Children leukemia prognosis evaluation and early warning method

Abstract

The invention discloses a children leukemia prognosis evaluation and early warning method, which comprises the following steps: acquiring historical clinical detection data of a children leukemia patient after treatment, and obtaining prognosis life limits of the children leukemia patient represented by different historical clinical detection data; constructing a prognosis evaluation model, training the prognosis evaluation model by using historical clinical detection data, and outputting a convergent prognosis evaluation model; collecting new clinical detection data of the child leukemia patient after treatment; respectively inputting the biological marker detection data set, the gene expression data set and the treatment reaction data set into a convergent prognosis evaluation model, calculating the probability of poor prognosis, constructing a risk model of the probability of poor prognosis, calculating the risk coefficient of poor prognosis, and evaluating the prognosis effect of the children leukemia patient. According to updating of prognosis clinical detection data of children leukemia patients, evaluation can be carried out in real time, and timeliness and high efficiency of big data analysis are embodied.

Description

Technical Field

[0001] The present invention relates to the field of big data prediction of childhood leukemia prognosis, and in particular to a childhood leukemia prognosis assessment and early warning method. Background Art

[0002] Childhood leukemia, especially acute lymphoblastic leukemia (ALL), is one of the most common malignant tumors in childhood. Although current treatment methods have made significant progress, there are still large differences in patient prognosis, and the risk of prognosis recurrence also needs to be monitored regularly. With the explosive growth of medical data, how to effectively use this data to predict disease progression and patient prognosis has become a hot topic in medical research. For the need for long-term and frequent prognosis monitoring, the use of medical big data analysis technology to assist in the prediction of patient prognosis recurrence can effectively reduce the management cost of childhood leukemia patients and improve the timeliness of prognosis effect evaluation of childhood leukemia patients. Therefore, the development of a childhood leukemia prognosis evaluation and early warning method is of great significance to improving the quality of life of patients and the survival rate of patients. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for early warning of prognosis assessment of childhood leukemia, which solves the problem of prognosis effect and recurrence risk assessment of childhood leukemia patients based on a mathematical model.

[0004] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0005] A method for early warning of childhood leukemia prognosis assessment comprises the following steps:

[0006] S1: Collect historical clinical test data of childhood leukemia patients after treatment. The historical clinical test data include biological marker test data, gene expression data and treatment response data, and obtain the prognosis survival years of childhood leukemia patients represented by different historical clinical test data;

[0007] S2: Construct a prognostic evaluation model, and use the prognostic survival years to calculate the probability of poor prognosis, use historical clinical test data to train the prognostic evaluation model, and output a converged prognostic evaluation model;

[0008] S3: Collect new clinical test data of childhood leukemia patients after treatment, and split the new clinical test data into a biological marker test data set, a gene expression data set, and a treatment response data set for calculating the probability of poor prognosis;

[0009] S4: Input the biological marker detection data group, gene expression data group and treatment response data group into the convergent prognostic evaluation model respectively, calculate the probability of poor prognosis of childhood leukemia patients at this time, and construct a risk model of the probability of poor prognosis, calculate the risk coefficient of poor prognosis, and evaluate the prognostic effect of childhood leukemia patients at this time.

[0010] Further, step S2 includes:

[0011] S21: Construct a prognostic evaluation model;

[0012]

[0013] Among them, f(.) is the prognostic evaluation function, p is the probability of poor prognosis of children with leukemia, X i is the clinical test data, X i ′ is the standard value or standard range of clinical test data, i is the type of clinical test data, β i is the weight coefficient of clinical test data, β0 is a constant term, and n is the number of clinical test data;

[0014] S22: Calculate the probability of poor prognosis p based on the prognostic survival years of childhood leukemia patients e ;

[0015] S23: Calculate the probability of poor prognosis of childhood leukemia patients corresponding to all historical clinical test data, and obtain the probability data set (p1, p2, ..., p E ), E is the number of children with leukemia, p E is the probability of poor prognosis for the Eth childhood leukemia patient;

[0016] S24: Establish three different training data sets based on the biological marker detection data, gene expression data, and treatment response data, and respectively assign the three training data sets and the corresponding probability data sets (p1, p2, ..., p E ) is input into the prognostic evaluation model, the prognostic evaluation model is trained, and the constant term β0 and weight coefficient β obtained during the training process are calculated. i , and take the average value to obtain a convergent prognostic evaluation model, including the biological marker prognostic evaluation function f1(.), the gene expression prognostic evaluation function f2(.) and the treatment response prognostic evaluation function f3(.).

[0017] Furthermore, the probability of poor prognosis p e The calculation method is:

[0018]

[0019] Among them, N eN0 is the average number of years of survival of childhood leukemia patients after treatment.

[0020] Further, step S4 includes:

[0021] S41: inputting the biological marker detection data set, the gene expression data set and the treatment response data set into the biological marker prognosis evaluation function f1(.), the gene expression prognosis evaluation function f2(.) and the treatment response prognosis evaluation function f3(.), respectively, and calculating the corresponding probabilities of poor prognosis p1, p2, p3;

[0022] S42: construct a risk model based on different probabilities of poor prognosis and calculate the risk coefficient P of poor prognosis;

[0023]

[0024] Among them, h0(.) is the basic risk function, N′ is the year after treatment of childhood leukemia patients when new clinical test data is collected, and p u is one of the probabilities of poor prognosis p1, p2, and p3, u is the number of the probabilities of poor prognosis p1, p2, and p3, α u is the weight coefficient corresponding to the probability of poor prognosis of the uth case;

[0025] S43: Setting a threshold value P0 of the risk factor;

[0026] If P≤P0, the prognosis of childhood leukemia patients at this time is good, and there is no need to produce early warning signals;

[0027] If P>P0, the prognosis of childhood leukemia patients at this time is poor, producing a warning signal, indicating that childhood leukemia patients are at risk of recurrence and need further examination.

[0028] The beneficial effects of the present invention are as follows: this scheme performs big data analysis based on a large amount of clinical test data, which is used to evaluate the effect of prognosis of childhood leukemia, and uses the clinical test data of historical childhood leukemia patients to train the prognosis evaluation model, thereby obtaining a data model that can accurately calculate the probability of poor prognosis, and can calculate the probability of poor prognosis from a variety of clinical test data, and at the same time, combine multiple probabilities of poor prognosis to calculate the risk coefficient through a risk model, which can accurately evaluate the prognosis effect or risk of childhood leukemia patients; it provides strong support for the prognosis evaluation or clinical decision-making of childhood leukemia patients, and can perform real-time evaluation according to the update of the clinical test data of the prognosis of childhood leukemia patients, reflecting the timeliness and efficiency of big data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1Flowchart of the early warning method for prognostic assessment of childhood leukemia. DETAILED DESCRIPTION

[0030] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0031] like Figure 1 As shown, a method for early warning of childhood leukemia prognosis assessment comprises the following steps:

[0032] S1: Collect historical clinical test data of childhood leukemia patients after treatment. The historical clinical test data include biological marker test data, gene expression data and treatment response data, and obtain the prognosis survival years of childhood leukemia patients represented by different historical clinical test data;

[0033] The biomarker detection data in the prognostic clinical detection data of childhood leukemia patients include peripheral blood cell counts, biomarker levels (such as sFas, CD56), etc., gene expression data include chromosomal abnormalities, immunophenotyping, specific gene expression (such as OPAL1), etc., and treatment response data include body temperature data (fever, hypothermia, etc.), vomiting data, diarrhea, etc.

[0034] S2: construct a prognosis evaluation model, and use the prognosis survival years to calculate the probability of poor prognosis, use historical clinical test data to train the prognosis evaluation model, and output a converged prognosis evaluation model; step S2 specifically includes:

[0035] S21: Construct a prognostic evaluation model;

[0036]

[0037] Among them, f(.) is the prognostic evaluation function, p is the probability of poor prognosis of children with leukemia, X i is the clinical test data, X i ′ is the standard value or standard range of clinical test data, i is the type of clinical test data, β i is the weight coefficient of clinical test data, which represents the importance of different clinical test data to prognosis evaluation. Generally, β i =(0,1), β0 is a constant term, and n is the number of clinical test data;

[0038] S22: Calculate the probability of poor prognosis p based on the prognostic survival years of childhood leukemia patients e; Poor prognosis probability p e The calculation method is:

[0039]

[0040] Among them, N e N0 is the average number of years of survival of childhood leukemia patients after treatment.

[0041] S23: Calculate the probability of poor prognosis of childhood leukemia patients corresponding to all historical clinical test data, and obtain the probability data set (p1, p2, ..., p E ), E is the number of children with leukemia, p E is the probability of poor prognosis for the Eth childhood leukemia patient;

[0042] S24: Establish three different training data sets based on the biological marker detection data, gene expression data, and treatment response data, and respectively assign the three training data sets and the corresponding probability data sets (p1, p2, ..., p E ) is input into the prognostic evaluation model, the prognostic evaluation model is trained, and the constant term β0 and weight coefficient β obtained during the training process are calculated. i , and take the average value to obtain a convergent prognostic evaluation model, including the biological marker prognostic evaluation function f1(.), the gene expression prognostic evaluation function f2(.) and the treatment response prognostic evaluation function f3(.).

[0043] S3: Collect new clinical test data of childhood leukemia patients after treatment, and split the new clinical test data into a biological marker test data set, a gene expression data set, and a treatment response data set for calculating the probability of poor prognosis;

[0044] S4: Input the biological marker detection data set, the gene expression data set and the treatment response data set into the converged prognostic evaluation model, calculate the probability of poor prognosis of the child leukemia patient at this time, and construct a risk model of the probability of poor prognosis, calculate the risk coefficient of poor prognosis, and evaluate the prognostic effect of the child leukemia patient at this time. Step S4 specifically includes:

[0045] S41: inputting the biological marker detection data set, the gene expression data set and the treatment response data set into the biological marker prognosis evaluation function f1(.), the gene expression prognosis evaluation function f2(.) and the treatment response prognosis evaluation function f3(.), respectively, and calculating the corresponding probabilities of poor prognosis p1, p2, p3;

[0046] S42: construct a risk model based on different probabilities of poor prognosis and calculate the risk coefficient P of poor prognosis;

[0047]

[0048] Among them, h0(.) is the basic risk function, N′ is the year after treatment of childhood leukemia patients when new clinical test data is collected, and p u is one of the probabilities of poor prognosis p1, p2, and p3, u is the number of the probabilities of poor prognosis p1, p2, and p3, α u is the weight coefficient corresponding to the u-th probability of poor prognosis, representing the importance of three different probabilities of poor prognosis to risk assessment;

[0049] S43: Setting a threshold value P0 of the risk factor;

[0050] If P≤P0, the prognosis of childhood leukemia patients at this time is good, and there is no need to produce early warning signals;

[0051] If P>P0, the prognosis of childhood leukemia patients at this time is poor, producing a warning signal, indicating that childhood leukemia patients are at risk of recurrence and need further examination.

[0052] This solution is based on big data analysis of a large amount of clinical test data to evaluate the prognosis of childhood leukemia. The clinical test data of historical childhood leukemia patients are used to train the prognosis evaluation model to obtain a data model that can accurately calculate the probability of poor prognosis. The probability of poor prognosis can be calculated from a variety of clinical test data. At the same time, the risk coefficient is calculated by combining multiple probabilities of poor prognosis through a risk model, which can accurately evaluate the prognosis effect or risk of childhood leukemia patients. It provides strong support for the prognosis evaluation or clinical decision-making of childhood leukemia patients, and can be evaluated in real time according to the update of the clinical test data of the prognosis of childhood leukemia patients, reflecting the timeliness and efficiency of big data analysis.

Claims

1. A method for early warning of childhood leukemia prognosis assessment, characterized in that: The following steps are involved: S1: Collect historical clinical test data of childhood leukemia patients after treatment. The historical clinical test data include biological marker test data, gene expression data and treatment response data, and obtain the prognosis survival years of childhood leukemia patients represented by different historical clinical test data; S2: Construct a prognostic evaluation model, and use the prognostic survival years to calculate the probability of poor prognosis, use historical clinical test data to train the prognostic evaluation model, and output a converged prognostic evaluation model; S3: Collect new clinical test data of childhood leukemia patients after treatment, and split the new clinical test data into a biological marker test data set, a gene expression data set, and a treatment response data set for calculating the probability of poor prognosis; S4: Input the biological marker detection data group, gene expression data group and treatment response data group into the convergent prognostic evaluation model respectively, calculate the probability of poor prognosis of childhood leukemia patients at this time, and construct a risk model of the probability of poor prognosis, calculate the risk coefficient of poor prognosis, and evaluate the prognostic effect of childhood leukemia patients at this time.

2. The method for early warning of childhood leukemia prognosis assessment according to claim 1, characterized in that: The step S2 comprises: S21: Construct a prognostic evaluation model; Among them, f(.) is the prognostic evaluation function, p is the probability of poor prognosis of children with leukemia, X i is the clinical test data, X i ′ is the standard value or standard range of clinical test data, i is the type of clinical test data, β i is the weight coefficient of clinical test data, β0 is a constant term, and n is the number of clinical test data; S22: Calculate the probability of poor prognosis based on the prognostic survival years of childhood leukemia patients e ; S23: Calculate the probability of poor prognosis of childhood leukemia patients corresponding to all historical clinical test data, and obtain the probability data set (p1, p2, ..., p E ), E is the number of children with leukemia, p E is the probability of poor prognosis for the Eth childhood leukemia patient; S24: Establish three different training data sets based on the biological marker detection data, gene expression data, and treatment response data, and respectively assign the three training data sets and the corresponding probability data sets (p1, p2, ..., p E ) is input into the prognostic evaluation model, the prognostic evaluation model is trained, and the constant term β0 and weight coefficient β obtained during the training process are calculated. i , and take the average value to obtain a convergent prognostic evaluation model, including the biological marker prognostic evaluation function f1(.), the gene expression prognostic evaluation function f2(.) and the treatment response prognostic evaluation function f3(.).

3. The method for early warning of childhood leukemia prognosis assessment according to claim 2, characterized in that: The probability of poor prognosis p e The calculation method is: Among them, N e N0 is the number of years in which the e-th childhood leukemia patient survives after treatment, and N1 is the average number of years in which the childhood leukemia patient survives after treatment.

4. The method for early warning of childhood leukemia prognosis assessment according to claim 1, characterized in that: The step S4 comprises: S41: inputting the biological marker detection data set, the gene expression data set and the treatment response data set into the biological marker prognosis evaluation function f1(.), the gene expression prognosis evaluation function f2(.) and the treatment response prognosis evaluation function f3(.), respectively, and calculating the corresponding probabilities of poor prognosis p1, p2, p3; S42: construct a risk model based on different probabilities of poor prognosis and calculate the risk coefficient P of poor prognosis; Among them, h0(.) is the basic risk function, N′ is the year after treatment of childhood leukemia patients when new clinical test data is collected, and p u is one of the probabilities of poor prognosis p1, p2, and p3, u is the number of the probabilities of poor prognosis p1, p2, and p3, α u is the weight coefficient corresponding to the probability of poor prognosis of the uth case; S43: Setting a threshold value P0 of the risk factor; If P≤P0, the prognosis of childhood leukemia patients is good at this time, and there is no need to produce early warning signals; If P>P0, the prognosis of childhood leukemia patients at this time is poor, producing a warning signal, indicating that childhood leukemia patients are at risk of recurrence and need further examination.

Images