Artificial intelligence-assisted cognitive assessment system and method

Abstract

The present disclosure provides an artificial intelligence (AI)-assisted cognitive assessment system, suitable for being installed in a computer device. The AI-assisted cognitive assessment system comprises: an AI task module, used for providing a multi-step AI vision task; an assessment module, used for obtaining a corresponding cognitive score on the basis of an execution result of a response state or an interaction state of an assessed individual for steps of the task; and a processing module, electrically connected to the AI task module and the assessment module, the processing module being used for controlling the operation of the AI task module and the assessment module, and obtaining a cognitive assessment result on the basis of the cognitive score.

Description

Artificial Intelligence Cognitive Assessment System and Method

[0001] Cross-reference of related applications

[0002] This application claims priority to U.S. patent applications filed June 18, 2025, with priority number 63 / 825,718, and July 17, 2024, with priority number 63 / 672,265, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This disclosure relates to a system and method, and more particularly to a task-oriented artificial intelligence cognitive assessment system and method based on visual AI. Background Technology

[0004] [Amended according to Rule 26, July 28, 2025] As dementia, including Alzheimer's disease (AD), has become an increasingly serious public health challenge in aging societies, including Taiwan, the prevalence of dementia is rising, placing a heavy burden on families and healthcare systems. Neuropsychiatric testing plays an important role in identifying and tracking patients who progress from mild cognitive impairment (MCI) to dementia. Summary of the Invention

[0005] Under the current healthcare system, traditional cognitive assessment tools, such as the MMSE, can typically only be administered to individuals with cognitive impairments once every few months, making it impossible to dynamically track cognitive changes. Furthermore, the limited number of qualified healthcare professionals makes it difficult to provide frequent neuropsychological testing, further restricting the accessibility and frequency of cognitive performance monitoring.

[0006] Therefore, the applicant devoted himself to research and development, and developed an artificial intelligence cognitive assessment system and method that can provide high-frequency and objective cognitive assessments to improve early detection capabilities and shorten the assessment time.

[0007] This disclosure provides an artificial intelligence cognitive assessment system suitable for installation on a computer device. The artificial intelligence cognitive assessment system includes: an AI task module for providing multi-step AI vision tasks; an assessment module for obtaining corresponding cognitive scores based on the performance results of the assessed person's response or interaction to the steps of the task; and a processing module electrically connected to the AI ​​task module and the assessment module, the processing module controlling the operation of the AI ​​task module and the assessment module, and obtaining cognitive assessment results based on the cognitive scores.

[0008] In one embodiment, the assessment module calculates a cognitive score based on the accuracy of the subject's execution of the steps and whether the running time is within a specified time.

[0009] In one embodiment, the processing module categorizes cognitive scores into scores for several cognitive functions, including long memory, short memory, judgment, practical application, and computational ability.

[0010] In one embodiment, the processing module divides the cognitive assessment results into several levels according to several pre-set score intervals, and each level corresponds to a different level of cognitive ability.

[0011] In one embodiment, the processing module converts the cognitive assessment results into scores for at least one of the following: a short intelligence test, an Alzheimer's disease scale, and a Cohen-Mansfield emotional behavior scale, so that the AI ​​task module can use it as a basis for providing steps in the AI ​​vision task.

[0012] In one embodiment, the processing module automatically selects a suitable task and its step configuration from a large number of tasks provided by the AI ​​task module based on the past task execution history of the person being evaluated.

[0013] In one embodiment, the evaluation module analyzes and identifies the evaluated person's operation of a multi-object teaching aid set through images acquired by a camera device connected to a computer device, and determines whether the evaluated person has correctly performed the steps based on the object identification results and movement changes.

[0014] In one embodiment, the evaluation module uses a machine learning model to analyze whether the steps are executed correctly and records the running time of completing the steps.

[0015] In one embodiment, the machine learning model has the functions of both an image segmentation model and an object detection model.

[0016] In one embodiment, the artificial intelligence cognitive assessment system of this disclosure further includes a difficulty adjustment module, adapted to be installed in a computer device, the difficulty adjustment module being electrically connected to the processing module and used to adjust the difficulty of the task based on the accuracy of the assessed person's execution of the steps.

[0017] In one embodiment, the artificial intelligence cognitive assessment system of this disclosure further includes a report generation module, adapted to be installed in a computer device, the report generation module being electrically connected to the processing module and used to automatically generate an assessment report based on the cognitive assessment results.

[0018] In one embodiment, the artificial intelligence cognitive assessment system of this disclosure further includes a database module and an electrical connection processing module for storing assessment reports.

[0019] In one embodiment, the database module is a multimodal database and supports retrieval enhancement generation technology as well as retrieval of external knowledge bases.

[0020] In one embodiment, the artificial intelligence cognitive assessment system of this disclosure further includes a generation module and an electrical connection processing module. The generation module is used to generate a task scenario and present steps in the task scenario.

[0021] In one embodiment, the generation module generates a task scenario using at least one of a photograph, life events, and a personal profile related to the person being evaluated.

[0022] In one embodiment, the task scene displays prompts for the steps via images and / or voice.

[0023] In one embodiment, the generation module uses a large language model and a text-to-image model to generate the task scene.

[0024] This disclosure provides an artificial intelligence cognitive assessment method, comprising the following steps: providing a multi-step AI vision task using an AI task module; obtaining a corresponding cognitive score using an assessment module based on the execution results of the assessed person's response state or interaction state to the steps of the task; and obtaining a cognitive assessment result using a processing module based on the cognitive score.

[0025] In one embodiment, the assessment module calculates a cognitive score based on the accuracy of the subject's execution of the steps and whether the running time is within a specified time.

[0026] In one embodiment, the processing module categorizes cognitive scores into scores for several cognitive functions, including long memory, short memory, judgment, practical application, and computational ability.

[0027] In one embodiment, the processing module divides the cognitive assessment results into several levels according to several pre-set score intervals, and each level corresponds to a different level of cognitive ability.

[0028] In one embodiment, the processing module converts the cognitive assessment results into scores for at least one of the following: a short intelligence test, an Alzheimer's disease scale, and a Cohen-Mansfield emotional behavior scale, so that the AI ​​task module can use it as a basis for providing steps in the AI ​​vision task.

[0029] In one embodiment, the processing module automatically selects a suitable task and its step configuration from a large number of tasks provided by the AI ​​task module based on the past task execution history of the person being evaluated.

[0030] In one embodiment, the evaluation module analyzes and identifies the operator's actions on a multi-object teaching aid set through images acquired by a photographic device, and judges whether the operator has correctly executed the steps based on the object identification results and movement changes.

[0031] In one embodiment, the evaluation module uses a machine learning model to analyze whether the steps are executed correctly and records the running time of completing the steps.

[0032] In one embodiment, the machine learning model has the functions of both an image segmentation model and an object detection model.

[0033] In one embodiment, the artificial intelligence cognitive assessment method disclosed herein further includes a difficulty adjustment module, which adjusts the difficulty of the task based on the accuracy of the assessed person in performing the steps.

[0034] In one embodiment, the artificial intelligence cognitive assessment method disclosed herein further includes a report generation module, which automatically generates an assessment report based on the cognitive assessment results.

[0035] In one embodiment, the artificial intelligence cognitive assessment method disclosed herein further includes storing assessment reports in a database module.

[0036] In one embodiment, the database module is a multimodal database and supports retrieval enhancement generation technology as well as retrieval of external knowledge bases.

[0037] In one embodiment, the task scenario is generated using at least one of a photograph, a life event, and a personal profile related to the person being evaluated.

[0038] In one embodiment, the AI ​​task module displays step prompts via images and / or voice.

[0039] In one embodiment, a large language model and a text-to-image model are used to generate the task scene.

[0040] In summary, the AI-based cognitive assessment system and method disclosed herein can provide multi-step AI visual tasks for the assessed person to perform through an AI task module. The assessment module obtains corresponding cognitive scores based on the assessed person's performance of each step. Finally, the processing module obtains the cognitive assessment results based on the cognitive scores, thereby assessing whether the assessed person has cognitive impairment. Since the tasks are generated in the form of visual AI, it can provide high-frequency and objective cognitive assessments, thereby improving early detection capabilities and shortening the assessment time.

[0041] Overview of the attached figures

[0042] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 is a block diagram of the first embodiment of the artificial intelligence cognitive assessment system disclosed herein.

[0044] Figures 2A to 2C are usage diagrams of the first embodiment of the artificial intelligence cognitive assessment system disclosed herein.

[0045] Figure 3 is a block diagram of a second embodiment of the artificial intelligence cognitive assessment system disclosed herein.

[0046] Figure 4 is a block diagram of a third embodiment of the artificial intelligence cognitive assessment system disclosed herein.

[0047] Figure 5 is a block diagram of the fourth embodiment of the artificial intelligence cognitive assessment system disclosed herein.

[0048] Figure 6 is a flowchart of the steps of the first embodiment of the artificial intelligence cognitive assessment method disclosed herein.

[0049] Figure 7 is a flowchart of the steps of the second embodiment of the artificial intelligence cognitive assessment method disclosed herein.

[0050] Figure 8 is a flowchart of the steps of the third embodiment of the artificial intelligence cognitive assessment method disclosed herein.

[0051] Figure 100: Artificial Intelligence Cognitive Assessment System; Figure 200: Artificial Intelligence Cognitive Assessment Method; 1. Computer Device; 11. AI Task Module; 12. Assessment Module; 13. Processing Module; 14. Difficulty Adjustment Module; 15. Report Generation Module; 16. Database Module; 17. Generation Module; 2. Photography Device; 3. Teaching Aids Group; 31. Fruits and Vegetables; 32. Storage Basket; 33. Cashier Counter; 34. Coins; 4. Display Device; S201, S202, S203, S204, S205: Steps

[0052] Preferred embodiments of this disclosure

[0053] To fully understand the purpose, features, and effects of this disclosure, the following specific embodiments, in conjunction with the accompanying drawings, will provide a detailed description of this disclosure, as follows:

[0054] Please refer to Figure 1, which is a block diagram of a first embodiment of the artificial intelligence cognitive assessment system 100 disclosed herein, suitable for installation on a computer device 1, such as a desktop computer or a tablet computer.

[0055] In this embodiment, the artificial intelligence cognitive assessment system 100 includes: an AI task module 11, an assessment module 12, and a processing module 13. The AI ​​task module 11 provides multi-step AI vision tasks; the assessment module 12 obtains corresponding cognitive scores based on the response or interaction status of the assessed individual at each step of the task; the processing module 13 is electrically connected to the AI ​​task module 11 and the assessment module 12, and controls the operation of the AI ​​task module 11 and the assessment module 12, obtaining cognitive assessment results based on the cognitive scores. Therefore, the artificial intelligence cognitive assessment system 100 of this disclosure can provide high-frequency and objective cognitive assessments, thereby improving early detection capabilities and shortening the assessment time.

[0056] In this embodiment, the AI ​​task module 11 can provide relevant tasks based on the task scenario, which can display step prompts through images and / or voice. For example, when the task scenario involves the person being evaluated making purchases in a fruit and vegetable store, the task can instruct the person to select specific fruits and vegetables, calculate the total amount of the selected fruits and vegetables, or remember which fruits and vegetables are sold in the store, but is not limited to these. In other embodiments, the task can also instruct the person being evaluated to complete the task by operating a mouse, keyboard, and / or touch screen, which are electrically connected to the computer device 1.

[0057] In this embodiment, the evaluation module 12 can calculate the cognitive score based on the accuracy of the evaluated person's execution of the steps and whether the running time is within the specified time. That is, if the evaluated person executes the steps correctly and the running time is within the specified time, the corresponding score can be obtained. In other embodiments, the evaluation module 12 can also calculate the cognitive score based on whether the evaluated person's reaction time to the steps is within the specified time. Specifically, reaction time refers to the time elapsed from when the steps are displayed in the task scenario to when the evaluated person begins to act, and may also include the time it takes for the evaluated person to complete the steps.

[0058] Please refer to Figure 2 as well. The assessment module 12 analyzes and identifies the assessed person's operation of the multi-object teaching aid set 3 through images acquired by the photography device 2 connected to the computer device 1. Based on the object identification results and movement changes, it judges whether the assessed person has correctly executed the steps. For example, the photography device 2 can be a camera, and the teaching aid set 3 can be a fruit and vegetable teaching aid set, a cooked food teaching aid set, or a butcher shop teaching aid set.

[0059] Specifically, the evaluation module 12 uses a machine learning model to analyze whether the steps are executed correctly and can also record the running time of completing the steps. In this embodiment, the machine learning model can be constructed using the YOLO framework (e.g., YOLO v5), and the machine learning model can simultaneously have the functions of an image segmentation model and an object detection model. For example, the machine learning model can use the Grounded SAM 2 model, which combines the Grounding DINO object detection model and the SAM 2 image segmentation model to detect and extract objects from images based on text. This is common knowledge in the relevant field of this disclosure and will not be elaborated further here.

[0060] In this embodiment, the processing module 13 can summarize the cognitive scores into scores for several cognitive functions, including long memory, short memory, judgment, praxis, and calculation. That is, the steps can be used to assess whether the user has cognitive impairments in long memory, short memory, judgment, praxis, and calculation.

[0061] In this embodiment, the processing module 13 can divide the cognitive assessment results into several levels according to several pre-set score intervals, each corresponding to a different level of cognitive level. Preferably, the processing module 13 can also convert the cognitive assessment results into scores of at least one of the following: a Mini-Mental State Examination (MMSE), a Geriatric Depression Scale (GDS), and a Cohen-Mansfield Agitation Inventory (CMAI), so that the AI ​​task module 11 can use it to provide the basis for the AI ​​vision task in the steps.

[0062] For example, the full score of the cognitive assessment result is 100. Several score intervals can be divided into the first interval, the second interval, and the third interval. The first interval can be 0-70, the second interval can be 71-90, and the third interval can be 91-100. The first interval corresponds to severe cognitive impairment (MMSE score of 0-21, including the endpoint), the second interval corresponds to mild cognitive impairment (MMSE score of 22-27, including the endpoint), and the third interval corresponds to cognitive integrity (MMSE score of 28-30, including the endpoint).

[0063] Preferably, the processing module 13 can also automatically select the most suitable task and its step configuration from among the many tasks provided by the AI ​​task module 11 based on the past task execution history of the person being evaluated.

[0064] Please refer to Figures 2A to 2C, which are usage diagrams of the first embodiment of the artificial intelligence cognitive assessment system 100 disclosed herein. The vegetable and fruit teaching aid set 3 (Figure 2A) is used as an example for illustration, and the tasks provided by the AI ​​task module 11 are displayed on the display device 4 (Figure 2B) electrically connected to the computer device 1.

[0065] In Figure 2A, the fruit and vegetable teaching aid set includes multiple fruit and vegetable objects 31, a storage basket 32, a cash register 33, and coins 34 of different denominations. The fruit and vegetable objects 31 have a price tag, and the coins 34 can have denominations of five, ten, and fifty yuan. The person being assessed performs corresponding operations on the fruit and vegetable objects 31 according to the steps displayed on the display device 4, such as picking up a particular fruit and vegetable object 31 or moving any fruit and vegetable object 31 to the storage basket 32; or selecting coins 34 of the corresponding denomination and moving them to the cash register 33. In Figure 2B, the process of the person being assessed performing the steps is recorded by the camera device 2. In Figure 2C, the assessment module 12 analyzes and judges whether the person being assessed has correctly performed the steps based on the recognition results and movement changes of the fruit and vegetable objects 31, the storage basket 32, the cash register 33, and the coins 34, to obtain a corresponding cognitive score. Subsequently, the processing module 13 obtains the cognitive assessment result of the person being assessed based on the cognitive score to determine whether the person being assessed has a cognitive impairment.

[0066] Please refer to Figure 3, which is a block diagram of a second embodiment of the artificial intelligence cognitive assessment system 100 of this disclosure. In this embodiment, compared with the first embodiment, it may further include a difficulty adjustment module 14, which is adapted to be installed on the computer device 1 and electrically connected to the processing module 13. The difficulty adjustment module 14 is used to adjust the difficulty of the task according to the accuracy of the assessed person in performing the steps.

[0067] For example, the steps of a task can include several types of cognitive functions such as long memory, short memory, judgment, application, and calculation. Corresponding cognitive scores can be awarded based on the correctness of the answer, and the tasks can have four levels of difficulty (e.g., level one is the easiest, and level four is the most difficult). Specifically, the difficulty of long memory tasks, from easiest to hardest, can be: "remembering two objects between one question," "remembering three objects between one question," "remembering two objects between two questions," and "remembering three objects between two questions." The difficulty of short memory tasks, from easiest to hardest, can be: "remembering two objects," "remembering three objects," "remembering four objects," and "remembering five objects." The difficulty of judgment tasks, from easiest to hardest, can be: "determining the correct answer from two options," "determining the correct answer from three options," "determining the correct answer from four options," and "adding a condition to the correct answer." The difficulty levels of the tasks, from easiest to hardest, are: "Move one object", "Move two objects simultaneously", "Move two objects to different areas", and "Add a conditional clause and only move objects that meet the condition". The difficulty levels of the calculation tasks, from easiest to hardest, are: "Add two numbers, the last two digits are 0 or 5, and the total does not exceed 100", "Add three numbers, the last two digits are 0 or 5, and the total does not exceed 1000", "Select at most two from four options, and the total does not exceed 100", and "Select at most three from four options, and the total sum after selection according to the conditional clause does not exceed 100".

[0068] For example, in the steps of a task, when the execution accuracy of steps of the same type reaches 60% or more, the difficulty adjustment module 14 increases the difficulty of the corresponding steps in the task by one level. For example, if long memory was originally at level one difficulty, it is adjusted to level two difficulty. On the other hand, when the execution accuracy of steps of the same type reaches 40% or less, the difficulty adjustment module 14 decreases the difficulty of the corresponding steps in the task by one level. For example, if long memory was originally at level two difficulty, it is adjusted to level one difficulty. The processing module 13 can set the score for any step to 10 points. When the score of a step exceeds 8 points, the processing module 13 judges the step as a correct answer, and when the score of a step is 0 points, the processing module 13 judges the step as an incorrect answer. Furthermore, when the original difficulty is level one, but the evaluator's execution accuracy for a certain type of step is below 40%, the difficulty adjustment module 14 removes steps of that type from the task and adds steps of other types to the task.

[0069] Please refer to Figure 4, which is a block diagram of a third embodiment of the artificial intelligence cognitive assessment system 100 disclosed herein. In this embodiment, compared with the first and second embodiments, it may further include a report generation module 15, which is adapted to be installed on the computer device 1 and electrically connected to the processing module 13. The report generation module 15 is used to automatically generate an assessment report based on the cognitive assessment results. Preferably, the assessment report may also be presented in a graphical format.

[0070] In this embodiment, a database module 16 and an electrical connection processing module 13 may also be included for storing evaluation reports. Preferably, the database module 16 is a multimodal database and supports retrieval-augmented generation (RAG) technology as well as retrieval of external knowledge bases (e.g., dementia treatment guidelines).

[0071] Please refer to Figure 5, which is a block diagram of the fourth embodiment of the artificial intelligence cognitive assessment system 100 disclosed herein. In this embodiment, compared with the first, second, and third embodiments, it may further include a generation module 17 and an electrical connection processing module 13. The generation module 17 is used to generate a task scene and present steps in the task scene. Preferably, the generation module 17 generates the task scene using at least one of the following: photos, life events, and personal profiles related to the person being assessed. For example, medical record data, core memory interview data, cognitive assessment information, caregiver-related information, social media information, institutional physiological monitoring data, past photos and videos of relatives and friends, photos of past life scenes, daily video data, and institutional / monitoring videos. Preferably, it may also include information generated during the steps of the person being assessed performing the task (e.g., past error types, reaction patterns, preferred topics). The polymorphic data such as photos, life events, personal profiles, cognitive assessment results, and task scenes related to the person being assessed can be stored in the database module 16. One implementation of this embodiment involves using life objects or past life experiences related to the caregiver (e.g., the person being assessed) to create an environment that evokes memories or emotions in the caregiver, allowing them to feel as if they are returning to a familiar or memorable environment, thereby enhancing the caregiver's interaction or connection / adhesion with the system. In other words, this embodiment can be used to recreate or reshape the caregiver's living environment or life experiences to facilitate interaction between the caregiver and the system.

[0072] Specifically, the generation module 17 uses a large-scale language model and a text-to-image model to generate the task scene. Preferably, it can also be used to generate an image of the avatar of the person being evaluated. For example, the generation module 17 can use an LLaMA large-scale language model and a FLUX text-to-image model to generate the task scene.

[0073] Please refer to Figure 6, which is a flowchart of the steps in the first embodiment of the artificial intelligence cognitive assessment method 200 of this disclosure, including the following steps:

[0074] Step S201: The AI ​​task module 11 provides a multi-step AI vision task. In this embodiment, the AI ​​task module 11 can display prompts for the steps through images and / or voice.

[0075] In one embodiment, the task scenario can be generated through the aforementioned generation module 17, using at least one of the following related to the person being assessed: photographs, life events, and personal profiles. Examples of sources include: medical record data, core memory interview data, cognitive assessment information, caregiver-related information, social media information, institutional physiological monitoring data, past photos and videos of relatives and friends, photos of past life scenes, daily video data, and institutional / monitoring images. Preferably, the input data for the task scenario may also include information generated during the steps the person being assessed takes to perform the task (e.g., past error types, response patterns, preferred topics).

[0076] Specifically, the generation module 17 uses a large-scale language model and a text-to-image model to generate the task scene. Preferably, it can also be used to generate an image of the avatar of the person being evaluated. For example, the generation module 17 can use an LLaMA large-scale language model and a FLUX text-to-image model to generate the task scene.

[0077] In one embodiment, the AI ​​task module 11 can automatically select a suitable task and its step configuration from among a number of tasks provided by the AI ​​task module 11, based on the past task execution history of the person being evaluated.

[0078] Step S202: The aforementioned assessment module 12 obtains the corresponding cognitive score based on the performance results of the assessed person's response or interaction with the steps of the task.

[0079] Specifically, the evaluation module 12 can calculate the cognitive score based on the accuracy of the evaluated person's execution of the steps and whether the running time is within the specified time. In other embodiments, the evaluation module 12 can also calculate the cognitive score based on whether the evaluated person's reaction time to the steps is within the specified time. In detail, reaction time refers to the time elapsed from when the steps are displayed in the task scenario to when the evaluated person begins to act, and may also include the time it takes for the evaluated person to complete the steps.

[0080] The evaluation module 12 can analyze and identify the evaluated person's operation on the multi-object teaching aid set through the images acquired by the aforementioned photographic device 2, and determine whether the evaluated person has correctly executed the steps based on the object recognition results and movement changes. In this embodiment, the evaluation module 12 can use a machine learning model to analyze whether the steps are correctly executed, and record the running time of completing the steps. The machine learning model has the functions of both an image segmentation model and an object detection model. For example, the machine learning model can use the Grounded SAM 2 model, which combines the Grounding DINO object detection model and the SAM 2 image segmentation model to detect and extract objects from images based on text.

[0081] In this embodiment, cognitive scores can be categorized into scores for several cognitive functions, including long memory, short memory, judgment, praxis, and calculation.

[0082] Step S203: Obtain the cognitive assessment result based on the cognitive score using the aforementioned processing module 13.

[0083] In one embodiment, the cognitive assessment results can be divided into several levels according to several pre-set score intervals, each corresponding to a different level of cognitive ability. Preferably, the cognitive assessment results can also be converted into scores of at least one of the following: a short intelligence test, an Alzheimer's disease scale, and the Cohen-Mansfield Agitated Behavior Scale, so that the AI ​​task module 11 can provide a basis for the AI ​​vision task of the steps.

[0084] Please refer to Figure 7, which is a flowchart of the steps in the second embodiment of the artificial intelligence cognitive assessment method 200 of this disclosure, including the following steps:

[0085] Step S204: Adjust the difficulty of the task based on the accuracy of the evaluated person's execution of the steps using the aforementioned difficulty adjustment module 14.

[0086] Please refer to Figure 8, which is a flowchart of the steps in the third embodiment of the artificial intelligence cognitive assessment method 200 of this disclosure, including the following steps:

[0087] Step S205: The aforementioned report generation module 15 automatically generates an assessment report based on the cognitive assessment results. Preferably, the assessment report can also be presented in the form of charts.

[0088] In one embodiment, the evaluation report can also be stored in the aforementioned database module 16. The database module 16 is a multimodal database and supports retrieval enhancement generation techniques and retrieval of external knowledge bases.

[0089] In summary, the AI-based cognitive assessment system and method disclosed herein can provide multi-step AI visual tasks for the assessed person to perform through an AI task module. The assessment module obtains corresponding cognitive scores based on the assessed person's performance of each step. Finally, the processing module obtains a cognitive assessment result based on the cognitive scores, thereby assessing whether the assessed person has cognitive impairment. Since the tasks are generated in a visual AI manner, it can provide high-frequency and objective cognitive assessments, thereby improving early detection capabilities and shortening the assessment time.

[0090] This disclosure has been described above with reference to preferred embodiments. However, those skilled in the art should understand that the embodiments are for illustrative purposes only and should not be construed as limiting the scope of this disclosure. It should be noted that all variations and substitutions equivalent to the embodiments should be included within the scope of this disclosure. Therefore, the scope of protection of this disclosure is defined by the claims.

Claims

1. An artificial intelligence cognitive assessment system adapted to be installed in a computer device, characterized by, The artificial intelligence cognitive assessment system comprises: an AI task module configured to provide a multi-step AI visual task; an assessment module configured to obtain a corresponding cognitive score according to an execution result of a response state or an interaction state of an evaluated person to the steps of the task; and a processing module electrically connected to the AI task module and the assessment module, the processing module being configured to control the operation of the AI task module and the assessment module, and obtain a cognitive assessment result according to the cognitive score. 2.The artificial intelligence cognitive assessment system of claim 1, wherein, The assessment module calculates the cognitive score according to whether the execution accuracy and the running time of the evaluated person to the steps are within a specified time. 3.The artificial intelligence cognitive assessment system of claim 1, wherein, The processing module induces the cognitive score into scores of several cognitive functions including long-term memory, short-term memory, judgment, dexterity and calculation. 4.The artificial intelligence cognitive assessment system of claim 1, wherein, The processing module divides the cognitive assessment result into several levels according to a plurality of score intervals set in advance, and respectively corresponds to different levels of cognitive degree. 5.The artificial intelligence cognitive assessment system according to claim 1, wherein, The processing module converts the cognitive assessment result into a score of at least one of a short intelligence test, a geriatric depression scale and a Cohen-Mansfield agitation inventory, so that the AI task module is used to provide the AI visual task of the steps. 6.The artificial intelligence cognitive assessment system of claim 1, wherein, The processing module automatically selects a task and its step configuration suitable for the evaluated person from a plurality of tasks provided by the AI task module according to the execution history of the evaluated person. 7.The artificial intelligence cognitive assessment system of claim 1, wherein, The assessment module analyzes and identifies the operation of the evaluated person to a plurality of object sets of teaching aids through images obtained by a camera connected to the computer device, to determine whether the evaluated person correctly executes the steps according to the identification result and movement change of the objects. 8.The artificial intelligence cognitive assessment system of claim 7, wherein, The assessment module uses a machine learning model to analyze whether the steps are correctly executed, and records the running time of completing the steps. 9.The artificial intelligence cognitive assessment system of claim 8, wherein, The machine learning model has the functions of image segmentation model and target detection model. 10.The artificial intelligence cognitive assessment system according to any one of claims 1 to 9, characterized in that, Further comprising a difficulty adjustment module adapted to be installed in the computer device, the difficulty adjustment module being electrically connected to the processing module and being configured to adjust the difficulty of the task according to the execution accuracy of the evaluated person to the steps. 11.The artificial intelligence cognitive assessment system according to any one of claims 1 to 9, characterized in that, Further comprising a report generation module adapted to be installed in the computer device, the report generation module being electrically connected to the processing module and being configured to automatically generate an assessment report according to the cognitive assessment result.

12. The artificial intelligence cognitive assessment system of claim 11, wherein, Further comprising a database module electrically connected to the processing module and being configured to store the assessment report.

13. The artificial intelligence cognitive assessment system of claim 12, wherein, The database module is a multi-modal database, and supports retrieval enhancement generation technology and retrieval external knowledge base. 14.The artificial intelligence cognitive assessment system according to any one of claims 1 to 9, wherein, Further comprising a generation module electrically connected to the processing module, the generation module being configured to generate a task scene and present the steps in the task scene.

15. The artificial intelligence cognitive assessment system of claim 14, wherein, The generation module generates the task scene with at least one of a photo related to the evaluated person, a life event and a personal profile.

16. The artificial intelligence cognitive assessment system of claim 15, wherein, The task scene displays the prompts of the steps through images or / and voice.

17. The artificial intelligence cognitive assessment system of claim 14, wherein, The generation module employs a large language model and a text-to-image model to generate the task scenario.

18. An artificial intelligence cognitive assessment method, comprising: The method comprises the following steps: The AI task module provides a multi-step AI visual task. The evaluation module calculates a cognitive score according to the response state or interaction state of the evaluated person to the steps of the task. The processing module obtains a cognitive evaluation result according to the cognitive score. The evaluation module calculates the cognitive score according to the execution accuracy and the running time of the evaluated person to the steps.

19. The artificial intelligence cognitive assessment method of claim 18, wherein, The processing module induces the cognitive score into scores of several cognitive functions including long-term memory, short-term memory, judgment, dexterity, and calculation ability.

20. The artificial intelligence cognitive assessment method of claim 18, wherein, The processing module divides the cognitive evaluation result into several levels according to pre-set score intervals, and respectively corresponds to different levels of cognitive degree.

21. The artificial intelligence cognitive assessment method of claim 18, wherein, The processing module converts the cognitive evaluation result into scores of at least one of a short intelligence test, a geriatric depression scale, and a Cohen-Mansfield agitation inventory behavior scale, so as to make the AI task module used for providing the AI visual task of the steps.

22. The artificial intelligence cognitive assessment method of claim 18, wherein, The processing module automatically selects a task and its step configuration suitable for the evaluated person from a plurality of tasks provided by the AI task module according to the execution history of the evaluated person.

23. The artificial intelligence cognitive assessment method of claim 18, wherein, The evaluation module analyzes and identifies the operation of the evaluated person to the multi-object teaching aid set through the image obtained by the camera, so as to judge whether the evaluated person correctly executes the steps according to the identification result and the movement change of the objects.

24. The artificial intelligence cognitive assessment method of claim 18, wherein, The evaluation module employs a machine learning model to analyze whether the steps are correctly executed, and records the running time of completing the steps.

25. The artificial intelligence cognitive assessment method of claim 24, wherein, The machine learning model has the functions of image segmentation model and target detection model.

26. The artificial intelligence cognitive assessment method of claim 25, wherein, The difficulty adjustment module adjusts the difficulty of the task according to the execution accuracy of the evaluated person to the steps.

27. The artificial intelligence cognitive assessment method of any of claims 18-26, wherein, The report generation module automatically generates an evaluation report according to the cognitive evaluation result.

28. The artificial intelligence cognitive assessment method of any one of claims 18-26, wherein, The database module stores the evaluation report.

29. The artificial intelligence cognitive assessment method of claim 28, wherein, The database module is a multi-modal database, and supports retrieval enhancement generation technology and retrieval external knowledge base.

30. The artificial intelligence cognitive assessment method of claim 29, wherein, The task scenario is generated by at least one of a photo, a life event, and a personal profile related to the evaluated person.

31. The artificial intelligence cognitive assessment method of claim 18, wherein, The AI task module displays the prompts of the steps through images or / and voices.

32. The artificial intelligence cognitive assessment method of claim 18, wherein, The generation module employs a large language model and a text-to-image model to generate the task scenario.

33. The artificial intelligence cognitive assessment method of claim 18, wherein, ​

Images