Dental disease area detection method based on image segmentation and reconstruction algorithm

Abstract

The invention provides a dental disease area detection method based on an image segmentation and reconstruction algorithm, and the method comprises the steps: inputting an original image, carrying out the area sampling of the original image through a target detection algorithm, and generating N detection frames; the detection frames with low lesion probability are filtered out through threshold suppression processing; the overlapping degree of the detection frames is obtained through non-maximum suppression processing, and whether the overlapping degree of the detection frames exceeds a suppression threshold value or not is judged; segmenting the reserved n detection frames through a segmentation algorithm to obtain n Boolean matrixes belonging to the focus area; reconstructing the n Boolean matrixes according to the regional information and generating a discrete array; and analyzing the pixel position represented by the discrete array and the lesion type of the position according to the value of the discrete array. According to the invention, by predicting the high-probability lesion region in the image and carrying out refined segmentation and reconstruction on the high-probability lesion region, the effects of reducing wrong segmentation and noisy points, avoiding interference of irrelevant regions in the image and improving the accuracy and overall performance of lesion region segmentation are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for detecting dental diseased areas based on image segmentation and reconstruction algorithms. Background Art

[0002] With the rapid development of artificial intelligence technology in the medical industry, oral diseases can also be analyzed by analyzing oral imaging data, such as panoramic films, CBCT images, and intraoral photos, using AI algorithms to automatically identify oral problems in a short period of time, thereby assisting dentists in making accurate diagnoses, greatly improving diagnostic efficiency while reducing errors caused by human factors.

[0003] The existing solution for processing oral images is usually to use a deep learning image segmentation algorithm and input the complete image into a convolutional neural network similar to the U-Net structure, which extracts feature information through a series of image processing steps. Specifically, the input oral image is first passed through a convolution layer to extract low-level features, and then the batch normalization layer is used to balance the data distribution, and the activation function is nonlinearly transformed to enhance the network expression ability. Then, the downsampling layer is used to reduce the size of the feature map and extract higher-level features. In order to avoid the overfitting problem in deep learning, a Dropout layer is inserted to reduce overfitting by randomly discarding neurons to improve the generalization ability of the model. Then, these high-dimensional feature maps are further processed, and the convolution layer, batch normalization layer, and activation function enhancement are successively performed, and then the size of the feature map is restored by the upsampling layer to make it consistent with the original image, and finally a segmentation result of the same size as the input image is generated.

[0004] However, the above-mentioned scheme based on equal pixel-by-pixel image processing has some obvious shortcomings. First, since the convolutional neural network treats every pixel of each image equally in the above process, it is easy to cause mis-segmentation in the prediction results, that is, pixels that do not belong to the target area are mistakenly classified as target objects. In addition, noise points in the image, such as low-quality or blurred areas, may be mistaken for part of the target area without distinction, which leads to a large number of noise points and errors in the final segmentation results. In summary, the robustness and accuracy of the existing technology when dealing with image processing in more complex or changing scenes still need to be improved.

[0005] Therefore, there is an urgent need for a dental disease area detection method that can selectively process key information in the image and ignore irrelevant areas when dealing with complex or changing scenes, thereby helping to reduce missegmentation and noise and improve overall performance. Summary of the invention

[0006] In order to solve the common problems in the prior art, the purpose of the present invention is to provide a method for detecting dental diseased areas based on image segmentation and reconstruction algorithm. The invention evaluates and processes the diseased area in advance, so as to selectively process key information in the image and ignore irrelevant areas during image processing, thereby reducing mis-segmentation and noise, avoiding interference from irrelevant areas in the image, and improving the accuracy of diseased area segmentation and the overall performance.

[0007] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0008] A method for detecting dental diseased areas based on image segmentation and reconstruction algorithm, comprising:

[0009] S1: Input an original image, perform region sampling on the original image through a target detection algorithm, and generate N detection frames, wherein the N detection frames contain region information of all suspected diseased regions in the original image.

[0010] S2: Using threshold suppression processing to filter out the detection frame with a low probability of lesions.

[0011] S3: Setting a suppression threshold, obtaining the overlap of the detection frame through non-maximum suppression processing, and determining whether the overlap of the detection frame exceeds the suppression threshold. If so, filtering out the detection frame; if not, retaining the detection frame.

[0012] S4: Segment the n detection frames that are retained using a segmentation algorithm, and determine whether the pixels in the segmented area belong to the lesion area, thereby obtaining n Boolean matrices belonging to the lesion area.

[0013] S5: Reconstruct the n Boolean matrices according to the region information to generate a discrete array with the same size as the original image.

[0014] S6: Analyze the pixel position represented by the value of the discrete array and the lesion type at the position.

[0015] According to the tooth disease area detection method based on image segmentation and reconstruction algorithm provided by the invention, image preprocessing is performed before inputting the original image, and the processing process includes:

[0016] S11: Read the original image, and perform pixel normalization and data enhancement processing on the original image.

[0017] S12: Input the processed original image into a neural network, and process the original image into a predicted data format through the neural network and then output it.

[0018] According to the invention, a method for detecting dental diseased areas based on image segmentation and reconstruction algorithm is provided. In step S1, the target detection algorithm adopts a region proposal network, including:

[0019] S13: Generate a candidate box containing all region information of the original image through a region proposal network.

[0020] S14: Extract features and classify each candidate box.

[0021] S15: Mapping the candidate frame to a position and size close to the real detection frame through frame regression to predict the position and size of the detection frame.

[0022] According to the invention, a method for detecting dental diseased areas based on image segmentation and reconstruction algorithm is provided. In step S1, the target detection algorithm adopts selective search, including:

[0023] S131: performing region segmentation on the original image.

[0024] S141: Create a bounding box for the segmented area.

[0025] S151: Setting a search rule, and merging similar regions according to the search rule to obtain the detection frame.

[0026] According to the invention, a method for detecting dental diseased areas based on image segmentation and reconstruction algorithm is provided, and the threshold suppression processing method in step S2 includes:

[0027] S21: Setting a minimum threshold value according to the image content of the dental disease.

[0028] S22: Compare the pixels in the detection frame with the minimum threshold, retain the detection frames that are greater than or equal to the minimum threshold, and filter out the detection frames that are less than the minimum threshold.

[0029] S23: Repeat step S22 until all N detection frames are processed.

[0030] According to the invention, a method for detecting dental diseased areas based on image segmentation and reconstruction algorithm is provided, and the non-maximum suppression processing method in step 3 includes:

[0031] S31: Calculate the confidence of each detection box.

[0032] S32: Sorting the detection boxes by scores according to the confidence values.

[0033] S33: Select the detection frame with the highest score, and calculate the overlap between the detection frame and the remaining detection frames.

[0034] S34: Determine whether the overlap is greater than the suppression threshold, if so, retain the detection frame, if not, set the score of the detection frame to 0.

[0035] S35: Repeat steps S33 to S34 until all detection frames are processed.

[0036] According to the dental disease area detection method based on image segmentation and reconstruction algorithm provided by the invention, the confidence calculation method in step S31 includes:

[0037] S311: Obtaining the probability of the lesion existing in the detection frame.

[0038] S312: Calculate the overlap between the detection frame and the real detection frame using the YOLO algorithm.

[0039] S313: Estimate the confidence of the detection frame according to the maximum overlap.

[0040] According to the dental disease area detection method based on image segmentation and reconstruction algorithm provided by the invention, after two suppression processes of threshold suppression and non-maximum suppression, the number of the detection frames remaining is n, which is consistent with the number of the real target areas.

[0041] According to the invention, a method for detecting dental diseased areas based on an image segmentation and reconstruction algorithm is provided, wherein the segmentation algorithm comprises:

[0042] S41: performing binary segmentation on the area of ​​the detection frame.

[0043] S42: Setting the lesion area threshold.

[0044] S43: Compare the pixels of the segmented area with the lesion area threshold, if the pixel is greater than or equal to the lesion area threshold, it means that the pixel belongs to the lesion area, and set the pixel as True. If the pixel is less than the lesion area threshold, set the pixel as False.

[0045] S44: Pixels with the value of Ture form a two-dimensional array.

[0046] S45: Repeat steps S43 to S44 until all segmented regions are compared and n Boolean matrices are obtained.

[0047] According to the method for detecting dental diseased areas based on image segmentation and reconstruction algorithm provided by the invention, step S6 further includes:

[0048] S61: Define a set of color palettes corresponding to different lesion types.

[0049] S62: Map the discrete array into an overlay layer of corresponding color according to the color palette.

[0050] S63: Add the pixel values ​​of the lesion area to the overlay layer in proportion, so as to analyze the area where the lesion is located through the displayed color.

[0051] It can be seen that compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. The present invention pre-processes the image and evaluates the location of the lesion area through a specific algorithm or strategy. Compared with traditional image processing solutions, the present invention avoids interference with irrelevant areas in the entire image through priority screening, thereby improving the accuracy of lesion area segmentation.

[0053] 2. The present invention locates the lesion area through efficient segmentation of local areas, and then divides it into multiple sub-areas for more refined segmentation processing. Compared with traditional image processing solutions, the segmentation strategy of the present invention enables the model to focus on the details of key areas, thereby reducing the impact of mis-segmentation and noise points.

[0054] 3. The present invention reconstructs and integrates the segmentation results, splices the segmentation results of multiple sub-regions back to the original image, and realizes the visualization of the overall prediction results. Compared with traditional image processing solutions, the present invention ensures segmentation accuracy and global consistency through image reconstruction.

[0055] 4. The present invention uses a dental disease area detection method based on an image segmentation and reconstruction algorithm to efficiently segment and detect common dental diseases, thereby being able to achieve a preliminary screening function for dental diseases.

[0056] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flow chart of an embodiment of a method for detecting dental diseased areas based on an image segmentation and reconstruction algorithm of the present invention.

[0058] Figure 2 It is an image processing process diagram of steps S1 to S3 of an embodiment of a method for detecting dental diseased areas based on an image segmentation and reconstruction algorithm of the present invention.

[0059] Figure 3 It is an image processing process diagram of S4 to S5 in an embodiment of a method for detecting dental diseased areas based on an image segmentation and reconstruction algorithm of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0062] See also Figure 1-3 The present invention provides a method for detecting a dental disease area based on an image segmentation and reconstruction algorithm, comprising:

[0063] S1: Input an original image, perform region sampling on the original image through a target detection algorithm, and generate N detection frames, wherein the N detection frames contain region information of all suspected diseased regions in the original image.

[0064] S2: Using threshold suppression processing to filter out the detection frame with a low probability of lesions.

[0065] S3: Setting a suppression threshold, obtaining the overlap of the detection frame through non-maximum suppression processing, and determining whether the overlap of the detection frame exceeds the suppression threshold. If so, filtering out the detection frame; if not, retaining the detection frame.

[0066] S4: Segment the n detection frames that are retained using a segmentation algorithm, and determine whether the pixels in the segmented area belong to the lesion area, thereby obtaining n Boolean matrices belonging to the lesion area.

[0067] S5: Reconstruct the n Boolean matrices according to the region information to generate a discrete array with the same size as the original image.

[0068] S6: Analyze the pixel position represented by the discrete array and the type of lesion at the position according to the value of the discrete array. In this embodiment, image preprocessing is performed before inputting the original image, and the processing process includes:

[0069] S11: Read the original image, and perform pixel normalization and data enhancement processing on the original image.

[0070] S12: Input the processed original image into a neural network, and process the original image into a predicted data format through the neural network and then output it.

[0071] Specifically, the pixel value normalization described in this embodiment is used to map the pixel values ​​within the interval [0, 1] so that the pixel value changes in the image maintain a relatively stable proportional relationship within a certain range, so as to improve the numerical stability and convergence speed of the image processing algorithm.

[0072] Specifically, the pixel value normalization described in this embodiment adopts a maximum and minimum value normalization method, and the process includes:

[0073] S111: Obtaining the minimum and maximum values ​​of all pixels in the original image;

[0074] S112: Calculate the difference between the value of each pixel in the original image and the minimum value;

[0075] S113: Compare the difference with the difference between the maximum value and the minimum value, so as to obtain a mapping of the pixel value to the interval [0, 1].

[0076] Specifically, the data enhancement processing of the image in step S11 of this embodiment includes image dimension adjustment, dimension swapping, scaling and format conversion, etc., which are used to improve the diversity of image data, thereby improving the generalization ability of the image data model. The above-mentioned image enhancement processing scheme is only exemplary and not the only processing method. For example, it also includes image contrast adjustment, cropping, filtering, etc., so as to achieve the following effects according to the actual dental image: (1) Eliminate background interference of the image and improve the clarity and contrast of the image; (2) Enhance image features, thereby reducing the consumption of computing resources and reducing the complexity of subsequent processing; (3) Remove irrelevant parts, focus on key areas, and adjust the image size to adapt to the application scenarios and algorithm requirements of dental diagnosis; (4) Enhance the detectability of information about the lesion area in the original image, and improve the reliability of feature extraction, image segmentation, matching and recognition.

[0077] Specifically, the prediction data format of the graph neural network in step S12 of this embodiment is in the form of a multidimensional array, and its specific format depends on the color mode and dimension of the original image. For example, if the original image is a color image, it adopts a three-dimensional array format, and the representation of the array is: (H, W, C), where H is the height of the original image, W is the width of the original image, and C is the number of channels of the original image. If the original image is an RGB image, C is 3.

[0078] Specifically, the target detection algorithm in step S1 of this embodiment adopts a deep learning convolutional neural network (CNN), which includes an R-CNN algorithm, a Fast R-CNN algorithm, and a Faster R-CNN algorithm. The target detection process using the Faster R-CNN algorithm in this embodiment includes:

[0079] S13: Introduce a region proposal network (RPN), and generate a candidate box containing all region information of the original image through the RPN.

[0080] S14: Extract features and classify each candidate box.

[0081] S15: Mapping the candidate frame to a position and size close to the real detection frame through frame regression to predict the position and size of the detection frame.

[0082] Specifically, this embodiment can quickly generate high-quality candidate boxes through the RPN network, thereby avoiding the tedious process of using external algorithms to generate candidate regions, thereby improving the speed and accuracy of target detection.

[0083] Specifically, another embodiment uses the R-CNN algorithm to detect the target, including:

[0084] S131: performing region segmentation on the original image.

[0085] S141: Create a bounding box for the segmented area.

[0086] S151: Setting a search rule, and merging similar regions according to the search rule to obtain the detection frame.

[0087] Specifically, another embodiment uses the Fast R-CNN algorithm to detect objects, including:

[0088] S132: Input the entire original image into the neural network and calculate its feature map;

[0089] S142: extracting corresponding candidate regions from the feature map and performing RoI pooling processing to obtain a feature vector of a fixed size;

[0090] S152: Perform category prediction and position prediction respectively according to the feature vector.

[0091] In this embodiment, the threshold suppression processing method in step S2 includes:

[0092] S21: Setting a minimum threshold value according to the image content of the dental disease.

[0093] S22: Compare the pixels in the detection frame with the minimum threshold, retain the detection frames that are greater than or equal to the minimum threshold, and filter out the detection frames that are less than the minimum threshold.

[0094] S23: Repeat step S22 until all N detection frames are processed.

[0095] Specifically, the setting of the minimum threshold in this embodiment is based on the data characteristics of the original image and combined with actual dental disease diagnosis applications. After multiple clinical trials and comparisons, multiple adjustments and optimizations are performed to determine the most appropriate threshold. By setting the minimum threshold, the characteristics of the dental disease area can be screened out, ensuring that the probability of the existence of lesions in all detection frames in the filtered target detection results is higher than the minimum threshold, which is conducive to more accurate dental disease diagnosis.

[0096] In this embodiment, the non-maximum suppression processing method in step 3 includes:

[0097] S31: Calculate the confidence of each detection box.

[0098] S32: Sort the detection boxes in descending order of scores according to the confidence values.

[0099] S33: Select the detection frame with the highest score, and calculate the overlap between the detection frame and the remaining detection frames.

[0100] S34: Determine whether the overlap is greater than the suppression threshold, if so, retain the detection frame, if not, set the score of the detection frame to 0.

[0101] S35: Repeat steps S33 to S34 until all detection frames are processed.

[0102] In this embodiment, the confidence calculation method in step S31 includes:

[0103] S311: Obtaining the probability of the lesion existing in the detection frame.

[0104] S312: Calculate the overlap between the detection frame and the real detection frame using the YOLO algorithm.

[0105] S313: Estimate the confidence of the detection frame according to the maximum overlap.

[0106] Specifically, in the target detection of this embodiment, step S3 also includes filtering the detection results by confidence, specifically setting a confidence threshold. Only when the confidence of the detection box is higher than this threshold, the detection result will be considered valid, so as to improve the accuracy and reliability of the detected target.

[0107] Specifically, in another embodiment, the non-maximum suppression processing method in step 3 includes:

[0108] S3-1: Set the IOU threshold of the detection box and sort it in descending order according to the category classification probability of the classifier.

[0109] S3-2: Starting from the maximum probability, determine in turn whether the overlap of the detection frame exceeds the IOU threshold. If so, suppress the detection frame, thereby completing the traversal of a detection frame;

[0110] S3-3: Find the detection frame with the highest probability among the remaining detection frames and repeat step S3-2 until all detection frames are processed.

[0111] Specifically, this embodiment uses a non-maximum suppression method to set the scores of detection frames that overlap in the same target area but have non-maximum scores to 0, thereby eliminating overlapping detection frames to achieve secondary suppression, improving the overall detection accuracy and reducing computational complexity, which helps to improve the speed and efficiency of target detection.

[0112] In this embodiment, after the two suppression processes of the threshold suppression and the non-maximum suppression, the number of the detection frames remaining is n, which is consistent with the number of the real target areas.

[0113] In this embodiment, the segmentation algorithm includes:

[0114] S41: performing binary segmentation on the area of ​​the detection frame.

[0115] S42: Setting the lesion area threshold.

[0116] S43: Compare the pixels of the segmented area with the lesion area threshold, if the pixel is greater than or equal to the lesion area threshold, it means that the pixel belongs to the lesion area, and set the pixel as True. If the pixel is less than the lesion area threshold, set the pixel as False.

[0117] S44: Pixels with the value of Ture form a two-dimensional array.

[0118] S45: Repeat steps S43 to S44 until all segmented regions are compared and n Boolean matrices are obtained, where the n Boolean matrices represent the specific shape of each lesion.

[0119] Specifically, the binarization segmentation in step S41 of this embodiment adopts a local binarization algorithm, including:

[0120] S411: Select a pixel point of the detection frame as the center, and dynamically calculate the threshold of the pixel point according to the grayscale mean and standard deviation of the current pixel point in the r*r neighborhood.

[0121] S412: Repeat step S411 and use the above r*r template to traverse the pixels of the detection frame.

[0122] S413: Perform binarization calculation using Sauvola algorithm formula.

[0123] Specifically, in step S43 of this embodiment, if the pixel value is True, it indicates that it is a target pixel, and if the pixel value is False, it indicates that it is a background pixel.

[0124] Specifically, this embodiment efficiently segments the local area and, after locating the lesion area, divides it into multiple sub-areas for more refined segmentation processing. This segmentation strategy enables the computational model to focus more on the details of key areas, thereby reducing the impact of mis-segmentation and noise points.

[0125] Specifically, the image reconstruction in step S5 of this embodiment includes:

[0126] S51: The coordinate information and disease category information of each region are obtained through the target detection algorithm.

[0127] S52: Reconstruct the n Boolean matrices back into the original image according to the above region information, and generate a discrete array with the same size as the original image.

[0128] Specifically, in the discrete array described in this embodiment, 0 represents that the current pixel is a background pixel, and other numbers represent the types of lesions, such as: 1 represents that the current pixel position is gingivitis, 2 represents that the current pixel position is caries, 3 represents that the current pixel is tartar, etc., wherein the types of dental lesions specifically include: gingivitis, caries, tartar, wear plane, deformed lingual pits, residual roots, missing teeth, tartar, wedge-shaped defects and other nine dental diseases.

[0129] Specifically, this embodiment reconstructs the image to rejoin the segmentation results of multiple sub-regions back to the original image, thereby realizing visualization of the overall prediction result and ensuring the segmentation accuracy and global consistency.

[0130] In this embodiment, step S6 further includes:

[0131] S61: Define a set of color palettes corresponding to different lesion types.

[0132] S62: Map the discrete array into an overlay layer of corresponding color according to the color palette.

[0133] S63: Add the pixel values ​​of the lesion area to the overlay layer in proportion, so as to analyze the area where the lesion is located through the displayed color.

[0134] Specifically, this embodiment takes an RGB color image as an example. First, a set of color palettes is customized, for example, 1 represents red and 2 represents green. The discrete array in the format of ori_H*ori_W is mapped to a color overlay layer according to the color palette, where different colors represent areas corresponding to different diseases. The pixel values ​​of the background area remain unchanged, and the pixel values ​​of the disease area are added with the overlay layer color according to a certain ratio, so that the disease area can be displayed in the original image with a customized color.

[0135] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0136] The above-mentioned embodiments are only preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantial changes and substitutions made by technicians in this field on the basis of the present invention shall fall within the scope of protection required by the present invention.

Claims

1. A method for detecting dental diseased areas based on image segmentation and reconstruction algorithm, characterized in that: include: S1: Input an original image, perform region sampling on the original image through a target detection algorithm, and generate N detection frames, wherein the N detection frames contain region information of all suspected diseased regions in the original image; S2: Using threshold suppression processing to filter out the detection frame with a low probability of lesions; S3: setting a suppression threshold, obtaining the overlap of the detection frame by non-maximum suppression processing, and determining whether the overlap of the detection frame exceeds the suppression threshold, if so, filtering out the detection frame; if not, retaining the detection frame; S4: segmenting the n detection frames retained by a segmentation algorithm, and determining whether pixels in the segmented area belong to the lesion area, thereby obtaining n Boolean matrices belonging to the lesion area; S5: reconstructing the n Boolean matrices according to the region information to generate a discrete array having the same size as the original image; S6: Analyze the pixel position represented by the value of the discrete array and the lesion type at the position.

2. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to claim 1, characterized in that: Image preprocessing is performed before inputting the original image, and the processing process includes: S11: reading the original image, and performing pixel normalization and data enhancement processing on the original image; S12: Input the processed original image into a neural network, and process the original image into a predicted data format through the neural network and then output it.

3. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to claim 1, characterized in that: The target detection algorithm in step S1 adopts a region proposal network, including: S13: Generate a candidate box containing all region information of the original image through a region proposal network; S14: extracting features from each candidate frame and classifying it; S15: Mapping the candidate frame to a position and size close to the real detection frame through frame regression to predict the position and size of the detection frame.

4. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to claim 1, characterized in that: The target detection algorithm in step S1 adopts selective search, including: S131: performing region segmentation on the original image; S141: Creating a bounding box for the segmented area; S151: Setting a search rule, and merging similar regions according to the search rule to obtain the detection frame.

5. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to claim 3 or 4, characterized in that: The threshold suppression processing method in step S2 includes: S21: setting a minimum threshold value according to the image content of the dental disease; S22: comparing pixels in the detection frame with the minimum threshold, retaining the detection frames that are greater than or equal to the minimum threshold, and filtering out the detection frames that are less than the minimum threshold; S23: Repeat step S22 until all N detection frames are processed.

6. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to claim 5, characterized in that: The non-maximum suppression processing method in step 3 includes: S31: Calculate the confidence of each detection frame; S32: Sorting the detection frames by scores according to the confidence values; S33: Select the detection frame with the highest score, and calculate the overlap between the detection frame and the remaining detection frames; S34: determining whether the overlap is greater than the suppression threshold, if so, retaining the detection frame; if not, setting the score of the detection frame to 0; S35: Repeat steps S33 to S34 until all detection frames are processed.

7. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to claim 6, characterized in that: The confidence calculation method in step S31 includes: S311: Obtaining the probability of the lesion existing in the detection frame; S312: Calculate the overlap between the detection frame and the real detection frame by using the YOLO algorithm; S313: Estimate the confidence of the detection frame according to the maximum overlap.

8. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to any one of claims 1 to 7, characterized in that: After the two suppression processes of the threshold suppression and the non-maximum suppression, the number of the detection frames remaining is n, which is consistent with the number of the real target areas.

9. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to claim 8, characterized in that: The segmentation algorithm includes: S41: performing binary segmentation on the area of ​​the detection frame; S42: Setting the lesion area threshold; S43: Compare the pixel of the segmented area with the lesion area threshold, if it is greater than or equal to the lesion area threshold, it means that it belongs to the lesion area, and then set the pixel to True; if it is less than, set it to False; S44: pixels whose values ​​are Ture form a two-dimensional array; S45: Repeat steps S43 to S44 until all segmented regions are compared and n Boolean matrices are obtained.

10. The method for detecting dental diseased areas based on image segmentation and reconstruction algorithm according to claim 9, characterized in that: Step S6 also includes: S61: Define a set of color palettes corresponding to different lesion types; S62: Mapping the discrete array into an overlay layer of corresponding color according to the color palette; S63: Add the pixel values ​​of the lesion area to the overlay layer in proportion, so as to analyze the area where the lesion is located through the displayed color.

Images