CN111292313B

CN111292313B - Dental filling quality evaluation method and device

Info

Publication number: CN111292313B
Application number: CN202010133411.3A
Authority: CN
Inventors: 彭刚; 晏峰; 王浩; 刘北村; 郑煊
Original assignee: Enshi Jingzhi Yiya Dental Hospital Co ltd
Current assignee: Enshi Jingzhi Yiya Dental Hospital Co ltd
Priority date: 2020-02-28
Filing date: 2020-02-28
Publication date: 2023-04-28
Anticipated expiration: 2040-02-28
Also published as: CN111292313A

Abstract

The invention belongs to the field of dental filling quality evaluation based on image technology, and discloses a dental filling quality evaluation method and device. The method comprises the following steps: according to the oral cavity picture to be tested, respectively obtaining the rectangular frame of the tooth filling position, the rectangular frame of the adjacent tooth position of the tooth to be filled, the single tooth filling picture, the adjacent tooth picture and the picture of the area filled on the tooth filling surface through a preset target detection convolutional neural network model, a preset single tooth segmentation convolutional neural network model and a preset area segmentation convolutional neural network model, respectively obtaining corresponding tooth filling information evaluation results according to tooth information, and carrying out tooth filling quality evaluation on the tooth filling information evaluation results according to tooth filling quality evaluation standards. By the method, rapid tooth filling quality evaluation is realized, and an accurate tooth filling quality evaluation result is obtained.

Description

Dental filling quality evaluation method and device

Technical Field

The invention relates to the field of dental filling quality evaluation based on image technology, in particular to a dental filling quality evaluation method and device.

Background

Currently, in the oral cavity dental filling industry, dental filling effects need to be evaluated for dental filling quality control. In the prior art, the dental filling quality evaluation is carried out by manually consulting dental filling images, and dentists need to score the tooth condition of each patient in each period of preoperative, intraoperative and postoperative so as to judge the dental filling effect. However, as patients continue to grow, medical images facing doctors continue to grow, and problems exposed by manual analysis are also increasingly prominent. There are the following disadvantages: the method comprises the following steps of (1) large image information data volume and low efficiency; (2) Due to visual fatigue, doctors are unavoidably making erroneous judgment; (3) The doctor mainly judges by experience, has strong subjectivity and lacks quantitative standard.

The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present invention and is not intended to represent an admission that the foregoing is prior art.

Disclosure of Invention

The invention mainly aims to provide a method and a device for evaluating the quality of filling teeth, and aims to solve the technical problems of low working efficiency and accuracy of the quality evaluation of filling teeth in the prior art.

In order to achieve the above object, the present invention provides a dental filling quality evaluation method, which comprises the steps of:

acquiring an oral cavity picture to be tested, and inputting the oral cavity picture to be tested into a preset target detection convolutional neural network model to obtain a position rectangular frame and a tooth position number corresponding to each tooth;

performing target matching on the tooth position numbers of the teeth to be filled according to the tooth position numbers to obtain a rectangular frame of the tooth position to be filled and a rectangular frame of the adjacent tooth position of the teeth to be filled;

obtaining a tooth position rectangular frame picture of the repaired tooth and an adjacent tooth position rectangular frame picture of the repaired tooth according to the tooth position rectangular frame of the repaired tooth and the adjacent tooth position rectangular frame of the repaired tooth;

inputting the rectangular frame picture of the tooth position of the filled tooth and the rectangular frame picture of the adjacent tooth position of the filled tooth into a preset single-tooth segmentation convolutional neural network model to obtain a single filled tooth picture and an adjacent tooth picture;

Inputting the single compensated tooth picture into a preset area segmentation convolutional neural network model to obtain a compensated area picture on the surface of the compensated tooth;

taking the oral cavity picture to be tested, the tooth position rectangular frame of the teeth to be filled, the adjacent tooth position rectangular frame of the teeth to be filled, the single tooth picture to be filled, the adjacent tooth picture and the area picture to be filled on the tooth surface to be filled as tooth filling information;

and obtaining a dental filling information evaluation result according to the dental filling information, and evaluating the dental filling quality of the dental filling information evaluation result according to a dental filling quality evaluation standard.

Preferably, before the step of obtaining the position rectangular frame and the tooth position number corresponding to each tooth, the step of obtaining the oral cavity picture to be tested, inputting the oral cavity picture to be tested into a preset target detection convolutional neural network model, further includes:

obtaining a tooth detection sample picture, and scaling the tooth detection sample picture according to a preset size to obtain an initial tooth detection sample picture;

training a preset target detection convolutional neural network according to the initial tooth detection sample picture, and establishing a preset target detection convolutional neural network model;

Dividing the initial tooth detection sample picture according to a user operation instruction to obtain a single tooth division sample picture;

training a preset single-tooth segmentation convolutional neural network according to the single-tooth segmentation sample picture, and establishing a preset single-tooth segmentation convolutional neural network model;

dividing the single-tooth divided sample picture according to a user instruction to obtain a region divided sample;

training a preset area segmentation convolutional neural network according to the area segmentation sample, and establishing a preset area segmentation convolutional neural network model.

Preferably, the dental filling information comprises the oral cavity picture to be tested;

the step of obtaining the dental filling information evaluation result according to the dental filling information and evaluating the dental filling quality of the dental filling information evaluation result according to the dental filling quality evaluation standard comprises the following steps:

acquiring a picture format, a picture proportion and a resolution corresponding to the oral cavity picture to be tested;

detecting whether the picture format corresponding to the oral cavity picture to be tested accords with a preset picture format or not, and obtaining a picture format detection result;

according to the dental filling quality evaluation standard, performing dental filling quality evaluation on the picture format detection result;

detecting whether the picture proportion corresponding to the oral cavity picture to be tested accords with a preset picture proportion or not, and obtaining a picture proportion detection result;

According to the dental filling quality evaluation standard, performing dental filling quality evaluation on the picture proportion detection result;

detecting whether the resolution corresponding to the oral cavity picture to be tested accords with the preset picture resolution or not, and obtaining a picture resolution detection result;

and evaluating the dental filling quality of the picture resolution detection result according to the dental filling quality evaluation standard.

Preferably, the tooth filling information comprises a rectangular frame of the tooth filling position of the filled tooth and a rectangular frame of the adjacent tooth position of the filled tooth;

calculating the center coordinates of the teeth to be filled according to the rectangular frame of the tooth position to be filled;

detecting the offset corresponding to the center coordinates of the teeth to be filled, and obtaining an offset detection result;

according to the dental filling quality evaluation standard, performing dental filling quality evaluation on the deviation detection result;

calculating the area of the repaired teeth and the adjacent teeth according to the rectangular frame of the repaired teeth and the rectangular frame of the adjacent teeth;

calculating the area proportion of the picture according to the areas of the compensated teeth and the adjacent teeth;

Detecting whether the area proportion of the picture meets a preset area threshold proportion or not, and obtaining a picture area proportion detection result;

according to the dental filling quality evaluation standard, performing dental filling quality evaluation on the image area proportion detection result;

calculating a rectangular frame center point according to the rectangular frame of the tooth position of the repaired tooth and the rectangular frame of the adjacent tooth position of the repaired tooth;

obtaining a straight line angle according to the center point;

and evaluating the tooth filling quality of the straight line angle according to the tooth filling quality evaluation standard.

Preferably, the dental filling information includes a rectangular frame of adjacent tooth positions of the filled teeth;

acquiring the number of adjacent teeth corresponding to the rectangular frame of the adjacent tooth position of the compensated tooth;

detecting whether the number of the adjacent teeth meets a preset adjacent tooth threshold range or not, and obtaining an adjacent tooth number detection result;

and carrying out tooth filling quality evaluation on the adjacent tooth quantity detection result according to the tooth filling quality evaluation standard.

Preferably, the dental filling information comprises a picture of the area filled on the surface of the filled tooth;

converting the image of the region of the surface of the teeth to be filled into a gray image of the teeth to be filled;

dividing the gray level picture of the tooth filling by using a picture caries threshold value to obtain a picture of caries which is not cleaned;

acquiring a lowest gray value of the picture according to the uncleaned caries picture;

according to the tooth filling quality evaluation standard, carrying out tooth filling quality evaluation on the lowest gray value of the picture, wherein the lowest gray value of the picture is the tooth filling information evaluation result;

obtaining the number of picture pixels according to the uncleaned caries picture;

and evaluating the tooth filling quality of the picture pixel number according to the tooth filling quality evaluation standard.

Preferably, the tooth filling information comprises the single tooth filling picture and the area picture filled on the tooth filling surface;

Comparing the color corresponding to the tooth filling surface region picture with the color corresponding to the single tooth filling picture to obtain a first color comparison result;

and evaluating the filling quality of the first color comparison result according to the filling quality evaluation standard.

Preferably, the tooth filling information comprises a picture of a region filled in the surface of the filled tooth and a picture of adjacent teeth;

acquiring the color corresponding to the picture of the region on the surface of the tooth to be filled and the color corresponding to the picture of the adjacent tooth;

performing color comparison on the color corresponding to the picture of the region of the surface of the tooth to be filled and the color corresponding to the adjacent tooth picture to obtain a second color comparison result;

and according to the dental filling quality evaluation standard, performing dental filling quality evaluation on the second color comparison result.

Preferably, the filling information includes the single filled tooth picture;

Obtaining a standard gully line template according to the single tooth picture;

extracting the lines of the tooth filling picture through an edge detection algorithm to obtain tooth filling lines;

matching the tooth filling lines with the standard gully line template to obtain line matching results;

and evaluating the quality of the filling teeth according to the quality evaluation standard of the filling teeth.

In addition, in order to achieve the above object, the present invention also provides a dental filling quality evaluation device, which includes: the acquisition module is used for acquiring an oral cavity picture to be tested, inputting the oral cavity picture to be tested into a preset target detection convolutional neural network model, and acquiring a position rectangular frame and a tooth position number corresponding to each tooth;

the acquisition module is also used for carrying out target matching on the tooth position numbers of the teeth to be filled according to the tooth position numbers to obtain a rectangular frame of the tooth position to be filled and a rectangular frame of the adjacent tooth position of the teeth to be filled;

the acquisition module is also used for acquiring a rectangular frame picture of the tooth position of the repaired tooth and a rectangular frame picture of the adjacent tooth position of the repaired tooth according to the rectangular frame of the tooth position of the repaired tooth and the rectangular frame of the adjacent tooth position of the repaired tooth;

The computing module is used for inputting the rectangular frame picture of the tooth position of the filled tooth and the rectangular frame picture of the adjacent tooth position of the filled tooth into a preset single-tooth segmentation convolutional neural network model to obtain a single tooth filled tooth picture and an adjacent tooth picture;

the calculation module is also used for inputting the single compensated tooth picture into a preset area segmentation convolutional neural network model to obtain a compensated area picture on the surface of the compensated tooth;

the information module is used for taking the oral cavity picture to be tested, the rectangular frame of the tooth filling position, the rectangular frame of the adjacent tooth position of the tooth filling, the single tooth filling picture, the adjacent tooth picture and the picture of the area filling the tooth surface of the tooth filling as tooth filling information;

the evaluation module is used for obtaining the tooth filling information evaluation result according to the tooth filling information and evaluating the tooth filling quality of the tooth filling information evaluation result according to the tooth filling quality evaluation standard.

According to the method, an oral cavity picture to be tested is obtained, the oral cavity picture to be tested is input into a preset target detection convolutional neural network model, a position rectangular frame and a tooth position number corresponding to each tooth are obtained, then target matching is carried out on the tooth position number of the filled tooth according to the tooth position number, a tooth position rectangular frame and an adjacent tooth position rectangular frame of the filled tooth are obtained, an oral cavity picture to be tested and an adjacent tooth position rectangular frame picture of the filled tooth are obtained according to the tooth position rectangular frame and the adjacent tooth position rectangular frame picture of the filled tooth, then the tooth position rectangular frame picture and the adjacent tooth position rectangular frame picture of the filled tooth are input into a preset single tooth segmentation convolutional neural network model, a single tooth picture and an adjacent tooth picture are obtained, the single tooth picture is input into a preset region segmentation convolutional neural network model, a tooth surface filled region picture is obtained, the oral cavity picture to be tested, the adjacent tooth position rectangular frame picture, the single tooth picture and the adjacent tooth region filled tooth quality information are evaluated according to the tooth quality evaluation result information, and the tooth quality information is obtained according to the tooth quality evaluation result. By the mode, the defect that the traditional method relies on the experience of doctors to evaluate the quality of the filling teeth is overcome, and therefore the accuracy of evaluating the quality of the filling teeth is improved.

Drawings

FIG. 1 is a schematic flow chart of a first embodiment of a method for evaluating dental filling quality according to the present invention;

fig. 2 is a block diagram of a first embodiment of the dental filling quality evaluation apparatus of the present invention.

The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

The embodiment of the invention provides a method for evaluating the quality of teeth filling, and referring to fig. 1, fig. 1 is a schematic flow chart of a first embodiment of the method for evaluating the quality of teeth filling.

In this embodiment, the method for evaluating the quality of the dental filling includes the following steps:

step S10: and acquiring an oral cavity picture to be tested, and inputting the oral cavity picture to be tested into a preset target detection convolutional neural network model to obtain a position rectangular frame and a tooth position number corresponding to each tooth.

It should be understood that the execution body of the scheme may be a computer with an oral cavity picture processing function to be tested and a tooth filling quality evaluating function, and the computer may receive an oral cavity picture to be tested sent by a user.

Before the step of obtaining a position rectangular frame and a tooth position number corresponding to each tooth, obtaining a tooth detection sample picture, scaling the tooth detection sample picture according to a preset size to obtain an initial tooth detection sample picture, training a preset target detection convolutional neural network according to the initial tooth detection sample picture, establishing a preset target detection convolutional neural network model, segmenting the initial tooth detection sample picture according to a user operation instruction to obtain a single tooth segmentation sample picture, training a preset single tooth segmentation convolutional neural network according to the single tooth segmentation sample picture, establishing a preset single tooth segmentation convolutional neural network model, segmenting the single tooth segmentation sample picture according to a user instruction to obtain a region segmentation sample, training a preset region segmentation convolutional neural network according to the region segmentation sample, and establishing a preset region segmentation convolutional neural network model.

The above-mentioned oral cavity picture to be tested is an original picture, the original picture needs to be scaled according to a preset picture proportion, the preset picture proportion is user-defined, and the person skilled in the art does not limit the present invention.

The method is characterized in that standard original size positions which are not covered after the original pictures are scaled in equal proportion are filled with black, standard original size pictures are obtained, then tooth detection samples are manufactured, a target detection convolutional neural network based on deep learning is built based on a YOLO v3 convolutional neural network architecture, and the target detection convolutional neural network is trained by utilizing the tooth detection samples, so that a target detection convolutional neural network model is obtained.

In the embodiment, firstly, a single-tooth segmentation sample is manufactured according to a user operation instruction, then a single-tooth segmentation convolutional neural network based on deep learning is constructed based on a U-net convolutional neural network architecture, and finally the single-tooth segmentation convolutional neural network is trained by utilizing the single-tooth segmentation sample to obtain a single-tooth segmentation convolutional neural network model; firstly, making a region segmentation sample according to a user operation instruction, then constructing a region segmentation convolutional neural network based on deep learning based on a U-net convolutional neural network architecture, and finally training the region segmentation convolutional neural network by using the region segmentation sample to obtain a region segmentation convolutional neural network model.

Step S20: and performing target matching on the tooth position numbers of the teeth to be filled according to the tooth position numbers to obtain a rectangular frame of the tooth position to be filled and a rectangular frame of the adjacent tooth position of the teeth to be filled.

It should be noted that, according to the known tooth position numbers of the teeth to be repaired, performing target matching, finding out the corresponding teeth to be repaired, and obtaining the rectangular frame of the position of the teeth to be repaired in the picture, and the rectangular frame of the position of the adjacent teeth of the teeth to be repaired in the picture.

And arranging the tooth position numbers of all the position rectangular frames in the original picture from left to right according to the positions of the position rectangular frames in the picture, correcting the individual error numbers to obtain a tooth position number sequence of the original picture, and obtaining the position rectangular frames of the teeth to be repaired and the adjacent tooth position rectangular frames of the teeth to be repaired according to the known tooth position numbers and tooth position number sequences of the teeth to be repaired.

Step S30: and obtaining a rectangular frame picture of the position of the teeth to be filled and a rectangular frame picture of the position of the adjacent teeth according to the rectangular frame of the position of the teeth to be filled and the rectangular frame of the position of the adjacent teeth of the teeth to be filled.

And carrying out ring drawing and processing according to the position rectangular frame according to user operation so as to obtain the adjacent tooth position rectangular frame picture of the teeth complemented by the tooth position rectangular frame picture set.

It should be understood that the neural network cannot complete the predetermined task before training, so that the samples of the training neural network need to be manually marked, and the model obtained after training can already complete the corresponding task, thereby obtaining the rectangular frame picture of the tooth position of the repaired tooth and the rectangular frame picture of the adjacent tooth position of the repaired tooth.

Step S40: and inputting the rectangular frame picture of the tooth position of the filled tooth and the rectangular frame picture of the adjacent tooth position of the filled tooth into a preset single-tooth segmentation convolutional neural network model to obtain a single filled tooth picture and an adjacent tooth picture.

Step S50: and inputting the single compensated tooth picture into a preset area segmentation convolutional neural network model to obtain the compensated area picture on the surface of the compensated tooth.

Step S60: taking the oral cavity picture to be tested, the rectangular frame at the tooth filling position, the rectangular frame at the adjacent tooth position of the tooth to be filled, the single tooth filling picture, the adjacent tooth picture and the picture of the area filled on the tooth filling surface as tooth filling information.

Step S70: obtaining a dental filling information evaluation result according to the dental filling information, and evaluating the dental filling quality of the dental filling information evaluation result according to a dental filling quality evaluation standard

When the dental filling information comprises the oral cavity picture to be tested, the steps of obtaining a dental filling information evaluation result according to the dental filling information and evaluating the dental filling quality of the dental filling information evaluation result according to a dental filling quality evaluation standard are that a picture format, a picture proportion and a resolution corresponding to the oral cavity picture to be tested are obtained, whether the picture format corresponding to the oral cavity picture to be tested accords with a preset picture format or not is detected, a picture format detection result is obtained, and dental filling quality evaluation is performed on the picture format detection result according to the dental filling quality evaluation standard.

In addition, it is understood that whether the picture proportion corresponding to the oral cavity picture to be tested accords with a preset picture proportion is detected, a picture proportion detection result is obtained, and tooth filling quality evaluation is carried out on the picture proportion detection result according to the tooth filling quality evaluation standard; detecting whether the resolution corresponding to the oral cavity picture to be tested accords with the preset picture resolution, obtaining a picture resolution detection result, and evaluating the dental filling quality of the picture resolution detection result according to the dental filling quality evaluation standard.

When the tooth filling information comprises the tooth filling position rectangular frame and the adjacent tooth position rectangular frame of the filled tooth, the step of obtaining a tooth filling information evaluation result according to the tooth filling information and evaluating the tooth filling quality of the tooth filling information evaluation result according to a tooth filling quality evaluation standard comprises the steps of calculating the center coordinates of the filled tooth according to the tooth filling position rectangular frame, detecting the offset corresponding to the center coordinates of the filled tooth, obtaining an offset detection result, and evaluating the tooth filling quality of the offset detection result according to the tooth filling quality evaluation standard.

It is to be understood that, according to the rectangular frame of the tooth filling position and the rectangular frame of the tooth filling position of the tooth filling, calculating the area of the tooth filling and the area of the adjacent tooth, according to the area of the tooth filling and the area of the adjacent tooth, calculating the area ratio of the picture, detecting whether the area ratio of the picture meets the preset area threshold ratio, obtaining the detection result of the area ratio of the picture, and according to the evaluation standard of the quality of the tooth filling, evaluating the quality of the tooth filling of the detection result of the area ratio of the picture; calculating a rectangular frame center point according to the rectangular frame of the tooth filling position and the rectangular frame of the adjacent tooth position of the tooth to be filled, obtaining a straight line angle according to the center point, and evaluating the tooth filling quality of the straight line angle according to the tooth filling quality evaluation standard.

When the tooth filling information comprises adjacent tooth position rectangular frames of the filled teeth, the step of obtaining a tooth filling information evaluation result according to the tooth filling information and evaluating the tooth filling quality of the tooth filling information evaluation result according to a tooth filling quality evaluation standard comprises the steps of obtaining the adjacent tooth quantity corresponding to the adjacent tooth position rectangular frames of the filled teeth, detecting whether the adjacent tooth quantity meets a preset adjacent tooth threshold range, obtaining an adjacent tooth quantity detection result, and evaluating the tooth filling quality of the adjacent tooth quantity detection result according to the tooth filling quality evaluation standard.

When the tooth filling information comprises a picture of a region filled on the tooth filling surface, the step of obtaining a tooth filling information evaluation result according to the tooth filling information and evaluating the tooth filling quality of the tooth filling information evaluation result according to a tooth filling quality evaluation standard comprises the steps of converting the picture of the region filled on the tooth filling surface into a gray picture of the tooth filling, carrying out picture caries threshold segmentation on the gray picture of the tooth filling to obtain an uncleaned caries picture, obtaining a picture minimum gray value according to the uncleaned caries picture, and carrying out tooth filling quality evaluation on the picture minimum gray value according to the tooth filling quality evaluation standard, wherein the picture minimum gray value is the tooth filling information evaluation result; and obtaining the number of picture pixels according to the unclean caries picture, and evaluating the tooth filling quality of the number of picture pixels according to the tooth filling quality evaluation standard.

In addition, when the tooth filling information comprises the single tooth filling picture and the region picture on the tooth filling surface, the step of obtaining a tooth filling information evaluation result according to the tooth filling information and evaluating the tooth filling quality of the tooth filling information evaluation result according to a tooth filling quality evaluation standard comprises the steps of comparing the color corresponding to the tooth filling surface region picture with the color corresponding to the single tooth filling picture to obtain a first color comparison result, and evaluating the tooth filling quality of the first color comparison result according to the tooth filling quality evaluation standard.

When the tooth filling information comprises the region picture filled on the tooth filling surface and the adjacent tooth picture, the step of obtaining a tooth filling information evaluation result according to the tooth filling information and evaluating the tooth filling quality of the tooth filling information evaluation result according to a tooth filling quality evaluation standard comprises the steps of obtaining the color corresponding to the region picture filled on the tooth filling surface and the color corresponding to the adjacent tooth picture, comparing the color corresponding to the region picture filled on the tooth filling surface with the color corresponding to the adjacent tooth picture to obtain a second color comparison result, and evaluating the tooth filling quality of the second color comparison result according to the tooth filling quality evaluation standard.

When the tooth filling information comprises the single tooth filling picture, the step of obtaining a tooth filling information evaluation result according to the tooth filling information and evaluating the tooth filling quality of the tooth filling information evaluation result according to a tooth filling quality evaluation standard comprises the steps of obtaining a standard gully grain template according to the single tooth filling picture, extracting grains of the tooth filling picture through an edge detection algorithm to obtain tooth filling grains, matching the tooth filling grains with the standard gully grain template to obtain a grain matching result, and evaluating the tooth filling quality of the grain matching result according to the tooth filling quality evaluation standard.

Further, for ease of understanding, the following is illustrative:

(1) Scaling the original picture, unifying the original picture to a standard original size, inputting the original picture into a target detection convolutional neural network model based on deep learning, and performing target detection to obtain a position rectangular frame and a tooth position number of each tooth;

(2) Performing target matching according to the known tooth position numbers of the teeth to be repaired, finding out the corresponding teeth to be repaired, and obtaining a rectangular frame of the position of the teeth to be repaired in the picture and a rectangular frame of the position of the adjacent teeth of the teeth to be repaired in the picture;

(3) Scaling the rectangular frame picture of the tooth filling position and the rectangular frame picture of the adjacent tooth position of the filled tooth to a uniform single tooth standard size, and inputting the uniform single tooth standard size into a single tooth segmentation convolutional neural network model based on deep learning to obtain a single tooth filling picture and an adjacent tooth picture after segmentation;

(4) Scaling the single segmented teeth to a uniform single tooth standard size, and inputting the single segmented teeth into a region segmentation convolutional neural network model based on deep learning to obtain a region picture of the surface of the teeth;

(5) Performing photo quality evaluation according to the pixel size and the file type of the original picture in the step (1), the rectangular frame at the tooth position of the filled tooth in the step (2) and the rectangular frame at the adjacent tooth position of the filled tooth and the tooth quality evaluation standard and method;

(6) And (3) carrying out treatment quality evaluation according to the dental position number of the teeth to be repaired in the step (2), the single teeth to be repaired and the adjacent teeth in the step (3), and the region pictures to be repaired on the surfaces of the teeth to be repaired in the step (4) and the quality evaluation standard and method of teeth to be repaired.

Further, the specific implementation manner of the step (1) is as follows:

(1-1) scaling the original picture to 416 x 416, and filling the uncovered standard original size position of the original picture to black after the original picture is scaled to obtain a standard original size picture;

(1-2) manufacturing a tooth detection sample, constructing a target detection convolutional neural network based on deep learning based on a YOLO v3 convolutional neural network architecture, and training the target detection convolutional neural network by using the tooth detection sample to obtain a target detection convolutional neural network model;

and (1-3) inputting the standard original size picture into a target detection convolutional neural network model based on deep learning to obtain a tooth position rectangular frame and a tooth position number.

Further, the specific implementation manner of the step (2) is as follows:

(2-1) arranging the tooth numbers of all the position rectangular frames in the original picture from left to right according to the positions of the position rectangular frames in the picture, and correcting individual error numbers to obtain a tooth number sequence of the original picture;

(2-2) obtaining the rectangular frame of the tooth position of the repaired tooth and the rectangular frame of the adjacent tooth position of the repaired tooth according to the known tooth position number and tooth position number sequence of the repaired tooth.

Further, the specific implementation manner of the deep learning-based single-tooth segmentation convolutional neural network model establishment in the step (3) is as follows:

firstly, a single-tooth segmentation sample is manufactured, then a single-tooth segmentation convolutional neural network based on deep learning is constructed based on a U-net convolutional neural network architecture, and finally the single-tooth segmentation convolutional neural network is trained by utilizing the single-tooth segmentation sample, so that a single-tooth segmentation convolutional neural network model is obtained.

Further, the specific implementation manner of the deep learning-based region segmentation convolutional neural network model establishment in the step (4) is as follows:

firstly, making a region segmentation sample, then constructing a region segmentation convolutional neural network based on deep learning based on a U-net convolutional neural network architecture, and finally training the region segmentation convolutional neural network by using the region segmentation sample to obtain a region segmentation convolutional neural network model.

Further, the single tooth standard size pixel size in the step (3) and the step (4) is 255×255;

further, the dental filling quality evaluation criteria and the method in the step (5) and the step (6) comprise two major parts of photo quality evaluation and treatment quality evaluation:

The photo quality evaluation is used for evaluating whether the picture uploaded by a doctor meets the requirements or not, and comprises three parts, namely photo definition, reasonable composition and the inclusion of the repaired teeth and adjacent teeth or not;

the treatment quality evaluation is used for evaluating whether the dental filling operation of a doctor on the patient is qualified or not, and comprises four parts of caries cleaning or not, whether the color is consistent with that of the tooth, whether the color is consistent with that of the adjacent tooth and the tooth cuspid form or not.

Further, in the embodiment of the present invention, the photo definition evaluation criteria and method are for scoring all of the preoperative, intra-operative and post-operative pictures, and the total score A1 is A1 score, and the total score A1 is 10 score, including:

(1) Whether the original picture is in a JPG format or not, and the A1-1 is fully divided into 4 parts; the evaluation method is to read the original picture file type, if the original picture file type is in a JPG format, the original picture file type is fully divided, otherwise, the original picture file type is 0 divided;

(2) Whether the proportion of the original pictures is 16:9 or not, and the A1-2 is fully divided into 4 points; the evaluation method calculates the aspect ratio of the original picture according to the pixel size of the original picture, and if the aspect ratio is 16:9, fully dividing, otherwise, dividing by 0;

(3) Whether the resolution of the original picture is more than 560 x 315, wherein A1-3 is divided into 2 minutes; the evaluation method calculates the resolution of the original picture according to the pixel size of the original picture, if the resolution is more than 560 x 315, the resolution is full, otherwise, every 100K is deducted by one, and the deduction is completed.

Whether the composition reasonably evaluates the standard and the method is to score all the pictures before, during and after the operation, wherein the total A2 score and the total A2 score are 10 scores, and the method comprises the following steps:

(1) Whether the tooth is positioned in the center area of the original picture or not, wherein A2-1 is divided into 4 minutes; the evaluation method comprises the steps of calculating the center point coordinates of the teeth according to the rectangular frame of the tooth filling position, calculating the center point coordinates of an original picture according to the pixel size of the original picture, calculating the deviation between the center point coordinates of the teeth filling and the center point coordinates of the original picture, and if the deviation is within 3 percent, not buckling the teeth, buckling the teeth for 1 percent by 3 to 8 percent, buckling the teeth for 2 percent by 8 to 15 percent, buckling the teeth for 3 percent by 15 to 30 percent and buckling the teeth for 4 percent by more than 30 percent;

(2) Whether the area proportion of the tooth filling teeth and the adjacent teeth of the tooth filling teeth accounting for the whole original picture is reasonable or not, and the A2-2 is fully divided into 4 parts; the evaluation method comprises the steps of calculating the sum of the areas of the rectangular frames according to the rectangular frames at the tooth filling positions and the rectangular frames at the adjacent tooth positions of the filled teeth to obtain the areas of the filled teeth and the adjacent teeth, calculating the area of an original picture according to the pixel size of the original picture, and finally calculating the proportion of the areas of the filled teeth and the adjacent teeth to the area of the original picture, wherein if the proportion is 40-70%, the proportion is fully divided, otherwise, every 5% higher or lower is 1 part, and the proportion is completely divided;

(3) Whether the shooting angle is consistent with preoperative or not, and the A2-3 is divided into 2 minutes; according to the evaluation method, the center points of the rectangular frames are calculated according to the rectangular frames at the tooth filling positions and the rectangular frames at the adjacent tooth positions of the teeth, a straight line is fitted according to the center points of the rectangular frames by using a linear fitting method, the angle of the straight line is obtained, the full-scale angle is calculated before operation, the intra-operation angle and the post-operation angle are compared, 1-scale buckling is carried out every 15 degrees, and the buckling is completed.

The evaluation standard and method for whether the teeth to be repaired and the adjacent teeth are contained are to score all the pictures before, during and after the operation, wherein the total A3 score and the total A3 score are 10 scores, and the method comprises the following steps:

whether the tooth to be filled in the original picture contains adjacent teeth or not, wherein A3 is divided into 10 parts; the evaluation method calculates the number of adjacent teeth of the teeth to be filled according to the rectangular frame of the adjacent teeth positions of the teeth to be filled, wherein the number of the adjacent teeth is 2, if not, the number of the adjacent teeth is 3, and if not, every 1 adjacent tooth is 3.

The caries cleaning evaluation standard and method only score the intraoperative pictures, and the total score is B1, and the total score of B1 is 20, comprising the following steps:

(1) Whether the tooth to be filled contains unclean caries or not, B1-1 is divided into 10; according to the method, the image of the region complemented by the tooth surface is converted into a gray image, a threshold value for dividing unclean caries is set to be 120, caries in the gray image of the region complemented by the tooth surface is divided according to the threshold value, unclean caries images are obtained, the number of caries pixels in the unclean caries images is calculated, and 1 minute is buckled for every 30 black pixels until the caries pixels are buckled; the method comprises the steps of carrying out a first treatment on the surface of the

(2) The color depth of caries which is not cleaned up by the tooth filling teeth is fully divided into 10 minutes by B1-2; the evaluation method calculates the minimum value of the gray value of caries in the uncleaned caries picture, and the minimum value is buckled for 1 minute when the minimum value is lower than the uncleaned caries threshold value by 10 minutes until the caries is buckled.

And (3) evaluating whether the color is consistent with the evaluation standard and method of the tooth, and grading the postoperative picture, wherein the total score is B2, and the total score of B2 is 20. Comprising the following steps:

whether the color of the region complemented by the surface of the tooth complemented is consistent with that of the surrounding region of the tooth complemented is that of the region complemented by the surface of the tooth complemented, and the B2 is fully divided into 20 minutes; according to the method, according to the single teeth to be repaired and the images of the areas repaired on the surfaces of the teeth to be repaired, the positions of the areas of the surfaces of the teeth to be repaired in the operation are mapped to the single teeth to be repaired after the operation in an equal proportion mode, the areas repaired on the surfaces of the teeth to be repaired after the operation are obtained, the areas before expansion are subtracted from the areas after expansion, the surrounding areas of the areas repaired on the surfaces of the teeth to be repaired after the operation are obtained, and the mean value difference of H values of the areas repaired on the surfaces of the teeth to be repaired and the surrounding areas after the operation under the HSV color space is calculated, wherein each time the mean value difference exceeds 0.5, the buckling is completed.

(3) And (3) evaluating whether the color is consistent with the adjacent teeth or not according to the evaluation standard and the evaluation method, and grading the images after operation, wherein the total score of B3 is 20. Comprising the following steps:

Whether the color of the region complemented by the surface of the tooth to be complemented is consistent with that of the corresponding region of the adjacent tooth of the tooth to be complemented, wherein B3 is fully divided into 20 minutes; according to the evaluation method, the positions of the areas of the surface of the teeth to be repaired in the operation are mapped into the single teeth to be repaired and the adjacent teeth to be repaired in the operation in an equal proportion according to the single teeth to be repaired and the images of the areas of the surface of the teeth to be repaired, the areas of the surface of the teeth to be repaired and the adjacent teeth to be repaired are obtained, the mean value difference of H values of the areas of the surface of the teeth to be repaired and the areas of the adjacent teeth to be repaired in the HSV color space after the calculation is calculated, and each time the mean value difference exceeds 0.5, the number of the points is 1, and the points are completed.

(4) The tooth cuspid morphology evaluation standard only scores the pictures of the teeth which are subjected to operation and are compensated for tooth grinding, and the total score is B4, and the total score of B4 is 10. Comprising the following steps:

the matching degree of the gully lines of the repaired teeth and the standard gully line template is that B4 is divided into 10 minutes; according to the evaluation method, a known corresponding standard gully line template is obtained according to the tooth position number of the repaired tooth and a single tooth picture to be repaired, then an edge detection algorithm is used for carrying out line extraction on the single tooth picture to be repaired to obtain a single tooth line to be repaired, and then the single tooth line to be repaired is matched with the standard gully line template to obtain a matching degree which is more than 70%, otherwise, 1.5 points are obtained every 10%.

In addition, it should be noted that the beneficial effects mentioned in the above detailed description are:

(1) The invention utilizes the target detection convolutional neural network based on deep learning to complete the positioning and classification of all teeth of the picture, positions the repaired teeth, and then carries out the quality scoring of the repaired teeth, thereby improving the evaluation efficiency.

(2) According to the invention, the region of interest in the picture is segmented by utilizing the single-tooth segmentation convolutional neural network and the region segmentation convolutional neural network based on deep learning, namely, the region concerned in the doctor tooth filling quality evaluation is segmented, so that the algorithm is more concerned in the tooth filling region, and error judgment is avoided.

(3) The invention converts the doctor's dental filling quality evaluation thought into quantized 7 indexes, provides a quantized dental filling quality evaluation standard, provides a dental filling quality evaluation method, overcomes the defect of traditional dental filling quality evaluation based on doctor experience, and improves the evaluation accuracy.

According to the method, an oral cavity picture to be tested is obtained, the oral cavity picture to be tested is input into a preset target detection convolutional neural network model, a position rectangular frame and a tooth position number corresponding to each tooth are obtained, then target matching is carried out on the tooth position number of the filled tooth according to the tooth position number, a tooth position rectangular frame and an adjacent tooth position rectangular frame of the filled tooth are obtained, an oral cavity picture to be tested and an adjacent tooth position rectangular frame picture of the filled tooth are obtained according to the tooth position rectangular frame and the adjacent tooth position rectangular frame picture of the filled tooth, then the tooth position rectangular frame picture and the adjacent tooth position rectangular frame picture of the filled tooth are input into a preset single tooth segmentation convolutional neural network model, a single tooth picture and an adjacent tooth picture are obtained, the single tooth picture to be filled is input into a preset region segmentation convolutional neural network model, a tooth surface to be filled region picture is obtained, the oral cavity picture to be tested, the adjacent tooth position rectangular frame picture, the single tooth picture to be filled tooth, the adjacent tooth region to be filled tooth picture are evaluated and the tooth quality information of the tooth is obtained according to the evaluation result information, and the tooth quality information is evaluated according to the tooth quality information. By the mode, the defect that the traditional method relies on the experience of doctors to evaluate the quality of the filling teeth is overcome, and therefore the accuracy and the working efficiency of the quality evaluation of the filling teeth are improved.

Referring to fig. 2, fig. 2 is a block diagram showing the construction of a first embodiment of the dental filling quality evaluation apparatus according to the present invention.

As shown in fig. 2, the dental filling quality evaluation device provided by the embodiment of the invention includes:

the acquisition module 4001 is used for acquiring an oral cavity picture to be tested, inputting the oral cavity picture to be tested into a preset target detection convolutional neural network model, and acquiring a position rectangular frame and a tooth position number corresponding to each tooth; the obtaining module 4001 is further configured to perform target matching on the tooth position number of the tooth to be filled according to the tooth position number, so as to obtain a rectangular frame of the tooth position to be filled and a rectangular frame of the adjacent tooth position of the tooth to be filled; the obtaining module 4001 is further configured to obtain a picture of the rectangular frame of the position of the teeth to be repaired and a picture of the rectangular frame of the position of the teeth to be repaired according to the rectangular frame of the position of the teeth to be repaired and the rectangular frame of the position of the adjacent teeth to be repaired; the calculation module 4002 is configured to input the rectangular frame picture of the tooth position of the compensated tooth and the rectangular frame picture of the adjacent tooth position of the compensated tooth into a preset single-tooth segmentation convolutional neural network model to obtain a single picture of the compensated tooth and an adjacent tooth picture; the calculation module 4002 is further configured to input the single compensated tooth picture into a preset area segmentation convolutional neural network model, so as to obtain a compensated area picture of the surface of the compensated tooth; the information module 4003 is configured to use the oral cavity picture to be tested, the rectangular frame of the tooth filling position, the rectangular frame of the adjacent tooth position of the tooth filling, the single tooth filling picture, the adjacent tooth picture and the picture of the area filling the surface of the tooth filling as tooth filling information; the evaluation module 4004 is used for obtaining a dental filling information evaluation result according to the dental filling information, and evaluating the dental filling quality of the dental filling information evaluation result according to a dental filling quality evaluation standard.

The obtaining module 4001 obtains an oral cavity picture to be tested, inputs the oral cavity picture to be tested into a preset target detection convolutional neural network model, and obtains operations of a rectangular frame and a tooth position number corresponding to each tooth.

The obtaining module 4001 performs target matching on the tooth position number of the tooth to be filled according to the tooth position number, and obtains operations of the rectangular frame of the tooth position to be filled and the rectangular frame of the adjacent tooth position of the tooth to be filled.

The obtaining module 4001 obtains the operation of the rectangular frame picture of the position of the teeth to be filled and the rectangular frame picture of the position of the adjacent teeth to be filled according to the rectangular frame of the position of the teeth to be filled and the rectangular frame of the position of the adjacent teeth to be filled.

In addition, it should be understood that the neural network cannot complete the predetermined task before training, so that the samples of the training neural network need to be manually marked, and the model obtained after training can already complete the corresponding task, thereby obtaining the rectangular frame picture of the tooth position of the repaired tooth and the rectangular frame picture of the adjacent tooth position of the repaired tooth.

The calculation module 4002 inputs the rectangular frame picture of the tooth position of the filled tooth and the rectangular frame picture of the adjacent tooth position of the filled tooth into a preset single-tooth segmentation convolutional neural network model to obtain an operation of a single filled tooth picture and an adjacent tooth picture.

The calculation module 4002 inputs the single compensated tooth picture to a preset area division convolutional neural network model to obtain the operation of the compensated area picture on the surface of the compensated tooth.

The information module 4003 performs an operation of taking the oral cavity picture to be tested, the rectangular frame of the tooth filling position, the rectangular frame of the adjacent tooth position of the tooth filling, the single tooth filling picture, the adjacent tooth picture and the picture of the area filling the tooth surface as tooth filling information.

The evaluation module 4004 obtains a dental filling information evaluation result according to the dental filling information, and performs a dental filling quality evaluation operation on the dental filling information evaluation result according to a dental filling quality evaluation standard.

The method comprises the steps of calculating the areas of the teeth to be repaired and the adjacent teeth according to the rectangular frames of the teeth to be repaired and the rectangular frames of the adjacent teeth to be repaired, calculating the area proportion of pictures according to the areas of the teeth to be repaired and the adjacent teeth, detecting whether the area proportion of pictures meets the preset area threshold proportion, obtaining the detection result of the area proportion of pictures, and evaluating the quality of the teeth to be repaired according to the quality evaluation standard of the teeth to be repaired; calculating a rectangular frame center point according to the rectangular frame of the tooth filling position and the rectangular frame of the adjacent tooth position of the tooth to be filled, obtaining a straight line angle according to the center point, and evaluating the tooth filling quality of the straight line angle according to the tooth filling quality evaluation standard.

Further, for ease of understanding, the following is illustrative:

Further, the specific implementation manner of the step (1) is as follows:

Further, the specific implementation manner of the step (2) is as follows:

It should be understood that the foregoing is illustrative only and is not limiting, and that in specific applications, those skilled in the art may set the invention as desired, and the invention is not limited thereto.

It should be noted that the above-described working procedure is merely illustrative, and does not limit the scope of the present invention, and in practical application, a person skilled in the art may select part or all of them according to actual needs to achieve the purpose of the embodiment, which is not limited herein.

In addition, technical details which are not described in detail in the present embodiment can be referred to the method for evaluating dental filling quality provided in any embodiment of the present invention, and are not described herein.

Furthermore, it should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.

The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.

From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. Read Only Memory)/RAM, magnetic disk, optical disk) and including several instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present invention.

The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.

Claims

1. A method of evaluating dental filling quality, the method comprising:

obtaining a picture of the rectangular frame of the tooth position direction of the repaired tooth and a picture of the rectangular frame of the adjacent tooth position of the repaired tooth according to the rectangular frame of the tooth position of the repaired tooth and the rectangular frame of the adjacent tooth position of the repaired tooth;

Obtaining a dental filling information evaluation result according to the dental filling information, and evaluating the dental filling quality of the dental filling information evaluation result according to a dental filling quality evaluation standard;

the step of taking the oral cavity picture to be tested, the rectangular frame of the tooth filling position, the rectangular frame of the adjacent tooth position of the tooth filling, the single tooth filling picture, the adjacent tooth picture and the picture of the area filling the surface of the tooth filling as tooth filling information comprises the following steps:

taking a picture format, a picture proportion and a resolution corresponding to the oral cavity picture to be tested, a straight line angle of a center point of a rectangular frame corresponding to a rectangular frame at the adjacent tooth position of the teeth to be compensated, the number of adjacent teeth corresponding to the rectangular frame at the adjacent tooth position of the teeth to be compensated, a color corresponding to the single teeth to be compensated picture, a color corresponding to the adjacent teeth picture and a color corresponding to the region picture to be compensated on the surface of the teeth to be compensated as teeth compensation information.

2. The method of claim 1, wherein the step of obtaining an oral picture to be tested, inputting the oral picture to be tested into a predetermined target detection convolutional neural network model, and obtaining a position rectangular frame and a tooth position number corresponding to each tooth further comprises:

3. The method of claim 1, wherein the dental filling information comprises the oral cavity picture to be tested;

4. The method of claim 1, wherein the dental filling information comprises a rectangular box of the filled tooth position and a rectangular box of the adjacent tooth position of the filled tooth;

obtaining a straight line angle according to the center point;

5. The method of claim 1, wherein the dental filling information comprises a rectangular box of adjacent tooth positions of the filled teeth;

6. The method of claim 1, wherein the dental filling information comprises a picture of the area filled up on the surface of the filled up tooth;

7. The method of claim 1, wherein the dental filling information comprises the single filled tooth picture and a filled area picture of the filled tooth surface;

8. The method of claim 1, wherein the dental filling information includes a picture of a filled region of the surface of the filled tooth and the picture of the adjacent tooth;

9. The method of claim 1, wherein the dental filling information comprises the single filled-in tooth picture;

obtaining a standard gully line template according to the single tooth picture;

10. A dental filling quality evaluation device, the device comprising:

The acquisition module is used for acquiring an oral cavity picture to be tested, inputting the oral cavity picture to be tested into a preset target detection convolutional neural network model, and acquiring a position rectangular frame and a tooth position number corresponding to each tooth;

The evaluation module is used for obtaining a dental filling information evaluation result according to the dental filling information and evaluating the dental filling quality of the dental filling information evaluation result according to a dental filling quality evaluation standard;

the information module is further configured to use, as the tooth filling information, a picture format, a picture proportion and a resolution corresponding to the oral cavity picture to be tested, a straight line angle of a center point of the rectangular frame corresponding to the rectangular frame at the tooth filling position of the tooth, the number of adjacent teeth corresponding to the rectangular frame at the tooth filling position of the tooth, a color corresponding to the single tooth filling picture, a color corresponding to the adjacent tooth picture, and a color corresponding to the picture of the region filled on the tooth filling surface.