Artificial intelligence-based color guide automated prosthetic manufacturing process machining method and system

Abstract

The application discloses an artificial intelligence-based shadow guide and an automatic dental restoration manufacturing process processing system, which can include a memory and a processor to store instructions. When the processor executes the instructions, the system receives input of a patient's tooth image, extracts color information of the tooth from the input image, uses a pre-trained artificial intelligence model to compare and analyze the extracted color information with pre-stored color data of the shadow guide, automatically determines the closest shadow value, generates processing data for manufacturing the restoration tooth according to the determined shadow value, and controls a restoration tooth manufacturing device according to the generated processing data to control the manufacturing of the restoration tooth of the corresponding color.

Description

Technical Field

[0001] This invention relates to the field of manufacturing dental restorations, and more specifically, to a process method and system for automatically determining shading direction using artificial intelligence-based tooth color analysis technology and manufacturing dental restorations with accurate colors based on the same process. Background Technology

[0002] In traditional dental prosthesis manufacturing, the color of a patient's teeth is primarily determined by a manual color guide. In practice, the dentist places a color guide with multiple shades next to the patient's teeth and visually compares them to select the most similar color. However, this method has significant drawbacks. Its accuracy is affected by numerous factors, such as the clinic's lighting conditions, ambient light reflection, and the observer's subjective judgment. Furthermore, even when using photography to convey tooth color information, differences in camera settings and shooting environments often result in discrepancies between the captured color and the actual tooth color, leading to inaccurate prosthesis colors. Summary of the Invention

[0003] The core objective of this invention is to provide an automatic light-shielding system. This system fully utilizes artificial intelligence technology to determine the color of a patient's teeth in an objective and accurate manner. Furthermore, this invention has another important objective: to provide an integrated system. This system can directly link the determined color information with the manufacturing process of the restorative teeth, thereby achieving the production of restorations with consistent quality.

[0004] The present invention solves the technical problem by adopting the following technical solution:

[0005] An AI-based shading guide and automated dental prosthesis manufacturing process system includes memory and a processor for storing instructions. When the processor executes these instructions, the system receives input information about the patient's teeth in an image. The system then extracts the tooth color information from the input image. Next, using a pre-trained AI model, the extracted color information is compared and analyzed with pre-stored shading guide color data. Through this process, the system can automatically determine the closest shading value.

[0006] After determining the shading value, the system generates processing data specifically for manufacturing dental restorations. Furthermore, the system controls the dental restoration manufacturing apparatus based on this data, thereby achieving precise control over the manufacturing process of dental restorations of the corresponding color.

[0007] It is worth mentioning that this system can be replaced by electronic devices (such as terminals, computers, etc.). The system operations executed by instructions can be categorized as a method.

[0008] The present invention has the following beneficial effects:

[0009] The present invention relates to an artificial intelligence-based color guide, which enables automated processing methods and systems for the manufacture of dental restorations. This color guide, utilizing artificial intelligence technology, can objectively and accurately determine tooth color, thereby significantly improving the color accuracy of dental restorations.

[0010] Furthermore, AI-based shading guidelines enable automated manufacturing processes and systems for dental restorations. This approach establishes a direct link between the determined color information and the manufacturing process, resulting in the production of restorations with consistent quality and significantly improved production efficiency.

[0011] Furthermore, this AI-based color guide and related system can minimize communication errors between dental clinics and laboratories. This reduction in errors shortens restoration production time and lowers costs. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of a system using an artificial intelligence model according to a single embodiment.

[0013] Figure 2 An illustration to illustrate the learning of a neural network according to a single embodiment.

[0014] Figure 3 A schematic diagram illustrating the composition of an artificial intelligence model according to an embodiment.

[0015] Figure 4 This demonstrates the configuration of a dental prosthesis manufacturing process handling system, which can be automated using AI-based shadow guidance in a single embodiment.

[0016] Figure 5 The flowchart illustrates a dental prosthesis manufacturing process that can be automated using an AI-based shading guide, according to one embodiment.

[0017] Figure 6 The flowchart illustrates a process method for automating the manufacturing process of dental prostheses using AI-based shadow guidance, according to one embodiment.

[0018] Figure 7The flowchart illustrates a process method for automating the manufacturing process of dental prostheses using AI-based shadow guidance, according to one embodiment.

[0019] Figure 8 To illustrate the process of creating a dental prosthesis through a dental prosthesis manufacturing process handling system, which can automate AI-based shadow guidance according to one embodiment. Detailed Implementation

[0020] To make the technical problems, solutions, and advantages of this invention clearer, various embodiments are described in detail below with reference to the accompanying drawings. However, it should be understood that these embodiments have many possible modifications. Therefore, the scope of this patent application is not limited or constrained by these embodiments. It is emphasized that all changes made to these embodiments and equivalent substitutions should be considered within the scope of this patent application.

[0021] Furthermore, the descriptions of the specific structures or functions involved in each embodiment are merely illustrative. In reality, these embodiments can be transformed into many different forms. Therefore, these embodiments are not limited to any particular form. Moreover, the scope covered by this specification should include various variations, equivalents, or alternatives arising from the technical concept of this invention.

[0022] The terms “first” or “second” may be used to describe various components, but they are only for distinguishing purposes. For example, the first component may be named the second component, and vice versa. When it is said that one component is “connected” to another component, it means that it can be directly connected, or that there may be another component between them. The terminology used is for illustrative purposes only and is not intended to be limiting. Singular expressions include plurals unless the context otherwise requires. “Including” or “having” in this specification should be understood as indicating the presence of a feature or element described in the specification, rather than excluding the presence of other features or elements.

[0023] Unless otherwise defined, all terms used herein have the meanings commonly understood by one of those skilled in the art. Terms defined in common dictionaries should be interpreted in a meaning appropriate to the relevant art context and should not be interpreted in an overly formal sense unless explicitly defined in this application. Furthermore, when describing drawings, the same reference numerals should be used for the same components, and repetitive descriptions should be omitted. Detailed descriptions may be omitted if the description of the relevant art is more ambiguous than necessary when describing embodiments.

[0024] Implementations can take many forms, including personal computers, laptops, tablets, smartphones, televisions, smart home appliances, smart cars, self-service terminals, and wearable devices. Artificial intelligence (AI) systems are computer systems that achieve human-level intelligence. Unlike existing rule-based intelligent systems, they are systems where machines learn and make decisions autonomously. As AI systems are used more extensively, their recognition rates improve, and user preferences are better understood; therefore, existing systems are gradually being replaced by deep learning-based AI.

[0025] Artificial intelligence technology consists of machine learning and several elemental technologies that use it. Machine learning refers to algorithms that classify and learn from the features of input data. Elemental technologies are techniques that replicate human cognitive and judgment functions through machine learning algorithms such as deep learning. These skills include language understanding, visual understanding, reasoning and prediction, knowledge representation, and motor control.

[0026] The various fields where artificial intelligence technology is applied are as follows: Language understanding is the technology of recognizing and processing human language, including natural language processing, machine translation, dialogue systems, question answering, speech recognition, and synthesis. Visual understanding is the art of recognizing and processing objects, including object recognition, tracking, image search, human recognition, and scene understanding. Reasoning and prediction is a technology of judging and predicting information, including knowledge-based reasoning, optimization prediction, and recommendation systems. Knowledge representation is a technology of automatically processing information about human experiences, including knowledge construction and management. Motion control is a technology of controlling the movement of autonomous vehicles or robots, including navigation, collision avoidance, and driving control. Generally, to apply machine learning algorithms to real life, they are trained using trial and error. Deep learning, in particular, requires hundreds of thousands of iterations. If it is difficult to implement in a real environment, a virtual environment is implemented on a computer, and learning is achieved through simulation.

[0027] In this invention, Artificial Intelligence (AI) refers to the technology of mimicking human learning, reasoning, and perception abilities and implementing them in computers, including concepts such as machine learning and symbolic logic. Machine learning (ML) is an algorithmic technique that learns the characteristics of input data on its own. Artificial intelligence technology can analyze input data, learn results, and make judgments or predictions based on these machine learning algorithms. Furthermore, techniques that use machine learning algorithms to simulate human cognition and judgment can also be categorized under artificial intelligence. Examples include language understanding, visual understanding, reasoning and prediction, knowledge representation, and motion control. Machine learning refers to the process of training neural network models based on experience in processing data. This allows computer software to improve its data processing capabilities on its own. Neural networks model the correlations between data, which can be represented by multiple parameters. The core of machine learning is that the neural network model extracts features from given data, analyzes them, derives relationships between the data, and repeats this process to optimize the model's parameters. For example, a neural network model can learn the relationship between inputs and outputs, or it can learn relationships simply by deriving patterns from the input data.

[0028] Artificial intelligence learning models, or neural network models, aim to replicate the structure of the human brain in a computer and include network nodes that simulate neurons. These nodes are interconnected to send and receive signals, and the AI ​​learning model processes data through layers of varying depths. These models may include artificial neural networks, convolutional neural networks (CNNs), and more. For example, supervised learning, unsupervised learning, and reinforcement learning methods can be used to perform machine learning on AI learning models. Machine learning algorithms include decision trees, Bayesian networks, support vector machines, artificial neural networks, iterative algorithms, perceptrons, genetic programming, clustering, and more.

[0029] A CNN (Convolutional Neural Network) is a multilayer perceptron designed with minimal preprocessing, consisting of one or more convolutional layers and a typical artificial neural network layer. Thanks to this structure, CNNs can efficiently utilize two-dimensional input data and perform well in video and audio domains. CNNs are trained using standard backpropagation, making them easier to train than other feedforward neural network techniques and offering the advantage of using fewer parameters. A convolutional network is a neural network containing a set of nodes bound to parameters; many computer vision tasks have improved with increasing training data volume and computational power. In today's large datasets, overfitting is not critical, and increasing the network size can improve test accuracy. Optimal use of computational resources becomes a limiting factor, necessitating distributed and scalable deep neural network implementations.

[0030] Figure 1 This is a schematic diagram illustrating the use of an artificial intelligence model to describe a system, as provided in an embodiment of the present invention.

[0031] The system 100 utilizing artificial intelligence models may include multiple user terminals 110-1, a server 120, and a database 130. According to one embodiment, the database 130 is configured separately from the server 120, but this is not necessary; the database 130 may reside within the server 120. For example, the server 120 may contain multiple artificial intelligences to execute machine learning algorithms. In another embodiment, the multiple user terminals 110-1, the server 120, and the database 130 may be connected via a network N to communicate with each other.

[0032] The network can perform wireless or wired communication between multiple user terminals (110-1, 01), a server (120), and a database (130). For example, the network can support wireless communication via LTE, subsequent evolutions of LTE, Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WDMA), wireless broadband, Bluetooth, Near Field Communication (NFC), and Global Positioning System (GPS). Alternatively, the network can also perform wired communication via Universal Serial Bus (USB), High Definition Multimedia Interface (HDM), Standard Interface (SMI), or conventional telephone service. Database 130 has the capability to store various types of data. The data stored in database 130 is data acquired, processed, or used by at least one component of the multiple user terminals 110-1 and server 120, and these user terminals 110-1 may include software (e.g., programs). Database 130 may include volatile and / or non-volatile memory.

[0033] Figure 2 This is an illustration provided by an embodiment of the present invention to illustrate learning a neural network according to a single embodiment.

[0034] The learning device can train neural network 123 to list comment responses received from multiple user terminals 110-1 item by item. Furthermore, the learning device can learn neural network 123 to extract user stop records based on user travel route information. According to one embodiment, the learning device may be an entity independent of server 120, but this is not limiting.

[0035] The neural network 123 includes an input layer 121 and an output layer 125, wherein training samples are input to generate a training output, and the network is trained based on the difference between the output and the labels. Labels can be defined based on comment response-related items and breadcrumb information in the user's dwell history, and the neural network 123 consists of multiple groups of nodes defined by weights and activation functions between nodes.

[0036] Learning devices can learn neural networks using either gradient descent (GD) or stochastic gradient descent (SGD) techniques.123 GD uses the entire dataset to tune the model's parameters, while SGD uses only a small amount of randomly selected data to speed up training. Learning devices can calculate the training error using a loss function designed in the form of mean squared error (MSE) or entropy, and use backpropagation to find and optimize the weights that influence the training error.

[0037] According to one embodiment, the learning device extracts a first object from the review response, obtains a first label, and applies it to a first neural network to perform training together with a generated first training output. Furthermore, a second object is extracted from trace navigation information to obtain a second label, and a second neural network is applied to perform training together with a generated second training output.

[0038] The learning device can generate a first training feature vector based on the composition, location, and pattern characteristics of the review response, and a second training feature vector based on the composition, length, and pattern characteristics of the motion path information. These feature vectors are applied to each neural network to generate training outputs, which, along with their labels, are used to train the comment item extraction algorithm and the user dwell history acquisition algorithm. In this context, backpropagation refers to a learning method that propagates the error generated in the output layer along the direction of the input layer and adjusts the weights of each layer.

[0039] Figure 3 This is a schematic diagram provided by an embodiment of the present invention, showing the composition of an artificial intelligence model according to one embodiment.

[0040] According to one embodiment, the AI ​​model may consist of an input layer, a hidden layer, and an output layer.

[0041] The input layer is the layer associated with the values ​​input into the AI ​​model. In the hidden layers, you can perform multiplication, accumulation, and activation operations on the input values ​​to create feature maps. A MAC operation is the process of multiplying the input value by its corresponding weight and then summing the results. An activation operation is the process of taking the result of the MAC operation and feeding it into an activation function, which can be of various forms. For example, activation functions may include, but are not limited to, the sigmoid function, the tangent function, the Lelu function, the Ricky Lelu function, the Maxout function, and the Eleutherococcus senticosus function.

[0042] A hidden layer can consist of at least one layer. For example, if it is divided into a first hidden layer and a second hidden layer, the first hidden layer can perform MAC and activation operations based on the input of the input layer to generate a feature map, which can be used as the input of the second hidden layer. The second hidden layer can perform MAC and activation operations based on the feature map generated by the first hidden layer.

[0043] The output layer can be a layer associated with the result of an operation performed in the hidden layer.

[0044] In one embodiment, the learning model learns commonly used syllable (letter) patterns in a given corpus to automatically identify boundaries between compound words and entity names, and generates an object information file for training by integrating object information from a first UI source with object information rendered in the browser. This learned object information file is used to generate data for training the deep learning network, and based on data received from various domains, the data is standardized in a uniform format according to one or more standardization methods specific to each domain. It then learns and infers data from specific domains, allowing you to determine what information is needed for standardization in that domain and perform post-processing on the data received from each domain.

[0045] The first UI source contains an XML file, and the object information file for training includes an input JSON file for learning features and an output JSON file used as label data during training. This output JSON file contains information about the HTML DOM tree, implemented according to Web standards. Each domain includes at least one of the following: Radio Access Network (RAN), Transport, or Core, and post-processing may include correlation functions.

[0046] Figure 4 This is the structure of the dental restoration manufacturing process processing system provided in the embodiments of the present invention. The system can be automated through artificial intelligence-based shadow guidance according to a single embodiment.

[0047] System 400 according to one embodiment may include processor 420 and memory 430, and some of the configurations shown may be omitted or replaced. System 400 according to one embodiment may be a server or a terminal. According to one embodiment, processor 420 is configured to perform operations or data processing related to the control and / or communication of each component of system 400. It may consist of one or more processors. Memory 430 may store information related to the methods described above, and may also store programs implementing the methods described above. Memory 430 may be volatile memory or non-volatile memory. Memory 430 may store various types of file data, and the stored file data may be updated according to the operation of processor 120.

[0048] In one embodiment, processor 420 can execute programs and control device 400. The code of the program executed by processor 420 can be stored in memory 430. Operations of processor 420 can be performed by loading instructions stored in memory 430. System 400 is connected to external devices (e.g., a personal computer or a network) and can exchange data.

[0049] System 400 according to one embodiment includes a processor 420 and a memory 430. System 400 according to one embodiment may be the server or terminal described above. Processor 420 may include... Figures 1 to 3 At least one of the aforementioned devices, or by means of Figures 1 to 3 At least one of the above methods may be executed. Memory 430 may store information related to the above methods, or it may store programs that implement the above methods. Memory 430 may be volatile memory or non-volatile memory.

[0050] System 400 according to one embodiment includes a processor 420 and a memory 430. System 400 according to one embodiment may be the server or terminal described above. Processor 420 may include... Figures 1 to 3 At least one of the aforementioned devices, or by means of Figures 1 to 3 At least one of the above methods may be executed. Memory 430 may store information related to the above methods, or it may store programs that implement the above methods. Memory 430 may be volatile memory or non-volatile memory.

[0051] Figure 5 This is a flowchart of a dental prosthesis manufacturing process that can be automated through an artificial intelligence-based shading guide, according to one embodiment of the present invention.

[0052] exist Figure 5 In the flowcharts, process steps, method steps, and algorithms are described sequentially, but these processes, methods, and algorithms can be configured to run in any suitable order. In other words, the steps of the processes, methods, and algorithms described in the various embodiments of the invention do not need to be executed in the order described in the invention. Furthermore, although some steps are described as being executed asynchronously, in other embodiments, some of these steps may be executed concurrently. Moreover, the illustration of the processes in the accompanying drawings does not imply that the example processes exclude other changes and modifications thereto, nor does it imply that any of the illustrated processes or their steps are essential to one or more of the various embodiments of the invention, nor does it imply that the example processes are preferred.

[0053] In operation 510, system 400 can receive digital images of a patient's teeth taken at a dental clinic. The input image can be a high-resolution digital image taken by a standard dental camera, and system 400 can automatically extract color features such as RGB color information, brightness, and saturation from the input image. This can serve as basic data for accurately analyzing the actual color of the patient's teeth.

[0054] According to one embodiment, system 400 can receive images of a patient's teeth from a high-resolution dental digital camera (e.g., at least 20 megapixels). The camera used is an ISO 9001 certified dental camera equipped with a macro lens to capture subtle color differences in the teeth and a polarizing filter to minimize surface reflections. System 400 is capable of distinguishing over 16.7 million colors by processing an input image with 24-bit color depth and can extract precise color values ​​from 0 to 255 for each RGB channel of each pixel. Furthermore, system 400 converts to an HSV (color, saturation, brightness) color space to measure color angle (0-360 degrees), saturation (0-100%), and brightness (0-100%). Additionally, system 400 quantifies the translucency of the teeth in five steps and can also analyze the optical properties of the enamel layer.

[0055] In operation 520, system 400 can use an AI model previously trained on 16 color levels from A1 to D4 in a classic color selection system. The AI ​​model, based on a convolutional neural network (CNN) architecture, calculates similarity by comparing and analyzing the color information of the extracted tooth with standard color data from a previously stored shading guide. System 400 can automatically determine the chromaticity value closest to the patient's tooth based on the calculated similarity, achieving a more objective and accurate color matching than existing visual comparison methods.

[0056] According to one embodiment, system 400 can use a deep learning-based color analysis model. This model, based on the ResNet-152 architecture, can be pre-trained using over 10,000 real tooth images and standard color data from classic color selection systems. System 400 encodes the color features extracted from each image into a 256-dimensional vector and calculates the similarity to standard shading values ​​between 0 and 1 using cosine similarity measurement. At this point, independent similarity analysis can be performed for each specific region of the tooth (section, center, cervical region), and shading values ​​with a similarity greater than 0.95 for each region can be automatically selected. System 400 can perform 5x validation to cross-validate the reliability of the selected shading values, potentially providing more than 30% higher accuracy than visual judgment.

[0057] In operation 530, system 400 can generate milling or additive manufacturing data for 3D printing required to manufacture dental prostheses based on determined chromaticity values. The generated machining data can be linked to computer-aided design / computer-aided systems to control the automated manufacturing process, and system 400 can optimize the operating parameters of the manufacturing equipment to achieve precise prosthesis manufacturing. Furthermore, system 400 can monitor quality in real time during manufacturing and automatically adjust manufacturing parameters as needed to ensure the production of high-quality prostheses.

[0058] According to one embodiment, system 400 can generate data for a computer-aided design / computer-aided manufacturing process based on selected shading values. For milling, system 400 generates a 3D model in a standard template library format and converts it into G-code to control the toolpath of a 5-axis milling machine. In this case, the toolpath can be optimized in increments of 0.01 mm, achieving high-quality machining with a surface finish of less than 0.2 μm Ra. In the case of 3D printing, system 400 can slice into layers of 50 micrometers in thickness to create an exposure pattern for each layer. System 400 can monitor tool load, spindle speed, feed rate, etc., in real time during manufacturing to maintain machining quality, and can control dimensional accuracy to within 20 micrometers using a laser displacement sensor. Furthermore, system 400 measures the color of the manufactured restoration using spectrophotometry and can automatically adjust manufacturing parameters to maintain a color difference from the target shading value at or below 1.5.

[0059] Figure 6 This is a flowchart of a dental prosthesis manufacturing process that can be automated through shadow guidance based on artificial intelligence, as provided in an embodiment of the present invention.

[0060] exist Figure 6 In the flowcharts, process steps, method steps, algorithms, etc., are described sequentially, but such processes, methods, and algorithms can be configured to run in any suitable order. In other words, the steps of the processes, methods, and algorithms described in the various embodiments of the present invention do not need to be executed in the order described in the present invention. Furthermore, although some steps are described as being executed asynchronously, in other embodiments, some of these steps may be executed concurrently. Moreover, the illustration of the processes in the accompanying drawings does not imply that the example processes exclude other changes and modifications thereto, nor does it imply that any of the illustrated processes or their steps are essential to one or more of the various embodiments of the present invention, nor does it imply that the example processes are preferred.

[0061] In operation 610, system 400 receives patients' personal information from the database and can set personalized color judgment criteria. By comprehensively analyzing tooth color characteristics by patient age group, color differences by gender, and tooth color statistics by ethnicity, personalized reference data can be generated. In this way, system 400 may be able to make more accurate color judgments taking into account the characteristics of each patient.

[0062] According to one embodiment, system 400 can receive patients' personal information from a database and set personalized color judgment criteria. The input information may include information about the patient's age, gender, and ethnicity, as well as lifestyle information that may affect tooth color, such as the frequency of smoking or coffee consumption. System 400 may default to a light tone (A3-A30) for 1-20 seconds, a medium tone (A3-A3) for 40-50 seconds, and a dark tone (A3-A4) for people over 60 years of age. Furthermore, system 400 may recommend a tone 1-2 levels lighter than the primary hue, reflecting the tendency for women to choose relatively lighter tones based on gender, and the possibility that, based on racial statistics, A-series tones for Asians and B-series tones for Westerners may be considered preferred.

[0063] In operation 620, system 400 can use a color measurement chart specifically designed for standardized color measurement. This chart can contain 16 levels of tooth patina from A1 to D4 and 18 levels of grayscale patina, divided in 5% increments from 0% to 100%. System 400 can obtain reference points by photographing this chart along with the teeth to compensate for color distortion caused by the imaging environment.

[0064] According to one embodiment, system 400 is capable of using a standard color measurement chart that meets the ISO 13485 standard. This chart includes a 16-color classic swatch guide with 16 color levels from A1 to D4, widely used in the manufacture of dental prostheses. Each swatch is 8 mm x 12 mm in size, allowing for accurate color measurements. Grayscale swatches can be used as a standard for light intensity and luminance correction, divided into 18 levels in 5% increments from 0% to 100%. System 400 recognizes the QR code embedded in each swatch and can automatically input the correct reference value. It can also automatically correct for geometric distortions in the image via alignment marks located in the corners of the chart.

[0065] In operation 630, system 400 can mathematically analyze the difference between the actual measured values ​​and known reference values ​​of the captured standard image. In this way, the degree of color distortion caused by lighting conditions, camera settings, etc., during shooting can be quantified, and coefficients can be automatically calculated for compensation. System 400 can then apply the calculated correction coefficients to the actual tooth image and automatically convert it to an accurate tooth color under standard lighting conditions.

[0066] According to one embodiment, system 400 can derive light intensity correction factors by analyzing the gamma curves of grayscale patches to compensate for distortions caused by lighting conditions, and can also automatically compensate for color temperature deviations caused by camera white balance settings. System 400 integrates these correction factors into a 3x3 color conversion matrix and applies them to each pixel of an actual tooth image, thereby enabling accurate conversion to the actual tooth color observed in a standard D65 lighting environment of 6500K.

[0067] In operation 640, system 400 can automatically segment and analyze tooth regions using state-of-the-art deep learning techniques (convolutional neural networks). The segmented tooth regions can be further subdivided into key regions, such as incision, midsection, and cervical spine regions, so that the color characteristics of each region can be measured independently. System 400 generates a color profile for each region by synthesizing measured RGB color values, brightness, saturation, and transparency, and compares it with standard shading guide information stored in a database to automatically select the optimal shading value.

[0068] According to one embodiment, system 400 can perform precise pixel-by-pixel segmentation of tooth regions using a convolutional neural network based on a deep residual network-50 structure. Based on the anatomical characteristics of the teeth, the segmented tooth regions can be automatically subdivided into the cutting surface (upper 30%), middle (middle 40%), and cervix (lower 30%). The average and standard deviation of the RGB color values ​​for each region are calculated, and luminance (0-100), saturation (0-128), and color vision (0-360 degrees) in the color space can be measured. Furthermore, system 400 can quantify and record the sharpness of the teeth within a range of 0-5, and select the closest shading value for each region by calculating the color difference with 16 standard colors from a classic color selection system. If the color difference value is less than 1.5, it can be judged as a clinically indistinguishable level, and the corresponding chromaticity value can be selected.

[0069] Figure 7 This is a flowchart of a dental prosthesis manufacturing process that can be automated through shadow guidance based on artificial intelligence, as provided in an embodiment of the present invention.

[0070] exist Figure 7In the flowcharts, process steps, method steps, and algorithms are described sequentially, but these processes, methods, and algorithms can be configured to run in any suitable order. In other words, the steps of the processes, methods, and algorithms described in the various embodiments of the invention do not need to be executed in the order described in the invention. Furthermore, although some steps are described as being executed asynchronously, in other embodiments, some of these steps may be executed concurrently. Moreover, the illustration of processes in the accompanying drawings does not imply that the example processes exclude other changes and modifications thereto, nor does it imply that any of the illustrated processes or their steps are essential to one or more of the various embodiments of the invention, nor does it imply that the example processes are preferred.

[0071] In operation 710, system 400 receives order data, including patient tooth number, treatment plan, restoration type and material information, through a web browser-based order management interface. It automatically records the color value corresponding to each tooth position in the database by matching the tooth count information contained in the order data with the captured tooth images.

[0072] In operation 720, system 400 is able to generate 3D data for milling and manufacturing process work instructions based on the recorded chromaticity values, repair type, and material information.

[0073] System 400 can generate precise manufacturing process data based on recorded chromaticity values ​​and restoration-related information. This may include 3D machining data from a 5-axis milling machine, as well as detailed manufacturing process instructions from the operator. The work order can specify in detail the exact hue number and size of the zirconia block to be used, the milling machine toolpath coordinates, sintering temperature conditions, sintering time for each stage, and the 16 color levels of the coloring process.

[0074] In operation 730, system 400 can measure milling accuracy, sintering temperature and coloring results in real time through manufacturing process status monitoring sensors, and record them periodically in the quality control database.

[0075] System 400 can monitor the entire manufacturing process in real time via sensors. The precision of the grinding process, the temperature changes during sintering, and the results of the coloring process can be continuously measured and automatically recorded in the quality control database. This ensures optimal quality control at every stage of the manufacturing process.

[0076] In operation 740, system 400 can register quality control data in the blockchain distributed ledger and apply a hash function hash algorithm to prevent the forgery and falsification of manufacturing history.

[0077] According to one embodiment, system 400 can securely manage manufacturing history data by utilizing blockchain technology that employs a highly secure hash function hash algorithm. Each manufacturing process data point can be recorded on a distributed ledger and stored in a tamper-proof format, providing a reliable foundation for product quality assurance and traceability.

[0078] In operation 750, system 400 uses a deep learning-based image processing algorithm to analyze the color difference between the final color and the initial shading value of the finished restoration to determine whether the error range is less than or equal to the standard color difference standard 1.5. If the standard is met, a completion report containing 3D scan data, color analysis report, manufacturing history information and restoration tooth transport tracking number can be generated and registered in PDF format in the database of the order management system.

[0079] According to one embodiment, system 400 can utilize advanced image processing technology to verify the quality of completed dental restorations. Through deep learning-based image processing algorithms, the color difference between the final restoration's color and the initially set shading value can be precisely analyzed, ensuring that the difference is within a standard allowable range of 1.5 or less. For restorations that have passed quality verification, the 3D scan data can generate a comprehensive completion report in PDF format, including a detailed color analysis report, complete manufacturing history information, and shipping tracking information, and be registered in the management system.

[0080] According to one embodiment, system 400 can determine the prosthesis manufacturing method based on the shade value determined above, and in the case of a single-color shade, perform 5-axis milling operation using a zirconia block of that color, and in the case of a mixture of multiple colors, use data loss-protected 3D printing additive manufacturing with ceramic powder containing photocurable binder.

[0081] According to one embodiment, system 400 applies primer as an initial adhesive to a metal frame at a specified thickness, prints a first laminate mixed with recyclable adhesive on the primer layer at a specified first thickness in a 7:3 ratio, prints main component ceramic powder on the primer layer, and can print a second laminate on the first laminate, selectively applying ceramic powder of different colors to achieve color in an area of ​​a second thickness that is relatively thinner than the first thickness.

[0082] According to one embodiment, the system 400 performs preliminary curing with a specified range of UV LEDs for 10 seconds immediately after each laminate is printed. After the entire lamination is completed, a two-step sintering process is performed, which is maintained at 800 degrees for 1 hour and at 1200 degrees for 2 hours. The system can also select the color and formulate the ceramic powder according to the determined color value and the ceramic powder formulation database.

[0083] In one embodiment, system 400 can acquire three-dimensional oral data of a patient via an intraoral scanner. The intraoral scanner employs high-resolution optical scanning technology to scan the shape of teeth and gums with micron-level precision. System 400 can automatically identify defective areas of teeth by applying a deep learning algorithm with a convolutional neural network architecture for image segmentation to the acquired three-dimensional data. At this point, the precise boundaries of the defective areas can be identified using an AI model that learns the morphological characteristics of normal teeth. System 400 can automatically generate a three-dimensional model of a metal substructure with a natural tooth shape by analyzing the shape of the gingival line and adjacent teeth around the identified defective areas. When creating the metal substructure model, the height, lateral curvature, and edge shape of the occlusal surface can be designed in harmony with the surrounding teeth.

[0084] In another embodiment, system 400 can perform a special surface treatment on the connection surface between the metal cap and the zirconia restoration. A lattice-type porous pattern with a depth of 50–200 micrometers can be created on the mating surface. This pattern can include microscopic inhomogeneities to maximize mechanical adhesion. System 400 can verify the structural stability of a 3D model generated by an AI algorithm based on finite element analysis. In this process, stress dispersion under occlusal pressures up to 900 N, shear bond strength on the connection surface, and resistance to fracture under lateral forces can be comprehensively analyzed. Based on a validated metal cap model, system 400 can progressively process dental soft metals (such as Co-Cr alloys). The sintering process can be performed in three stages: preheating at 400°C, main sintering at 900°C, and final sintering at 1100°C. Precise temperature control optimizes the microstructure of the metal at each stage.

[0085] In an additional embodiment, System 400 can precision machine zirconia blocks based on a validated restoration model. Complex tooth geometries can be precisely achieved using a 5-axis milling machine, and color uniformity can be improved by using a specialized milling tip coated with ceramic powder tailored to chromaticity values. System 400 allows the use of a special ceramic adhesive mixed with an inorganic colorant at a ratio of 7-12 wt% to minimize color deviations when bonding the metal inner cap and the zirconia restoration. The shape of the finished prosthesis can be inspected using a precision 3D scanner, and color uniformity can be measured using a spectrophotometer. All measured quality data can be automatically recorded in a manufacturing history management system and used as a basis for future quality improvements.

[0086] Figure 8 This is the process of creating a dental restoration through a dental restoration manufacturing process processing system provided in an embodiment of the present invention, which can automate artificial intelligence-based shadow guidance according to one embodiment.

[0087] According to one embodiment, system 400 can use the color information of the shadow guide as a database and as reference data for training an AI model, and can use a spectrophotometer to measure the exact optical properties of each color sample to construct quantified data.

[0088] According to one embodiment, the color guide contains a total of 16 standardized color samples from A1 to D4, each sample being made in the same shape and size as an actual tooth so that a natural comparison can be made. The base of each sample is clearly engraved with a color code (A1, A2, B1, B2, etc.) so that it can be easily identified by healthcare professionals, and the system 400 can automatically identify this code and match it with a database.

[0089] According to one embodiment, system 400 can manage each color of the shading guide by classifying them in the CIE Lab* color space, as follows: Group A is reddish-brown (A1-A4), with L values ​​ranging from 65 to 80, A values ​​from 1 to 4, and B values ​​from 12 to 18, likely the most common hue in natural teeth. Group B is yellow (B1-B4), distributed in the range of L values ​​70-85, A values ​​0-2, and B values ​​15-20, representing a bright, transparent hue primarily observed in the teeth of younger patients. Group C is gray (C1-C4), with L values ​​ranging from 60-75, A values ​​0-2, and B values ​​8-14, likely representing a cloudy hue primarily seen in the teeth of older patients. Group D is a series of reddish-gray (D2-D4), distributed in the range of L values ​​55-70, A values ​​2-5, and B values ​​10-15, where darker hues may be observed in special cases.

[0090] According to one embodiment, system 400 can independently measure the color and transparency of each shaded sample by cut, center, and cervical region to construct a location-specific color database. In this case, the cut surface exhibits relatively high transparency (0.7-0.9) and low saturation, while the cervix may exhibit low transparency (0.2-0.4) and high saturation. System 400 can reflect these region-specific characteristics in an artificial intelligence model to achieve more complex color matching.

[0091] According to one embodiment, system 400 is able to select a color that most closely resembles the patient's natural teeth with over 95% accuracy, using standardized hue guidelines and a precise color database. This process can greatly contribute to improving the aesthetics of the final restoration. Furthermore, system 400 continuously updates its artificial intelligence model based on accumulated clinical data and has the potential to recommend the optimal hue, taking into account individual characteristics and preferences.

[0092] According to one embodiment, the system 400 automatically identifies the defect location by applying a deep learning-based tooth region segmentation algorithm to the patient's three-dimensional oral data acquired by an intraoral scanner, analyzes the gingival shape and anatomical shape information of the identified defect location and the surrounding teeth, generates a three-dimensional model of a metal crown that resembles a natural tooth, and can generate a three-dimensional model of a zirconia restoration with enhanced aesthetics based on the generated metal crown model and the aforementioned determined shadow values.

[0093] According to one embodiment, the system 400 automatically generates a lattice-type porous surface pattern and non-uniform structure of a specified depth on the connection surface between the metal cap and the zirconia restoration to improve the bonding strength. The system also uses an artificial intelligence-based stress analysis algorithm to simulate the occlusal pressure dispersion, bonding strength, and fracture resistance of the generated three-dimensional model to verify the structural stability. Based on the verified metal cap model, a metal cap with enhanced durability can be produced by gradually sintering dental soft metal at a specified temperature.

[0094] According to one embodiment, system 400 uses a validated prosthetic model to machine a zirconia block using a multi-axis milling machine, and improves color uniformity by using a specialized milling tip with a heat-treated coating of ceramic powder corresponding to the aforementioned determined shading values. When joining the manufactured metal crown and zirconia prosthesis, a ceramic binder mixed with an inorganic colorant corresponding to the aforementioned determined shading values ​​is used to correct color deviations, and the shape accuracy and color uniformity of the finished prosthesis are measured using a 3D scanner and spectrophotometer. The measurement data can be recorded in a database and controlled to manage the quality history.

[0095] In one embodiment, system 400 can perform a three-dimensional scan of a patient's oral cavity using a high-resolution intraoral scanner. The intraoral scanner used in this case can digitize the shape of teeth and gums with an accuracy of less than 20 micrometers, and in particular, it can also acquire color information. System 400 can process the acquired three-dimensional data in real time, converting it into point cloud data and reconstructing it into a mesh-like three-dimensional model. System 400 can automatically segment and identify defects by applying a deep learning algorithm based on the U-Net architecture to the generated three-dimensional model. This algorithm, pre-trained using data from over 10,000 dental scans, can identify defects with an accuracy exceeding 95%. System 400 can comprehensively analyze the gingival line around the identified defect site, the occlusal relationship with the teeth, and the morphological features of adjacent teeth. Based on this analysis, system 400 can automatically design a three-dimensional model of a metal inner cap that resembles the appearance of natural teeth, and can generate a three-dimensional model of the final zirconia restoration by applying selected shading values ​​to the designed model.

[0096] In one embodiment, system 400 can perform a special surface treatment on the mating surfaces between the metal top plate and the zirconia prosthesis. Specifically, a lattice pattern 100 to 150 micrometers deep can be arranged at 0.5 mm intervals on the mating surfaces, and a microporous structure of 30 to 50 micrometers in size can be formed within each lattice. This surface treatment can increase the mechanical bond strength of the adhesive by up to 40%. System 400 can perform ANSYS-based finite element analysis to verify the structural stability of the design model. During this process, interlocking forces up to 1000 N can be used to analyze the stress distribution relative to lateral forces and determine whether shear stress (especially shear stress at the joint surface) remains below 50 MPa. System 400 can use a cobalt-chromium alloy to manufacture a proven metal inner cap. The sintering process can be carried out in three stages: initial heating at 450 degrees Celsius for 30 minutes, primary sintering at 950 degrees Celsius for 40 minutes, and final sintering at 1150 degrees Celsius for 20 minutes.

[0097] In one embodiment, System 400 can machine zirconia blocks using a 5-axis synchronously controlled milling machine. During milling, a ceramic-coated milling tip specifically designed for the selected chromaticity value can be used. This milling tip is coated with nanoscale ceramic powder, allowing fine ceramic particles to penetrate the zirconia surface during machining, thereby improving color uniformity. System 400 is used to bond the manufactured metal crown and zirconia prosthesis, using a special ceramic adhesive mixed with 10 wt% inorganic colorant, depending on the selected chromaticity value. This adhesive does not change color even during firing at 1500 degrees Celsius, ensuring the color stability of the final prosthesis. System 400 can use a blue light 3D scanner to measure shape accuracy and use spectrophotometry to assess color uniformity in the International Commission on Illumination (ICI) color system color space to check the quality of the finished restoration. All measurement data is automatically recorded in an SQL database for future quality control and improvement.

[0098] The examples above can be implemented using hardware and software components or a combination thereof. For instance, the devices, methods, and components mentioned can be implemented using general-purpose or special-purpose computers, such as processors, controllers, arithmetic logic units, digital signal processors, microcomputers, field-programmable gate arrays, programmable logic units, microprocessors, or other devices capable of executing and responding to instructions. A processing unit is an operating system and one or more software applications running on that operating system. Furthermore, the device can perform functions such as accessing, storing, manipulating, processing, and generating data, depending on the execution of the software. For ease of understanding, a processing unit can be described as a single processing unit, but someone with normal knowledge of this field will recognize that it may contain multiple processing elements or different types of processing elements. For example, a processing unit may consist of multiple processors or one processor and one controller. You can also use other processing configurations, such as parallel processors.

[0099] This software may include computer programs, code, instructions, or combinations thereof, through which the processing unit can be configured independently or collectively in any desired manner or command. The software and data may be interpreted by the processing unit and may be implemented permanently or temporarily on various machines, components, physical or virtual devices, computer storage media, or transmission signals to provide instructions and data. This software may be distributed, stored, or executed on networked computer systems, and may be stored on one or more computer-readable recording media.

[0100] The examples above are illustrated with specific drawings, but experts in the field can apply various technical modifications and variations based on them. For example, appropriate results may be obtained even if the described techniques are performed differently in the order they are presented, or if the components of the described system, structure, device, or circuit are combined in different forms or replaced by different components or equivalents.

Claims

1. A manufacturing process and system for automated color guide restorations based on artificial intelligence, characterized in that, The AI-powered automated dental restoration manufacturing system relies on memory-stored instructions, executed by a processor. When the processor runs, the instruction execution flow is as follows: First, it receives an image of the patient's teeth; then, it accurately extracts the color information of the teeth from the input image. Next, using a pre-trained AI model, it compares and analyzes the extracted color information with previously stored shading guide color data to automatically determine the closest shading value. Based on the determined shading value, it generates processing data for manufacturing the dental restoration. Finally, according to the generated processing data, the control system operates the dental restoration manufacturing device to produce the corresponding color dental restoration.

Images