Method and device for predicting analysis results based on deep learning

Abstract

The method and device for predicting analytical results based on deep learning according to a preferred embodiment of the present invention can shorten the time it takes to confirm results by predicting analytical results of immune reaction analysis-based kits such as lateral flow assay (LFA) and antigen-antibody-based diagnostic kits based on deep learning.

Description

[Technical Field]

[0001] The present invention relates to a method and apparatus for predicting analysis results based on deep learning, and more particularly to a method and apparatus for predicting analysis results based on deep learning. [Background technology]

[0002] Diagnostic tests using lateral flow assays (LFA), which involve collecting samples from specimens and using the collected samples, take a long time (10 to 30 minutes). Lateral flow assays exhibit different behaviors depending on the sample concentration and reaction time, and diagnostic tests using LFA can be judged after approximately 15 minutes when the reaction has fully progressed. However, for certain diseases such as myocardial infarction, results are often required within 10 minutes, and recently, there has been a significant increase in the need for quick diagnoses within 5 minutes from the perspective of both hospitals and patients. Summary of the Invention [Problem to be solved by the invention]

[0003] The object of the present invention is to provide a method and device for predicting analytical results based on deep learning, which predicts analytical results of immune reaction analysis-based kits such as lateral flow assays (LFAs), antigen-antibody-based diagnostic kits, etc. based on deep learning.

[0004] Other objects not explicitly stated in the present invention can be considered within the scope that can be easily inferred from the following detailed description and its effects. [Means for solving the problem]

[0005] To achieve the above object, a method for predicting an analytical result based on deep learning according to a preferred embodiment of the present invention includes the steps of acquiring a reaction image for a predetermined initial period of interaction between a sample obtained from a specimen and an optical-based kit, and predicting a concentration for a predetermined result time based on the reaction image for the predetermined initial period using an analytical result prediction model that has been previously trained and constructed.

[0006] Here, the reaction image acquiring step may include acquiring a plurality of the reaction images at predetermined time intervals during the predetermined initial period.

[0007] Here, the analysis result prediction model includes an image generator, which includes a convolutional neural network (CNN), a long short-term memory (LSTM), and a generative adversarial network (GAN), and generates a predicted image corresponding to a predetermined result time based on an input response image, and outputs the generated predicted image; and a regression model, which includes the convolutional neural network (CNN), and outputs a predicted concentration for the predetermined result time based on the predicted image generated by the image generator, and the regression model may be trained using the training data so that a difference between a predicted concentration for the predetermined result time obtained based on the response image of the training data and an actual concentration for the predetermined result time of the training data is minimized.

[0008] Here, the image generator may include an encoder that acquires a feature vector from an input response image using the convolutional neural network (CNN), acquires a latent vector based on the acquired feature vector using the long short-term memory (LSTM), and outputs the acquired latent vector, and a decoder that generates the predicted image using the generative adversarial neural network (GAN) based on the latent vector acquired through the encoder, and outputs the generated predicted image.

[0009] Here, the decoder includes a generator that generates the predicted image based on the latent vector and outputs the generated predicted image, and a discriminator that compares the predicted image generated by the generator with an actual image corresponding to a predetermined result time of the learning data and outputs the comparison result, and can be trained using the learning data so that the predicted image obtained based on the latent vector is determined to be the actual image.

[0010] Here, the reaction image acquisition step may comprise acquiring the reaction image of an area corresponding to the test line when the optical base kit includes a test line and a control line.

[0011] Here, the reaction image acquiring step may comprise acquiring the reaction image of an area corresponding to one or more predetermined test lines among the plurality of test lines when the optical base kit includes a plurality of test lines.

[0012] Here, the reaction image acquiring step may involve acquiring the reaction image including all areas corresponding to one or more predetermined test lines among the plurality of test lines, or acquiring the reaction image for each test line such that the areas corresponding to one or more predetermined test lines among the plurality of test lines are divided by test line.

[0013] A computer program according to a preferred embodiment of the present invention for achieving the above technical objectives is stored on a computer-readable recording medium and causes a computer to execute any one of the above-mentioned methods for predicting analysis results based on deep learning.

[0014] To achieve the above object, a deep learning-based analysis result prediction device according to a preferred embodiment of the present invention is an analysis result prediction device based on deep learning that predicts analysis results based on deep learning, and includes a memory that stores one or more programs for predicting analysis results, and one or more processors that perform operations for predicting analysis results using the one or more programs stored in the memory, and the processor uses a pre-trained and constructed analysis result prediction model to predict concentrations for a predetermined result time based on a reaction image of a predetermined initial period of interaction between a sample obtained from a specimen and an optical-based kit.

[0015] Here, the processor may acquire a plurality of the reaction images at predetermined time intervals during the predetermined initial period.

[0016] Here, the analysis result prediction model includes an image generator, which includes a convolutional neural network (CNN), a long short-term memory (LSTM), and a generative adversarial network (GAN), and generates a predicted image corresponding to a predetermined result time based on an input response image, and outputs the generated predicted image; and a regression model, which includes the convolutional neural network (CNN), and outputs a predicted concentration for the predetermined result time based on the predicted image generated by the image generator, and the regression model may be trained using the training data so that a difference between a predicted concentration for the predetermined result time obtained based on the response image of the training data and an actual concentration for the predetermined result time of the training data is minimized.

[0017] Here, the image generator may include an encoder that acquires a feature vector from an input response image using the convolutional neural network (CNN), acquires a latent vector based on the acquired feature vector using the long short-term memory (LSTM), and outputs the acquired latent vector, and a decoder that generates the predicted image using the generative adversarial neural network (GAN) based on the latent vector acquired through the encoder, and outputs the generated predicted image. [Effects of the Invention]

[0018] According to a preferred embodiment of the present invention, a method and apparatus for predicting analytical results based on deep learning can shorten the time required to confirm the results by predicting analytical results of immune reaction analysis-based kits such as lateral flow assays (LFAs) and antigen-antibody-based diagnostic kits based on deep learning.

[0019] The effects of the present invention are not limited to those mentioned above, and other effects not mentioned above will be clearly understood by those skilled in the art from the following description. [Brief explanation of the drawings]

[0020] [Figure 1] 1 is a block diagram illustrating an analysis result prediction device based on deep learning according to a preferred embodiment of the present invention. FIG. [Figure 2] 1 is a diagram illustrating a process of predicting an analysis result according to a preferred embodiment of the present invention; [Figure 3] 1 is a diagram illustrating a process of predicting a change in color intensity over time based on a reaction image according to a preferred embodiment of the present invention. [Figure 4] 1 is a diagram illustrating a process of predicting concentration with respect to result time based on a reaction image for an initial period according to a preferred embodiment of the present invention. [Figure 5] 1 is a flow chart illustrating a method for predicting analysis results based on deep learning according to a preferred embodiment of the present invention. [Figure 6] 1 is a diagram illustrating an example of the structure of an analysis result prediction model according to a preferred embodiment of the present invention; [Figure 7] 7 is a diagram illustrating an example of an analysis result prediction model shown in FIG. 6; [Figure 8] 10 is a diagram illustrating another example of the structure of an analysis result prediction model according to a preferred embodiment of the present invention. [Figure 9] 1 is a diagram illustrating learning data used in a learning process of an analysis result prediction model according to a preferred embodiment of the present invention. [Figure 10] 10 is a diagram for explaining the configuration of the learning data shown in FIG. 9. [Figure 11] 11 is a diagram for explaining an example of a reaction image shown in FIG. 10. [Figure 12]1 is a diagram illustrating an example of a pre-processing process for a reaction image according to a preferred embodiment of the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0021] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. Advantages and features of the present invention, as well as methods for achieving them, will become apparent from the following detailed description of the embodiments in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments set forth below and may be embodied in various different forms. However, the present embodiments are provided to fully disclose the present invention and to fully convey the scope of the invention to those skilled in the art, and the present invention is defined only by the scope of the claims. Like reference numerals refer to like elements throughout the specification.

[0022] Unless otherwise defined, all terms, including technical / scientific terms, used herein may be used in the sense that they can be commonly understood by those skilled in the art to which the present invention belongs. Furthermore, terms defined in commonly used dictionaries should not be interpreted ideally or excessively unless they are clearly and specifically defined.

[0023] In this specification, terms such as "first" and "second" are used to distinguish one component from another and should not be used to limit the scope of rights. For example, a first component may be called a second component, and similarly, a second component may be called a first component.

[0024] In this specification, the identification numbers (e.g., a, b, c, etc.) used in each step are for convenience of explanation, and do not describe the order of each step. Each step may occur in a different order from the specified order unless a specific order is clearly stated in the context. That is, each step may be performed in the same order as specified, substantially simultaneously, or in the reverse order.

[0025] In this specification, the terms "have," "may have," "include," or "may include" indicate the presence of a given feature (e.g., a value, function, operation, or component such as a part) and do not exclude the presence of additional features.

[0026] Hereinafter, preferred embodiments of the method and apparatus for predicting analysis results based on deep learning according to the present invention will be described in detail with reference to the accompanying drawings.

[0027] First, an analysis result prediction device based on deep learning according to a preferred embodiment of the present invention will be described with reference to FIGS.

[0028] FIG. 1 is a block diagram illustrating an analysis result prediction device based on deep learning according to a preferred embodiment of the present invention, FIG. 2 is a diagram illustrating a process of predicting an analysis result according to a preferred embodiment of the present invention, FIG. 3 is a diagram illustrating a process of predicting a change in color intensity over time based on a reaction image according to a preferred embodiment of the present invention, and FIG. 4 is a diagram illustrating a process of predicting concentration over result time based on a reaction image for an initial period according to a preferred embodiment of the present invention.

[0029] Referring to FIG. 1, a deep learning-based analysis result prediction device (hereinafter referred to as the "analysis result prediction device") 100 according to a preferred embodiment of the present invention predicts analysis results of immune reaction analysis-based kits such as lateral flow assays (LFAs), antigen-antibody-based diagnostic kits, etc. based on deep learning.

[0030] Meanwhile, the operation of predicting an analysis result based on deep learning according to the present invention can be applied not only to a lateral flow assay that derives a result based on color intensity, but also to other analyses that derive a result based on fluorescence intensity. However, for the sake of convenience, the following description will be given assuming that the present invention predicts the result of a lateral flow assay.

[0031] To this end, the analysis result prediction device 100 may include one or more processors 110 , a computer-readable storage medium 130 , and a communication bus 150 .

[0032] The processor 110 can control the operation of the analysis result prediction device 100. For example, the processor 110 can execute one or more programs 131 stored on a computer-readable recording medium 130. The one or more programs 131 may include one or more computer-executable instructions, which, when executed by the processor 110, may be configured to cause the analysis result prediction device 100 to perform operations for predicting the outcome of an analysis (e.g., a lateral flow assay, etc.).

[0033] The computer-readable recording medium 130 is configured to store computer-executable instructions or program code, program data, and / or other suitable information for predicting the outcome of an assay (e.g., a lateral flow assay, etc.). The program 131 stored on the computer-readable recording medium 130 includes a set of instructions executable by the processor 110. In one embodiment, the computer-readable recording medium 130 may be memory (volatile memory such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, other forms of storage media that can be accessed by the assay result prediction apparatus 100 and store desired information, or a suitable combination thereof.

[0034] The communication bus 150 interconnects various other components of the analysis result prediction device 100, including the processor 110, the computer-readable storage medium 130, and the like.

[0035] The analysis result prediction apparatus 100 may also include one or more input / output interfaces 170 that provide interfaces for one or more input / output devices, and one or more communication interfaces 190. The input / output interfaces 170 and the communication interfaces 190 are coupled to the communication bus 150. The input / output devices (not shown) may be coupled to other components of the analysis result prediction apparatus 100 via the input / output interfaces 170.

[0036] Referring to FIG. 2, the processor 110 of the analysis result prediction device 100 uses a previously trained and constructed analysis result prediction model to predict the concentration for a predetermined result time based on a reaction image of a predetermined initial period of interaction between a sample obtained from a specimen and an optical-based kit (e.g., an immune reaction analysis-based kit such as a lateral flow assay kit, an antigen-antibody-based diagnostic kit such as an LSPR, SPR, or fluorescence-based assay).

[0037] Here, the preset initial period and the preset result time may vary depending on the type or model of the optical base kit, and in particular, the preset result time refers to the time during which the final result for the sample can be confirmed. For example, the preset result time may be set to "15 minutes," and the preset initial period may be set to "0 to 5 minutes."

[0038] In this case, the processor 110 may acquire a plurality of reaction images in a predetermined time unit during a predetermined initial period. For example, if the predetermined initial period is '0 minutes to 5 minutes' and the predetermined time unit is '10 seconds', the processor 110 may acquire '6 x 5 = 30' reaction images.

[0039] The analysis result prediction models include convolutional neural networks (CNN), long short-term memory (LSTM), and generative adversarial networks (GAN), which are described in more detail below.

[0040] That is, as shown in Figure 3, the change in color intensity over time can be predicted from the reaction image of the interaction between an optical-based kit including one test line and one control line and a sample at a specific time point. As a result, as shown in Figure 4, the concentration for a predetermined result time (e.g., 15 minutes) can be predicted from the reaction image for a predetermined initial period.

[0041] Now, with reference to FIG. 5, a method for predicting analysis results based on deep learning according to a preferred embodiment of the present invention will be described.

[0042] FIG. 5 is a flow chart illustrating a method for predicting analysis results based on deep learning according to a preferred embodiment of the present invention.

[0043] Referring to FIG. 5, the processor 110 of the analysis result prediction device 100 acquires a reaction image of a predetermined initial period of interaction between the sample obtained from the specimen and the optical-based kit (S110).

[0044] In this case, the processor 110 may acquire a plurality of reaction images at preset time intervals during a preset initial period.

[0045] The processor 110 may also perform a pre-processing step on the reaction image before inputting the reaction image into the analysis result prediction model.

[0046] That is, if the optical base kit includes a test line and a control line, the processor 110 can acquire a reaction image of the area corresponding to the test line. Here, the size of the area may be preset, such as "200 x 412 size."

[0047] Meanwhile, if the optical base kit includes a plurality of test lines, the processor 110 can acquire a reaction image of an area corresponding to one or more predetermined test lines among the plurality of test lines.

[0048] In this case, the processor 100 may acquire a reaction image including all areas corresponding to one or more predetermined test lines among the plurality of test lines, or may acquire a reaction image for each test line such that the areas corresponding to one or more predetermined test lines among the plurality of test lines are divided by test line.

[0049] Thereafter, the processor 110 predicts the concentration for a predetermined result time based on the reaction image for a predetermined initial period using the analysis result prediction model that has been previously trained and constructed (S130).

[0050] Now, the structure of the analysis result prediction model according to a preferred embodiment of the present invention will be described with reference to FIGS.

[0051] FIG. 6 is a diagram illustrating an example of the structure of an analysis result prediction model according to a preferred embodiment of the present invention, and FIG. 7 is a diagram illustrating an example of implementing the analysis result prediction model shown in FIG. 6.

[0052] Referring to FIG. 6, an example of an analysis result prediction model according to the present invention may include an image generator and a regression model.

[0053] The image generator includes a convolutional neural network (CNN), a long-short-term memory (LSTM), and a generative adversarial neural network (GAN), and can generate a predicted image corresponding to a predetermined result time based on multiple input reaction images, and output the generated predicted image.

[0054] To this end, the image generator may include an encoder and a decoder.

[0055] The encoder acquires a feature vector for each of the multiple input response images using a convolutional neural network (CNN), acquires a latent vector based on the acquired feature vectors using a long-short-term memory (LSTM), and outputs the acquired latent vector. In other words, the encoder can generate a latent vector by calculating the relationship between density, color intensity change, and time from multiple feature vectors.

[0056] The decoder can generate a predicted image using a generative adversarial neural network (GAN) based on the latent vector obtained by the encoder and output the generated predicted image.

[0057] That is, the decoder may include a generator that generates a predicted image based on the latent vector and outputs the generated predicted image, and a discriminator that compares the predicted image generated by the generator with an actual image corresponding to a predetermined result time of the learning data and outputs the comparison result.

[0058] In this case, the decoder can be trained using training data so that the predicted image obtained based on the latent vector is determined to be the actual image.

[0059] The regression model includes a convolutional neural network (CNN) and can output a predicted density for a predetermined result time based on the predicted image generated by the image generator. That is, the regression model acquires a feature vector of the predicted image and passes the acquired feature vector through two linear layers to obtain a predicted density.

[0060] In this case, the regression model may be trained using the training data so that the difference between the predicted concentration for a predetermined result time obtained based on the reaction image of the training data and the actual concentration for the predetermined result time of the training data is minimized.

[0061] Meanwhile, the classifier is a module necessary for the learning process of the analysis result prediction model, and the classifier may be removed from the analysis result prediction model after the learning is completed.

[0062] For example, as shown in Figure 7, when multiple reaction images (Image 1 to Image k in Figure 7) acquired every 10 seconds during the initial period (10 to 200 seconds) are input into the analysis result prediction model, the analysis result prediction model can acquire feature vectors (Feature Vector_1 to Feature Vector_k in Figure 7) for each of the multiple reaction images (Image 1 to Image k in Figure 7) using the convolutional neural network (CNN) ResNet-18.The analysis result prediction model can then input each of the acquired feature vectors (Feature Vector_1 to Feature Vector_k in Figure 7) into the corresponding long-short-term memory (LSTM).

[0063] The analysis result prediction model can then input the latent vector output from the long-short-term memory (LSTM) into the generator, SRGAN, a generative adversarial neural network (GAN). The analysis result prediction model can then generate a predicted image corresponding to the result time (15 minutes) based on the latent vector. The analysis result prediction model can then input the generated predicted image into the classifier, ResNet, a convolutional neural network (CNN), and SRGAN. The classifier then compares the predicted image with the actual image corresponding to the result time (15 minutes) and provides the comparison result to the generator.

[0064] Therefore, the analysis result prediction model can output a predicted concentration for the result time (15 minutes) based on the predicted image.

[0065] Here, the analysis result prediction model can be trained using training data to minimize two losses. The first loss (Loss #1 in FIG. 7) is to determine whether the predicted image obtained based on the latent vector is the actual image. The second loss (Loss #2 in FIG. 7) is to minimize the difference between the predicted concentration for the result time (15 minutes) and the actual concentration for the result time (15 minutes).

[0066] FIG. 8 is a diagram illustrating another example of the structure of an analysis result prediction model according to a preferred embodiment of the present invention.

[0067] Referring to Figure 8, another example of the analysis result prediction model according to the present invention is substantially the same as the example of the analysis result prediction model described above (see Figure 6), and may be a model that omits the process of generating an image from the example of the analysis result prediction model (see Figure 6).

[0068] That is, another example of the analysis result prediction model may include an encoder and a regression model by removing the decoder from the example of the analysis result prediction model (see FIG. 6).

[0069] The encoder includes a convolutional neural network (CNN) and a long short-term memory (LSTM), and can generate latent vectors based on multiple input response images and output the generated latent vectors. More specifically, the encoder can obtain feature vectors using a convolutional neural network (CNN) for each of the multiple input response images, obtain latent vectors using a long short-term memory (LSTM) based on the obtained feature vectors, and output the obtained latent vectors.

[0070] The regression model includes a neural network (NN) and can output a predicted concentration for a predetermined result time based on a latent vector acquired through an encoder. In this case, the regression model can be trained using training data so that a difference between a predicted concentration for a predetermined result time acquired based on a reaction image of the training data and an actual concentration for the predetermined result time of the training data is minimized.

[0071] Now, the learning data used in the learning process of the analysis result prediction model according to the preferred embodiment of the present invention will be described with reference to FIGS.

[0072] FIG. 9 is a diagram for explaining the learning data used in the learning process of the analysis result prediction model according to a preferred embodiment of the present invention, FIG. 10 is a diagram for explaining the configuration of the learning data shown in FIG. 9, FIG. 11 is a diagram for explaining an example of the reaction image shown in FIG. 10, and FIG. 12 is a diagram for explaining an example of the preprocessing process of the reaction image according to a preferred embodiment of the present invention.

[0073] The learning data used in the learning process of the analysis result prediction model according to the present invention may consist of a plurality of pieces of learning data, as shown in FIG.

[0074] That is, each learning data may include all reaction images over time (reaction image 1 to reaction image n in FIG. 10) and all actual concentrations over time (actual concentration 1 to actual concentration n in FIG. 10), as shown in FIG. 10. Here, "reaction image 1 to reaction image k" in FIG. 10 indicate reaction images in the initial period. For example, reaction images can be acquired in 10-second increments from the reaction start time (0 minutes) to the result time (15 minutes), as shown in FIG. 11.

[0075] In this case, some of all the reaction images over time (reaction image 1 to reaction image n in FIG. 10) and all the actual concentrations over time (actual concentration 1 to actual concentration n in FIG. 10) may be used as training data, and the rest may be used as test data and validation data. For example, odd-numbered reaction images and actual concentrations may be used as training data to train an analysis result prediction model. And even-numbered reaction images and actual concentrations may be used as test data and validation data to test and validate the analysis result prediction model.

[0076] A pre-processing process for the reaction image may be performed before inputting the reaction image into the analysis result prediction model. For example, as shown in FIG. 12, a reaction image of the region corresponding to the test line in the entire image may be obtained.

[0077] Now, an example of an implementation of the analysis result prediction device 100 based on deep learning according to a preferred embodiment of the present invention will be described.

[0078] First, a reaction image of an initial period (e.g., 0 to 5 minutes) of the interaction between a sample obtained from a specimen and an optical-based kit (e.g., a lateral flow assay kit) is captured by a camera (not shown). In this case, the camera can capture images at preset time intervals (e.g., 10 seconds) to capture multiple reaction images. Of course, the camera can also capture a video of the initial period and extract image frames from the captured video at preset time intervals to capture multiple reaction images.

[0079] The camera then provides the reaction image to the analysis result prediction device 100 according to the present invention directly or via an external server via wireless / wired communication. Of course, if the analysis result prediction device 100 according to the present invention includes a photography module, the reaction image can also be obtained directly.

[0080] Therefore, the analysis result prediction device 100 according to the present invention can predict the concentration for the result time (e.g., 15 minutes) based on the reaction image for the initial period using a pre-stored analysis result prediction model, and output the result. Of course, the analysis result prediction device 100 according to the present invention can also provide the reaction image for the initial period to an external server storing the analysis result prediction model via wireless / wired communication, receive the predicted concentration for the result time from the external server, and output the result.

[0081] The operations according to the present embodiment may be embodied in the form of program instructions that can be executed by various computer means and recorded on a computer-readable recording medium. A computer-readable recording medium refers to any medium that participates in providing instructions to a processor for execution. The computer-readable recording medium may include program instructions, data files, data structures, or combinations thereof. Examples include magnetic media, optical recording media, and memory. The computer program may be distributed over computer systems connected via a network, so that computer-readable code is stored and executed in a distributed manner. Functional programs, codes, and code segments for implementing the present embodiment should be easily construed by programmers skilled in the art to which the present embodiment pertains.

[0082] The present embodiment is intended to explain the technical idea of ​​the present embodiment, and does not limit the scope of the technical idea of ​​the present embodiment. The scope of protection of the present embodiment should be interpreted by the appended claims, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of rights of the present embodiment. [Explanation of symbols]

[0083] 100 Analysis result prediction device 110 Processor 130 Computer-readable recording medium 131 Program 150 Communication Bus 170 Input / Output Interface 190 Communication Interface

Claims

1. acquiring a reaction image for a predetermined initial period based on an immune reaction between a sample obtained from a specimen and an optical-based diagnostic kit; predicting the concentration of the sample for a predetermined result time based on the reaction image for the predetermined initial period using a previously trained and constructed analysis result prediction model; Including, The analysis result prediction model includes a convolutional neural network (CNN), a long short-term memory (LSTM), and a generative adversarial neural network (GAN), and an image generator that generates a predicted image corresponding to a predetermined result time based on an input reaction image and outputs the generated predicted image; and a regression model including the convolutional neural network (CNN) that outputs a predicted concentration for a predetermined result time based on the predicted image generated by the image generator. Including, The regression model is trained using the training data so that the difference between the predicted concentration for a predetermined result time obtained based on the reaction image of the training data and the actual concentration for the predetermined result time of the training data is minimized.

2. The reaction image acquisition step includes: The method of claim 1 , further comprising acquiring a plurality of the reaction images at predetermined time intervals during the predetermined initial period.

3. The image generator an encoder that acquires a feature vector from an input response image using the convolutional neural network (CNN), acquires a latent vector based on the acquired feature vector using the long short-term memory (LSTM), and outputs the acquired latent vector; a decoder that generates the predicted image using the generative adversarial neural network (GAN) based on the latent vector obtained by the encoder and outputs the generated predicted image; The deep learning-based analysis result prediction method of claim 1, comprising:

4. The decoder a generator that generates the predicted image based on the latent vector and outputs the generated predicted image; a discriminator that compares the predicted image generated by the generator with an actual image corresponding to a predetermined result time of the learning data and outputs a comparison result; Including, The method of claim 3 , wherein the method is trained using the training data so that the predicted image obtained based on the latent vector is determined to be the actual image.

5. The reaction image acquisition step includes:

2. The deep learning-based analysis result prediction method according to claim 1, further comprising acquiring the reaction image of an area corresponding to a test line when the optical-based diagnostic kit includes a test line and a control line.

6. The reaction image acquisition step includes:

6. The deep learning-based analysis result prediction method of claim 5, further comprising acquiring the reaction image of an area corresponding to one or more predetermined test lines among the plurality of test lines when the optical-based diagnostic kit includes a plurality of test lines.

7. The reaction image acquisition step includes: obtaining the reaction image including all areas corresponding to one or more predetermined test lines among the plurality of test lines; or 7. The method of claim 6, further comprising acquiring the response image for each test line so that an area corresponding to one or more predetermined test lines among the plurality of test lines is divided by the test line.

8. A computer program stored on a computer-readable recording medium for executing the analysis result prediction method based on deep learning described in any one of claims 1, 2, and 3 to 7.

9. An analysis result prediction device based on deep learning that predicts an analysis result based on deep learning, a memory storing one or more programs for predicting analytical results; one or more processors that perform operations to predict analysis results according to the one or more programs stored in the memory; Including, The processor: By using a previously trained and constructed analysis result prediction model, a concentration of the sample for a predetermined result time is predicted based on a reaction image of a predetermined initial period based on an immune reaction between a sample obtained from a specimen and an optical-based diagnostic kit; The analysis result prediction model includes a convolutional neural network (CNN), a long short-term memory (LSTM), and a generative adversarial neural network (GAN), and an image generator that generates a predicted image corresponding to a predetermined result time based on an input reaction image and outputs the generated predicted image; a regression model including the convolutional neural network (CNN) that outputs a predicted concentration for a predetermined result time based on the predicted image generated by the image generator; Including, The regression model is a deep learning-based analysis result prediction device that is trained using the training data so as to minimize the difference between the predicted concentration for a predetermined result time obtained based on the reaction image of the training data and the actual concentration for the predetermined result time of the training data.

10. The processor: The deep learning-based analysis result prediction device of claim 9 , wherein a plurality of the reaction images are acquired at predetermined time intervals during the predetermined initial period.

11. The image generator an encoder that acquires a feature vector from an input response image using the convolutional neural network (CNN), acquires a latent vector based on the acquired feature vector using the long short-term memory (LSTM), and outputs the acquired latent vector; a decoder that generates the predicted image using the generative adversarial neural network (GAN) based on the latent vector obtained by the encoder and outputs the generated predicted image; The deep learning-based analysis result prediction device according to claim 9, comprising:

Images