Automated assessment method of perioperative ASA grading based on deep learning

Abstract

The present invention relates to the technical field of image processing, and discloses a perioperative ASA grading automated assessment method based on deep learning, comprising first obtaining anesthesia type, calculating a perioperative assessment coefficient according to basic information of a patient, when the anesthesia type is local anesthesia, determining Q initial injection area images based on H oral cavity images, segmenting the Q initial injection area images to generate Q intermediate injection area images, determining a target injection position according to the Q intermediate injection area images, continuously obtaining real-time patient status information during surgery, calculating an anxiety coefficient according to the real-time patient status information and the perioperative assessment coefficient, and judging whether to stop the surgery based on the anxiety coefficient and the anesthesia type; the present invention performs preoperative and intraoperative assessments, and when selecting an anesthesia position, selects a position that causes less pain, thereby minimizing the psychological burden of anesthesia on the patient.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and more specifically, to a preoperative ASA grading automated assessment method based on deep learning. Background Art

[0002] The ASA physical status classification is a method for evaluating all body systems. It originated from the systemic assessment method of the American Society of Anesthesiologists and has become the most valuable method for judging surgical and anesthesia risks before surgery. In the field of oral medicine, the ASA classification is used to evaluate the patient's systemic condition and tolerance before oral treatment, understand their systemic medication, and judge the potential risks of oral local anesthetics and treatment operations. In the existing technology, the ASA classification technology is mostly used for preoperative or postoperative evaluation, such as using the ASA classification technology to predict the degree of surgical risk, or predict the quality of life after surgery, etc., ignoring the intraoperative evaluation.

[0003] For example, the Chinese patent application with publication number CN115381429A provides an airway assessment terminal based on artificial intelligence. The patent evaluates lung function by analyzing the audio frequency of the respiratory airflow and using machine learning to analyze the airflow changes produced during the entire breathing process. It also performs glottis assessment by performing glottis recognition on the collected image data of the target object's throat anatomical structure, thereby determining whether the patient will experience breathing difficulties after anesthesia.

[0004] Although the above patent was evaluated through the ASA grading technology, it still belongs to the preoperative evaluation. In the field of oral medicine, some patients are anxious mainly because they see oral treatment instruments, while other patients cannot tolerate the sound of dental drills, the sound of hammers and mechanical stimulation during the treatment process, which leads to severe anxiety. Therefore, only preoperative evaluation cannot accurately reduce the potential risks brought by the treatment operation, and preoperative anesthesia is equally important. The pain caused by anesthesia will also bring psychological burden to patients.

[0005] In view of this, the present invention proposes an automated perioperative ASA grading assessment method based on deep learning to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides an automated assessment method for perioperative ASA grading based on deep learning.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The deep learning-based automated assessment method for perioperative ASA grading includes:

[0009] S10: Obtain H oral sub-images of the area to be treated and basic information of the patient, and input the H oral sub-images into a pre-built anesthesia classification model to obtain anesthesia types, where the anesthesia types include local anesthesia and intravenous sedation, and H is an integer greater than 1;

[0010] S20: Calculate the perioperative assessment coefficient according to the patient's basic information, input the anesthesia type and the perioperative assessment coefficient into the pre-built anesthesia dose model to obtain the target anesthesia dose;

[0011] S30: when the anesthesia type is local anesthesia, Q initial injection area images are determined based on H oral sub-images, the Q initial injection area images are segmented to generate Q intermediate injection area images, and the target injection position is determined according to the Q intermediate injection area images, H>Q;

[0012] S40: Real-time patient status information is continuously obtained during the operation, an anxiety coefficient is calculated according to the real-time patient status information and a perioperative assessment coefficient, and whether to stop the operation is determined based on the anxiety coefficient and anesthesia type.

[0013] Furthermore, the method for obtaining H mouth images includes:

[0014] Obtain M oral region images, segment the M oral region images using a region growing algorithm to obtain S oral sub-images, calculate the similarities between the S oral sub-images and the standard sub-image, remove the oral sub-images whose similarity is greater than a preset similarity threshold, and generate H oral sub-images, where S>M.

[0015] Furthermore, the method for constructing the anesthesia classification model includes:

[0016] Obtain i groups of training data, where i is a positive integer greater than 1, and the training data includes historical oral sub-images and historical anesthesia types. The historical oral sub-images and historical anesthesia types are used as sample sets, and the sample sets are divided into training sets and test sets. A classifier is constructed, and the historical oral sub-images in the training set are used as input data, and the historical anesthesia types in the training set are used as output data. The classifier is trained to obtain an initial classifier, and the initial classifier is tested using the test set, and the initial classifier having an accuracy greater than a preset threshold is output as an anesthesia classification model.

[0017] Furthermore, the method for calculating the perioperative assessment coefficient according to the patient's basic information includes:

[0018]

[0019] Where PAS is the perioperative assessment coefficient, Pag is the patient's age, log3(·) is the logarithmic function with base 3, Pwt is the patient's weight, φ(·) is the standard normal distribution function, sin(·) is the sine function, and Pcn is r is the patient's creatinine value on day r, Pbn is the patient's bilirubin value, Pbg is the patient's blood glucose difference, ln(·) is a logarithmic function with e as the base, L and t are constants, e is a natural constant, is the rounding function.

[0020] Furthermore, the method for constructing the anesthetic dosage model includes:

[0021] A sample data set is obtained, wherein the sample data set includes historical anesthesia types, historical perioperative assessment coefficients, and historical anesthesia doses. The sample data set is divided into a sample training set and a sample test set, and a regression network is constructed. The historical anesthesia types and historical perioperative assessment coefficients in the sample training set are used as input data of the regression network, and the historical anesthesia doses in the sample training set are used as output data of the regression network. The regression network is trained to obtain an initial regression network for predicting real-time anesthesia doses. The initial regression network is tested using the sample test set, and the initial regression network that satisfies a preset error value is output as an anesthesia dose model.

[0022] Further, the method for determining Q initial injection area images based on H oral cavity sub-images includes:

[0023] H oral sub-images are identified respectively to determine whether there is a periodontal ligament in the oral sub-image. When the periodontal ligament exists in the oral sub-image, it is retained; when the periodontal ligament does not exist in the oral sub-image, it is removed, so as to generate Q initial injection area images.

[0024] Further, the method of generating Q intermediate injection area images includes:

[0025] The Q initial injection area images are converted into corresponding grayscale images, and the grayscale images are segmented based on the watershed algorithm to generate Q intermediate injection area images, which are represented as periodontal ligament images.

[0026] Further, the method for determining the target injection position according to the Q intermediate injection area images includes:

[0027] The real-time average grayscale values ​​of Q intermediate injection area images are obtained respectively, and the absolute value of the difference between the real-time average grayscale value and the standard average grayscale value is calculated. The intermediate injection area image corresponding to the minimum absolute value of the difference is used as the target injection area image, and the periodontal ligament in the target injection area image is used as the target injection position.

[0028] Furthermore, the method for calculating the anxiety coefficient based on the real-time patient status information and the perioperative assessment coefficient includes:

[0029]

[0030] Where ASE is the anxiety index, Bpv is the patient's blood pressure, Hrv is the patient's heart rate, Hry is the patient's heart rate variability, Gsr is the patient's skin conductance, PAS is the perioperative assessment coefficient, log2(·) is the logarithmic function with base 2, cosh(·) is the hyperbolic cosine function, and tan -1 (·) is the inverse tangent function, ln(·) is the logarithmic function with e as the base, e is a natural constant, and f1 and f2 are both weight factors.

[0031] Further, the method for determining whether to stop the operation based on the anxiety index and the anesthesia type includes:

[0032] When the anesthesia type is local anesthesia and the anxiety coefficient is greater than the preset first coefficient threshold, the operation is stopped;

[0033] When the anesthesia type is intravenous sedation and the anxiety coefficient is greater than the preset second coefficient threshold, the operation is stopped, and the preset first coefficient threshold is greater than the preset second coefficient threshold.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The present invention first obtains the anesthesia type, calculates the perioperative assessment coefficient according to the patient's basic information, and when the anesthesia type is local anesthesia, determines Q initial injection area images based on H oral cavity images, and divides the Q initial injection area images into Q intermediate injection area images, and determines the target injection position according to the Q intermediate injection area images. During the operation, real-time patient status information is continuously obtained, and the anxiety coefficient is calculated according to the real-time patient status information and the perioperative assessment coefficient. It is determined whether to stop the operation based on the anxiety coefficient and the anesthesia type. The present invention performs preoperative and intraoperative evaluations to accurately reduce the potential risks brought by the treatment operation. When selecting the anesthesia position, the position that brings less pain is selected, thereby minimizing the psychological burden of anesthesia on the patient. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flowchart of the deep learning-based perioperative ASA grading automated assessment method of the present invention;

[0037] Figure 2 This is a schematic diagram of the structure of the deep learning-based perioperative ASA grading automated assessment system of the present invention.

[0038] Figure 3A schematic diagram of a computer-readable storage medium in the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0040] Example 1

[0041] See also Figure 1 As shown, this embodiment discloses a method for automated perioperative ASA grading assessment based on deep learning, including:

[0042] S10: Obtain H oral sub-images of the area to be treated and basic information of the patient, and input the H oral sub-images into a pre-built anesthesia classification model to obtain anesthesia types, where the anesthesia types include local anesthesia and intravenous sedation, and H is an integer greater than 1;

[0043] In this embodiment, the area to be treated refers to the oral area of ​​the patient, and the basic information of the patient includes but is not limited to the patient's age, patient weight, patient blood sugar difference, patient bilirubin value and patient creatinine value; the patient's blood sugar value, patient bilirubin value and patient creatinine value are all obtained by testing the patient's blood, the patient's bilirubin value represents the patient's liver function, and the patient's creatinine value represents the patient's kidney function. Generally, the lower the patient's bilirubin value, the stronger the patient's liver function, and the lower the patient's creatinine value, the stronger the patient's kidney function. The above-mentioned patient's blood sugar difference refers to the difference between the real-time blood sugar value and the standard blood sugar value.

[0044] The method of obtaining H oral images includes:

[0045] Acquire M oral region images, segment the M oral region images by using a region growing algorithm to obtain S oral sub-images, calculate the similarity between the S oral sub-images and the standard sub-image, remove the oral sub-images with similarity greater than a preset similarity threshold, and generate H oral sub-images, S>M, wherein the similarity threshold is determined by a person skilled in the art according to actual needs and a large amount of experimental data;

[0046] It should be noted that the oral region image can be a left mandibular oral image or a right mandibular oral image. Taking the left mandibular oral image as an example, the left mandibular oral image has a row of teeth images of the left mandibular. Due to the damage of the teeth, a concave region image, i.e., a gingival tissue image, will also appear in the left mandibular oral image. Then, after segmentation by the region growing algorithm, the S oral sub-images have both images representing teeth and images representing gingival tissue.

[0047] It should be added that the standard sub-image mentioned above refers to the standard denture image. The purpose of calculating the similarity and removing the oral cavity sub-images whose similarity is greater than the preset similarity threshold is to remove the denture images. For example, if the patient is an elderly person, it is very likely that dentures are implanted in the elderly's mouth. In order to eliminate the interference of the denture image on the subsequent process, the denture image needs to be removed. Since the real tooth image is not similar to the standard denture image, the gingival tissue image is also not similar to the standard denture image, so the denture images in the S oral cavity sub-images can be removed;

[0048] It is understandable that the region growing algorithm starts from one or more seed points and gradually expands the region according to the similarity between pixels until the region boundary no longer meets the expansion conditions. The region growing algorithm is a prior art and will not be described in detail in this embodiment.

[0049] Methods for calculating the similarity between the oral cavity sub-image and the standard sub-image include:

[0050] Obtaining a first average grayscale value of the oral cavity sub-image and a second average grayscale value of the standard sub-image, calculating the Euclidean distance between the first average grayscale value and the second average grayscale value, and taking the reciprocal of the Euclidean distance as the similarity between the oral cavity sub-image and the standard sub-image;

[0051] It can be understood that the above calculation of the Euclidean distance is only an exemplary description and this embodiment does not limit this.

[0052] The method for constructing an anesthesia classification model includes:

[0053] Obtain i groups of training data, where i is an integer greater than 1, and the training data includes historical oral sub-images and historical anesthesia types. The historical oral sub-images and historical anesthesia types are used as sample sets, and the sample sets are divided into a training set and a test set. A classifier is constructed, and the historical oral sub-images in the training set are used as input data, and the historical anesthesia types in the training set are used as output data. The classifier is trained to obtain an initial classifier, and the initial classifier is tested using the test set, and the initial classifier having a value greater than a preset accuracy threshold is output as an anesthesia classification model. The classifier is preferably a naive Bayes model or a support vector machine model, and the accuracy threshold is preset according to the accuracy required by the anesthesia classification model;

[0054] It should be noted that the oral sub-image represents a tooth image or a gum tissue image, and the severity of the inside of the oral cavity shown by the oral sub-image is used to determine whether local anesthesia or intravenous sedation is required. For example, when it is determined based on the oral sub-image that the patient needs a tooth filling or root canal treatment, only local anesthesia is required. When it is determined based on the oral sub-image that the patient needs a dental implant or complex wisdom tooth extraction, intravenous sedation is required. The determination of which surgery the patient needs based on the oral sub-image can be based on expert experience or can be analyzed with the help of existing dental diagnostic systems, such as the DentAl diagnostic system or the 0verjet diagnostic system.

[0055] S20: Calculate the perioperative assessment coefficient according to the patient's basic information, input the anesthesia type and the perioperative assessment coefficient into the pre-built anesthesia dose model to obtain the target anesthesia dose;

[0056] Methods for calculating perioperative assessment coefficients based on basic patient information include:

[0057]

[0058] Where PAS is the perioperative assessment coefficient, Pag is the patient's age, log3(·) is the logarithmic function with base 3, Pwt is the patient's weight, φ(·) is the standard normal distribution function, Sin(·) is the sine function, and Pcn is r is the patient's creatinine value on day r, Pbn is the patient's bilirubin value, Pbg is the patient's blood glucose difference, ln(·) is a logarithmic function with e as the base, L and t are constants, e is a natural constant, is the rounding function.

[0059] It should be noted that creatinine is a product of muscle metabolism and is mainly excreted through the kidneys. It is used to evaluate renal function. The patient's creatinine value needs to be tested long-term and continuously to obtain more accurate data. Taking the patient's creatinine value as an example, the lower the patient's creatinine value, the stronger the patient's renal function, and the larger the dose that can be used for anesthesia. Therefore, in this embodiment, the perioperative assessment coefficient is positively correlated with the anesthetic dose.

[0060] The methods for constructing anesthesia dosage models include:

[0061] Acquire a sample data set, wherein the sample data set includes historical anesthesia types, historical perioperative assessment coefficients, and historical anesthesia doses, divide the sample data set into a sample training set and a sample test set, construct a regression network, use the historical anesthesia types and historical perioperative assessment coefficients in the sample training set as input data of the regression network, use the historical anesthesia doses in the sample training set as output data of the regression network, train the regression network to obtain an initial regression network for predicting real-time anesthesia doses, test the initial regression network using the sample test set, and output the initial regression network that satisfies a preset error value as an anesthesia dose model, wherein the initial regression network is preferably a neural network model;

[0062] It should be added that the type of anesthesia directly affects the subsequent anesthetic dose. Taking local anesthesia and intravenous sedation as examples, the anesthetic dose of local anesthesia is less than that of intravenous sedation. This is because different types of anesthesia need to achieve different effects and depths, and therefore different anesthetic doses are required. Therefore, this embodiment combines the anesthesia type and the perioperative assessment coefficient to determine the final target anesthetic dose.

[0063] S30: when the anesthesia type is local anesthesia, Q initial injection area images are determined based on H oral sub-images, the Q initial injection area images are segmented to generate Q intermediate injection area images, and the target injection position is determined according to the Q intermediate injection area images, H>Q;

[0064] The method for determining Q initial injection area images based on H oral cavity images includes:

[0065] Recognize H oral sub-images respectively to determine whether there is a periodontal ligament in the oral sub-image. If there is a periodontal ligament in the oral sub-image, retain it; if there is no periodontal ligament in the oral sub-image, remove it, so as to generate Q initial injection area images;

[0066] The recognition of H oral images can be achieved through the YOLO algorithm, which is a real-time target detection algorithm. The periodontal ligament features can be input into the YOLO algorithm, and the YOLO algorithm specifically used for detecting the periodontal ligament can be obtained through transfer learning. It can be understood that transfer learning is a machine learning technology, which is usually used to apply knowledge learned in one task to another related task. In the prior art, there is already a YOLO algorithm for identifying medical images, and the YOLO algorithm specifically used for detecting the periodontal ligament can be obtained more quickly through transfer learning.

[0067] The method of generating Q intermediate injection area images includes:

[0068] The Q initial injection area images are converted into corresponding grayscale images, and the grayscale images are segmented based on the watershed algorithm to generate Q intermediate injection area images, which are represented as periodontal ligament images;

[0069] It should be noted that the initial injection area image is obtained by screening the oral sub-image, and the initial injection area image is an oral sub-image with a periodontal ligament. Taking the initial injection area image as an example to represent the tooth image, there are both teeth and periodontal ligament in the initial injection area image, so the target injection area image mentioned above refers to the periodontal ligament image. The watershed algorithm is an image segmentation technology, which is commonly used for the segmentation of grayscale images. It regards the image as terrain and simulates the process of water expanding from low to high. The watershed algorithm is a prior art and this embodiment will not go into details.

[0070] The method for determining the target injection position according to Q intermediate injection area images includes:

[0071] The real-time average grayscale values ​​of Q intermediate injection area images are respectively obtained, and the absolute value of the difference between the real-time average grayscale value and the standard average grayscale value is calculated, and the intermediate injection area image corresponding to the minimum absolute value of the difference is used as the target injection area image, and the periodontal ligament in the target injection area image is used as the target injection position;

[0072] It is worth noting that the standard average grayscale value mentioned above is the average grayscale value of the standard injection area image, and the standard injection area image refers to the image of the periodontal ligament in a good state. It can be understood that the periodontal ligament is pink in a good state. In this embodiment, in order to minimize the pain caused by anesthesia, the periodontal ligament in a good state is selected when selecting the anesthesia position, so as to minimize the psychological burden of anesthesia on the patient.

[0073] S40: continuously obtaining real-time patient status information during the operation, calculating an anxiety coefficient according to the real-time patient status information and a perioperative assessment coefficient, and determining whether to stop the operation based on the anxiety coefficient and anesthesia type;

[0074] Real-time patient status information includes but is not limited to patient blood pressure, patient heart rate, patient skin conductance, and patient heart rate variability. Skin conductance reflects sweat gland activity. When a patient is anxious or fearful, skin conductance usually increases. Heart rate variability is the time difference between two consecutive heartbeats. When a patient is anxious or fearful, skin conductance usually decreases. The method for obtaining the patient's skin conductance is to place a skin conductance sensor on a finger or palm and obtain the skin conductance value by measuring the change in skin resistance. The patient's heart rate variability can be directly obtained through a smart device.

[0075] Methods for calculating anxiety coefficients based on real-time patient status information and perioperative assessment coefficients include:

[0076]

[0077] Where ASE is the anxiety index, Bpv is the patient's blood pressure, Hrv is the patient's heart rate, Hry is the patient's heart rate variability, Gsr is the patient's skin conductance, PAS is the perioperative assessment coefficient, log2(·) is the logarithmic function with base 2, cosh(·) is the hyperbolic cosine function, and tan -1 (·) is the inverse tangent function, ln(·) is the logarithmic function with e as the base, e is a natural constant, f1 and f2 are weight factors, and are set according to experience.

[0078] Methods for deciding whether to stop surgery based on anxiety level and type of anesthesia include:

[0079] When the anesthesia type is local anesthesia and the anxiety coefficient is greater than the preset first coefficient threshold, the operation is stopped;

[0080] When the anesthesia type is intravenous sedation, and the anxiety coefficient is greater than the preset second coefficient threshold, the operation is stopped, and the preset first coefficient threshold is greater than the preset second coefficient threshold;

[0081] It should be noted that the first coefficient threshold and the second coefficient threshold are determined by technical personnel in this field based on actual needs and a large number of experiments. Local anesthesia is to block nerve conduction in a specific part by injecting anesthetic drugs, thereby making the area lose sensation without affecting the patient's state of consciousness. Intravenous sedation is to make the patient enter a relaxed and sedated state by intravenous injection of sedatives. So when the patient becomes anxious because of the sound of a dental drill, etc., usually, the real-time patient status information detected by the two is different. The patient's blood pressure value and patient's heart rate value detected after intravenous sedation are significantly smaller, and the corresponding patient's skin conductance value and patient's heart rate variability value are larger. Therefore, they need to be considered separately to accurately judge the patient's anxiety level at this time and reduce the risk of accidents.

[0082] In this embodiment, H oral sub-images of the area to be treated and basic information of the patient are obtained, and the H oral sub-images are input into a pre-constructed anesthesia classification model to obtain the anesthesia type, and the perioperative assessment coefficient is calculated according to the basic information of the patient. When the anesthesia type is local anesthesia, Q initial injection area images are determined based on the H oral sub-images, and the Q initial injection area images are segmented to generate Q intermediate injection area images. The target injection position is determined according to the Q intermediate injection area images, and real-time patient status information is continuously obtained during the operation. The anxiety coefficient is calculated according to the real-time patient status information and the perioperative assessment coefficient, and whether to stop the operation is determined based on the anxiety coefficient and the anesthesia type. This embodiment performs preoperative and intraoperative evaluations to accurately reduce the potential risks brought about by the treatment operation. When selecting the anesthesia position, the position that causes less pain is selected to minimize the psychological burden of anesthesia on the patient.

[0083] Example 2

[0084] like Figure 2 As shown, this embodiment provides a preoperative ASA grading automated assessment system based on deep learning on the basis of embodiment 1, including:

[0085] Classification module: used to obtain H oral sub-images of the area to be treated and basic information of the patient, and input the H oral sub-images into the pre-built anesthesia classification model to obtain the anesthesia type, which includes local anesthesia and intravenous sedation, and H is an integer greater than 1;

[0086] In this embodiment, the area to be treated refers to the oral area of ​​the patient. The basic information of the patient includes but is not limited to the patient's age, patient weight, patient blood sugar difference, patient bilirubin value and patient creatinine value. The patient's blood sugar value, patient bilirubin value and patient creatinine value are all obtained by testing the patient's blood. The patient's bilirubin value represents the patient's liver function, and the patient's creatinine value represents the patient's kidney function. Generally, the lower the patient's bilirubin value, the stronger the patient's liver function, and the lower the patient's creatinine value, the stronger the patient's kidney function. The above-mentioned patient's blood sugar difference refers to the difference between the real-time blood sugar value and the standard blood sugar value.

[0087] The method of obtaining H oral images includes:

[0088] Obtain M oral region images, segment the M oral region images by using a region growing algorithm to obtain S oral sub-images, calculate the similarity between the S oral sub-images and the standard sub-image, remove the oral sub-images whose similarity is greater than a preset similarity threshold, and generate H oral sub-images, S>M;

[0089] It should be noted that the oral region image can be a left mandibular oral image or a right mandibular oral image. Taking the left mandibular oral image as an example, the left mandibular oral image has a row of teeth images of the left mandibular. Due to the damage of the teeth, a concave region image, i.e., a gingival tissue image, will also appear in the left mandibular oral image. Then, after segmentation by the region growing algorithm, the S oral sub-images have both images representing teeth and images representing gingival tissue.

[0090] It is understandable that the region growing algorithm starts from one or more seed points and gradually expands the region according to the similarity between pixels until the region boundary no longer meets the expansion conditions. The region growing algorithm is a prior art and will not be described in detail in this embodiment.

[0091] Methods for calculating the similarity between the oral cavity sub-image and the standard sub-image include:

[0092] Obtaining a first average grayscale value of the oral cavity sub-image and a second average grayscale value of the standard sub-image, calculating the Euclidean distance between the first average grayscale value and the second average grayscale value, and taking the reciprocal of the Euclidean distance as the similarity between the oral cavity sub-image and the standard sub-image;

[0093] It can be understood that the above calculation of the Euclidean distance is only an exemplary description and this embodiment does not limit this.

[0094] Dose generation module: used to calculate the perioperative assessment coefficient based on the patient's basic information, input the anesthesia type and perioperative assessment coefficient into the pre-built anesthesia dose model, and obtain the target anesthesia dose;

[0095] Methods for calculating perioperative assessment coefficients based on basic patient information include:

[0096]

[0097] Where PAS is the perioperative assessment coefficient, Pag is the patient's age, log3(·) is the logarithmic function with base 3, Pwt is the patient's weight, φ(·) is the standard normal distribution function, sin(·) is the sine function, and Pcn is r is the patient's creatinine value on day r, Pbn is the patient's bilirubin value, Pbg is the patient's blood glucose difference, ln(·) is a logarithmic function with e as the base, L and t are constants, e is a natural constant, is the rounding function.

[0098] It should be noted that creatinine is a product of muscle metabolism and is mainly excreted through the kidneys. It is used to evaluate renal function. The patient's creatinine value needs to be tested long-term and continuously to obtain more accurate data. Taking the patient's creatinine value Pcn as an example, the lower the patient's creatinine value Pcn, the stronger the patient's renal function, and the larger the dose that can be used for anesthesia. Therefore, in this embodiment, the perioperative assessment coefficient PAS is positively correlated with the anesthetic dose.

[0099] Position determination module: when the anesthesia type is local anesthesia, Q initial injection area images are determined based on H oral sub-images, the Q initial injection area images are segmented to generate Q intermediate injection area images, and the target injection position is determined according to the Q intermediate injection area images, H>Q.

[0100] The method for determining Q initial injection area images based on H oral cavity images includes:

[0101] Recognize H oral sub-images respectively to determine whether there is a periodontal ligament in the oral sub-image. If there is a periodontal ligament in the oral sub-image, retain it; if there is no periodontal ligament in the oral sub-image, remove it, so as to generate Q initial injection area images;

[0102] The recognition of H oral images can be achieved through the YOLO algorithm. The YOLO algorithm is a real-time target detection algorithm. The periodontal ligament features can be input into the YOLO algorithm, and the YOLO algorithm specifically used for detecting the periodontal ligament can be obtained through transfer learning. It can be understood that transfer learning is a machine learning technology, which is usually used to apply knowledge learned in one task to another related task. In the prior art, there is already a YOLO algorithm for identifying medical images, and the Y0LO algorithm specifically used for detecting the periodontal ligament can be obtained more quickly through transfer learning.

[0103] The method of generating Q intermediate injection area images includes:

[0104] The Q initial injection area images are converted into corresponding grayscale images, and the grayscale images are segmented based on the watershed algorithm to generate Q intermediate injection area images, which are represented as periodontal ligament images;

[0105] It should be noted that the initial injection area image is obtained by screening the oral sub-image, and the initial injection area image is an oral sub-image with a periodontal ligament. Taking the initial injection area image as an example to represent the tooth image, there are both teeth and periodontal ligament in the initial injection area image, so the target injection area image mentioned above refers to the periodontal ligament image. The watershed algorithm is an image segmentation technology, which is commonly used for the segmentation of grayscale images. It regards the image as terrain and simulates the process of water expanding from low to high. The watershed algorithm is a prior art and this embodiment will not go into details.

[0106] The method for determining the target injection position according to Q intermediate injection area images includes:

[0107] The real-time average grayscale values ​​of Q intermediate injection area images are respectively obtained, and the absolute value of the difference between the real-time average grayscale value and the standard average grayscale value is calculated, and the intermediate injection area image corresponding to the minimum absolute value of the difference is used as the target injection area image, and the periodontal ligament in the target injection area image is used as the target injection position;

[0108] It is worth noting that the standard average grayscale value mentioned above is the average grayscale value of the standard injection area image, and the standard injection area image refers to the image of the periodontal ligament in a good state. It can be understood that the periodontal ligament is pink in a good state. In this embodiment, in order to minimize the pain caused by anesthesia, the periodontal ligament in a good state is selected when selecting the anesthesia position, so as to minimize the psychological burden of anesthesia on the patient.

[0109] Termination module: used to continuously obtain real-time patient status information during the operation, calculate the anxiety coefficient based on the real-time patient status information and the perioperative assessment coefficient, and determine whether to stop the operation based on the anxiety coefficient and anesthesia type;

[0110] Real-time patient status information includes but is not limited to patient blood pressure, patient heart rate, patient skin conductance, and patient heart rate variability. Skin conductance reflects sweat gland activity. When a patient is anxious or fearful, skin conductance usually increases. Heart rate variability is the time difference between two consecutive heartbeats. When a patient is anxious or fearful, skin conductance usually decreases. The method for obtaining the patient's skin conductance is to place a skin conductance sensor on a finger or palm and obtain the skin conductance value by measuring the change in skin resistance. The patient's heart rate variability can be directly obtained through a smart device.

[0111] Methods for calculating anxiety coefficients based on real-time patient status information and perioperative assessment coefficients include:

[0112]

[0113] Where ASE is the anxiety index, Bpv is the patient's blood pressure, Hrv is the patient's heart rate, Hry is the patient's heart rate variability, Gsr is the patient's skin conductance, PAS is the perioperative assessment coefficient, log2(·) is the logarithmic function with base 2, cosh(·) is the hyperbolic cosine function, and tan -1 (·) is the inverse tangent function, ln(·) is the logarithmic function with e as the base, e is a natural constant, f1 and f2 are weight factors, which are set according to experience.

[0114] Methods for deciding whether to stop surgery based on anxiety level and type of anesthesia include:

[0115] When the anesthesia type is local anesthesia and the anxiety coefficient is greater than the preset first coefficient threshold, the operation is stopped;

[0116] When the anesthesia type is intravenous sedation, and the anxiety coefficient is greater than the preset second coefficient threshold, the operation is stopped, and the preset first coefficient threshold is greater than the preset second coefficient threshold;

[0117] It should be noted that local anesthesia is the process of blocking nerve conduction in a specific area by injecting anesthetic drugs, thereby making the area lose sensation without affecting the patient's state of consciousness. Intravenous sedation is the process of injecting sedatives intravenously to make the patient enter a relaxed and calm state. So when the patient becomes anxious due to the sound of a dental drill, etc., usually, the real-time patient status information detected by the two is different. The blood pressure and heart rate values ​​detected by the patients after intravenous sedation are significantly lower, and the corresponding skin conductance and heart rate variability values ​​are larger. Therefore, they need to be considered separately to accurately judge the patient's anxiety level at this time and reduce the risk of accidents.

[0118] Example 3

[0119] The present embodiment discloses an electronic device, including a power supply, an interface, a keyboard, a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the deep learning-based perioperative ASA grading automated assessment method provided by the above-mentioned methods is implemented. The interface includes a network interface and a data interface. The network interface includes a wired or wireless interface, and the data interface includes an input or output interface.

[0120] Since the electronic device introduced in this embodiment is an electronic device used to implement the perioperative ASA grading automated assessment method based on deep learning in the embodiment of the present application, based on the perioperative ASA grading automated assessment method based on deep learning introduced in the embodiment of the present application, a person skilled in the art can understand the specific implementation of the electronic device of the present embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as a person skilled in the art implements the electronic device used by the perioperative ASA grading automated assessment method based on deep learning in the embodiment of the present application, it belongs to the scope of protection of this application.

[0121] Example 4

[0122] like Figure 3 As shown, the present embodiment discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the above-mentioned deep learning-based automated perioperative ASA grading assessment method is implemented.

[0123] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters, weights and thresholds in the formula are set by technicians in this field according to actual conditions.

[0124] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired network or a wireless network. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD) or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0125] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0126] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0127] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0128] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0129] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0130] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0131] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An automated perioperative ASA grading assessment method based on deep learning, characterized in that: include: S10: Obtain H oral sub-images of the area to be treated and basic information of the patient, and input the H oral sub-images into a pre-built anesthesia classification model to obtain anesthesia types, where the anesthesia types include local anesthesia and intravenous sedation, and H is an integer greater than 1; S20: Calculate the perioperative assessment coefficient according to the patient's basic information, input the anesthesia type and the perioperative assessment coefficient into the pre-built anesthesia dose model to obtain the target anesthesia dose; S30: when the anesthesia type is local anesthesia, Q initial injection area images are determined based on H oral sub-images, the Q initial injection area images are segmented to generate Q intermediate injection area images, and the target injection position is determined according to the Q intermediate injection area images, H>Q; S40: continuously obtaining real-time patient status information during the operation, calculating an anxiety coefficient according to the real-time patient status information and a perioperative assessment coefficient, and determining whether to stop the operation based on the anxiety coefficient and anesthesia type; The method of obtaining H oral images includes: Obtain M oral region images, segment the M oral region images by using a region growing algorithm to obtain S oral sub-images, calculate the similarity between the S oral sub-images and the standard sub-image, remove the oral sub-images whose similarity is greater than a preset similarity threshold, and generate H oral sub-images, S>M; The method for determining Q initial injection area images based on H oral sub-images comprises: Recognize H oral sub-images respectively to determine whether there is a periodontal ligament in the oral sub-image. If there is a periodontal ligament in the oral sub-image, retain it; if there is no periodontal ligament in the oral sub-image, remove it, so as to generate Q initial injection area images; The method for generating Q intermediate injection area images comprises: The Q initial injection area images are converted into corresponding grayscale images, and the grayscale images are segmented based on the watershed algorithm to generate Q intermediate injection area images, which are represented as periodontal ligament images; The method for determining the target injection position according to Q intermediate injection area images comprises: The real-time average grayscale values ​​of Q intermediate injection area images are obtained respectively, and the absolute value of the difference between the real-time average grayscale value and the standard average grayscale value is calculated. The intermediate injection area image corresponding to the minimum absolute value of the difference is used as the target injection area image, and the periodontal ligament in the target injection area image is used as the target injection position.

2. The deep learning-based perioperative ASA grading automated assessment method according to claim 1, characterized in that: The method for constructing the anesthesia classification model includes: Obtain i groups of training data, where i is a positive integer greater than 1, and the training data includes historical oral sub-images and historical anesthesia types. The historical oral sub-images and historical anesthesia types are used as sample sets, and the sample sets are divided into training sets and test sets. A classifier is constructed, and the historical oral sub-images in the training set are used as input data, and the historical anesthesia types in the training set are used as output data. The classifier is trained to obtain an initial classifier, and the initial classifier is tested using the test set, and the initial classifier having an accuracy greater than a preset threshold is output as an anesthesia classification model.

3. The deep learning-based automated perioperative ASA grading assessment method according to claim 1, characterized in that: The method for calculating the perioperative assessment coefficient according to the patient's basic information includes: Where PAS is the perioperative assessment coefficient, Pag is the patient's age, log3(·) is the logarithmic function with base 3, Pwt is the patient's weight, is the standard normal distribution function, sin(·) is the sine function, Pcn r is the patient's creatinine value on day r, Pbn is the patient's bilirubin value, Pbg is the patient's blood glucose difference, ln(·) is a logarithmic function with e as the base, L and t are constants, e is a natural constant, is the rounding function.

4. The deep learning-based automated perioperative ASA grading assessment method according to claim 1, characterized in that: The method for constructing the anesthetic dosage model comprises: A sample data set is obtained, wherein the sample data set includes historical anesthesia types, historical perioperative assessment coefficients, and historical anesthesia doses. The sample data set is divided into a sample training set and a sample test set, and a regression network is constructed. The historical anesthesia types and historical perioperative assessment coefficients in the sample training set are used as input data of the regression network, and the historical anesthesia doses in the sample training set are used as output data of the regression network. The regression network is trained to obtain an initial regression network for predicting real-time anesthesia doses. The initial regression network is tested using the sample test set, and the initial regression network that satisfies a preset error value is output as an anesthesia dose model.

5. The deep learning-based perioperative ASA grading automated assessment method according to claim 1, characterized in that: The method for calculating the anxiety coefficient based on real-time patient status information and perioperative assessment coefficient includes: Where ASE is the anxiety index, Bpv is the patient's blood pressure, Hrv is the patient's heart rate, Hry is the patient's heart rate variability, Gsr is the patient's skin conductance, PAS is the perioperative assessment coefficient, log2(·) is the logarithmic function with base 2, cosh(·) is the hyperbolic cosine function, and tan -1 (·) is the inverse tangent function, ln(·) is the logarithmic function with e as the base, e is a natural constant, and f1 and f2 are both weight factors.

6. The deep learning-based automated perioperative ASA grading assessment method according to claim 5, characterized in that: The method for determining whether to stop surgery based on the anxiety coefficient and the anesthesia type includes: When the anesthesia type is local anesthesia and the anxiety coefficient is greater than the preset first coefficient threshold, the operation is stopped; When the anesthesia type is intravenous sedation and the anxiety coefficient is greater than the preset second coefficient threshold, the operation is stopped, and the preset first coefficient threshold is greater than the preset second coefficient threshold.

Images