computer, a notebook computer, a handheld game console, smart glasses, a smart watch, a wearable device, a virtual display device, a display enhancement device, or the like.

Referring to fig. 2, a structural diagram of a tooth segmentation apparatus for dental medical images according to a second embodiment of the present application is shown.

The present embodiment provides a tooth segmentation apparatus for dental medical images, including: a first determining module 201, configured to determine, according to an oral medical image sample obtained by scanning an oral cavity by an oral cavity scanning device, a first training set used for training a tooth classification prediction module in a tooth classification prediction model and a second training set used for training a tooth partition prediction module in a tooth partition prediction model; a first training module 202, configured to generate, by a first adaptive module in the tooth classification prediction model, a first pipeline fingerprint for training the tooth classification prediction module according to the first data fingerprint of the first training set, and train the tooth classification prediction module according to the first training set and the first pipeline fingerprint; a second training module 203, configured to generate a second pipeline fingerprint for training the tooth partition prediction module adaptively according to a second data fingerprint of the second training set through a second adaptive module in the tooth partition prediction model, and train the tooth partition prediction module according to the second training set and the second pipeline fingerprint; the prediction module 204 is configured to predict the classification of teeth in the target oral medical image through the trained tooth classification prediction module, and predict the partition of teeth in the target oral medical image through the trained tooth partition prediction module; the second determining module 205 is configured to determine a tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image.

Optionally, the apparatus further comprises: the first selection module is used for adaptively selecting a preprocessing mode for the first training set according to the first data fingerprint of the first training set through the first self-adaptive module; a first preprocessing module, configured to preprocess the first training set according to the preprocessing manner to obtain the preprocessed first training set, where the first training module 202 is specifically configured to: and training the tooth classification prediction module according to the preprocessed first training set and the preprocessed first pipeline fingerprint.

Optionally, the preprocessing mode includes clipping, resampling and normalization, and the first preprocessing module is specifically configured to: cutting the oral medical image samples in the first training set to obtain the cut oral medical image samples; resampling the cut oral medical image samples according to the voxel spacing of the cut oral medical image samples to obtain the resampled oral medical image samples; normalizing the resampled oral medical image sample according to the voxel size of the resampled oral medical image sample to obtain the normalized oral medical image sample, and obtaining the preprocessed first training set according to the normalized oral medical image sample.

Optionally, the apparatus further comprises: the second selection module is used for adaptively selecting a preprocessing mode for the second training set according to the second data fingerprint of the second training set through the second self-adaptive module; a second preprocessing module, configured to preprocess the second training set according to the preprocessing manner to obtain the preprocessed second training set, where the second training module 203 is specifically configured to: and training the tooth partition prediction module according to the preprocessed second training set and the second pipeline fingerprint.

Optionally, the preprocessing mode includes clipping, resampling and normalization, and the second preprocessing module is specifically configured to: clipping the oral medical image samples in the second training set to obtain the clipped oral medical image samples; resampling the cut oral medical image samples according to the voxel spacing of the cut oral medical image samples to obtain the resampled oral medical image samples; normalizing the resampled oral medical image sample according to the voxel size of the resampled oral medical image sample to obtain the normalized oral medical image sample, and obtaining the preprocessed second training set according to the normalized oral medical image sample.

Optionally, the apparatus further comprises: a third selection module, configured to adaptively select a data enhancement mode for the first training set through the first adaptive module; a first data enhancement module, configured to perform data enhancement on the first training set according to the data enhancement mode to obtain the first training set after data enhancement, where the first training module 202 is specifically configured to: and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint after data enhancement.

Optionally, the data enhancement mode includes rotation scaling, gaussian noise adding, gaussian blurring, brightness processing, or contrast processing, and the first data enhancement module is specifically configured to: rotationally scaling the oral medical image samples in the first training set to obtain rotationally scaled oral medical image samples, and obtaining the first training set with enhanced data according to the rotationally scaled oral medical image samples; or, performing gaussian noise on the oral medical image samples in the first training set to obtain gaussian noise-added oral medical image samples, and obtaining data-enhanced first training set according to the gaussian noise-added oral medical image samples; or, performing gaussian blurring on the oral medical image samples in the first training set to obtain the oral medical image samples after gaussian blurring, and obtaining the first training set after data enhancement according to the oral medical image samples after gaussian blurring; or, performing brightness processing on the oral medical image samples in the first training set to obtain brightness-processed oral medical image samples, and obtaining the first training set with enhanced data according to the brightness-processed oral medical image samples; or, performing contrast processing on the oral medical image samples in the first training set to obtain the oral medical image samples after the contrast processing, and obtaining the first training set after data enhancement according to the oral medical image samples after the contrast processing.

Optionally, the apparatus further comprises: a fourth selection module, configured to adaptively select a data enhancement mode for the second training set through the second adaptive module; a second data enhancement module, configured to perform data enhancement on the second training set according to the data enhancement manner to obtain the second training set after data enhancement, where the second training module 203 is specifically configured to: and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint after data enhancement.

Optionally, the data enhancement mode includes rotation scaling, gaussian noise adding, gaussian blurring, brightness processing, or contrast processing, and the second data enhancement module is specifically configured to: rotationally scaling the oral medical image samples in the second training set to obtain rotationally scaled oral medical image samples, and obtaining the second training set with enhanced data according to the rotationally scaled oral medical image samples; or, performing gaussian noise on the oral medical image samples in the second training set to obtain gaussian noise-added oral medical image samples, and obtaining data-enhanced second training set according to the gaussian noise-added oral medical image samples; or, performing gaussian blurring on the oral medical image samples in the second training set to obtain the oral medical image samples after gaussian blurring, and obtaining the second training set after data enhancement according to the oral medical image samples after gaussian blurring; or, performing brightness processing on the oral medical image samples in the second training set to obtain the oral medical image samples after the brightness processing, and obtaining the second training set after data enhancement according to the oral medical image samples after the brightness processing; or, performing contrast processing on the oral medical image samples in the second training set to obtain the oral medical image samples after the contrast processing, and obtaining the second training set after data enhancement according to the oral medical image samples after the contrast processing.

Optionally, the first pipeline fingerprint includes identification information of a network to be trained in the tooth classification prediction module and a network hyper-parameter of the network to be trained, and the first training module 202 is specifically configured to: determining the network to be trained according to the identification information, and setting the network to be trained according to the network hyper-parameter so as to obtain the set network to be trained; predicting the classification of teeth in the oral medical image samples in the first training set through the set network to be trained so as to obtain tooth classification prediction results of the oral medical image samples; and training the set network to be trained according to the tooth classification prediction result of the oral medical image sample and the tooth classification labeling result of the oral medical image sample.

Optionally, the second pipeline fingerprint includes identification information of a network to be trained in the tooth partition prediction module and a network hyper-parameter of the network to be trained, and the second training module 203 is specifically configured to: determining the network to be trained according to the identification information, and setting the network to be trained according to the network hyper-parameter so as to obtain the set network to be trained; predicting the tooth partition in the oral medical image sample in the second training set through the set network to be trained so as to obtain the tooth partition prediction result of the oral medical image sample; and training the set network to be trained according to the tooth partition prediction result of the oral medical image sample and the tooth partition marking result of the oral medical image sample.

Optionally, the apparatus further comprises: a fifth selecting module, configured to adaptively select, through the first adaptive module, an inference mode for the trained tooth classification prediction module, where the prediction module 204 is specifically configured to: and predicting the classification of the teeth in the target oral medical image according to the reasoning mode through the trained tooth classification prediction module so as to obtain a tooth classification prediction result of the target oral medical image.

Optionally, the apparatus further comprises: a sixth selecting module, configured to adaptively select, through the second adaptive module, an inference mode for the trained tooth partition prediction module, where the prediction module 204 is specifically configured to: and predicting the dental subareas in the target oral medical image according to the reasoning mode by the trained dental subarea prediction module to obtain a dental subarea prediction result of the target oral medical image.

Optionally, the prediction module 204 is specifically configured to: downsampling the target oral medical image through a plurality of cascaded downsampling units in the trained tooth classification prediction module to obtain a plurality of first feature maps with different scales of the target oral medical image; and performing upsampling on the last first feature map obtained by downsampling the target oral medical image according to a plurality of first feature maps with different scales of the target oral medical image through a plurality of cascaded upsampling units in the trained tooth classification prediction module to obtain a tooth classification prediction result of the target oral medical image.

Optionally, the prediction module 204 is specifically configured to: downsampling the target oral medical image through a plurality of cascaded downsampling units in the trained tooth partition prediction module to obtain a plurality of second feature maps with different scales of the target oral medical image; and performing upsampling on the last second feature map obtained by downsampling the target oral medical image according to a plurality of second feature maps with different scales of the target oral medical image through a plurality of cascaded upsampling units in the trained tooth partition prediction module to obtain a tooth partition prediction result of the target oral medical image.

Optionally, the apparatus further comprises: a seventh selecting module, configured to adaptively select, through the first adaptive module, a first post-processing manner for the tooth classification prediction result; an eighth selecting module, configured to adaptively select a second post-processing manner for the tooth partition prediction result through the second adaptive module, where the second determining module 205 includes: the first post-processing submodule is used for performing post-processing on the tooth classification prediction result of the target oral medical image through the first post-processing mode to obtain a post-processed tooth classification prediction result; the second post-processing submodule is used for performing post-processing on the tooth partition prediction result of the target oral medical image in the second post-processing mode to obtain the post-processed tooth partition prediction result; and the fusion submodule is used for fusing the post-processed tooth classification prediction result and the post-processed tooth partition prediction result so as to obtain a tooth segmentation result of the target oral medical image.

Optionally, the first post-processing manner includes connected component analysis, the tooth classification prediction result includes probability data of a tooth classification label to which a voxel in the target oral medical image belongs, and the post-processed tooth classification prediction result includes a tooth classification label to which a tooth region of the target oral medical image belongs, and the first post-processing sub-module is specifically configured to: and performing connected domain analysis on the probability data of the tooth classification label to which the voxel in the target oral medical image belongs to obtain the tooth classification label to which the tooth area of the target oral medical image belongs.

Optionally, the second post-processing manner includes connected component analysis, the tooth partition prediction result includes probability data of a tooth partition label to which a voxel in the target oral medical image belongs, and the tooth partition prediction result of post-processing includes a tooth partition label to which a tooth area of the target oral medical image belongs, and the second post-processing sub-module is specifically configured to: and performing connected domain analysis on the probability data of the tooth zone label to which the voxel in the target oral medical image belongs to obtain the tooth zone label to which the tooth area of the target oral medical image belongs.

Optionally, the post-processed tooth classification prediction result includes a tooth classification label to which a tooth region of the target oral medical image belongs, and the post-processed tooth partition prediction result includes a tooth partition label to which a tooth region of the target oral medical image belongs, and the fusion sub-module is specifically configured to: fusing a tooth classification label to which the tooth region of the target oral medical image belongs and a tooth partition label to which the tooth region of the target oral medical image belongs to obtain a tooth segmentation label to which the tooth region of the target oral medical image belongs; and segmenting the target oral medical image according to the tooth segmentation label of the tooth area of the target oral medical image so as to obtain the tooth segmentation result of the target oral medical image.

The tooth segmentation apparatus for an oral medical image provided in this embodiment is used to implement the tooth segmentation method for an oral medical image in the foregoing method embodiments, and has the beneficial effects of the corresponding method embodiments, which are not described herein again.

Fig. 3 is a schematic structural diagram of an electronic device in a third embodiment of the present application; the electronic device may include:

one or more processors 301;

a computer-readable medium 302, which may be configured to store one or more programs,

when executed by the one or more processors, cause the one or more processors to implement the method for dental segmentation of an oral medical image as described in the above embodiment.

Fig. 4 is a hardware structure of an electronic device according to a fourth embodiment of the present application; as shown in fig. 4, the hardware structure of the electronic device may include: a processor 401, a communication interface 402, a computer-readable medium 403, and a communication bus 404;

wherein the processor 401, the communication interface 402, and the computer-readable medium 403 are in communication with each other via a communication bus 404;

alternatively, the communication interface 402 may be an interface of a communication module, such as an interface of a GSM module;

the processor 401 may be specifically configured to: according to an oral medical image sample obtained by scanning an oral cavity by an oral cavity scanning device, determining a first training set for training a tooth classification prediction module in a tooth classification prediction model and a second training set for training a tooth partition prediction module in a tooth partition prediction model; adaptively generating a first pipeline fingerprint used for training the tooth classification prediction module according to a first data fingerprint of the first training set through a first adaptive module in the tooth classification prediction model, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint; adaptively generating a second pipeline fingerprint used for training the tooth partition prediction module according to a second data fingerprint of the second training set through a second adaptive module in the tooth partition prediction model, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint; predicting the classification of teeth in the target oral medical image through the trained tooth classification prediction module, and predicting the subareas of the teeth in the target oral medical image through the trained tooth subarea prediction module; and determining the tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image.

Processor 401 may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components. The various methods, steps, and logic blocks disclosed in the embodiments of the present application may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The computer-readable medium 403 may be, but is not limited to, a Random Access Memory (RAM), a Read-Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like.

In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code configured to perform the method illustrated by the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication section, and/or installed from a removable medium. The computer program, when executed by a Central Processing Unit (CPU), performs the above-described functions defined in the method of the present application. It should be noted that the computer readable medium described herein can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access storage media (RAM), a read-only storage media (ROM), an erasable programmable read-only storage media (EPROM or flash memory), an optical fiber, a portable compact disc read-only storage media (CD-ROM), an optical storage media piece, a magnetic storage media piece, or any suitable combination of the foregoing. In the present application, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In this application, however, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.

Computer program code configured to carry out operations for the present application may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may operate over any of a variety of networks: including a Local Area Network (LAN) or a Wide Area Network (WAN) -to the user's computer, or alternatively, to an external computer (e.g., through the internet using an internet service provider).

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions configured to implement the specified logical function(s). In the above embodiments, specific precedence relationships are provided, but these precedence relationships are only exemplary, and in particular implementations, the steps may be fewer, more, or the execution order may be modified. That is, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The modules described in the embodiments of the present application may be implemented by software or hardware. The described modules may also be provided in a processor, which may be described as: a processor includes a first determination module, a first training module, a second training module, a prediction module, and a second determination module. The names of these modules do not constitute a limitation to the module itself in some cases, for example, the first determination module may also be described as a module that determines a first training set for training the tooth classification prediction module in the tooth classification prediction model and a second training set for training the tooth partition prediction module in the tooth partition prediction model according to the oral medical image samples obtained by scanning the oral cavity with the oral cavity scanning device.

As another aspect, the present application further provides a computer readable medium, on which a computer program is stored, which when executed by a processor, implements the tooth segmentation method of the medical image of the oral cavity as described in the first embodiment.

As another aspect, the present application also provides a computer-readable medium, which may be contained in the apparatus described in the above embodiments; or may be present separately and not assembled into the device. The computer readable medium carries one or more programs which, when executed by the apparatus, cause the apparatus to: according to an oral medical image sample obtained by scanning an oral cavity by an oral cavity scanning device, determining a first training set for training a tooth classification prediction module in a tooth classification prediction model and a second training set for training a tooth partition prediction module in a tooth partition prediction model; adaptively generating a first pipeline fingerprint used for training the tooth classification prediction module according to a first data fingerprint of the first training set through a first adaptive module in the tooth classification prediction model, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint; adaptively generating a second pipeline fingerprint used for training the tooth partition prediction module according to a second data fingerprint of the second training set through a second adaptive module in the tooth partition prediction model, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint; predicting the classification of teeth in the target oral medical image through the trained tooth classification prediction module, and predicting the subareas of the teeth in the target oral medical image through the trained tooth subarea prediction module; and determining the tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image.

The expressions "first", "second", "said first" or "said second" used in various embodiments of the present disclosure may modify various components regardless of order and/or importance, but these expressions do not limit the respective components. The above description is only configured for the purpose of distinguishing elements from other elements. For example, the first user equipment and the second user equipment represent different user equipment, although both are user equipment. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of the present disclosure.

When an element (e.g., a first element) is referred to as being "operably or communicatively coupled" or "connected" (operably or communicatively) to "another element (e.g., a second element) or" connected "to another element (e.g., a second element), it is understood that the element is directly connected to the other element or the element is indirectly connected to the other element via yet another element (e.g., a third element). In contrast, it is understood that when an element (e.g., a first element) is referred to as being "directly connected" or "directly coupled" to another element (a second element), no element (e.g., a third element) is interposed therebetween.

The above description is only a preferred embodiment of the application and is illustrative of the principles of the technology employed. It will be appreciated by those skilled in the art that the scope of the invention herein disclosed is not limited to the particular combination of features described above, but also encompasses other arrangements formed by any combination of the above features or their equivalents without departing from the spirit of the invention. For example, the above features may be replaced with (but not limited to) features having similar functions disclosed in the present application.

Claims

1. A method for dental segmentation of an oral medical image, the method comprising:

according to an oral medical image sample obtained by scanning an oral cavity by an oral cavity scanning device, determining a first training set for training a tooth classification prediction module in a tooth classification prediction model and a second training set for training a tooth partition prediction module in a tooth partition prediction model;

adaptively generating a first pipeline fingerprint used for training the tooth classification prediction module according to a first data fingerprint of the first training set through a first adaptive module in the tooth classification prediction model, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint;

adaptively generating a second pipeline fingerprint used for training the tooth partition prediction module according to a second data fingerprint of the second training set through a second adaptive module in the tooth partition prediction model, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint;

predicting the classification of teeth in the target oral medical image through the trained tooth classification prediction module, and predicting the subareas of the teeth in the target oral medical image through the trained tooth subarea prediction module;

and determining the tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image.

2. The method for dental segmentation of an oral medical image according to claim 1, further comprising:

adaptively selecting a preprocessing mode for the first training set according to the first data fingerprint of the first training set through the first adaptive module;

preprocessing the first training set according to the preprocessing mode to obtain a preprocessed first training set,

training the tooth classification prediction module according to the first training set and the first pipeline fingerprint, including:

and training the tooth classification prediction module according to the preprocessed first training set and the preprocessed first pipeline fingerprint.

3. The dental segmentation method of claim 2, wherein the preprocessing includes clipping, resampling and normalization,

the preprocessing the first training set according to the preprocessing mode to obtain the preprocessed first training set includes:

cutting the oral medical image samples in the first training set to obtain the cut oral medical image samples;

resampling the cut oral medical image samples according to the voxel spacing of the cut oral medical image samples to obtain the resampled oral medical image samples;

normalizing the resampled oral medical image sample according to the voxel size of the resampled oral medical image sample to obtain the normalized oral medical image sample, and obtaining the preprocessed first training set according to the normalized oral medical image sample.

4. The method for dental segmentation of an oral medical image according to claim 1, further comprising:

adaptively selecting a data enhancement mode for the first training set through the first adaptive module;

performing data enhancement on the first training set according to the data enhancement mode to obtain the first training set after data enhancement,

and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint after data enhancement.

5. The dental segmentation method of claim 4, wherein the data enhancement mode comprises a rotation scaling, a Gaussian noise adding mode, a Gaussian blur mode, a brightness processing mode, or a contrast processing mode,

the data enhancement of the first training set according to the data enhancement mode to obtain the data-enhanced first training set includes:

rotationally scaling the oral medical image samples in the first training set to obtain rotationally scaled oral medical image samples, and obtaining the first training set with enhanced data according to the rotationally scaled oral medical image samples; alternatively, the first and second electrodes may be,

performing Gaussian noise on the oral medical image samples in the first training set to obtain the oral medical image samples subjected to Gaussian noise, and obtaining the first training set subjected to data enhancement according to the oral medical image samples subjected to Gaussian noise; alternatively, the first and second electrodes may be,

performing Gaussian blur on the oral medical image samples in the first training set to obtain the oral medical image samples after Gaussian blur, and obtaining the first training set after data enhancement according to the oral medical image samples after Gaussian blur; alternatively, the first and second electrodes may be,

performing brightness processing on the oral medical image samples in the first training set to obtain the oral medical image samples after the brightness processing, and obtaining the first training set after data enhancement according to the oral medical image samples after the brightness processing; alternatively, the first and second electrodes may be,

contrast processing is carried out on the oral medical image samples in the first training set to obtain the oral medical image samples after the contrast processing, and the first training set after data enhancement is obtained according to the oral medical image samples after the contrast processing.

6. The dental segmentation method of the oral medical image according to claim 1, wherein the first pipeline fingerprint includes identification information of a network to be trained in the dental classification prediction module and a network hyper-parameter of the network to be trained,

determining the network to be trained according to the identification information, and setting the network to be trained according to the network hyper-parameter so as to obtain the set network to be trained;

predicting the classification of teeth in the oral medical image samples in the first training set through the set network to be trained so as to obtain tooth classification prediction results of the oral medical image samples;

and training the set network to be trained according to the tooth classification prediction result of the oral medical image sample and the tooth classification labeling result of the oral medical image sample.

7. The method for dental segmentation of an oral medical image according to claim 1, further comprising:

adaptively selecting an inference mode for the trained tooth classification prediction module through the first adaptive module,

the predicting the classification of the teeth in the target oral medical image by the trained tooth classification predicting module comprises:

and predicting the classification of the teeth in the target oral medical image according to the reasoning mode through the trained tooth classification prediction module so as to obtain a tooth classification prediction result of the target oral medical image.

8. The method for dental segmentation of an oral medical image according to claim 1, wherein the predicting the classification of the tooth in the target oral medical image by the trained tooth classification prediction module comprises:

downsampling the target oral medical image through a plurality of cascaded downsampling units in the trained tooth classification prediction module to obtain a plurality of first feature maps with different scales of the target oral medical image;

and performing upsampling on the last first feature map obtained by downsampling the target oral medical image according to a plurality of first feature maps with different scales of the target oral medical image through a plurality of cascaded upsampling units in the trained tooth classification prediction module to obtain a tooth classification prediction result of the target oral medical image.

9. The method for dental segmentation of an oral medical image according to claim 1, further comprising:

adaptively selecting a first post-processing mode for the tooth classification prediction result through the first adaptive module;

adaptively selecting a second post-processing mode for the tooth partition prediction result through the second adaptive module,

the determining the tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image comprises:

carrying out post-processing on the tooth classification prediction result of the target oral medical image through the first post-processing mode to obtain a post-processed tooth classification prediction result;

performing post-processing on the tooth partition prediction result of the target oral medical image through the second post-processing mode to obtain a post-processed tooth partition prediction result;

and fusing the post-processed tooth classification prediction result and the post-processed tooth partition prediction result to obtain a tooth segmentation result of the target oral medical image.

10. The method of dental segmentation of an oral medical image according to claim 9, wherein the first post-processing manner includes connected component analysis, the dental classification prediction result includes probability data of a dental classification label to which a voxel in the target oral medical image belongs, and the post-processed dental classification prediction result includes a dental classification label to which a dental region of the target oral medical image belongs,

the post-processing the tooth classification prediction result of the target oral medical image by the first post-processing mode to obtain a post-processed tooth classification prediction result, including:

and performing connected domain analysis on the probability data of the tooth classification label to which the voxel in the target oral medical image belongs to obtain the tooth classification label to which the tooth area of the target oral medical image belongs.

11. The method of dental segmentation of an oral medical image according to claim 9, wherein the second post-processing means includes connected component analysis, the dental segmentation prediction result includes probability data of a dental segmentation label to which a voxel in the target oral medical image belongs, and the post-processed dental segmentation prediction result includes a dental segmentation label to which a dental region of the target oral medical image belongs,

the post-processing the tooth partition prediction result of the target oral medical image by the second post-processing method to obtain the post-processed tooth partition prediction result includes:

and performing connected domain analysis on the probability data of the tooth zone label to which the voxel in the target oral medical image belongs to obtain the tooth zone label to which the tooth area of the target oral medical image belongs.

12. The method of dental segmentation in oral medical images according to claim 9, wherein the post-processed dental classification prediction result includes a dental classification label to which a dental region of the target oral medical image belongs, and the post-processed dental segmentation prediction result includes a dental segmentation label to which a dental region of the target oral medical image belongs,

the fusing the post-processed tooth classification prediction result and the post-processed tooth partition prediction result to obtain a tooth segmentation result of the target oral medical image comprises:

fusing a tooth classification label to which the tooth region of the target oral medical image belongs and a tooth partition label to which the tooth region of the target oral medical image belongs to obtain a tooth segmentation label to which the tooth region of the target oral medical image belongs;

and segmenting the target oral medical image according to the tooth segmentation label of the tooth area of the target oral medical image so as to obtain the tooth segmentation result of the target oral medical image.

13. An apparatus for dental segmentation of an oral medical image, the apparatus comprising:

the first determination module is used for determining a first training set used for training a tooth classification prediction module in a tooth classification prediction model and a second training set used for training a tooth partition prediction module in a tooth partition prediction model according to an oral medical image sample obtained by scanning an oral cavity by an oral cavity scanning device;

the first training module is used for adaptively generating a first pipeline fingerprint used for training the tooth classification prediction module according to the first data fingerprint of the first training set through a first self-adaptive module in the tooth classification prediction model, and training the tooth classification prediction module according to the first training set and the first pipeline fingerprint;

the second training module is used for adaptively generating a second pipeline fingerprint used for training the tooth partition prediction module according to a second data fingerprint of the second training set through a second self-adaptive module in the tooth partition prediction model, and training the tooth partition prediction module according to the second training set and the second pipeline fingerprint;

the prediction module is used for predicting the classification of the teeth in the target oral medical image through the trained tooth classification prediction module and predicting the subareas of the teeth in the target oral medical image through the trained tooth subarea prediction module;

and the second determination module is used for determining the tooth segmentation result of the target oral medical image according to the tooth classification prediction result and the tooth partition prediction result of the target oral medical image.

14. An electronic device, characterized in that the device comprises:

one or more processors;

a computer readable medium configured to store one or more programs,

when executed by the one or more processors, cause the one or more processors to implement the method for dental segmentation of oral medical images of any one of claims 1-12.

15. A computer-readable medium, on which a computer program is stored which, when being executed by a processor, carries out a method for dental segmentation of an oral medical image according to any one of claims 1 to 12.