Mental disease prediction and diagnosis classification apparatus by means of natural language expression and method therefor

Abstract

The present invention relates to a mental disease prediction and diagnosis classification apparatus by means of a natural language expression and a method therefor. According to the present invention, the mental disease prediction and diagnosis classification apparatus comprises: an input unit that receives, as an input from a medical personnel terminal, mental disease natural language expression information of a patient to be diagnosed; a preprocessing unit that tokenizes the mental disease natural language expression information by using a natural language processing (NLP) algorithm; and a disease prediction unit that inputs the tokenized mental disease natural language expression information into a pre-built diagnosis classification model to classify the corresponding patient to be diagnosed as having any one mental disease from among a generalized anxiety disorder, a panic disorder, a post-traumatic stress disorder, obsessive-compulsive thought or rumination, a remarkable obsessive-compulsive behavior, a social phobia, a specific phobia, agoraphobia, an adjustment disorder, an acute stress response, and a mixed obsessive-compulsive thought and behavior.

Description

A device and method for predicting and diagnosing mental illness using natural language expressions

[0001] The present invention relates to a device and method for predicting and diagnosing mental illness using natural language expressions, and more particularly, to a device and method for predicting and diagnosing mental illness using natural language expressions of patients expected to have mental illness among unstructured medical data, and performing diagnosis and classification.

[0002] Traditionally, psychiatrists diagnosed patients with mental illness based on their knowledge and experience, using standardized questionnaires and the patient's primary complaints. Because the diagnosis of mental illness relies on the specialist's knowledge and experience, different specialists can reach different diagnoses of mental illness, even when diagnosing the same patient.

[0003] Moreover, even when the same patient is treated by the same psychiatrist, the diagnosis given at the initial visit often differs from the diagnosis given at the follow-up visit approximately three months later. This can lead to confusion for patients, as medications administered at the initial visit may differ from those administered at the follow-up visit depending on the diagnosis.

[0004] Additionally, it takes a lot of time and resources to train psychiatrists with a wealth of knowledge and experience, and the number of patients with mental illness has been rapidly increasing recently, which may lead to an imbalance between supply and demand due to the increasing number of mental health diagnoses.

[0005] The technology underlying the present invention is disclosed in Republic of Korea Patent No. 10-2548241 (announced on June 28, 2023).

[0006] The technical task to be achieved by the present invention is to provide a mental illness prediction and diagnosis classification device and method for predicting mental illness and performing diagnosis classification through natural language expressions of patients expected to have mental illness among non-standard medical data.

[0007] According to an embodiment of the present invention for achieving such a technical task, a device for predicting and diagnosing and classifying mental illness using natural language expressions includes an input unit that receives natural language expression information of mental illness of a patient to be diagnosed from a medical professional terminal, a preprocessing unit that tokenizes the natural language expression information of mental illness using a natural language processing (NLP) algorithm, and a disease prediction unit that inputs the tokenized natural language expression information of mental illness into a pre-established diagnosis classification model, and classifies the corresponding patient to be diagnosed as having any one of the following mental illnesses: generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive-compulsive thoughts or rumination, prominent obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress response, and mixed obsessive-compulsive thoughts and behaviors.

[0008] The method may further include a data collection unit that collects clinical information of patients with mental illness from a clinical data warehouse, an analysis unit that classifies the collected clinical information according to diagnosis and labels it, and applies a natural language processing (NLP) algorithm to the labeled clinical information to analyze natural language expression information included in the clinical information, and a learning unit that builds a diagnosis classification model based on an artificial intelligence algorithm and inputs the analyzed natural language expression information into the built diagnosis classification model to learn it.

[0009] The above clinical information may include patient identification information, natural language expression information, and diagnosis.

[0010] The above analysis unit can classify the mental illness as one of generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, marked obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress reaction, and mixed obsessive-compulsive thoughts and behaviors using the diagnosis included in the collected clinical information, and can label the mental illness with a value of 1 to 11 according to the classified mental illness.

[0011] The above analysis unit can perform tokenization by applying a natural language processing (NLP) algorithm to the natural language expression information, analyze the tokenized words to extract words corresponding to mental illness expressions, and count the frequency of each extracted word.

[0012] The above analysis unit can calculate the length of each sentence included in the tokenized natural language expression information using a natural language processing (NLP) algorithm, and then analyze the calculated sentences to calculate the maximum length and average length.

[0013] The above learning unit can train a diagnostic classification model using the frequency of mental illness expressions, maximum sentence length, and average sentence length acquired in response to generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, prominent obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress reaction, and mixed obsessive-compulsive thoughts and behaviors.

[0014] The above diagnostic classification model analyzes the input natural language expression information to calculate the probability of each of generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, prominent obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress response, and mixed obsessive-compulsive thoughts and behaviors, and can predict the disease with the largest value among the calculated probabilities as the final mental illness.

[0015] In addition, according to an embodiment of the present invention, a method for predicting and diagnosing mental illness using a mental illness prediction and diagnosis classification device includes a step of receiving mental illness natural language expression information of a patient to be diagnosed from a medical professional terminal, a step of tokenizing the mental illness natural language expression information using a natural language processing (NLP) algorithm, and a step of inputting the tokenized mental illness natural language expression information into a pre-established diagnosis classification model, thereby classifying the corresponding patient to be diagnosed as having any one of the following mental illnesses: generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, prominent obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress response, and mixed obsessive-compulsive thoughts and behaviors.

[0016] In this way, according to the present invention, mental illness can be accurately predicted and diagnosed and classified using an artificial intelligence algorithm, thereby improving the efficiency of the diagnostic judgment work of psychiatrists, thereby enabling them to focus on essential treatment of mental illness patients.

[0017] Figure 1 is a schematic diagram of a mental illness prediction and diagnosis classification device according to an embodiment of the present invention.

[0018] FIG. 2 is a flowchart for explaining a mental illness prediction and diagnosis classification method using a mental illness prediction and diagnosis classification device according to an embodiment of the present invention.

[0019] Figure 3 is an example diagram for explaining clinical information collected in step S210 illustrated in Figure 2.

[0020] FIG. 4 is an example diagram for explaining the tokenized state of natural language expression information in step S130 illustrated in FIG. 2.

[0021] Figure 5 is an exemplary diagram for explaining a method for calculating the maximum length and average length in step S130 illustrated in Figure 2.

[0022] Figure 6 is an exemplary diagram for explaining step S270 illustrated in Figure 2.

[0023] Below, with reference to the attached drawings, embodiments of the present invention are described in detail so that those skilled in the art can easily implement them. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In addition, in the drawings, parts irrelevant to the description are omitted for clarity of description, and similar parts are designated with similar reference numerals throughout the specification.

[0024] Throughout the specification, whenever a part is said to "include" a component, this does not mean that it excludes other components, but rather that it may include other components, unless otherwise stated.

[0025] Then, with reference to the attached drawings, an embodiment of the present invention will be described in detail so that a person having ordinary knowledge in the technical field to which the present invention pertains can easily practice the present invention.

[0026] Hereinafter, a mental illness prediction and diagnosis classification device according to an embodiment of the present invention will be described in more detail using FIG. 1.

[0027] Figure 1 is a schematic diagram of a mental illness prediction and diagnosis classification device according to an embodiment of the present invention.

[0028] As illustrated in FIG. 1, a mental illness prediction and diagnosis classification device (100) according to an embodiment of the present invention includes a data collection unit (110), an analysis unit (120), a learning unit (130), an input unit (140), a preprocessing unit (150), and a disease prediction unit (160).

[0029] First, the data collection unit (110) collects clinical information of patients with mental illness from a clinical data warehouse (CDW).

[0030] The analysis unit (120) labels clinical information according to the diagnosis. In detail, the analysis unit (120) classifies the collected clinical information according to the 11 subtypes of anxiety disorders and then labels the classified clinical information from 1 to 11. Then, the analysis unit (120) applies a natural language processing (NLP) algorithm to the labeled clinical information to obtain tokenized natural language expression information. Then, the analysis unit (120) extracts words corresponding to the expression of neurological disorder from the tokenized natural language expression information and calculates the frequency. In addition, the analysis unit (120) calculates the maximum length and average length of sentences from the tokenized natural language expression information.

[0031] The learning unit (130) constructs a diagnostic classification model and trains the diagnostic classification model using the frequency of mental illness expressions, maximum sentence length, and average length obtained in response to anxiety disorders classified according to 11 subtypes.

[0032] The input unit (140) receives clinical information of the patient to be diagnosed from the medical terminal.

[0033] Next, the preprocessing unit (150) extracts natural language expression information included in the clinical information of the patient to be diagnosed, and performs tokenization by applying a natural language processing (NLP) algorithm to the extracted natural language expression information.

[0034] Finally, the disease prediction unit (160) inputs the tokenized natural language expression information into a diagnostic classification model whose training has been completed. Then, the diagnostic classification model determines whether the patient in question has a mental illness, and if determined to have a mental illness, classifies the patient as having one of the following mental illnesses: generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, marked obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress response, and mixed obsessive-compulsive thoughts and behaviors.

[0035] Hereinafter, a mental illness prediction and diagnosis classification method using a mental illness prediction and diagnosis classification device (100) according to an embodiment of the present invention will be described in more detail using FIGS. 2 to 7.

[0036] FIG. 2 is a flowchart for explaining a mental illness prediction and diagnosis classification method using a mental illness prediction and diagnosis classification device according to an embodiment of the present invention.

[0037] As illustrated in Fig. 2, the method for predicting and diagnosing mental illness includes a step of constructing a diagnostic classification model and a step of predicting whether a patient has a mental illness using the constructed diagnostic classification model and classifying it with a specific diagnosis.

[0038] First, the steps for building a diagnostic classification model will be described. The prediction and diagnostic classification system (100) collects clinical information of patients with mental illness from a clinical data warehouse (S210).

[0039] To elaborate, the data collection unit (110) collects clinical information on patients diagnosed with anxiety disorders among mental illness patients from a clinical data warehouse (CDW) based on clinical big data. Here, the clinical information includes patient identification information, natural language expression information, and a diagnosis.

[0040] Figure 3 is an example diagram for explaining clinical information collected in step S210 illustrated in Figure 2.

[0041] As illustrated in Figure 3, anxiety disorders in mental illness encompass 11 subtypes. These subtypes include generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or ruminations, marked obsessive-compulsive behaviors, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress response, and mixed obsessive-compulsive thoughts and behaviors.

[0042] In the present invention, mental illness was classified using clinical information corresponding to anxiety disorders including 11 subtypes, but data on other types of mental illness, such as depression as well as anxiety disorders, can be secured and classified.

[0043] Next, the analysis unit (130) labels the collected clinical information according to the types of anxiety disorders classified into 11 categories (S220).

[0044] The analysis unit (120) classifies the collected clinical information into 11 types according to the type of anxiety disorder, and labels the classified clinical information from 1 to 11. For example, clinical information classified as generalized anxiety disorder is labeled as 1, clinical information classified as panic disorder is labeled as 2, and clinical information classified as post-traumatic stress disorder is labeled as 3. In addition, clinical information classified as obsessive thoughts or rumination is labeled as 4, clinical information classified as marked obsessive-compulsive behavior is labeled as 5, and clinical information classified as social phobia is labeled as 6. In addition, clinical information classified as specific phobia is labeled as 7, clinical information classified as agoraphobia is labeled as 8, and clinical information classified as adjustment disorder is labeled as 9. Finally, clinical information classified as acute stress reaction is labeled as 10, and clinical information classified as mixed obsessive-compulsive thoughts and behaviors is labeled as 11.

[0045] When step S220 is completed, the analysis unit (120) extracts natural language expression information from the classified clinical information, and analyzes the extracted natural language expression information to obtain the frequency of mental illness expressions, maximum sentence length, and average length (S230).

[0046] FIG. 4 is an example diagram for explaining a state in which natural language expression information is tokenized in step S130 illustrated in FIG. 2, and FIG. 5 is an example diagram for explaining a method for calculating the maximum length and average length in step S130 illustrated in FIG. 2.

[0047] As illustrated in FIG. 4, the analysis unit (120) extracts natural language expression information from clinical information and performs tokenization by applying a natural language processing (NLP) algorithm to the extracted natural language expression information.

[0048] Next, the analysis unit (120) extracts words corresponding to mental illness expressions from the tokenized natural language expression information and counts the frequency of each extracted word.

[0049] In addition, as illustrated in FIG. 5, the analysis unit (120) calculates the length of each sentence included in the tokenized natural language expression information, and then analyzes the calculated sentences to calculate the maximum length and average length.

[0050] When step S230 is completed, the learning unit (130) trains a diagnostic classification model using the frequency of mental illness expressions, maximum sentence length, and average length (S240).

[0051] First, the learning unit (130) constructs a diagnostic classification model. The diagnostic classification model is a machine learning algorithm based on supervised learning, and is constructed using any one of the following algorithms: the k-nearest neighbors algorithm, the linear regression algorithm, the logistic regression algorithm, the support vector machine (SVM) algorithm, the decision tree, the random forest algorithm, and the neural network algorithm.

[0052] Next, the learning unit (130) trains the constructed diagnostic classification model using the frequency of mental illness expressions, maximum sentence length, and average sentence length acquired in step S230 as input data and the mental illness diagnosis as output data. At this time, the diagnostic classification model may output the probability of each of the 11 anxiety disorders as output values.

[0053] When learning of the diagnosis classification model is completed using steps S210 to S240, the mental illness prediction and diagnosis classification device (100) receives natural language expression information on the mental illness of the patient to be diagnosed (S250).

[0054] The input unit (140) receives natural language expression information written by a medical professional from a medical professional terminal. At this time, the natural language expression information includes the main complaint symptoms of the patient to be diagnosed.

[0055] Next, the preprocessing unit (150) tokenizes the input natural language expression information (S260).

[0056] To elaborate, the preprocessing unit (150) performs tokenization by applying a natural language processing (NLP) algorithm to the input natural language expression information.

[0057] Next, the disease prediction unit (160) inputs tokenized natural language expression information into a diagnostic classification model whose learning has been completed to classify the mental illness of the patient to be diagnosed (S270).

[0058] To elaborate, the disease prediction unit (160) inputs tokenized natural language expression information into a trained diagnostic classification model. The diagnostic classification model then analyzes the frequency of mental illness expressions, maximum sentence length, and average sentence length for the input natural language expression information, and determines whether the patient being diagnosed has a mental illness based on the analyzed results.

[0059] If a mental illness is determined to exist, the diagnostic classification model calculates the probability of being diagnosed with generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, prominent obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress reaction, and mixed obsessive-compulsive thoughts and behaviors, and predicts the disorder with the largest value among the calculated probabilities as the final mental illness.

[0060] Next, the disease prediction unit (160) transmits the predicted results to the medical terminal.

[0061] Figure 6 is an exemplary diagram for explaining step S270 illustrated in Figure 2.

[0062] As illustrated in FIG. 6, the input unit (140) assumes that the medical professional has received natural language expression information from the terminal, “I keep feeling anxious and wandering around. I feel anxious all the time. I feel like I can’t sit still and want to wander around.”

[0063] Then, the disease prediction unit (160) inputs tokenized natural language expression information into a diagnosis classification model, calculates the frequency of the mental illness expression “anxiety” and the maximum and average sentence lengths, and predicts the disability of the patient to be diagnosed using the calculated frequency of mental illness expressions, maximum and average sentence lengths.

[0064] And, assuming that among the 11 anxiety disorders from the diagnostic classification model, panic disorder obtained the highest probability value with a probability of 56%, as shown in Fig. 6, the disease prediction unit (160) diagnoses the mental illness of the patient to be diagnosed as panic disorder and outputs it.

[0065] In this way, the mental illness prediction and diagnosis classification device according to the present invention can precisely predict and diagnose and classify mental illness using an artificial intelligence algorithm, thereby improving the efficiency of the diagnosis and judgment work of psychiatrists, thereby enabling them to focus on essential treatment of mental illness patients.

[0066] While the present invention has been described with reference to the embodiments illustrated in the drawings, these are merely exemplary, and those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible. Therefore, the true scope of technical protection of the present invention should be determined by the technical spirit of the appended claims.

Claims

1. In a device for predicting and diagnosing mental illness using natural language expression, An input unit that receives natural language information on mental illness of a patient to be diagnosed from a medical professional terminal; A preprocessing unit that tokenizes the above mental illness natural language expression information using a natural language processing (NLP) algorithm, and A mental illness prediction and diagnosis classification device including a disease prediction unit that inputs the tokenized mental illness natural language expression information into a pre-established diagnosis classification model and classifies the corresponding diagnosis target patient as having any one of the following mental illnesses: generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, marked obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress reaction, and mixed obsessive-compulsive thoughts and behaviors.

2. In paragraph 1, A data collection department that collects clinical information on patients with mental illness from a clinical data warehouse. An analysis unit that classifies the collected clinical information according to diagnosis, performs labeling, and applies a natural language processing (NLP) algorithm to the labeled clinical information to analyze natural language expression information included in the clinical information, and A device for predicting and classifying mental illness, comprising a diagnostic classification model built based on an artificial intelligence algorithm, and further comprising a learning unit for inputting analyzed natural language expression information into the built diagnostic classification model to learn it.

3. In paragraph 2, The above clinical information is, A mental illness prediction and diagnosis classification device that includes patient identification information, natural language expression information, and diagnosis name.

4. In paragraph 2, The above analysis unit, Using the diagnosis included in the clinical information collected above, classify it as one of the following mental disorders: generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or ruminations, marked obsessive-compulsive behaviors, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress reaction, and mixed obsessive-compulsive thoughts and behaviors. A mental illness prediction and diagnosis classification device that labels values ​​from 1 to 11 according to the above-mentioned mental illness.

5. In paragraph 2, The above analysis unit, Tokenization is performed by applying a natural language processing (NLP) algorithm to the above natural language expression information, A mental illness prediction and diagnosis classification device that analyzes tokenized words to extract words corresponding to mental illness expressions and counts the frequency of each extracted word.

6. In paragraph 2, The above analysis unit, Using a natural language processing (NLP) algorithm, the length of each sentence contained in the tokenized natural language expression information is calculated, A mental illness prediction and diagnosis classification device that analyzes the generated sentences to calculate the maximum and average lengths.

7. In paragraph 2, The above learning department, A mental illness prediction and diagnosis classification device that trains a diagnostic classification model using the frequency of mental illness expressions, maximum sentence length, and average sentence length acquired in response to generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, prominent obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress reaction, and mixed obsessive-compulsive thoughts and behaviors.

8. In paragraph 7, The above diagnostic classification model is, By analyzing the input natural language expression information, the probability of each of generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, prominent obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress response, and mixed obsessive-compulsive thoughts and behaviors is calculated. A mental illness prediction and diagnosis classification device that predicts the disease with the largest value among the produced probabilities as the final mental illness.

9. In a method for predicting and diagnosing mental illness using a mental illness prediction and diagnosis classification device, A step for receiving natural language expression information on the mental illness of a patient to be diagnosed from a medical professional terminal. A step of tokenizing the above mental illness natural language expression information using a natural language processing (NLP) algorithm, and A method for predicting and diagnosing mental illness, comprising a step of inputting the above tokenized mental illness natural language expression information into a pre-established diagnostic classification model, and classifying the corresponding diagnostic target patient as having any one of the following mental illnesses: generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, marked obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress reaction, and mixed obsessive-compulsive thoughts and behaviors.

10. In paragraph 9, The step of collecting clinical information of patients with mental illness from a clinical data warehouse. A step of classifying the collected clinical information according to diagnosis and performing labeling, and applying a natural language processing (NLP) algorithm to the labeled clinical information to analyze natural language expression information included in the clinical information, and A method for predicting and classifying mental illness, further comprising the step of constructing a diagnostic classification model based on an artificial intelligence algorithm and inputting analyzed natural language expression information into the constructed diagnostic classification model to allow learning.

11. In paragraph 10, The above clinical information is, A method for predicting and classifying mental illness diagnosis, including patient identification information, natural language expression information, and diagnosis name.

12. In paragraph 10, The step of analyzing the above natural language expression information is: Using the diagnosis included in the clinical information collected above, classify it as one of the following mental disorders: generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or ruminations, marked obsessive-compulsive behaviors, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress reaction, and mixed obsessive-compulsive thoughts and behaviors. A method for predicting and diagnosing mental illness by labeling values ​​from 1 to 11 according to the above-mentioned mental illness.

13. In paragraph 10, The step of analyzing the above natural language expression information is: Tokenization is performed by applying a natural language processing (NLP) algorithm to the above natural language expression information, A method for predicting and diagnosing mental illness by analyzing tokenized words to extract words that represent mental illness and counting the frequency of each extracted word.

14. In paragraph 10, The step of analyzing the above natural language expression information is: Using a natural language processing (NLP) algorithm, the length of each sentence contained in the tokenized natural language expression information is calculated, A method for predicting and diagnosing mental illness by analyzing the generated sentences and calculating the maximum and average lengths.

15. In paragraph 10, The above learning steps are: A method for predicting and classifying mental illnesses, wherein a diagnostic classification model is trained using the frequency of mental illness expressions, maximum sentence length, and average sentence length obtained in response to generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or ruminations, prominent obsessive-compulsive behaviors, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress response, and mixed obsessive-compulsive thoughts and behaviors.

16. In paragraph 15, The above diagnostic classification model is, By analyzing the input natural language expression information, the probability of each of generalized anxiety disorder, panic disorder, post-traumatic stress disorder, obsessive thoughts or rumination, prominent obsessive-compulsive behavior, social phobia, specific phobia, agoraphobia, adjustment disorder, acute stress response, and mixed obsessive-compulsive thoughts and behaviors is calculated. A mental illness prediction and diagnosis classification method that predicts the disease with the largest value among the produced probabilities as the final mental illness.

Images