Artificial Intelligence Systems

Abstract

To disclose an artificial intelligence system and prediction method of the same.SOLUTION: An artificial intelligence system includes: a data module that receives personal health record (PHR) data on a patient, and performs preprocessing of the PHR data to generate final data; a model module receives the final data and an instruction, and derives a pre-prediction of the patient via at least one artificial intelligence model selected according to the instruction; and a Reliable and Interpretable AI System (RIAS) that operates separately and independently the data module and the model module, receives the pre-prediction and the final data, and converts the pre-prediction into a final prediction.SELECTED DRAWING: Figure 6

Description

[Technical Field]

[0001] The present invention relates to an artificial intelligence system and a prediction method therefor. Specifically, the present invention relates to an artificial intelligence system and a prediction method therefor that can use various data and artificial intelligence models. [Background technology]

[0002] PHR (Personal Health Record) data is a record of an individual's health and can contain a wide range of information. Using such PHR data to derive medical decisions, such as patient mortality rates, is now expanding from the domain of doctors to the realm of artificial intelligence. Medical diagnostic AI typically develops a single AI model for a specific type of data and draws a conclusion. In other words, an independent pipeline is built according to the data, and a new pipeline is built when the AI ​​model is updated.

[0003] This method has the inefficiency of discarding overlapping parts of each pipeline without utilizing them, and having to start from scratch with new pipelines. Therefore, there has been continuous research to eliminate this inefficiency. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Korean Patent Publication No. 10-2023-0140774 Summary of the Invention [Problem to be solved by the invention]

[0005] An object of the present invention is to provide an artificial intelligence system that does not need to be modified even when data and artificial intelligence models change.

[0006] Another object of the present invention is to provide a prediction method for an artificial intelligence system that does not need to be modified even when data and artificial intelligence models change.

[0007] Another object of the present invention is to provide a computer program recorded on a recording medium for executing a prediction method for an artificial intelligence system that does not need to be modified even when data and artificial intelligence models change.

[0008] The objects of the present invention are not limited to those mentioned above, and other objects and advantages of the present invention not mentioned above will be understood from the following description and will become more clearly understood by the examples of the present invention. Furthermore, it will be easily understood that the objects and advantages of the present invention can be achieved by the means and combinations thereof as claimed. [Means for solving the problem]

[0009] To solve the above problem, an artificial intelligence system according to some embodiments of the present invention includes a data module that receives a patient's PHR (Personal Health Record) data and performs pre-processing of the PHR data to generate final data, a model module that receives the final data and instructions and derives a pre-prediction of the patient through at least one artificial intelligence model selected according to the instructions, and a Reliable and Interpretable AI System (RIAS) that is separate from the data module and the model module and operates independently, receives the pre-prediction and the final data, and converts the pre-prediction into a final prediction.

[0010] The patient may also include a patient with myocardial infarction.

[0011] The final prediction may also include the patient's mortality rate after a specified period of time.

[0012] The final prediction may also include a weighting of factors that positively contribute to the mortality rate and a weighting of factors that negatively contribute to the mortality rate.

[0013] The PHR data can also include mortality rate after acute myocardial infarction in a patient or reverse remodeling data after acute myocardial infarction in a patient.

[0014] Additionally, the PHR data may further include data that alters or manipulates the mortality or remodeling data.

[0015] The data module may also include a preprocessing module that receives instructions including a preset data frame and preprocesses the PHR data according to the data frame to generate preprocessed data, and an imputation module that processes missing values ​​in the preprocessed data to generate final data.

[0016] The imputation module can also receive the instructions and process the missing values ​​according to the data frame.

[0017] Additionally, the imputation module can handle the missing values ​​via at least one of mean imputation, median imputation, mode imputation, regression imputation, K-nearest neighbor imputation, and multiple imputation.

[0018] The instructions may also include information about a predetermined artificial intelligence model.

[0019] In addition, the at least one artificial intelligence model can be classified into one type if they have a common function, and the model module can objectify the common function according to the type.

[0020] Also, the type may include at least one of GBDT (Gradient Boosted Decision Trees), pytorch tabula, and Saint.

[0021] The RIAS may also include at least one of a training module that trains the artificial intelligence model and tunes and evaluates hyperparameters of the artificial intelligence model; a calibration module that corrects the preliminary prediction to derive a final prediction; a local explanation module that performs a local explanation method for the contribution of each feature to the final data; and a counterfactual explanation module that performs a counterfactual explanation method on the final data.

[0022] In addition, the pre-prediction is a predicted value between 0 and 1, and the calibration module can adjust the pre-prediction to mimic the likelihood of an actual event occurring to derive a final prediction.

[0023] To solve the above problem, an artificial intelligence system according to some embodiments of the present invention includes a memory and a processor that performs calculations in conjunction with the memory, wherein the processor preprocesses a patient's PHR data to generate final data, inputs the final data into a selected artificial intelligence model to derive a pre-prediction, and corrects the pre-prediction to derive a final prediction.

[0024] In order to solve the other problems described above, a prediction method of an artificial intelligence system according to some embodiments of the present invention includes the artificial intelligence system preprocessing the patient's PHR data to generate final data, inputting the final data into a selected artificial intelligence model to derive a pre-prediction, and correcting the pre-prediction to derive a final prediction.

[0025] In addition, generating the final data may include converting the PHR data into a predetermined data frame to generate pre-processed data, and processing missing values ​​in the pre-processed data to generate the final data.

[0026] The method may further include the artificial intelligence system receiving instructions, and the instructions may include the above-mentioned preset data frame.

[0027] Also, deriving the pre-prediction may include selecting an artificial intelligence model according to the instructions, and deriving the pre-prediction via the selected artificial intelligence model. [Effects of the Invention]

[0028] The artificial intelligence system and prediction method of the present invention do not require modification of the pipeline even if the input data changes.

[0029] Furthermore, even if the artificial intelligence module changes, there is no need to modify the pipeline, making it possible to implement an efficient pipeline system.

[0030] Furthermore, since the degree of contribution of factors that contribute to the prediction can be expressed, the prediction can be easily understood and explained to the patient.

[0031] In addition to the above, specific effects of the present invention will be described together with the following explanation of specific matters for carrying out the invention. [Brief explanation of the drawings]

[0032] [Figure 1] FIG. 1 is a block diagram illustrating an artificial intelligence system according to some embodiments of the present invention. [Figure 2] FIG. 2 is a block diagram for explaining the PHR data of FIG. [Figure 3]FIG. 3 is a block diagram for explaining the instructions of FIG. [Figure 4] FIG. 4 is an exemplary diagram showing the predictions when the beta-blocker of FIG. 1 is not administered. [Figure 5] FIG. 5 is an exemplary diagram showing the prediction when the beta blocker of FIG. 1 is administered. [Figure 6] FIG. 6 is a block diagram for explaining the configuration of the artificial intelligence system of FIG. [Figure 7] FIG. 7 is a block diagram for explaining the structure of the data module of FIG. [Figure 8] FIG. 8 is a block diagram for explaining the structure of the RIAS of FIG. [Figure 9] FIG. 9 is a block diagram for explaining the model module of FIG. [Figure 10] FIG. 10 is a flowchart illustrating a prediction method of an artificial intelligence system according to some embodiments of the present invention. [Figure 11] FIG. 11 is a flowchart for explaining in detail the final data generation stage of FIG. [Figure 12] FIG. 12 is a flowchart illustrating in detail the pre-prediction derivation step of FIG. [Figure 13] FIG. 13 is a diagram illustrating the hardware configuration of an artificial intelligence system according to some embodiments of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0033] The terms and words used in this specification and claims should not be interpreted as being limited to their general or dictionary meanings. They should be interpreted as meanings and concepts consistent with the technical idea of ​​the present invention, based on the principle that an inventor can define the concept of a term or word to best describe his or her invention. In addition, the embodiments described in this specification and the configurations shown in the drawings are merely examples in which the present invention is realized, and do not represent the entire technical idea of ​​the present invention. It should be understood that there may be various equivalents, modifications, and applicable examples that can replace them at the time of filing this application.

[0034] Terms such as "first," "second," "A," and "B" used in this specification and claims may be used to describe various elements, but the elements should not be limited by these terms. These terms are used only to distinguish one element from another. For example, a first element can be designated as a second element, and similarly, the second element can be designated as a first element, without departing from the scope of the present invention. The term "and / or" includes a combination of multiple related listed items or any of multiple related listed items.

[0035] The terms used in this specification and claims are merely used to describe specific embodiments and are not intended to limit the present invention. The singular expressions include the plural expressions unless the context clearly indicates otherwise. It should be understood that the terms "comprise" or "have" in this application do not preclude the presence or additional possibility of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification.

[0036] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0037] Terms such as those defined in commonly used dictionaries should be construed to have a meaning consistent with the contextual meaning of the relevant art, and should not be construed as having an idealized or overly formal meaning unless expressly defined in this application.

[0038] Furthermore, the configurations, processes, steps, or methods included in the embodiments of the present invention may be shared to the extent that they are not technically inconsistent with each other.

[0039] Hereinafter, artificial intelligence systems according to some embodiments of the present invention will be described with reference to FIGS.

[0040] FIG. 1 is a block diagram illustrating an artificial intelligence system according to some embodiments of the present invention.

[0041] Referring to Figure 1, an artificial intelligence system (10) according to some embodiments of the present invention can receive PHR data (DATA_PHR) and instructions (Inst). The artificial intelligence system (10) can internally generate a prediction (Prd) via the PHR data (DATA_PHR) and instructions (Inst). The artificial intelligence system (10) can transmit the generated prediction (Prd) to an external device.

[0042] FIG. 2 is a block diagram for explaining the PHR data of FIG.

[0043] Referring to FIG. 2 , the PHR data (DATA_PHR) may be data for various personal health records of a patient. The PHR data (DATA_PHR) may be, for example, personal health record data of a myocardial infarction patient. However, this embodiment is not limited thereto. The PHR data (DATA_PHR) may include, for example, at least one of first data (d1) on mortality within six months after the onset of acute myocardial infarction (AMI), second data (d2) on reverse remodeling within twelve months after the onset of myocardial infarction, third data (d3) on mortality within five years after the onset of myocardial infarction, and fourth data (d4) on modified data of the above data.

[0044] In this regard, reverse remodeling refers to the process by which the heart recovers from damage caused by myocardial infarction, whereby the size, shape, function, etc. of the heart return to normal. That is, the reverse remodeling process can include medical strategies to minimize damage to cardiac tissue, optimize cardiac function, and improve the patient's quality of life.

[0045] The periods mentioned in the first data (d1), second data (d2), and third data (d3) are all exemplary periods, and any number of different periods may be adopted depending on the need and purpose.

[0046] FIG. 3 is a block diagram for explaining the instructions of FIG.

[0047] Referring to FIGS. 1 to 3, the instruction (Inst) can include a plurality of pieces of information necessary for the artificial intelligence system (10) according to this embodiment to operate.

[0048] The instruction (Inst) may include, for example, at least one of first information (I1) for a data frame, second information (I2) for an artificial intelligence model, and third information (I3) for evaluation parameters for evaluation of the artificial intelligence model.

[0049] In this case, the first information (I1) may be information on how to preprocess the data frame of the PHR data (DATA_PHR). That is, since the PHR data (DATA_PHR) can exist in various formats, if it is not converted into one data frame, it may be necessary to newly configure a new pipeline system, which may result in inefficiency. The artificial intelligence system (10) according to this embodiment converts various PHR data (DATA_PHR) into a pre-defined data frame included in the instruction (Inst) and transmits it to the artificial intelligence model, thereby maximizing efficiency without the need to newly configure a pipeline system.

[0050] The second information (I2) may be information specifying an artificial intelligence model. That is, the second information (I2) may include information regarding which artificial intelligence model to use to derive the prediction (Prd). In existing prediction systems, in order to newly introduce the latest artificial intelligence model, a new pipeline system must be newly constructed, which may result in inefficiency. The artificial intelligence system (10) according to this embodiment can improve efficiency by selecting the artificial intelligence model as a preset model included in the instruction (Inst) and using functions overlapping with other artificial intelligence models as they are.

[0051] The third information (I3) may be necessary when presetting parameters necessary for evaluating an AI model. In this case, the third information (I3) may be pre-set so that it can be applied to any AI model selected from various AI models.

[0052] FIG. 4 is an exemplary diagram showing the prediction when a beta-blocker of FIG. 1 is not administered, and FIG. 5 is an exemplary diagram showing the prediction when a beta-blocker of FIG. 1 is administered.

[0053] 1, 4, and 5, prediction (Prd) can represent a probabilistic value, such as the patient's mortality rate 12 months after a myocardial infarction. As an example of prediction (Prd), FIG. 4 shows that the patient's mortality rate is 95% if a beta-blocker is not administered after a myocardial infarction.

[0054] In Figure 4, the red part indicates the patient's condition that contributes to the patient's death, and the blue part indicates the patient's condition that contributes to the patient's survival. The prediction (Prd) of the artificial intelligence system (10) of this embodiment can provide the contribution to the current prediction (Prd) value in this way, making it easier to analyze and understand.

[0055] Referring to Figure 5, it can be seen that the mortality rate is reduced to 38% when a beta-blocker is administered. In other words, medical professionals in charge of treating patients can be helped in diagnosis and prescription by comparing the prediction (Prd) in Figure 4 with the prediction (Prd) in Figure 5.

[0056] FIG. 6 is a block diagram for explaining the configuration of the artificial intelligence system of FIG.

[0057] Referring to FIG. 6, the artificial intelligence system (10) may include a data module (100), a Reliable and Interpretable AI System (RIAS) (200), and a model module (300).

[0058] The data module (100) can receive PHR data (DATA_PHR) and instructions (Inst) and perform preprocessing. The data module (100) can preprocess the PHR data (DATA_PHR) according to the instructions (Inst) to generate final data (Data_F). The data module (100) can transmit the final data (Data_F) to the RIAS (200) and the model module (300).

[0059] The model module (300) can receive instructions (Inst) and final data (Data_F) and train an artificial intelligence model to perform prediction. The model module (300) can generate a pre-prediction (Prd_P) through the artificial intelligence model. The model module (300) can transmit the pre-prediction (Prd_P) to the RIAS (200). In this case, the pre-prediction (Prd_P) can be a value between 0 and 1.

[0060] The RIAS (200) can receive instructions (Inst) and final data (Data_F). The RIAS (200) can also receive a pre-prediction (Prd_P) from the model module (300). The RIAS (200) can use the final data (Data_F) and instructions (Inst) to derive the pre-prediction (Prd_P) into a final prediction (Prd). The RIAS (200) can evaluate and tune the pre-prediction (Prd_P).

[0061] In this case, the data module 100, the model module 300, and the RIAS 200 can operate in a pipeline system regardless of the type of data and the type of model. That is, a general artificial intelligence system generates predictions using an artificial intelligence model determined according to the type of data, and when the data or the artificial intelligence model changes, a new pipeline system must be implemented instead of the existing pipeline system.

[0062] In contrast, in the artificial intelligence system according to this embodiment, the data module (100) converts changed data into the same data frame and provides it, and the model module (300) objectifies it so that common functions can be utilized among various artificial intelligence models, and the RIAS (200) can derive the final prediction (Prd) independently of the data and artificial intelligence model without configuring a new pipeline system.

[0063] FIG. 7 is a block diagram for explaining the structure of the data module of FIG.

[0064] Referring to FIG. 7, the data module (100) may include a preprocessing module (110) and an imputation module (120).

[0065] The pre-processing module 110 can receive PHR data (DATA_PHR) and instructions (Inst). At this time, the instructions (Inst) can include information on a data frame. The pre-processing module 110 can pre-process the PHR data (DATA_PHR) using the information on the data frame included in the instructions (Inst). As a result, the pre-processing module 110 can generate pre-processed data (Data_P).

[0066] The imputation module (120) can receive preprocessed data (Data_P) from the preprocessing module (110). Additionally, the imputation module (120) can receive instructions (Inst) and determine a method for handling missing values ​​contained in the instructions (Inst).

[0067] The imputation module 120 may process missing values ​​using at least one of the following methods: mean substitution, median substitution, mode substitution, regression substitution, K-nearest neighbor substitution, and multiple imputation. The method selected by the imputation module 120 may be a preset method or a method specified by an instruction (Inst). The imputation module 120 may remove missing values ​​from the preprocessed data (Data_P) to generate the final data (Data_F).

[0068] FIG. 8 is a block diagram for explaining the structure of the RIAS of FIG.

[0069] Referring to FIG. 8, the RIAS (200) may include at least one of a training module (210), a calibration module (220), a local explanation module (230), and an alternative factual explanation module (240).

[0070] The training module (210) can inherit the pre-prediction (Prd_P) model received by the RIAS (200) from the model module (300) and tune, learn, and evaluate the hyperparameters of the model. The training module (210) can continue to operate even if the artificial intelligence model selected by the model module (300) changes.

[0071] The calibration module 220 can convert the pre-prediction (Prd_P) into a final prediction (Prd). The calibration module 220 can convert the pre-prediction (Prd_P) into a prediction (Prd) that is acceptable to a user as a predicted probability of an actual event occurring by processing the pre-prediction (Prd_P) to mimic the likelihood of an actual event occurring.

[0072] The locality explanation module 230 can inform how much the features of the final data (Data_F) used in the prediction (Prd) contributed to the prediction (Prd). That is, the locality explanation module 230 can generate the contribution portion of FIGS. 4 and 5.

[0073] The alternative factual explanation module (240) can provide insights about what the user should do to achieve the desired outcome based on the pre-prediction (Prd_P). Thus, the prediction (Prd) can include the insights. For example, the alternative factual explanation module (240) can confirm the mortality rates with and without beta blocker administration in Figures 4 and 5 and provide an insight suggesting beta blocker administration.

[0074] FIG. 9 is a block diagram for explaining the model module of FIG.

[0075] Referring to Figure 9, the model module (300) can select one of various artificial intelligence models to perform learning and inference. Models that share common functions can be classified into the same type. Specifically, the artificial intelligence models can include a first type (310), a second type (320), and a third type (330). While Figure 9 shows three types, this embodiment is not limited to this.

[0076] The first type (310), the second type (320), and the third type (330) may be, for example, Gradient Boosted Decision Trees (GBDT), Pytorch tabula, and Saint, respectively, but this is merely an example and the present embodiment is not limited thereto.

[0077] Therefore, the first model (311), the second model (312), and the third model (313) belonging to the first type (310) may be XGBoost, LightGBM, and CatBoost, respectively, but this embodiment is not limited thereto. The fourth model (321), the fifth model (322), the sixth model (323), the seventh model (324), and the eighth model (325) belonging to the second type (320) may be FTTransformer, TabTransformer, TabNet, AutoInt, and MLP, respectively, but this embodiment is not limited thereto.

[0078] In this way, the model module (300) can specify common objects for the same type of artificial intelligence models having the same functions, so that the pipeline can be shared as much as possible.

[0079] This embodiment operates independently of the data or the AI ​​model, and the same flow can be used regardless of the data or model used. This makes it possible to derive predictions much more efficiently than existing AI systems.

[0080] Hereinafter, a prediction method for an artificial intelligence system according to some embodiments of the present invention will be described with reference to Figures 6 to 8 and Figures 10 to 12. Portions that overlap with the above-described embodiments will be simplified or omitted.

[0081] The prediction method of the AI ​​system according to some embodiments of the present invention may be stored in a recording medium and executed as a computer program. In this case, it may be organically combined with the AI ​​system to operate, but the present embodiments are not limited thereto.

[0082] Figure 10 is a flowchart illustrating a prediction method of an artificial intelligence system according to some embodiments of the present invention, and Figure 11 is a flowchart illustrating in detail the final data generation step of Figure 10. Figure 12 is a flowchart illustrating in detail the pre-prediction derivation step of Figure 10.

[0083] Referring to FIG. 10, the PHR data is pre-processed to generate final data (S100).

[0084] Referring to FIG. 11 in detail, the PHR data is converted into a preset data frame to generate pre-processed data (S110).

[0085] 7, the pre-processing module 110 can receive PHR data (DATA_PHR) and instructions (Inst). At this time, the instructions (Inst) can include information on a data frame. The pre-processing module 110 can pre-process the PHR data (DATA_PHR) using the information on the data frame included in the instructions (Inst). As a result, the pre-processing module 110 can generate pre-processed data (Data_P).

[0086] Referring again to FIG. 11, the missing values ​​of the pre-processed data are processed to generate the final data (S120).

[0087] 7, the imputation module 120 can receive preprocessed data (Data_P) from the preprocessing module 110. Furthermore, the imputation module 120 can receive instructions (Inst) and determine a missing value processing method included in the instructions (Inst). The imputation module 120 can remove missing values ​​from the preprocessed data (Data_P) to generate final data (Data_F).

[0088] Referring again to FIG. 10, the final data is input to the selected artificial intelligence model to derive a preliminary prediction (S200).

[0089] In detail, referring to FIG. 12, an artificial intelligence model is selected according to an instruction (S210), and a pre-prediction is derived through the selected artificial intelligence model (S220).

[0090] Specifically, referring to Figure 6, the model module (300) can receive instructions (Inst) and final data (Data_F) and train an artificial intelligence model to perform prediction. The model module (300) can generate a pre-prediction (Prd_P) through the artificial intelligence model. The model module (300) can transmit the pre-prediction (Prd_P) to the RIAS (200).

[0091] Referring again to FIG. 10, the pre-prediction is corrected to derive the final prediction (S300).

[0092] 8, the calibration module 220 can convert the preliminary prediction (Prd_P) into a final prediction (Prd). The calibration module 220 can process the preliminary prediction (Prd_P) to mimic the likelihood of an actual event occurring, and convert it into a prediction (Prd) that the user can accept as the predicted probability of the actual event occurring.

[0093] FIG. 13 is a diagram illustrating the hardware configuration of an artificial intelligence system according to some embodiments of the present invention.

[0094] An artificial intelligence system according to some embodiments of the present invention may be implemented as an electronic device 1000. The electronic device 1000 may include a processor 1010, an input / output device 1020 (I / O), a memory 1030, an interface 1040, and a bus 1050. The processor 1010, the input / output device 1020, the memory 1030, and / or the interface 1040 may be coupled to one another via a bus 1050. The bus 1050 corresponds to a path through which data travels.

[0095] Specifically, the processor (1010) may include at least one of a CPU (Central Processing Unit), an MPU (Micro Processor Unit), an MCU (Micro Controller Unit), a GPU (Graphic Processing Unit), a microprocessor, digital signal processing, a microcontroller, a financial application processor (AP), and logic elements capable of performing similar functions.

[0096] The processor (1010) and memory (1030) in FIG. 1 may be configured as software on the electronic device (1000) or as separate hardware.

[0097] The input / output device (1020) may include at least one of a keypad, a keyboard, a touch screen, and a display device.

[0098] The memory 1030 can load data and / or programs. In this case, the memory 1030 can include high-speed DRAM and / or SRAM as operating memory to improve the operation of the processor 1010. The memory 1030 can include one or more volatile memory devices, such as Double Data Rate Static DRAM (DDR SDRAM) and Single Data Rate SDRAM (SDR SDRAM), and / or one or more non-volatile memory devices, such as Electrical Erasable Programmable ROM (EEPROM) and flash memory.

[0099] The interface 1040 may transmit data to or receive data from a communication network. The interface 1040 may be wired or wireless. For example, the interface 1040 may include an antenna or a wired / wireless transceiver.

[0100] The storage can store and preserve data and / or programs. The storage can include one or more non-volatile memory devices such as a solid state drive (SSD), a hard drive, or a flash memory. In the present invention, the storage can store a computer program configured with instructions for executing the method for providing a digital will service described above.

[0101] The input / output device 1020 may be a personal digital assistant (PDA), a portable computer, a web tablet, a wireless phone, a mobile phone, a digital music player, a memory card, or any electronic product capable of transmitting and / or receiving information in a wireless environment.

[0102] Furthermore, the quantum circuit simulation device according to the embodiment of the present invention may be a device formed by connecting a plurality of electronic devices 1000 to each other via a network. In such a case, each module or a combination of modules may be embodied as the electronic device 1000. However, the present embodiment is not limited thereto.

[0103] Additionally, the quantum circuit simulation apparatus may be embodied as at least one of a workstation, a data center, an internet data center (IDC), a DAS (direct attached storage) system, a SAN (storage area network) system, a NAS (network attached storage) system, and a RAID (redundant array of inexpensive disks, or redundant array of independent disks) system, but the present embodiment is not limited thereto.

[0104] Furthermore, the quantum circuit simulation device can transmit data via a network. The network can include a network using wired Internet technology, wireless Internet technology, and short-range communication technology. The wired Internet technology can include, for example, at least one of a local area network (LAN) and a wide area network (WAN).

[0105] The above description merely exemplifies the technical concept of the present embodiment, and various modifications and variations may be made by a person skilled in the art without departing from the essential characteristics of the present embodiment. Therefore, the present embodiment is intended to illustrate, not limit, the technical concept of the present embodiment, and the scope of the technical concept of the present embodiment is not limited by such an embodiment. The scope of protection of the present embodiment should be interpreted by the following claims, and all technical concepts within the scope equivalent thereto should be interpreted as being included in the scope of the present embodiment.

Claims

1. a data module that receives patient PHR (Personal Health Record) data and performs pre-processing of the PHR data to generate final data; a model module that receives the final data and instructions and derives a pre-prediction for the patient via at least one artificial intelligence model selected in accordance with the instructions; and a Reliable and Interpretable AI System (RIAS) that is separated from the data module and the model module and operates independently, receives the pre-prediction and the final data, and converts the pre-prediction into a final prediction; The data module includes an imputation module that processes missing values ​​in the pre-processed data to generate final data. Artificial intelligence system.

2. The artificial intelligence system of claim 1 , wherein the patient comprises a myocardial infarction patient.

3. The artificial intelligence system of claim 1 , wherein the final prediction includes a mortality rate for the patient after a specified period of time.

4. The artificial intelligence system of claim 3 , wherein the final prediction includes a weighting of factors that positively contribute to the mortality rate and a weighting of factors that negatively contribute to the mortality rate.

5. The PHR data may be used to assess the mortality rate of a patient after acute myocardial infarction or the reversal rate of a patient after acute myocardial infarction. The artificial intelligence system of claim 1 including remodeling data.

6. The PHR data is data that has been modified or manipulated from the mortality or remodeling data. The artificial intelligence system of claim 5 further comprising:

7. The data module includes a pre-processing module that receives instructions including a preset data frame, and pre-processes the PHR data according to the data frame to generate the pre-processed data. The artificial intelligence system of claim 1 .

8. The artificial intelligence system of claim 7 , wherein the imputation module receives the instructions and processes the missing values ​​according to the data frame.

9. 10. The artificial intelligence system of claim 1, wherein the imputation module handles the missing values ​​via at least one of mean imputation, median imputation, mode imputation, regression imputation, K-nearest neighbor imputation, and multiple imputation.

10. A data module that receives patient PHR (Personal Health Record) data and performs pre-processing of the PHR data to generate final data; a model module that receives the final data and instructions and derives a pre-prediction for the patient via at least one artificial intelligence model selected in accordance with the instructions; and and a Reliable and Interpretable AI System (RIAS) that is separate from the data module and the model module and operates independently, receives the preliminary prediction and the final data, and converts the preliminary prediction into a final prediction, wherein the instructions include information for a predetermined artificial intelligence model.

11. The artificial intelligence system of claim 10 , wherein the patient comprises a myocardial infarction patient.

12. The artificial intelligence system of claim 10 , wherein the final prediction includes a mortality rate for the patient after a specified period of time.

13. 13. The artificial intelligence system of claim 12, wherein the final prediction includes a weighting of factors that contribute positively to the mortality rate and a weighting of factors that contribute negatively to the mortality rate.

14. The artificial intelligence system of claim 10 , wherein the PHR data includes a patient's post-acute myocardial infarction mortality rate or a patient's post-acute myocardial infarction reverse remodeling data.

15. The artificial intelligence system of claim 14 , wherein the PHR data further includes data that alters or manipulates the mortality or remodeling data.

16. The artificial intelligence system of claim 10 , wherein the at least one artificial intelligence model can be classified into one type if they have common functions, and the model module objectifies the common functions according to the type.

17. 17. The artificial intelligence system of claim 16, wherein the types include at least one of Gradient Boosted Decision Trees (GBDT), pytorch tabula, and Saint.

18. A data module that receives a patient's PHR (Personal Health Record) data and performs pre-processing of the PHR data to generate final data; a model module that receives the final data and instructions and derives a pre-prediction for the patient via at least one artificial intelligence model selected in accordance with the instructions; and The system further includes a Reliable and Interpretable AI System (RIAS), which is separated from the data module and the model module and operates independently, receives the pre-prediction and the final data, and converts the pre-prediction into a final prediction, and the RIAS comprises: a training module that trains the artificial intelligence model, tunes hyperparameters of the artificial intelligence model, and evaluates the training module; a calibration module for correcting the pre-prediction to derive a final prediction; a local explanation module that performs a local explanation method on the contribution of each feature to the final data; and at least one counterfactual explanation module that performs a counterfactual explanation method on the final data. Artificial intelligence system.

19. The artificial intelligence system of claim 18 , wherein the patient comprises a myocardial infarction patient.

20. 20. The artificial intelligence system of claim 18, wherein the final prediction includes a mortality rate for the patient after a specified period of time.

21. 21. The artificial intelligence system of claim 20, wherein the final prediction includes a weighting of factors that contribute positively to the mortality rate and a weighting of factors that contribute negatively to the mortality rate.

22. 20. The artificial intelligence system of claim 18, wherein the PHR data comprises a patient's post-acute myocardial infarction mortality rate or a patient's post-acute myocardial infarction reverse remodeling data.

23. 23. The artificial intelligence system of claim 22, wherein the PHR data further includes data that alters or manipulates the mortality or remodeling data.

24. the pre-prediction is a predicted value between 0 and 1; the calibration module adjusts the pre-prediction to mimic the likelihood of an actual event occurring, to derive a final prediction; 19. The artificial intelligence system of claim 18.

25. memory; and a processor in conjunction with the memory to perform operations; the processor: Preprocessing the patient's PHR data and processing missing values ​​in the preprocessed data to generate final data; inputting the final data into a selected artificial intelligence model to derive a preliminary prediction; correcting the pre-prediction to derive a final prediction; Artificial intelligence system.

Images