WO2018147653A1 - Method, device and computer program for generating survival rate prediction model - Google Patents

Abstract

An embodiment of the present invention provides a method for generating a survival rate prediction model comprising: a first step of generating a Nth interval survival rate prediction model using clinical data and Nth interval survival rate data of a subject providing the clinical data; and a second step of generating an (N+1)th interval survival rate prediction model using the Nth interval survival rate data, the Nth interval survival rate prediction model, and (N+1)th interval survival rate data.

Description

Methods, devices, and computer programs for generating survival prediction models

Embodiments of the present invention relate to a method, apparatus and computer program for generating a survival prediction model.

Five-year cancer survival rate is a representative indicator of the effectiveness of medical technology and the performance of the medical system. Cancer 5-year survival rate refers to the probability that a cancer patient diagnosed with cancer and started treatment will not die from cancer that has occurred for at least 5 years. The high five-year survival rate indicates that cancer treatment is progressing relatively smoothly, and that the medical system is suitable for managing severe diseases such as cancer.

The OECD compares international 5-year survival rates for colorectal cancer, uterine cancer and breast cancer. OECD cancer survival statistics are inconsistent with domestic cancer survival statistics due to differences in calculations. On a OECD basis, Korea's five-year survival rate for colorectal cancer was 70.9% in 2008-2013, the highest among the major OECD countries, followed by Sweden (65.4%). The 5-year survival rate for breast cancer is 85.9% in Korea, slightly higher than the OECD average (84.9%) and about 5 percentage points lower than Sweden (89.4%), which has the highest survival rate.

Embodiments of the present invention provide a method, apparatus, and computer program for generating a survival prediction model capable of predicting annual survival rates from medical data of a patient.

According to an embodiment of the present invention, a first step of generating an N-th section survival prediction model using clinical data and N-th section survival rate data of a subject who provided the clinical data; And a second step of generating an N + 1th interval survival prediction model using the Nth interval survival rate data, the Nth interval survival prediction model, and the N + 1st interval survival rate data. do.

And assigning a score according to the survival period to each of the N-th section survival rate data, wherein the second step comprises: N + 1 year survival rate by using a score assigned to each of the N-th section survival rate data A prognostic prediction model generation method for generating a prediction model is disclosed.

The score discloses a method for generating a prognostic prediction model that is proportional to the survival period.

The survival period discloses a method for generating a prognostic prediction model divided at least monthly.

The N + 1 th section survival prediction model discloses a method for generating a prognostic prediction model, which is an input / output function for inputting the clinical data and the N th section survival rate data and outputting the N + 1 th section survival rate data.

The generating step discloses a prognostic prediction model generation method using a Recurrent Neural Network (RNN) algorithm.

Another embodiment of the present invention comprises a clinical data acquisition unit for obtaining clinical data; A survival rate data acquisition unit for obtaining survival rate data; And generating the N-th section survival rate prediction model using the clinical data and the N-th section survival rate data of the subject who provided the clinical data, and the N-th section survival rate data, the N-th section survival rate prediction model, and the N + 1st And a prediction model generator configured to generate an N + 1th interval survival prediction model using interval survival data.

Survival data processing unit for giving a score according to the survival period for each of the N-th section survival rate data; further comprising, wherein the prediction model generator, N + using the scores assigned to each of the N-th section survival rate data Disclosed is a prognostic prediction model generating device for generating a 1 year survival prediction model.

The score discloses an apparatus for generating a prognostic prediction model that is proportional to the survival period.

The survival period discloses a prognostic prediction model generating device divided at least monthly.

The N + 1th section survival rate prediction model discloses a prognostic prediction model generating device which is an input / output function for inputting the clinical data and the Nth section survival rate data and outputting the N + 1th section survival rate data.

The generating step discloses a prognostic prediction model generation device using a Recurrent Neural Network (RNN) algorithm.

Another embodiment of the present invention discloses a computer program stored in a medium for executing the above-described prognostic prediction model generation method using a computer.

Other aspects, features, and advantages other than those described above will become apparent from the following drawings, claims, and detailed description of the invention.

Survival prediction model generation method, apparatus and computer program according to embodiments of the present invention can predict the annual survival rate from the medical data of the patient.

The method, apparatus and computer program for generating a survival prediction model according to embodiments of the present invention generate a prediction model with high accuracy by generating a survival prediction model this year using survival data of a previous year's patient.

Survival prediction model generation method, apparatus and computer program according to embodiments of the present invention can obtain a large number of significant data by assigning a ranked score in using the survival data of the previous year patient, and thus accuracy Produces a high prediction model.

1 schematically illustrates a configuration of an apparatus for generating a predictive model according to an embodiment of the present invention.

2 is a flowchart illustrating a method of generating a prediction model according to an embodiment of the present invention.

3 is a flowchart illustrating a method of generating a prediction model according to another embodiment of the present invention.

4 is a view for explaining a prediction model according to an embodiment of the present invention.

5 is another example of a diagram for describing a prediction model according to an embodiment of the present invention.

6 is a graph illustrating a prediction model according to an embodiment of the present invention.

As the invention allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description. Effects and features of the present invention, and methods of achieving them will be apparent with reference to the embodiments described below in detail together with the drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various forms.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings, and the same or corresponding components will be denoted by the same reference numerals, and redundant description thereof will be omitted. .

In the following examples, the singular forms "a", "an" and "the" include plural forms unless the context clearly indicates otherwise. In the following examples, the terms including or having have meant that there is a feature or component described in the specification and does not preclude the possibility of adding one or more other features or components.

Hereinafter, with reference to the preferred embodiment of the present invention shown in the accompanying drawings will be described in detail the present invention.

1 schematically illustrates a configuration of an apparatus for generating a predictive model according to an embodiment of the present invention.

The predictive model generating apparatus 10 shown in FIG. 1 shows only the components related to the present embodiment in order to prevent the features of the present embodiment from being blurred. Accordingly, it will be understood by those skilled in the art that other general purpose components may be further included in addition to the components shown in FIG. 1.

The prediction model generating apparatus 10 according to the present embodiment may correspond to at least one processor or may include at least one processor. Accordingly, the predictive model generating device 10 may be driven in a form included in another hardware device such as a microprocessor or a general purpose computer system.

The invention can be represented by functional block configurations and various processing steps. Such functional blocks may be implemented in various numbers of hardware or / and software configurations that perform particular functions. For example, the present invention is an integrated circuit configuration such as memory, processing, logic, look-up table, etc., capable of executing various functions by the control of one or more microprocessors or other control devices. You can employ them. Similar to the components in the present invention may be implemented in software programming or software elements, the present invention includes various algorithms implemented in data structures, processes, routines or other combinations of programming constructs, including C, C ++ It may be implemented in a programming or scripting language such as Java, an assembler, or the like. The functional aspects may be implemented with an algorithm running on one or more processors. In addition, the present invention may employ the prior art for electronic environment setting, signal processing, and / or data processing. Terms such as "mechanism", "element", "means", "configuration" may be used widely, and the components of the present invention are not limited to mechanical and physical configurations. The term may include the meaning of a series of routines of software in conjunction with a processor or the like.

Referring to FIG. 1, the prediction model generating device 10 includes a clinical data acquisition unit 11, a survival rate data acquisition unit 12, a survival rate data processing unit 14, and a prediction model generation unit 13.

According to an embodiment, the clinical data acquisition unit 11 acquires medical data of a patient, for example, clinical data. Clinical data may be obtained from a medical image of a patient or may be obtained from a patient's specimen test result, but is not limited thereto.

The survival rate data acquisition unit 12 according to an embodiment acquires survival rate data of a subject (patient) who provided clinical data. Survival data is data showing survival or not. Survival data may include only survival, but may further include information about survival.

The prediction model generator 13 generates a survival prediction model that can predict the survival rate for each section of the patient from the clinical data of the patient. The interval may be a year. For example, the prediction model generator 13 may generate a survival prediction model that predicts annual survival rate of gastric cancer patients from clinical data of gastric cancer patients. In detail, the prediction model generator 13 may generate a survival prediction model that may predict annual survival rates of gastric cancer patients from 1 year to 5 years from the first year clinical data of gastric cancer patients.

According to an embodiment, the prediction model generator 13 may generate a survival prediction model based on clinical data and actual survival data of patients using a machine learning technique. The prediction model generator 13 may use a deep learning algorithm among machine learning techniques, and among them, may use a recurrent neural network (RNN) algorithm. The RNN algorithm refers to a neural network in which a connection between units constituting an artificial neural network constitutes a directed cycle.

The prediction model generator 13 generates the N-th section survival prediction model using the clinical data and the N-th section survival rate data of the patient who provided the clinical data. The prediction model generator 13 generates the N + 1st section survival prediction model using the clinical data and the N + 1st section survival rate data of the patient who provided the clinical data. The prediction model generator 13 may be based on the Nth interval survival prediction model using the RNN algorithm to generate the N + 1th interval survival prediction model.

The prediction model generator 13 according to an embodiment further uses the N-th section survival rate data to generate the N + 1-th section survival rate prediction model. That is, the N + 1th section survival rate prediction model may be an input / output function that inputs clinical data and Nth section survival rate data and outputs N + 1th section survival rate. As such, when the N-th section survival rate data is further used as an input value of the N + 1 th section survival prediction model, an accurate prediction model may be generated.

In relation to generating the N-th section survival prediction model using the N-th section survival rate data, the prediction model generator 13 according to an embodiment generates the aforementioned N + 1-th section survival prediction model when N is 2 or more. The method applies equally. For example, when N is 2 or more, the prediction model generator 13 may be based on the N-1th interval survival prediction model using the RNN algorithm to generate the Nth interval survival prediction model, and N-1 Second interval survival data may be further used.

On the other hand, when N is 1, the prediction model generator 13 does not use the N-1th section survival rate prediction model, and uses the clinical data and the Nth section survival rate data of the patient who provided the clinical data. A survival prediction model may be generated, and a predetermined initial survival rate P_0 may be further used. Details related to this will be described later with reference to FIG. 5.

The prediction model generator 13 may use the processed data in using the N-th section survival rate data.

The survival rate data processor 14 according to an embodiment processes the survival rate data. For example, each survival data is scored over survival. The prediction model generator 13 generates a N + 1 year survival prediction model using scores assigned to each of the N-th section survival data. The survival rate data processing unit 14 may assign a score to the survival rate data in proportion to the survival period. Survival periods may be divided at least monthly. According to this, the survival rate of the deceased patient is not counted as 0, but the ranked score is given as much as the survival period, thereby increasing the number of significant data used to generate the model, and consequently contributing to the generation of a highly accurate model. .

The prediction model generator 13 generates a survival rate prediction model for each section, for example, annually. For example, in the case of a model for predicting the survival rate of a cancer patient, a 1 year survival prediction model, a 2 year survival prediction model, a 3 year survival prediction model, a 4 year survival prediction model and a 5 year survival prediction model may be generated. Each annual model predicts the patient's corresponding annual survival rate from the patient's clinical data.

Each annual model may include a multi-layered matrix that connects i nodes corresponding to clinical data of the patient to two survival nodes. Each annual model generated by the predictive model generator 13 according to an embodiment may connect two nodes corresponding to i nodes corresponding to clinical data of a patient and two nodes corresponding to a previous year survival rate to two survival rate nodes. It may include a matrix of multiple layers, that is, a matrix of multiple layers connecting i + 2 input nodes to two output nodes.

The input node contains nodes corresponding to previous year's survival rate data. The previous year's survival rate data nodes may be two, each of which may be a survival node and a death node. When the two nodes corresponding to the survival rate data are [survival node and death node], the survival rate data may be [0, 1] when the patient dies or [1, 0] when the patient survives.

However, according to one embodiment of the present invention, a treatment method capable of giving a score by ranking it is proposed without treating the survival rate data when the patient dies with [0, 1]. Survival data of patients who died in this example may be [p, 1-p], where p is assigned a non-zero score value. According to one example, the score may be given in proportion to the survival of the deceased patient. In this case, the survival period may be divided into at least monthly units. For example, for patients who survived for 3 months at N years, the N year survival score may be 3/12, and the survival data may be [3/12, 1-3 / 12] = [0.25, 0.75]. Table 1 below is an example of N-year survival rate scores by N-year survival period.

Survival	Score
One	1/12 = 0.08
2	2/12 = 0.17
3	3/12 = 0.25
4	4/12 = 0.33
5	5/12 = 0.42
6	6/12 = 0.5
7	7/12 = 0.58
8	8/12 = 0.67
9	9/12 = 0.75
10	10/12 = 0.83
11	11/12 = 0.92
12	12/12 = 1

Table 1 described an example in which survival periods are divided into monthly units, but the present invention is not limited thereto, and the survival periods may be divided into various units such as semi-annual, quarterly, monthly, and day depending on the design.

2 is a flowchart illustrating a method of generating a prediction model according to an embodiment of the present invention.

Referring to FIG. 2, in step 21, the prediction model generator 13 of FIG. 1 generates the N-th section survival prediction model using the clinical data of the patient and the N-th section survival rate data.

In operation 22, the prediction model generator 13 of FIG. 1 generates the N + 1th section survival prediction model using the Nth section survival rate data, the Nth section survival rate prediction model, and the N + 1st section survival rate data.

3 is a flowchart illustrating a method of generating a prediction model according to another embodiment of the present invention.

Referring to FIG. 3, in step 31, the prediction model generator 13 of FIG. 1 generates the N-th section survival prediction model using the clinical data of the patient and the N-th section survival rate data.

In step 32, the survival rate data processing unit 14 of FIG. 1 assigns a score according to the survival period to the Nth section survival rate data.

In operation 33, the prediction model generator 13 of FIG. 1 generates the N + 1th section survival prediction model using the score of the Nth section survival rate data, the Nth section survival prediction model, and the N + 1st section survival rate data. .

On the other hand, the flow chart shown in Figures 2 and 3 is composed of steps that are processed in time series in the predictive model generating device 10 shown in FIG. Therefore, even if omitted below, it can be seen that the above description of the components shown in FIG. 1 also applies to the flowcharts shown in FIGS. 2 and 3.

Meanwhile, the method of generating a predictive model according to an embodiment of the present invention shown in FIGS. 2 and 3 may be written as a program that can be executed in a computer, and the general purpose of operating the program using a computer-readable recording medium. It can be implemented in a digital computer. The computer-readable recording medium may include a storage medium such as a magnetic storage medium (eg, a ROM, a floppy disk, a hard disk, etc.) and an optical reading medium (eg, a CD-ROM, a DVD, etc.).

4 is a view for explaining a prediction model according to an embodiment of the present invention.

Referring to FIG. 4, the N-1 year survival rate prediction model PM_N-1 and the N year survival rate prediction model PM_N are illustrated. Referring to FIG. 4, the N-1 year survival rate prediction model PM_N-1 predicts the N-1 year survival rate P_N-1 from clinical data (X). For example, the N-1 year survival rate prediction model PM_N-1 is an input / output function capable of outputting the N-1 year survival rate P_N-1 when clinical data X is input. The N-1 year survival rate prediction model PM_N-1 may be generated by machine learning techniques based on clinical data (X) and N-1 year survival rate (P_N-1) data for a plurality of patients. Meanwhile, the N-year survival rate prediction model PM_N may be generated based on the N-year survival rate prediction model PM_N-1.

Referring to FIG. 4, the N year survival prediction model PM_N is generated based on the N-1 year prediction model PM_N-1 to generate a survival rate P_N from the clinical data X. Referring to FIG. According to an embodiment of the present invention, in generating the N-year survival rate prediction model PM_N, not only the N-1 year survival rate prediction model PM_N-1, but also the N-1 year survival rate P_N-1 is used. . On the other hand, the N-1 year survival rate (P_N-1) is the actual data for each patient case, it is used to generate the N-1 year survival rate prediction model (PM_N-1). Survival prediction models can be generated from year 1 to year 5.

In FIG. 4, the clinical data X input to each model may be first year clinical data, which are all initial values. According to this, an annual survival rate prediction model may be generated using the annual clinical data and the annual survival rate (P) for a plurality of patients, and inputting the first year clinical data of any patient into the generated model, The annual survival rate of the patient can be predicted.

5 is another example of a diagram for describing a prediction model according to an embodiment of the present invention.

Referring to FIG. 5, the first year survival rate prediction model PM_1 and the subsequent N year year survival rate prediction model PM_N are illustrated. Referring to FIG. 4, the first-year survival rate prediction model PM_1, which is an input / output function capable of outputting the first-year survival rate P_1 when the clinical data X and the initial survival rate P_0 is inputted, is machine-learned. Is generated. Next, when the clinical data (X) and the first year survival rate (P_1) are input, a second year survival rate prediction model (PM_2), which is an input / output function capable of outputting the second year survival rate (P_2), is generated by the machine learning technique. At this time, the first year survival rate prediction model PM_1 is used to generate the second year survival rate prediction model PM_2. As this process is repeated (N = N + 1), when the clinical data (X) and the N-1 year survival rate (P_N-1) are input, N, the input / output function that can output the N year survival rate (P_N) The annual survival rate prediction model PM_N is generated by the machine learning technique. The N-1 year survival rate prediction model PM_N-1 is used to generate the N year survival rate prediction model PM_N. For example, when the N-year survival rate prediction model (PM_N) is generated by the machine learning method based on the input and output of the N-year survival rate prediction model (PM_N), the N-1 year survival rate prediction model (PM_N−) is used as an initial model. 1) can be used. The first year survival rate prediction model PM_1 is used for the second year survival rate prediction model PM_2, and the second year survival rate prediction model PM_2 is the basis for generation of the third year survival rate prediction model PM_3.

The survival rate initial value P_0 may be set to [0, 1]. The first year survival rate P_1 may include values of two nodes, such as [1-p, p].

6 is a graph illustrating a prediction model according to an embodiment of the present invention.

The vertical axis of the heatmaps shown in FIG. 6 is the serial number of the subject, and the vertical axis corresponds to the node. The value corresponding to each node is represented by the intensity of the color. Referring to FIG. 6, graphs 611, 612, and 613 show a 1-year survival rate prediction model. Graph 611 shows clinical data for a plurality of subjects at concentrations corresponding to each of 30 nodes, with initial values of survival expressed as 2 nodes. Graph 613 shows the first year survival rate for a plurality of subjects as two nodes. That is, the 1-year survival prediction model converges a total of 32 node values (shown in graph 611) into a total of 2 node values (shown in graph 613). However, since the number of specific nodes is only an example, the present invention is not limited thereto.

Similarly, graphs 621, 622, and 623 show a second year survival rate prediction model. The graph 621 shows clinical data for a plurality of subjects at concentrations corresponding to each of 30 nodes, and 1 year survival rate as 2 nodes. Graph 623 shows the second year survival rate for two subjects as two nodes. In other words, the second year survival prediction model converges a total of 32 node values (shown in graph 621) into a total of two node values (shown in graph 613). The second year survival prediction model is generated by the first year survival prediction model, clinical data, the first year survival rate, and the second year survival rate.

Similarly, graphs 651, 652, and 653 show a five-year survival rate prediction model. Graph 651 shows clinical data for a plurality of subjects at concentrations corresponding to each of 30 nodes, with 4 year survival rates as 2 nodes. Graph 653 shows five-year survival rates for two subjects as two nodes. In other words, the 5 year survival prediction model converges a total of 32 node values (shown in graph 651) into a total of 2 node values (shown in graph 653).

So far I looked at the center of the preferred embodiment for the present invention. Although the present invention has been described with reference to the embodiments illustrated in the drawings, this is merely exemplary, and a person skilled in the art to which the present invention pertains may modify the present invention without departing from the essential characteristics of the present invention. It may be implemented in the form, it will be understood that other equivalent embodiments are possible. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

Embodiments of the present invention can be applied to a method, apparatus and computer program for generating a survival prediction model capable of predicting annual survival rate from medical data of a patient.

Although the above has been described with reference to exemplary embodiments of the present invention, those skilled in the art may vary the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. It will be understood that modifications and changes can be made.

Claims (13)

1. A first step of generating an N-th section survival prediction model using clinical data and N-th section survival rate data of the subject who provided the clinical data;

   A second step of generating an N + 1th interval survival prediction model using the Nth interval survival rate data, the Nth interval survival prediction model, and the N + 1th interval survival rate data;

   How to generate a survival prediction model.
2. According to claim 1,

   For each of the N-th section survival rate data, assigning a score according to the survival period;

   In the second step, the N + 1 year survival rate prediction model is generated using the scores assigned to each of the Nth interval survival rate data.

   How to create a prognostic prediction model.
3. The method of claim 2,

   The score is proportional to the survival

   How to create a prognostic prediction model.
4. The method of claim 2,

   The survival period is divided at least monthly

   How to create a prognostic prediction model.
5. According to claim 1,

   The N + 1th section survival rate prediction model is an input / output function that inputs the clinical data and the Nth section survival rate data and outputs the N + 1th section survival rate data.

   How to create a prognostic prediction model.
6. According to claim 1,

   The generating step using a Recurrent Neural Network (RNN) algorithm

   How to create a prognostic prediction model.
7. Clinical data acquisition unit for acquiring clinical data;

   A survival rate data acquisition unit for obtaining survival rate data; And

   The Nth section survival rate prediction model is generated using the clinical data and the Nth section survival rate data of the subject who provided the clinical data, and the Nth section survival rate data, the Nth section survival rate prediction model, and the N + 1st section A prediction model generator configured to generate an N + 1th interval survival prediction model using survival data;

   Survival prediction model generation device.
8. The method of claim 7, wherein

   And a survival rate data processing unit for giving a score according to the survival period for each of the N-th section survival rate data.

   The prediction model generator generates an N + 1 year survival prediction model using scores assigned to each of the Nth interval survival data.

   Prognostic prediction model generator.
9. The method of claim 8,

   The score is proportional to the survival

   Prognostic prediction model generator.
10. The method of claim 8,

    The survival period is divided at least monthly

    Prognostic prediction model generator.
11. The method of claim 7, wherein

    The N + 1th section survival rate prediction model is an input / output function that inputs the clinical data and the Nth section survival rate data and outputs the N + 1th section survival rate data.

    Prognostic prediction model generator.
12. The method of claim 7, wherein

    The generating step using a Recurrent Neural Network (RNN) algorithm

    Prognostic prediction model generator.
13. A computer program stored in a medium for carrying out the method of any one of claims 1 to 6 using a computer.

Images