Oral root canal treatment filling quality assessment method and system based on deep learning

Abstract

The present invention belongs to the field of image processing technology. Provided is a method and system for evaluating the quality of oral root canal treatment and filling based on deep learning, obtaining an X-ray dental film image of oral root canal treatment; converting the spatial information of the tooth into frequency domain information and storing it in the loss function of the first deep learning model, using the first deep learning model to segment the teeth of the root canal treatment part in the X-ray dental film image, and using the second deep learning model to obtain the classification result of the root canal treatment effect according to the segmentation result; the present invention converts the spatial information of the tooth into frequency domain information and stores it in the loss function, which helps the segmentation model to more accurately identify the boundary information of the tooth, thereby improving the accuracy and robustness of the segmentation, and helps to achieve automatic and accurate evaluation of the quality of root canal treatment and filling.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for evaluating the quality of oral root canal treatment and filling based on deep learning. Background Art

[0002] The statements in this section merely provide background art related to the present invention and do not necessarily constitute prior art.

[0003] As one of the main means of treating periodontal diseases, root canal treatment plays an important role in the patient's later recovery stage. However, the quality of treatment has a huge impact on the patient's health. Insufficient filling may lead to bacterial residues and cause inflammation, while excessive filling may cause complications such as pain or tissue necrosis. Therefore, it is particularly important to accurately evaluate the root canal after treatment, which helps to promptly detect and remedy insufficient or excessive filling. However, there are some challenges in the traditional root canal treatment result evaluation process. First, it is highly dependent on the doctor's expertise and rich experience. Secondly, the long-term manual evaluation of multiple X-rays is cumbersome and prone to subjective errors; in addition, the evaluation criteria between different doctors may differ, resulting in inconsistent results.

[0004] In the process of evaluating the results of root canal treatment, the first step is to segment the teeth for detailed analysis. However, the tooth boundary areas are often unclear in X-ray films, which may be caused by the shooting angle or the obstruction of the tissues around the teeth. In addition, the roots of the teeth may be obscured by the surrounding soft tissues or other teeth, making the structure of the root canal system difficult to observe, thereby limiting the visual information that can be obtained during the evaluation process. In this case, the doctor needs to rely on experience and imagination to infer the actual shape of the roots, which makes traditional manual evaluation methods face great difficulties. Summary of the invention

[0005] In order to address the shortcomings of the prior art, the present invention provides an oral root canal treatment filling quality assessment method based on deep learning, which converts the spatial information of the tooth into frequency domain information and stores it in the loss function, which helps the segmentation model to more accurately identify the boundary information of the tooth, thereby improving the accuracy and robustness of the segmentation, and helping to achieve automatic and accurate root canal treatment filling quality assessment.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for evaluating the quality of oral root canal treatment filling based on deep learning.

[0008] A method for evaluating the quality of oral root canal treatment and filling based on deep learning, including the following processes:

[0009] Obtain X-ray dental images for oral root canal treatment;

[0010] The spatial information of the teeth is converted into frequency domain information and stored in the loss function of the first deep learning model. The first deep learning model is used to segment the root canal treatment part of the teeth in the X-ray dental image. The second deep learning model is used according to the segmentation result to obtain the classification result of the root canal treatment effect.

[0011] As a further limitation of the first aspect of the present invention, converting the spatial information of the tooth into frequency domain information and storing it in the loss function of the first deep learning model includes:

[0012] The loss function of the first deep learning model is the original loss function and The product of , represents the frequency domain representation of the prediction boundary a, Represents the frequency domain representation of the manually labeled boundary b. The loss function of the first deep learning model is a quantitative representation of the closed shape.

[0013] As a further limitation of the first aspect of the present invention, the first deep learning model is a symmetrical U-shaped structure, the first branch of the U-shaped structure is a convolutional layer and four Mamba modules connected in sequence, and the second branch of the U-shaped structure is four Mamba modules and a convolutional layer connected in sequence; wherein the Mamba module is composed of a VSS module and a depth-separable convolution combination;

[0014] Two convolutional layers are used to extract shallow features, and eight Mamba modules introduce global visual information through depthwise separable convolutions and VSS modules while processing local features to segment teeth and root canal fillings.

[0015] As a further limitation of the first aspect of the present invention, the Mamba modules on the first branch are sequentially a first Mamba module, a second Mamba module, a third Mamba module, and a fourth Mamba module, and the Mamba modules on the second branch are sequentially a fifth Mamba module, a sixth Mamba module, a seventh Mamba module, and an eighth Mamba module;

[0016] The first Mamba module is connected to the eighth Mamba module through a spatial attention mechanism, the second Mamba module is connected to the seventh Mamba module through a spatial attention mechanism, the third Mamba module is connected to the sixth Mamba module through a spatial attention mechanism, and the fourth Mamba module is connected to the fifth Mamba module through a spatial attention mechanism.

[0017] As a further limitation of the first aspect of the present invention, the Mamba module includes two branches, for each branch, a depth-wise separable convolution is used to extract shallow features, and then two VSS modules are sequentially passed through the VSS modules, the VSS modules capture the global semantic information in the image by introducing global information and cross-channel interaction, and after each branch extracts features, the features extracted by each branch and the original input features are integrated and fused to form a final feature representation;

[0018] The VSS module consists of two branches. The first branch includes a linear layer, a depth-wise separable convolutional layer, a SiLU activation function, an SS2D module, and a normalization layer, which is used to extract and model the long-range spatial features of the image. The second branch includes a linear layer and a SiLU activation function, which is used to extract the local features of the image.

[0019] The outputs of the two branches are multiplied to aggregate the features of the two branches, and then processed through a linear layer to obtain the fused features. The fused features are weighted summed with the original input of the VSS module to obtain the final output of the VSS module.

[0020] As a further limitation of the first aspect of the present invention, the classification results of root canal treatment effects include just filling and underfilling. The distance between the top pixel points of the tooth and the filling is calculated. If the distance is greater than a set threshold, it is underfilled. If the distance is less than or equal to the set threshold, it is just fill. The distance is a positive number.

[0021] In a second aspect, the present invention provides an oral root canal treatment filling quality assessment system based on deep learning.

[0022] A deep learning-based oral root canal treatment filling quality assessment system, comprising:

[0023] An image acquisition unit is configured to: acquire an X-ray dental film image of an oral root canal treatment;

[0024] The evaluation and classification unit is configured to: convert the spatial information of the teeth into frequency domain information and store it in the loss function of the first deep learning model, use the first deep learning model to segment the root canal treatment part of the teeth in the X-ray dental image, and use the second deep learning model to obtain the root canal treatment effect classification result based on the segmentation result.

[0025] In a third aspect, the present invention provides a computer device, comprising: a processor and a computer-readable storage medium;

[0026] a processor adapted to execute a computer program;

[0027] A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, the method for evaluating the quality of oral root canal treatment and filling based on deep learning as described in the first aspect of the present invention is implemented.

[0028] In a fourth aspect, the present invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing the oral root canal treatment filling quality assessment method based on deep learning as described in the first aspect of the present invention.

[0029] In a fifth aspect, the present invention provides a computer program product, comprising a computer program. When the computer program is executed by a processor, it implements the oral root canal treatment filling quality assessment method based on deep learning as described in the first aspect of the present invention.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. The present invention introduces the concept of prior knowledge loss function, converts the spatial information of teeth into frequency domain information, and stores it in the loss function, which helps the segmentation model to more accurately identify the boundary information of teeth, thereby improving the accuracy and robustness of segmentation, enabling the model to better understand the spatial relationship of tooth structure, and then effectively distinguish the boundaries between teeth and surrounding tissues.

[0032] 2. The present invention adopts two deep learning models (i.e., the first deep learning model and the second deep learning model) to effectively evaluate the quality of oral root canal treatment. It combines image segmentation and classification techniques. In the first stage, teeth are segmented, and in the second stage, root canal treatment effects are classified. Through this two-stage design, the results of root canal treatment can be evaluated more comprehensively, thereby improving the accuracy and reliability of the evaluation.

[0033] 3. The present invention realizes feature fusion and integration by combining the semantic information of edge local features of convolutional neural network with the semantic information context of global features of the main body of Mamba module, so that features of different scales can complement each other better, thereby improving the model's recognition ability for complex tooth structures and enhancing the fusion effect of features of different scales.

[0034] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0036] Figure 1 A schematic flow chart of a method for evaluating oral root canal treatment filling quality based on deep learning provided in Example 1 of the present invention;

[0037] Figure 2 A schematic diagram of a first deep learning model provided in Example 1 of the present invention;

[0038] Figure 3 A schematic diagram of a Mamba module provided in Example 1 of the present invention;

[0039] Figure 4 A schematic diagram of a VSS module provided in Embodiment 1 of the present invention;

[0040] Figure 5 A schematic diagram of an oral root canal treatment filling quality assessment system based on deep learning provided in Example 2 of the present invention;

[0041] Figure 6 This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0043] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0044] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0045] Embodiment 1:

[0046] In an embodiment, a method for evaluating the quality of oral root canal treatment filling based on deep learning is proposed, such as Figure 1 As shown, the following process is included:

[0047] S1: Obtain X-ray dental images of oral root canal treatment;

[0048] S2: Convert the spatial information of the teeth into frequency domain information and store it in the loss function of the first deep learning model. Use the first deep learning model to segment the root canal treatment part of the teeth in the X-ray dental image. Use the second deep learning model to obtain the root canal treatment effect classification result based on the segmentation result.

[0049] In step S2 of the present implementation method, specifically, it includes: by introducing a priori knowledge Loss function, the spatial information of the teeth can be converted into frequency domain information, and embedded into the loss function to help the segmentation model better identify the boundary information of the teeth. Specifically, the process of converting spatial information into frequency domain information can enable the model to more effectively capture the global features and local details of the tooth structure. Such a priori knowledge Loss function can not only improve the accuracy of the segmentation model for tooth boundaries, but also enhance the model's ability to understand the tooth structure, thereby improving the quality and stability of the segmentation results.

[0050] According to the prior knowledge of anatomy, tooth roots have similar shapes. In order to better guide the model to segment the tooth roots, the shape information is added to the Loss function.

[0051] Assume that (xm, ym) is one of the coordinates on the root boundary containing N pixels, and the boundary shape can form a complex number z(m)=xm+jym, where j is an imaginary unit. This shape can be described by a priori knowledge Loss function. Specifically, it can be used to represent the boundary shape. By using the prior knowledge Loss function, the shape information of the root boundary can be converted into frequency domain information, so as to better capture its structural features. Adding this shape information to the loss function can make the model pay more attention to the shape characteristics of the root during the segmentation process, thereby improving the accuracy and stability of the segmentation. This method can effectively utilize the anatomical prior knowledge to provide stronger guidance and constraints for the segmentation model, thereby improving the segmentation effect of complex structures such as roots.

[0052] Further, the formula is as follows:

[0053] (1);

[0054] (2);

[0055] (3);

[0056] in, Z(i) Represents input i The frequency domain representation of z(l) Represents a point l The complex number representation of BCE represents the original loss function based on the predicted boundary a and the true boundary b. N Represents the number of pixels, m Representative m pixels, is the input for predicting boundary a i The frequency domain representation of is the input of the real boundary bi The frequency domain representation of is the input for predicting boundary a i The frequency domain representation of the input and the real boundary b i The modulus of the difference in frequency domain representation, is the final prior knowledge loss function.

[0057] Furthermore, the prior knowledge Loss function is a quantitative representation of a closed shape, which is independent of its starting point, scale, position and rotation. The frequency domain representation of the predicted boundary a and the true boundary b can be calculated according to formula (1). The shape difference between the predicted boundary a and the manually marked boundary b can be quantified according to formula (2). Formula (3) is used to add to the training process of the segmentation network.

[0058] In step S2 of this implementation, the training of the first deep learning model and the second deep learning model both require the construction of data sets and data annotation, specifically including:

[0059] A certain amount of X-ray dental image data is obtained, all of which are from patients who have undergone root canal treatment, and the dental data with extremely poor exposure and poor quality are eliminated; multiple professional dentists will mark the treated teeth and root canal fillings, and multiple doctors who did not participate in the labeling will evaluate it to ensure the authenticity of the label.

[0060] More specifically, we collect X-ray dental image data of patients who have gone through the root canal treatment stage (the X-ray dental data here are all data that excludes patient privacy and is used with the consent of the patients) to build a comprehensive root canal treatment evaluation data set. These X-ray dental data will cover various treatment situations and the diversity of different cases, thereby providing a rich source of information for model training and evaluation, ensuring the quality and integrity of the data. Through careful screening and analysis, the accuracy and reliability of each data sample are confirmed. Each data set is crucial for the development and verification of the evaluation model. It will provide dentists with a powerful tool for accurately evaluating the effect of root canal treatment and the patient's dental health status. After a rigorous process, senior dentists will accurately annotate the teeth and fillings that have undergone root canal treatment. This process requires careful review by multiple professional doctors to ensure the accuracy and consistency of the annotations. Each doctor will conduct a detailed review of the images and actual conditions in order to accurately capture any possible problems or potential treatment plans. The annotation results will be a highly reliable and accurate data set, providing a solid basis for subsequent treatment decisions.

[0061] like Figure 2As shown, the first deep learning model focuses on the tooth segmentation model of X-ray images. This model will accurately separate the tooth area from the complex X-ray image through the deep learning algorithm and accurately identify its boundary and structure. Considering the differences in tooth morphology between different patients and the possible ambiguity of X-ray images, the present invention will focus on the robustness and generalization ability of the model to ensure that the segmentation task can be reliably completed.

[0062] More specifically, the first deep learning model is a symmetrical U-shaped structure, and the first branch of the U-shaped structure (i.e. Figure 2 The left branch in the figure is composed of convolutional layers and four Mamba modules connected in sequence. The second branch of the U-shaped structure (i.e. Figure 2 The right branch in the figure is composed of four Mamba modules and convolutional layers connected in sequence. These modules cooperate with each other to efficiently capture the semantic information of global and local attention. Among them, the Mamba module is composed of a VSS module and a depth-wise separable convolution.

[0063] In this implementation, two convolutional layers are used to extract shallow features, and eight Mamba modules are responsible for introducing global visual information through depthwise separable convolution and VSS modules, while effectively processing local features, thereby achieving accurate segmentation of teeth and root canal fillings.

[0064] In this implementation, further, the Mamba modules on the first branch are sequentially a first Mamba module, a second Mamba module, a third Mamba module, and a fourth Mamba module, and the Mamba modules on the second branch are sequentially a fifth Mamba module, a sixth Mamba module, a seventh Mamba module, and an eighth Mamba module;

[0065] The first Mamba module is connected to the eighth Mamba module through a spatial attention mechanism, the second Mamba module is connected to the seventh Mamba module through a spatial attention mechanism, the third Mamba module is connected to the sixth Mamba module through a spatial attention mechanism, and the fourth Mamba module is connected to the fifth Mamba module through a spatial attention mechanism, which means that shallow spatial information and deep spatial information can be effectively exchanged. This jump connection mechanism helps to improve the model's perception of features at different levels, thereby enhancing the model's representation ability and segmentation accuracy.

[0066] In this implementation, preferably, the Mamba module, such as Figure 3 As shown, it includes two branches (to achieve multi-scale extraction of image features and modeling of long-range dependencies). For each branch, a depth-separable convolution is used to extract shallow features (the first convolution of the above branch is , the last convolution of the above branch is , the first convolution of the following branch is , the first convolution of the following branch is ), this convolution operation can effectively capture the local details of the image, and extract features of different scales by setting symmetrical and different-sized convolution kernels. Such a design enables the model to have a wider range of perception capabilities when processing images, and can simultaneously pay attention to feature information at different scales, thereby improving the model's understanding and representation capabilities of images; then two VSS modules are used to model long-range dependencies. The VSS module can better capture the global semantic information in the image by introducing global information and cross-channel interactions, and effectively model long-range contextual dependencies. This design can make the model more accurate and comprehensive in understanding the overall structure and semantic information of the image. At the same time, in order to maintain the importance of shallow features and promote the fusion of features of different scales, combined with jump connections, it can maintain the effectiveness of shallow features and ensure information exchange between features of different scales, thereby enhancing the model's representation ability and segmentation accuracy; after extracting features from each branch, the features extracted from each branch and the original input features are integrated and fused to form the final feature representation. Such a fusion operation can further improve the model's understanding of the image and segmentation performance, and provide a more reliable feature representation for subsequent classification tasks.

[0067] In this implementation, preferably, Figure 4 As shown, the VSS module includes two branches. The first branch includes a linear layer, a depth-separable convolution layer, a SiLU activation function, an SS2D module (i.e., 2D-Selective-Scan) and a normalization layer, which are used to extract and model the long-range spatial features of the image; the second branch includes a linear layer and a SiLU activation function, which are used to extract the local features of the image.

[0068] The outputs of the two branches are multiplied to aggregate the features of the two branches, and then a linear layer is processed to obtain the fused features. The fused features are weighted summed with the original input of the VSS module to obtain the final output of the VSS module. This design enables the VSS module to comprehensively utilize global and local features, fully mine the information in the image, and provide the model with richer and more accurate feature representation, thereby improving the performance of the model in image processing tasks.

[0069] The second deep learning model in this implementation mainly uses CNN and Vision Transformer modules as the backbone network to extract local and global features of the root canal treatment part and finally complete the classification.

[0070] In step S2 of the present implementation method, the classification results of the root canal treatment effect, including just filling and underfilling, are calculated. The distance between the top pixel points of the tooth and the filling is calculated. If the distance is greater than a set threshold, it is underfilled. If the distance is less than or equal to the set threshold, it is just filled. The distance is a positive number.

[0071] More specifically, by accurately segmenting the teeth and root canal fillings, the pixel points of the root apex and the filling apex can be obtained, and the accurate distance between the root and the filling can be obtained by calculating the vertical distance between these two pixel points. This process is crucial for evaluating the effectiveness of root canal treatment, because the distance between the root apex and the filling apex can reflect the success of the root canal treatment. A smaller distance usually means a smaller gap between the filling and the root apex and a better treatment effect, while a larger distance may imply that the filling is not in an ideal position or there are other problems. Therefore, by accurately segmenting the teeth and root canal fillings, more reliable data can be obtained, which can then more accurately evaluate the quality of root canal treatment and the patient's dental health.

[0072] For example, after classification, two categories, just-filled and under-filled, will be screened out. The pixel positions at the top of the segmented teeth and fillings can be obtained, and then the distance between the two points can be calculated by calling np.sqrt through the numpy package in python. If it is greater than 0.02 (that is, the set threshold) mm, it is under-filled, and if it is between 0 and 0.02 mm (that is, the set threshold), it is just-filled.

[0073] In this implementation, after the first deep learning model and the second deep learning model are trained, they need to be tested. The specific testing process includes:

[0074] The input X-ray image is sent to the trained first deep learning model and the second deep learning model for processing. The first deep learning model and the second deep learning model have been fully trained and have the ability to efficiently process X-ray images. During the processing, the first deep learning model will first perform tooth segmentation on the image, accurately distinguish the areas of teeth and root canal fillings, and obtain clear boundary information of teeth and fillings, laying the foundation for subsequent evaluation work; for the root canal treatment part, the second deep learning model will further evaluate the quality of the filling. It will accurately identify the location, shape, texture and other features of the filling, calculate the vertical distance of the key points, that is, the distance between the top of the root and the top of the filling, and set the threshold for evaluation according to the pre-set distance.

[0075] Embodiment 2:

[0076] like Figure 5As shown, this implementation method provides an oral root canal treatment filling quality assessment system based on deep learning, including:

[0077] An image acquisition unit is configured to: acquire an X-ray dental image of an oral root canal treatment; the specific working method of this unit is the same as the process of step S1 provided in embodiment 1, and will not be repeated here;

[0078] The evaluation and classification unit is configured to: convert the spatial information of the teeth into frequency domain information and store it in the loss function of the first deep learning model, use the first deep learning model to segment the root canal treatment part of the teeth in the X-ray dental image, and use the second deep learning model to obtain the root canal treatment effect classification result according to the segmentation result; the specific working method of this unit is the same as the process of step S2 provided in Example 1, and will not be repeated here.

[0079] It is understandable that the above-mentioned units can be separately or completely combined into one or several other units to constitute, or one (some) of the units can be further divided into multiple functionally smaller units to constitute, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above-mentioned units are divided based on logical functions. In practical applications, the function of one unit can also be implemented by multiple units, or the functions of multiple units can be implemented by one unit. In other embodiments of the present application, the image processing device may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of multiple units.

[0080] According to another embodiment of the present application, the system described in this embodiment can be constructed and the image processing method of the embodiment of the present application can be implemented by running a computer program (including program code) capable of executing the steps involved in the corresponding method described in Example 1 on a general computing device such as a computer including processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, and loaded into the above-mentioned computing device through the computer-readable recording medium and run therein.

[0081] Embodiment 3:

[0082] like Figure 6As shown, this implementation provides an electronic device, which includes a processor 1001, a communication interface 1002, and a computer-readable storage medium 1003. The processor 1001, the communication interface 1002, and the computer-readable storage medium 1003 may be connected via a bus or other means.

[0083] Among them, the communication interface 1002 is used to receive and send data, the computer-readable storage medium 1003 can be stored in the memory of the electronic device, the computer-readable storage medium 1003 is used to store a computer program, the computer program includes program instructions, and the processor 1001 is used to execute the program instructions stored in the computer-readable storage medium 1003.

[0084] The processor 1001 (or CPU (Central Processing Unit)) is the computing core and control core of the electronic device, which is suitable for implementing one or more instructions, and specifically suitable for loading and executing one or more instructions to implement corresponding method processes or corresponding functions.

[0085] The processor 1001 is configured to execute the following process:

[0086] Acquire an X-ray dental image of oral root canal treatment; the specific working method of this unit is the same as the process of step S1 provided in Example 1, and will not be repeated here;

[0087] The spatial information of the teeth is converted into frequency domain information and stored in the loss function of the first deep learning model. The first deep learning model is used to segment the teeth of the root canal treatment part in the X-ray dental image. The second deep learning model is used to obtain the classification result of the root canal treatment effect based on the segmentation result. The specific working method of this unit is the same as the process of step S2 provided in Example 1, which will not be repeated here.

[0088] Embodiment 4:

[0089] This implementation provides a computer-readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the electronic device and, of course, extended storage media supported by the electronic device. The computer-readable storage medium provides a storage space that stores the processing system of the electronic device.

[0090] In addition, the storage space also stores one or more instructions suitable for being loaded and executed by the processor, and these instructions may be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here may be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage; optionally, it may also be at least one computer-readable storage medium located away from the aforementioned processor.

[0091] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to implement the following process:

[0092] Acquire an X-ray dental image of oral root canal treatment; the specific working method of this unit is the same as the process of step S1 provided in Example 1, and will not be repeated here;

[0093] The spatial information of the teeth is converted into frequency domain information and stored in the loss function of the first deep learning model. The first deep learning model is used to segment the teeth of the root canal treatment part in the X-ray dental image. The second deep learning model is used to obtain the classification result of the root canal treatment effect based on the segmentation result. The specific working method of this unit is the same as the process of step S2 provided in Example 1, which will not be repeated here.

[0094] Embodiment 5:

[0095] This embodiment provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device performs the following process:

[0096] Acquire an X-ray dental image of oral root canal treatment; the specific working method of this unit is the same as the process of step S1 provided in Example 1, and will not be repeated here;

[0097] The spatial information of the teeth is converted into frequency domain information and stored in the loss function of the first deep learning model. The first deep learning model is used to segment the teeth of the root canal treatment part in the X-ray dental image. The second deep learning model is used to obtain the classification result of the root canal treatment effect based on the segmentation result. The specific working method of this unit is the same as the process of step S2 provided in Example 1, which will not be repeated here.

[0098] A person skilled in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0099] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from a website site, a computer, a server or a data center to another website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server, a data center, etc. that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating the quality of oral root canal treatment and filling based on deep learning, characterized in that: The process includes: Obtain X-ray dental images for oral root canal treatment; The spatial information of the teeth is converted into frequency domain information and stored in the loss function of the first deep learning model, the first deep learning model is used to segment the root canal treatment part of the teeth in the X-ray dental image, and the second deep learning model is used to obtain the root canal treatment effect classification result according to the segmentation result; The loss function of converting the spatial information of the tooth into frequency domain information and storing it in the first deep learning model is specifically: The loss function of the first deep learning model is the original loss function and The product of , represents the frequency domain representation of the prediction boundary a, represents the frequency domain representation of the manually marked boundary b. The loss function of the first deep learning model is a quantitative representation of the closed shape. The original loss function is , is the original loss function based on the predicted boundary a and the true boundary b.

2. The oral root canal treatment filling quality assessment method based on deep learning according to claim 1, characterized in that: The first deep learning model is a symmetrical U-shaped structure, with the first branch of the U-shaped structure consisting of sequentially connected convolutional layers and four Mamba modules, and the second branch of the U-shaped structure consisting of sequentially connected four Mamba modules and a convolutional layer; wherein the Mamba module is composed of a VSS module and a depthwise separable convolution; Two convolutional layers are used to extract shallow features, and eight Mamba modules introduce global visual information through depthwise separable convolutions and VSS modules while processing local features to segment teeth and root canal fillings.

3. The oral root canal treatment filling quality assessment method based on deep learning according to claim 2, characterized in that: The Mamba modules on the first branch are the first Mamba module, the second Mamba module, the third Mamba module and the fourth Mamba module, and the Mamba modules on the second branch are the fifth Mamba module, the sixth Mamba module, the seventh Mamba module and the eighth Mamba module; The first Mamba module is connected to the eighth Mamba module through a spatial attention mechanism, the second Mamba module is connected to the seventh Mamba module through a spatial attention mechanism, the third Mamba module is connected to the sixth Mamba module through a spatial attention mechanism, and the fourth Mamba module is connected to the fifth Mamba module through a spatial attention mechanism.

4. The oral root canal treatment filling quality assessment method based on deep learning as claimed in claim 2, characterized in that: The Mamba module consists of two branches. For each branch, a depth-wise separable convolution is used to extract shallow features. Then, the features are passed through two VSS modules in sequence. The VSS modules capture the global semantic information in the image by introducing global information and cross-channel interactions. After extracting features from each branch, the features extracted from each branch are integrated and fused with the original input features to form the final feature representation. The VSS module includes two branches, the first branch includes a linear layer, a depth-separable convolution layer, a SiLU activation function, an SS2D module and a normalization layer, which are used to extract and model the long-range spatial features of the image; the second branch includes a linear layer and a SiLU activation function, which are used to extract the local features of the image; wherein the SS2D module refers to a 2D selective scanning module; The outputs of the two branches are multiplied to aggregate the features of the two branches, and then processed through a linear layer to obtain the fused features. The fused features are weighted summed with the original input of the VSS module to obtain the final output of the VSS module.

5. The oral root canal treatment filling quality assessment method based on deep learning according to claim 1, characterized in that: The classification results of root canal treatment effects include just-filled and under-filled. The distance between the top pixel points of the tooth and the filling is calculated. If the distance is greater than a set threshold, it is under-filled. If the distance is less than or equal to the set threshold, it is just-filled. The distance is a positive number.

6. A deep learning-based oral root canal treatment filling quality assessment system, characterized in that: include: An image acquisition unit is configured to: acquire an X-ray dental film image of an oral root canal treatment; The evaluation and classification unit is configured to: convert the spatial information of the tooth into frequency domain information and store it in the loss function of the first deep learning model, use the first deep learning model to segment the root canal treatment part of the tooth in the X-ray dental image, and use the second deep learning model to obtain the root canal treatment effect classification result according to the segmentation result; The loss function of converting the spatial information of the tooth into frequency domain information and storing it in the first deep learning model is specifically: The loss function of the first deep learning model is the original loss function and The product of , represents the frequency domain representation of the prediction boundary a, represents the frequency domain representation of the manually marked boundary b. The loss function of the first deep learning model is a quantitative representation of the closed shape. The original loss function is , is the original loss function based on the predicted boundary a and the true boundary b.

7. A computer device, characterized in that: include: a processor and a computer readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, the method for evaluating the quality of oral root canal treatment and filling based on deep learning as described in any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which is suitable for being loaded by a processor and executing the oral root canal treatment filling quality assessment method based on deep learning as described in any one of claims 1 to 5.

9. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, it implements the oral root canal treatment filling quality assessment method based on deep learning as described in any one of claims 1 to 5.

Images