CN104392145A - Placental implantation prediction method based on hidden Markov model - Google Patents

Abstract

The present invention relates to a method for predicting placenta accreta based on a hidden Markov model, comprising the following steps: the first step: extracting maternal data, obtaining x pieces of maternal data from a hospital or health organization, and the observation object of a piece of maternal data includes cesarean section Birth history, placenta previa, maternal age, multiparity history and placental position, x=200; the second step: the initial data of the hidden Markov model; the third step: the initial data of the hidden Markov model Based on the prediction of the type of placenta accreta in the case. The placenta accreta prediction method based on the hidden Markov model described in the present invention makes up for the limitations and deficiencies of the existing MRI diagnosis, such as thermal effect, susceptibility to fetal movement interference, no unified diagnostic standard, high cost, etc., so that the type of placenta accreta in pregnant women The prediction is more accurate, making a great contribution to protecting the lives of pregnant women and fetuses.

Description

Prediction method of placenta accreta based on hidden Markov model

technical field

The invention relates to a method for predicting placenta accreta based on a hidden Markov model.

Background technique

In recent years, due to the increase in the rate of cesarean section and other factors, the incidence of placenta accreta is gradually increasing, and the rate of death of pregnant women caused by complications of placenta accreta is also increasing, which seriously threatens the lives of pregnant women. , The diagnosis and treatment of placenta accreta has always been a hot and difficult issue in obstetrics and gynecology research.

Placenta accreta refers to abnormal invasive implantation of the placenta, where the placental villi directly penetrate the decidua basalis to reach or invade the myometrium. Placenta accreta is a pathologically diagnosed disease, and prepartum diagnosis mainly relies on medical history (high risk factors) and imaging examinations. In recent years, MRI diagnosis has been used. There are some limitations and shortcomings in MRI diagnosis. ① Thermal effect ② Easy to be disturbed by fetal movement ③ No unified diagnostic standard ④ Poor repeatability ⑤ High cost high.

Contents of the invention

The purpose of the present invention is to provide a method for predicting placenta accreta based on a hidden Markov model, so as to make up for the limitation and deficiency of MRI diagnosis and discrimination of placenta accreta.

To achieve the above object, the present invention adopts the following technical solutions: a method for predicting placenta accreta based on hidden Markov model, characterized in that it comprises the following steps:

The first step: extract maternal data, obtain x pieces of maternal data from hospitals or health organizations, and the observation objects of a piece of maternal data include history of cesarean section, placenta previa, maternal age, history of multiple births and placental position, 200≤x≤400 ;

The second step: establish the initial data of the hidden Markov model;

Step 3: Predict the type of placenta accreta of the case on the basis of the initial data of the hidden Markov model.

In one embodiment of the present invention, a piece of maternal data in the first step includes 15 stages, and two adjacent stages are separated by one week, which is recorded as , where 1≤n≤x.

In one embodiment of the present invention, the specific steps of establishing the initial data of the hidden Markov model in the second step are as follows:

Observation sequence: {Cesarean section history, placenta previa, maternal age, multiparity history, placental position} recorded as ;

State sequence: {normal type, adhesion type, implantation type, penetration type} is recorded as ;

Initial state probability distribution: denoted as ,in , , , ;

State transition probability matrix: , ,take out , extract the feature data of each stage, for Medium state The state transition of the state transition is counted, and the remaining n-1 pieces of data are processed in the same way, and the obtained state transition is carried out for probability statistics, so as to obtain the state transition probability matrix;

Observe the state transition matrix: , ,take out , extract the characteristic data of each stage, divide them into four categories with the state sequence q as the division standard, and perform the same processing on the remaining n-1 pieces of maternal data, and make statistics on the data of the four categories to obtain Observation state transition matrix;

Model parameters: .

In one embodiment of the present invention, the specific steps for the third step to predict the type of placenta accreta of the case are as follows: Perform iterative operations, and finally obtain the probability distribution of the initial state The probability of placenta accreta is used to assist in judging the type of placenta accreta. In order to realize the type prediction process, it is necessary to prepare the state transition probability matrix A and the value of the observed state transition matrix B to perform iterative calculations to obtain the final transition state probability. The initial value of the process is:

, ;

Then perform iterative operation to get , where N is the number of child elements of state q;

Finally, the probability of the sequence is calculated by the above formula, and the type of placenta accreta of the examiner is predicted by this.

Compared with the prior art, the present invention has the following beneficial effects: compared with MRI diagnosis, the method of the present invention does not produce thermal effects, does not have unpredictable effects on the fetus in the abdomen, and does not cost a lot, and the result is also more precise.

Description of drawings

Fig. 1 is the algorithm flow chart of the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

Please refer to Fig. 1, the present invention provides a kind of placenta accreta prediction method based on Hidden Markov Model, it is characterized in that comprising the following steps:

The first step: extract maternal data, obtain x pieces of maternal data from hospitals or health organizations, and the observation objects of a piece of maternal data include history of cesarean section, placenta previa, maternal age, history of multiple births and placental position, 200≤x≤400 ;

The second step: establish the initial data of the hidden Markov model;

Step 3: Predict the type of placenta accreta of the case on the basis of the initial data of the hidden Markov model.

In this embodiment, a piece of maternal data in the first step includes 15 stages, and two adjacent stages are separated by one week, which is recorded as , where 1≤n≤x.

In this embodiment, the specific steps for establishing the initial data of the hidden Markov model in the second step are as follows:

Observation sequence: {Cesarean section history, placenta previa, maternal age, multiparity history, placental position} recorded as ;

State sequence: {normal type, adhesion type, implantation type, penetration type} is recorded as ;

Initial state probability distribution: denoted as ,in , , , ;

State transition probability matrix: ,,take out , extract the feature data of each stage, for Medium state The state transition of the state transition is counted, and the remaining n-1 pieces of data are processed in the same way, and the obtained state transition is carried out for probability statistics, so as to obtain the state transition probability matrix;

Observe the state transition matrix: , ,take out , extract the characteristic data of each stage, divide them into four categories with the state sequence q as the division standard, and perform the same processing on the remaining n-1 pieces of maternal data, and make statistics on the data of the four categories to obtain Observation state transition matrix;

Model parameters: .

In this embodiment, the specific steps for the third step to predict the type of placenta accreta of the case are as follows: the observation sequence Perform iterative operations, and finally obtain the probability distribution of the initial state The probability of placenta accreta is used to assist in judging the type of placenta accreta. In order to realize the type prediction process, it is necessary to prepare the state transition probability matrix A and the value of the observed state transition matrix B to perform iterative calculations to obtain the final transition state probability. The initial value of the process is:

, ;

Then perform iterative operation to get , where N is the number of child elements of state q;

Finally, the probability of the sequence is calculated by the above formula, and the type of placenta accreta of the examiner is predicted by this.

In order to allow those skilled in the art to better understand the technical solutions of the present invention, the following describes the hidden Markov model self-learning process in detail.

O: The O set is a set of observation states. The settings in it only select a few groups of eigenvalues with large impact factors in the current scientific research, which have a large impact on the final probability judgment. Through continuous update The observable state of the O set makes predictions more accurate.

q: state set q is the placenta accreta classification proposed by Maldjian, which is a relatively recognized classification method at present.

: The initial value of the initial state probability distribution needs to be saved. The current data is partial and one-sided. When a new authoritative organization releases new statistics on placenta accreta cases, the new state probability distribution can be updated.

A: The probability distribution matrix of state transition also needs to rely on the case statistics of authoritative organizations to continuously update the probability matrix to achieve the desired effect.

B: According to the baum-welch algorithm, the observation sequence O and the initial model , to obtain a new parameter by re-formulating , , A new model can be obtained, namely , and then use the new model parameters to iteratively calculate the probability. At the same time, the model parameters are constantly changing, and the new model parameter values are updated regularly.

Proceed as follows:

Initialize: give , , assign the initial value to get ;

Calculation by EM algorithm: calculating expected value ;

Expected value: Right now ,

;

Re-estimate model parameters: re-estimate the resulting expected value , , , get new ;

               

               

.

The purpose of the hidden Markov model self-learning process is to provide a parameter re-estimation method that reduces data errors and uses algorithms to approach accurate initial values.

The above descriptions are only preferred embodiments of the present invention, and all equivalent changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope of the present invention.

Claims (4)

1. a method for predicting placenta accreta based on hidden Markov model, is characterized in that comprising the following steps: The first step: extract maternal data, obtain x pieces of maternal data from hospitals or health organizations, and the observation objects of a piece of maternal data include history of cesarean section, placenta previa, maternal age, history of multiple births and placental position, 200≤x≤400 ; The second step: establish the initial data of the hidden Markov model; Step 3: Predict the type of placenta accreta of the case on the basis of the initial data of the hidden Markov model. 2. the method for predicting placenta accreta based on hidden Markov model according to claim 1, is characterized in that: in the first step, a parturient data note comprises 15 stages, and adjacent two stages are separated by one week, and are recorded as , where 1≤n≤x. 3. the placenta accreta prediction method based on hidden Markov model according to claim 1, is characterized in that: in the described second step, the concrete steps of setting up the initial data of hidden Markov model are as follows: Observation sequence: {Cesarean section history, placenta previa, maternal age, multiparity history, placental position} recorded as ; State sequence: {normal type, adhesion type, implantation type, penetration type} is recorded as ; Initial state probability distribution: denoted as ,in , , , ; State transition probability matrix: , ,take out , extract the feature data of each stage, for Medium state Statistical state transition of the remaining n-1 pieces of maternal data is processed in the same way, and the obtained state transition is statistically calculated to obtain a state transition probability matrix; Observe the state matrix: , ,take out , extract the characteristic data of each stage, divide them into four categories with the state sequence q as the division standard, and perform the same processing on the remaining n-1 pieces of maternal data, and make statistics on the data of the four categories to obtain Observation state transition matrix; Model parameters: . 4. the placenta accreta prediction method based on hidden Markov model according to claim 1, is characterized in that: the concrete step that described 3rd step carries out the type prediction to the placenta accreta of case is as follows: to described observation sequence Perform iterative operations, and finally obtain the probability distribution of the initial state The probability of placenta accreta is used to assist in judging the type of placenta accreta. In order to realize the type prediction process, it is necessary to prepare the state transition probability matrix A and the value of the observed state transition matrix B to perform iterative calculations to obtain the final transition state probability. The initial value of the process is: , ; Then perform iterative operation to get , where N is the number of child elements of state q; Finally, the probability of the sequence is calculated by the above formula, and the type of placenta accreta of the examiner is predicted by this.