WO2024065052A1 - Biomarqueurs de neuro-imagerie fonctionnelle de la récupération après lésion cérébrale et leurs utilisations - Google Patents

Biomarqueurs de neuro-imagerie fonctionnelle de la récupération après lésion cérébrale et leurs utilisations Download PDF

Info

Publication number
WO2024065052A1
WO2024065052A1 PCT/CA2023/051284 CA2023051284W WO2024065052A1 WO 2024065052 A1 WO2024065052 A1 WO 2024065052A1 CA 2023051284 W CA2023051284 W CA 2023051284W WO 2024065052 A1 WO2024065052 A1 WO 2024065052A1
Authority
WO
WIPO (PCT)
Prior art keywords
patient
rehabilitation
data
recovery
biomarkers
Prior art date
Application number
PCT/CA2023/051284
Other languages
English (en)
Inventor
Christopher FRIESEN
Tony Joseph Gerald INGRAM
Michael Lawrence
Original Assignee
Axem Neurotechnology Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Axem Neurotechnology Inc. filed Critical Axem Neurotechnology Inc.
Publication of WO2024065052A1 publication Critical patent/WO2024065052A1/fr

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising

Definitions

  • This disclosure relates generally to functional neuroimaging biomarkers.
  • the disclosure relates to functional neuroimaging biomarkers of brain injury recovery, the use of such biomarkers in methods to predict brain injury recovery and methods for personalized brain rehabilitation.
  • Brain activity metrics may be used to predict recovery, track progress, and compare the effects of different exercises, potentially allowing clinicians to better tailor therapy to individual patients. See for example US8380314B2 which describes a system whereby brain activity is used to dictate what treatments are provided to a patient.
  • Electroencephalography which measures electrical activity. See for example US9532748 which teaches portable systems for brain activity recording, storage, analysis and neurofeedback.
  • Near infrared spectroscopy which measures relative changes in oxygen concentration in the brain.
  • Brain activity requires oxygen to use energy, which is known as the hemodynamic response and is the basis for many brain imaging technologies.
  • the concentration of oxygen will increase in the right motor cortex in the area responsible for initiating motor commands for that hand. The more muscle recruitment and the more complex the movement, the greater the oxygen change. See for example, W02020006647A1 (incorporated herein by reference) which teaches a method and apparatus for monitoring brain activity of a user the apparatus includes a plurality of spatially separated emitters operable to generate infrared radiation.
  • An object of the present invention is to provide functional neuroimaging biomarkers and uses thereof.
  • a method of adaptive data collection for extraction of biomarkers of post stroke hemiparesis and/or hemiplegia recovery comprising: a. providing a rehabilitation program to one or more patients recovering from a stroke and having hemiparesis and/or hemiplegia; b. collecting data related to one or more functional neuroimaging biomarkers from said one or more patients and optionally one or more metrics of recovery based on predetermined collection parameters before, during and/or after said rehabilitation program; and c. determining if collected data is sufficient for extraction of biomarkers, if insufficient, optionally automatically, adjusting data collection parameters and collecting data related to one or more biomarkers from said one or more patients based on adjusted data collection parameters.
  • a method of identifying functional neuroimaging biomarkers of post stroke hemiparesis and/or hemiplegia recovery comprising: a. providing a rehabilitation program comprising one or more rehabilitation exercises and/or activities to one or more patients recovering from a stroke and having hemiparesis and/or hemiplegia; b. collecting data related to one or more functional neuroimaging biomarkers from said one or more patients based on predetermined collection parameters before, during and/or after said rehabilitation program; c.
  • a method of predicting a stroke patient’s movement recovery comprising: a.
  • a movement prediction engine comprising one or more probabilistic models; and b. generating a prediction set for one or more metrics of recovery for a given time or times.
  • a method of modifying a rehabilitation plan for a stroke patient hemiparesis and/or hemiplegia comprising: a. providing a rehabilitation program comprising one or more rehabilitation exercises and/or activities to one or more patients recovering from a stroke and having hemiparesis and/or hemiplegia; b. collecting data related to one or more functional neuroimaging biomarkers from said one or more patients based on predetermined collection parameters before, during and/or after said rehabilitation program; c.
  • a method of modifying a rehabilitation plan for a stroke patient having hemiparesis and/or hemiplegia comprising: a. providing a rehabilitation program comprising one or more rehabilitation exercises and/or activities to one or more patients recovering from a stroke and having hemiparesis and/or hemiplegia; b. collecting data related to one or more functional neuroimaging biomarkers from said one or more patients based on predetermined collection parameters before, during and/or after said rehabilitation program; c. generating a prediction set for one or more metrics of recovery for a given time, d. modifying said patient’s prescribed rehabilitation plan based at least in part on the prediction set generated and pre-defined rules.
  • Figure 1 provides correlation plots between functional near-infrared spectroscopy (fNIRS) and upper extremity movement ability.
  • fNIRS functional near-infrared spectroscopy
  • Figure 2 provides a graphical presentation of the correlation of Latent fNIRS, Latent Fugal Meyer (FM) and Latent Stroke Impact Scale (SIS_ with Latent Function.
  • the present invention provides functional neuroimaging biomarkers of brain injury recovery, methods of identifying such biomarkers and the use of such biomarkers in methods to predict brain injury recovery and methods for personalized brain rehabilitation.
  • brain injury may include brain injury resulting from strokes including but not limited to ischemic and hemorrhagic strokes and/or traumatic brain injuries.
  • the brain injury is a stroke and the patient is suffering from post-stroke hemiparesis and/or hemiplegia.
  • a biomarker is “a defined characteristic that is measured as an indicator of normal biological processes, pathogenic processes, or biological responses to an exposure or intervention, including therapeutic interventions.” (FDA-NIH Biomarker Working Group. BEST (Biomarkers, Endpoints, and other Tools) Resource. (Silver Spring, MD) (2016)).
  • the one or more biomarkers are functional neuroimaging biomarkers. In certain embodiments, the one or more functional neuroimaging biomarkers are patterns of brain activity. In specific embodiments, the brain activity is brain activity in the primary motor cortex.
  • the present invention provides methods of data collection for the extraction of biomarkers such as functional neuroimaging biomarkers relevant to recovery post-brain injury, including but not limited to recovery from post-stroke hemiparesis and/or hemiplegia.
  • the method comprises collecting data related to one or more biomarkers from one or more patients post-brain injury, including but not limited to post-stroke patients having hemiparesis and/or hemiplegia and optionally one or more metrics of recovery based on predetermined collection parameters before, during and/or after a rehabilitation program.
  • the rehabilitation program may include a health care provider (HCP)-led (either in-person or remote, e.g. telephone or online) rehabilitation sessions, independent rehabilitation sessions (see for example WO2022/036447, herein incorporated by reference) or a combination thereof.
  • HCP health care provider
  • Data related to functional neuroimaging biomarkers may be measured using a variety of monitoring devices.
  • Exemplary non-invasive monitoring devices include but are not limited to EEG-based brain activity monitoring devices, near infrared spectroscopy-(NIRS)-based devices and MRI devices.
  • the monitoring device is a NIRS-based device.
  • the patient wears a monitoring device to collect the data related to the functional neuroimaging biomarkers.
  • the patient may wear/use the device in one or more locations including but not limited to the hospital, the clinic and the home.
  • the neuroimaging biomarkers are collected in multiple locations (e.g., at the hospital, then later at the home and during outpatient visits).
  • the monitoring device measures and transmits a signal which includes data on brain activity to a computing device.
  • the computing device is a local computing device such as a tablet, computer or smart phone capable of video display.
  • the video display may be used to (1) guide patients to move in accordance with any experimental procedures and/or rehabilitation exercises, and/or (2) enable video conference in support of synchronous remote-rehabilitation (such as tele-rehabilitation) sessions (3) and/or other remote rehabilitation and/or remote-experimental (tele-expreimental) procedures.
  • Data for the one or more biomarkers may be collected while the patient is at rest, while performing one or more activities/exercises of a rehabilitation program and/or while performing a testing procedure.
  • the testing procedure may be a standardized movement assessments common to the field, or a bespoke testing procedure.
  • data is collected while the patient is performing rehabilitation exercises.
  • An exemplary non-limiting method of collecting data is described in WO2022/036447 (incorporated by reference).
  • Rehabilitation exercises may be performed in a temporal sequence resembling a block-designed experiment. Alternately rehabilitation exercises might be performed in a more naturalistic way (with the patient moving with less guidance as to how they ought to move from moment to moment).
  • data is collected while the patient is performing standardized movement assessments.
  • standardized testing procedures commonly used in the stroke rehabilitation field to quantify various aspects of function and/or impairment related to recovery from hemiparesis and/or hemiplegia (herein known as movement assessments, or MA).
  • data is collected via a bespoke testing procedure.
  • the bespoke testing procedure may include but is not limited to: a movement experiment (where the patient is asked to repetitively repeat a given movement), a resting experiment (where the participant is simply asked to rest), and/or an experiment wherein the patient is subject to tactile stimulation.
  • the testing procedure may take place either in-person, with someone facilitating the experiment (i.e., an ‘experimenter’, including but not limited to a Health care provider (HCP) or HCP-support-staff), or remotely.
  • Remote testing procedures may include a patient conducting the procedure independently, and/or guided via video conference by a HCP or HCP-support-staff (i.e., a “teleexperiment”).
  • motion capture may be used to synchronize the movements of the patient with the data being collected.
  • Methods of motion capture include but are not limited to image filtering (e.g., Gaussian), Optical flow, Fisher kernels, Harris corner detector, Random sample consensus, Convolutional neural networks.
  • the video for use in motion capture may be obtained with a device that interfaces with the brain imaging device (ie, the tablet or phone the patient is using the app associated with the brain imaging device on) and/or via sensor fusion with, or derived exclusively from additional peripheral motion-capturing devices such as a body-worn inertial movement unit, depth-sensing camera systems, or any other related technologies.
  • the methods of the present invention optionally further comprise collection of one or more metrics of recovery (MORs).
  • MORs are collected using standardized movement assessments. Exemplary standardized movement assessments are known in the art and include but are not limited to Fugl-Meyer, Wolf Motor Function Test, Action Research Arm Test). MORs may also be collected using other movement-ability-based testing procedures such as those requested by the patient’s HCP.
  • the procedures for collection of MORs may be conducted in the presence of the patient’s HCP or support staff (either in-person or via tele-conference) or else independently, being guided by software (in a preferred embodiment via a tablet application employing exemplar videos to guide the patient). Collection of the MORs may be at the same time or separate from the collection of the data related to functional neuroimaging biomarkers.
  • movement assessment based metrics of recovery are scores from standardized movement assessments (including but not limited to Fugl-Meyer) — either as recorded by a HCP or support staff and/or via motion capture.
  • the metrics of recovery are rehabilitation exercise-based metrics of recovery (Rehab-MORs).
  • the MORs may be derived from motion capture data captured while the patient is performing rehabilitation exercises.
  • the rehabilitation exercises may be performed in a temporal sequence resembling a block-designed experiment. Alternately rehabilitation exercises might be performed in a more naturalistic way (with the patient moving with less guidance as to how they ought to move from moment to moment).
  • automated computer-vision and/or motion capture technology-based technique is used to score standardized movement assessments.
  • personalized algorithms e.g., optical flow
  • personalized algorithms which are uniquely tuned to judge an individual patient’s performance on a specific assessment procedure are used to generate an accurate assessment of the patient’s performance on a given sub-item of a standardized movement assessment).
  • the output is a probability density function expressing the relative probabilities of scores across the spectrum of possible grades pertinent to that scoring item.
  • Novel MORs based on MC data may include (but are not limited to) measures (i.e., discrete and/or assessing the variability over time) or derivations of: range of motion, movement smoothness, movement speed and/or movement economy/parsimony.
  • the metrics of recovery are based on/derived from self-report assessments (including those conducted via a structured interview-based forms).
  • Self report assessments may be assessments well known by the stroke rehabilitation field, or custom assessments. Such assessments focus on ascertaining patients’ perceptions of their movement abilities and/or their ability to live independently and/or participate in their desired activities of daily living (e.g., stroke impact scale (i.e., measures of disability)).
  • biomarker data may be collected at the same time as the neurofunctional biomarker data and/or metrics of recovery is collected or independent from collection of the neurofunctional biomarker data and/or metrics of recovery.
  • biomarker data may be collected from additional or other biosensors worn during any of the activities outlined above and/or biomarker data taken via procedures outside the scope of the activities specified above (e.g., a patient’s CT scan that was taken upon admission for stroke may be used).
  • biosensors that may be worn during the activities outlined above may include but are not limited to: inertial measurement units (potentially contained within a wrist-mounted device such as a smart watch including but not limited to an Apple® watch); electrocardiogram (potentially contained within a wrist-mounted device such as a smart watch); electrodermal activity (e.g., galvanic skin response; potentially acquired within a wrist-mounted device).
  • inertial measurement units potentially contained within a wrist-mounted device such as a smart watch including but not limited to an Apple® watch
  • electrocardiogram potentially contained within a wrist-mounted device such as a smart watch
  • electrodermal activity e.g., galvanic skin response; potentially acquired within a wrist-mounted device.
  • Examples of procedures that may be undertaken to extract additional biomarker data aside from those described above may include but are not limited to: structural magnetic resonance imaging; diffusion magnetic resonance imaging and magnetoencephalography
  • the present invention also provides a method of adaptive data collection for extraction of biomarkers, the method comprising determining if previously collected data is sufficient for extraction of biomarkers, if insufficient, adjusting, optionally automatically, data collection parameters and collecting data related to one or more biomarkers from the one or more patients based on adjusted data collection parameters.
  • the system/method periodically makes determinations as to whether the patient is providing adequate brain activity data for the purposes of biomarker extraction.
  • the method/system sends a notification that insufficient data is being collected and requests that the patient to perform more exercises or other testing procedures.
  • the notification may further include details with respect to the amount and type of exercises required.
  • the requirements for biomarker-extraction is adaptive based on uncertainty estimates generated via Bayesian modelling — i.e., the requirement of the system for data at any particular time with any particular patient might change according to the situation, based on the individual patient’s historical data, as well as the data that is currently being collected (e.g., if a patient’s data deviated dramatically from their past data, the method/system would request more data because the uncertainty associated with any conclusions that might be drawn from the data would be too high).
  • the system/method makes a determination whether/how frequently functional assessments are required.
  • the system/method of the present invention cross-references a patient’s brain activity data with their movement assessment data, for the purposes of better identifying bad data and/or identifying how to optimize the formulation of a particular biomarker for a particular patient (e.g., which subset of brain activity measurement channels to use).
  • the assessment schedule of the system/method is fully automated.
  • the system /method may automatically send a notification when the next assessment is required.
  • the present invention provides methods of predicting recovery from brain injury. Accordingly, in certain embodiments, the present invention provides a movement recovery prediction engine. In such embodiments, the data collected by the methods of the present invention are input in the movement recovery engine.
  • the data collected is accessible via a single dataset.
  • Details informing various parameters of biomarker and/or MOR definition, as well as predictive model specification and testing rely on norms derived from a combination of data from the system’s dataset, information derived from other open-access publicly accessible datasets, and/or domain specific know-how.
  • raw functional neuroimaging data e.g., from functional near-infrared spectroscopy, electroencephalography, or any other modality
  • the system will first be pre- processed in order to attenuate noise and allow for a better isolation of the signal (i.e., brain activity). See for example, W02020006647A1 incorporated by reference.
  • the ratio between the magnitude of brain activity from either hemisphere’s cortical motor system during paretic arm movement may be utilized as a functional neuroimaging biomarker.
  • the system may extract such metrics from fNIRS data, utilizing the relative increase in oxyhemoglobin as a proxy for brain activity.
  • the functional neuroimaging device used may have >1 measurement locations located within either cortical motor system; in such an embodiment, functional neuroimaging biomarker metrics of motor cortex laterality might be based on an average of all these measurement locations (i.e., a ratio of the average level of brain activity in one hemisphere to the average in the other), and/or extracted using only the measurement location which demonstrates the greatest amount of brain activity during the period during which the patient is performing movement, as well as its homotopic pair in the other hemisphere.
  • metrics reflecting the connectivity between (i.e., inter-hemispheric connectivity) and/or within (i.e., intra-hemispheric connectivity) the hemisphere’s cortical motor system, at rest and/or during paretic arm movement may be utilized as a functional neuroimaging biomarker.
  • Such metrics of connectivity may be based on the correlation between brain activity at different measurement locations (i.e., functional connectivity) and/or the casual interdependence between brain activity at different measurement locations (i.e., effective connectivity).
  • the functional neuroimaging device used may have >1 measurement locations located within either cortical motor system.
  • functional neuroimaging biomarker metrics of motor cortex connectivity might be based on an average of all these measurement locations, and/or utilize an algorithm to determine the optimal subset of measurement locations (e.g., if measuring connectivity during movement, the locations containing the greatest magnitude of brain activity might be utilized; if measuring at rest, the measurement locations with the highest signal-to-noise ratio (or above a pre-determined level of signal-to-noise) may be used).
  • the average magnitude of motor cortex activity may be utilized as an functional neuroimaging biomarker.
  • Such metrics may sub-segment between ipsi- and contra- lesional motor cortex brain activity — showing them as separate metrics.
  • such functional neuroimaging biomarkers may be presented subsegmented by rehabilitation exercise — thus providing a functional neuroimaging biomarker metric associated with each rehabilitation exercise the patient has completed as part of their program.
  • the functional neuroimaging device used may have >1 measurement locations located within either cortical motor system.
  • functional neuroimaging biomarker metrics of average motor cortex activity might be based on an average of all these measurement locations, and/or utilize an algorithm to determine the optimal subset of measurement locations.
  • the cumulative amount of motor cortex activity evoked throughout a given period of time may be utilized as a functional neuroimaging biomarker.
  • Such metrics may sub-segment between ipsi- and contra- lesional motor cortex brain activity — showing them as separate metrics.
  • Such metrics may be sub-segmented between different rehabilitation exercises, such that the cumulative amount of brain activity
  • biomarker forms including but not limited to metrics of laterality and/or connectivity from other areas of the brain and the level of activity at a particular location e.g., the degree of activity at the ipsi-lesional motor cortex alone may be used.
  • the system may combine brain-activity data collected (and/or metrics derived therefrom) with data from other biosensors used to collect physiological data from the patient (either during rehabilitation or during their daily life (e.g., via a smart watch such as an Apple Watch)), and/or transform data from other biosensors exclusively into novel processed data inputs to be provided to the movement recovery engine.
  • the system may extract data from patient’s electronic health record (EHR), including but not limited to: demographic data (age, sex, etc.), pertinent medical history details extracted from their electronic health record (e.g., past disease history including but not limited to stroke details extracted from electronic medical record (lesion side, ischemic vs. hemorrhagic).
  • EHR electronic health record
  • demographic data age, sex, etc.
  • pertinent medical history details extracted from their electronic health record e.g., past disease history including but not limited to stroke details extracted from electronic medical record (lesion side, ischemic vs. hemorrhagic).
  • scores from standardized movement assessments (or the probability density functions on sub-items based on an assessment of an individual’s ability to move relative to cues defined by the assessment) are used, these data may be analyzed via Bayesian models (that may or may not have hierarchical properties), whereby all sub-items within any assessment used are themselves modeled as probabilistic constructs (e.g., ordinal constructs in the case of items within the Fugl-Meyer).
  • motion capture is used to generate MAs, it may do so based either on data from the patient performing functional movement assessments, and/or from the patient performing rehabilitation exercises during the course of their rehabilitation program. Moreover, if > 1 standardized assessment and/or metric of recovery derived from direct movement data is used, these data sources may be combined within a single model that seeks to model movement recovery as a latent factor (perhaps though not necessarily via structural equation modelling (Bayesian or maximum-likelihood-based)).
  • the present invention further provides methods for generating a probabilistic model of patients’ movement recovery trajectories. Accordingly, in certain embodiments, the present invention provides a method of predicting a stroke patient’s movement recovery, the method comprising: a. entering data related to one or more functional neuroimaging biomarkers and optionally one or more metrics of post stoke hemiparesis and/or hemiplegia recovery into a movement prediction engine comprising one or more probabilistic models; and b. generating a prediction set for one or more metrics of recovery for a given time or times.
  • a prediction set refers to a weighted set of predictions made by the system’s generalized prediction engine for one or more MORs, for a given time or times (i.e., one or more prospective time points or ranges of timepoints), for one or more sets of assumptions about what the future of the patient’s rehabilitation program will entail (henceforward referred to as an “assumption set”).
  • the prediction set’s “predictions” are probability density functions (PDF) for each combination of MOR/time-point/assumption-set the prediction set definition specifies.
  • PDF probability density functions
  • these assumption sets can then be modified.
  • the assumption sets can be modified by the HCP (e.g., the patient will continue completing the same volume of exercises they have been to date; or the patient will increase or decrease their volume of rehabilitation exercises by 50%) in order to best predict what the optimal course of action might be for the patient.
  • the possibility space of all potential assumption sets might be explored in an automated way
  • a prediction set may relate to a single MOR or multiple MORs.
  • a prediction set can request a prediction for a single future time-point, or multiple future timepoints.
  • a prediction set may include a single assumption set or multiple assumption sets.
  • a non-limiting example of a single MOR, time point, and a single assumption set is predicting Fugl-Meyer score for a patient 8 weeks from present, assuming the currently prescribed rehabilitation program details.
  • a non-limiting example of multiple MORs, with a single time point and assumption set is predicting Fugl-Meyer and ARAT scores for a patient at the end of each week from now till 8 weeks from present, assuming the currently prescribed rehabilitation program details.
  • a non-limiting example of a single MOR, multiple time points, and a single assumption set is predicting Fugl-Meyer score for a patient at the end of each week from now till 8 weeks from present, assuming the currently prescribed rehabilitation program details.
  • a non-limiting example of a single MOR, a single time point, and multiple assumption sets is predicting Fugl-Meyer score for a patient 8 weeks from present, assuming (1) the currently prescribed rehabilitation program details as well as (2) a rehabilitation program where all currently-prescribed exercises have their volume prescriptions increased by 15%.
  • prediction sets can have single or multiple MORs, single or multiple timepoints, and single or multiple assumption sets.
  • a non-limiting example of multiple MORs/time points/assumption sets is predicting Fugl-Meyer and ARAT scores for a patient at the end of each week from now till 8 weeks from present, assuming (1) the currently prescribed rehabilitation program details as well as (2) a rehabilitation program where all currently-prescribed exercises have their volume prescriptions increased by 15%.
  • the methods of the present invention may generate prediction sets in response to user input or automatically.
  • a health care provider can manually request a prediction set.
  • the system in an automated fashion may periodically (optionally at a frequency approved by the HOP; this frequency may be periodic or dynamic, adapting to the results of the resulting prediction set requests) generate prediction sets for various MORs (either all MORs or a subset defined by the HOP) using that patient’s present default assumptions (i.e., the currently prescribed rehabilitation plan, conditioned on the patient’s compliance/performance of their plan to date), as well as for prediction sets for a subset of all possible assumptions (henceforward known as a patient’s ‘assumption exploration space’).
  • a patient’s Assumption Exploration Space may be defined by the HOP directly, or else the HOP may approve the use of an automated method that conditions the possibility space of all potential assumptions on the patient’s historical data (both of all patients and of the patient of interest).
  • the method might automatically constrain the possibility space for assumptions regarding rehabilitation volume, to ensure that unrealistic assumptions (e.g., large amounts of rehab that the patient would not be capable of completing) would not be explored.
  • the results of an automated prediction may be saved to the patient profile, and/or trigger further actions such as a notification being sent to the health care provider or patient, and/or a change to the patient’s rehabilitation program.
  • an HCP could specify that, if the prediction set for a particular assumption set were to generate a prediction for a particular MOR that was a pre-specified level of improvement over the prediction generated by the default assumption case (with ‘default assumption case’ here meaning the rehabilitation program the patient is currently projected to be carrying out), a triggering event (such as the HCP being notified, and/or a change to the patient’s rehabilitation program) could occur (e.g. if a prediction set predicts a Fugl-Meyer upper-extremity score >8 points greater than that of the default assumption case.)
  • the system will enter all its data inputs collected from the patient of interest (data collected from the rehabilitation system, both about the details of their rehabilitation, the data collected during rehabilitation as well as during standardized assessments and/or other testing procedures, as well as from their EHR) into a movement recovery prediction engine.
  • the movement recovery prediction engine contains probabilistic models
  • historical data from all relevant prior prediction set requests (both from this particular patient, as well as all other patients) accessible will be used to condition any priors of said models.
  • prediction sets are generated via the application of a probabilistic model to all data inputs, resulting in an estimate (in the form of a probability density function) on the MOR(s) requested.
  • the probabilistic model used may be comprised of several sub-models whose composition and/or weighting towards the final prediction set is determined by a model specification/selection and weighting processes that take place within the single predictive model. Otherwise, said predictive model may be of a singular form (i.e., may not contain multiple sub-models within it that the model selects between, evaluates, weights etc.).
  • the predictive sub-model subset consists of all predictive model types (i.e., with one model from each model type contained within the overall model being included) whose outputs will be passed forward to a weighting procedure.
  • the inclusion of different predictive sub- model types may be determined by an automated process of model evaluation, or a manual or pseudo-manual iterative process of model evaluation.
  • the generation of the predictive sub-model subset requires a determination of the optimal model permutation for all potential sub-model types (i.e., all types of predictive sub-models utilized by the system that have their data input requirements met for this given prediction set request).
  • the optimal parameter/hyper-parameter settings for a given model type may be determined via a process whereby all (or a subset of all potential parameters/hyper-parameters, with the limiting of this space being either automated or performed in advance through manually defined decision rules based on past iteration on model composition) possible permutations of that model type are run, whereupon the permutation of that model type which generates the most accurate prediction in response will be selected for inclusion in the predictive sub-model subset.
  • the system includes probabilistic sub-model types utilizing simulationbased inference; wherein this step may involve the choice of priors and/or posterior sampling used.
  • the system includes one or more model types based on structural equation modelling (potentially though not necessarily implemented within a probabilistic model inference), wherein this step may involve varying model structures.
  • the system includes one or more sub-model types based on a particular type of machine learning, wherein this step may involve various hyper-parameter-selection algorithms (e.g, grid search, random search), and/or varying optimization methods.
  • hyper-parameter-selection algorithms e.g, grid search, random search
  • Such methods may include but are not limited to: Supervised deep learning via techniques including but not limited to convolutional neural networks, recurrent neural networks; Unsupervised deep learning via techniques including but not limited to Boltzmann machines and autoencoders; Partially supervised reinforcement learning; Deep reinforcement learning; and a deep learning technique that outputs estimates on a probability distribution (e.g., mixture density networks).
  • the output of the process of predictive subset generation is the definition of a set of fully specified models (i.e., the predictive sub-model subset).
  • the prediction set is a weighted combination of all probability density functions (PDF) generated by all models contained in the predictive sub-model subset.
  • PDF probability density functions
  • the weighting of a particular PDF may be determined via a model evaluation process.
  • Theoretically model weighting may range from 0 (contributing nothing towards the final prediction set) 1 equivalent to the final prediction set (i.e., determining it completely). In certain embodiments, this will be a relative weighting — models will be rated WRT their accuracy given the formulation of this particular prediction set request, relative to all other models being utilized.
  • This weighting may be linear (i.e., a linear combination of all PDFs with weightings determined by accuracy in a linear manner), nonlinear, binary (i.e., either the model is included or not, and all included models are weighted equally), or winner-take-all (only the top-performing model is included).
  • Accuracy in this case refers to an estimate of how accurately these models perform when asked to make informative test predictions using subsets of previously-collected data from the system’s database. How ‘informative’ any given test prediction is to the current prediction set request will be unique (i.e., if test prediction X and Y are identical, except test prediction X involves data from subject X, and test prediction Y from subject Y, then subsequent test prediction X will be more informative for subsequent prediction set requests involving subject X than for prediction set requests for subject Y), and thus this process will be unique for each prediction set request.
  • the process of determining which (and how many) test predictions to run may itself be defined by an adaptive algorithm (in certain embodiments such a model might seek to optimize various factors, including but not limited to: variance in test prediction accuracy), or else defined via static decision rules.
  • Model evaluation may include the use of many techniques, for example it may include (but is not limited to) the use of cross validation (e.g., leave one out or leave some out estimation) or information criteria (e.g., widely applicable information criteria, Akaike Information Criteria, or Watanabe-Akaike Information Criteria);
  • cross validation e.g., leave one out or leave some out estimation
  • information criteria e.g., widely applicable information criteria, Akaike Information Criteria, or Watanabe-Akaike Information Criteria
  • the output of the sub-model evaluation process is a weighted ranking of models (according to their performance across all evaluation techniques used), used to determine the relative influence of these model’s PDFs towards the generation of the final prediction set.
  • the system may include a variety of options to allow HCPs (1) to determine whether they are to be notified of the results of any given prediction set; (2) to determine if any automatic triggering action is supposed to take place as a result of any given prediction set (as well as their notification preferences in response to any triggering action); (3) customize the manner in which they visually explore the results of prediction sets.
  • HCPs (1) to determine whether they are to be notified of the results of any given prediction set; (2) to determine if any automatic triggering action is supposed to take place as a result of any given prediction set (as well as their notification preferences in response to any triggering action); (3) customize the manner in which they visually explore the results of prediction sets.
  • the engine will define a PDF to represent its predictions.
  • HCPs can view this PDF itself, or request to only see summary metrics derived from it (e.g., they may simply want to know the score that holds the largest probability weight, which may be 18.5, and thus they would simply see the prediction “18.5”. Or they may want to know the range of values for a particular score that subsume 90% of the PDF - maybe that is scores ranging from 15.5-21.2).
  • a given prediction set becomes retrospective (i.e., once the patient profile contains data past the time point or time range specified by the prediction set) it will automatically be re-analyzed with the goal of improving the prediction abilities of the generalized prediction engine.
  • This procedure may include but may not be limited to, the optimization of the prediction sets’ original predictive models (including model priors and pre-processing specifications), its priors/pre-processing specifications; in the case where this reanalysis reveals insights that might improve model performance in the case of this prediction set (i.e., that original formulation of the predictive models underlying the final prediction set was not optimal), these insights may be used to update the generalized prediction engine’s default operations.
  • all or else a subset of past prediction sets may also be re-analyzed.
  • the subset to be re-analyzed is determined based on a metric of how ‘informative’ a given newly-retrospective prediction is projected to be for a given retrospective prediction set (with the operation of ‘projecting’ in this way also, in certain embodiments, being subject to an automatic adaptive algorithm, and/or a manual or pseudo-manual iteration process).
  • data related to the performance of various sub-models and permutation details therein may be provided to a model performance dataset (to be contained within the overall system dataset.
  • the model performance dataset may be structured in a manner in which it can provide insight on the strength/weakness of various aspects of model structure for all types of sub-models utilized by the system.
  • the model performance dataset may have automated analyses performed on it according to a static or adaptive schedule, and/or it may be analyzed manually in order to explore and verify conclusions regarding what types of model structure have been found to perform better/worse given the system’s dataset and history of all past prediction set request. Updates to the model performance dataset may trigger an action such as: An automated alteration to the sub-model accuracy estimation processes or a notification sent to individuals responsible for manually updating said system.
  • the present invention further provides a method of modifying a patient’s rehabilitation program based on the data collected and/or generated predictions.
  • a method of modifying a rehabilitation plan comprising: a. providing a rehabilitation program comprising one or more rehabilitation exercises and/or activities to one or more patients recovering from a brain injury including but not limited to stroke; b. collecting data related to one or more functional neuroimaging biomarkers from the one or more patients based on predetermined collection parameters before, during and/or after said rehabilitation program; c. collecting data related to one or more metrics of recovery from said one or more patients based on based on predetermined collection parameters; and d. modifying the patient’s prescribed rehabilitation plan based at least in part on data collected in part b and/or c and predefined rules and/or a manual or pseudo-manual process involving the patient’s HCP.
  • a method of modifying a rehabilitation plan comprising a. providing a rehabilitation program comprising one or more rehabilitation exercises and/or activities to one or more patients recovering from a brain injury, including but not limited to stroke; b. collecting data related to one or more functional neuroimaging biomarkers from said one or more patients based on predetermined collection parameters before, during and/or after said rehabilitation program; c. generating a prediction set for one or more metrics of recovery for a given time; d. modifying said patient’s prescribed rehabilitation plan based at least in part on the prediction set generated and pre-defined rules.
  • the initial rehabilitation program is based on standard of care guidelines.
  • AHA/ASA 2016 Stroke Rehab & Recovery Guidelines https://pubmed.ncbi.nlm.nih.gov/27145936/; Canadian Stroke Best Practice Guidelines: https://pubmed.ncbi.nlm.nih.gov/31983296/ & https://pubmed.ncbi.nlm.nih.gov/31983292/.
  • the initial rehabilitation program is based on standard of care guidelines.
  • the rehabilitation program is developed based on data collected for the patient including but not limited to brain-activity based or other biomarkers extracted from an initial baseline data collection, data from movement assessments; medical co-morbidities and complications; patient-specific goals and values; prediction sets available at the outset of the patient’s program and/or prediction sets (generated either via manual requests or through an automated process) based on the data provided to the system about the patient prior to the commencement of their rehabilitation via the system may be utilized to make decisions about the details of the patient’s prescription.
  • data collected for the patient including but not limited to brain-activity based or other biomarkers extracted from an initial baseline data collection, data from movement assessments; medical co-morbidities and complications; patient-specific goals and values; prediction sets available at the outset of the patient’s program and/or prediction sets (generated either via manual requests or through an automated process) based on the data provided to the system about the patient prior to the commencement of their rehabilitation via the system may be utilized to make decisions about the details of the patient’s prescription.
  • the present invention further comprises a recovery monitoring scheme for the patient.
  • the recovery monitoring scheme may comprise (1) a list of all aspects of the patient’s rehabilitation program to-be-monitored (2) the details specifying the nature of this monitoring, and (3) decision rules determining what actions should be taken based on the monitoring process.
  • the monitoring scheme may be automatically generated or defined by the HCP.
  • the monitoring may be monitoring of rehabilitation program details, MOR or functional neuroimaging biomarkers.
  • the monitoring includes but is not limited to:
  • the monitoring includes but is not limited to:
  • Specifying the recovery monitoring scheme for a patient may result in a customized data display for the HCP user, such that when they access a patient’s historical view, the data visualizations displayed by default will be related to the details of the recovery monitoring scheme the HCP chose for that patient.
  • the method further comprises an initiation of an action, based on predefined decision rules and the results of the monitoring.
  • Actions may include but are not limited to (1) notifying the HCP, (2) notifying the HCP and suggesting a change to the details of the patient’s rehabilitation program, or (3) notifying the HCP that a change to the details of a patient’s rehabilitation program. Examples of this might be (though are not limited to):
  • the decision rules may be simple heuristics (e.g., if they have X score or lower on the upperextremity Fugl-Meyer, assign a pre-defined set of general upper-extremity exercises (and a similar rule for lower-extremity)), or else could be more complicated, involving an automated evaluation of all (or a subset of) relevant rehabilitation program details, MORs and/or biomarkers.
  • the system may present the HCP with suggested decision frameworks - i.e., combinations of metrics-to-be-monitored and actions-to-be-taken.
  • the HCPs select and confirm the decision framework before their implementation
  • a decision framework may be preloaded and then personalized for a particular patient.
  • the present invention provides a method for an HCP to review patients’ progress on an ongoing basis.
  • the HCP uses the system of the present invention (in specific embodiments there is a web portal) to (1) access/review data from their patient on their progress in their rehabilitation program (including but not limited to: patient compliance with their prescribed independent rehabilitation exercises, observed brain activity-based (or other) biomarkers, changes in movement ability, (2) access/review predictions made by the system on the potential future trajectory of their patient with respect to MORs of interest; (3) review and/or makes changes to their patients’ rehabilitation program, and (4) communicate (either synchronously via video conferencing, or asynchronously via text or voice messaging) with the patient.
  • access/review data from their patient on their progress in their rehabilitation program including but not limited to: patient compliance with their prescribed independent rehabilitation exercises, observed brain activity-based (or other) biomarkers, changes in movement ability, (2) access/review predictions made by the system on the potential future trajectory of their patient with respect to MORs of interest; (3) review and/or makes changes to their patients’ rehabilitation program, and (4) communicate (either synchronously via video conferencing, or asynchronous
  • the system of the present invention triggers evaluations of a patient’s progress.
  • the system may send a communication to the HCP.
  • the schedule of evaluations may be based on a pre-determined scheduled (e.g., weekly), based on a pre-determined triggering event, including but not limited to: the patient providing a particular answer to a self-report question or a particular change in the brain-activity biomarker data for a particular session and/or based on the HCP’s decision to do so at any given time.
  • a pre-determined scheduled e.g., weekly
  • a pre-determined triggering event including but not limited to: the patient providing a particular answer to a self-report question or a particular change in the brain-activity biomarker data for a particular session and/or based on the HCP’s decision to do so at any given time.
  • the present invention further provides methods of modifying a rehabilitation plan.
  • the decision to change a patient’s rehabilitation program may be performed manually or automatically.
  • the system may contain decision frameworks for altering a patient’s rehabilitation program in response to their progress; when done manually, these would appear as suggestions to the HCP; when done automatically these would act as decision rules according to which the patient’s rehabilitation program would change).
  • a method of modifying a rehabilitation plan for a patient suffering a brain injury comprising: a. providing a rehabilitation program comprising one or more rehabilitation exercises and/or activities to one or more patients recovering from a brain injury; b. collecting data related to one or more functional neuroimaging biomarkers from said one or more patients based on predetermined collection parameters before, during and/or after said rehabilitation program; c. collecting data related to one or more metrics of post brain injury recovery from said one or more patients based on based on predetermined collection parameters; and d. modifying said patient’s prescribed rehabilitation plan based at least in part on data collected in part b and/or c and predefined rules. In certain embodiments, the modifications are based in part on the patient’s compliance with the rehabilitation plan.
  • a method of modifying a rehabilitation plan for a brain injury patient comprising a. providing a rehabilitation program comprising one or more rehabilitation exercises and/or activities to one or more patients recovering from a brain injury; b. collecting data related to one or more functional neuroimaging biomarkers from said one or more patients based on predetermined collection parameters before, during and/or after said rehabilitation program; c. generating a prediction set for one or more metrics of recovery for a given time; d. modifying said patient’s prescribed rehabilitation plan based at least in part on the prediction set generated and pre-defined rules
  • the example detailed below demonstrates a correlation between fNIRS and upper extremity movement ability.
  • the correlation plots set forth in FIG. 1 are outputs from a Bayesian structural equation model, where input is data from 28 stroke survivors who had behavioural and functional neuroimaging measures taken (prior to an intervention).
  • the behavioural measures (Fugl Meyer Upper Extremity and Stroke Impact Scale Hand and Arm sections) both reflect stroke survivor’s upper-extremity movement abilities.
  • these measures are ‘combined’ in a single construct called “Latent Function” herein.
  • the neuroimaging measures were taken with the functional near-infrared spectroscopy (fNIRS) device (capable of being utilized independently by stroke survivors in their homes) described in W02020006647 (incorporated by reference), and reflect changes in motor cortex activation during paretic upper-extremity movement.
  • fNIRS functional near-infrared spectroscopy
  • Multiple aspects of the fNIRS data were similarly ‘combined’ in the model, most prominently the laterality ratio of activation between the two motor cortices (or motor cortex laterality); this neuroimaging-derived construct is called “Latent fNIRS” herein.
  • Results show a definitive correlation between fNIRS and upper extremity movement ability (95% credible intervals for the correlation between Latent Function and Latent fNIRS do not include zero (meaning the model does not find a non-correlation between these constructs to be credible) (see FIG. 2).
  • HCP develops rehabilitation program.
  • HCP performs weekly re-assessment to gather evidence that rehabilitation is effective.
  • a Positive change in behavioural data.
  • i This is evidence that the program is effective.
  • ii. Continue progressing program (ensuring “just right” challenge) as planned.
  • HCP develops rehabilitation program.
  • Hi Total units of brain activation across each session, and across a period (e.g., a week).
  • HCP uses a combination of this data and their clinical reasoning to adjust exercise program. For example: a. If an exercise resulted in a low average brain activation, consider either assigning a more challenging version of that exercise or replacing the exercise with another that is more challenging. b. If an exercise resulted in a high average brain activation, but the patient did not adhere, the HCP may inquire as to why. It is likely that exercise is too challenging and should be modified to be less challenging or replaced with a less challenging exercise that the patient is able to perform for greater volumes. c.
  • HCP re-assesses the patient program weekly, optimizing for the highest possible units of brain activation per week — and not necessarily a large volume (hours) of rehab.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Molecular Biology (AREA)
  • Veterinary Medicine (AREA)
  • Physics & Mathematics (AREA)
  • Primary Health Care (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Epidemiology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Physiology (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Physical Education & Sports Medicine (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

La présente invention concerne des biomarqueurs de neuro-imagerie fonctionnelle de la récupération après lésion cérébrale, l'utilisation de tels biomarqueurs dans des procédés pour prédire une récupération après lésion cérébrale et des procédés de rééducation cérébrale personnalisée.
PCT/CA2023/051284 2022-09-29 2023-09-28 Biomarqueurs de neuro-imagerie fonctionnelle de la récupération après lésion cérébrale et leurs utilisations WO2024065052A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US202263411240P 2022-09-29 2022-09-29
US63/411,240 2022-09-29

Publications (1)

Publication Number Publication Date
WO2024065052A1 true WO2024065052A1 (fr) 2024-04-04

Family

ID=90475046

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CA2023/051284 WO2024065052A1 (fr) 2022-09-29 2023-09-28 Biomarqueurs de neuro-imagerie fonctionnelle de la récupération après lésion cérébrale et leurs utilisations

Country Status (1)

Country Link
WO (1) WO2024065052A1 (fr)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150018664A1 (en) * 2013-07-12 2015-01-15 Francisco Pereira Assessment of Traumatic Brain Injury
US20160019693A1 (en) * 2014-07-15 2016-01-21 The Brigham And Women's Hospital Systems and methods for generating biomarkers based on multivariate classification of functional imaging and associated data
US20210319867A1 (en) * 2018-09-03 2021-10-14 Nec Corporation Systems and methods for predicting recovery of a patient

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150018664A1 (en) * 2013-07-12 2015-01-15 Francisco Pereira Assessment of Traumatic Brain Injury
US20160019693A1 (en) * 2014-07-15 2016-01-21 The Brigham And Women's Hospital Systems and methods for generating biomarkers based on multivariate classification of functional imaging and associated data
US20210319867A1 (en) * 2018-09-03 2021-10-14 Nec Corporation Systems and methods for predicting recovery of a patient

Similar Documents

Publication Publication Date Title
CN108780663B (zh) 数字个性化医学平台和系统
JP7031953B2 (ja) 膝手術の患者の管理
US10925533B2 (en) Systems and methods for determining human performance capacity and utility of a person-utilized device
CN104717921B (zh) 自学性认知训练装置及其方法
Kim et al. The use of wearable sensors to assess and treat the upper extremity after stroke: a scoping review
AU2016278356A1 (en) Method and system for assessing mental state
US20170329905A1 (en) Life-Long Physiology Model for the Holistic Management of Health of Individuals
JP7311637B2 (ja) 認知トレーニング及び監視のためのシステム及び方法
EP3960071A1 (fr) Systèmes et dispositifs d'évaluation de la qualité d'intervention sur le sommeil
US20210204853A1 (en) Systems and Methods for Cognitive Diagnostics in Connection with Major Depressive Disorder and Response to Antidepressants
US20200114115A1 (en) Cognitive platform for deriving effort metric for optimizing cognitive treatment
EP3958275A1 (fr) Systèmes, procédés et dispositifs pour l'estimation de qualité d'intervention durant le sommeil
TW202333005A (zh) 用於經電腦化之認知訓練程式之自適應組態之系統及方法
CN117059274A (zh) 一种脑卒中康复训练系统
CA3157380A1 (fr) Systemes et methodes de traitement de troubles neurologiques : maladie de parkinson et depression comorbide
CN117133404A (zh) 一种针对胸廓出口综合征的智能化康复护理装置
WO2024065052A1 (fr) Biomarqueurs de neuro-imagerie fonctionnelle de la récupération après lésion cérébrale et leurs utilisations
EP4113535A1 (fr) Procédés et systèmes de surveillance à distance pour surveiller des patients souffrant de maladies inflammatoires chroniques
EP4044193A1 (fr) Système de surveillance d'état psychologique
US20220230755A1 (en) Systems and Methods for Cognitive Diagnostics for Neurological Disorders: Parkinson's Disease and Comorbid Depression
US12026636B2 (en) Cognitive platform for deriving effort metric for optimizing cognitive treatment
CN113905663B (zh) 监测注意力缺陷伴多动障碍的诊断和有效性
US20220398477A1 (en) Cognitive platform for deriving effort metric for optimizing cognitive treatment
US11938275B2 (en) Systems, methods, and devices for custom sleep implementation
Liu et al. Improving Mental Health Support in Engineering Education using Machine Learning and Eye-Tracking

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23869376

Country of ref document: EP

Kind code of ref document: A1