Disease classification method using speech analysis and recording medium having recorded thereon same

Abstract

The present invention relates to a disease classification method using speech analysis and comprises the steps of: if speech information is input from a patient, extracting an acoustic component including a predetermined parameter from the input speech information; performing speech analysis by using the acoustic component; and as the result of the speech analysis, diagnosing and classifying a potential disease including a brain disease. According to the present invention, a disease may be prevented and prepared for in advance by diagnosing a potential disease by using speech information.

Description

Disease classification method using voice analysis and recording medium recording the same

[0001] The present invention relates to a disease classification technique using voice analysis.

[0002] As South Korea's aging population accelerates, the number of dementia patients is expected to rapidly increase. As of 2020, the number of dementia patients aged 65 and older was approximately 840,000, accounting for approximately 10.3% of the 8.13 million elderly population aged 65 and older. Furthermore, the number of dementia patients is rapidly increasing, projected to exceed 3 million by 2050. In addition to the number of dementia patients, the cost of dementia care is estimated at 20.61 million won per person per year in 2020. At the national level, the cost is projected to increase from approximately 17.3 trillion won in 2020 to approximately 56.9 trillion won in 2040.

[0003] In the United States, demand for elder care services is rapidly increasing due to the aging baby boomer generation. Consequently, the US senior home care market is projected to grow from $100 billion in 2016 to $225 billion in 2024. By 2025, the US elder care industry is expected to face a shortage of approximately 500,000 home helpers and 100,000 nursing assistants. Digitalization is expected to address this labor shortage.

[0004] Currently, dementia diagnosis and progression monitoring involve comprehensive assessments, including medical history taking, brain MRI, amyloid PET or cerebrospinal fluid amyloid testing, and cognitive function tests. These tests have been criticized for their cost, analysis time, accuracy, and risks (radiation exposure, invasiveness). Neuropsychological testing, in particular, requires interpretation and interpretation by a qualified neuropsychologist, requiring approximately two hours of testing and two hours of analysis time. However, due to a lack of sufficient resources, waiting times for testing can extend to several months.

[0005] Moreover, behavioral and psychological symptoms of dementia (BPSD), such as depression, anxiety, and aggression, are serious and significant problems in dementia patients. Currently, these symptoms are measured through questionnaires administered to patients or their caregivers. However, patients often have limited cognitive function and insight due to the nature of dementia, making it difficult for them to accurately understand and respond to symptom-related questionnaires. Caregivers often only have indirect access to the patient's emotional state, and if they don't interact frequently with the patient, they may not be able to fully grasp the patient's condition, significantly limiting their ability to assess the patient's symptoms.

[0006]

[0007] Therefore, for professional dementia management through prevention, diagnosis, treatment, and management, organic collaboration between related departments such as neurology and psychiatry is necessary in hospitals. In addition, for early detection or prevention of dementia, comprehensive management is necessary by linking hospitals, homes, social facilities, public medical care, and welfare facilities, not just the efforts of hospitals.

[0008] Recently, with the importance of early diagnosis and prevention of dementia and the complex healthcare system that connects dementia patients' families and society, research is being conducted on efficient diagnostic methods utilizing digital technologies such as artificial intelligence. In addition to research using medical markers in existing hospitals, research is also being conducted on the prevention and diagnosis of various dementias using natural language tools such as ChatGPT and artificial intelligence. Research using speech and language is particularly gaining attention due to its efficiency. Although these terms "speech" and "language" are similar, they are used distinctly in neurolinguistics. In other words, speech refers to the production of meaningful sounds through neural activity, whereas language is related to the processing process of encoding specific concepts according to certain rules to express individual thoughts and emotions or understand the intentions of others.

[0009] The process of producing speech sounds occurs when air from the lungs through respiration passes through the vocal cords in the larynx, causing them to vibrate. This process is called phonation, and these sounds are amplified and resonate as they pass through the pharynx. The resonated sounds then pass through the oral and nasal cavities. When the movements of the soft palate, tongue, and lips—the organs of articulation—are altered, various consonants and vowels are created. When intonation is added to the continuous production of these consonant and vowel sounds, prosody is formed. Therefore, if any of the above speech subsystems (respiration, phonation, resonance, articulation, prosody) are disrupted, problems with pronunciation and articulation arise, which is called a speech disorder. Therefore, speech disorders can be caused not only by damage to the central nervous system related to speech movement, but also by damage to the peripheral nerves that control the organs where speech is produced (e.g., vocal cords, tongue).

[0010] The present invention has been devised to solve the above problems, and its purpose is to provide a method for diagnosing and classifying potential diseases, including brain diseases, using a patient's voice information.

[0011] The purpose of the present invention is not limited to the purposes mentioned above, and other purposes not mentioned will be clearly understood by those skilled in the art from the description below.

[0012] To achieve the above purpose, the present invention relates to a disease classification method using voice analysis, comprising the steps of: when voice information is input from a patient, extracting an acoustic component including a predetermined parameter from the input voice information; performing voice analysis using the acoustic component; and diagnosing and classifying a potential disease including a brain disease based on the results of the voice analysis.

[0013] The parameters of the above acoustic component may include phonation, resonance, articulation, prosody, respiration, rate and regularity.

[0014] By performing a predetermined task on a patient, voice information can be input, and an acoustic component including a predetermined parameter can be extracted from the input voice information.

[0015] In the step of performing the above voice analysis, the voice analysis can be performed by diagnosing a speech disorder using the above acoustic component and diagnosing a disease causing the diagnosed speech disorder.

[0016] In one embodiment of the present invention, voice analysis can be performed by comparing each parameter with a normal distribution of parameters of the same age and gender as the patient.

[0017] According to the present invention, by diagnosing potential diseases using voice information, there is an effect of being able to prevent and prepare for diseases in advance.

[0018] In particular, the present invention has the effect of preventing potential dementia-related brain diseases and reducing their prevalence by analyzing voice information to diagnose and classify brain diseases caused by nervous system damage.

[0019] Figure 1 is a flowchart showing a disease classification method using voice analysis according to one embodiment of the present invention.

[0020] Figure 2 is a flowchart showing in more detail a disease classification method using voice analysis according to one embodiment of the present invention.

[0021] Figures 3 to 9 are flowcharts illustrating a detailed disease classification process in a disease classification method using voice analysis according to one embodiment of the present invention.

[0022] Figure 10 is a graph illustrating a normal distribution.

[0023] Figure 11 is a diagram showing the standard deviation values ​​in Figure 10.

[0024] FIG. 1 is a flowchart showing a disease classification method using voice analysis according to one embodiment of the present invention, and FIG. 2 is a flowchart showing a disease classification method using voice analysis according to one embodiment of the present invention in more detail.

[0025] Referring to FIGS. 1 and 2, when voice information is input (S110), an acoustic component is extracted from the voice information (S120).

[0026] More specifically, a predetermined task is performed on a patient to input voice information (S130), and an acoustic component including predetermined parameters is extracted from the input voice information and analyzed (S140).

[0027] To describe the analysis execution step (S140) in more detail, for each predetermined category item for diagnosing a disease, each parameter is compared with a normal distribution of parameters of the same age and gender as the target patient (S142), and a weight is applied to each category item (S144).

[0028] And, based on the analysis results, the disease is diagnosed and classified (S150).

[0029] In one embodiment of the present invention, parameters of the acoustic component may include phonation, resonance, articulation, prosody, respiration, rate, and regularity.

[0030] In the present invention, one or more tasks can be performed.

[0031] In one embodiment of the present invention, a total of four tasks (task 1, task 2, task 3, task 4) can be performed.

[0032] Task 1 involves inducing the patient to pronounce / a / as long and evenly as possible. This task is intended to determine the maximum phonation time.

[0033] Task 2 involves prompting the patient to rapidly and evenly pronounce / keo / , / teo / , and / peo / for 5 seconds each. This task is intended to examine oral mechanism, allowing for the assessment of the structure, range of motion, speed, and force of the articulatory organs, including the lips, tongue, jaw, and soft palate.

[0034] Task 3 involves prompting the subject to pronounce / ker-ter-per / in rapid and even succession. This task ensures that rate and regularity are maintained.

[0035] Task 4 is a task designed to elicit connected speech, assessing whether the patient can speak in sentences of five or more words. For example, the task encourages connected speech by asking questions like, "Tell me freely about today's weather," "Tell me about what happened this morning," and "Describe the picture you see." This task assesses parameters such as respiration, phonation, resonance, articulation, and prosody.

[0036] In the present invention, analysis is conducted by comparing each parameter with a normal distribution of parameters of the same age and gender as the patient.

[0037] The present invention is susceptible to various modifications and embodiments. Specific embodiments are illustrated and described in detail in the drawings. However, this is not intended to limit the present invention to specific embodiments, but rather to encompass all modifications, equivalents, and alternatives falling within the spirit and technical scope of the present invention.

[0038] The terminology used in this application is only used to describe specific embodiments and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, it should be understood that the terms "comprise" or "have" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but do not preclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0039] 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 pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and will not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0040] In addition, when describing with reference to the attached drawings, identical components will be assigned the same reference numerals regardless of the drawing numbers, and redundant descriptions thereof will be omitted. When describing the present invention, if a detailed description of a related known technology is judged to unnecessarily obscure the gist of the present invention, the detailed description will be omitted.

[0041] The subject performing the disease classification method using voice analysis of the present invention may be any computer device or server system performing the disease classification method using voice analysis, or may be a control unit or processor that generally controls the system or device performing the disease classification method using voice analysis. That is, the disease classification method using voice analysis of the present invention may be composed of an algorithm, which is a type of software, and the software may be executed in the control unit or processor of the server system or device performing the disease classification method using voice analysis.

[0042] The present invention relates to a disease classification method using voice analysis, and comprises the steps of: when voice information is input from a patient, extracting an acoustic component including a predetermined parameter from the input voice information; performing voice analysis using the acoustic component; and diagnosing and classifying a potential disease including a brain disease based on the results of the voice analysis.

[0043] The parameters of the above acoustic component may include phonation, resonance, articulation, prosody, respiration, rate and regularity.

[0044] By performing a predetermined task on a patient, voice information can be input, and an acoustic component including a predetermined parameter can be extracted from the input voice information.

[0045] In the step of performing the above voice analysis, the voice analysis can be performed by diagnosing a speech disorder using the above acoustic component and diagnosing a disease causing the diagnosed speech disorder.

[0046] In one embodiment of the present invention, voice analysis can be performed by comparing each parameter with a normal distribution of parameters of the same age and gender as the patient.

[0047] FIG. 1 is a flowchart showing a disease classification method using voice analysis according to one embodiment of the present invention, and FIG. 2 is a flowchart showing a disease classification method using voice analysis according to one embodiment of the present invention in more detail.

[0048] Referring to FIGS. 1 and 2, when voice information is input (S110), an acoustic component is extracted from the voice information (S120).

[0049] More specifically, a predetermined task is performed on a patient to input voice information (S130), and an acoustic component including predetermined parameters is extracted from the input voice information and analyzed (S140).

[0050] To describe the analysis execution step (S140) in more detail, for each predetermined category item for diagnosing a disease, each parameter is compared with a normal distribution of parameters of the same age and gender as the target patient (S142), and a weight is applied to each category item (S144).

[0051] And, based on the analysis results, the disease is diagnosed and classified (S150).

[0052] In one embodiment of the present invention, parameters of the acoustic component may include phonation, resonance, articulation, prosody, respiration, rate, and regularity.

[0053] In the present invention, one or more tasks can be performed.

[0054] In one embodiment of the present invention, a total of four tasks (task 1, task 2, task 3, task 4) can be performed.

[0055] Task 1 involves inducing the patient to pronounce / a / as long and evenly as possible. This task is intended to determine the maximum phonation time.

[0056] Task 2 involves prompting the patient to rapidly and evenly pronounce / keo / , / teo / , and / peo / for 5 seconds each. This task is intended to examine oral mechanism, allowing for the assessment of the structure, range of motion, speed, and force of the articulatory organs, including the lips, tongue, jaw, and soft palate.

[0057] Task 3 involves prompting the subject to pronounce / ker-ter-per / in rapid and even succession. This task ensures that rate and regularity are maintained.

[0058] Task 4 is a task designed to elicit connected speech, assessing whether the patient can speak in sentences of five or more words. For example, the task encourages connected speech by asking questions like, "Tell me freely about today's weather," "Tell me about what happened this morning," and "Describe the picture you see." This task assesses parameters such as respiration, phonation, resonance, articulation, and prosody.

[0059] In the present invention, analysis is conducted by comparing each parameter with a normal distribution of parameters of the same age and gender as the patient.

[0060] Figure 10 is a graph illustrating a normal distribution.

[0061] For example, in the case of normal distribution data such as Figure 10, the standard deviation (SD) is a value that indicates how far the normal distribution data is from the mean.

[0062] Figure 11 is a diagram showing the standard deviation values ​​in Figure 10.

[0063] Referring to Figure 11, it can be confirmed that the standard deviation of ±1SD of the mean corresponds to ±1σ and includes 68.2%. In this case, the standard deviation is used only for data that is normally distributed.

[0064]

[0065] Figures 3 to 9 are flowcharts illustrating a detailed disease classification process in a disease classification method using voice analysis according to one embodiment of the present invention.

[0066] Referring to FIGS. 3 to 9, when voice information is input (S110), an acoustic component is extracted from the voice information (S120).

[0067] More specifically, a predetermined task is performed on a patient to input voice information (S130), and an acoustic component including predetermined parameters is extracted from the input voice information and analyzed (S140).

[0068] Figure 4 is a flowchart showing the analysis process for diagnosing and classifying Lou Gehrig's disease and cranial nerve palsy.

[0069] Referring to Figure 4, there are three categories: 1) when the resonance parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient, suggesting hypernasality (category item 1); 2) when the articulation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 2); and 3) when the phonation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 3). Here, hypernasality refers to hypernasality, hypernasality, and hypernasality. It refers to a change in voice caused by excessive air being emitted through the nose due to insufficiency of the velopharynx.

[0070] In one embodiment of the present invention, in FIG. 4, category item 1 is given a weight of 60%, category item 2 is given a weight of 20%, and category item 3 is given a weight of 20%. That is, out of 100 points, category item 1 is given a maximum of 60 points, category item 2 is given a maximum of 20 points, and category item 3 is given a maximum of 20 points.

[0071] Here, if two or more items from the three categories apply and the total score is 60 or higher out of 100, it is classified as flaccid dysarthria and diagnosed as Lou Gehrig's disease or cranial nerve palsy (S201-S205). Here, flaccid dysarthria exhibits symptoms such as hypernasality, imprecise consonants, breathy voice, and monotony.

[0072] If diagnosed with Lou Gehrig's disease or cranial nerve palsy, a message recommending neurological examination and brain imaging can be displayed on the computer device.

[0073] Figure 5 is a flowchart showing the analysis process for diagnosing and classifying stroke and vascular dementia.

[0074] Referring to Figure 5, there are four categories: 1) when the articulation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 1), 2) when the phonation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 2), 3) when the prosody parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 3), and 4) when the rate parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 4).

[0075] In one embodiment of the present invention, in FIG. 5, category item 1 is given a weight of 40%, category item 2 is given a weight of 25%, category item 3 is given a weight of 20%, and category item 4 is given a weight of 15%. That is, for a full score of 100, category item 1 is given a maximum of 40 points, category item 2 is given a maximum of 25 points, category item 3 is given a maximum of 20 points, and category item 4 is given a maximum of 15 points.

[0076] Here, if two or more items from the four categories apply and the total score is 60 or higher out of 100, it is classified as spastic dysarthria and diagnosed as stroke or vascular dementia (S207-S211). Here, spastic dysarthria refers to a disorder that occurs when one or more muscles used for speaking are not properly controlled, and is manifested by symptoms such as imprecise consonants, monotony, reduced stress, rough voice, monotony, low pitch, slow rate, hypernasality, suppressed voice, and short phrases.

[0077] If a patient is classified as having a stroke or vascular dementia, a message recommending neurological examination and brain imaging can be displayed on the computer device.

[0078] Figure 6 is a flowchart showing the analysis process for diagnosing and classifying ataxia, brainstem lesion, and cerebellar lesion.

[0079] Referring to Figure 6, there are three categories: 1) when the articulation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 1), 2) when the prosody parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 2), and 3) when the phonation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 3).

[0080] In one embodiment of the present invention, in FIG. 6, category item 1 is given a weight of 50%, category item 2 is given a weight of 30%, and category item 3 is given a weight of 20%. That is, out of 100 points, category item 1 is given a maximum of 50 points, category item 2 is given a maximum of 30 points, and category item 3 is given a maximum of 20 points.

[0081] Here, if two or more items from the three categories apply and the total score is 60 or higher out of 100, it is diagnosed as ataxic dysarthria and classified as ataxia, brainstem lesion, or cerebellar lesion (S213-S217). Here, ataxic dysarthria presents symptoms such as imprecise consonants, excessive or equal stress, irregular articulation errors, vowel distortion, and rough voice.

[0082] If the diagnosis is classified as ataxia, brainstem lesion, or cerebellar lesion, a message recommending neurological examination and brain imaging can be displayed on the computer device.

[0083] Figure 7 is a flowchart showing the analysis process for diagnosing and classifying Parkinson's disease.

[0084] Referring to Figure 7, there are three categories: 1) when the phonation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 1), 2) when the prosody parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 2), and 3) when the rate parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 3).

[0085] In one embodiment of the present invention, in FIG. 7, category item 1 is given a weight of 50%, category item 2 is given a weight of 30%, and category item 3 is given a weight of 20%. That is, out of 100 points, category item 1 is given a maximum of 50 points, category item 2 is given a maximum of 30 points, and category item 3 is given a maximum of 20 points.

[0086] Here, if two or more items from the three categories apply and the total score is 60 or higher out of 100, it is diagnosed as hypokinetic dysarthria and classified as Parkinson's disease (S219-S223). Hypokinetic dysarthria here presents with symptoms such as monotony, reduced stress, monotony, imprecise consonants, inappropriate pauses, and short releases.

[0087] If diagnosed with Parkinson's disease, a message can be displayed on the computer device recommending that the patient consult a neurologist and undergo a brain imaging test.

[0088] Figure 8 is a flowchart showing the analysis process for diagnosing and classifying Huntington's disease (chorea) and dystonia.

[0089] Referring to Figure 8, there are two category items: 1) when the articulation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 1), and 2) when the prosody parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient (category item 2).

[0090] In one embodiment of the present invention, in FIG. 8, a weight of 60% is assigned to category item 1, and a weight of 40% is assigned to category item 2. That is, out of 100 points, category item 1 is worth up to 60 points, and category item 2 is worth up to 40 points.

[0091] Here, if both categories are met and the total score is 60 or higher out of 100, it is classified as hyperkinetic dysarthria and diagnosed as Huntington's disease (chorea) or dystonia (S225-S229). Hyperkinetic dysarthria is characterized by symptoms such as aphonic consonants, vowel distortion, rough voice, monotony, prolonged gaps, excessive intensity changes, and inconsistent rates.

[0092] If classified as Huntington's disease (chorea) or dystonia, a message recommending neurological examination and brain imaging may be displayed on the computer device.

[0093] Figure 9 is a flowchart showing the analysis process for diagnosing and classifying auditory organ dysfunction (hearing impairment) or speech organ damage (head and neck cancer).

[0094] Referring to Figure 9, there is one category item (category item 1) where 1) the articulation parameter is less than -1SD compared to the normal distribution of parameters corresponding to the same age and gender as the patient.

[0095] In Figure 9, if all other parameters are within the normal range and category item 1 is applicable, it is diagnosed and classified as a disorder of the auditory organ (hearing impairment) or damage to the speech organ (head and neck cancer) (S231, S233).

[0096] If the patient is classified as having a hearing impairment (hearing impairment) or speech organ damage (head and neck cancer), a message recommending an otolaryngology examination and hearing test, or recommending an otolaryngology examination and head and neck examination, can be displayed on the computer device.

[0097] Meanwhile, the disease classification method using voice analysis according to an embodiment of the present invention can be implemented as computer-readable code on a computer-readable recording medium. A computer-readable recording medium includes any type of recording device that stores data that can be read by a computer system.

[0098] For example, computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, hard disk, floppy disk, removable storage device, non-volatile memory (flash memory), and optical data storage device.

[0099] Additionally, the computer-readable recording medium can be distributed across computer systems connected to a computer network, and stored and executed as readable code in a distributed manner.

[0100] While the present invention has been described using several preferred embodiments, these embodiments are illustrative and not limiting. Those skilled in the art will appreciate that various changes and modifications can be made without departing from the spirit and scope of the invention as set forth in the appended claims.

Claims

1. When voice information is input from a patient, a step of extracting an acoustic component including a predetermined parameter from the input voice information; A step of performing voice analysis using the above acoustic component; and A step of diagnosing and classifying potential diseases including brain diseases based on the above voice analysis results. A disease classification method using voice analysis including .

2. In claim 1, A disease classification method using voice analysis, characterized in that the parameters of the above acoustic components include phonation, resonance, articulation, prosody, respiration, rate, and regularity.

3. In claim 1, A disease classification method using voice analysis, characterized in that a predetermined task is performed on a patient to input voice information and an acoustic component including predetermined parameters is extracted from the input voice information.

4. In claim 1, A disease classification method using voice analysis, characterized in that, in the step of performing the above voice analysis, the voice analysis is performed in a manner of diagnosing a speech disorder using the acoustic component and diagnosing a disease causing the diagnosed speech disorder.

5. In claim 4, A disease classification method using voice analysis, characterized in that voice analysis is performed by comparing each parameter with a normal distribution of parameters of the same age and gender as the patient.

6. A computer-readable recording medium having recorded thereon a program capable of executing the method of any one of claims 1 to 5 by a computer.

Images