CN108538393B - Bone quality assessment expert system based on big data and prediction model establishing method - Google Patents
Bone quality assessment expert system based on big data and prediction model establishing method Download PDFInfo
- Publication number
- CN108538393B CN108538393B CN201810283270.6A CN201810283270A CN108538393B CN 108538393 B CN108538393 B CN 108538393B CN 201810283270 A CN201810283270 A CN 201810283270A CN 108538393 B CN108538393 B CN 108538393B
- Authority
- CN
- China
- Prior art keywords
- data
- prediction
- user
- osteoporosis
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H80/00—ICT specially adapted for facilitating communication between medical practitioners or patients, e.g. for collaborative diagnosis, therapy or health monitoring
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention discloses a bone quality assessment expert system based on big data and a prediction model establishing method, wherein the bone quality assessment expert system comprises a server, a database and an intelligent terminal, wherein the server is internally provided with a prediction model module, a data acquisition module and a bone expert module; the database stores user identity data, characteristic data of historical osteoporosis fracture patients and detection data of patients to be predicted; the prediction model module establishes an osteoporosis fracture prediction model; the data acquisition module is used for acquiring a prediction request and detection data of a patient to be predicted, and the osteoporosis fracture prediction model obtains an osteoporosis fracture prediction grade according to the detection data; the bone expert module outputs a prevention suggestion to the patient to be predicted. Has the advantages that: and the fracture prediction of osteoporosis patients is realized. Is convenient. No queuing is required. And all can operate anytime and anywhere, intelligence, it is convenient.
Description
Technical Field
The invention relates to the technical field of big data, in particular to a bone quality assessment expert system based on big data and a prediction model establishing method.
Background
Osteoporosis is a disease characterized by decreased bone mass, degeneration of microstructure of bone tissue, increased bone fragility, decreased bone strength, and increased risk of fracture, and has become a serious problem threatening physical and mental health of the elderly. With the advent of aging society, patients with osteoporosis are gradually increased, and corresponding osteoporotic fractures are increased. The increase in mortality, disability rate, teratogenicity rate and medical costs due to osteoporotic fracture and its complications are becoming more and more serious for the burden on the family and society, and are therefore particularly important for the prevention of osteoporotic fracture.
Currently, Bone Mineral Density (BMD) is an important means for detecting osteoporosis and predicting osteoporotic fractures. However, many patients have observed that some patients have bone density that fails to meet the predictive criteria for osteoporosis, but have suffered from brittle fractures. Some patients have low bone density and no fracture. Studies have shown that only 10% to 44% of fractures occur in people with BMD that meet the criteria for osteoporosis prediction. Therefore, the BMD alone cannot completely predict the osteoporosis fracture, so that a plurality of patients at high risk of fracture are caused, the serious hidden danger of the life safety of the patients is not realized, the long treatment time is missed, and finally the fracture is caused.
In summary, the prediction quoted about the bone fracture caused by osteoporosis not only includes bone density, but also includes many internal and external factors, however, in the prior art, no prediction standard is provided for unknown patients to refer to, and the individual differentiation of the patients is large, and the prediction standard of the osteoporosis fracture is difficult to convince the patients.
Disclosure of Invention
Aiming at the problems, the invention provides a bone quality assessment expert system based on big data and a prediction model establishing method, based on data of patients with osteoporosis fracture, the bone characteristic data is analyzed and trained to obtain an osteoporosis fracture model, any user analyzes the individual bone quality through the osteoporosis fracture model and obtains corresponding fracture prevention suggestions, reference suggestions are provided for the patients, and the patients can conveniently analyze and monitor the bone in real time.
In order to achieve the purpose, the invention adopts the following specific technical scheme:
a bone quality assessment expert system based on big data, the key technology of which lies in: the bone mass prediction system comprises a server, wherein the server is provided with a prediction model module, a data acquisition module and a bone mass expert module, and a database and N intelligent terminals are connected to the server; n is a positive integer greater than or equal to 1.
The database is used for storing identity data of all users, characteristic data of historical osteoporosis fracture patients and detection data of all patients to be predicted;
the prediction model module establishes an osteoporosis fracture prediction model according to the characteristic data;
the data acquisition module is used for acquiring a prediction request sent by any user, acquiring detection data of a patient to be predicted from the database according to the prediction request, and inputting the detection data into the osteoporosis fracture prediction model to obtain the osteoporosis fracture prediction grade of the patient to be predicted;
the bone expert module gives a patient prevention suggestion to the patient to be predicted according to the osteoporosis fracture prediction grade of the patient to be predicted;
the intelligent terminal is provided with an osteoporosis fracture prediction APP, any user transmits detection data of a patient to be predicted to the data acquisition module through the osteoporosis fracture prediction APP, and acquires an anti-illness suggestion output by the bone expert module.
Through the scheme, any user inputs the detection data of the user into the osteoporosis fracture prediction model established by the prediction model module through the osteoporosis fracture prediction APP or the data acquisition module to obtain the osteoporosis fracture prediction grade of the user, and the bone quality of the user is monitored in real time. And the bone expert module provides a fracture prevention suggestion for the user according to the osteoporosis fracture prediction grade to remind the user of self-protection. The user can acquire the self bone information at any time, and can remotely acquire expert suggestions for prevention in advance. The fracture risk of the user is reduced, the whole process time for detecting and obtaining the suggestion is short, and queuing is not needed. And all can operate anytime and anywhere, intelligence, it is convenient.
Further, the identity data of the user at least comprises the name and the certificate number of the user;
the characteristic data and the detection data at least comprise sex, age, height, weight, bone density, albumin content, calcium content, phosphorus content, alkaline phosphatase content, hemoglobin content, lymphocyte content, grade, motion amount and 3D walking track.
The identity data of the user is communicated with the public security system and the medical system, when the user registers the system user, the identity of the user needs to be verified, and the medical information of the user can be acquired under the condition that the user agrees. And the characteristic data and the detection data contain surgical data and medical data, and the medical data are applied to surgical bone quality detection, so that the surgical detection data and the medical detection data are combined, and the model prediction model is more accurate. And all data acquisition is convenient and fast, and the detection cost is low.
Further, the amount of motion and the 3D walking trajectory are obtained by the intelligent terminal;
the amount of exercise includes at least a number of walking steps;
the 3D walking track at least comprises a horizontal moving track and an altitude moving track.
By adopting the scheme, the application program for acquiring the user motion amount is installed on the intelligent terminal; or the authority of the amount of exercise is obtained in the osteoporosis fracture prediction APP zone, the amount of exercise of the user is obtained through the intelligent terminal, and data acquisition of the amount of exercise is achieved. The intelligent and convenient.
Further, any user or through the osteoporosis fracture prediction APP sends a prediction request to the data acquisition module;
either user or by sending a prediction request directly to the data collection module.
The user can send a prediction request to the system through the data acquisition module or the osteoporosis fracture prediction APP, and the prediction channel selected by the user is more.
Further describing, the osteoporosis fracture prediction APP at least comprises a user registration login unit, a detection data query unit, a detection data entry unit, a terminal prediction request unit and an expert suggestion unit; the user registration login unit is used for the intelligent terminal user to perform user registration or login; the detection data query unit is used for the intelligent terminal user to query the detection data of the intelligent terminal user; the detection data entry unit is used for the intelligent terminal user to enter detection data which is relative to the characteristic data loss; the terminal prediction request unit is used for an intelligent terminal user to send a prediction request to the data acquisition module; the expert suggestion unit is used for acquiring and displaying the prevention suggestions given by the bone expert module.
By adopting the scheme, the user can realize user login, detection data entry and bone prediction request and acquire prevention suggestions of experts through the user registration login unit, the detection data query unit, the detection data entry unit, the terminal prediction request unit and the expert suggestion unit.
Further described, the osteoporosis fracture prediction grade is graded according to the size of the fracture probability;
the prevention advice at least comprises exercise amount advice, diet advice, medication advice and auxiliary treatment advice, and the given standard of the prevention advice is the prediction grade of the osteoporosis fracture.
And (4) formulating the probability standard of osteoporosis and fracture according to the professional opinion of the bone medicine experts. And grading the osteoporosis fracture prediction of the patient according to the osteoporosis fracture probability standard to obtain the osteoporosis fracture prediction grade. And according to the prevention suggestions of the bone medical experts, which are specified aiming at the sick severity grades of the patients with different grades, the related prevention suggestions are provided for the corresponding clients. Through the system, the patient does not need to touch the doctor, and corresponding expert suggestions can be obtained by only inputting all detection data and sending out corresponding prediction requests at any place. The speed is fast, fast and convenient.
The big data-based bone quality assessment expert system prediction model building method according to claim 1, comprising the following steps:
s1: acquiring characteristic data of M groups of historical osteoporosis and fracture patients in the database;
s2: processing the acquired M groups of characteristic data to obtain processed data, and dividing the processed data into M1Training data and m2Group assessment data; m is more than or equal to M1+m2(ii) a The data division method adopts a 10-fold cross verification method.
S3: setting model parameters and an osteoporosis fracture prediction initial model based on a neural network;
s4: setting the training times to m1Substituting the training data into the osteoporosis fracture prediction initial model obtained in the step S3, and performing model training to obtain an osteoporosis fracture prediction training model;
s5: setting the evaluation conditions of the model, and calculating2Substituting the group evaluation data into the osteoporosis fracture prediction training model obtained in the step S4, and evaluating the obtained osteoporosis fracture prediction training model;
s6: if the model evaluation condition is met, the osteoporosis fracture prediction training model obtained in the step S4 is used as an osteoporosis fracture prediction model and output; if the evaluation result does not meet the model evaluation condition, the process returns to step S1.
The most common single hidden layer neural network is used for establishing an osteoporosis fracture prediction model, monitoring the fracture condition of an osteoporosis patient in real time, and reducing the osteoporosis probability. The life quality of the osteoporosis patients is ensured.
Further, the step of performing data processing on the M groups of feature data in step S2 at least includes data cleaning, data integration, data reduction, and data transformation;
the data cleaning at least comprises characteristic elimination and missing value filling, and the missing value filling adopts M-M3Carrying out shape filling on the corresponding characteristic mean value of the group characteristic data; m is3The number of the vacancies in any characteristic type;
and the data integration is data standardization processing, and the feature data of any feature type is scaled according to the data proportion, so that all the data fall into a specific interval. Wherein, M, m1、m2、m3Are all positive integers.
Further still, the setting of the model parameters in step S3 includes at least: the method comprises the following steps that input layer unit setting, hidden layer unit setting and output layer unit setting of a neural network are carried out; setting a hidden layer activation function and an output layer activation function; setting a characteristic data cost function; setting a characteristic data weight division function; the feature value of the feature data is set.
The invention has the beneficial effects that: the invention predicts the fracture probability of the osteoporosis patient by combining the medical detection data and the surgical detection data and applies the medical detection data to surgical medical prediction. By establishing a bone quality assessment expert system based on big data, osteoporosis patients can remotely realize bone quality prediction, and prevention suggestions of experts are obtained according to prediction grades, so that the patients can be prevented as early as possible, and the fracture risk of the patients is reduced. And establishing a prediction model by combining the internal and external detection data and adopting a neural network. The prediction effect is good, the precision is high, and the whole prediction system is reliable and safe. Provides a remote orthopedist for the patient.
Drawings
FIG. 1 is a system block diagram of the present invention;
FIG. 2 is a flow chart of the modeling of the present invention;
FIG. 3 is a graph of ROC plots for different nodes of the hidden layer of the present invention;
FIG. 4 is model prediction accuracy at different input data;
FIG. 5 is model prediction accuracy for different input data;
FIG. 6 is the area under the model prediction ROC curve for different input data;
FIG. 7 is a model prediction true positive rate at different input data;
FIG. 8 is a model predicted false positive rate for different input data;
FIG. 9 is a ROC curve for model prediction at different input data.
Detailed Description
The following provides a more detailed description of the embodiments and the operation of the present invention with reference to the accompanying drawings.
As can be seen from fig. 1, the bone quality assessment expert system based on big data comprises a server 1, wherein the server 1 is provided with a prediction model module 1a, a data acquisition module 1b and a bone expert module 1c, and a database 2 and N intelligent terminals 3 are connected to the server 1;
in this embodiment, the database 2 is used for storing the identity data of all users, the feature data of patients with historical osteoporosis and fracture, and the detection data of all patients to be predicted; the prediction model module 1a establishes an osteoporosis fracture prediction model according to the characteristic data; the data acquisition module 1b is used for acquiring a prediction request sent by any user, acquiring detection data of a patient to be predicted from the database 2 according to the prediction request, and inputting the detection data into the osteoporosis fracture prediction model to obtain the osteoporosis fracture prediction grade of the patient to be predicted; the bone expert module 1c gives a prevention suggestion to the patient to be predicted according to the osteoporosis fracture prediction grade of the patient to be predicted; an osteoporosis fracture prediction APP is installed on the intelligent terminal 3, any user transmits detection data of a patient to be predicted to the data acquisition module 1b through the osteoporosis fracture prediction APP, and an illness prevention suggestion output by the bone expert module 1c is obtained.
Further, the identity data of the user at least comprises the name and the certificate number of the user;
the characteristic data and the test data include at least Gender (Gender), Age (Age), Height (Height), Weight (Weight), bone density (BMD), albumin content (Alb), calcium content (Ca), phosphorus content (P), alkaline phosphatase content (Alp), hemoglobin content (Hb), lymphocyte content (lym), class (class), exercise amount, and 3D walking trajectory.
In the embodiment, the amount of motion and the 3D walking track are obtained through the intelligent terminal (3);
the amount of exercise includes at least a number of walking steps; the 3D walking track at least comprises a horizontal moving track and an altitude moving track. In the present embodiment, the motion amount is motion amount data acquired through third-party software, such as qq, a hundred-degree map, a running APP, and the like. In this embodiment, the collected data is derived from the electronic health record of the seventh people hospital in Chongqing City. The envelope includes 21631 cases. After data cleaning, data integration, data reduction and data transformation are carried out on the data, 5230 cases are obtained, wherein 788 cases are osteoporosis fracture cases. The final data set contained 5230 cases, most of which included 11 physical examination indices. As can be seen in table 1, for some of the characteristic data in the database:
TABLE 1 characteristic data Table
| Name (I) | Certificate number | Gender | Age | Height | Weight | BMD | Alb | Ca | P | Alp | Hb | Lym | Class |
| * | *** | 1 | 69 | 168 | 66 | -0.6346 | 40.8 | 2.14 | 0.96 | 107.68 | 1.42 | 115.63 | - |
| * | *** | 0 | 51 | -1.2825 | 46.4 | 2.49 | 1.09 | 151.6 | 1.59 | 148.5 | - | ||
| * | *** | 1 | 71 | 165 | 74 | -0.7323 | 108.5 | 1.82 | 141 | - | |||
| * | *** | 0 | 69 | 172 | 79 | -1.5923 | 40.5 | 2.34 | 1.18 | 89.7 | 2.17 | 137 | - |
| * | *** | 0 | 84 | 150 | 42 | -3.2427 | 42.8 | 2.44 | 1.3 | 140.5 | 1.01 | 117 | - |
| * | *** | 0 | 76 | -3.3952 | 37.23 | 2.11 | 0.94 | 94.5 | 1.76 | 122.75 | - | ||
| * | *** | 0 | 85 | -3.2941 | 29.77 | 1.98 | 0.67 | 98.5 | 0.73 | 88.89 | - | ||
| * | *** | 0 | 64 | 158 | 55 | -3.6052 | 39.75 | 98.1 | 0.88 | 97 | - | ||
| * | *** | 0 | 71 | 150 | 50 | -3.0520 | 46.5 | 2.34 | 1.14 | 77.2 | 1.66 | 130 | - |
| * | *** | 1 | 74 | 167 | 59 | -1.0450 | 40.9 | 2.28 | 0.75 | 84.7 | 1.39 | 125.67 | - |
| * | *** | 1 | 86 | 160 | 65 | -0.8104 | 43.3 | 2.32 | 1.16 | 71.1 | 1.49 | 119.67 | - |
| * | *** | 0 | 89 | 150 | 35 | -2.8167 | 41.5 | 2.27 | 1.16 | 97.4 | 1.52 | 113 | - |
| * | *** | 1 | 57 | 44.6 | 2.32 | 0.73 | 114.5 | ||||||
| * | *** | 0 | 82 | 29.75 | 1.98 | 1.95 | 129.88 | 2.1 | 110.14 | - | |||
| ... | ... | ... | ... | ... | ... | ... | .... | ... | ... | .... | .... | ... | ... |
The name and the certificate number are user privacy information, and the authority to be inquired needs to be acquired. The data has missing and obvious error data, and data processing is needed.
In the table, gender 0 is gender girl, and gender 1 is gender male.
In this embodiment, any user sends a prediction request to the data acquisition module 1b, either through the osteoporosis fracture prediction APP;
either user or by sending a prediction request directly to the data acquisition module 1 b.
In this embodiment, the osteoporosis fracture prediction APP at least includes a user registration login unit, a detection data query unit, a detection data entry unit, a terminal prediction request unit, and an expert suggestion unit;
the user registration login unit is used for the intelligent terminal user to perform user registration or login;
the detection data query unit is used for the intelligent terminal user to query the detection data of the intelligent terminal user;
the detection data entry unit is used for the intelligent terminal user to enter detection data which is relative to the characteristic data loss; the terminal prediction request unit is used for an intelligent terminal user to send a prediction request to the data acquisition module 1 b; the expert advice unit is used for acquiring and displaying the prevention advice given by the bone expert module 1 c.
In the embodiment, the osteoporosis fracture prediction grade is graded according to the size of fracture probability;
the prevention advice comprises exercise amount advice, diet advice, medication advice and auxiliary treatment advice, and the given standard of the prevention advice is the osteoporosis fracture prediction grade.
In this example, the osteoporotic fracture prediction levels were classified into four grades: mild, moderate, severe, and dangerous;
and specific prevention suggestion:
lightness: strengthening nutrition and balancing diet; sufficient sunshine; regular movement; stop smoking, limit alcohol consumption, and avoid or use less medicine for affecting bone metabolism.
Medium: ingestion of calcium agents and vitamin D; and to enhance nutrition and balance diet; sufficient sunshine; regular movement; stop smoking, limit alcohol consumption, and avoid or use less medicine for affecting bone metabolism.
And (3) severe degree: taking bone absorption resisting medicine, bone formation promoting medicine, and other medicines. Ingestion of calcium agents and vitamin D; and to enhance nutrition and balance diet; sufficient sunshine; regular movement; stop smoking, limit alcohol consumption, and avoid or use less medicine for affecting bone metabolism.
The degree of risk; monitoring bone metabolism markers, monitoring bone density and monitoring bone imaging; anti-bone resorption drugs, bone formation promoting drugs, other drugs; ingestion of calcium agents and vitamin D; and has effects in supplementing nutrition, balancing diet, providing sufficient sunshine, promoting regular exercise, stopping smoking, limiting alcohol consumption, and avoiding or reducing drug influence on bone metabolism.
A big data-based prediction model building method of a bone quality assessment expert system is disclosed, and can be seen by combining figure 2, and the method comprises the following steps:
s1: acquiring characteristic data of M groups of historical osteoporosis and fracture patients in the database 2;
s2: processing the acquired M groups of characteristic data to obtain processed data, and dividing the processed data into M1Training data and m2Group assessment data; m is more than or equal to M1+m2；
S3: setting model parameters and an osteoporosis fracture prediction initial model based on a neural network; in the present embodiment, a BP neural network is employed.
S4: setting the training times to m1Substituting the training data into the osteoporosis fracture prediction initial model obtained in the step S3, and performing model training to obtain an osteoporosis fracture prediction training model;
s5: setting the evaluation conditions of the model, and calculating2Substituting the group evaluation data into the osteoporosis fracture prediction training model obtained in the step S4, and evaluating the obtained osteoporosis fracture prediction training model;
s6: if the model evaluation condition is met, the osteoporosis fracture prediction training model obtained in the step S4 is used as an osteoporosis fracture prediction model and output; if the evaluation result does not meet the model evaluation condition, the process returns to step S1.
The step of performing data processing on the M groups of feature data in step S2 at least includes data cleaning, data integration, data reduction, and data transformation;
the data cleaning at least comprises characteristic elimination and missing value filling, and the missing value filling adopts M-M3Carrying out shape filling on the corresponding characteristic mean value of the group characteristic data; m is3The number of the vacancies in any characteristic type.
And the data integration is data standardization processing, and the feature data of any feature type is scaled according to the data proportion, so that all the data fall into a specific interval.
Setting the model parameters in step S3 includes:
the method comprises the following steps that input layer unit setting, hidden layer unit setting and output layer unit setting of a neural network are carried out;
setting a hidden layer activation function and an output layer activation function;
setting a characteristic data cost function;
setting a characteristic data weight division function;
the feature value of the feature data is set.
In this embodiment, the input features have 11 dimensions, and the classes to be classified have two classes, so the structure of the neural network is 11-100-1, that is, the input layer has 11 nodes, the hidden layer has 80-120 nodes, and the output layer has 1 node. The hidden layer activation function is a ReLu function, and the output layer activation function is a Logistic function. The learning rate was set to 0.001 and the number of training was 10000. A 10-fold cross-validation method is used to partition the training data and the evaluation data.
In the present embodiment, in the proposed model, the number of inputs is set to 11, the neurons in the hidden layer are 100 and the output neurons are 1.
Definition VijAs a weighting parameter between the input layer and the hidden layer; w is ajkAs a weighting parameter between the hidden layer and the output layer; wherein i is 0,1,2,. 11; 1,2, 100; k is 1.
aj(j ═ 1, 2.., 11) denotes the threshold of the neurons of the input layer; bj(j ═ 1, 2.., 100) represents the threshold of the neurons of the hidden layer; c. Cj(j ═ 1) represents a threshold value of a neuron of the output layer; for training sample X (X)1，x2，...，x11) We can get the output as follows:
wherein ZkIs the output value corresponding to sample x; f. of1And f2Is the activation function. When calculating ZkWe can adjust the back propagation of weight parameter errors in the model. After training is completed, the model can be used.
Since the number of cases in the dataset is not large enough, a 10-fold cross-validation strategy is required for the experiments. In a BP neural network, the Relu function is set as a hidden layer activation function f1(ii) a The Relu function is calculated as: f (x) max (0, x); the learning rate and the number of training times were set to 0.001 and 1000, respectively.
Activation function f of output layer2Selecting a sigmoid function, wherein the calculation formula is as follows:
the classification threshold is 0.5. The number of nodes in the hidden layer is determined by a trial and error method. Different nodes are set for the hidden layer and corresponding ROCs are calculated. As can be seen from fig. 3, the hidden layers are 80, 90, 100, 110 and 120 nodes, respectively, and the ROC curves (false positive rate on the horizontal axis and true positive rate on the vertical axis) are used to evaluate the models. In general, when the hidden layer nodes are set to be 100, the generalization performance of the model is better, and the number of the hidden layer nodes is set to be 100.
To protect patient privacy, we also predict outcomes by deleting patient gender and age information. The prediction results show that the prediction results for NA, NG and NAG are still present and the results are acceptable. See figures 4-9 for details.
In fig. 4 to 9, NA is a prediction result in which the input information does not include age information; NG is a prediction result that the input information does not contain sex information; NGA is a prediction result of input information which does not contain age information and gender information; normal is the prediction result containing all the input information.
Wherein: the five metrics are: accuracy Prediction Accuracy of Accuracy; area under AUC ROC curve: TPR (true positive rate); FPR (false positive rate) false positive rate; and the calculation formula is:
wherein, TP (true Positives): feature number of predicted positive sample and actually positive sample
FP (false positives): the characteristic number of the positive sample is predicted and the characteristic number of the negative sample is actually predicted;
TN (true neurons): predicting as a negative sample, actually as a characteristic number of the negative sample;
FN (false negatives): features of predicted negative samples and actually positive samples;
it should be noted that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make variations, modifications, additions or substitutions within the spirit and scope of the present invention.
Claims (4)
1. A bone quality assessment expert system based on big data is characterized in that: the bone mass prediction system comprises a server (1), wherein the server (1) is provided with a prediction model module (1a), a data acquisition module (1b) and a bone mass expert module (1c), and a database (2) and N intelligent terminals (3) are connected to the server (1);
the database (2) is used for storing identity data of all users, characteristic data of historical osteoporosis fracture patients and detection data of all patients to be predicted;
the prediction model module (1a) establishes an osteoporosis fracture prediction model according to the characteristic data;
the data acquisition module (1b) is used for acquiring a prediction request sent by any user, and acquiring detection data of a patient to be predicted from the database (2) according to the prediction request, wherein the detection data is input into the osteoporosis fracture prediction model to obtain the osteoporosis fracture prediction grade of the patient to be predicted;
the bone expert module (1c) gives a patient to be predicted a prevention advice according to the osteoporosis fracture prediction grade of the patient to be predicted;
an osteoporosis fracture prediction APP is installed on the intelligent terminal (3), any user transmits detection data of a patient to be predicted to the data acquisition module (1b) through the osteoporosis fracture prediction APP, and an illness prevention suggestion output by the bone expert module (1c) is obtained;
any user inputs the detection data of the user into the osteoporosis fracture prediction model established by the prediction model module through the osteoporosis fracture prediction APP or the data acquisition module to obtain the osteoporosis fracture prediction grade of the user, and monitors the bone quality of the user in real time; the bone expert module provides a fracture prevention suggestion for the user according to the osteoporosis fracture prediction grade to remind the user of self-protection; a user acquires self bone information and can remotely acquire expert advice to prevent in advance;
the identity data of the user at least comprises the name and the certificate number of the user;
the characteristic data and the detection data at least comprise sex, age, height, weight, bone density, albumin content, calcium content, phosphorus content, alkaline phosphatase content, hemoglobin content, lymphocyte content, grade, motion amount and 3D walking track;
the amount of motion and the 3D walking track are obtained through the intelligent terminal (3);
the amount of exercise includes at least a number of walking steps;
the 3D walking track at least comprises a horizontal moving track and an altitude moving track;
-any user or through an osteoporotic fracture prediction APP sends a prediction request to the data acquisition module (1 b);
any user or by sending a prediction request directly to the data acquisition module (1 b);
the osteoporosis fracture prediction APP at least comprises a user registration login unit, a detection data query unit, a detection data entry unit, a terminal prediction request unit and an expert suggestion unit;
the user registration login unit is used for the intelligent terminal user to perform user registration or login;
the detection data query unit is used for the intelligent terminal user to query the detection data of the intelligent terminal user;
the detection data entry unit is used for the intelligent terminal user to enter detection data which is relative to the characteristic data loss;
the terminal prediction request unit is used for an intelligent terminal user to send a prediction request to the data acquisition module (1 b);
the expert suggestion unit is used for acquiring and displaying the prevention suggestions given by the bone expert module (1 c);
the osteoporosis fracture prediction grade is graded according to the fracture probability;
the prevention advice at least comprises exercise amount advice, diet advice, medication advice and auxiliary treatment advice, and the given standard of the prevention advice is the prediction grade of the osteoporosis fracture.
2. The big data-based bone quality assessment expert system prediction model building method according to claim 1, characterized by the following steps:
s1: acquiring characteristic data of M groups of history osteoporosis and fracture patients in the database (2);
s2: processing the acquired M groups of characteristic data to obtain processed data, and dividing the processed data into M1Training data and m2Group assessment data; m is more than or equal to M1+m2；
S3: setting model parameters and an osteoporosis fracture prediction initial model based on a neural network;
s4: setting the training times to m1Substituting the training data into the osteoporosis fracture prediction initial model obtained in the step S3, and performing model training to obtain an osteoporosis fracture prediction training model;
s5: setting the evaluation conditions of the model, and calculating2Substituting the group evaluation data into the osteoporosis fracture prediction training model obtained in the step S4, and evaluating the obtained osteoporosis fracture prediction training model;
s6: if the model evaluation condition is met, the osteoporosis fracture prediction training model obtained in the step S4 is used as an osteoporosis fracture prediction model and output; if the evaluation result does not meet the model evaluation condition, the process returns to step S1.
3. The big-data-based bone quality assessment expert system prediction model building method according to claim 2, wherein the step of performing data processing on the M sets of feature data in step S2 at least comprises data cleaning, data integration, data reduction, data transformation;
the data cleaning at least comprises characteristic elimination and missing value filling, and the missing value filling adopts M-M3Carrying out shape filling on the corresponding characteristic mean value of the group characteristic data; m is3The number of the vacancies in any characteristic type;
and the data integration is data standardization processing, and the feature data of any feature type is scaled according to the data proportion, so that all the data fall into a specific interval.
4. The method for building a prediction model of an expert system for bone quality assessment based on big data according to claim 2, wherein the setting of model parameters in step S3 includes:
the method comprises the following steps that input layer unit setting, hidden layer unit setting and output layer unit setting of a neural network are carried out;
setting a hidden layer activation function and an output layer activation function;
setting a characteristic data cost function;
setting a characteristic data weight division function;
the feature value of the feature data is set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810283270.6A CN108538393B (en) | 2018-04-02 | 2018-04-02 | Bone quality assessment expert system based on big data and prediction model establishing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810283270.6A CN108538393B (en) | 2018-04-02 | 2018-04-02 | Bone quality assessment expert system based on big data and prediction model establishing method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108538393A CN108538393A (en) | 2018-09-14 |
| CN108538393B true CN108538393B (en) | 2021-12-10 |
Family
ID=63482730
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810283270.6A Active CN108538393B (en) | 2018-04-02 | 2018-04-02 | Bone quality assessment expert system based on big data and prediction model establishing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108538393B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111523038B (en) * | 2020-04-27 | 2022-12-27 | 重庆邮电大学 | Natural noise fuzzy management method and storage medium applied to recommendation system |
| CN111950636A (en) * | 2020-08-14 | 2020-11-17 | 暨南大学附属第一医院（广州华侨医院） | Deep learning-based lumbar X-ray image classification method |
| CN112164464A (en) * | 2020-11-03 | 2021-01-01 | 上海市同济医院 | Establishment method and application of osteoporosis risk prediction model for type 2 diabetes patients |
| CN112309578B (en) * | 2020-11-03 | 2021-06-15 | 南通市第一人民医院 | Method and system for improving recovery efficiency of osteoporotic vertebral fracture patient |
| CN113689078A (en) * | 2021-07-27 | 2021-11-23 | 中国科学院地理科学与资源研究所 | Survey data verification method and device |
| CN115331825A (en) * | 2022-07-26 | 2022-11-11 | 上海交通大学医学院附属新华医院 | Fracture risk prediction model based on big data multi-level feature mining |
| CN115661052B (en) * | 2022-10-13 | 2023-09-12 | 高峰医疗器械(无锡)有限公司 | Alveolar bone detection method, device, equipment and storage medium |
| CN116564526B (en) * | 2023-07-11 | 2023-09-19 | 天津市滨海新区中医医院 | Bone health assessment method, device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1194045A (en) * | 1995-07-25 | 1998-09-23 | 好乐思治疗公司 | Computer assisted methods for diagnosing diseases |
| CN103617110A (en) * | 2013-11-11 | 2014-03-05 | 国家电网公司 | Server device condition maintenance system |
| CN104146732A (en) * | 2014-07-29 | 2014-11-19 | 哈尔滨工业大学 | Sparse representation-based multi-parameter osteoporosis class judgment method for bone densitometer |
| CN104504297A (en) * | 2015-01-21 | 2015-04-08 | 甘肃百合物联科技信息有限公司 | Method for using neural network to forecast hypertension |
| CN107205709A (en) * | 2015-02-13 | 2017-09-26 | 株式会社岛津制作所 | Bone analytical equipment |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8852128B2 (en) * | 2008-03-12 | 2014-10-07 | University Of Cincinnati | Computer system and method for assessing dynamic bone quality |
| RU2511430C2 (en) * | 2012-04-19 | 2014-04-10 | Общество С Ограниченной Ответственностью "Парафарм" | Diagnostic technique for osteoporosis, method for testing cavitary lesion closure behaviour for assessing effective use of various osteoprotectors |
| CN104123445B (en) * | 2014-07-14 | 2017-01-11 | 中国科学院合肥物质科学研究院 | Exercise goal reasoning system based on multi-source information |
-
2018
- 2018-04-02 CN CN201810283270.6A patent/CN108538393B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1194045A (en) * | 1995-07-25 | 1998-09-23 | 好乐思治疗公司 | Computer assisted methods for diagnosing diseases |
| CN103617110A (en) * | 2013-11-11 | 2014-03-05 | 国家电网公司 | Server device condition maintenance system |
| CN104146732A (en) * | 2014-07-29 | 2014-11-19 | 哈尔滨工业大学 | Sparse representation-based multi-parameter osteoporosis class judgment method for bone densitometer |
| CN104504297A (en) * | 2015-01-21 | 2015-04-08 | 甘肃百合物联科技信息有限公司 | Method for using neural network to forecast hypertension |
| CN107205709A (en) * | 2015-02-13 | 2017-09-26 | 株式会社岛津制作所 | Bone analytical equipment |
Non-Patent Citations (2)
| Title |
|---|
| "单中心MHD患者骨质疏松患病情况、危险因素调查及骨折和跌倒风险预测";张燕妮;《中国优秀硕士学位论文全文数据库 医药卫生科技辑》;20151215(第12期);E067-30，正文第6-8、30、47页 * |
| "骨质疏松治疗仪的诊断及疗效评估***的研究";杨紫娟;《中国优秀硕士学位论文全文数据库 信息科技辑》;20110915(第09期);I140-79，正文第9、22-24、28页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108538393A (en) | 2018-09-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108538393B (en) | Bone quality assessment expert system based on big data and prediction model establishing method | |
| US20230054751A1 (en) | Method and apparatus for classifying electroencephalogram signal, method and apparatus for training classification model, and electronic device and computer-readable storage medium | |
| CN114830132A (en) | System and method for processing human-related data including physiological signals to make context-aware decisions using distributed machine learning of edges and clouds | |
| US20140316220A1 (en) | Personal Health Monitoring System | |
| Gholipour et al. | Using an artificial neural networks (ANNs) model for prediction of intensive care unit (ICU) outcome and length of stay at hospital in traumatic patients | |
| WO2012063166A1 (en) | Method of continuous prediction of patient severity of illness, mortality, and length of stay | |
| CN116597954B (en) | Intelligent home-based remote diagnosis and treatment appointment consultation system | |
| CN116472591A (en) | Techniques for generating predictive outcomes associated with spinal muscular atrophy using artificial intelligence | |
| CN111000569A (en) | Intelligent cognitive monitoring system for abnormal blood sugar | |
| Deepika et al. | Iot based elderly monitoring system | |
| CN115662631A (en) | AI intelligence discrimination-based nursing home management system | |
| McGregor | A platform for real-time online health analytics during spaceflight | |
| Liu et al. | Health care data analysis and visualization using interactive data exploration for sportsperson | |
| CN116453641B (en) | Data processing method and system for auxiliary analysis information of traditional Chinese medicine | |
| WO2024027438A1 (en) | Personalized state-space progression model-based assisted decision-making system for disease | |
| Stone et al. | Automated health alerts from kinect-based in-home gait measurements | |
| Frize et al. | Effectiveness of a hybrid pattern classifier for medical applications | |
| Chang et al. | Using machine learning algorithms in medication for cardiac arrest early warning system construction and forecasting | |
| AU2020235557B2 (en) | Digital solutions for differentiating asthma from COPD | |
| Verma et al. | Methodologies, Applications, and Challenges of Pneumonia Detection of Chest X-Ray images for COVID-19 using IoT-enabled Deep Learning | |
| Su et al. | Cloud computing management architecture for digital health remote patient monitoring | |
| Chinniah et al. | ICD 10 Based Medical Expert System Using Fuzzy Temporal Logic | |
| CN117379009B (en) | Early warning and monitoring system for critical cardiovascular and cerebrovascular diseases of old people in community | |
| Agarwal et al. | Diseases prediction and diagnosis system for healthcare using IoT and machine learning | |
| Verkoeijen | Identifying Risk Factors and Readmission to the Hospital Using Joint Modeling in Telemonitored Data of Patients with Heart Failure |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |