Intelligent monitoring system for postoperative rehabilitation of ophthalmic patient

Abstract

The invention discloses an intelligent monitoring system for postoperative rehabilitation of ophthalmic patients, and relates to the technical field of intelligent monitoring of postoperative rehabilitation. The method comprises the following steps: shooting postoperative parts of eyes of a patient at a fixed sampling frequency through a wearable image acquisition module, acquiring a postoperative part image # imgabs1 # at the current # imgabs0 # moment, acquiring basic information and chief complaint information of the patient at the same time, uploading the acquired real-time image to an image processing server by utilizing a client, and transmitting the real-time image to the image processing server; the method comprises the following steps: extracting a first text feature and a second text feature of chief complaint information of a to-be-detected patient through a natural language processing technology, extracting image features through a convolutional neural network, carrying out information data fusion on obtained postoperative part image information, basic information and chief complaint information, and constructing a neural network model; the postoperative part recovery state level corresponding to the feature quantity is recognized through a neural network, weighting processing is conducted on the postoperative part recovery state level through the weighting coefficient, and the postoperative part recovery state level of the to-be-detected patient is determined.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent postoperative rehabilitation monitoring, and particularly relates to an intelligent postoperative rehabilitation monitoring system for ophthalmic patients. Background Art

[0002] With the continuous development of ophthalmic surgery techniques, more and more ophthalmic diseases can be effectively treated through surgery. However, the monitoring and management during the postoperative rehabilitation stage are crucial for the vision recovery of patients and the prevention of complications. Traditional postoperative rehabilitation monitoring for ophthalmology mainly relies on regular reexaminations of patients and the subjective judgments of medical staff. This method has the following problems: Discontinuous monitoring: Postoperative patients usually need to travel back and forth between the hospital and home for reexaminations, resulting in time intervals in monitoring and making it impossible to grasp the patient's recovery situation in real time; Lack of real-time feedback: It is difficult for medical staff to obtain the patient's physiological data and recovery status in real time, making it impossible to detect abnormal situations in time and intervene; Strong subjectivity: The evaluation of the postoperative recovery situation mostly depends on the experience of medical staff and the subjective descriptions of patients, lacking objective and quantitative evaluation methods; Poor patient compliance: Patients may fail to detect abnormalities during the recovery process in a timely manner due to their own negligence or lack of medical knowledge, thus delaying treatment. Therefore, there is an urgent need for an intelligent monitoring system that can monitor the postoperative recovery of ophthalmic patients in real time, comprehensively, and objectively to solve the problem in the prior art that the postoperative recovery situation cannot be determined in a timely manner. For this reason, we propose an intelligent postoperative rehabilitation monitoring system for ophthalmic patients. Summary of the Invention

[0003] The main object of the present invention is to provide an intelligent postoperative rehabilitation monitoring system for ophthalmic patients, which can effectively solve the problems in the background art.

[0004] To achieve the above object, the technical solution adopted by the present invention is as follows: An intelligent postoperative rehabilitation monitoring system for ophthalmic patients, comprising: A wearable image acquisition module, configured to take pictures of the postoperative part of the patient's eye at a fixed sampling frequency to obtain the postoperative part image at the current moment; ; A client, which is communicatively connected to the wearable image acquisition module, configured to read the postoperative part image obtained by the wearable image acquisition module, or control the taking of pictures of the postoperative part of the patient's eye to obtain the postoperative part image at the current moment, and upload the obtained postoperative part image to an image processing server; the client is further configured to obtain the basic information and chief complaint information of the patient; An image processing server is used to perform information data fusion on the acquired postoperative part image information, basic information, and chief complaint information, obtain the recognition result of the postoperative part recovery situation of the patient, and feedback the obtained recognition result to the client.

[0005] The system includes a memory, a processor, and a computer program stored on the memory and executable on the processor.

[0006] Further, the process of information data fusion includes the following steps: Extract the first text feature of the chief complaint information of the patient to be detected through natural language processing technology , Denoted as the th text feature of the chief complaint information; Extract the second text feature of the basic information of the patient to be detected through natural language processing technology , Denoted as the th text feature of the basic information; Extract the image feature of the postoperative part image at the current moment through a convolutional neural network of the postoperative part image , Denoted as the th image feature of the postoperative part image at the moment; Perform weighted fusion on the image feature and the text feature to form a unified feature representation.

[0007] Further, the processing formula for the weighted fusion is: In the formula, is the feature vector after weighted fusion; , , are the weighting coefficients of the first text feature, the second text feature, and the image feature respectively, and + + = 1.

[0008] Further, the weighting coefficient is dynamically adjusted and set according to the following formula, specifically: = ; = ; = ; In the formula, is the Sigmoid function; is expressed as concatenating the first text feature, the second text feature, and the image feature together; , , are the weight matrices of the first text feature, the second text feature, and the image feature respectively; , , are the bias terms of the first text feature, the second text feature, and the image feature respectively.

[0009] Furthermore, the recognition process of the postoperative recovery status of the patient includes the following steps: Obtain the basic information, the chief complaint information, and the postoperative part image of the patients who have undergone the same type of eye surgery, and construct an information set of the same type of eye surgery; Extract the first text feature in the basic information of the patients who have undergone the same type of surgery in the said information set; Extract the second text feature in the chief complaint information of the patients who have undergone the same type of surgery in the said information set; Extract the image feature of the postoperative part image of the patients who have undergone the same type of surgery in the said information set through a convolutional neural network; Establish a first neural network model with the said first text feature as the input and the postoperative part recovery status level as the output, train the established first neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the first expected value; Establish a second neural network model with the said second text feature as the input and the postoperative part recovery status level as the output, train the established second neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the second expected value; Establish a third neural network model with the said image feature as the input and the postoperative part recovery status level as the output, train the established third neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the third expected value; Input the first text feature of the basic information of the patient to be detected into the constructed first neural network model to obtain the postoperative part recovery status level of the basic information of the patient to be detected ; Input the second text feature of the basic information of the patient to be detected into the constructed second neural network model to obtain the postoperative part recovery status level of the chief complaint information of the patient to be detected ; Input the postoperative part image of the patient to be detected at the current moment The image features are input into the constructed third neural network model to extract and obtain the postoperative site image of the patient to be detected at the current moment of the postoperative site recovery status level ; The weighted coefficient is used to perform weighted processing on the postoperative site recovery status level , , , and according to the processing result, the postoperative site recovery status level of the patient to be detected is determined , where the formula for the status level is: = , represents the floor operation on the value of × + × + × .

[0010] Furthermore, the formula for the expected value is, where represents the th expected value; represents the number of input samples of the th neural network model; represents the output function of the th neural network model; represents the th output sample of the th neural network model, = 1, 2, 3.

[0011] The present invention has the following beneficial effects: Compared with the prior art, through the wearable image acquisition module, the postoperative site of the patient's eye is photographed at a fixed sampling frequency to obtain the postoperative site image at the current moment Meanwhile, collect the basic information and chief complaint information of the patient, use the client to upload the acquired real-time images to the image processing server, extract the first text feature and the second text feature of the chief complaint information of the patient to be detected through natural language processing technology, extract image features through a convolutional neural network, perform information data fusion on the acquired postoperative part image information, basic information and chief complaint information, and construct a neural network model. Identify the postoperative part recovery status level corresponding to the feature quantity through the neural network, perform weighted processing on the postoperative part recovery status level using a weighting coefficient, determine the postoperative part recovery status level of the patient to be detected, and realize real-time, comprehensive and objective monitoring of the postoperative recovery of ophthalmic patients by using big data and machine learning technologies, avoid the phenomenon of inaccurate evaluation results caused by subjective judgment, and at the same time, keep track of the patient's recovery situation in real time and detect abnormalities in the recovery process of the patient's postoperative part in a timely manner. Brief Description of the Drawings

[0012] Figure 1 It is a schematic working flow diagram of an intelligent postoperative rehabilitation monitoring system for ophthalmic patients according to the present invention; Figure 2 It is a schematic structural diagram of an intelligent postoperative rehabilitation monitoring system for ophthalmic patients according to the present invention. Detailed Embodiments

[0013] The present invention will be further described below in conjunction with the detailed embodiments. Among them, the drawings are only used for illustrative purposes, showing only schematic diagrams, not physical diagrams, and should not be construed as a limitation to the present invention. In order to better illustrate the detailed embodiments of the present invention, some components in the drawings will be omitted, enlarged or reduced, and do not represent the dimensions of the actual product.

[0014] An intelligent postoperative rehabilitation monitoring system for ophthalmic patients includes: A wearable image acquisition module, which is used to photograph the postoperative part of the patient's eye at a fixed sampling frequency to obtain the postoperative part image at the current moment ; A client, which is communicatively connected to the wearable image acquisition module, is used to read the postoperative part image acquired by the wearable image acquisition module, or control the photographing of the postoperative part of the patient's eye to obtain the postoperative part image at the current moment, and upload the acquired postoperative part image to the image processing server; the client is also used to obtain the basic information and chief complaint information of the patient; An image processing server, which is used to perform information data fusion on the acquired postoperative part image information, basic information and chief complaint information, obtain the recognition result of the postoperative part recovery situation of the patient, and feedback the obtained recognition result to the client.

[0015] The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0016] The specific implementation process of the technical solution of the present invention includes the following steps: Step 1: Obtain the basic information, chief complaint information, and postoperative part images of patients undergoing the same type of eye surgery, and construct an information set of the same type of eye surgery; Among them, the basic information may include the patient's gender, age, surgery time, surgery type, postoperative time, types of medications, dosage of medications, etc.; the chief complaint information parameters include the onset situation, disease process, diagnosis and treatment situation, etc.

[0017] Step 2: Extract the first text features in the basic information of patients undergoing the same type of surgery in the information set; extract the second text features in the chief complaint information of patients undergoing the same type of surgery in the information set; extract the image features of the postoperative part images of patients undergoing the same type of surgery in the information set through a convolutional neural network.

[0018] Step 3: Establish a first neural network model with the first text features as the input and the postoperative part recovery status level as the output, train the established first neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the first expected value.

[0019] It should be noted that the postoperative part recovery status level can be divided according to the clinical experience of medical staff. By selecting representative information in the information set, taking the image information as an example, the postoperative process is divided into several stages according to the clinical experience of medical staff. Representative postoperative part images in each stage are respectively selected in the information set. By extracting the feature quantities in the images and performing mathematical processing on the obtained feature quantities, such as calculating the average value or weighted value, as the representative feature quantity of this recovery stage, corresponding to a recovery status level; the division of the recovery status level for the chief complaint information and basic information is the same, which will not be elaborated here.

[0020] Step 4: Establish a second neural network model with the second text features as the input and the postoperative part recovery status level as the output, train the established second neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the second expected value.

[0021] Step 5: Establish a third neural network model with the image features as the input and the postoperative part recovery status level as the output, train the established third neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the third expected value.

[0022] Step 6: Take pictures of the patient's postoperative eye area at a fixed sampling frequency to obtain the postoperative area image at the current moment, and obtain the basic information and chief complaint information of the patient when the postoperative area image is acquired.

[0023] Step 7: Use the client to upload the obtained postoperative area image , the basic information and chief complaint information of the patient to the image processing server.

[0024] Step 8: Extract the first text feature of the chief complaint information of the patient to be detected through natural language processing technology , denoted as the th text feature of the chief complaint information; extract the second text feature of the basic information of the patient to be detected through natural language processing technology , denoted as the th text feature of the basic information; extract the image feature of the postoperative area image at the current moment through a convolutional neural network, , denoted as the th image feature of the postoperative area image at the moment.

[0025] Step 9: Perform weighted fusion on the image feature and the text feature to form a unified feature representation. Among them, the processing formula for weighted fusion is: In the formula, is the feature vector after weighted fusion; , , are the weighted coefficients of the first text feature, the second text feature, and the image feature respectively, and + + = 1.

[0026] The weighted coefficients are dynamically adjusted and set according to the following formula, specifically: = ; = ; = ; In the formula, is the Sigmoid function; It is expressed as splicing the first text feature, the second text feature, and the image feature together; , , are the weight matrices of the first text feature, the second text feature, and the image feature respectively; , , are the bias terms of the first text feature, the second text feature, and the image feature respectively.

[0027] Step 10: Input the first text feature of the basic information of the patient to be detected into the constructed first neural network model to obtain the postoperative part recovery status level of the basic information of the patient to be detected .

[0028] Step 11: Input the second text feature of the basic information of the patient to be detected into the constructed second neural network model to obtain the postoperative part recovery status level of the chief complaint information of the patient to be detected .

[0029] Step 12: Input the image feature of the postoperative part image of the patient to be detected at the current moment into the third neural network model constructed, and extract and obtain the postoperative part recovery status level of the postoperative part image of the patient to be detected at the current moment of the postoperative part image of the patient to be detected .

[0030] Step 13: Use the weighting coefficient to perform weighted processing on the postoperative part recovery status levels , , , and determine the postoperative part recovery status level of the patient to be detected according to the processing result , where the calculation formula of the status level is: = , represents the floor operation on the value of × + × + × .

[0031] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and what is described in the above embodiments and the specification is only to illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent monitoring system for postoperative rehabilitation of ophthalmic patients, characterized in that, Including: A wearable image acquisition module is used to photograph the postoperative part of the patient's eye at a fixed sampling frequency to obtain the postoperative part image at the current moment ; The client, which is communicatively connected to the wearable image acquisition module, is used to read the postoperative site image obtained by the wearable image acquisition module, or control the photographing of the postoperative site of the patient's eye to obtain the postoperative site image at the current moment, and upload the obtained postoperative site image to the image processing server; the client is also used to obtain the basic information and the chief complaint information of the patient; An image processing server, configured to perform information data fusion on the acquired postoperative part image information, basic information, and chief complaint information, obtain an identification result of the postoperative part recovery situation of the patient, and feedback the obtained identification result to the client.

2. The intelligent monitoring system for postoperative rehabilitation of ophthalmic patients according to claim 1, characterized in that: The process of information data fusion includes the following steps: Extract the first text feature of the patient's chief complaint information to be detected through natural language processing technology , Indicates the th text feature of the chief complaint information; Extracting the second text feature of the basic information of the patient to be detected through natural language processing technology , which is expressed as the th text feature of the basic information; Extract the current postoperative site image at the moment image features of , denoted as the postoperative site image at the moment the th item of image features; Perform weighted fusion on the image features and text features to form a unified feature representation.

3. An intelligent postoperative rehabilitation monitoring system for ophthalmic patients according to claim 2, characterized in that: The processing formula for the weighted fusion is: In the formula, is the feature vector after weighted fusion; , , are the weighted coefficients of the first text feature, the second text feature, and the image feature respectively, and + + = 1.

4. An intelligent monitoring system for postoperative rehabilitation of ophthalmic patients according to claim 3, characterized in that: The weighting coefficient is dynamically adjusted and set according to the following formula, specifically: = ； = ； = ； In the formula, is the Sigmoid function; represents the concatenation of the first text feature, the second text feature, and the image feature; , , are the weight matrices of the first text feature, the second text feature, and the image feature respectively; , , are the bias terms of the first text feature, the second text feature, and the image feature respectively.

5. An intelligent monitoring system for postoperative rehabilitation of ophthalmic patients according to claim 1 or 2 or 3, characterized in that: The identification process of the postoperative part recovery situation of the patient includes the following steps: Obtain the basic information, chief complaint information, and postoperative part images of patients who have undergone the same type of eye surgery, and construct an information set of the same type of eye surgery; Extract the first text feature in the basic information of the patients who have undergone the same type of surgery in the information set; Extract the second text feature in the chief complaint information of the patients who have undergone the same type of surgery in the information set; Extract the image features of the postoperative part images of the patients who have undergone the same type of surgery in the information set through a convolutional neural network; Establish a first neural network model with the first text feature as the input and the postoperative part recovery status level as the output, train the established first neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the first expected value; Establish a second neural network model with the second text feature as the input and the postoperative part recovery status level as the output, train the established second neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the second expected value; Establish a third neural network model with the image feature as the input and the postoperative part recovery status level as the output, train the established third neural network model, and adjust the number of hidden layers of the model according to the error between the training result and the actual alarm result until the accuracy is not lower than the third expected value; Input the first text feature of the basic information of the patient to be detected into the constructed first neural network model to obtain the postoperative part recovery status level of the basic information of the patient to be detected ; Input the second text feature of the basic information of the patient to be detected into the constructed second neural network model to obtain the postoperative part recovery status level of the chief complaint information of the patient to be detected ; Input the image features of the postoperative site of the patient to be detected at the current moment into the third neural network model constructed, and extract and obtain the postoperative site recovery status level of the postoperative site image of the patient to be detected at the current moment of the postoperative site image ; ; Using a weighting coefficient to grade the recovery status of the postoperative site , , Perform weighted processing, and determine the recovery status grade of the postoperative site of the patient to be tested according to the processing result , where the status grade The calculation formula is: = , Indicates the floor operation on the value of × + × + × .

6. The intelligent monitoring system for postoperative rehabilitation of ophthalmic patients according to claim 5, wherein: The calculation formula for the expected value is in, Indicates Expected value; Indicates The number of input samples for a neural network model; Indicates The output function of a neural network model; Indicates The neural network model output samples, =1,2,3.

7. An intelligent monitoring system for postoperative rehabilitation of ophthalmic patients according to claim 1, characterized in that: The system includes a memory, a processor, and a computer program stored on the memory and executable on the processor. Among them, when the processor executes the program, it can implement the steps of the process described in any one of claims 2 and 5.

Images