System and method for detecting vascular mild cognitive impairment based on digital biomarkers

Abstract

This disclosure relates to a computing module that is configured to detect vascular mild cognitive impairment (MCI) by generating an MCI assessment score. The score is determined by a trained machine learning model, which is configured to process inputs generated by the user as the user completes a plurality of games that are accessible through an interface module communicatively connected to the computing module. The machine learning model is trained based on correlations between clinical data of subjects and gameplay data acquired from the subjects when the subjects complete a series of predesigned games. In various embodiments, the plurality of games are developed to obtain behavioural and cognitive markers including executive function, processing speed, apathy and disinhibition.

Description

SYSTEM AND METHOD FOR DETECTING VASCULAR MILD COGNITIVE IMPAIRMENT BASED ON DIGITAL BIOMARKERSCROSS REFERENCE TO RELATED APPLICATION

[0001] This application claims the benefit of priority to Singapore patent application no. 10202303300R which was filed on 22 November 2023, the contents of which are hereby incorporated by reference in its entirety for all purposes.TECHNICAL FIELD

[0002] This application relates to a computing module that is configured to assess a user’s vascular mild cognitive impairment (MCI). The score is determined by a trained machine learning model, which is configured to process inputs generated by the user as the user completes a plurality of games that are accessible through an interface module communicatively connected to the computing module.BACKGROUND

[0003] Mild Cognitive Impairment (MCI) may be classified as a prodromal stage of dementia, characterized in that it has a heightened risk of progressing to overt dementia. When MCI is related to silent strokes, it is then classified as vascular MCI, which is often associated with Vascular Risk Factors (VRFs) such as hypertension, diabetes, and high cholesterol. It is noted that individuals with one or more of these risk factors are at an increased likelihood of developing vascular MCI. In order to diagnose vascular MCI, a patient would have to typically undergo a comprehensive neuropsychological evaluation lasting 2.5 to 3 hours followed by an MRI scan to detect silent strokes. These diagnostic tests are quite costly and may be prohibitive to most patients. In addition to these high costs, the patient would also typically have to wait around three to six months before the patient is able to be evaluated as mentioned above.

[0004] Despite the pressing need for early detection, there are currently no mobile applications available to assist clinicians in diagnosing vascular MCI. The costly and time- consuming nature of traditional diagnostic methods creates a significant bottleneck in identifying patients with VRFs before they develop advanced cognitive impairment. Consequently, most patients at risk are not screened for vascular MCI until it is too late, resulting in more severe cognitive impairment and a higher burden on healthcare resources.This delay in diagnosis exacerbates the cost of care and makes treatment less effective as cognitive decline progresses.

[0005] The current gold standard for diagnosing vascular MCI includes both neuropsychological evaluations and the detection of cerebrovascular disease (CVD) on MRI scans, making the process time-intensive and expensive. Neuropsychological assessments, typically conducted using pen-and-paper methods, are critical for detecting cognitive and behavioral changes associated with brain function and daily capabilities. While existing research has shown that cognitive domains such as Executive Function (EF) and Processing Speed (PS) are commonly impaired in patients with vascular MCI, the role of behavioral domains has not been fully explored or quantified as diagnostic markers for this condition. This gap underscores the need for scalable, affordable, and accessible tools to aid in early diagnosisSUMMARY

[0006] In one aspect, the present application discloses a module for detecting vascular Mild Cognitive Impairment (MCI). The module includes a processing unit and a non-transitory media readable by the processing unit, the media storing instructions that when executed by the processing unit causes the processing unit to receive and analyze a plurality of action- related tensors from an input / output ( I / O) interface module communicatively coupled to the module, the I / O interface module being configured to record real-time inputs from a user interacting with the I / O interface module, wherein the plurality of action-related tensors comprises a first and a second action-related tensor that arc each generated in response to cognitive stimuli designed to assess executive function and processing speed of the user, a third and a fourth action-related tensor that are each generated in response to behavioral stimuli designed to assess apathy and disinhibition of the user. The module then proceeds to generate a multi-domain feature set based on the plurality of action-related tensors and to compute a vascular MCI assessment score by providing the multi-domain feature set to a trained machine learning model, whereby the machine learning model was trained based on correlations between a plurality of training sets of action-related tensors and corresponding clinical data comprising neuropsychological and neuroimaging data.

[0007] In embodiments of this one aspect, the first action-related tensor comprises the inputs recorded by the I / O interface module in response to a user of the I / O interface modulebeing displayed a set of identifiers on an interface of the I / O interface module and being prompted to match the displayed identifiers in a predetermined order, wherein the recorded inputs comprise a number of correctly matched identifiers, a duration of time taken to complete each match attempt, a total time taken to complete an entire matching of the set of identifiers, touch locations on the interface for each attempt, reaction times between presentation of an identifier and the user’s response, or patterns of correct matched identifiers.

[0008] In embodiments of this one aspect, the second action-related tensor comprises the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of symbols on an interface of the I / O interface module and being prompted to connect the symbols in a predetermined order, wherein the recorded inputs comprise a number of correctly connected symbols, a duration of time taken to complete each connection, a total time taken to complete an entire connection of the set of symbols, a number of taps recorded on the interface, time intervals between consecutive taps on the interface, or patterns of correct connections.

[0009] In embodiments of this one aspect, the third action-related tensor comprises the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of visual stimuli comprising a target object and a plurality of obstacles on an interface of the I / O interface module and being prompted to perform an input action on the interface of the I / O interface module based on a direction of the target object, wherein the recorded inputs comprise a number of correct input actions in relation to variations of the target object, a duration of time taken to complete each input action, a total time taken to complete a level, a difficulty-value obtained based on a number of the obstacles and a number of variations of the target object, or patterns of correct input actions.

[0010] In embodiments of this one aspect, the fourth action-related tensor comprises the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a visual list of items and being prompted to memorize objects contained in the list of items, wherein the recorded inputs comprise a number of correctly selected items, a duration of time taken to complete an item selection task, a number of times the user accesses the visual list of items, a duration of time taken by the user to select an item, or patterns of correct selections.

[0011] In another aspect, the present application discloses a method for detecting vascular Mild Cognitive Impairment (MCI) using a computing module. The disclosed method comprises the steps of receiving and analysing a plurality of action-related tensors from an input / output (I / O) interface module communicatively coupled to the computing module, the I / O interface module being configured to record inputs provided by a user, wherein the plurality of action-related tensors comprises a first and a second action-related tensors that are each generated in response to cognitive stimuli, and wherein the plurality of action-related tensors comprises a third and a fourth action-related tensors that are each generated in response to behavioral stimuli. The method then generates a set of features derived from the plurality of action -related tensors and computes a vascular MCI assessment score by providing the generated set of features to a trained machine learning model, whereby the machine learning model was trained based on correlations between a plurality of training sets of action-related tensors and corresponding clinical data, the clinical data comprising neuropsychological and ncuroimaging data.BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Various embodiments of the present disclosure are described below with reference to the following drawings:Figure 1 illustrates a block diagram of a computing device configured to perform the steps for detecting vascular' mild cognitive impairment (MCI) based on digital biomarkers in accordance with embodiments of the present disclosure;Figure 2 illustrates a block diagram of a processing system for performing embodiments of the present disclosure;Figure 3 illustrates plots that show that individuals with high white matter hyperintensities (WMH) have greater cognitive impairments as compared to those with low WMH;Figure 4 illustrates plots that show that individuals with high white matter hyperintensities (WMH) have greater mild behavioural impairment-checklist (MBI-C) domain scores as compared to those with low WMH;Figure 5 illustrates plots that show the correlation between the scores of individuals who played the games and the associated WMH of the individuals who played the games;Figure 6 illustrates plots that show the correlation between the scores of individuals who played the games with neuropsychological gold standard pen-and-paper tests carried out to test the executive function of the individuals who played the games;Figure 7 illustrates plots that show the correlation between the scores of individuals who played the games with neuropsychological gold standard pen-and-paper tests carried out to test the processing speed of the individuals who played the gamesFigure 8 illustrates plots that show the correlation between the scores of individuals who played the games with the MBI-C domain scores of the individuals who played the games;Figure 9 illustrates a flow diagram showing the training of a machine learning model with ground truth vascular' MCI diagnosis in accordance with embodiments of the disclosure;Figure 10 illustrates the inner layers of the machine learning model used to perform the classification of vascular MCI in accordance with embodiments of the disclosure;Figure 11 illustrates a flow chart showing the process for detecting vascular MCI in accordance with embodiments of the disclosure;Figure 12 illustrates a Receiver Operating Characteristic (ROC) curve that shows the performance of the trained machine learning model for detecting mild cognitive impairment (MCI) in accordance with embodiments of the disclosure; andFigure 13 illustrates a Receiver Operating Characteristic (ROC) curve that shows the performance of the trained machine learning model for detecting vascular mild cognitive impairment (MCI) in accordance with embodiments of the disclosure.DETAILED DESCRIPTION

[0013] The following detailed description is made with reference to the accompanying drawings, showing details and embodiments of the present disclosure for the purposes of illustration. Features that arc described in the context of an embodiment may correspondingly be applicable to the same or similar features in the other embodiments, even if not explicitly described in these other embodiments. Additions and / or combinations and / or alternatives as described for a feature in the context of an embodiment may correspondingly be applicable to the same or similar feature in the other embodiments.

[0014] In the context of various embodiments, the articles “a”, “an” and “the” as used with regard to a feature or clement include a reference to one or more of the features or elements.

[0015] In the context of various embodiments, the term “about” or “approximately” as applied to a numeric value encompasses the exact value and a reasonable variance as generally understood in the relevant technical field, e.g., within 10% of the specified value.

[0016] As used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items.

[0017] As used herein, “comprising” means including, but not limited to, whatever follows the word “comprising”. Thus, use of the term “comprising” indicates that the listed elements are required or mandatory, but that other elements are optional and may or may not be present.

[0018] As used herein, “consisting of’ means including, and limited to, whatever follows the phrase “consisting of’. Thus, use of the phrase “consisting of’ indicates that the listed elements are required or mandatory, and that no other elements may be present.

[0019] As used herein, an action-related tensor relates to a multidimensional array used to store and encode information or sets of data related to actions, tasks or interactions as recorded by an input / output interface module.

[0020] As used herein, cognitive stimuli refers to any external stimulation that engages and challenges an individual's mental processes, such as attention, memory, problem-solving, or decision-making. This stimulus is usually used to assess or enhance cognitive function of an individual.

[0021] As used herein, behavioral stimuli refers to any external stimulation that evokes an individual's observable actions, reactions, or behavioral responses. This stimulus is often used to assess emotional regulation, motivation, impulse control, or social interactions of the individual.

[0022] One skilled in the art will recognize that certain functional units in this description have been labelled as modules throughout the specification. The person skilled in the art will also recognize that a module may be implemented as circuits, logic chips or any sort of discrete component. Still further, one skilled in the art will also recognize that a module may be implemented in software which may then be executed by a variety of processor architectures. In embodiments of the disclosure, a module may also comprise computer instructions or executable code that may instruct a computer processor to carry out a sequence of events basedon instructions received. The choice of the implementation of the modules is left as a design choice for a person skilled in the art and does not limit the scope of the claimed subject matter in any way.

[0023] Through a comprehensive evaluation process, which included pen-and-paper neuropsychological assessments, detailed MRI brain scans, and blood-based evaluations for each subject, it was observed that dementia and cognitive impairment manifested particularly in populations where Executive Function (EF) and Processing Speed (PS) were diminished. Notably, it was found that EF and PS were significantly impaired in subjects with Vascular Mild Cognitive Impairment (MCI), a subtype commonly associated with silent strokes. Furthermore, this evaluation revealed that, in addition to cognitive impairments in EF and PS of these subjects, there were also behavioral impairments in Apathy and Disinhibition, in individuals with vascular MCI compared to those with non-vascular MCI.

[0024] Based on the above, a diagnostic approach is described in this disclosure where specially designed digital games or digital tasks are utilized to assess both cognitive domains (EF and PS) and behavioral domains (Apathy and Disinhibition) of subjects. This combination allows for the detection of vascular MCI and for vascular MCI to be distinguished from other types of MCI. Unlike traditional MCI assessments that focus solely on cognitive domains, this disclosure integrates behavioral markers into the analysis, offering a more comprehensive and accurate evaluation of vascular MCI in subjects.

[0025] First, behavioral markers such as, but are not limited to, Apathy and Disinhibition, are obtained based on stimulus provided to a user as the user plays the games or performs the digital tasks. Second, both cognitive and behavioral assessments are obtained based on stimuli experienced by the user as the user plays the games. Through this approach, the system is able to obtain a more holistic diagnostic picture of the user or players of the games.

[0026] A block diagram of a computing device 100 for detecting vascular mild cognitive impairment (MCI) based on digital biomarkers in accordance with embodiments of the disclosure is illustrated in Figure 1. Computing device 100 comprises input / output ( I / O) interface module 102 that is communicatively connected to computing module 104. In embodiments of the disclosure, computing device 100 may comprise, but is not limited to, asmartphone, a tablet, a laptop, a server, and any type of computing device. As for I / O interface module 102, this module may comprise a touchscreen, a keyboard, a mouse, a controller or any other sort of module that may be configured to display information to a user of the device, and to record data / information entered by the user.

[0027] In accordance with embodiments of the present disclosure, a block diagram representative of components of processing system 200 that may be provided within computing module 104, and / or any of the modules shown in Figure 1 to carry out the computing and processing functions in accordance with embodiments of the disclosure is shown in Figure 2. One skilled in the art will recognize that the exact configuration of each processing system provided within these modules may be different and the exact configuration of processing system 200 may vary and the arrangement illustrated in Figure 2 is provided by way of example only.

[0028] In embodiments of the disclosure, processing system 200 may comprise controller201 and user interface 202. User interface 202 is arranged to enable manual interactions between a user and the computing module as required and for this purpose includes the input / output components required for the user to enter instructions to provide updates to each of these modules. A person skilled in the art will recognize that components of user interface202 may vary from embodiment to embodiment but will typically include one or more of display 240, keyboard 235 and optical device 236.

[0029] Controller 201 is in data communication with user interface 202 via bus 215 and includes memory 220, processing unit or processor 205 mounted on a circuit board that processes instructions and data for performing the method of this embodiment, an operating system 206, an input / output ( I / O) interface 230 for communicating with user interface 202 and a communications interface, which in this embodiment is in the form of a network card 250. Network card 250 may, for example, be utilized to send data from these modules via a wired or wireless network to other processing devices or to receive data via the wired or wireless network. Wireless networks that may be utilized by network card 250 include, but are not limited to, Wireless-Fidelity (Wi-Fi), Bluetooth, Near' Field Communication (NFC), cellular networks, satellite networks, telecommunication networks, Wide Area Networks (WAN) etc.

[0030] Memory 220 and operating system 206 arc in data communication with processor 205 via bus 210. The memory components include both volatile and non-volatile memory and more than one of each type of memory, including Random Access Memory (RAM) 223, Read Only Memory (ROM) 225 and a mass storage device 245, the last comprising one or more solid-state drives (SSDs). One skilled in the art will recognize that the memory components described above comprise non-transitory computer-readable media and shall be taken to comprise all computer-readable media except for a transitory, propagating signal. Typically, the instructions arc stored as program code in the memory' components but can also be hardwired. Memory 220 may include a kernel and / or programming modules such as a software application that may be stored in either volatile or non-volatile memory.

[0031] Herein the term “processor” or “processing unit” is used to refer generically to any device or component that can process such instructions and may include: a microprocessor, a processing unit, a microcontroller, a programmable logic device or other computational device. That is, processor 205 may be provided by any suitable logic circuitry for receiving inputs, processing them in accordance with instructions stored in memory and generating outputs (for example to the memory components or on display 240). In this embodiment, processor 205 may be a single core or multi-core processor with memory addressable space. In one example, processor 205 may be multi-core, comprising — for example — an 8 core CPU. In another example, it could be a cluster of CPU cores operating in parallel to accelerate computations.

[0032] In operation, a selection of games, i.c., Games 1 - 4 or (A) - (B), will be presented to a user of computing device 100. As the user plays these games, the user’s inputs to computing device 100 and the stimulus presented to the user will be recorded and stored by I / O interface module 102. The data recorded may be stored in the form of action -related tensors or datasets where these action-related tensors may be generated in response to cognitive and / or behavioral stimuli, where the type of stimuli is dependent on the type of game that is being played. Computing module 104 then utilizes these action-related tensors to generate a set of features. For example, the set of features associated with the games may include, but is not limited to, the motivation level, the perseverance, the consistency and effort put in, and the adaptability of the subject; accuracy and control over users’ actions, speed of cognitive processing, reaction times and decision-making speed of the subject; the impulse control and attention span, reaction time and visuospatial processing speed, emotional control andcognitive flexibility of the subject; and the subject’s memory capabilities and willingness to put in more effort, memory retrieval speed and task engagement, and motivation and focus over time. More details about the workings of these games will be provided in the subsequent sections below.[0033 J Computing module 104 then proceeds to compute a vascular MCI assessment score by providing the generated set of features to a trained machine learning model. In embodiments of the disclosure, the machine learning model may be trained based on correlations between training sets of action-related tensors that were generated based on ground truth data and corresponding clinical data, where the clinical data comprises neuropsychological and neuroimaging data.

[0034] Figure 3 illustrates the comparative performance of individuals with vascular MCI and non-vascular MCI across key cognitive domains. Individuals with vascular- MCI were found to have higher WMH as compared to individuals with non-vascular MCI. The data plotted in Figure 3 illustrates that individuals with vascular MCI perform significantly worse in the cognitive domains of executive function (EF) and processing speed (PS) when compared to those with non-vascular MCI. It should be noted that the p-value shown in Figure 3 represents a statistical measure that is used to determine the significance of the results where the p-value indicates the probability that the observed differences between individuals with vascular MCI and those with non-vascular MCI occurred by chance. A low p-value (typically less than 0.05) suggests that the observed difference is statistically significant and unlikely to have occurred by random chance.

[0035] Figure 3 highlights that EF impairment is a key characteristic of Vascular Cognitive Impairment, which is often associated with silent strokes, which in turn contributes to a subject’s PS deficits. Specifically, individuals with vascular-related cognitive issues are particularly vulnerable to declining EF and PS abilities, primarily due to the impact of silent strokes and cerebrovascular disease on brain connectivity and processing capabilities. However, despite this recognition, it should be noted that there are currently no specialized diagnostic tools that target vascular MCI, nor are there comprehensive assessments that simultaneously evaluate both cognitive and behavioral domains in this population. Current tests predominantly focus on general cognitive impairments, without addressing the specificnuances of vascular MCI, as shown in Figure 3, where the cognitive impairments in EF and PS are more pronounced.

[0036] The Mild Behavioral Impairment-Checklist (MB1-C) is a tool designed to detect sustained and significant neuropsychiatric symptoms (NPS) over a six-month period in individuals at a prc-dcmcntia stage. These symptoms, which arc not typically captured by traditional psychiatric diagnostic systems, are assessed across the five behavioral domains shown in Figure 4. These MBI symptoms may manifest before or alongside MCI, offering an early indicator of potential cognitive decline. Figure 4 illustrates MBI-C domain scores of individuals with vascular MCI (i.e., individuals with high WMH) and non- vascular MCI (i.e., individuals with low WMH) across various behavioral domains. The domains assessed include Decreased Motivation, Emotional Dysregulation, Impulse Dyscontrol, Social Inappropriateness, and Abnormal Thoughts.

[0037] The plots in Figure 4 show that individuals with High WMH perform worse in the Decreased Motivation and Impulse Dyscontrol domains compared to those with Low WMH. This suggests a trend where individuals with vascular MCI may have greater impairment in the Decreased Motivation (i.e., Apathy) and Impulse Dyscontrol (i.e., Disinhibition) domains.

[0038] For the remaining behavioral domains such as Emotional Dysregulation, Social Inappropriateness, and Abnormal Thoughts, the differences between individuals with High and Low WMH can be seen to be minimal. This indicates that these domains do not show a significant difference between individuals with vascular MCI and non-vascular MCI, regardless of WMH levels.

[0039] In summary, Decreased Motivation and Impulse Dyscontrol are the only behavioral domains that show a statistically significant worsening in individuals with High WMH compared to Low WMH. The data for other behavioral domains like Emotional Dysregulation, Social Inappropriateness, and Abnormal Thoughts suggest that these are less affected by WMH levels in individuals with vascular MCI. These findings underscore the importance of focusing on Impulse Dyscontrol and Decreased Motivation as key behavioral markers in vascular MCI.

[0040] First Embodiment of Games (i.e., Games 1 - 4)

[0041] In a first embodiment of the disclosure, the following games as set out in Table 1 were developed in order to obtain the behavioral and cognitive markers as described above.Table 1

[0042] Game 1 - Symbol Matching

[0043] In embodiments of the disclosure, as Game 1 is played by a user, Game 1 will elicit responses from the user to cognitive stimuli presented during gameplay. These stimuli generated by Game 1, prompt the player to engage and respond in a manner described as follows.

[0044] Game 1 involves matching a set of identifiers (e.g., numbers and letters) to corresponding symbols (e.g., shapes and colors) from a reference list. The objective is to match as many symbols to the identifiers as accurately and quickly as possible within a specified number of attempts. The game is time-limited, and various data points are captured during gameplay, including:• The number of correctly matched symbols and identifiers within the given time.• The number of incorrect matches relative to the total number of attempts.• Each instance when the player interacts with the screen.• The time interval between each matching attempt.• Patterns of successful and unsuccessful attempts.• The reaction time, measured as the delay between the presentation of a symbol and the player’s response.• A comparison of performance at the beginning of the game versus the end, particularly under time pressure.• The time taken to observe the symbols and initiate each task.• The performance differences when switching between various sets of symbols.• Patterns of mismatches, such as incorrect symbol-to-color associations.

[0045] The data recorded in relation to Game 1 is stored as an action-related tensor in computing module 104. The number of correctly matched symbols and identifiers provides insight into the player’s processing speed, while the number of incorrectly matched pairs provides insight about the player’s disinhibition. The recorded data may then be used by the computing module 104 to generate a memory and processing speed-related score.

[0046] Game 2 - Connect Em

[0047] In embodiments of the disclosure, as Game 2 is played by a user, Game 2 will elicit responses from the user to cognitive stimuli presented during gameplay. These stimuli generated by Game 2, prompt the player to engage and respond in a manner described as follows.

[0048] A player is first tasked with matching one symbol (e.g., numbers, letters, colors, shapes) to another set of corresponding symbols (e.g., numbers, letters, colors, shapes) in a specified sequence. The objective is to accurately connect the symbols in the correct order. The player begins the game with a predetermined number of lives. The following data is captured during gameplay:• The number of correct matches between symbols.• The time taken to complete the sequence.• The number of incorrect matches relative to the number of attempts.• The total number of taps made by the player.• The time interval between consecutive taps.• The time taken to initiate each tap.• Patterns of incorrect taps.• Differences in gameplay performance between the beginning and end of the game.• Patterns of screen interactions, including the location of each tap.

[0049] The data recorded in relation to Game 2 is stored as an action-related tensor in computing module 104. The speed and accuracy with which the player matches numbers to letters or shapes serve as indicators of executive functioning where the speed and accuracy of the player implies that the player has stronger executive function abilities. The recorded data may then be used by the computing module to generate an executive function-related score.

[0050] Game 3 - Airplane Game

[0051] In embodiments of the disclosure, as Game 3 is played by a user, Game 3 will elicit responses from the user to behavioral stimuli presented during gameplay. These stimuligenerated by Game 3, prompt the player to engage and respond in a manner described as follows.

[0052] In this game, the player swipes in the direction indicated by the airplane of a specific color. The objective is to swipe correctly through as many levels as possible. With each level, the color of the target airplane changes, and the number of distractor airplanes also varies. The game collects the following data:• The number of correct swipes.• The number of incorrect swipes.• The time taken to complete each swipe.• Patterns of disinhibition, as reflected in impulsive swipes on incorrect or distractor airplanes of different colors.• Differences in player reactions between earlier, easier levels and more challenging levels.• The time taken to initiate each movement.• Overall reaction time.

[0053] The data recorded in relation to Game 3 is stored as an action-related tensor in computing module 104. The player's ability to complete higher levels successfully provides an insight to the player’s disinhibition and visuospatial abilities, with faster and more accurate swipes reflecting stronger behavioral abilities. The recorded data may then be used by the computing module to generate visuospatial-rclatcd score.

[0054] Game 4 - Grocery Shopping

[0055] In embodiments of the disclosure, as Game 4 is played by a user, Game 4 will elicit responses from the user to behavioral stimuli presented during gameplay. These stimuli generated by Game 4, prompt the player to engage and respond in a manner described as follows.

[0056] In this game, the player is initially presented with a list of visual items to memorize. The list is then removed, and the player is tasked with selecting the correct items from memory.The player may rcchcck the list as often as necessary. The game progresses through multiple rounds, each increasing in difficulty. The following data is captured during gameplay:• The total number of visual items selected within a given time.• The number of correct visual items selected within the allotted time.• The number of incorrect visual items selected within the same time frame.• The time taken to select each item.• The time required to complete the entire list of items.• The patterns of item selection and memory retention as the game advances.• The number of times the original list is reopened in each round.• Patterns of reopening the list as the difficulty increases.

[0057] The data recorded in relation to Game 4 is stored as an action-related tensor in computing module 104. The game provides insights into the player’ s motivation and apathy by analysing these features across multiple rounds, providing insights into the player's behavioral engagement throughout the task. The recorded data may then be used by the computing module to generate a motivation and apathy-related score.

[0058] Second Embodiment of Games (i.e., Games (A) - (Dj)

[0059] In a second embodiment of the disclosure, the following games as set out in Table 2 were developed in order to obtain the behavioral and cognitive markers as described above.Table 2

[0060] Game (A) - Keep the Ball Alive

[0061] In embodiments of the disclosure, as Game (A) is played by a user, Game (A) will elicit responses from the user to behavioral stimuli presented during gameplay. These stimuli generated by Game (A), prompt the player to engage and respond in a manner described as follows.

[0062] In Game (A), the player taps the screen to manoeuvre an object while avoiding obstacles. The objective is to remain in the game for as long as possible without colliding with any obstacles. The player begins with a set number of lives and the data captured during gameplay includes:• The number of times the object collides with an obstacle.• The time elapsed before hitting an obstacle.• The location of each tap on the screen.• Changes in the tap location throughout the game.• The frequency of tapping.• Patterns of tapping as the game progresses.• The number of collisions relative to the number of attempts.• The total time the object stays alive.The player's response to various obstacle combinations (i.c., whether they successfully navigate them).The point of contact where the object hits an obstacle (e.g., front, top, bottom, or back).

[0063] The data recorded in relation to Game (A) is stored as an action-related tensor in computing module 104. The duration of time the object remains alive is correlated with the player's level of motivation and this provides insights into the player’s motivation and apathy when these features arc analysed across multiple rounds.

[0064] Game (B) - Symbol Matching

[0065] In embodiments of the disclosure, as Game (B) is played by a user, Game (B) will elicit responses from the user to behavioral stimuli presented during gameplay. These stimuli generated by Game (B), prompt the player to engage and respond in a manner described as follows.

[0066] As Game (B) progresses, players are required to match identifiers (e.g., numbers and letters) to symbols (e.g., shapes and colors) from a reference list. The goal is to match as many symbols as possible accurately and quickly within the set number of attempts within a time limit. The data captured includes:• The number of correctly matched symbols and numbers within the allotted time.• The number of incorrect matches and total attempts.• Each interaction with the screen (i.e., every time the player touches the screen).• The time interval between each attempt.• Patterns of successful and unsuccessful attempts.• The reaction time, measured as the delay between showing the symbol and the touch response.• A comparison of performance at the beginning versus the end of the game under time pressure.• The time taken to observe and initiate each task.• Performance differences when switching between different sets of symbols.• Mismatch patterns (e.g., incorrect symbol-to-color associations).

[0067] The data recorded in relation to Game (B) is stored as an action-related tensor in computing module 104. A high number of correctly matched symbols indicates quick processing speed, while incorrect matches indicate disinhibition.

[0068] Game (C) - Connect Em

[0069] In embodiments of the disclosure, as Game (C) is played by a user, Game (C) will elicit responses from the user to cognitive stimuli presented during gameplay. These stimuli generated by Game (C), prompt the player to engage and respond in a manner described as follows.

[0070] This game involves matching one symbol (e.g., numbers, letters, colors, or shapes) to another in a predefined sequence. The goal is to correctly connect symbols in the appropriate order, with players starting with a set number of lives. The data captured includes:• The number of correct symbol matches.• The time taken to complete the sequence.• The number of incorrect matches relative to total attempts.• The total number of taps.• The time interval between each tap.• The time taken to initiate the first tap.• Patterns of incorrect taps.• Differences in performance between the beginning and end of the game.• Patterns of touch locations on the screen.

[0071] The data recorded in relation to Game (C) is stored as an action-related tensor in computing module 104. Faster and more accurate matches during gameplay indicates that the user has strong executive functioning abilities.

[0072] Game (D) - Maze

[0073] In embodiments of the disclosure, as Game (D) is played by a user, Game (D) will elicit responses from the user to cognitive stimuli presented during gameplay. These stimuligenerated by Game (D), prompt the player to engage and respond in a manner described as follows.

[0074] In this game, the player moves an object to avoid obstacles that appear randomly on the screen. The difficulty increases as the game progresses. The goal is to avoid obstacles as quickly as possible. The player begins with a set number of lives. The data captured includes:• The number of times the object collides with obstacles relative to total attempts.• The time taken to complete each task at various levels.• The highest level completed.• Patterns of incorrect moves.• The distance between the object and the correct path.• The time interval between each move.• Patterns of responses after a blocked move.• Reaction differences between earlier, easier levels and later, more challenging ones.• The time taken to initiate each move.• Overall reaction time for each move.

[0075] The data recorded in relation to Game (D) is stored as an action-related tensor in computing module 104. Completion of higher levels indicates better processing speed and visuospatial abilities.

[0076] Based on the above games, it can be noted that upon completion of these games, computing module 104 would have recorded a plurality of action -related tensors that were each generated in response to cognitive and / or behavioral stimuli. Computing module 104 may then generate from the plurality of action -related tensors a mcmory-rclaied score, an attention- related score, an executive function-related score, a visuospatial-related score, a language related-score, a processing speed-related score, a motor task-related score, a sensory-motor integration-related score, a social-cognition-related score, a problem solving -related score, a decision making-related score, a motivation and apathy -related score or a learning-related score based on a part of the behavioral stimuli and a part of the cognitive stimuli as contained within each of the plurality of action-related tensors. The generation of the respective scores are based on the data captured for each of the games. For example, a memory-related score may be computed based on the total number of visual items selected within a given time, the numberof correct visual items selected within the allotted time, the time taken to select each of the items and the time required to complete the entire list of items.

[0077] The following sections describe the methods utilized to obtain clinical data comprising neuroimaging and / or neuropsychological data of subjects.

[0078] Neuroimaging

[0079] Tn embodiments of the disclosure, MRT scans of subjects were obtained using two systems: the 3T Prisma fit System (Siemens, Erlangen, Germany) and the 3T Ingenia System (Philips Medical Systems). Each of the subjects of the MRI scans using the 3T Prisma fit System underwent high-resolution T1 -weighted Magnetization Prepared Rapid Gradient Echo (MPRAGE) imaging based on the following settings: 192 continuous sagittal slices, TR / TE / n = 2300 / 2.28 / 900ms, flip angle = 8°, FOV = 256 * 240 mm2, matrix = 256 x 240, isotropic voxel size = 1.0 * 1.0 x 1.0 mm3, and a bandwidth = 200 Hz / pixel), as well as Fluid Attenuated Inversion Recovery (FLAIR) sequences with the following settings: 192 continuous sagittal slices, TR / TE / TI= 5000 / 387.0 / 1800ms, flip angle = 120°, FOV = 256 x 256 mm2, matrix = 256 x 256, isotropic voxel size = 1.0 * 1.0 * 1.0 mm3, and a bandwidth = 750 Hz / pixel. In summary, imaging parameters, including slice thickness, repetition time (TR), echo time (TE), inversion time (TI), field of view (FOV), and voxel size, were adjusted accordingly.

[0080] When the scans were performed using the 3T Ingenia System, each participant underwent high resolution Tl-weighted Magnetization Prepared Rapid Gradient Echo imaging based on the following settings: 180 continuous sagittal slices, TR / TE = 7.46 / 3.4ms, flip angle = 8°, FOV = 256 x 256 mm2, matrix = 256 * 256, isotropic voxel size = 1.0 x 1.0 x 1.0 mm3) and FLAIR sequences based on the following settings: 200 continuous sagittal slices, TR / TE / TI = 4800 / 378.15 / 1650ms, flip angle = 90°, FOV = 240 x 240 mm2, matrix = 240 x 240, and a isotropic voxel size = 1.0 x 1.0 x 2.0 mm3.

[0081] Once the scans were complete, the obtained images were reviewed for motion artifacts and significant pathological findings, and those with such issues were excluded from further analysis.

[0082] Image Pre-processing

[0083] Image pre-processing on the generated images were performed using the Computational Anatomy Toolbox (CAT12) within Statistical Parametric Mapping (SPM12). The Tl-wcightcd images were normalized via affine transformation and non-linear registration, followed by bias field inhomogeneity collection. The images were then segmented into grey matter (GM), white matter (WM), and cerebrospinal fluid (CSF) components for voxel-based morphometry analysis.

[0084] Measurement of Cerebrovascular Disease Burden

[0085] Cerebrovascular disease burden was assessed using Tl-wcightcd and FLAIR images rated according to the modified Fazekas scale for White Matter Hyperintensity (WMH) severity. Periventricular and deep subcortical WMH were scored separately on a scale of 0-3 for both hemispheres. Specifically, the scoring criteria were as follows: for periventricular WMH, the absence of any WMH = 0; the presence of caps or pencil-thin lining = 1; a smooth halo along the edges of the lateral ventricle = 2; and irregular hyperintensities extending into deep white matter = 3. For deep subcortical WMH, the absence of any WMH = 0; the presence of non-conflucnt foci of WMH in the deep subcortical region = 1 ; beginning confluence of WMH foci = 2; and the presence of large confluent areas = 3. The modified Fazekas scale allows for quantification of white matter lesions in four brain regions, namely right periventricular, left periventricular, right deep subcortical, and left deep subcortical to provide a score range of 0-12. The visual ratings were performed by two independent raters, with any discrepancies resolved by consensus.

[0086] Neuropsychological Assessments

[0087] Participants underwent global cognitive assessments, including the Clinical Dementia Rating Scale, Montreal Cognitive Assessment, and Visual Cognitive Assessment Test. A detailed neuropsychological test evaluated cognitive domains such as episodic memory, executive function, language, processing speed, and visuospatial function. Additionally, participants completed behavioral questionnaires, including the Mild BehavioralImpairment-Checklist (MBI-C), covering domains such as Interest, Mood, Control, Social, and Beliefs.

[0088] The subjects that underwent the neuropsychological and neuroimaging assessments described above were then tasked to play Games 1 - 4. Data collected for 50 such subjects are then plotted in Figures 5-8 whereby these figures illustrate the relationships between the scores obtained from these games and the clinical data obtained from the clinical assessments for associated subjects.

[0089] Figure 5 illustrates the correlation between WMH, a marker of silent strokes, and game scores across four different games, i.e., Games 1-4. The WMH of the subjects (which were obtained using the neuroimaging methods described above) are plotted on the x-axis, representing a range from 0 to 12, where higher values indicate greater white matter damage, often linked to silent strokes. The game scores obtained by the subjects are then plotted on the y-axis, ranging from 0 to 1, with higher values indicating better performance.

[0090] The downward- sloping trend lines across all games suggest an inverse relationship between levels of WMH and game performance whereby as the levels of WMH increases, game scores decrease. This relationship is statistically significant for all four games, as indicated by the p-valucs, with Game 3 showing the strongest correlation (p < 0.001), followed by Game 1 (p = 0.017), Game 4 (p = 0.043), and Game 2 (p = 0.049). This pattern implies that individuals with higher WMH, indicative of more severe silent strokes, tend to perform worse across the games. The results suggest that these games can effectively capture cognitive or behavioral deficits related to silent strokes or Vascular Mild Cognitive Impairment (MCI).

[0091] Figure 6 illustrates the relationship between executive function scores (which were obtained using the neuropsychological methods described above) and game scores for four different games, i.e., Games 1 -4. The executive function scores are plotted on the x-axis, ranging from 0.0 to 1.0, representing varying levels of cognitive performance in this domain, where higher executive function scores correlate to lower executive function as the neuropsychological method performed in this example comprised an inverse test. Game scores arc plotted on the y-axis, ranging from 0.0 to 1.0, where higher scores reflect better performance in the respective games.

[0092] The negative slopes of all regression lines indicate an inverse correlation: whereby as the executive function scores improve, the game scores of associated subjects decrease. This suggests that individuals with weaker executive function tend to perform worse in these specific games. Hence, the game scores correlate directly to the executive function of individuals.

[0093] Figure 7 illustrates the relationship between processing speed scores (which were obtained using the neuropsychological methods described above) and game scores for two different games, i.e., Games 2 and 3. The processing speed scores are plotted on the x-axis, and ranges from 0.0 to 1 .0, with higher values indicating faster cognitive processing. Game scores of the associated subjects are plotted on the y-axis, also ranging from 0.0 to 1.0, where higher games scores indicate better performance in the respective games.

[0094] Both games show a positive correlation between processing speed and game performance, as indicated by the upward-sloping lines. This suggests that individuals with faster processing speeds tend to achieve higher game scores.

[0095] Figure 8 illustrates the relationship between behavioral impairment scores (which were obtained using the neuropsychological methods described above) and game scores for associated subjects across four different games, i.e., Games 1-4. The behavioral impairment scores are plotted on the x-axis, ranging from 1 to 3, where higher values indicate greater impairment. Game scores of associated subjects are plotted on the y-axis, ranging from 0 to 1 , with higher values representing better performance in the respective games.

[0096] The negative slopes for all four games suggest an inverse relationship between behavioral impairment scores and game performance of associated subjects whereby as behavioral impairment increases, game performance of the associated subjects tends to decrease.

[0097] Based on the information obtained from Figures 5-8, it can be seen that the gameplay of the four games produces game scores that arc highly correlated with silent strokes and someof the games arc either correlating or trending towards the two cognitive domains and two behavioural domains specific to vascular MCI.

[0098] Figure 9 illustrates a flow diagram that shows the process for diagnosing MCI, based on clinical data of a subject in accordance with embodiments of the disclosure. This diagram shows that in addition to diagnosing MCI, the process is also configured to determine whether the diagnosed MCI is of the vascular or non-vascular subtype. The process begins with the subject undergoing clinical evaluation which may comprise of neuropsychological assessments and / or neuroimaging evaluation to assesses the cognitive abilities of the subject to determine whether the individual is cognitively normal - 906 or has MCI - 904. If the individual's cognitive scores fall below a certain threshold, they are diagnosed with MCI - 904.

[0099] Once MCI is diagnosed, MRI imaging is used to assess the presence or absence of silent strokes, which helps differentiate between the two subtypes of MCI. If the MRI reveals silent strokes, the individual is classified as having vascular MCI - 908. Conversely, if no silent strokes are detected in the MRI, the individual is diagnosed with non-vascular MCI - 910. This decision tree demonstrates the integration of cognitive assessments and neuroimaging in diagnosing and classifying MCI subtypes.

[0100] Training of the Machine Learning Model

[0101] A training data set, i.e., a ground truth set, may then be generated based on the correlations between the clinical data, comprising neuropsychological and ncuroimaging data of a group of subjects, and the game scores of these subjects as they complete some and / or all of the games mentioned in this disclosure, i.e., the game scores that were generated in response to behavioral and cognitive stimuli. Table 3 below sets out the demographic and cognitive characteristics of two groups whereby the information obtained from the subjects in these two groups may be used to generate the training data set.Table 3

[0102] The table above shows that there are 132 individuals in the cognitively normal group and 1 18 individuals in the MCI group. The other parameters of the two groups are the average age of the individuals, the average number of years of education of the individuals, the average Montreal Cognitive Assessment (MoCA) score of the individuals and the average ReCOGnALze score of the individuals. ReCOGnAIze Score was computed using machine learning features derived from the 4 games. This shows that the ReCOGnAIze Score is able to significantly differtiate between MCI and Cognitively Normal.

[0103] Figure 10 illustrates a machine learning model designed for the classification of vascular MCI and non-vascular MCI based on input features derived from four distinct games, i.e.. Games 1 - 4 in accordance with embodiments of the disclosure. As shown, the input layer comprises features extracted from Game 1, Game 2, Game 3, and Game 4, which serve as the primary data points for the model. These features are passed through several hidden layers of neurons, where complex relationships and patterns are learned through weighted connections. Each hidden layer processes the input data in increasingly abstract ways to capture the underlying features relevant to the classification task.

[0104] As the data progresses through the hidden layers, the model attempts to distinguish, based on its training, between individuals with vascular MCI and those without. In embodiments of the disclosure, the model first distinguishes between normal and MCI, and within the MCI data, the final output layer then presents two possible outcomes: vascular' MCI or non-vascular MCI, indicating whether the individual likely has vascular MCI based on the game data. The training data set mentioned above may be used to train this machine learning model to generate a vascular MCI assessment score which can be used to classify whether the features provided represent a user with vascular MCI or non-vascular MCI. In embodiments of the disclosure, the machine learning model may comprise models such as, but is not limited to, a feedforward neural network architecture, a fully connected neural network or a multilayer perceptron model.

[0105] In embodiments of the disclosure, the model illustrated in Figure 10 may comprise a deep learning-based neural network regression model that may be employed to integrate various game parameters as input features, with the classification of vascular MCI or non- vascular MCI being generated as the output of the model. The training labels used to train the model may be obtained from subjects who have underwent cognitive neuropsychological test results to assess their MCI statuses, along with the Fazekas Scale and / or vascular cognitive impairment markers on MRI to evaluate vascular load, which will be stratified into different categories. One skilled in the art will recognize that other neuropsychological tests may be performed to classify the subjects according to their MCI classification.

[0106] In embodiments of the disclosure, vascular MCI may be defined by a White Matter Hyperintensity (WMH) score > 4 on the Fazekas Scale, a cognitive performance below 1.5 standard deviations in at least one of the five neuropsychological domains relative to the cognitively normal population, and a Clinical Dementia Rating (CDR) of 0.5 or more.

[0107] On the other hand, non-vascular MCI is characterized by a WMH score < 4 on the Fazekas Scale, a cognitive performance below 1.5 standar d deviations in at least one of the five neuropsychological domains, and a CDR of 0.5 or more. The model will then utilize these criteria to classify and predict vascular-related cognitive impairment of a subject based on the integration of game-derived features.

[0108] A flowchart which sets out the process for detecting vascular' mild cognitive impairment (MCI) using a computing module in accordance with embodiments of the present disclosure is illustrated in Figure 11. In embodiments of the disclosure, process 1100 as illustrated in Figure 11 may be performed by computing module 104 or any combination of modules provided within computing device 100.

[0109] Process 1100 begins at step 1102 with process 1100 receiving and analyzing a plurality of action-related tensors from an input / output (I / O) interface module communicatively coupled to the computing module. The I / O interface module was configured to record inputs provided by a user of the I / O module. In embodiments of the disclosure, the first and second plurality tensor are each generated in response to cognitive stimuli designedto assess executive function and processing speed of the user, and a third and a fourth action- related tensor are each generated in response to behavioral stimuli designed to assess apathy and disinhibition of the user.

[0110] Process 1100 then proceeds to generate a set of multi-domain features derived from the plurality of action-related tensors, and this takes place at step 1104. A vascular MCI assessment score is then computed by process 1100 at step 1106 by providing the generated set of features to a trained machine learning model, whereby the machine learning model was trained based on correlations between a plurality of training sets of action -related tensors that were generated in response to behavioral and cognitive stimuli and corresponding clinical data, the clinical data comprising neuropsychological and neuroimaging data.

[0111] In embodiments of the disclosure, the cognitive stimuli comprises at least one type of visuomotor task. In further embodiments, the first action-related tensor may comprise the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of identifiers on an interface of the I / O interface module and being prompted to match the displayed identifiers in a predetermined order, wherein the recorded inputs comprise a number of correctly matched identifiers, a duration of time taken to complete each match attempt, a total time taken to complete an entire matching of the set of identifiers, touch locations on the interface for each attempt, reaction times between presentation of an identifier and the user’s response, or patterns of correct matched identifiers.[001 12] In further embodiments, the second action-related tensor may comprise the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of symbols on an interface of the I / O interface module and being prompted to connect the symbols in a predetermined order, wherein the recorded inputs comprise a number of correctly connected symbols, a duration of time taken to complete each connection, a total time taken to complete an entire connection of the set of symbols, a number of taps recorded on the interface, time intervals between consecutive taps on the interface, or patterns of correct connections.

[0113] In embodiments of the disclosure, the behavioral stimuli may comprise at least one type of reaction task. In further embodiments, the third action-related tensor may comprise theinputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of visual stimuli comprising a target object and a plurality of obstacles on an interface of the I / O interface module and being prompted to perform an input action on the interface of the I / O interface module based on a direction of the target object, wherein the recorded inputs comprise a number of correct input actions in relation to variations of the target object, a duration of time taken to complete each input action, a total time taken to complete a level, a difficulty value obtained based on a number of the obstacles and a number of variations of the target object, or patterns of correct input actions.

[0114] In further embodiments, the fourth action-related tensor may comprise the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a visual list of items and being prompted to memorize objects contained in the list of items, wherein the recorded inputs comprise a number of correctly selected items, a duration of time taken to complete an item selection task, a number of times the user accesses the visual list of items, a duration of time taken by the user to select an item, or patterns of correct selections.

[0115] In embodiments of the disclosure, the step of generating the set of features derived from the plurality of action-related tensors further comprises the steps of generating a memory- related score, an attention-related score, an executive function-related score, a visuospatial- related score, a language related-score, a processing speed-related score, a motor task-related score, a sensory-motor integration-related score, a social-cognition-rclatcd score, a problem solving-related score, a decision making-related score or a learning -related score based on a part of the behavioral stimuli and a part of the cognitive stimuli as contained within each of the plurality of action-related tensors.

[0116] In embodiments of the disclosure, the computed vascular MCI assessment score is expressed in terms of a score on a Fazekas Scale, vascular cognitive impairment markers on MRI (c.g., ncuroimaging features associated with vascular contributions to cognitive impairment), a score from one or more neuropsychological domains, and a Clinical Dementia Rating score. The machine learning model may also comprise a deep learning based neural network regression model.

[0117] Figure 12 illustrates a Receiver Operating Characteristic (ROC) curve that shows the performance of a trained machine learning model for detecting MCI in accordance with embodiments of the disclosure when the model was used on a sample size of 250 subjects. The x-axis represents a false positive rate (1 - specificity), and the y-axis represents a true positive rate (sensitivity). Plot 1202 represents the ROC curve while plot 1204 represents the performance of a random guess. As can be seen from Figure 12, the area under the curve (AUC) was found to be 0.89 which indicates that the model performed well in distinguishing between individuals with and without MCI. The model was also able to achieve a balanced sensitivity and specificity (83% and 85%) indicating that the model has a reliable performance in both identifying MCI cases and ruling out non-MCI cases.

[0118] Figure 13 illustrates a ROC curve that shows the performance of a trained machine learning model for detecting vascular MCI in accordance with embodiments of the disclosure when the model was used on a subset of 118 MCI subjects, with 59 subjects having vascular type MCI. The x-axis represents a false positive rate (1 - specificity), and the y-axis represents a true positive rate (sensitivity). Plot 1302 represents the ROC curve while plot 1304 represents the performance of a random guess. As can be seen from Figure 13, the AUC was found to be 0.88, which indicates that the model performed well in distinguishing between individuals with and without vascular MCI, i.e. the model shows strong diagnostic capability specifically for vascular subtype identification. The model was also able to achieve a balanced sensitivity and specificity (82% and 82%) indicating that the model has a reliable performance in both identifying vascular MCI cases and ruling out non-vascular MCI cases.

[0119] Numerous other changes, substitutions, variations, and modifications may be ascertained by the skilled in the art and it is intended that the present application encompass all such changes, substitutions, variations, and modifications as falling within the scope of the appended claims.

Claims

CLAIMS:

1. A computing module for detecting vascular mild cognitive impairment (MCI) comprising: a processing unit; and a non-transitory media readable by the processing unit, the media storing instructions that when executed by the processing unit causes the processing unit to: receive and analyse a plurality of action-related tensors from an input / output (VO) interface module communicatively coupled to the module, the I / O interface module being configured to record real-time inputs from a user interacting with the I / O interface module, wherein the plurality of action-related tensors comprises: a first and a second action-related tensor that are each generated in response to cognitive stimuli designed to assess executive function or processing speed of the user, a third and a fourth action-related tensor that are each generated in response to behavioral stimuli designed to assess apathy or disinhibition of the user, whereby the cognitive and behavioral stimuli are presented to the user through the I / O interface module, generate a multi-domain feature set based on the plurality of action-related tensors; compute a vascular MCI assessment score by providing the multi-domain feature set to a trained machine learning model, whereby the machine learning model was trained based on correlations between a plurality of training sets of action-related tensors and corresponding clinical data comprising neuropsychological and neuroimaging data.

2. The computing module according to claim 1, wherein the cognitive stimuli comprise at least one type of visuomotor task.

3. The computing module according to claim 1, wherein the first action-related tensor comprise the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of identifiers on an interface of the I / O interface module and being prompted to match the displayed identifiers in a predetermined order, wherein the recorded inputs comprise a number of correctly matched identifiers, a duration of time taken to complete each match attempt, a total time taken to complete an entire matching of the set of identifiers, touch locations on the interface for each attempt, reaction times between presentation of an identifier and the user’s response, or patterns of correct matched identifiers.

4. The computing module according to claim 1, wherein the second action -related tensor comprise the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of symbols on an interface of the I / O interface module and being prompted to connect the symbols in a predetermined order, wherein the recorded inputs comprise a number of correctly connected symbols, a duration of time taken to complete each connection, a total time taken to complete an entire connection of the set of symbols, a number of taps recorded on the interface, time intervals between consecutive taps on the interface, or patterns of correct connections.

5. The computing module according to claim I, wherein the behavioral stimuli comprise at least one type of reaction task.

6. The computing module according to claim 1, wherein the third action-related tensor comprise the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of visual stimuli comprising a target object and a plurality of obstacles on an interface of the I / O interface module and being prompted to perform an input action on the interface of the I / O interface module based on a direction of the target object, wherein the recorded inputs comprise a number of correct input actions in relation to variations of the target object, a duration of time taken to complete each input action, a total time taken to complete a level, a difficulty-value obtained based on a number of the obstacles and a number of variations of the target object, or patterns of correct input actions.

7. The computing module according to claim 1, wherein the fourth action-related tensor comprise the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a visual list of items and being prompted to memorize objects contained in the list of items, wherein the recorded inputs comprise a number of correctly selected items, a duration of time taken to complete an item selection task, a number of times the user accesses the visual list of items, a duration of time taken by the user to select an item, or patterns of correct selections.

8. The computing module according to claim 1, whereby the instructions that causes the processing unit to generate the set of features derived from the plurality of action-related tensors further comprises instructions for directing the processing unit to: generate a memory and processing speed-related score, an executive function-related score, a visuospatial-related score, or a motivation and apathy-related score based on a part of the behavioral stimuli and a part of the cognitive stimuli as contained within each of the plurality of action-related tensors.

9. The computing module according to claim 1, wherein the computed vascular MCI assessment score is expressed in terms of a score based on vascular cognitive impairment markers on magnetic resonance imaging (MRI), and a score from one or more neuropsychological domains.

10. The computing module according to claim 1, wherein the machine learning model comprise a deep learning based neural network regression model.

11. A method for detecting vascular mild cognitive impairment (MCI) using a computing module, the method comprising: receiving and analysing a plurality of action-related tensors from an input / output (I / O) interface module communicatively coupled to the module, the I / O interface module being configured to record real-time inputs from a user interacting with the I / O interface module, wherein the plurality of action-related tensors comprises: a first and a second action-related tensor that are each generated in response to cognitive stimuli designed to assess executive function and processing speed of the user, a third and a fourth action -related tensor that are each generated in response to behavioral stimuli designed to assess apathy and disinhibition of the user, whereby the cognitive and behavioral stimuli are presented to the user through the I / O interface module, generating a multi-domain feature set based on the plurality of action-related tensors; computing a vascular MCI assessment score by providing the multi-domain feature set to a trained machine learning model, whereby the machine learning model was trained based on correlations between a plurality of training sets of action -related tensors and corresponding clinical data comprising neuropsychological and neuroimaging data.

12. The method according to claim 11 , wherein the cognitive stimuli comprise at least one type of visuomotor task.

13. The method according to claim 11, wherein the first action-related tensor comprises the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of identifiers on an interface of the I / O interface module and being prompted to match the displayed identifiers in a predetermined order, wherein the recorded inputs comprise a number of correctly matched identifiers, a duration of time taken to complete each match attempt, a total time taken to complete an entire matching of the set of identifiers, touch locations on the interface for each attempt, reaction times between presentation of an identifier and the user’s response, or patterns of correct matched identifiers.

14. The method according to claim 11, wherein the second action-related tensor comprises the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of symbols on an interface of the I / O interface module and being prompted to connect the symbols in a predetermined order, wherein the recorded inputs comprise a number of correctly connected symbols, a duration of time taken to complete each connection, a total time taken to complete an entire connection of the set of symbols, a number of taps recorded on the interface, time intervals between consecutive taps on the interface, or patterns of correct connections.

15. The method according to claim 11, wherein the behavioral stimuli comprise at least one type of reaction task.

16. The method according to claim 11, wherein the third action-related tensor comprises the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a set of visual stimuli comprising a target object and a plurality of obstacles on an interface of the I / O interface module and being prompted to perform an input action on the interface of the I / O interface module based on a direction of the target object, wherein the recorded inputs comprise a number of correct input actions in relation to variations of the target object, a duration of time taken to complete each input action, a total time taken to complete a level, a difficulty value obtained based on a number of the obstacles and a number of variations of the target object, or patterns of correct input actions.

17. The method according to claim 11, wherein the fourth action-related tensor comprises the inputs recorded by the I / O interface module in response to a user of the I / O interface module being displayed a visual list of items and being prompted to memorize objects contained in the list of items, wherein the recorded inputs comprise a number of correctly selected items, a duration of time taken to complete an item selection task, a number of times the user accesses the visual list of items, a duration of time taken by the user to select an item, or patterns of correct selections.

18. The method according to claim 11, whereby the step of generating the set of features derived from the plurality of action-related tensors further comprises the steps of: generating a memory and processing speed-related score, an executive function-related score, a visuospatial-related score, or a motivation and apathy-related score based on a part of the behavioral stimuli and a part of the cognitive stimuli as contained within each of the plurality of action-related tensors.

19. The method according to claim 11, wherein the computed vascular MCI assessment score is expressed in terms of a score based on vascular cognitive impairment markers on magnetic resonance imaging (MRI), and a score from one or more neuropsychological domains20. The method according to claim 11, wherein the machine learning model comprises a deep learning based neural network regression model.

Images