MODELOS DE INTELIGÊNCIA ARTIFICIAL (IA) E DE GÊMEOS DIGITAIS PARA PERSONALIZAÇÃO E GERENCIAMENTO DE ASSISTÊNCIA DE RESPIRAÇÃO

Abstract

Methods, devices and systems are described for adjusting airflow provided by a breathing assistance device to a user. In one aspect, personalized predictive models may be determined and used by the breathing assistance device to adjust the airflow to the user to improve various aspects including the user's sleep. In another aspect, models are developed for the user and the breathing assistance device to provide a digital twin to simulate the user and the device and determine if the user may develop any conditions and / or whether the device is operating properly. In another aspect, the personalized predictive model may be used in digital twin simulation results to validate performance before actual deployment.

Description

[001] This application claims the benefit of U.S. Provisional Patent Application 63 / 453,093, filed March 18, 2023. The full description of U.S. Provisional Patent Application 63 / 453,093 is incorporated herein by reference. FIELD

[002] The various modalities described in this document generally refer to systems and methods for generating a customized predictive model for airway pressure management. Several modalities relating to systems and methods for simulating the operation of a respiratory assistance device and / or a user's health status are also described in this document. FUNDAMENTALS OF THE INVENTION

[003] Individuals suffering from acute or chronic respiratory conditions (Chronic Obstructive Pulmonary Disease (COPD), asthma, Acute Respiratory Distress Syndrome (ARDS)) or conditions related to the respiratory system (e.g., sleep apnea) may require assistive devices to maintain respiratory function at normal levels. Devices of Petition 870250101201, dated 05 / 11 / 2025, page 10 / 192 2 / 172 assistive devices, such as mechanical ventilators, Positive Airway Pressure (PAP) devices, or Continuous Positive Airway Pressure (CPAP) devices, are common for providing respiratory assistance. However, while such assistive devices are critical in maintaining normal respiratory function, they can also cause harm and discomfort to a user due to stress or strain from the amount of pressure or flow transmitted to the user's respiratory system. Furthermore, most devices are currently reactive rather than proactive in predicting and preventing respiratory discomfort or distress. As such, there is a need for methods and systems to identify and minimize harm to the user.

[004] It is known in technology that there are several levels of mechanical support for different types of respiratory failure. In the most basic form, the inspired oxygen concentration can be increased to percentages above 21%, which is the normal atmospheric oxygen content. This helps a patient in need to meet the metabolic oxygen requirement for their body.

[005] The next highest level of ventilatory support addresses the problem of when the contents of Petition 870250101201, dated 05 / 11 / 2025, page 11 / 192 3 / 172 oxygen from the inhaled gas mixture is insufficient to maintain the patient's body homeostasis. This means that CO2 retention is also becoming a problem. For these types of respiratory failures, a more invasive method of ventilation, including actively raising airway pressure above atmospheric pressure, is involved to eliminate CO2 as the final product of the body's metabolism. Since this involves a tightly sealed mask, there are limits to the pressure that can be applied to the system.

[006] If the increase in inhaled oxygen and airway pressure facilitated by the mask are no longer sufficient, then so-called mechanical ventilation using a hermetically sealed endotracheal tube and / or tracheostomy along with a ventilator is used for ventilation. The parameters that are controlled with a ventilator include the volume of each breath delivered to the patient, the respiratory rate per minute which, when taken together, allows the patient's ventilation volume to be controlled over certain time intervals, for example, every minute. Furthermore, mechanical ventilation typically also controls the fraction of inhaled oxygen from 21% to 100% in the air and the inspiratory to expiratory ratio of the breathing cycle. If these measures are not sufficient to maintain blood oxygen and CO2 levels within safe physiological limits, then positive end-ventilation pressure (PEP) is used. Petition 870250101201, dated 05 / 11 / 2025, page 12 / 192 4 / 172 expiratory pressure (PEEP) and an inspiratory to expiratory ratio (I / E) are applied. Within the limits that ventilation monitoring is related, there are a variety of methods known in technology that include end-expiratory CO2, inspired CO2, inspired O2, expired O2, blood gas analysis of arterial blood pressure / volume diagrams, and volumetric measurements of inspired and expired ventilation volumes in the patient.

[007] In sleep apnea, the “gold standard” diagnostic test for Obstructive Sleep Apnea (OSA) is polysomnography (PSG), in which respiratory, cardiac, muscular, and neurological parameters are monitored during sleep. Monitoring these various physiological and neurological parameters allows for the assessment of blood oxygen saturation, ventilation pauses, EEG activity to determine the sleep phase, and EMG to determine spontaneous muscle activity. SUMMARY OF VARIOUS MODALITIES

[008] According to a broad aspect of the precepts set forth in this document, in at least one embodiment described in this document, there is a method for generating a customized predictive model to adjust airflow provided by a breathing assistance device for a user, wherein the method comprises: implementing a model Petition 870250101201, dated 05 / 11 / 2025, page 13 / 192 5 / 172 predictive model trained on a processor of a respiratory assistance device controller, the trained predictive model being able to generate, based on sensor data, a real-time estimate of the user's current breathing state and a prediction of the user's future breathing state over a predicted period of time; receive sensor data measured from one or more sensors; operate the processor of the respiratory assistance device controller to apply the trained predictive model to the sensor data to generate the real-time estimate and prediction; identify one or more false negative predictions when the sensor data corresponding to a current period indicate a breathing event and the real-time estimate and / or prediction corresponding to the current period indicate a normal breathing state;Receive false negative data comprising, for each of one or more false negative predictions, the real-time estimate, the prediction, and a portion of the sensor data extending from an initial point in time before the start of the false negative prediction to a second point in time after a shift in the false negative prediction; generate the custom predictive model by retraining the trained predictive model using the false negative data so that the custom predictive model is customized for the user; and implement the custom predictive model in; Petition 870250101201, dated 05 / 11 / 2025, page 14 / 192 6 / 172 processor of the respiratory assistance device controller.

[009] In at least one modality, generating the custom predictive model occurs after a minimum number of false negative predictions have been identified.

[0010] In at least one embodiment, the method further comprises generating simulated false negative data to retrain the trained predictive model when an insufficient number of false negative data occur during use by: applying one or more signal processing techniques to the sensor data, the one or more signal processing techniques comprising: random perturbation, noise addition, magnitude scaling, magnitude distortion, filtering, phase distortion, phase scaling, or segment truncation.

[0011] In at least one embodiment, the minimum number of false negative predictions is in a range from one to a total number of time points at which real-time estimation and prediction are generated during a monitoring time period in which airflow is provided by the breathing assistance device to the user.

[0012] In at least one mode, the generation of the custom predictive model occurs automatically at a predetermined frequency.

[0013] In at least one form, the method includes Petition 870250101201, dated 05 / 11 / 2025, page 15 / 192 7 / 172 additionally determine one or more of a sensitivity, precision, F1 score and / or adjusted F1 score of the trained predictive model, wherein generating the custom predictive model occurs based on one or more of the sensitivity, precision, F1 score and / or adjusted F1 score.

[0014] In at least one modality, the first point in time and the second point in time are within a range of 20 seconds to 60 seconds.

[0015] In at least one embodiment, the method further comprises preprocessing the false negative data using one or more of: normalization, sensor data weighting, principal component analysis, independent component analysis, downsampling, upsampling, frequency filtering, or manual inspection of the sensor data.

[0016] In at least one modality, retraining the trained predictive model includes one or more of: transfer learning, fine-tuning of one or more parameters of the trained predictive model, adding a layer to the trained predictive model, reinforcement learning, or any combination thereof.

[0017] In at least one mode, retraining the trained predictive model additionally uses sensor data.

[0018] In at least one modality, the predictive model Petition 870250101201, dated 05 / 11 / 2025, page 16 / 192 8 / 172 comprises one or more nodes and one or more layers, and the one or more parameters of the trained predictive model include, which must be adjustable, a node type, a node selection, a node weight, a node activation, a node memory, a number of connections between nodes, an orientation of connections between nodes, an orientation of connections between layers, a layer type, a number of layers, a connection between layers, a number of inputs, a number of outputs, or an operable combination thereof.

[0019] In at least one embodiment, operating the processor of the respiratory assistance device controller to apply the trained predictive model comprises: generating the real-time estimate by determining, based on sensor data, a first plurality of probabilities, each of the first plurality of probabilities corresponding to a respective current respiratory state of the user; and generating the prediction by determining, based on sensor data, a second plurality of probabilities, each of the second plurality of probabilities corresponding to a respective predicted future respiratory state of the user.

[0020] In at least one embodiment, the user's current breathing state and the user's predicted future breathing state comprise components that include normal breathing or one or more Petition 870250101201, dated 05 / 11 / 2025, page 17 / 192 9 / 172 respiratory failure events.

[0021] In at least one embodiment, the one or more respiratory failure events comprise components that include obstructive apnea, central apnea, central hypopnea, obstructive hypopnea, respiratory effort-related stimulus, an unclassified event, or any operable combination thereof.

[0022] In at least one modality, one or more respiratory failure events include stimuli related to respiratory effort including flow limitation, wheezing, oxygen desaturation, fragmentation, heart rate abnormality, or any combination thereof.

[0023] In at least one embodiment, the real-time estimate corresponding to the current period indicates the breathing event based on a first comparison of one or more of the first plurality of probabilities with one or more first limits and the prediction corresponding to the current period indicates the normal breathing state based on a second comparison of one or more of the second plurality of probabilities with one or more second limits.

[0024] In at least one mode, one or more first limits and / or one or more second limits are customized for the user and adjustable.

[0025] In at least one modality, one or more first limits and / or one or more second limits are Petition 870250101201, dated 05 / 11 / 2025, page 18 / 192 10 / 172 updated in real time, near real time, hourly, daily, weekly, and / or monthly.

[0026] In at least one embodiment, the one or more sensors comprise user sensors, environmental sensors, or device sensors.

[0027] In at least one modality, the predicted time period falls within a range of 20 seconds to 60 seconds.

[0028] In at least one embodiment, false negative data is received on a remote processor located remote from the respiratory assistance device, the custom predictive model is generated on the remote processor, and the generated custom predictive model is transmitted for implementation by the respiratory assistance device controller processor via a network connection between the respiratory assistance device controller processor and the remote processor.

[0029] In at least one embodiment, the custom predictive model determines one or more of a pressure rise rate, a pressure fall rate, and / or a pressure amplitude to adjust the airflow provided by the breathing assistance device to the user.

[0030] In another aspect, at least one form of controller is provided to control the operation of a Petition 870250101201, dated 05 / 11 / 2025, page 19 / 192 11 / 172 breathing assistance device that provides breathing assistance to a user, wherein the controller comprises: a memory unit comprising software instructions and parameters for at least one trained predictive model, the trained predictive model being able to generate, based on sensor data, a real-time estimate of the user's current breathing status and a prediction of the user's future breathing status over a predicted period of time;and a processor that is electronically coupled to the memory unit, the processor being configured to generate a control signal to control the respiratory assistance device for a current monitoring time period by: receiving sensor data obtained by one or more sensors, the sensor data including measurements of at least one airflow parameter of the user's airflow during the current monitoring time period when the user is wearing the respiratory assistance device; applying the trained predictive model to generate the real-time estimate and forecast; identifying one or more false negative predictions when the sensor data corresponding to the current monitoring time period indicate a breathing event and the real-time estimate and / or forecast corresponding to the current monitoring time period indicate a normal breathing state; extracting data from; Petition 870250101201, dated 05 / 11 / 2025, page 20 / 192 12 / 172 false negative comprising, for each of one or more false negative predictions, the real-time estimate, the prediction, and a portion of the sensor data extending from a first point in time before a false negative prediction begins to a second point in time after a false negative prediction shift; generating a custom predictive model by retraining the trained predictive model using the false negative data, the custom predictive model being customized for the user; and the memory units.

[0031] In at least one embodiment, the controller is additionally configured to perform the method, as previously described.

[0032] In another aspect, a method is provided for generating a customized predictive model to adjust an airflow provided by a breathing assistance device for a user, the method comprising: implementing a predictive model trained on a processor of a breathing assistance device controller, the trained predictive model being able to generate, based on sensor data, a real-time estimate of the user's current breathing state and a prediction of the user's future breathing state over a predicted period of time; receiving sensor data measured from one or more sensors; operating the device controller processor. Petition 870250101201, dated 05 / 11 / 2025, page 21 / 192 13 / 172 of respiratory assistance to apply the trained predictive model to generate real-time estimation and prediction; generate a summary representation of the user, the summary representation comprising user data; generate the custom predictive model by conditioning the trained predictive model using the summary representation, the custom predictive model being customized for the user; and implement the custom predictive model in the processor of the respiratory assistance device controller.

[0033] In at least one embodiment, user data comprises one or more of weight, height, gender, sex, age, body mass index, apnea-hypopnea index, SpO2, type of respiratory assistance device mask, prescribed pressure to be provided by the respiratory assistance device, type of user location, or user location elevation.

[0034] In at least one embodiment, the user data comprises one or more statistical representations of the user's breathing based on sensor data, the one or more statistical representations of the user's breathing comprising an average waveform of a user's breathing, a variance for each time point of the sample in the average waveform, or one or more of a minimum, a maximum, a mean, a median, or a variance of a Petition 870250101201, dated 05 / 11 / 2025, page 22 / 192 14 / 172 plus user airflow, air pressure, tidal volume, respiratory rate, SpO2, heart rate, sound, or movement.

[0035] In at least one embodiment, the user data comprises one or more statistical representations of the user's environment based on sensor data, or one or more statistical representations of the user's environment comprising one or more minimums, maximums, means, medians, or variances of one or more temperature, ambient CO2, or ambient O2.

[0036] In at least one modality, the summary representation is generated by the predictive model trained using one or more embedding layers.

[0037] In at least one embodiment, conditioning the predictive model trained using the summary representation comprises: providing the summary representation as input to the trained predictive model; adapting a feature representation based on cross-attention to the summary representation, the feature representation being based on sensor data; or providing the summary representation as input to a normalization block distinguished by a shift factor and a scale factor for each feature corresponding to the normalization block being determined by a machine learning model conditioned on the summary representation. Petition 870250101201, dated 05 / 11 / 2025, page 23 / 192 15 / 172 summary.

[0038] In another aspect, a controller is provided for controlling the operation of a breathing assistance device that provides breathing assistance to a user, wherein the controller comprises: a memory unit comprising software instructions and parameters for at least one trained predictive model, the trained predictive model being able to generate, based on sensor data, a real-time estimate of the user’s current breathing status and a prediction of the user’s future breathing status over a predicted period of time;and a processor that is electronically coupled to the memory unit, the processor being configured to generate a control signal to control the respiratory assistance device for a current monitoring time period by: receiving sensor data obtained by one or more sensors, the sensor data corresponding to measurements of at least one airflow parameter of the user's airflow during the current monitoring time period when the user is wearing the respiratory assistance device; applying the trained predictive model to generate the real-time estimate and forecast; generating a summary representation of the user, the summary representation comprising user data; generating the customized predictive model by conditioning the predictive model; Petition 870250101201, dated 05 / 11 / 2025, page 24 / 192 16 / 172 trained using the summary representation, the custom predictive model being customized for the user; and implement the custom predictive model in the processor of the respiratory assistance device controller.

[0039] In at least one embodiment, the processor is additionally configured to perform the method in the manner previously discussed.

[0040] In another aspect, a method is provided for simulating one or more operations of a respiratory assistance device and a health status of a user receiving assistance from the respiratory assistance device, the method comprising: receiving a user model from the user, the user model being adapted to model one or more physiological systems of the user; receiving a device model from the respiratory assistance device, the device model being adapted to model one or more components of the respiratory assistance device, one or more subsystems of the respiratory assistance device, or one or more functions of the respiratory assistance device, or any operable combination thereof; receiving sensor data from one or more sensors; and determining an expected state of the respiratory assistance device and / or an expected health status of the user based on the Petition 870250101201, dated 05 / 11 / 2025, page 25 / 192 17 / 172 application of sensor data in the device model and / or user model.

[0041] In at least one embodiment, the one or more sensors include any combination of: (a) one or more of the user sensors placed on the user, (b) device sensors that measure the properties of the breathing assistance device, and (c) environmental sensors.

[0042] In at least one embodiment, the user sensors include a device for measuring the user's blood parameters.

[0043] In at least one embodiment, the user model is generated, in part, based on one or more personal characteristics of the user, the one or more personal characteristics of the user received from one or more of: the user, a medical professional, a pharmacy system, a payer system, a health system, other home medical equipment, and an electronic medical records system.

[0044] In at least one embodiment, one or more personal characteristics comprise the user's alcohol consumption, the user's drug use, medication taken by the user, the user's height, the user's weight, the user's age, a blood test result, or any operable combination thereof.

[0045] In at least one embodiment, the user model is generated based on a model of the respiratory system of Petition 870250101201, dated 05 / 11 / 2025, page 26 / 192 18 / 172 user, a model of the user's cardiovascular system, a model of the user's nervous system, or any operable combination thereof.

[0046] In at least one embodiment, the method further comprises: determining a user’s current health status by applying sensor data to the user model; comparing the user’s current health status with the user’s expected health status; and generating a recommendation based on the comparison.

[0047] In at least one embodiment, the user model is associated with a physical model that models one or more of the user's physiological systems, and determining the user's current health status involves applying sensor data to the user's physical model.

[0048] In at least one embodiment, the user's health status includes sleep health and wherein the method further comprises: determining one or more expected sleep parameter values ​​for the user by performing a simulation by applying sensor data to the user model and, optionally, to the device; determining one or more actual sleep parameter values ​​for the user based on received sensor data; comparing the one or more expected sleep parameter values ​​with the one or more actual sleep parameter values; and generating a recommendation based on the comparison. Petition 870250101201, dated 05 / 11 / 2025, page 27 / 192 19 / 172

[0049] In at least one embodiment, the one or more sleep parameter values ​​comprise a sleep duration, a length of sleep stages, a sleep depth, a heart rate during sleep, or any operable combination thereof.

[0050] In at least one instance, the recommendation is a health recommendation, a recommendation to consult a medical professional, a diagnosis, a mask change, a tubing change, and / or an adjustment to the operation of the respiratory assistance device.

[0051] In at least one form, the expected state is a future state.

[0052] In at least one embodiment, the method further comprises: determining a device correction factor for the respiratory assistance device to improve the user's health status, wherein the determination is based on the user's expected health status; simulating an operation of the respiratory assistance device when the device correction factor is applied using the device model; and adjusting the operation of the respiratory assistance device according to the correction factor when the simulation indicates an improvement in the user's health status when the correction factor is applied.

[0053] In at least one modality, in which the factor of Petition 870250101201, dated 05 / 11 / 2025, page 28 / 192 20 / 172 correction is based, in part, on the sensor data received.

[0054] In at least one embodiment, the method further comprises: subsequent to adjusting the operation of the respiratory assistance device, determining an intervention index that distinguishes a probability of the user experiencing a respiratory event at a given time; and, when the intervention index exceeds a predetermined threshold of the intervention index, readjusting the operation of the respiratory assistance device by one of: reverting the operation of the respiratory assistance device to an operation prior to the adjustment of the operation of the respiratory assistance device or modifying the operation of the respiratory assistance device to provide reactive therapy, wherein the probability of the user experiencing the respiratory event at the given time is determined based on the user model.

[0055] In at least one embodiment, the method further comprises: subsequent to adjusting the operation of the respiratory assistance device, determining a respiratory event index that distinguishes a number of respiratory events experienced by the user in one hour; and determining an average historical number of respiratory events per hour experienced by the user; and, when the respiratory event index exceeds the average historical number of respiratory events by a limit of Petition 870250101201, dated 05 / 11 / 2025, page 29 / 192 21 / 172 Predetermined respiratory event index, readjust the operation of the respiratory assistance device by one of: reverting the operation of the respiratory assistance device to an operation prior to the adjustment of the respiratory assistance device operation or modifying the operation of the respiratory assistance device to provide reactive therapy.

[0056] In at least one embodiment, adjusting the operation of the respiratory assistance device comprises adjusting an operating mode of the respiratory assistance device, wherein the operating mode is adjusted by selecting a device profile from: a continuous positive airway pressure (CPAP), an automatic positive airway pressure (APAP) profile, a bilevel positive airway pressure (BiPAP) profile, an adaptive servoventilation (ASV) profile, and a non-invasive ventilator (NIV) profile.

[0057] In at least one embodiment, the method further comprises: determining a current state of the respiratory assistance device based on sensor data; comparing the current state of the respiratory assistance device and an expected state of the respiratory assistance device; in response to identifying a Petition 870250101201, dated 05 / 11 / 2025, page 30 / 192 22 / 172 difference in the current state compared to the expected state, determine that one or more components of the respiratory assistance device are defective; and, in response to determining that one or more components are defective, generate a device recommendation.

[0058] In at least one embodiment, the device recommendation is to replace one or more components or perform maintenance on one or more components.

[0059] In at least one embodiment, when the expected state of the respiratory assistance device and / or the expected health status of the user are acceptable, the method comprises implementing at least one model, calibration data, and / or an operational adjustment of the respiratory assistance device that was used during the simulation to update future operation of the respiratory assistance device.

[0060] In at least one embodiment, the received user model used during the simulation also includes a custom predictive model for the user, and the implementation includes sending the custom predictive model to update the future operation of the breathing assistance device.

[0061] In at least one embodiment, the method further comprises performing authentication by receiving a breath signature for the model of Petition 870250101201, dated 05 / 11 / 2025, page 31 / 192 23 / 172 user and / or sensor data, obtain a stored breath signature for the user, and perform the simulation when the received breath signature for the user is equal to the stored breath signature for the user.

[0062] In at least one embodiment, the method comprises providing an outlet to interrupt or adjust therapy in situations where simulation determines that a certain amount of air pressure is unsafe for a user who has undergone surgery.

[0063] In at least one embodiment, the method comprises obtaining an augmented quantity of data recording and analyzing augmented data to investigate whether sleep therapy caused and / or aggravated a medical condition for the user.

[0064] In another aspect, a system is provided for simulating one or more operations of a respiratory assistance device and a health status of a user receiving assistance from the respiratory assistance device, the system comprising: one or more sensors for measuring sensor data; a database storing a user model of the user, the user model being adapted to model one or more internal systems of the user, and a device model of the respiratory assistance device, the device model being Petition 870250101201, dated 05 / 11 / 2025, page 32 / 192 24 / 172 adapted to model one or more components of the respiratory assistance device, one or more subsystems of the respiratory assistance device, or one or more functions of the respiratory assistance device, or any operable combination thereof; a controller communicating with one or more sensors and the database, the controller comprising at least one processor configured to: receive sensor data from one or more sensors; and determine one or more of an expected state of the respiratory assistance device and an expected health state of the user based on one or more of the respiratory assistance device model and the user model, and the sensor data.

[0065] In at least one embodiment, the processor is additionally configured to perform the method, as previously described.

[0066] In another aspect, according to the precepts set forth in this document, at least one embodiment of a method is provided for adjusting an airflow provided by a breathing assistance device for a user, wherein the method comprises: determining a personalized predictive model for the user, the personal predictive model being adapted to control the operation of the specific breathing assistance device for the user; performing the simulation of the user and the device of Petition 870250101201, dated 05 / 11 / 2025, page 33 / 192 25 / 172 respiratory assistance when the respiratory assistance device is controlled by the personal predictive model; determine if the simulation indicates that the use of the personalized predictive model has a beneficial effect on the user's health; and, when there is a beneficial effect on the user's health, implement the personalized predictive model to adjust the future operation of the respiratory assistance device.

[0067] In at least one modality, the customized predictive model can be determined according to the precepts set forth in this document.

[0068] In at least one mode, the simulation is carried out in accordance with the principles set forth in this document.

[0069] In at least one modality, false negative data associated with the custom predictive model are used during the simulation.

[0070] In another aspect, according to the precepts set forth in this document, at least one embodiment of a system is provided for adjusting an airflow provided by a breathing assistance device for a user, wherein the system comprises a memory that stores program instructions for a method for determining a customized predictive model and a realization of a digital simulation and a processor that is coupled to the memory for Petition 870250101201, dated 05 / 11 / 2025, page 34 / 192 26 / 172 receive the program instructions, the processor being configured, during the execution of the program instructions, to perform the method previously described.

[0071] In another aspect, according to the precepts set forth in this document, a non-transient computer-readable medium is provided in at least one embodiment that stores program instructions that are executable by a processor to perform a method that is defined according to the precepts set forth in this document.

[0072] It will be perceived that the summary set forth presents representative aspects of embodiments to assist those skilled in the art in understanding the following detailed description. Other features and advantages of this application will become apparent from the following detailed description taken together with the accompanying drawings. It is understood, however, that the detailed description and specific examples, while indicating preferred embodiments of the application, are given for illustrative purposes only, as various changes and modifications in the spirit and scope of the application will become apparent to those skilled in the art from this detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] For a better understanding of the various modalities described in this document, and to show more Petition 870250101201, dated 05 / 11 / 2025, p. 35 / 192 27 / 172 clearly showing how these various modalities can be carried out, reference will be made, by way of example, to the attached drawings which show at least one example embodiment, and which are now described. The drawings are not intended to limit the scope of the precepts described in this document.

[0074] Figure 1A is a block diagram of an example embodiment of a breathing assistance system for controlling or fine-tuning a breathing assistance device during use by a user based on the determination and / or prediction of different sleep and / or respiratory behaviors in accordance with the precepts set forth in this document.

[0075] Figure 1B is a schematic diagram of another example embodiment of a respiratory assistance system for controlling or fine-tuning a respiratory assistance device during use by a user based on the determination and / or prediction of different sleep and / or respiratory behaviors in accordance with the precepts set forth in this document.

[0076] Figure 2 is a block diagram of another example embodiment of a respiratory assistance system for controlling or fine-tuning a respiratory assistance device during use by a user based on the determination and / or prediction of Petition 870250101201, dated 05 / 11 / 2025, page 36 / 192 28 / 172 different sleep and / or respiratory behaviors according to the principles set forth in this document.

[0077] Figure 3A is a flowchart of an example embodiment of a method for generating a custom predictive model that can be used to adjust an airflow provided by a breathing assistance device for a user in accordance with the principles set forth in this document.

[0078] Figure 3B is an example flowchart of the data flow through an example respiratory assistance system according to the principles set forth in this document.

[0079] Figure 4A shows example sensor data comprising pressure and airflow data.

[0080] Figure 4B shows sample real-time estimation and forecasting data generated by a sample predictive model trained according to the principles set forth in this document.

[0081] Figure 4C shows sample real-time estimation and forecasting data generated by a custom sample predictive model in accordance with the principles set forth in this document.

[0082] Figure 4D shows sample sensor data comprising pressure and airflow data and sample forecast data generated by a predictive model of Petition 870250101201, dated 05 / 11 / 2025, page 37 / 192 29 / 172 customized example according to the precepts set forth in this document.

[0083] Figure 5A shows sample sensor data comprising airflow data, sample forecast data generated by a sample general trained predictive model, and sample forecast data generated by a sample custom predictive model according to the principles set forth in this document.

[0084] Figure 5B additionally shows example sensor data comprising airflow data, example forecast data generated by an example general trained predictive model, and example forecast data generated by a customized example predictive model according to the principles set forth in this document.

[0085] Figure 5C shows further example sensor data comprising airflow data, example forecast data generated by a general example trained predictive model, and example forecast data generated by a custom example predictive model according to the principles set forth in this document.

[0086] Figure 6A shows example sensitivity data corresponding to a general example trained predictive model and a custom example predictive model according to the principles set forth in this document.

[0087] Figure 6B shows example precision data. Petition 870250101201, dated 05 / 11 / 2025, page 38 / 192 30 / 172 corresponding to a general trained predictive model example and a customized predictive model example according to the precepts set forth in this document.

[0088] Figure 7 is a flowchart of an example embodiment of a method for simulating the operation of a respiratory assistance device and / or the health status of a user receiving assistance from the respiratory assistance device.

[0089] Additional aspects and characteristics of the example embodiments described in this document will appear from the following description taken together with the attached drawings. DETAILED DESCRIPTION OF THE MODALITIES

[0090] The headings and Description Summary provided in this document are for convenience only and do not interpret the scope or meaning of the modalities.

[0091] Various embodiments of the principles set forth in this document will be described below to provide examples of at least one embodiment of the claimed subject matter. No embodiment described in this document limits any claimed subject matter. The claimed subject matter is not limited to devices, systems, or methods that have all the characteristics of any of the devices, systems, or methods described below or the characteristics common to multiple or all of the devices, systems, or methods described below. Petition 870250101201, dated 05 / 11 / 2025, p. 39 / 192 31 / 172 methods described in this document. It is possible that there may be a device, system or method described in this document that is not an embodiment of any claimed matter. Any matter described in this document that is not claimed in this document may be the subject matter of another protective instrument, for example, a continuing patent application, and the applicants, inventors or holders do not intend to abandon, decline or dedicate to the public any such matter by its description in this document.

[0092] Furthermore, it will be noted that, for simplicity and clarity of illustration, when deemed appropriate, reference numbers may be repeated between figures to indicate corresponding or analogous elements or steps. Moreover, numerous specific details are presented in order to provide a thorough understanding of the embodiments described in this document. However, it will be understood by those skilled in the art that the embodiments described in this document can be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the embodiments described in this document. Also, the description should not be considered as limiting the scope of the embodiments described in this document.

[0093] It should also be noted that the terms “coupled” Petition 870250101201, dated 05 / 11 / 2025, page 40 / 192 32 / 172 or coupling, as used in this document, can have several different meanings depending on the context in which these terms are used. For example, the terms coupled or coupling can have a mechanical, electrical, or communicative connotation. For example, as used in this document, the terms coupled or coupling can indicate that two elements or devices can be directly connected to each other or connected to each other through one or more intermediate elements or devices by means of an electrical element, electrical signal, or a mechanical element, depending on the particular context.

[0094] Unless the context requires otherwise, for all the specification and claims that follow, the word comprehends and variations thereof, such as comprehend and comprehending, should be interpreted in an open, inclusive sense, that is, as including but without limitations.

[0095] Several terms used throughout this description may be read and understood as follows, unless the context indicates otherwise: articles and pronouns in the singular form, as used throughout, include their plural forms, and vice versa; similarly, gender pronouns include their counterpart pronouns, so that pronouns should not be understood as limiting anything described in this document for use, implementation, Petition 870250101201, dated 05 / 11 / 2025, page 41 / 192 33 / 172 performance, etc. by a single gender. Additional definitions for the terms may be presented in this document; they may apply to prior and subsequent instances of these terms, as will be understood from a reading of this description.

[0096] It should also be noted that, as used in this document, the wording “and / or” is intended to represent an inclusive “or”. That is, you want “X and / or Y” to mean X or Y or both, for example. As a further example, you want “X, Y, and / or Z” to mean X or Y or Z or any combination thereof. As another example, the phrase “A, B, C or any operable combination thereof” or “any combination of A, B and C” is intended to cover any combination of the elements A, B and C that provides utility, which may, for example, include A, B, C, A and B, A and C, B and C, or A, B and C.

[0097] It should be noted that degree terms, such as “substantially”, “about”, and “approximately”, as used in this document, mean a reasonable amount of deviation from the modified term, such that the final result is not significantly altered. These degree terms may also be interpreted as including a deviation from the modified term, such as by 1%, 2%, 5%, or 10%, for example, if this deviation does not negate the meaning of the term it modifies. Petition 870250101201, dated 05 / 11 / 2025, p. 42 / 192 34 / 172

[0098] Furthermore, the citation of numerical ranges by periods set forth in this document includes all numbers and fractions subsumed within this range (e.g., 1 to 5 includes 1, 1.5, 2, 2.75, 3, 3.90, 4, and 5). It should also be understood that all numbers and fractions thereof are presumed to be modified by the term “about”, which means a variation of up to a certain amount from the number to which the reference is being made if the final result is not significantly altered, such as 1%, 2%, 5%, or 10%, for example.

[0099] Reference throughout this specification to a “some modality”, “at least one modality” or “some modalities” means that one or more particular features, structures, or characteristics may be combined in any suitable manner in one or more modalities, unless otherwise specified as not combinable or as alternative options.

[00100] As used in this specification and the accompanying claims, the singular forms “a”, “an”, “the” and “the” include plural referents, unless the content clearly indicates otherwise. It should also be noted that the term “or” is generally used in its broadest sense, i.e., meaning “and / or”, unless the content clearly indicates otherwise.

[00101] Similarly, for all this specification and Petition 870250101201, dated 05 / 11 / 2025, p. 43 / 192 35 / 172 As per the appended claims, the term “communicative,” as in “communicative pathway,” “communicative coupling,” and in variants such as “communicatively coupled,” is generally used to refer to any engineered arrangement for transferring and / or exchanging information. Examples of communicative pathways include, but are not limited to, electrically conductive pathways (e.g., electrically conductive wires, physiological signal conduction), electromagnetically radiative pathways (e.g., radio waves), or any combination thereof. Examples of communicative couplings include, but are not limited to, electrical couplings, magnetic couplings, radio couplings, or any combination thereof.

[00102] Throughout this specification and the attached claims, infinitive verb forms are frequently used. Examples include, but are not limited to: “to detect,” “to provide,” “to transmit,” “to communicate,” “to process,” “to rotate,” and the like. Unless the specific context requires otherwise, such infinitive verb forms are used in an open inclusive sense, i.e., as “to, at least, detect,” “to, at least, provide,” “to, at least, transmit,” and so forth.

[00103] A portion of the example embodiments of the systems, devices, or methods described in accordance with Petition 870250101201, dated 05 / 11 / 2025, page 44 / 192 36 / 172 The precepts set forth in this document may be implemented as a combination of hardware or software. For example, some of the embodiments described in this document may be implemented, at least in part, by the use of one or more computer programs running on one or more programmable devices comprising at least one processing element and at least one data storage element (including volatile and non-volatile memory). These devices may also have at least one input device (e.g., a keyboard, a mouse, a touch screen, and the like) and at least one output device (e.g., a display screen, a printer, a wireless radio, and the like), depending on the nature of the device.

[00104] It should also be noted that there may be some elements used to implement at least part of the modalities described in this document that can be implemented by means of software written in a high-level procedural language, such as object-oriented programming. The program code may be written in C, C++, or any other suitable programming language, and may comprise modules or classes, as is known to those versed in the art of object-oriented programming. Alternatively, or in addition, some of these elements implemented by means of software may be written in Petition 870250101201, dated 05 / 11 / 2025, page 45 / 192 37 / 172 assembly language, machine language or embedded software, as needed.

[00105] At least parts of the software programs used to implement at least one of the embodiments described in this document may be stored on a storage medium or device that is readable by a general-purpose or special-purpose programmable device. The software program code, when read by the programmable device, configures the programmable device to operate in a new, specific, and predefined manner in order to perform at least one of the methods described in this document.

[00106] Furthermore, at least some of the programs associated with the systems and methods of the embodiments described in this document may be capable of being distributed in a computer program product comprising a computer-readable medium that carries computer-usable instructions, such as program code, for one or more processors. The program code may be pre-installed and embedded during manufacturing and / or may be subsequently installed as an update to an already implemented computing system. The medium may be provided in various forms, including non-transient forms such as, but not limited to, one or more floppy disks, Compact Discs, tapes, chips, and magnetic and electronic storage. In alternative embodiments, the medium may be Petition 870250101201, dated 05 / 11 / 2025, page 46 / 192 38 / 172 transient in nature, such as, but not limited to, cable transmissions, satellite transmissions, internet transmissions (e.g., transfer), media, digital and analog signals, and the like. Computer-usable instructions may also be in various formats, including compiled and uncompiled code.

[00107] Thus, any module, unit, component, server, computer, terminal, or device described in this document that executes software instructions may include or otherwise have access to computer-readable media, such as storage media, computer storage media, or data storage devices (removable and / or non-removable), such as, for example, magnetic disks, optical disks, or tape. Computer storage media may include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data.Examples of computer storage media include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile discs (DVDs) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or. Petition 870250101201, dated 05 / 11 / 2025, page 47 / 192 39 / 172 any other media that can be used to store the desired information, and that can be accessed by an application, module, or both. Any such computer storage media may be part of the device or accessible or connectable to it.

[00108] It should be noted that the term user covers a person who is using a breathing assistance device. In some cases, the user may be an individual who is using the breathing assistance device in their home or a non-medical setting. In other cases, the user may be a patient who is using the breathing assistance device in a medical setting, such as a clinic or hospital, for example.

[00109] Respiratory failure can be understood to encompass all diseases and conditions that can result in a negative change in the respiratory system of a person or animal, such as an obstruction of breathing or peripheral airways. In some cases, respiratory failure may be a temporary respiratory event that occurs, such as during OSA or an asthma attack, or it may be due to a chronic respiratory condition, such as lung cancer, cystic fibrosis, or Chronic Obstructive Pulmonary Disease.

[00110] In accordance with the precepts set forth in this document, several modalities are provided for Petition 870250101201, dated 05 / 11 / 2025, p. 48 / 192 40 / 172 Intelligent monitoring, analysis, determination, and / or prevention of different respiratory and sleep behaviors of a user, such as generating a personalized predictive model that is specific to a particular user to detect and / or predict the user's breathing state, and using a digital model of the user (e.g., digital twin) and / or a digital model of a respiratory assistance device that provides assistance to the user to simulate the user's sleep health, which may include predicting the user's future sleep health. Various modalities can be used to adjust the operation of the respiratory assistance device. Several modalities for monitoring the operation of a respiratory assistance device and predicting future operation of the respiratory assistance device by simulating the operation of the respiratory assistance device using a digital model of the respiratory assistance device are also provided.

[00111] At least some of the methods involve using machine learning and other advanced computational techniques to build accurate supervised classification and prediction models. Some of these models will employ multiple measurements, some of which include, but are not limited to: (a) air pressure and airflow only; (b) airway impedance measured by FOT only; (c) plethysmography, such as EEG; (d) still other data from Petition 870250101201, dated 05 / 11 / 2025, page 49 / 192 41 / 172 sensor, as described in this document, or (e) a combination of two or more of (a), (b), (c), and other sensor data described in this document.

[00112] In one aspect, the precepts set forth in this document may utilize data associated with the forced oscillation technique in the detection and / or prediction of respiratory failure, which may be performed in various ways, such as, but not limited to, the techniques described in U.S. Patent 11,633,560 entitled “METHOD AND APPARATUS FOR CONTINUOUS MANAGEMENT OF AIRWAY PRESSURE FOR DETECTION AND / OR PREDICTION OF RESPIRATORY FAILURE”, and / or in U.S. Patent 11,612,708 entitled “METHOD & APPARATUS FOR DETERMINING AND / OR PREDICTING SLEEP AND RESPIRATORY BEHAVIOURS FOR MANAGEMENT OF AIRWAY PRESSURE”, each of which is hereby incorporated by reference in its entirety.

[00113] In one aspect, in at least one modality, the precepts set forth in this document provide techniques for generating a personalized predictive model that may be based, in some cases, on physiological measurements of the user, or values ​​calculated from such physiological measurements, such as a user's breathing signature, classification of the user's sleep stage and / or classification of the severity of the predicted sleep disruption. Based on the detection and / or prediction indicated by the personalized predictive model, the operating parameters of Petition 870250101201, dated 05 / 11 / 2025, page 50 / 192 42 / 172 respiratory assistance devices can then be adjusted to improve the chances that the user of the respiratory assistance device will not experience a respiratory failure event or at least will experience minimal respiratory failure events. In this way, at least one of these detection and / or prediction indications can be used to generate a feedback control signal that is used to control the operation of the respiratory assistance device. These indications can be determined for a given period of time and can be used to control the respiratory assistance device throughout that period of time.

[00114] Therefore, in one aspect, in at least one embodiment, the precepts set forth in this document provide real-time, customized detection and / or prediction of the user's breathing status based on a customized predictive model, which can then be used to perform at least one adjustment of the operation of the respiratory assistance device and provide data in a user report that can be used to monitor the user's breathing and / or diagnose a respiratory disorder for the user. For example, the report can be generated for data collected when the user slept at night, and the report can be provided to the user or a medical professional for review, such as in the morning, so that the user or Petition 870250101201, dated 05 / 11 / 2025, page 51 / 192 43 / 172 the medical professional can review data about your respiratory and sleep health, for example. Several example modalities of the personalized predictive model are provided in this document. The predictive model may employ machine learning models also referred to as AI models.

[00115] Previously, it was not possible to provide an automated warning of an impending respiratory failure event to the user of a respiratory assistance device before respiratory failure was about to occur. Therefore, respiratory assistance devices were conventionally controlled manually by a medical professional who set and then adjusted the operating parameters of the respiratory assistance device frequently. This was detrimental because, if the user began to experience respiratory failure, it was not conventionally possible to automatically adjust the respiratory assistance device to reduce the effect or amount of respiratory failure experienced by the user, which can be fatal in some situations where response time is critical for adjusting the operation of the respiratory assistance device.Furthermore, such conventional techniques do not even allow for the prediction of impending respiratory failure.

[00116] More recently, other techniques including Petition 870250101201, dated 05 / 11 / 2025, page 52 / 192 Traditional FOT / Oscillometry methods were used to automatically adjust the parameters of respiratory assistance devices. However, traditional FOT uses weighting and therefore there is a delay of several seconds before any detection can occur. This is also detrimental to the user's health when significant respiratory failure is imminent or occurring. Furthermore, automatic adjustment of respiratory assistance devices using techniques such as only sensing airflow or oxygen levels may not provide sufficient information about the health of the entire respiratory system in certain situations.

[00117] It is believed that techniques for generating a customized predictive model for detecting and / or preventing a user's breathing state, in order to generate a control signal to control the respiratory assistance device to maintain the user's respiratory health within a certain range where the user is not experiencing a respiratory failure event, in at least one modality provided in accordance with the precepts set forth in this document, will increase the adoption rate of use of respiratory assistance devices where use is voluntary (i.e., for sleep apnea devices). Such models also provide technical advantages, such as an increase in the speed of adaptation of the assistive device. Petition 870250101201, dated 05 / 11 / 2025, page 53 / 192 45 / 172 breathing assistance is crucial for detecting any respiratory failure event that is currently being encountered or, preferably, for preventing a respiratory failure event that the user may encounter soon, since methods using custom AI models described in this document can detect and / or predict respiratory failure relatively quickly, with increased confidence, and can also, preferably, take proactive or reactive steps quickly to control the breathing assistance device to reduce the level / amount of respiratory failure encountered by the user or even prevent the respiratory failure event from occurring. This can be critical in some cases where increased respiratory failure events can have significant, if not fatal, consequences for the user.

[00118] In another aspect, the precepts set forth in this document provide at least one method for simulating the operation of a respiratory assistance device and / or simulating a user's health status, for example, a user receiving assistance from the respiratory assistance device, using various models of the user (i.e., user model) and / or of the respiratory assistance device (i.e., device model). Simulating a user's health status using a user model through the use of a "digital twin" of the Petition 870250101201, dated 05 / 11 / 2025, page 54 / 192 46 / 172 user, which can be used to assess a user's current health status and compare the current health status with an expected health status to evaluate the therapy provided by the respiratory assistance device, and / or to predict a future condition of the user if the therapy should continue, such as a future health status of the user, a future sleep health of the user and / or a future respiratory health of the user and / or a future operation of the respiratory assistance device. By predicting health events and / or health conditions or diseases, it may be possible to avoid or reduce the likelihood of the events and / or conditions or diseases occurring or to seek early treatment. In some cases, the prediction(s) may be used to make recommendations and / or to modify the operation of the respiratory assistance device to adjust the therapy provided to the user, such as, for example, to improve the therapy provided to the user.In at least one example embodiment, the device model of the respiratory assistance device and / or the user together with (a) respiratory assistance device data, (b) sensor data from sensors on the user and / or in the user's room, (c) user data provided by the user and / or a medical professional supervising the user's therapy, and / or (d) test results can be used to simulate the effects of changes. Petition 870250101201, dated 05 / 11 / 2025, page 55 / 192 47 / 172 in the therapy provided by the respiratory assistance device, which may be useful in identifying ideal or improved adjustments in the operation of the respiratory assistance device. For example, (i) adjustments may be made to the operation of the respiratory assistance device, (ii) recommendations may be provided to the patient to adjust their mask, change the tubing used with the respiratory assistance device, change the type of mask the user uses, and / or change to another user-specific sleeping setup, (iii) reports and / or alerts may be sent to the patient and / or to an external user, such as a medical professional who monitors the patient's therapy, or (iv) a combination of two or more of (i), (ii), and (iii).As another example, simulating the operation of the respiratory assistance device through the use of a "digital twin" of the respiratory assistance device can help identify faulty components of the respiratory assistance device or make predictions about future failure conditions that the respiratory assistance device may experience.

[00119] The various custom models, user models, and device models described in this document can be developed using a combination of models, for example, analog models, digital models, models Petition 870250101201, dated 05 / 11 / 2025, page 56 / 192 48 / 172 mathematical, neural network models and / or other machine learning models, to model the user's physiological systems that may affect a user's health status, including sleep health, cardiovascular health, respiratory health and / or brain health, and to model the user's environment and / or environmental factors that may affect the user's health status. The models can be customized by adjusting model parameters so that the models can more accurately reflect a given user's health status based on the health of another user and / or reflect the user's current environment. The respiratory assistance device model can be similarly developed using a combination of models to model the operation and / or components of the respiratory assistance device.

[00120] Now, with regard to Figure 1A, a block diagram of a respiratory assistance system 100 is illustrated therein, which can use a custom predictive model to adjust an airflow provided by a respiratory assistance device to a user and / or to simulate an operation of the respiratory assistance device and / or a health status of the user receiving assistance from the respiratory assistance device in communication with external components via the network 114. In at least one mode, the system of Petition 870250101201, dated 05 / 11 / 2025, page 57 / 192 49 / 172 breathing assistance 100 can also be used to control or fine-tune the air pressure or airflow provided by a breathing assistance device to the user using the forced oscillation technique based on the detection and / or prediction of a respiratory failure event using one or more customized models according to at least one embodiment of the precepts set forth in this document. The system 100 comprises a breathing assistance device 102 that generates an airflow that is provided to a user 110 through air conduits 104 and 108 and, for example, a laryngeal tube, a breathing mask, or an endotracheal tube 109 (hereinafter collectively referred to as an “inlet element”). The airflow can be at least a pulse of air pressure, a continuous airflow, or another type of airflow, as is known to those skilled in the art.The airflow is controllable by adjusting at least one of the air pressure and flow rate of the respiratory assistance device 102 via corresponding inlet controls on the respiratory assistance device 102.

[00121] In some embodiments, the respiratory assistance device 102 may be a mechanical ventilator to provide respiratory support to the user. In other embodiments, the respiratory assistance device 102 may be a CPAP, APAP, BiPAP, PAP, ASV, NIV, or device. Petition 870250101201, dated 05 / 11 / 2025, page 58 / 192 50 / 172 High-Flow Oxygen (HFO) to provide respiratory support to the user. In other embodiments, the respiratory assistance device 102 may be a respiratory treatment delivery device, such as, but not limited to, respiratory treatment delivery devices that assist a user in clearing their lungs and expectorating secretions. In other instances, the respiratory assistance device 102 may be an anesthesia machine in the OR, an ICU ventilator, a residential COPD ventilator and oxygenator, and any other machine that provides respiratory assistance to a user who has a respiratory disease.Therefore, in general, the principles described in this document for the detection and / or prediction of a respiratory failure event and the proactive or reactive actions taken to reduce, remove, or prevent respiratory failure can be used with all types of ventilation, including invasive (tubed) and non-invasive (tubed) ventilation.

[00122] A respiratory failure event can include any event that deviates from normal breathing. For example, in sleep apnea-related breathing, a respiratory failure event can include flow limitation, snoring, obstructive apnea, central apnea, obstructive hypopnea, central hypopnea, respiratory effort-related stimuli (RERA). Petition 870250101201, dated 05 / 11 / 2025, page 59 / 192 51 / 172 Related Arousals), and / or excitations. For example, stimuli related to respiratory effort may include flow limitation, wheezing, oxygen desaturation, fragmentation, heart rate abnormality, or any combination thereof. As another example, in users with diseases such as Chronic Obstructive Pulmonary Disease, a respiratory failure event may include lung stiffness, exacerbation, and / or flow limitation. As another example, in users on an ICU ventilator, a respiratory failure event may include failure to wean the user from ventilation, pneumonia, and / or resistive or elastic lung events. Other respiratory failure events may be determined.

[00123] A respiratory assistance device controller 106 is coupled to the respiratory assistance device 102 via the air conduit 104 (which may also be called the flow passage 104) and receives airflow from the respiratory assistance device 102 and distributes the airflow via the air conduit 108 and the inlet element 109 to the user 110. It should be noted that the term air in this description is generally used to denote the flow of gas and other airborne particles through the system 100. For example, the output of a mechanical ventilator may include gases and / or vapors other than air, such as, but without Petition 870250101201, dated 05 / 11 / 2025, page 60 / 192 52 / 172 limitations, anesthetics, for example, which are typically vapors, but can also be gases. In a PAP device, water vapor can be combined with air. In some embodiments of the respiratory assistance device 102, gaseous medication (i.e., steroids, oxygen, nitrogen, etc.) can be added to the airflow and provided to the patient under ventilation based on respiratory health and / or measured comfort level. For example, the medication may include an appropriate amount of steroids that can be used daily to improve the CPAP experience for the user. The airflow can be delivered to the user 110 through the inlet element 109. In the present embodiment, the inlet element 109 can be a mask worn over the nose and mouth of the user 110, only over the nose or adjacent to or inside the nostrils of the user 110 for alternative masks.In other modalities, the entry element 109 may be an endotracheal tube inserted into the trachea via intubation or tracheostomy.

[00124] In embodiments where the breathing device 102 is a mechanical ventilator, there are actually two airways (not shown) instead of just the air transport pathway 104 (which may also be called a flow passage) where one of the pathways is used for inhalation and the other of these pathways is used for exhalation. The pathways shown in Figure 1 apply to the case where the Petition 870250101201, dated 05 / 11 / 2025, page 61 / 192 53 / 172 breathing assistance device 102 is a PAP device. It can be understood that the breathing assistance device 102 provides at least one pathway to allow air to flow from the air transport pathway 104 to the air transport pathway 108. It can be further understood that there may be embodiments in which the controller of the breathing assistance device 106 is at least partially or completely incorporated “in-line” with the airflow pathways from the breathing assistance device 102 to the user 110.

[00125] In the present embodiment, the controller of the respiratory assistance device 106 comprises one or more sensors (for example, see Figure 2) to measure various parameters of the airflow that is delivered to the user 110. For example, sensors may be attached to the mask worn by the user 110, which may result in improved SNR for the sensor data obtained from the sensors. Alternatively, sensors, such as ultrasonic sensors, for example, may be attached to the tubing pathway. In either case, these sensors can be used to measure the airflow parameters associated with both inspiration and expiration. However, in the case of a PAP machine, such sensors are located close to the mask because the tubing 108 only conducts inspiratory flow, whereas, in a mechanical ventilator, Petition 870250101201, dated 05 / 11 / 2025, page 62 / 192 54 / 172 The sensors can be attached to the mask or endotracheal tube, or they can be located anywhere along the tubes used for the inspiratory and expiratory pathways.

[00126] In some embodiments, the respiratory assistance device controller 106 may not include these sensors, but may instead read these parameters from the respiratory assistance device 102, since the respiratory assistance device 102 may also be equipped with sensors to measure airflow parameters. In some embodiments, the respiratory assistance device controller 106 transmits data, including sensor data, to an external system and / or receives data from an external system that is used to adjust the operation of the respiratory assistance device 102 (for example, see Figure 1B). The respiratory assistance device controller 106 may further comprise a device for providing a forced oscillation signal in order to provide changes in air pressure to the airflow provided to the user 110.In some embodiments, a sensor to measure both air pressure and airflow is present. In other embodiments, dedicated sensors may be used to measure airflow or air pressure, such that more than one sensor may be used with the device controller. Petition 870250101201, dated 05 / 11 / 2025, page 63 / 192 55 / 172 breathing assistance 106. For example, some sensor technologies use a laser to detect movement or ultrasound can be used to detect both pressure and flow using a sensor (since the measured flow can be determined from dividing the measured pressure by a known resistance).

[00127] The measured airflow parameters, such as air volume, air pressure and / or air flow rate, can be used by the respiratory assistance device controller 106 to generate a control signal 112 that can be used as feedback to adjust the operation of the respiratory assistance device 102. For example, the respiratory assistance device controller 106 may employ a control method wherein the control signal that is generated may be based on a trained predictive model, such as, but not limited to, a custom predictive model (e.g., according to method 300 in Figure 3A) and / or a predicted current or future sleep or respiratory health of the user and / or a predicted current or future operation of the respiratory assistance device (e.g., according to method 700 in Figure 7).The customized predictive model can be trained using user-specific data, and the user's predicted future sleep and respiratory health can be determined using user-specific model(s) and from this. Petition 870250101201, dated 05 / 11 / 2025, page 64 / 192 In this way, a user-specific control signal can be generated.

[00128] For example, the respiratory assistance controller 106 may employ a control method that uses multiple therapy models, including a trained predictive model and / or a custom predictive model, such as those described in relation to method 300 (see figure 3A), for example, to predict when a respiratory failure event will occur (for example, perhaps up to and including the next few minutes, such as, for example, from about a few milliseconds to about 5 minutes) and then generate the control signal to provide proactive action so that the user does not experience the predicted respiratory failure event.As another example, the respiratory assistance controller 106 may employ a control method that uses (a) a digital model of the user, (b) a digital model of the respiratory assistance device, such as those described in relation to method 700 (see figure 7) to predict an expected state of the respiratory assistance device and / or (c) an expected health state of the user and then generate the control signal to adjust the operation of the respiratory assistance device to improve the expected state of the respiratory assistance device and / or the user and provide preventive therapy. In such cases... Petition 870250101201, dated 05 / 11 / 2025, page 65 / 192 57 / 172 modalities, in at least one modality, additional measured signals may be used to implement the predictive method. For example, the additional measured signals may be one or more of the physiological and / or neurological signals that are obtained during polysomnography (PSG) and hereinafter referred to as PSG signals. PSG signals may include at least one physiological signal (such as, but not limited to, eye movements (EOG), muscle activity or musculoskeletal activation (EMG), and cardiac signals (ECG)) and / or at least one neurological signal (such as, but not limited to, EEG). PSG signals may additionally include air pressure, airflow, chest movement tracked by a belt worn around a user's chest, SpO2, pulse oximetry, audio signals, camera signals, temperature signals, heart rate, or any operable combination thereof.In some embodiments, the additional measured signals may include other types of data, such as those measured by sound sensors (e.g., microphones), motion detection and / or recording systems (e.g., image and video acquisition and / or recording systems, and / or radar and / or electromagnetic recording systems), humidity sensors, alcohol and / or substance sensors, position sensors (e.g., gyroscopes and / or accelerometers), light sensors, pressure sensors, or any operable combination thereof. Petition 870250101201, dated 05 / 11 / 2025, page 66 / 192 58 / 172 Sensors for measuring such signals are known to be skilled in the art and can be added to the system 100, depending on the particular mode that is used.

[00129] In several embodiments described in this document, the control signal 112 can be used to adjust one, two, some, or all of the adjustable parameters of the respiratory assistance device 102 to prevent or reduce the severity of a respiratory failure event. For example, parameters that can be adjusted include airflow rate, airflow volume, airflow pressure, rate of increase of airflow pressure, rate of decrease of airflow pressure, the frequency of certain changes in airflow (such as changes in airflow rate, volume, pressure, and amplitude), airflow amplitude and / or airflow phase that can be generated by the respiratory assistance device 102, or an operable combination thereof.

[00130] In the present embodiment, the respiratory assistance device controller 106 comprises a network communication module (not shown) that enables communication with a server 118 and a data storage device 116 via a network 114. Although a server 118 is shown, it is understood that one or more servers 118 may be understood to be distributed over a wide geographical area and able to communicate by Petition 870250101201, dated 05 / 11 / 2025, page 67 / 192 59 / 172 network medium 114 with the respiratory assistance device controller 106. For example, server 118 can be configured to receive data from the respiratory assistance device controller 106. Server 118 can be further configured to generate and / or update models, including a user model 120, a device model 122 of the respiratory assistance device 102, and various therapy models implemented by the therapy module 124, as shown in Figure 1B, including predictive models. In at least one modality, the predictive model(s) comprise(s) trained predictive model(s) using an aggregate of group data.In at least one embodiment, the predictive model comprises one or more custom predictive models of the user and / or the respiratory assistance device that are trained and / or developed using data received from the respiratory assistance device controller 106 and / or from sensors (not shown). In at least one embodiment, the server 118 is configured to generate the custom predictive model(s) using one or more retraining techniques using the data received from the respiratory assistance device controller 106. In at least one embodiment, the server 118 receives data from the data storage 116. For example, in at least one embodiment, the server 118 receives data from the data storage 116. Petition 870250101201, dated 05 / 11 / 2025, page 68 / 192 60 / 172 minus one modality, the predictive model(s) are generated using group data received from data store 116. Server 118 can also be configured to save data, such as one or more models, in data store 116. Server 118 can be a cloud-based server, as shown in Figure 1B.

[00131] Data storage 116 may include RAM, ROM, one or more hard disks, one or more flash drives, or some other suitable data storage elements, such as disk drives, etc. For example, data storage 116 may include one or more databases or files (both not shown) for storing, for example, sensor data, predictive models, training data, test data, and / or validation data. Data storage 116 may additionally be used to store information related to sensor data, such as, for example, user data relating to sensor data, predictive models, training data, test data, and / or validation data.The sensor data, predictive models, training data, test data, validation data, and / or information related to the sensor data stored in data storage 116 can be retrieved by the respiratory assistance device controller 106 and / or the server 118. Data storage 116 may also include database(s) for... Petition 870250101201, dated 05 / 11 / 2025, page 69 / 192 61 / 172 storage of user-related information that can be used to train and / or generate the predictive model(s) and information regarding the results obtained from the application of the model(s), for example, recommendations and reports. In some modalities, data storage 116 is cloud-based.

[00132] The computing device 150 may include any device capable of communicating with other devices through a network, such as the network 114. A network device may connect to the network 114 via a wired or wireless connection. The computing device 106 may include a processor and memory, and may be an electronic tablet, a personal computer, a workstation, a server, a portable computer, a mobile device, a personal digital assistant, a laptop, a smartphone, a WAP phone, an interactive television, video display terminals, game consoles, and portable electronic devices, or any combination thereof. The computing device 150 may be a device used to input and / or receive data about the user and / or to receive alerts and / or notifications associated with the user's therapy.Although only one computing device 150 is shown, it will be understood that more than one computing device 150 may be communicating with the respiratory assistance device. Petition 870250101201, dated 05 / 11 / 2025, page 70 / 192 62 / 172 102, the data storage 116 and the server 118. For example, a first computing device 150 may be associated with a medical professional who monitors the user's therapy 110 and a second computing device 150 may be associated with the user 110.

[00133] Network 114 may be any network capable of carrying data, including the Internet, Ethernet, standard analog telephone service line / system (POTS, Plain Old Telephone Service), public switched telephone network (PSTN, Public Switch Telephone Network), integrated services digital network (ISDN, Integrated Services Digital Network), digital subscriber line (DSL, Digital Subscriber Line), coaxial cable, fiber optic, satellite, mobile, wireless (e.g., Wi-Fi, WiMAX), SS7 signaling network, fixed line, local area network, wide area network, and others, including any combination thereof, capable of interfacing with, and enabling communication between, the respiratory assistance device controller 106, the server 118, and the data storage 116.

[00134] Now, with regard to Figure 2, a block diagram of an example embodiment of a respiratory assistance system 200 is shown therein, which can be used to generate user-customized predictive model(s) and / or simulate the operation of a device. Petition 870250101201, dated 05 / 11 / 2025, page 71 / 192 63 / 172 respiratory assistance 202 and / or a user's health status. The respiratory assistance device 200 can be used to control or tune the respiratory assistance device 202 during use by a user 210 based on the detection and / or prediction of respiratory failure, analysis of simulations of the operation of the respiratory assistance device 202 and / or analysis of simulations of the user's health status, in accordance with the precepts set forth in this document.

[00135] Elements in Figure 2 that correspond to elements in Figure 1A have been numbered similarly. Similar to the configuration of the breathing assistance system 100, a breathing assistance device 202 generates airflow that is provided to a user 210 via air conduits 204 and 204' and the breathing tube 208, and the airflow is monitored by a breathing assistance device controller 206 to modify the operation of the breathing assistance device 202 under certain conditions. Similar to Figure 1A, the airflow can be delivered to the user 210 via an inlet element 209. In this example embodiment, the inlet element 209 can be a mask worn over, to fluidically couple with, the user's nose 210 (i.e., nostrils) and, optionally, the user's mouth. In other embodiments, the inlet element 209 can be an endotracheal tube. Petition 870250101201, dated 05 / 11 / 2025, page 72 / 192 64 / 172 inserted into the trachea via intubation or tracheostomy.

[00136] Figure 2 provides further details regarding the various system components that may be employed. In some embodiments, the respiratory assistance device 202 may be a mechanical ventilator to provide respiratory support. In other embodiments, the respiratory assistance device 202 may be a PAP device to provide respiratory support. Other options are available for the respiratory assistance device 202, as explained for the respiratory assistance device 102.

[00137] In the present embodiment, the respiratory assistance device 202 is a mechanical ventilator; however, in other embodiments, the respiratory assistance device 202 may be a PAP or CPAP device. The respiratory assistance device 202 includes an inspiratory tube 204 and an expiratory tube 204' to provide an airflow pathway for airflow leaving and returning to the respiratory assistance device 202, respectively. The inspiratory tube 204 and the expiratory tube 204' can be connected to the respiratory assistance device controller 206 in an airflow pathway using the connector tube 214. The airflow can then flow to the user 210 through another airflow pathway from the respiratory assistance device controller. Petition 870250101201, dated 05 / 11 / 2025, page 73 / 192 65 / 172 breathing 206. The airflow of the inspiratory tube 204 may be disturbed by a forced oscillation produced by a motor or an actuator (hereinafter referred to as an “actuator” to refer to both cases) 216 which generates an air oscillation at a desired frequency and intensity. The actuator 216 may be a loudspeaker, an electromagnet, a piezoelectric device, a piston, or a motor, for example. The choice of actuator may depend on the design specifications, such as the physical size of the device 206, as well as the limitations imposed in the Bill of Materials (BOM). It should be noted that in some embodiments, the actuator 216 may be included in the breathing assistance device 202 and not in the controller of the breathing assistance device 206.Alternatively, in some embodiments, both the respiratory assistance device 202 and the respiratory assistance device controller 206 may include separate actuators.

[00138] In general, the generated oscillation pressure signal can also be controlled to deliver a desired pressure. In some cases, it may be preferable to produce pressures (i.e., amplitude of the generated oscillation signal) that do not exceed a peak-to-peak value of about 0.01 cm H2O to about 2 cm H2O. In some other cases, the pressure can be chosen based on the oscillation frequency or on Petition 870250101201, dated 05 / 11 / 2025, page 74 / 192 66 / 172 sensitivity and accuracy of the flow sensor and / or pressure sensor.

[00139] The inspiratory tube 204 and the expiratory tube 204' can be combined before reaching the user 210 at a junction using a tube adapter 218 connected to a breathing tube 208. Subsequent to the tube adapter 218, the combined airflow can be perceived to determine airflow parameters such as air flow rate and air pressure. In this example embodiment, the sensors used comprise a flow transducer 220 and a pressure transducer 221. It should be noted that the flow transducer 220 may also be called a flow transducer or an airflow transducer. The type of sensor used for the transducers 220 and 221 can be any suitable transducer device, including but not limited to, ultrasonic, pneumatic or piezoelectric transducers, for example. In some applications, airflow parameters can be measured and calculated by recording the pressure drop using a pneumotachograph, which acts as the sensor.

[00140] The outputs of the flow transducer 220 and the pressure transducer 221 can be pre-conditioned before being further processed and analyzed. For example, the output signals of transducers 220 and 221 can be amplified by an appropriate amplifier 224 to obtain Petition 870250101201, dated 05 / 11 / 2025, page 75 / 192 67 / 172 as desired signal amplitudes. For example, in some embodiments, amplifier 224 may be a synchronous detection amplifier that can be used to reduce signal noise to help focus on the frequency of interest. It should be noted that separate amplifiers may be used for each measured signal or a dual-channel amplifier may be used.

[00141] The amplified signal can then be filtered to remove extraneous frequency domain content. The filter can be a low-pass filter, a high-pass filter, a band-pass filter, or any other appropriate filter.

[00142] After the signals have been amplified and filtered, the signals are received by processor 228 for further processing and analysis in order to apply one or more predictive models in order to generate a control signal 212 which is provided to the respiratory assistance device 202 to adjust its operation, as described in more detail below.

[00143] In some embodiments, the 228 processor may be a programmable device, such as a programmable microcontroller or a field programmable gate array (FPGA). In other embodiments, the processor may be part of a single-board computer system platform, such as the Arduino platform or the Raspberry Pi platform. Petition 870250101201, dated 05 / 11 / 2025, page 76 / 192 68 / 172 In still other embodiments, signal filtering can be performed using processor 228, as well as by using digital signal processing (DSP) techniques, such that a separate filtering device 226 may not be necessary. In some embodiments, there may be multiple processors performing dedicated functions.

[00144] In the embodiment of Figure 2, the respiratory assistance device controller 206 includes the processor 228. In some embodiments, the respiratory assistance device controller 206 can be separated from the processor 228. For example, in some embodiments, the respiratory assistance device controller 206 can be physically coupled to the respiratory assistance device 202, while in other example embodiments, the respiratory assistance device controller 206 can be wirelessly coupled to the respiratory assistance device 202, for example, via the network 114. For example, in some embodiments, the respiratory assistance device controller 206 can be implemented in a mobile phone, in a separate device located in the same room as the respiratory assistance device 202, or in a remotely located device.

[00145] Control signal 212 can be provided for Petition 870250101201, dated 05 / 11 / 2025, page 77 / 192 69 / 172 the respiratory assistance device 202 using any method known to those skilled in the art. For example, the control signal 212 may be provided via a wired connection. However, in other implementations, the control signal may be wirelessly communicated to the respiratory assistance device 202, in which case the system 200 may include a transmitter or a transceiver (such as a Wi-Fi or Bluetooth transceiver).

[00146] The measured airflow parameters, certain indices and / or control signal 212, may also be displayed on an optional display 230 provided on the respiratory assistance device controller 206. The display 230 may be, but is not limited to, an LCD display, such as that for a tablet or smartphone type device. In some embodiments, certain indices may include indices of the user's body position, the user's average temperature and / or one or more of the indices described in US patents 11,633,560 and 11,612,708.

[00147] The 200 system also includes a memory unit 229 which may include RAM, ROM, one or more hard disks, one or more flash drives or some other suitable data storage elements, such as disk drives, etc. The memory unit 229 stores program instructions for an operating system 229a, a device control module 229b, one or more files Petition 870250101201, dated 05 / 11 / 2025, p. 78 / 192 The memory 229 may also include device profile data 229g, the user model 120, and the device model 122. The simulation module 229f can be used to perform digital twin simulations, while the device profile data 229g stores data that includes various settings and program instructions to configure the system 200 to operate the respiratory assistance device as if it were a CPAP device, an APAP device, and similar devices, as described in this document. It should be noted that some of these modules may be optional, depending on the modality.Although some of these components are shown residing on a cloud server, as in Figure 1B, in at least one embodiment they can also be stored locally on the system and uploaded to / transferred from the cloud server when it is being updated. Other software instructions may be included, as is known to those skilled in the art.

[00148] The 229b device control module comprises software instructions that, when executed, configure the 228 processor to operate in a certain way. Petition 870250101201, dated 05 / 11 / 2025, page 79 / 192 71 / 172 specifically designed to implement various functions, processes, and methods for system 200. For example, the device control module 229b may include program instructions to sense various data from sensors 223, perform various calculations or processing using the perceived data, and then apply one or more of the machine learning models 229e to generate a control signal in order to control the respiratory assistance device 202 to provide updated airflow to the user, which may be to prevent a respiratory event from occurring and / or provide updated therapy to the user.As another example, the 229b device control module may include program instructions to identify false negative predictions based on the performance of the 229e machine learning models and extract false negative data corresponding to the identified false negative predictions, which can be used to update the user's custom model.

[00149] Machine learning models 229e include different types of models that are used to classify or predict different phenomena related to sleep or breathing. Machine learning models 229e can be based on the use of different machine learning methods, such as one or more Random Forest classifiers, a classifier Petition 870250101201, dated 05 / 11 / 2025, page 80 / 192 72 / 172 linear logistic algorithm, the K-Nearest Neighbors algorithm, the Feature-Based Dissimilarity Space Classifier (FDSC), Neural Networks, and support vector machines. Alternatively, deep learning or Convolutional Neural Networks (CNNs) can be used. For reduced computational time, a Random Forest classifier may be preferable as the machine learning model.

[00150] One or more of the 229c machine learning models, in general, also include a trained predictive model that can, for example, be determined on the general population. In at least one modality, there may be multiple trained predictive models that may have been determined based on an age group (e.g., deciles such as 31 to 40, 41 to 50, etc.), user gender (e.g., male or female), respiratory condition (e.g., user has severe sleep apnea or moderate sleep apnea, etc.).

[00151] One or more of the machine learning models 229e generally also include a custom predictive model to adjust the airflow provided by the breathing assistance device 202 to the user 210, a user model and / or a device model for the Petition 870250101201, dated 05 / 11 / 2025, page 81 / 192 73 / 172 respiratory assistance device 202. An example embodiment of the personalized predictive model is described in more detail with respect to method 300 in Figure 3A and the example data in Figures 4A-4C. An example embodiment of the user and respiratory assistance device 202 models is described in more detail with respect to Figure 7.

[00152] Certain input features are provided for the machine learning model that has high predictive power. For example, input features may include the Power Spectral Density (PSD) of the measured airflow (e.g., airflow and air rate are synonyms). The PSD can be measured across the entire spectral range of the sampled sensor data. Alternatively, there may be cases where the PSD across a more specific frequency range can be measured, such as between about 0.01 Hz to about 20 Hz, which contains most of the information for the PSD. Other potential input features may include one or more of the PSD of the measured air pressure, the PSD of the measured resistance and reactance using FOT, and, for a given PSD, the number of peaks in the PSD, and the power at a certain frequency.In other cases, the input characteristics for a given measured or determined signal may be one or more of an index for the signal; coefficients of a... Petition 870250101201, dated 05 / 11 / 2025, page 82 / 192 74 / 172 linear least squares regression in windows along the signal; the absolute energy of the signal (e.g., the sum of the squared amplitudes of the signal); the minimum and / or maximum amplitude of the signal; the standard deviation, skewness, and / or kurtosis of the signal; the number of peaks in the signal; the autocorrelation of the signal and the absolute value of the FFT coefficients of the signal. The signal can be measured airflow, measured air pressure, reactance obtained from the FOT method, resistance obtained from the FOT method, or impedance obtained from the FOT method. The actual input characteristics depend on the particular type of machine learning model. For example, in some embodiments, the input provided to the machine learning model includes sensor data measured from one or more sensors described in this document.

[00153] The 229e machine learning models can be trained in several ways. For example, data obtained from patients can be pre-processed (as described in this document for Figure 2) and divided into a training set, a test set, and a validation set. For example, the training set, the test set, and the validation set can comprise using amounts of data in the proportion of 70%, 15%, and 15%, respectively. The training data is used to train the machine learning model, from Petition 870250101201, dated 05 / 11 / 2025, page 83 / 192 75 / 172 so that it accurately predicts a desired parameter. Machine learning models can be initialized with some initial parameters and then trained using the training set to determine which machine learning model parameters and input features provide the highest accuracy. The machine learning model can then be tested with the test dataset and the accuracy noted. The machine learning model parameters can then be adjusted to maximize the accuracy of the machine learning model on the test dataset. The machine learning model can then be tested with the validation dataset and the accuracy noted.

[00154] The data used to develop, test, and validate the 229e machine learning models described in this document were obtained from the QEII Health Sciences Center in Halifax. The data included thousands of respiratory failure events comprising sleep apneas of various categories, including obstructive sleep apnea, central sleep apnea, and hypopnea. The data were divided into test, training, and validation datasets. In some modes, the data used to generate the 229e machine learning models are customized for the user, for example, false negative data generated while the user was using the 202 respiratory assistance device. Events of Petition 870250101201, dated 05 / 11 / 2025, page 84 / 192 76 / 172 Obstructive Sleep Apnea data were extracted, and data were also obtained from baseline and pre-apnea periods. Data were also taken based on different sleep stages determined by EEG. The data include measured airflow pressure, measured airflow rate, and resistance and reactance determined using the pressure and airflow rates measured during the FOT method at different frequencies. The data also included other PSG measurements.

[00155] The input / output module 229d receives input data obtained from the sensors, pre-processes the input data, and / or provides output data (or signals, such as control signal 212 from processing by the device control module 229b) which are then sent to the corresponding hardware. The input / output module 229d can, for example, operate in conjunction with the device control module 229b to communicate data (or signals) between one or more of the processor 228 and one or more of the sensors 220 to 223 and the display 230. In various embodiments, the functionality of the input / output module 229d can be implemented, for example, using a combination of hardware, embedded software, and / or software.

[00156] Data files 229c can store any temporary data (for example, data that is not needed after the assistive device is removed). Petition 870250101201, dated 05 / 11 / 2025, page 85 / 192 77 / 172 breathing assistance device 202 was used) or permanent data (e.g., data saved for later use), such as user data (e.g., a user ID), settings for the breathing assistance device 202, pre-processing or other processing settings, including variables and calibration data, and various machine learning models, such as at least one custom model, at least one user model, and / or at least one device model. The 229c data files may also include various user data for each user who uses the breathing assistance device 202, such as identification data, respiratory physiological data, and user data recorded during the use of the breathing assistance device 202.For example, data files 229c may include files with information for user model 120, device model 122, and device profiles 229g, which are shown separately for clarity.

[00157] In at least one embodiment, the respiratory assistance device controller 206 can be configured to operate continuously to monitor the pressure and flow rate of the airflow supplied to the user 210 to allow detection and / or prediction of the user's breathing status for constant adjustment of the operation of the respiratory assistance device 202. Do this Petition 870250101201, dated 05 / 11 / 2025, page 86 / 192 78 / 172 may allow real-time or near real-time adaptive adjustments to be made to minimize or avoid any respiratory failure experienced by the user 210. In other embodiments, the respiratory assistance device controller 206 may alternatively be controlled to operate intermittently, for example, at an adjusted time interval. In at least one embodiment, the adjusted time interval may be every few milliseconds, every few seconds, every 30 seconds, or every hour. However, other adjusted time intervals may be used. Such operating conditions may be preferred if the respiratory assistance device controller 206 is battery operated to help extend the operational life of the respiratory assistance device 202.

[00158] The sensors may also include at least two additional sensors 222 and 223. Sensor 222 may be a CO2 gas sensor. Sensors 223 may include one or more user sensors that may be placed in certain locations on the user 210. For example, sensors 223 may include sensors that are used to obtain at least one physiological signal and / or at least one neurological signal. For example, physiological signals include one or more ECG, EOG, and EMG signals and blood parameters, including inflammatory markers (as determined by means of a test of Petition 870250101201, dated 05 / 11 / 2025, page 87 / 192 79 / 172 blood (which can be obtained from data sent from a laboratory or a sensor) and neurological signals include one or more EEG and Peripheral Neurophysiological Examination (PNE) signals. These signals can be measured using known electrodes that are placed at certain locations on the user 210, as is known to those skilled in the art. Because a user's position can affect the therapy provided by the respiratory assistance device, the sensors 223 may additionally include positional sensors placed at certain locations on the user 210 to determine the user's position. In at least one embodiment, a CO2 signal can be obtained and / or light sensors can also be placed on the user's skin and can also be used to measure blood-related parameters.In at least one embodiment, the sensors 223 may additionally include environmental sensors, placed in different locations in the room where the breathing assistance device 202 is used. For example, environmental sensors may include: (a) sound sensors, such as microphones that can record sounds from the user and / or the environment (e.g., snoring, ambient breathing noise, external noise); (b) motion detection and / or recording systems, including (I) image and video acquisition and / or recording systems, and / or (II) recording systems. Petition 870250101201, dated 05 / 11 / 2025, page 88 / 192 80 / 172 by radar and / or electromagnetic devices that can detect and / or record movement (e.g., breathing, sleeping position, etc.); (d) humidity sensors; (e) alcohol and / or substance sensors that can measure a concentration of alcohol or substance; (f) position sensors, such as gravity sensors, gyroscopes, accelerometers, that can detect a user's position; (g) light sensors that can detect ambient light; (h) pressure sensors placed, for example, on the surface on which the user is sleeping to measure movement and position; (i) any other sensor that can detect environmental parameters and user-related parameters; or any operable combination of items (a) through (i).In at least one embodiment, the sensors 223 may additionally include device sensors placed on, or within, the respiratory assistance device 202 to measure the properties of the respiratory assistance device 202, as previously described. These sensors may be used, for example, to monitor the health of physical components of the respiratory assistance device. For example, a humidity sensor may be placed on the respiratory assistance device 202 to monitor a humidity level of the device. The sensor data may be pre-processed, as is known, by the different channels of the amplifier 224 and the filter 226 to reduce noise for these signals. Petition 870250101201, dated 05 / 11 / 2025, page 89 / 192 81 / 172 particular before these signals are sent to processor 228 for further analysis. The adjustments for signal amplification and filtering are known to those skilled in the art.

[00159] The system 200 also includes, in general, a network communication module 232 that enables communication with other devices via a network 114. For example, the respiratory assistance device controller 206 can send data to remote devices via the network communication module 232 and the network 114. The data sent to remote devices may include, for example, sensor data, processed data and / or determined indices, false negative data from one or more machine learning models 229e and models used by the respiratory assistance device 202, as described in more detail in this document.

[00160] Now, with regard to Figure 3A, this document shows a flowchart of an example embodiment of a 300 method for generating a custom predictive model to adjust airflow provided by a breathing assistance device for a user. The 300 method can also be used to receive data from the sensor and use the data from the sensor to generate a real-time estimate and prediction of the user's breathing based on the custom predictive model for Petition 870250101201, dated 05 / 11 / 2025, pp. 90 / 192 82 / 172 adjust the airflow provided by the respiratory assistance device 202 to the user. Method 300 can be performed by the controller processor 206 during the execution of software instructions from the various modules described previously. However, in other embodiments, method 300 can be performed by other processors or another applicable device. For example, in some embodiments, method 300 can be performed by a combination of processors, including, for example, processor 228 of the respiratory assistance system 200 and a remote processor. For ease of explanation, the elements represented in the respiratory assistance system 200 should be used in the description of the various steps of method 300. For example, method 300 can be implemented by processor 228 of the respiratory assistance system 200.However, it is understood that this technique can be used in the controller of the integrated respiratory assistance device 206 or another applicable device.

[00161] Method 300 can begin when the respiratory assistance device 202 has been activated and is supplying an airflow to the user 210. Starting at act 302, a trained predictive model is implemented in the respiratory assistance device controller 206. For example, the trained predictive model can be saved in the memory unit 229 of the respiratory assistance device controller. Petition 870250101201, dated 05 / 11 / 2025, pp. 91 / 192 83 / 172 breathing 206. The trained predictive model can be generated according to the details discussed in relation to Figure 2. For example, the trained predictive model can be determined on the general population. In at least one modality, there may be multiple trained predictive models that may have been determined based on an age group (e.g., deciles such as 31 to 40, 41 to 50, etc.), user sex (e.g., male or female), respiratory condition (e.g., user has severe sleep apnea or moderate sleep apnea, etc.). In at least one modality, the trained predictive model can receive sensor data as input. In at least one modality, the trained predictive model can receive determined characteristics as input. The determined characteristics can be determined based on the sensor data.In at least one modality, the characteristics are determined based on sensor data from a specific population and used as input in one or more models for a general population. In at least one modality, the characteristics are determined based on sensor data from a specific user and used as input in one or more models for that specific user. For example, if it is determined that an 80% threshold in predicting obstructive sleep apnea performs better than 70% in patients after coronary revascularization, or another type of surgery, this characteristic... Petition 870250101201, dated 05 / 11 / 2025, page 92 / 192 84 / 172 of having surgery could be an input in the predictive model. This feature can be tested in a digital twin simulation (further details provided below) on other users who are from the same population and, if the results are also good, the update can be applied to the entire population of the same age range who underwent this surgery. Or, if pre-apnea intervention with 3 cm H2O for a period of 3 seconds results in a reduction in AHI for COPD patients in PAP, this feature can be used as an input in the predictive model and can be applied to other patients who fit the same category. This can be applied in the digital twin simulation. Patient categorization can occur through clustering, or limiting through signal processing techniques, or machine learning for classification of different populations to which the new features apply.This can be applied in digital twin simulation. Using data from the sensor, the trained predictive model can generate a real-time estimate of the user's current breathing state and a prediction of the user's future breathing state within a predicted time period. In at least one mode, the predicted time period is immediately before a breathing event. The predicted time period can be limited by the data sampling frequency period, 1 / f. Petition 870250101201, dated 05 / 11 / 2025, pp. 93 / 192 85 / 172 seconds, where f is the sampling frequency. In at least one mode, the predicted period is the maximum time in which an event prediction can reasonably be made. Other predicted time periods may be used. In some modes, the trained predictive model receives other types of data as input, such as processed data from the sensor or values ​​calculated and / or determined based on sensor data. This is discussed in more detail in relation to Acts 304 and 306 below.

[00162] In act 304, sensor data is received from one or more sensors coupled to the respiratory assistance device controller 206. In some example embodiments, the sensor data includes flow rate and pressure measured from the flow transducer 220 and the pressure transducer 221, respectively, of the airflow (including the disturbance) that is sent to the user 210. The sensor data may be amplified by the amplifier 224 and filtered, for example, by filters 226. The filter may include a low-pass filter, a high-pass filter, a band-pass filter, or another suitable filter. Also noted previously, in some embodiments, the passband may be made sufficiently narrow so that a notch filter may be used in substitution when a single frequency is used in the FOT measurement. If Petition 870250101201, dated 05 / 11 / 2025, pp. 94 / 192 86 / 172 If a FOT measurement is not being performed, filtering can be applied to generally remove noise, as is known to those skilled in the art.

[00163] In some embodiments, the sensor data may include data from other sensors, such as gas sensor 222 and / or one or more sensors 223, including PSG sensors. For example, the sensor data may include at least one physiological signal and / or at least one neurological signal, as described previously. As another example, the sensor data may include a signal from one or more environmental sensors, as described previously.

[00164] After the signals have been processed by applying amplification and filtering, the processed signals are received by the 228 processor for further processing and analysis. For example, in at least one mode, further processing may include processing the signals to have a zero mean and unit variance. In some modes, the 228 processor may use the sensor data to calculate and / or determine various values ​​or indices representative of the user's breathing state. For example, in some modes, a peak-to-peak value or associated PSD may be calculated. In some modes, indices such as indices of the user's body position or average temperature may be calculated. In Petition 870250101201, dated 05 / 11 / 2025, pages 95 / 192 87 / 172 In some modes, sensor data is provided as input to the predictive model. In other modes, calculated and / or determined values ​​or indices are provided as input to the predictive model.

[00165] In Act 306, the trained predictive model is applied to the input data to generate the real-time estimate and prediction of the user's breathing state. In some modalities, the trained predictive model generates a real-time estimate and a prediction of the user's breathing state for each point in time of the input data. For example, in at least one modality, the trained predictive model is trained on a large dataset to perform classification. The training data, for example, may be divided into training, test, and validation data in proportions of 70%, 15%, and 15%, respectively. At each point in time, the trained predictive model may use a memory of previous time points for context to classify the current breathing state and the predicted future breathing state based on the current point in time of the data.The trained predictive model can be developed using various means, such as, but not limited to, a convolutional neural network (CNN), a deep neural network (DNN), layers of long-term memory (LSTM), a reinforcement learning model (RL). Petition 870250101201, dated 05 / 11 / 2025, pp. 96 / 192 88 / 172 (from the English Reinforcement Learning) or an operable combination thereof, for example. Other implementations of the trained predictive model may be used. This generally involves optimizing the architecture of the AI / machine learning model that is selected by determining, for example, the number of LSTM layers and the number of units per layer, and the learning rate parameters (e.g., initial learning rate, maximum learning rate, and functional form of the learning rate (which decreases as training progresses)) that resulted in acceptable / best performance.In an example embodiment for an LSTM machine learning model, the following parameters were verified to result in acceptable performance: LSTM layers: 3 units per layer; 32 maximum learning rate; 1 initial learning rate; 10e-5; learning rate function (as the maximum learning rate decreases with training time); 1 / 2(x-1) where x = training epoch.

[00166] In some modalities, the prediction represents a probability of the user's predicted breathing state over a predicted period of time. For example, in some modalities, the predicted time period may be any of 20 seconds, 40 seconds, or 60 seconds; however, other predicted time periods may also be used. In some modalities, the prediction comprises a Petition 870250101201, dated 05 / 11 / 2025, p. 97 / 192 89 / 172 plurality of probabilities corresponding to a plurality of breathing states, such that each breathing state is represented by a probability that the user's future breathing state within the predicted time period will correspond to the respective type of breathing state. Types of future breathing states may include: (a) normal breathing, (b) a respiratory failure event including, but not limited to, obstructive apnea, central apnea, hypopnea, obstructive hypopnea, and / or respiratory effort-related stimulus, or (c) an unclassified event. Other breathing states and respiratory events may be included. In some embodiments, a normal breathing state comprises a state in which a respiratory failure event is not detected and / or predicted for a certain period of time.

[00167] In some modalities, the real-time estimate represents a probability of the user's current breathing state. In some modalities, the real-time estimate comprises a plurality of probabilities corresponding to a plurality of breathing states, such that each breathing state is represented by a probability that the user's current breathing state corresponds to the respective type of breathing state. As with prediction, types of Petition 870250101201, dated 05 / 11 / 2025, pp. 98 / 192 90 / 172 Real-time estimation of breathing states may include: (a) normal breathing, (b) a respiratory failure event including obstructive apnea, central apnea, hypopnea, obstructive hypopnea, and / or respiratory effort-related stimulus, or (c) an unclassified event. Other types of breathing states and respiratory failure events may be used. In some modalities, a normal breathing state comprises a state in which a respiratory failure event is not detected and / or predicted for a certain period of time. In some modalities, the real-time estimate is generated by determining, based on the input data, a probability that the user's current breathing state (e.g., corresponding to the current time period) comprises a breathing event and a probability that the user's current breathing state comprises normal breathing.In at least one modality, the current time period can be the last point in time, the last 10 points in time, the points in time of the last hour, the points in time of the last day, etc., where a point in time is a sample. Other current time periods may be used.

[00168] In some embodiments, an adjustment in the airflow provided by the breathing assistance device 202 to the user 210 may be made based on real-time estimation and / or prediction. For example, in some Petition 870250101201, dated 05 / 11 / 2025, pp. 99 / 192 91 / 172 modes, one or more of the predicted probabilities and / or the real-time estimated probabilities can be compared with a threshold. Now, with regard to Figure 3B, an example flowchart of the data flow through the respiratory assistance system 100 is shown in this document. The input data 316 may include, for example, user data 318, medical professional data (e.g., from a doctor, healthcare professional, hospital, pharmacy, etc.), respiratory assistance device data 320, environmental sensor data 322, and / or predetermined thresholds 324. User data 318 may include, for example, any suitable user-related data 110 from the respiratory assistance device 202, as described in this document.Some examples of user data include any combination of alcohol consumption, drug use, medication, sensations related to being under medication, insomnia, rest, fatigue, and the like. Data from the respiratory assistance device 320 may include any suitable data from the respiratory assistance device 202, such as, for example, air pressure and / or airflow data. Data from the environmental sensor 322 may include, for example, data from one or more environmental sensors, as described in this document, such as, but not limited to, Petition 870250101201, dated 05 / 11 / 2025, pages 100 / 192 92 / 172 room air humidity, sounds in and outside the room, and / or temperature in the room where the user is sleeping. Predetermined limits 324 may include, for example, limits that are pre-selected based on real-time estimates and / or expected prediction probabilities (e.g., based on data from the user group similar to user 110). Although some input data types 316 are shown in Figure 3B, it is understood that input data 316 are not so limited and may include other input data types as described in this document and / or any, all, and / or none of the input data types 316 shown.

[00169] Input data can be provided to the predictive model 326 to generate real-time estimation probabilities 328 and / or prediction probabilities 330, as described in this document. The predictive model 326 may include, for example, the trained predictive model and / or the custom predictive model, as described in this document (for example, the custom predictive model is the trained predictive model after being customized / further trained for a particular user). Input data 316 can be further provided to one or more threshold models 332, which can determine one or more calculated thresholds 334. For example, the threshold model 332 can apply known techniques of Petition 870250101201, dated 05 / 11 / 2025, pages 101 / 192 93 / 172 signal processing on the input data 316 to determine one or more calculated limits 334. In some embodiments, one or more of the calculated limits 334 can be customized for the user 110. For example, if the flow limitation limit, or snoring, or fragmentation, or oxygen drop reaches a certain amount (after filtering the data within the correct frequency range after reception from the sensors) an action may occur which may include reducing or increasing the air pressure that is provided to the user.

[00170] Real-time estimation probabilities 328, prediction probabilities 330, calculated limits 334, and / or predetermined limits 324 can be used to limit comparisons 336. In some embodiments, one or more limit comparisons 336 can be performed. For example, Figure 3B shows three different limit comparisons 336. In other example embodiments, zero, one, two, or more than three limit comparisons 336 can be performed. For each limit comparison 336, one or more of the real-time estimation probabilities 328 and / or prediction probabilities 330 are compared with the limit 338. Each of the limits 338a-338c can be associated with, for example, a different type of respiratory event. The limit model can operate continuously until the limits are customized for the user. For example, Petition 870250101201, dated 05 / 11 / 2025, p. 102 / 192 94 / 172 If the AHI is below 2 apnea events per hour and the patient's respiratory, cognitive, and cardiological health is acceptable, the limits can be considered personalized (i.e., optimized) for this particular user. Periodically, the limits can be checked to determine if changes are necessary, such as when the user's physiological health has changed in such a way that adjustments are needed for more effective personalization. For example, if health results or one of the sensors related to the user's heart, brain, or respiratory system health, or even the user's own feedback (e.g., through manual input) shows that they are not doing well, then the limits can be adjusted. Based on the comparison, the 206 controller can perform one or more actions.For example, if comparison with limits 338a-c indicates that the real-time estimation probability 328 and / or the prediction probability 330 meet the boundary condition, controller 206 can take actions 1A-1C shown in 340a-340c in Figure 3B, respectively. Alternatively, if comparison with limits 338a-338c indicates that the real-time estimation probability 328 and / or the prediction probability 330 do not meet the boundary condition, controller 206 can take actions 2A-2C shown in 342a-340c of Figure 3B, respectively. Although Figure 3B shows... Petition 870250101201, dated 05 / 11 / 2025, p. 103 / 192 95 / 172 only one action for each result of each threshold comparison, it is understood that more than one action can be performed. For example, action 1A (340a) may include one or more actions and / or sub-actions. In some modalities, the one or more actions 340a-340c and 342a-342c may be customized for the user 110. For example, if a user's / patient's apneas are predicted with a probability above 70%, then the air pressure for the airflow provided to the user increases to prevent apnea, but if the probability is not above 70% and the real-time probability estimate shows no change in breathing status, then the air pressure provided to the user remains the same. In another patient, this threshold may be 80%.As another example, if the real-time estimate shows that the probability of snoring is 20% higher than usual, then the air pressure for the airflow to the user may increase slightly to further open the user's airways and prevent snoring. As another example, if the probability of central apnea is increasing by more than 50%, the airflow pressure to the user may decrease to prevent further central apneas. These limits may differ for different users / patients. It should be noted that the terms user and patient may be used interchangeably.

[00171] In this way, in one or more modalities, the Petition 870250101201, dated 05 / 11 / 2025, pp. 104 / 192 Controller 206 can adjust the airflow provided by the respiratory assistance device 202 to the user 210 based on a comparison of prediction probabilities with a threshold (e.g., a probability threshold). In at least one mode, the threshold can be predefined as 50%, 70%, or 90% probability, for example. Other predefined thresholds can be used. In some modes, the threshold can be a calculated threshold 334, as described in this document. For example, in an example mode, an adjustment can be triggered when a predicted probability of at least one respiratory failure event is above a threshold of 75%. As another example, in an example mode, an adjustment can be triggered when a predicted probability of normal breathing is below a threshold of 50%.In at least one mode, the limit is optimized and personalized for the user (for example, several decisions for a specific user may require a specific limit for that user, and limits can be updated / modified if a particular user's health status changes). In at least one mode, the limit can be updated at various time intervals, for example, in real time (on the order of a few seconds), near real time (on the order of a few minutes), hourly, daily, weekly, monthly, etc. In this way, the limit can be adaptive. Petition 870250101201, dated 05 / 11 / 2025, page 105 / 192 97 / 172 to be more responsive to the user's current sleep behavior and / or physiological health.

[00172] Alternatively, in some embodiments, one or more of the real-time estimation probabilities may be compared to a threshold. Based on the comparison, the controller 206 may adjust the airflow provided by the breathing assistance device 202 to the user 210. In some embodiments, the threshold may be predefined as 50%, 70%, or 90% probability level, for example. Other predefined thresholds may also be used. In some embodiments, the threshold may be a calculated threshold 334, as described in this document. For example, in one example embodiment, an adjustment may be triggered when the real-time estimation probability corresponding to an obstructive apnea breathing event is above a threshold of 75%. As another example, in at least one example embodiment, an adjustment may be triggered when a real-time estimation probability of normal breathing is below a threshold of 50%.In at least one mode, the limit is optimized and personalized for the user in a manner similar to that described above. In at least one mode, the limit can be updated at various time intervals, for example, in real time (on the order of a few seconds), near real time (on the order of a few minutes), hourly, daily, weekly. Petition 870250101201, dated 05 / 11 / 2025, pp. 106 / 192 98 / 172 monthly, etc. In this way, the limit can be adaptive to be more reactive to the user's current sleep behavior and / or physiological health.

[00173] In at least one mode, the various limits that are used may be different for each breathing state, for example, the normal breathing state and each of the respiratory stages of breathing, and these limits may also be customized for each user. These various limits may also change over time. Adjusting the various limits in this way by the controller or remote process allows more personalized therapy to be provided to the user.

[00174] In at least one embodiment, both real-time estimation and prediction probabilities can be used to determine a control signal for the controller 206. For example, in at least one embodiment, real-time estimation and prediction can have respective weights representative of the importance of the respective value for the control signal. For example, in at least one embodiment, real-time estimation may be more critical in determining the airflow to be provided to the user by the breathing assistance device 202, and thus real-time estimation will have a greater weight than prediction during the determination of the control signal for the controller 206. As another example, in at least Petition 870250101201, dated 05 / 11 / 2025, page 107 / 192 99 / 172 In one modality, prediction may be more critical in determining the airflow to be provided to the user by the respiratory assistance device 202 and, in this way, prediction will have a greater weight than real-time estimation during the determination of the control signal for the controller 206. The combination of real-time estimation and weighted prediction can then be compared with a threshold for decision-making.

[00175] Returning now to Figure 3A, in Act 308, one or more false negative predictions can be identified. A false negative prediction occurs when the real-time estimate and / or the prediction fail to correctly indicate a current or future respiratory failure event when the user has, in fact, experienced a respiratory event. In some embodiments, identifying that the user has experienced a respiratory failure event can be determined, for example, based on sensor data. For example, a false negative prediction of the real-time estimate can be identified when the sensor data indicates that the user experienced a respiratory event in a current time period and the real-time estimate indicated a normal breathing state of the user in the current time period. In another example embodiment, a false negative prediction of the prediction can be identified when the sensor data indicates that the user Petition 870250101201, dated 05 / 11 / 2025, pp. 108 / 192 100 / 172 experienced a respiratory failure event in a current time period, and the prediction indicated a normal breathing state in a predicted time period, where the predicted time period corresponds to the current time period. A given breathing state can be indicated based, for example, on comparison with a threshold, according to the description provided.

[00176] In other example embodiments, a false negative prediction occurs when one or more of the real-time estimate or prediction fails to correctly indicate a particular type of respiratory failure event. For example, a false negative prediction of the real-time estimate might be identified when the sensor data indicates that the user experienced an obstructive apnea breathing event in a current time period and the real-time estimate indicated a central apnea breathing event in the current time period. In another example, a false negative prediction of the prediction might be identified when the sensor data indicates that the user experienced an obstructive apnea breathing event in a current time period and the prediction indicated a central apnea breathing event in a predicted time period, where the predicted time period corresponds to the current time period.A given respiratory state can be indicated based on, for example, the... Petition 870250101201, dated 05 / 11 / 2025, pp. 109 / 192 101 / 172 comparison with a limit, according to the description given.

[00177] In act 310, false negative data is received. In some embodiments, false negative data is received on a processor or device that is remote from the respiratory assistance device controller 206. For example, in some embodiments, the respiratory assistance device controller 206 sends false negative data to a server 118 via the network communication module 232 and the network 114. False negative data comprises, for each of one or more false negative predictions, the real-time estimate, the prediction, and a portion of the sensor data used to determine the real-time estimate and the prediction. In some embodiments, the portion of sensor data extends from a first point in time before the start of the false negative prediction to a second point in time after the false negative prediction shift.In some disciplines, the first and second time points fall within the range of 20 to 60 seconds. However, other first and second time points may be used.

[00178] In some modalities, the start of a false negative prediction is defined as the first point in time of a predetermined number of consecutive real-time estimates and / or false negative predictions. By Petition 870250101201, dated 05 / 11 / 2025, pages 110 / 192 102 / 172 For example, in some modes, the start of a false negative forecast is the first point in time of the first real-time estimate and / or false negative forecast of three consecutive real-time estimates and / or forecasts. Other predetermined numbers of consecutive real-time estimates and / or false negative forecasts may be used.

[00179] In other embodiments, the start of a false negative prediction is determined as the point in time at which the controller 206 generates a control signal 212, for example, to adjust or maintain the airflow provided to the breathing assistance device 202 based on the real-time estimate and / or prediction generated. In some embodiments, the offset of a false negative prediction is defined as the point in time at which a predetermined number of consecutive real-time estimates and / or correct predictions are generated. For example, in some embodiments, the predetermined number of consecutive real-time estimates and / or correct predictions may be three. Other predetermined numbers of consecutive real-time estimates and / or correct predictions may be used.

[00180] In some modes, false negative data is pre-processed. In at least one mode, false negative data is pre-processed to generate a Petition 870250101201, dated 05 / 11 / 2025, pp. 111 / 192 103 / 172 cleaner dataset for use in training or retraining predictive models. Preprocessing of false negative data may include one or more of: normalization using techniques known to those skilled in the art, weighting of the sensor data portion of the false negative data, principal component analysis, independent component analysis, top-down sampling, top-down sampling, frequency filtering, or manual inspection of the sensor data portion of the false negative data.

[00181] In Act 312, the custom predictive model is generated by retraining the trained predictive model using false negative data. The custom predictive model is customized for the user. The features identified in the sensor data may differ from user to user. In this way, a predictive model that is trained using group data only may not account for certain features of the sensor data that are unique to a given user. A custom predictive model that is retrained using data from a specific user may improve the model's detection and / or prediction performance for that user.

[00182] In some modalities, the custom predictive model is automatically generated / retrained at a predetermined frequency. For example, in some Petition 870250101201, dated 05 / 11 / 2025, pages 112 / 192 104 / 172 modes, the predetermined frequency can be real-time, every few minutes, every hour, daily, weekly, or monthly. Other predetermined frequencies can be used for how often the custom predictive model is retrained.

[00183] In some embodiments, the custom predictive model is generated after a minimum number of false negative predictions are identified. In one example embodiment, the minimum number of false negative predictions might be 500, and the custom predictive model is generated once at least 500 false negative predictions are identified. In some embodiments, the minimum number of false negative predictions might range from one to the total number of time points at which a real-time estimate and / or prediction is generated during a monitoring period in which the respiratory assistance device is used by the user. Another minimum number of false negative predictions may be used.

[00184] In some modes, the custom predictive model is generated after a predetermined period of time has elapsed. For example, in some modes, the custom predictive model is generated daily, weekly, or monthly. Other predetermined time periods may be used.

[00185] Overall performance of the predictive model, Petition 870250101201, dated 05 / 11 / 2025, pp. 113 / 192 105 / 172 improves with the amount of data used for training. In some modes, it is preferable to have larger amounts of false negative data to retrain the trained predictive model to generate the custom predictive model. In some modes, method 300 involves generating simulated false negative data to retrain the trained predictive model. For example, in some modes, simulated false negative data may be generated when the number of identified false negative predictions is less than the minimum number of false negative predictions. Simulated false negative data may be generated by applying one or more signal processing techniques to the sensor data. In some modes, one or more signal processing techniques are selected to preserve at least part of the sensor data information.For example, one or more signal processing techniques may include applying random perturbation, adding noise, magnitude scaling, or segment truncation. Those versed in the technique will realize that other data simulation techniques are possible.

[00186] However, in at least one modality, a larger amount of false negative data is not needed to retrain the trained predictive model. In some modalities, a portion of the false negative data for a specific user may no longer be relevant (for example, Petition 870250101201, dated 05 / 11 / 2025, pp. 114 / 192 106 / 172 due to a particular user classification (e.g., phenotype) and may be omitted during retraining. In at least one modality, such data omission occurs when the amount of false negative user data is sufficiently large.

[00187] In some embodiments, retraining the trained predictive model includes one or more transfer learning actions, fine-tuning of one or more parameters of the trained predictive model, adding one or more layers to the trained predictive model, performing reinforcement learning, or any combination thereof. In some embodiments, retraining the predictive model additionally uses at least some of the sensor data. For example, in some embodiments, retraining the predictive model uses sensor data (e.g., data from any of the pressure sensors, flow sensors, PSG sensors, environmental sensors, etc.) in addition to false negative data.

[00188] Transfer learning can include training the predictive model from scratch using false negative data. In example modalities that use transfer learning, the model weights are relearned, and the model architecture is preserved.

[00189] Fine-tuning one or more parameters of the trained predictive model may include continuation or Petition 870250101201, dated 05 / 11 / 2025, pages 115 / 192 107 / 172 resumption of training used to initially train the trained predictive model, but using only the false negative data. In example modalities using this technique, the model weights of the trained predictive model are not significantly altered, and the model architecture is preserved. The trained predictive model may include one or more nodes and one or more layers, and the one or more parameters that can be adjusted include a node type, a node selection, a node weight, a node activation, a node memory, a number of connections between nodes, an orientation of connections between nodes, an orientation of connections between layers, a layer type, a number of layers, a connection between layers, a number of inputs, a number of outputs, or an operable combination thereof, for example. Other parameters known to those skilled in the art may be used.

[00190] Adding one or more layers to the trained predictive model may include, for example, adding an output layer to the model, thereby altering the model architecture. In at least one modality, adding one or more layers to the trained predictive model may include adding one or more intermediate layers. In example modalities that use this technique, the model weights of the trained predictive model are not altered, and the weights associated with the additional output layer are learned. Petition 870250101201, dated 05 / 11 / 2025, pp. 116 / 192 108 / 172 through training using false negative data.

[00191] Reinforcement learning can involve training the predictive model to maximize a cumulative reward. For example, the predictive model might have one or more states represented by a time series window of physiological data, such as user airflow data. At each time step, a decision made by the predictive model (e.g., predicted value of the user's airflow data) can change the predictive model's state. The predictive model can receive a reward at each time step for the decision made by the predictive model. In this way, at each time step, the predictive model can use the user's airflow data (e.g., air pressure flow data) as input to ultimately output a probability distribution of all possible airflow data values ​​(e.g., all possible air pressure flow values) for the user.The airflow data value (e.g., air pressure flow value) with the highest probability is selected by the predictive model as the next time-step airflow data value (e.g., air pressure flow value). The predictive model then transitions to the next state (e.g., time step) and receives a reward for the completed time step. RL can be performed for training. Petition 870250101201, dated 05 / 11 / 2025, pp. 117 / 192 109 / 172 to maximize the cumulative reward received during the course of an episode with T time steps. Training can be done to obtain customized models that maintain normal breathing, improve health metrics, and / or decrease the air pressure needed for sleep therapy.

[00192] In some modalities, reinforcement learning may include implementing a Q-learning technique. In such cases, for pressure management, at any given time, Q-values ​​(airflow, set pressure) are estimated to optimize the metrics of interest. For example, for any given time, given an airflow window, there may be 100 (4.0-14.0, step size 0.1) possible pressure values ​​to choose from, from whose average for the current airflow state there are 100 Q-values ​​to estimate. Any airflow state in the pressure management dataset will have 100 possible Q-values ​​to estimate. As the model is trained, it will become better at estimating the Q-values ​​(a value to represent the expected future reward or a value to represent the probability that a certain pressure value is the ideal pressure).

[00193] In some modalities, reinforcement learning may include implementing a deep Q learning technique. Traditionally, Q values ​​are calculated and tracked in a large table (state, action). However, Petition 870250101201, dated 05 / 11 / 2025, pp. 118 / 192 110 / 172 since the number of possible states becomes incredibly large, this is not feasible. For pressure management, the number of possible airflow states / windows can become incredibly large. To address this problem, deep Q learning can be used, which involves using neural networks as Q-value estimators. For pressure management, a neural network takes the airflow window as input and feeds the probability distribution through all possible pressure values ​​and estimates the ideal pressure to maximize future reward / metrics of interest.

[00194] In some modalities, reinforcement learning may include implementing DDQN learning, which can be an improvement on the DQN approach. The DDQN approach uses two neural networks, where one of them is randomly updated to tackle overestimation problems with the original DQN.

[00195] In some modalities, reinforcement learning is performed offline (for example, training and / or retraining of the predictive model is based on data previously collected through offline reinforcement learning). Traditional / Online RL methods are based on an online learning paradigm, in which an algorithm / agent actively interacts with an environment. Petition 870250101201, dated 05 / 11 / 2025, pages 119 / 192 111 / 172 However, performing offline RL can be more efficient and ethical, since patients do not put themselves at risk by training and making adjustments to their personal predictive model in real time. Since there may be some ideal transitions in offline datasets, and many unseen transitions, CQL (Conservative Offline RL) training can be used to minimize the overestimation of Q values ​​in unseen transitions.

[00196] In some modalities, offline TL training can use a training dataset (state, action rewards) based on past airflow windows and pressure changes, from which rewards are determined for all (state, action) pairs. For apnea and flow limitations, negative rewards can be determined based on the reward function to optimize metrics of interest. For example, CQL with a deep learning model can be used for training to predict pressure values ​​that maximize future rewards. This training can also use different techniques to address trend in the direction of a policy used in existing data and improve generalization to unseen data.

[00197] Those versed in the technique will realize that other retraining techniques are possible.

[00198] In some modalities, the predictive model Petition 870250101201, dated 05 / 11 / 2025, pages 120 / 192 A customized 112 / 172 representation is generated by conditioning the trained predictive model using a summary representation for the user. The summary representation comprises user data. In some modes, the summary representation is generated based on data generated the previous day. In some modes, the summary representation is an average representation based on an exponential moving average of the summary representation. This allows for in-context adaptation of the trained predictive model using existing parameters, without requiring retraining or fine-tuning of the parameters. Conditioning the trained predictive model using the user's summary representation can include changing how the parameters used for inference and training can influence classification and / or learning. The summary representation can be used to alter the behavior of the trained predictive model without retraining the trained predictive model.

[00199] In some embodiments, the summary representation includes user characteristics such as weight, height, gender, sex, age, body mass index, apnea-hypopnea index, SpO2, type of respiratory assistance device mask, prescribed pressure to be provided by the respiratory assistance device, type of site, site elevation, or any operable combination thereof.

[00200] In some modalities, representation in Petition 870250101201, dated 05 / 11 / 2025, pages 121 / 192 113 / 172 Summary includes one or more statistical representations of the user's respiration based on sensor data. The one or more statistical representations of the user's respiration may include an average waveform of a user's breath, a variance for each time point in the sample in the average waveform, or one or more of a minimum, a maximum, a mean, a median, or a variance of airflow, air pressure, tidal volume, respiratory rate, SpO2, heart rate, sound and / or user movement, or any operable combination thereof.

[00201] In some embodiments, the summary representation includes one or more statistical representations of the user's environment based on sensor data. One or more statistical representations of the user's environment may include one or more minimums, maximums, averages, medians, or variances of one or more temperature, ambient CO2, or ambient O2 values.

[00202] In some modalities, the summary representation includes statistical representations of user characteristics, user breathing, and / or user environment.

[00203] In some modalities, the summary representation can be generated by the trained predictive model. An embedding layer of the model can be generated by a block in the model with learnable parameters based on the Petition 870250101201, dated 05 / 11 / 2025, pages 122 / 192 114 / 172 summary representation, where the block is optimized for the task of creating an embedding layer. The embedding layer can learn and change how the summary representation impacts the trained predictive model. A block in the model can include, for example, a single layer or a plurality of layers.

[00204] In some modalities, the trained predictive model is conditioned on the summary representation by providing the summary representation as input to the trained predictive model, in addition to the sensor data. In other modalities, the trained predictive model is conditioned on the summary representation by adapting a feature representation based on cross-attention to the summary representation. The feature representation is based on the sensor data. In other modalities, the trained predictive model is conditioned on the summary representation by providing the summary representation as input to a normalization block distinguished by a displacement factor and a scale factor for each feature corresponding to the normalization block. The displacement factor and the scale factor can be determined by a machine learning model conditioned on the summary representation.In at least one modality, the characteristics are abstracted, and may not be known prior to conditioning. In some modalities, the model... Petition 870250101201, dated 05 / 11 / 2025, pp. 123 / 192 115 / 172 machine learning is a multilayer perceptron.

[00205] In act 314, the custom predictive model is implemented in the respiratory assistance device controller 206. In embodiments where false negative data is received on a processor or device that is remote from the respiratory assistance device controller 206 and the custom predictive model is generated / updated on the device, the custom predictive model can then be sent / implemented on the respiratory assistance device controller 206 via the network communication module 232 and the network 114. For example, the custom predictive model can be generated on the server 118, sent via the network 114 to the network communication module 232, and implemented on the controller 206.

[00206] In some modes, the customized predictive model is updated in the 206 respiratory assistance device controller and implemented automatically at a predetermined update frequency. For example, in some modes, the predetermined update frequency may be real-time, near real-time, every few minutes, every hour, daily, weekly, or monthly. Other predetermined frequencies may also be used.

[00207] In some modalities, the predictive model Petition 870250101201, dated 05 / 11 / 2025, pages 124 / 192 116 / 172 customized determines one or more of a pressure rise rate, a pressure fall rate, and / or a pressure amplitude to adjust the airflow provided by the breathing assistance device to the user. In such modes, the pressure rise rate, the pressure fall rate, and / or the pressure amplitude can be customized for the user. For example, some users may require a faster pressure rise rate and / or a higher pressure amplitude than other users (e.g., some users may need a faster / slower ramp-up or ramp-down rate for the change in airflow and / or higher / lower airflow amplitudes).

[00208] In some embodiments, method 300 may include an additional step of monitoring the performance of the custom predictive model to determine if it is not operating at a safe operating rate and if it is not operating within a safe operating range, avoiding the use of the custom predictive model which may include operating based on OEM settings, while the custom predictive model is readjusted so that its use results in safe operation of the respiratory assistance device. This can be done by recording one or more operational characteristics and comparing the operational characteristic(s) Petition 870250101201, dated 05 / 11 / 2025, pages 125 / 192 117 / 172 recorded with a safety limit or safety range. For example, if the number of respiratory failure events is recorded, compared to a respiratory failure event safety limit and found to be above the limit, it can be determined that the custom predictive model results in the respiratory assistance device 202 being outside a safe operating range, and the use of the custom predictive model can be disabled. In alternative modes, two or more parameters can be monitored and compared to safety limits, such that when each of the two or more parameters is outside safe operating ranges. In at least one mode, the sensitivity, accuracy, F1 score and / or adjusted F1 score of the custom predictive model, which are described in more detail below, can be used for these safety comparisons.In at least one modality, if the digital twin simulations (described below) show that the associated sensitivities of impending apneas for the user are falling to below 50%, whereas the sensitivity for the user used to be above 65%, it can be determined that the custom predictive model results in the breathing assistance device being outside a safe operating range and the therapy provided by the breathing assistance device 202 can be reduced to one. Petition 870250101201, dated 05 / 11 / 2025, pages 126 / 192 118 / 172 version is safe until user customization is safely bringing the sensitivity back to 65%, as determined by digital twin simulations, at which point an over-the-air update can be promoted.

[00209] Now, with regard to Figure 4A, example sensor data 400 is shown, which is processed according to at least one embodiment described in this document, where a general trained prediction model is used. In this example, the sensor data comprises pressure sensor data 410a and airflow sensor data 410b. In other examples, embodiments may be used where the sensor data may comprise data measured by the other sensor types, as discussed previously. It can be seen from the example shown in Figure 4A that the user's breathing state includes an obstructive apnea breathing event during time period 412a. It can further be seen from the example embodiment shown in Figure 4A that the user's breathing state includes a normal breathing state during time periods 413a and 413b.

[00210] Now, with regard to Figure 4B, it shows an example of real-time estimation 402 and an example of prediction 404 generated by a general trained model using the sensor 200 data shown in Figure 4A. In the mode used for this example, the real-time estimation Petition 870250101201, dated 05 / 11 / 2025, pp. 127 / 192 119 / 172 comprises a probability of 414a, for each point in time of the sensor data, that the user's breathing state comprises normal breathing at each given point in time. In the mode used for this example, the real-time estimate additionally comprises a probability of 414b, for each point in time of the sensor data, that the user's breathing state comprises a respiratory failure event at each given point in time. In at least one mode, the probabilities of more than one of the breathing event types are close to zero and, in this way, can overlap in a visualization of the probabilities. The probability of a breathing type exceeding the threshold is the breathing type that is identified for this breathing event.

[00211] In this example, the prediction comprises a probability 416a, for each point in time of the sensor data, that the user's future breathing state in a predicted time period will comprise normal breathing. In this example embodiment, the prediction further comprises a probability 416b, for each point in time of the sensor data, that the user's future breathing state in a predicted time period will comprise a respiratory failure event.

[00212] In the example shown in Figure 4B, it can be seen that the real-time estimate 402 fails to detect Petition 870250101201, dated 05 / 11 / 2025, pp. 128 / 192 120 / 172 correctly predicts the obstructive apnea breathing event identified in the sensor data shown in Figure 4A. The time period 412b of the real-time estimate 402 indicates that the user's breathing state comprises normal breathing at the time of the obstructive apnea breathing event 412a identified in the sensor data 400. In the example shown in Figure 4B, it can also be seen that prediction 404 fails to correctly predict the obstructive apnea breathing event identified in the sensor data shown in Figure 4A. The time period 415 of prediction 404, which occurs a predicted time period before the obstructive apnea breathing event 412a identified in the sensor data, indicates that the user's breathing state will comprise normal breathing at the time of the obstructive apnea breathing event 412c.

[00213] Now, with regard to Figure 4C, an example of real-time estimation 406 and an example of prediction 408 generated by a custom predictive model using the sensor data shown in Figure 4A are shown therein. In this example, the real-time estimation comprises a probability 418a, for each point in time of the sensor data, that the user's breathing state comprises normal breathing at each given point in time. In this example, the real-time estimation additionally comprises a probability 418b, for each point in time of the sensor data. Petition 870250101201, dated 05 / 11 / 2025, pages 129 / 192 121 / 172 sensor, that the user's breathing state comprises an obstructive apnea breathing event at each given point in time. In this example, the real-time estimate additionally comprises the probabilities 418c-d, for each point in time of the sensor data, that the user's breathing state comprises other types of breathing events at each given point in time.

[00214] In this example, the prediction comprises a probability 420a, for each point in time of the sensor data, that the user's future breathing state in a predicted time period will comprise normal breathing. In this example, the prediction further comprises a probability 420b, for each point in time of the sensor data, that the user's future breathing state in a predicted time period will comprise an obstructive apnea breathing event. In this example, the prediction further comprises probabilities 420c, for each point in time of the sensor data, that the user's future breathing state in a predicted time period will comprise other types of breathing events at each given point in time.

[00215] In the example shown in Figure 4C, it can be seen that the real-time estimate correctly detects the obstructive apnea breathing event identified in the sensor data shown in Figure 4A. The time period 412d of Petition 870250101201, dated 05 / 11 / 2025, pp. 130 / 192 122 / 172 real-time estimate indicates that the user's breathing status comprises an obstructive apnea breathing event corresponding to the obstructive apnea breathing event identified in the sensor data in Figure 4A.

[00216] In the example embodiment shown in Figure 4C, it can be further seen that, at time point 422, the prediction correctly forecasts the obstructive apnea breathing event that will occur in the predicted time period corresponding to the obstructive apnea breathing event identified in the sensor data in Figure 4A.

[00217] In this way, it can be seen, from figures 4A to 4C, that the personalized predictive model provides improved detection and prediction of the user's breathing status compared to the general trained predictive model generated from non-personalized data (i.e., from general population data).

[00218] Now, with regard to Figure 4D, it is shown in the same example of flow data 432 and pressure data 434. It is also shown in the same example of probability data determined from a prediction example generated by a custom predictive model. This probability data example includes probabilities for each point in time of the flow and pressure data that the user's future breathing state in a predicted time period will comprise a breathing event of Petition 870250101201, dated 05 / 11 / 2025, pages 131 / 192 123 / 172 obstructive apnea 438, a central apnea event 442, a hypopnea event 440 or normal breathing 444. This example of probability data can be compared with a threshold 436 to make a decision.

[00219] Reference will now be made to figures 5A, 5B, 5C, 6A, and 6B. In some modalities, a sensitivity, an accuracy, an F1 score, and / or an adjusted F1 score can be determined for the general trained predictive model and / or the customized predictive model. Sensitivity may also be referred to as recall or true positive rate and is determined by true positives divided by the sum (or true positives and false negatives). For example, sensitivity can be determined by dividing the number of correct predictions of a respiratory failure event by a predictive model (e.g., true positives) by the total number of actual respiratory failure events (e.g., sum of true positives and false negatives). In this way, reducing the number of false negatives of a predictive model increases the sensitivity of the predictive model.Accuracy can be determined by dividing the number of correct predictions of a respiratory failure event by a predictive model (e.g., true positives) by the total number of respiratory failure events predicted by the predictive model (e.g., sum of true positives and false positives). Petition 870250101201, dated 05 / 11 / 2025, pages 132 / 192 124 / 172 In this way, reducing the number of false positives in a predictive model increases the model's accuracy. The F1 score can be determined based on the harmonic mean of the sensitivity and accuracy of a predictive model. For example, the F1 score can be determined using the following equation: F1 = 2 X precision x sensitivity precision+sensitivity (1)

[00220] In this way, a predictive model with low precision and / or low sensitivity will also have a low F1 score. The adjusted F1 score can be determined by applying one or more adjustments to the F1 score calculation, as described, based on a trend and / or imbalance in the data. For example, if, in a specific dataset, precision matters more than sensitivity, a weight can be applied to precision. An unbalanced dataset can result from having too much data in one class and not enough in the other class; therefore, weights can be applied in this case.

[00221] In this way, sensitivity, accuracy, F1 score and / or adjusted F1 score can provide an indication of the performance of a predictive model. In some modalities, the customized predictive model can be updated based on one or more of the sensitivity, accuracy, F1 score and / or adjusted F1 score of the general trained predictive model and / or the predictive model. Petition 870250101201, dated 05 / 11 / 2025, pages 133 / 192 125 / 172 customized. For example, in some modalities, the customized predictive model can be retrained and / or one or more parameters of the customized predictive model can be updated based on one or more of the sensitivity, accuracy, F1 score, and / or adjusted F1 score. In some modalities, the customized predictive model can be generated based on one or more of the sensitivity, accuracy, F1 score, and / or adjusted F1 score of the overall trained predictive model. For example, if a threshold changes from 40% to 60% for a user, and a higher adjusted F1 score is achieved, then this higher threshold can be used for this user, as it will likely result in better outcomes.

[00222] Now, with regard to Figure 5A, example sensor data comprising airflow sensor data 502a is shown therein. In other examples, embodiments may be used where the sensor data may comprise data measured by the other types of sensors, as discussed previously. It can be seen from the example shown in Figure 5A that the user's breathing state includes an apnea breathing event during time period 508a. It can further be seen from the example embodiment shown in Figure 5A that the user's breathing state includes a normal breathing state during time period 510a. Petition 870250101201, dated 05 / 11 / 2025, pp. 134 / 192 126 / 172

[00223] Figure 5A further shows example prediction data 504a generated by a general example trained predictive model using sensor data 502a and example prediction data 506a generated by a custom example predictive model using sensor data 502a. In the embodiment used for this example, prediction data 504a comprises a probability for each point in time of sensor data 502a that the user's future breathing state at a predicted time period will comprise a respiratory failure event based on the general example trained predictive model. In the embodiment used for this example, prediction data 506a comprises a probability for each point in time of sensor data 502a that the user's future breathing state at a predicted time period will comprise a respiratory failure event based on the custom example predictive model.

[00224] In this example, a 512a threshold of 0.5 is shown. In other example embodiments, other threshold levels may be used. A probability represented by the 504a or 506a prediction data reaching or exceeding the 512a threshold indicates that the respective predictive model predicts that the user's future breathing status in the predicted time period will comprise a respiratory failure event. As shown in Figure 5A, each of the Petition 870250101201, dated 05 / 11 / 2025, pages 135 / 192 127 / 172 prediction data points 504a and 506a exceed the 512a threshold before the respiratory failure event in time period 508a. In this way, each of the general trained predictive model and the custom predictive model correctly predicted the respiratory failure event occurring in time period 508a (i.e., true positive prediction).

[00225] Now, with regard to Figure 5B, the same example sensor data is shown, comprising data from the airflow sensor 502b. In other examples, embodiments may be used where the sensor data may comprise data measured by the other types of sensors, as discussed previously. It can be seen from the example shown in Figure 5B that the user's breathing state includes a normal breathing state for the duration of the time period 510b shown.

[00226] Figure 5B further shows example prediction data 504b generated by a general example trained predictive model using sensor data 502b and example prediction data 506b generated by a custom example predictive model using sensor data 502b. In the embodiment used for this example, prediction data 504b comprises a probability for each point in time of sensor data 502b that the user's future breathing state at a predicted time period will comprise a respiratory failure event based on Petition 870250101201, dated 05 / 11 / 2025, pages 136 / 192 128 / 172 general trained predictive model example. In the modality used for this example, the prediction data 506b comprises a probability for each point in time of the sensor data 502b that the user's future breathing state at a predicted time period will comprise a respiratory failure event based on the custom example predictive model.

[00227] In this example, a 512b threshold of 0.5 is shown. In other example embodiments, other threshold levels may be used. A probability represented by the 504b or 506b prediction data reaching or exceeding the 512b threshold indicates that the respective predictive model predicts that the user's future breathing status in the predicted time period will comprise a respiratory failure event. As shown in Figure 5B, the 504b prediction data exceeds the 512b threshold in several instances throughout the 510b time period, indicating a prediction that the user's future breathing status in the predicted time period will comprise a respiratory failure event. However, as shown by the 502b sensor data, the user's breathing status in this example does not comprise a respiratory failure event.In this way, the overall trained predictive model incorrectly predicted one or more respiratory failure events (e.g., false positive prediction). As shown in... Petition 870250101201, dated 05 / 11 / 2025, pages 137 / 192 129 / 172 figure 5B, the prediction data 506b does not reach or exceed the limit 512b for the entire time period 510b, indicating a prediction that the user's future breathing state in the predicted time period will comprise normal breathing. In this way, the custom predictive model correctly predicted normal breathing (i.e., true negative prediction).

[00228] Now, with regard to Figure 5C, example sensor data comprising airflow sensor data 502c is shown therein. In other examples, embodiments may be used in which the sensor data may comprise data measured by the other types of sensors, as discussed previously. It can be seen from the example shown in Figure 5C that the user's breathing state includes an apnea breathing event during time period 508c. It can further be seen from the example embodiment shown in Figure 5C that the user's breathing state includes a normal breathing state during time period 510c.

[00229] Figure 5C further shows example prediction data 504c generated by a general example trained predictive model using sensor data 502c and example prediction data 506c generated by a custom example predictive model using sensor data 502c. In the embodiment used for this example, the data of Petition 870250101201, dated 05 / 11 / 2025, pages 138 / 192 130 / 172 prediction 504c comprises a probability for each point in time of the sensor data 502c that the user's future breathing state within a predicted time period will comprise a respiratory failure event based on the general example trained predictive model. In the modality used for this example, prediction data 506c comprises a probability for each point in time of the sensor data 502c that the user's future breathing state within a predicted time period will comprise a respiratory failure event based on the custom example predictive model.

[00230] In this example, a 512c threshold of 0.5 is shown. In other example embodiments, other threshold levels may be used. A probability represented by the 504c or 506c prediction data reaching or exceeding the 512c threshold indicates that the respective predictive model predicts that the user's future breathing status in the predicted time period will comprise a respiratory failure event. As shown in Figure 5C, neither the 504c nor 506a prediction data reach the 512c threshold before the respiratory failure event in time period 508c. Thus, each of the general trained predictive models and the custom predictive models failed to correctly predict the respiratory failure event occurring in time period 508c (e.g., false negative prediction). Petition 870250101201, dated 05 / 11 / 2025, pp. 139 / 192 131 / 172

[00231] Now, with respect to Figure 6A, example sensitivity data 602a corresponding to a custom example predictive model and example sensitivity data 604a corresponding to a general trained example predictive model are shown. Sensitivity data 602a and 604a represent the number of correctly predicted respiratory failure events (i.e., true positives) by the custom predictive model and the general trained predictive model, respectively, divided by the total number of respiratory failure events that occurred in a measured time period. Figure 6A shows sensitivity data 602a and 604a as a function of the threshold, where the threshold represents the probability at or above which the corresponding predictive model predicts that the user's future breathing status in the predicted time period will comprise a respiratory failure event.As shown in Figure 6A, the custom predictive model has better sensitivity than the general trained predictive model across all limits.

[00232] Now, with regard to Figure 6B, the same shows example precision data 602b corresponding to a custom example predictive model and example precision data 604b corresponding to a general trained example predictive model, where these models Petition 870250101201, dated 05 / 11 / 2025, pages 140 / 192 132 / 172 are similar to those in Figure 6A. This is only an example of an ideal threshold that can be chosen to verify the best accuracy and sensitivity in the same patient. The accuracy data 602b and 604b represent the number of correctly predicted respiratory failure events (e.g., true positives) by the custom predictive model and the general trained predictive model, respectively, divided by the total number of respiratory failure events predicted by the custom predictive model and the general trained predictive model, respectively (e.g., sum of true positives and false positives). Figure 6B shows the accuracy data 602b and 604b as a function of the threshold, where the threshold represents the probability at or above which the corresponding prediction model predicts that the user's future breathing status in the predicted time period will comprise a respiratory failure event.As shown in Figure 6B, the custom predictive model has better accuracy than the general trained predictive model across all limits.

[00233] Now, with regard to Figure 7, a flowchart is shown of an example embodiment of a method 700 for simulating the operation of a respiratory assistance device and / or the health status of a user receiving assistance from the respiratory assistance device, such as, for example, the device of Petition 870250101201, dated 05 / 11 / 2025, pages 141 / 192 133 / 172 Respiratory Assistance 102. Method 700 can be used to implement a digital twin simulation of one or more aspects of the user and / or the operation of the respiratory assistance device. Method 700 can be used in combination with method 300 or as an alternative to method 300. Similar to method 300, method 700 can be performed by the controller processor 206 during the execution of the software instructions of the various modules described above. However, in other embodiments, method 700 can be performed by other processors or another applicable device. For example, in some embodiments, method 700 can be performed by a combination of processors, including, for example, the processor 228 of the respiratory assistance system 200 and a remote processor, such as the server processor(s) 118.For ease of explanation, the elements represented in Figures 1A-2 should be used in describing the various steps of method 700. For example, method 700 can be implemented by processor 228 of the respiratory assistance device controller 206. However, it is understood that this technique can be used in another applicable device.

[00234] Method 700 can begin when the breathing assistance device 202 has been activated, although in some cases some steps of Method 700 may be Petition 870250101201, dated 05 / 11 / 2025, pages 142 / 192 134 / 172 performed before the breathing assistance device 202 is activated. For example, as will be described, at least some information about the user can be obtained before the breathing assistance device 202 is activated.

[00235] In act 702, a user model is received. The user model 120 can be stored, for example, in a data store, such as data store 116, and / or generated by the server 118 and can be a custom model or a generic model (if the functionalities provided by method 700 are used for the first time), for example. As described above, the user model 120 is a digital user model that can be called a “digital twin” of the user and can be a model that can model a current state of the user and a future state of the user. In some cases, the parameters of the model(s) can be initially determined and / or updated from models associated with other individuals. For example, parameters of user models and / or user models associated with multiple other individuals can be collected and stored, for example, in data store 116 and can be distributed to initialize the user model 120.Stored user models can be categorized according to similarities between the individuals with whom these user models are associated. For example, the models of... Petition 870250101201, dated 05 / 11 / 2025, pages 143 / 192 135 / 172 users can be grouped according to user phenotypes. The parameters of the user model(s) can also be used to update the parameters of user models associated with other individuals who are grouped by similarity. User model 120 can be a single customized model or a combination of user models that model different internal systems of the user (i.e., models of the user's physiological system), including, but not limited to, the respiratory system, the cardiovascular system, the user's nervous system, or an operable combination thereof, for example, and this models the user's environment as it affects the user's health status (i.e., environmental models).

[00236] User model(s) 120 may be implementing the use of any type of model, including mathematical models and neural networks, other machine learning models, or a combination of models. For example, the respiratory system may be modeled as the airways open in series with a single compartment, two compartments, or a multi-branching respiratory tree using an RLC circuit, a mass-spring-damper system, a complex nonlinear network, etc. Similarly, the nervous system model may be an electrical and / or mechanical model, a mathematical model, a neural network, etc., that model stages of Petition 870250101201, dated 05 / 11 / 2025, pages 144 / 192 136 / 172 sleep, a respiratory rate, a ventilatory rate, cardiac function, and / or other brain-related functions of a user. The cardiovascular system model can be an electrical and / or mechanical model, a mathematical model, and / or a neural network that can model a user's cardiovascular functions. In some cases, the models may include models known in the technology, for example, models of the form described in “An integrative model of respiratory and cardiovascular control in sleep-disordered breathing”, Limei Cheng, Olga Ivanova, Hsing-Hua Fan, Michael CK Khoo, doi:10.1016 / j.resp.2010.06.001 and in “Analysis of Mathematical Models of the Human Lung”, Rachel L. Cooper, Master of Science, Virginia Commonwealth University.

[00237] In some embodiments, the user model 120 may be associated with a corresponding physical model (called a “physical twin”) that models one or more of the user’s physiological systems. For example, the user model 120 may be associated with a benchtop lung simulator and / or a benchtop lung-brain-cardiovascular simulator, which may simulate the user’s lung, brain, and / or cardiovascular functions. Examples of such physical simulators are provided in U.S. Patent 11,610,513 entitled “Benchtop within-breath dynamic lung simulator,” which is incorporated herein in its entirety. The user model 120 and the physical model associated with the user model Petition 870250101201, dated 05 / 11 / 2025, pages 145 / 192 137 / 172 user models can be used in combination to simulate a current and / or expected health status of the user or an expected user response to a therapy / change in therapy. For example, the physical model can be used to validate and / or supplement the user model 120.

[00238] In at least one embodiment, the user system models that are used in the user model may be common to all user models, but the parameters of the user system models may be modifiable in such a way that the user model 120 can be customized for a given user and model the health status of that given user. For example, when sensor data is received, as will be described in further detail below, the model parameters may be modified in such a way that an expected health status of the user can be determined through simulation. The parameters may be updated by transmitting the sensor data as inputs to the user model(s) 120 and / or the device model(s) 122 to train or retrain the models. The parameters may be updated frequently as sensor data is received.Alternatively, the parameters can be updated at slower, regular intervals, such as hourly, nightly, weekly, or at the request of the user or a healthcare professional supervising the user's treatment. Petition 870250101201, dated 05 / 11 / 2025, pages 146 / 192 138 / 172 For example, the sensor data can be recorded, optionally pre-processed and stored as it is received, and the model(s) can be trained to determine the updated parameters each time an update is scheduled and / or requested. The magnitude of the change in parameters can vary depending on the type(s) of model(s) used, the amount of data (e.g., sensor data) received by the model(s), the learning rate of the model(s), the weight of the training classes used by the model(s), and / or other characteristics of the model(s) used.

[00239] In at least one embodiment, the model parameters can be further customized based on the user's personal characteristics. For example, during the setup of the 202 respiratory assistance device or using the functionalities of method 700 for the first time, or in some cases, immediately before using the 202 respiratory assistance device, the user may be prompted to provide or update personal characteristics (also called user personal factors), for example, the user's alcohol consumption, the user's drug use, medication taken by the user, recent blood test results, the user's height, the user's weight, and / or the user's age. Because these factors can affect a user's health status, the user model 120 can be customized. Petition 870250101201, dated 05 / 11 / 2025, pages 147 / 192 139 / 172 using one or more of these personal user factors can allow more personalization of the user model 120, so that the user model 120 can be a more accurate reflection of the user's physiological systems. The user can provide the user's personal factors, for example, through the computing device 150 (user computing device 150a in Figure 1B). In some embodiments, one or more personal user factors can be provided by an external user or an external system, such as a medical professional who monitors the user's therapy, a pharmacy system, a payer system, a health system, other home medical equipment, or an electronic medical records system, can provide one or more personal user factors about the user, through external user computing device 150b (shown in Figure 1B).In at least one mode, before or immediately after the use of the respiratory assistance device 202, the user may be prompted to enter information about the user's current state, for example, the user's alcohol or substance use, a subjective assessment of the pressure exerted by the respiratory assistance device 202 on the user's face (e.g., nasal passages and / or mouth), a subjective assessment of the humidity and temperature of the air provided by the respiratory assistance device 202, an assessment of the user's sleep quality, or... Petition 870250101201, dated 05 / 11 / 2025, pages 148 / 192 140 / 172 any operable combination thereof. User-provided information may be used to vary the weighting of parameters and / or vary the range of parameters that may be used for the general user model 120 and / or one or more of the user's physiological system models.

[00240] In some cases, a generic model based on one or more of the user's personal characteristics may be used. For example, data storage 116 may include models associated with different categories, such as, but not limited to, age categories, sex categories, or weight categories, for example. Models may be derived, for example, based on existing models associated with users who have similar personal characteristics (e.g., physiological conditions, age, sex, etc.). For example, simulation results from users may be transmitted to a server, such as server 118, which analyzes the results to derive parameters for the user model.

[00241] In act 704, a device model 122 of the respiratory assistance device 202 is received. Similar to user model 120, the device model 122 of the respiratory assistance device 202 may be stored, for example, in a data storage such as data storage 116, and / or generated by the server 118 and may be a model that is customized for the device. Petition 870250101201, dated 05 / 11 / 2025, pages 149 / 192 141 / 172 of respiratory assistance device 202 that is being used or a generic model based on a general type of respiratory assistance device, for example, if the functionalities provided by method 700 are used for the first time. As described above, the device model can be called a “digital twin” of the respiratory assistance device 202. The device model 122 of the respiratory assistance device 202 can be a model or a combination of models that model the operation of the respiratory assistance device 202.For example, device model 122 of the respiratory assistance device 202 may be based on a combination of device submodels, such as: (a) individual device components of the respiratory assistance device 202, such as a piston, a blower, an electrical input of the device, (b) device subsystems of the respiratory assistance device 202, such as a humidifier subsystem of the respiratory assistance device 202, (c) device functionalities of the respiratory assistance device 202, such as the air pressure of the distributed air, the humidity of the distributed air and the airflow of the distributed air, or (d) any operable combination of (a) through (c). Each of these models may be any type of model that can model the operation, components and / or subsystems of a device. Petition 870250101201, dated 05 / 11 / 2025, pages 150 / 192 142 / 172 respiratory assistance 202, including, but not limited to, an electrical and / or mechanical model, complex transfer functions, neural networks, other machine learning models, etc.

[00242] Similar to user model 120, device models 122 used may be common to all respiratory assistance device models 122, but the parameters of device models 122 may be modifiable in such a way that device model 122 can be customized for the device being used. For example, when sensor data is received, as will be described in further detail below, parameters of device models 122 can be modified in such a way that an expected performance or operation of the respiratory assistance device 202 can be determined. It is possible that, for two devices that are of the same type (i.e., same manufacturer and model), the devices may behave differently over time due to the amount of use, the amount of maintenance, the amount of component variation, and the like.

[00243] In Act 706, sensor data is received from sensor(s), for example, sensors 220-223. As described in Figure 2, sensor data can be received from one or more sensors placed in different locations on the user, the sensors placed in Petition 870250101201, dated 05 / 11 / 2025, pages 151 / 192 143 / 172 various locations in the room where the respiratory assistance device 202 is located and / or sensors located on, or in, the respiratory assistance device 202. In some embodiments, sensor data can be received from all available sensors 220-23. In other embodiments, sensor data can be received from only some of the available sensors. For example, sensor data from some of the sensors 220-223 can be received only if selected by the user (or an external user). As another example, the processor can select sensors from which sensor data is received based on personal characteristics of the user, a previous operation of the respiratory assistance device 202, and / or user feedback.For example, before activating the breathing assistance device 202, the user can provide feedback on a previous experience with the breathing assistance device 202. Based on the user feedback received, one or more sensors 220-223 can be selected as sensors from which sensor data is received. Sensor data can be received as sensor data is collected, for example, sensor data can be continuously received in real time. Alternatively, sensor data can be received at predetermined time intervals. Petition 870250101201, dated 05 / 11 / 2025, pages 152 / 192 144 / 172

[00244] In Act 708, an expected state of the respiratory assistance device 202 and / or an expected health state of the user are determined based on the device model 122 of the respiratory assistance device 202 and / or the user model 120, as well as the sensor data received in Act 706. The expected state may correspond to an expected future state or an expected current state based at least on the sensor data received at a previous time. By determining an expected state of the user's health state and / or of the respiratory assistance device 202, changes in a user's health state or unexpected changes in the operation of the respiratory assistance device 202 can be identified.Remedial actions can then be taken if the expected states indicate that a problem is occurring or will occur, such as imminent device failure, in which case maintenance can be performed on the device, including replacing one or more device components, replacing one or more device subsystems, recalibrating the device, or any operable combination thereof.

[00245] By using the sensor data received in act 706, the parameters of the user model 120 and / or the parameters of the device model 122 of the respiratory assistance device 202 can be modified, in such a way that one or both models can be Petition 870250101201, dated 05 / 11 / 2025, pp. 153 / 192 145 / 172 customized / personalized for the user and / or the particular respiratory assistance device 202 that is actually used. For example, using sensor data, the parameters of the user model 120 can be modified to adjust the user model 120 to reflect the user's characteristics (e.g., the current state of these characteristics). Similarly, using sensor data, the parameters of the device model 122 can be modified to reflect a current state of the respiratory assistance device 202. For example, based on received sensor data, it can be determined that one or more components of the respiratory assistance device 202 are not operating according to their original manufacturer specifications.In this case, the parameters of device model 122 can be modified in such a way that device model 122 more accurately reflects / models the current state of the respiratory assistance device 202. The model(s) can then be used to simulate a current and / or future state of the respiratory assistance device and / or the user's health over time.

[00246] By modeling a current state of the respiratory assistance device 202, a future state of the respiratory assistance device 202 can be determined. For example, the average service life of a Petition 870250101201, dated 05 / 11 / 2025, pp. 154 / 192 146 / 172 component of the respiratory assistance device 202 can be known and, based on the component's current state, including, for example, the component's current age and current use, the component's operation over time can be simulated using the general device model 122 or a component model. For example, based on received sensor data and / or received user personal factors, the model(s) (e.g., device model 122, component model, user model 120, a combination of models) can generate synthetic data, which can be used to further train the model(s) (i.e., updating model parameters). The synthetic data can be generated by extrapolating the current data over time and / or by applying a model, such as a mathematical regression model, to the current data.Alternatively, synthetic data may not be generated, and a predictive model may be applied directly to the model parameters to update the model parameters based on predictions regarding the model parameters and / or the models themselves. In some cases, the expected state of the respiratory assistance device 202 may be determined based, in part, on the user model 120. For example, it may be determined that certain user characteristics affect the operation of the respiratory assistance device 202. An asthmatic user, for example, may... Petition 870250101201, dated 05 / 11 / 2025, pages 155 / 192 147 / 172 being more sensitive to air quality. In such cases, the service life of the air filter of the 202 breathing assistance device may be determined to be shorter than for a non-asthmatic user. Alternatively, or furthermore, the environment in which the user wears the 202 breathing assistance device may affect the expected state of the 202 breathing assistance device. For example, based on sensor data from environmental sensors, it may be determined that the room where the 202 breathing assistance device is placed is at a high elevation or has poor air quality (e.g., due to smoke, pets, etc.), which may affect the service life of some components of the 202 breathing assistance device.

[00247] The user's health status can be similarly determined using user model 120 and, optionally, device model 122 of the respiratory assistance device 202. For example, based on sensor data received in act 704 and user model 120, the user's health status can be determined by providing the sensor data as input to user model 120, since some measurements obtained from the sensors may indicate poor health or, conversely, good health. For example, some FOT measurements, sound measurements, and reactance measurements may be associated with individuals Petition 870250101201, dated 05 / 11 / 2025, pages 156 / 192 148 / 172 who have CPOD. Alternatively, or in addition, the user model 120 can be used to simulate the user's health status over a certain period of time, such as every hour (e.g., for very poor health users on a ventilator), weeks, months, or another time horizon. In the manner described above, based on received sensor data and / or received user personal factors, synthetic data can be used to further train the model(s) and / or predicted model parameters can also be generated and / or a predicted model behavior can also be determined. Using the trained models, the user's current and predicted future health states can then be determined.

[00248] In at least some modalities, the user's health status may include the user's sleep health. A user's sleep health may correspond to a measure of the user's sleep health and may be determined based on sleep parameters, including, but not limited to, sleep duration, duration of sleep stages, sleep depth in each stage, and / or the user's heart rate during sleep, and may be provided as inputs to the user model 120 which may then transmit a sleep health to the user. The measure may be a score, for example, a unified score based on two or more measured parameters, or a score Petition 870250101201, dated 05 / 11 / 2025, pages 157 / 192 149 / 172 multivariate (i.e., a score for each parameter). Alternatively, sleep health can correspond to a set of values ​​corresponding to the measured parameters. Using sensor data from air pressure and EEG sensors, for example, as parameters in the model(s), one or more of these factors can be evaluated to determine a user's sleep health. Based on the simulated assessment of the user's sleep health, the operation of the 202 breathing assistance device can be modified if the simulated sleep health is poor, as will be described in further detail below. The user's sleep health can also be determined by simulating the user's health status over time using the models. For example, it can be determined that, based on current sensor data, the user is unlikely to achieve deep sleep.

[00249] The user and / or respiratory assistance device models 202 can be used to make predictions about the future health status of the user and / or the respiratory assistance device 202. For example, using the user model 120 and, optionally, the device model 122, the user's health, which may include the user's sleep health, can be simulated over time and health indicators can be identified. For example, health indicators may correspond to data (e.g., sensor data, synthetic data), data patterns, and / or Petition 870250101201, dated 05 / 11 / 2025, pages 158 / 192 150 / 172 changes that are known to be associated with a respiratory disease. By simulating the user's health over time and identifying health indicators, the occurrence of a respiratory disease can be predicted. For example, based on the identified health indicators, a relative score (e.g., a probability) can be assigned to the health conditions or diseases associated with the health indicators. These health indicators may correspond to known indicators and / or indicators determined from the assessment of the health status of other users over time. For example, the variation over time of a model parameter, discrepancies between an actual and an expected parameter, and / or deviations from an expected parameter for an individual with the user's personal characteristics (e.g., age, sex, weight, respiratory resistance, etc.) may be an indication that the user is developing COPD or heart disease.An expected parameter value can be determined based on the user's personal characteristics and / or the user's previous health status, and deviations from the expected parameter value may be indicative of a health condition. When it is determined that the user is experiencing a respiratory failure event, or other health event sensor data received during the event, it can be transmitted to an external database. Petition 870250101201, dated 05 / 11 / 2025, pages 159 / 192 151 / 172 in such a way that health event indicators can be determined over time. In this way, in at least one modality, several diseases can be predicted or determined based on sensor data, in such a way that each disease can be given a relative probability score. Since data is obtained when users are using their respiratory assistance devices, parameters corresponding to these diseases can be adapted to the data, and these parameters can then be compared with those of the general population to determine a probability of at least one of these diseases. In modalities where the user model 120 is associated with a physical model, at least part of the simulation can be done on the physical model.

[00250] As another example, the performance of the breathing assistance device 202 can be simulated over time using the user model 120 and, optionally, the device model 122. Predicting the user's future health status may, in some cases, involve predicting a sleep disruption. For example, by using the user model 120 and, optionally, the device model 122 to simulate the user's sleep health over time, it can be determined that a sleep disruption is expected to occur. For example, the user's sleep depth and / or the user's REM sleep percentage, a determinant of Petition 870250101201, dated 05 / 11 / 2025, pages 160 / 192 152 / 172 An individual's sleep quality can be measured. By measuring these parameters, the user's expected percentage of REM sleep can be determined and compared to the expected percentage of REM sleep for individuals with similar or identical personal characteristics to the user, and a deviation between the user's actual percentage of REM sleep and the expected percentage of REM sleep can be determined. To prevent sleep disruption or decrease the likelihood of sleep disruption occurring, the operation of the 202 breathing assistance device can be modified, as will be described in further detail below.

[00251] A recommendation can be generated based on the simulation results. For example, a recommendation can be made to proactively replace a component of the respiratory assistance device 202 and / or perform maintenance if it is predicted that the component will break or a device subsystem will fail based on the simulation. Alternatively, or in addition, a report can be generated. The recommendation can be determined by the therapy module 124 which can analyze the results of the user model 120 and the device model 122. Several examples of the analysis that can be performed to provide recommendations are provided in this document and can be performed by the therapy module 124 by making comparisons between certain data and limits or providing certain data to Petition 870250101201, dated 05 / 11 / 2025, pages 161 / 192 153 / 172 models. In another example, therapy module 124 can make a recommendation, such as sending a message to a doctor / health system to prescribe medication or to request permission to make a change to the sleep therapy being provided to the user by the breathing assistance device, as further explained below.

[00252] In at least one embodiment, the user model 120 and, optionally, the device model 122 can be used to determine whether a change, for example, an unexpected change, has occurred in the user's health status and / or in the status of the respiratory assistance device 202. For example, method 700 may involve simulating the operation of the respiratory assistance device 202 over a period of time to determine (through simulation) expected parameters associated with the operation of the respiratory assistance device 202. Method 700 may then involve receiving sensor data from the sensor(s) and comparing the parameters expected from the simulation with actual parameters that are based on the received sensor data. Based on the comparison, a failure of a component of the respiratory assistance device 202 or a malfunction of the device 202 can be determined.For example, when the respiratory assistance device 202 is simulated as not performing as expected, this... Petition 870250101201, dated 05 / 11 / 2025, pages 162 / 192 154 / 172 may indicate that one or more components of device 202 will experience a failure. As another example, a difference between the expected distributed airflow, as determined by control signal 112, and the actual airflow of the distributed gases may indicate that a component is faulty. In such cases, a notification may be generated to alert the user of the fault. In some cases, the alert may indicate the source of the fault and / or identify the component(s) that are faulty.

[00253] In some cases, a change in a user's health status may be indicative of a respiratory disease, a heart disease, a brain disease, or any other disease that can be identified using sensor(s) and one or more user models 120. To identify a change in a user's health status, the method 700 may involve simulating a user's health status using user model 120 and, optionally, device model 122 of the respiratory assistance device 202 over time to determine expected parameters associated with the user's health status. The method may then involve receiving sensor data from the sensor(s) and comparing the expected parameters with parameters based on the received sensor data. Alternatively, based on the expected parameters determined by the simulations, expected sensor data Petition 870250101201, dated 05 / 11 / 2025, pages 163 / 192 155 / 172 (e.g., simulated) can be obtained, and the expected sensor data can be compared with the actual sensor data received. For example, a test such as FOT can be performed, which indicates that the patient is developing asthma or COPD. In another example, the patient's heart rate can be analyzed in the digital twin simulation for heart rate variation, and it can be determined that the patient's cardiovascular health is being influenced by the sleep therapy being provided by the breathing assistance device, and a change in sleep therapy may be necessary.

[00254] Based on the comparison, a recommendation can be generated. For example, if the sensor data received indicates that the user's health status is lower than the expected health status, for example, as determined by the expected parameter values, an alert notifying the user's health can be generated. Alternatively, or furthermore, a health recommendation to improve the user's health, a diagnosis of a respiratory condition or another physiological condition, a report indicating the user's health and / or a health recommendation to consult a medical professional can be generated. For example, if, before using the 102 respiratory assistance device, the user indicates alcohol consumption, it is determined that the user's sleep health is affected, and the Petition 870250101201, dated 05 / 11 / 2025, pages 164 / 192 Recommendation 156 / 172 may include a recommendation regarding the use of alcohol. In cases where the respiratory assistance device 202 is used as part of treatment supervised by a medical professional, reports may be generated and transmitted to a computing device used by the medical professional, for example, computing device 150. The recommendation may additionally include a recommendation to use a type of mask, a recommendation to change the fitting of an existing mask, a recommendation for a new tube for the respiratory assistance device 202, recommended adjustments for the respiratory assistance device 202, recommended environmental adjustments (e.g., room temperature, light exposure, mattress, etc.), or any operable combination thereof.For example, a message might be sent to a user via their computing device stating that their mask has aged and, as a result, air leakage during use has increased, reducing the effectiveness of sleep therapy. The message might also include information that the user's insurance covers the cost of a new mask and that the user should obtain a new mask. This message might be sent based on analysis performed to determine that the mask needs to be replaced. For example, based on digital twin simulations and analysis of sound and leakage data. Petition 870250101201, dated 05 / 11 / 2025, pages 165 / 192 157 / 172 mask air pressure, a recommendation can be made that the user should change their mask to another specific type of mask, which may result in less air leakage and better sleep therapy.

[00255] For example, during digital twin simulation, it may be determined that the breathing assistance device is functioning properly, but due to a poor mask fit, there is a high level of air leakage. In this example, a recommendation may be provided to the user and / or a medical facility (e.g., store) where the mask was obtained that the user should exchange or adjust their mask for a better fit and reduced air leakage. This can be important because, in the first few weeks of using a new breathing assistance device, users typically find that they need to return to the facility where they received the new device and continue requesting adjustments until they feel comfortable using the breathing assistance device, which can be inconvenient and cause some users to give up using the device.However, with the use of simulation that provides automated personalized recommendations, users will be able to more easily adjust the operation of their respiratory assistance devices and / or use more appropriate equipment (e.g. Petition 870250101201, dated 05 / 11 / 2025, pages 166 / 192 158 / 172 example, better masks with a better fit), which will increase the likelihood that the user will continue to use their respiratory assistance devices.

[00256] In at least one embodiment, the recommendation may involve adjusting the operation of the respiratory assistance device 202. For example, as previously described in relation to Figures 1A-1B, operating parameters of the respiratory assistance device 202 may be adjusted to adjust, correct, and / or improve a therapy provided to the user. Since the user model 120 is customized for the user, the adjustment is customized for the user. For example, based on the user's user model 120, it may be determined that the air pressure of the respiratory assistance device 202 is causing the user to experience cardiac inflammation. In this way, the air pressure of the respiratory assistance device 202 may be adjusted (e.g., reduced). As another example, it may be determined that the air pressure of the respiratory assistance device 202 is too low.However, it can be determined, based on user model 120, that the user recently underwent heart surgery (for example, based on medical records received from a medical professional treating the user or an electronic medical records system, or based on medication taken by the user) and that, consequently, Petition 870250101201, dated 05 / 11 / 2025, pages 167 / 192 159 / 172 moderate air pressure, rather than high air pressure, should be used.

[00257] It should be noted that, in general, in any embodiment described in this document, both real and simulated, adjustments can be made to factors other than airflow and air pressure flow, where examples of these factors include, but are not limited to, humidification level, heated air temperature, dispensing of medication doses in the airflow, for oxygen therapy, pressure, flow and frequency of oxygen application. In the example of medication doses, a medical professional may send a message to the user to take a certain dose of medication, which can be included in the airflow by the device, such as including statin when Ang2 levels for the user are considered high.

[00258] As another example, adjusting, correcting, and / or improving a therapy provided to the user may include changing an operating mode of the respiratory assistance device 202 by using a device profile that will be more effective in improving the user's sleep health. For example, a respiratory assistance device, with the same hardware, may be configured to operate using a different device profile (e.g., implemented by software programs and parameter values), so that the respiratory assistance device Petition 870250101201, dated 05 / 11 / 2025, pages 168 / 192 The 160 / 172 breathing assist device, with no hardware changes, can be configured to operate as several different breathing assistance devices based on the selected device profile. For example, device profiles include, but are not limited to, CPAP, APAP, BiPAP, ASV, and non-invasive ventilator (NIV) device profiles. The 202 breathing assist device may be operating according to a given device profile, and then the breathing assist device can be reconfigured via software to operate according to a different device profile. For example, the device profile of the 202 breathing assist device can be switched so that it no longer operates as a CPAP device but now operates as a BiPAP device. In this way, in at least one mode, the control signal 112 can be automatically generated according to the selected device profile.In some cases, to adjust the operation of the respiratory assistance device 202, a correction factor can be determined, and the operation of the respiratory assistance device 202, with the correction factor applied, can be simulated using the device model 122 of the respiratory assistance device 202 and, optionally, the user model 120.

[00259] As described previously, the models Petition 870250101201, dated 05 / 11 / 2025, pages 169 / 192 161 / 172 (e.g., device models 122, user models 120) can be configured in such a way that, during simulation, an expected value of a parameter based on current sensor data, synthetic data, and / or predicted model parameters is determined. In at least one embodiment, by simulating the operation of the respiratory assistance device 202, expected parameter values ​​can be obtained and a simulated effect of changing the correction factor for the control signal, either by using a particular device profile or by making some other change, such as updating the custom prediction model, in the respiratory assistance device and / or the user, can be evaluated. If it is determined that applying the correction factor and / or other changes achieves the desired effect, then the change to the correction factor can be applied via the control signal 112 and / or the other changes can be made.Conversely, if it is determined that the correction factor does not achieve the desired effect, the correction factor can be modified and the effect of the modified correction factor and / or other changes can be simulated again, using the model(s), until a satisfactory correct factor is found. For example, various correction factors applied to inspiratory and expiratory air pressure levels, humidity and / or gases can be simulated using the assistive device models. Petition 870250101201, dated 05 / 11 / 2025, pages 170 / 192 162 / 172 breathing 202 and the user and the impact of these correction factors can be determined based on simulations.

[00260] In at least one embodiment, as sensor data can be continuously received, the adjustment of the operation of the respiratory assistance device 202 can take into account recently received sensor data. For example, based on positional sensor data, it can be determined that the user is on their back and, in this way, the model(s) can simulate the effect of the therapy provided by the respiratory assistance device when a user is on their back. In some embodiments, as shown in Figure 1B, before adjusting the operation of the respiratory assistance device 202, the respiratory assistance device 202 can transmit a request to an external device, for example, the external user computing device 150b associated with a medical professional who supervises the user's therapy and requires an indication of approval of the adjustment.In some embodiments where the user model 120 is associated with a physical model, the adjustment can be simulated on the physical model to determine the effect of the adjustment before operation of the respiratory assistance device 202 being adjusted.

[00261] Once the operation of the respiratory assistance device 202 is adjusted, the response of Petition 870250101201, dated 05 / 11 / 2025, pages 171 / 192 163 / 172 The user's response to the change in operation of the respiratory assistance device 202 can be monitored. The user's response to the change in operation can be monitored by evaluating an Intervention Index (II) that distinguishes the probability of the user experiencing an imminent respiratory event at a given time and / or a Respiratory Event Index (REI), which distinguishes the number of respiratory events experienced by the user at a given time. If the intervention index exceeds a predetermined limit and / or the respiratory event index exceeds a historical average number of respiratory events per hour experienced by the user by a predetermined limit, the respiratory assistance controller 206 can revert the setting.Alternatively, the respiratory assistance controller 206 can switch the operation of the respiratory assistance device 202 so that the respiratory assistance device 202 provides reactive therapy instead of preventive therapy. The intervention index can be determined based on the user model 120, and the average historical number of respiratory events can be retrieved from memory, for example, from memory unit 229 or from data storage 116. The activities described in this paragraph can be performed by digital twin simulation, and changes can also be made to the personal prediction model by sending one. Petition 870250101201, dated 05 / 11 / 2025, pages 172 / 192 164 / 172 Over-the-air update for any user's personal predictive model to improve security.

[00262] In some modalities, the change in the operation of the respiratory assistance device 202 and the user's response to the change in operation can be used to train user models associated with other individuals, so that similar changes can be applied to individuals associated with similar user models such as user model 120 with the expectation that these similar changes for other users will achieve similar results. For example, as explained previously, individuals can be classified according to phenotypes.A change applied to the operation of a user's respiratory assistance device 202 that achieves favorable results (e.g., improved sleep health) can be implemented in respiratory assistance devices associated with other individuals associated with the same phenotype and / or can be used to train user models associated with other individuals associated with the same phenotype. The activities described in this paragraph can be performed by digital twin simulation, and changes can also be made to the personal predictive model by sending an over-the-air update to any user's personal predictive model to improve safety. Petition 870250101201, dated 05 / 11 / 2025, pp. 173 / 192 165 / 172

[00263] For example, patient phenotypes can be developed based on similarities in a combination of received data, age, height, and outcomes. For example, if a patient with a certain height, age, medication intake, inflammatory biomarker level, and BMI is treated and their outcomes are similar to another patient with similar height, weight, etc., then these patients can be grouped in terms of these phenotypes, and digital twin simulation can allow model and operational parameters, which can be saved in a database (e.g., a digital twin database), to apply the same update to patients / users who fit this phenotype. Therefore, personalization can occur with patient groups in some cases. However, further personalization can also be performed according to the precepts set forth in this document.

[00264] In this way, in at least one modality, for a given user, the digital twin simulation can be configured to search through the digital twin database to identify an equivalent patient and replicate their breathing solution for this user based on similarities between the given user and the equivalent patient.

[00265] In another aspect, according to the precepts set forth in this document, user identification (e.g., patient) can be incorporated into Petition 870250101201, dated 05 / 11 / 2025, pp. 174 / 192 166 / 172 Digital twin simulation for security / authentication purposes. For example, the user's breathing signature can be determined. This can be done, for example, according to the techniques described in US patent 11,612,708. The digital twin simulation can then use the user's breathing signature as a security measure to confirm that the patient data and the custom models used for digital twin simulation for a given user actually correspond to that user. For example, the breathing signature can be included with the user data and the custom predictive model provided for digital twin simulation. Separately, part of the digital twin simulation may involve accessing a database to obtain a stored breathing signature for the given user.The stored breath signature from the database is compared to the received breath signature that was included with the user data and the custom predictive model to ensure a match before the simulation continues. Similarly, the user's breath signature may be saved with any simulation results for safety purposes to ensure that future data use is with the user who has a matching breath signature. Petition 870250101201, dated 05 / 11 / 2025, pages 175 / 192 167 / 172

[00266] In another aspect, according to the precepts set forth in this document, digital twin simulation can be performed for a given user using a custom predictive model to simulate how well the custom predictive model will perform when it is actually implemented for use by the user's respiratory assistance device. In this way, a custom predictive model can first be determined for the user using any of the techniques described in this document. The digital twin simulation is then performed by incorporating the custom predictive model into the device models that are used to simulate the operation of the respiratory assistance device.Physiological data for the user derived from the simulation data can then be analyzed to determine the effect on the user's health during the use of the custom predictive model and, optionally, simulate the operation of the respiratory assistance device according to a selected device profile, as explained previously. If the health effect is determined to be beneficial, which may involve determining that the user is not expected to develop a health condition or have a certain quality of sleep compared to normative data, then the custom predictive model is determined to be beneficial and is then implemented for use. Petition 870250101201, dated 05 / 11 / 2025, pages 176 / 192 168 / 172 with the user's actual breathing assistance device. In at least one mode, the false negative data that are determined for the user, which can be done according to the precepts set forth in this document, are also included in the digital twin simulation, so that the simulation can be more realistic.

[00267] In at least one embodiment, the digital twin simulation can simulate the ongoing use of the custom predictive model, the selected device profile and / or any other operational parameters to determine when breathing assistance is no longer effective (e.g., the user's sleep health declines), in which case the use of another custom predictive model, the selected device profile and / or any other operational parameters can be simulated to determine changes to one or more of these items that will achieve improved sleep health for the user.

[00268] In at least one embodiment, the result of a digital twin simulation may also include, or be solely, an implementation of data and / or adjustments, such as, but not limited to, at least one model, calibration data, one or more operational adjustments of the respiratory assistance device, or any operational combination thereof. For example, the at least one model may be a custom predictive model or another model that may be used Petition 870250101201, dated 05 / 11 / 2025, pages 177 / 192 169 / 172 by the controller used with the respiratory assistance device. Once the implemented data is received, the operation of the respiratory assistance device is updated in this way. This may involve updating the embedded software used by the respiratory assistance device and / or the controller.

[00269] In at least one embodiment, as explained, the input to a digital twin simulation may come from a doctor / physician / healthcare professional, an HME (Home Medical Equipment) reseller, a healthcare system, an EMR, a pharmacy system, or a payer (e.g., reimbursement), via, for example, the external user's computing device 150b. The user may also provide their own input via a software application on the user's computing device 150a to indicate how they are feeling due to a medical condition, medication use, and / or sleep therapy. This input may then allow the simulation of the effects of taking certain medication on the patient / user, so that the therapy module 124 can provide a recommendation for applying the correct sleep therapy to the user. The user's computing device 150a may be a smartphone.

[00270] In at least one modality, an exit from Petition 870250101201, dated 05 / 11 / 2025, pages 178 / 192 170 / 172 Digital twin simulation can interrupt or adjust therapy under certain situations. For example, raising the pressure to the ceiling (using high pressures) in the airflow is not appropriate for a cardiac surgery patient. Thus, if a user has had surgery and is taking medication, this can be provided as input in the digital twin simulation, and a decision can be made based on this input to alter the therapy.

[00271] In at least one mode, different levels of data logging can be adjusted during initialization for digital twin simulation. For example, in the event of a patient complaint, the ability to activate more active logging can be useful for investigating the complaint. For example, the data logging frequency can change in the event of an investigation (e.g., if the data collection sampling rate was 20 Hz, it can increase to 200 Hz in this or other situations). Alternatively, or furthermore, other extra sensors that are available but not normally used can be activated to provide additional data for the investigation. For example, a sound sensor that is not usually active can be activated. With the use of extra data logging, the causes of complaints can be excluded to narrow down to the real cause. Extra data logging can be used in other situations, such as when a user is Petition 870250101201, dated 05 / 11 / 2025, pages 179 / 192 171 / 172 experiencing a particular medical condition, for example, heart failure, and the extra data can be analyzed to determine if the sleep therapy being provided by the breathing assistance device is aggravating or causing the medical condition, or if the sleep therapy can be ruled out as a cause / aggravating factor of the medical condition. In this way, performing the simulation involves obtaining an increase in the amount of data recording and analyzing the increased amount of data to investigate whether the sleep therapy caused and / or aggravated a medical condition for the user.

[00272] Although the applicant's precepts described in this document are presented together with various embodiments for illustrative purposes, the applicant's precepts are not intended to be limited to such embodiments. On the contrary, the applicant's precepts described and illustrated in this document encompass various alternatives, modifications, and equivalents, without, in general, departing from the embodiments described in this document. For example, although the precepts described and shown in this document may comprise certain elements / components and steps, modifications may be made, as is known to those skilled in the art. For example, selected features from one or more of the example embodiments described in this document in accordance with the precepts set forth in this document may be Petition 870250101201, dated 05 / 11 / 2025, pp. 180 / 192 172 / 172 combined to create alternative modalities that are not explicitly described. All values ​​and sub-ranges within the described ranges are also described. The material described in this document is intended to cover and embrace all appropriate changes in technology.

Claims

1. A method for generating a customized predictive model to adjust airflow provided by a breathing assistance device for a user, the method being characterized in that it comprises: implementing a predictive model trained on a processor of a breathing assistance device controller, the trained predictive model being able to generate, based on sensor data, a real-time estimate of the user's current breathing state by determining a first plurality of probabilities, each of the first plurality of probabilities corresponding to a respective current breathing state of the user, and a prediction of the user's future breathing state by determining a second plurality of probabilities, each of the second plurality of probabilities corresponding to a respective predicted future breathing state of the user, within a predicted time period;receive sensor data measured from one or more sensors; operate the respiratory assistance device controller processor to apply the predictive model trained on the sensor data to generate real-time estimation and forecasting; identify one or more false negative predictions Petition 870250083544, dated 09 / 17 / 2025, p. 67 / 96 2 / 29 when the sensor data corresponding to a current period indicate a breathing event and the real-time estimate and / or forecast corresponding to the current period indicate a normal breathing state; receive false negative data comprising, for each of the one or more false negative predictions, the real-time estimate, the forecast, and a portion of the sensor data extending from a first point in time before a false negative prediction begins to a second point in time after a false negative prediction shift;Generate the custom predictive model by retraining the trained predictive model using false negative data so that the custom predictive model is tailored to the user; and implement the custom predictive model in the processor of the respiratory assistance device controller.

2. Method, according to claim 1, characterized in that the generation of the customized predictive model occurs after a minimum number of false negative predictions have been identified.

3. Method, according to claim 2, the method being characterized in that it further comprises generating simulated false negative data for Petition 870250083544, dated 09 / 17 / 2025, page 68 / 96 3 / 29 retraining the trained predictive model when an insufficient number of false negative data occurs during use by: applying one or more signal processing techniques to the sensor data, the one or more signal processing techniques comprising: random perturbation, noise addition, magnitude scaling, magnitude distortion, filtering, phase distortion, phase scaling, or segment truncation.

4. Method, according to claim 2 or 3, characterized in that the minimum number of false negative predictions is in a range from one to a total number of time points at which real-time estimation and prediction are generated during a monitoring period in which airflow is provided by the breathing assistance device to the user.

5. Method, according to claim 1, characterized in that generating the customized predictive model occurs automatically at a predetermined frequency.

6. Method, according to claim 1, characterized in that it further comprises determining one or more of a sensitivity, precision, F1 score and / or adjusted F1 score of the trained predictive model, wherein generating the custom predictive model Petition 870250083544, dated 09 / 17 / 2025, page 69 / 96 4 / 29 occurs based on one or more of the sensitivity, precision, F1 score and / or adjusted F1 score.

7. A method, according to any one of claims 1 to 6, characterized in that the first point in time and the second point in time lie within a range of 20 seconds to 60 seconds.

8. A method according to any one of claims 1 to 7, the method being characterized in that it further comprises preprocessing the false negative data using one or more of: normalization, sensor data weighting, principal component analysis, independent component analysis, downsampling, upsampling, frequency filtering, or manual inspection of the sensor data.

9. A method, according to any one of claims 1 to 8, characterized in that the retraining of the trained predictive model includes one or more of: transfer learning, fine-tuning of one or more parameters of the trained predictive model, addition of a layer to the trained predictive model, reinforcement learning, or any combination thereof.

10. Method, according to any one of claims 1 to 9, characterized in that the retraining of the trained predictive model additionally uses sensor data. Petition 870250083544, dated 09 / 17 / 2025, p. 70 / 96 5 / 29 11. A method according to claim 9, characterized in that the predictive model comprises one or more nodes and one or more layers, and the one or more parameters of the trained predictive model that are adjustable include a node type, a node selection, a node weight, a node activation, a node memory, a number of connections between nodes, an orientation of connections between nodes, an orientation of connections between layers, a layer type, a number of layers, a connection between layers, a number of inputs, a number of outputs, or an operable combination thereof.

12. A method, according to any one of claims 1 to 11, characterized in that operating the processor of the respiratory assistance device controller to apply the trained predictive model comprises: generating the real-time estimate by determining, based on sensor data, the first plurality of probabilities; and generating the prediction by determining, based on sensor data, the second plurality of probabilities.

13. Method, according to any one of claims 1 to 12, characterized in that the user's current breathing state and the user's predicted future breathing state comprise components that include normal breathing or one or more respiratory failure events.

14. A method according to claim 13, characterized in that one or more respiratory failure events comprise components that include obstructive apnea, central apnea, central hypopnea, obstructive hypopnea, a stimulus related to respiratory effort, an unclassified event, or any operable combination thereof.

15. A method according to claim 13, characterized in that one or more respiratory failure events include stimuli related to respiratory effort, including flow limitation, wheezing, oxygen desaturation, fragmentation, heart rate abnormalities, or any combination thereof.

16. Method, according to any one of claims 12 to 15, characterized in that the real-time estimate corresponding to the current period indicates the breathing event based on a first comparison of one or more of the first plurality of probabilities with one or more first limits, and the prediction corresponding to the current period indicates the normal breathing state based on a second comparison of one or more of the second plurality of probabilities with one or more second limits. Petition 870250083544, dated 09 / 17 / 2025, p. 72 / 96 7 / 29 17. Method, according to claim 16, characterized in that one or more first limits and / or one or more second limits are customized for the user and adjustable.

18. Method, according to claim 16 or 17, characterized in that one or more first limits and / or one or more second limits are updated in real time, near real time, hourly, daily, weekly and / or monthly.

19. A method according to any one of claims 1 to 18, characterized in that one or more sensors comprise user sensors, environmental sensors, or device sensors.

20. A method, according to any one of claims 1 to 19, characterized in that the predicted time period is in a range of 20 seconds to 60 seconds.

21. A method, according to any one of claims 1 to 20, characterized in that false negative data is received on a remote processor located remote from the respiratory assistance device, the custom predictive model is generated on the remote processor, and the generated custom predictive model is transmitted for implementation by the processor of the respiratory assistance device controller via a network connection between the processor of the respiratory assistance device controller and the remote processor.

22. A method, according to any one of claims 1 to 21, characterized in that the customized predictive model determines one or more of a pressure increase rate, a pressure decrease rate, and / or a pressure amplitude to adjust the airflow provided by the breathing assistance device to the user.

23. A controller for controlling the operation of a respiratory assistance device that provides respiratory assistance to a user, the controller being characterized in that it comprises: a memory unit comprising software instructions and parameters for at least one trained predictive model, the trained predictive model being capable of generating, based on sensor data, a real-time estimate of the user's current respiratory state and a prediction of the user's future respiratory state by determining a first plurality of probabilities, each of the first plurality of probabilities corresponding to a respective current respiratory state of the user, and a prediction of the user's future respiratory state by determining a second plurality of probabilities, each of the second Petition 870250083544, dated 09 / 17 / 2025,p. 74 / 96 9 / 29 plurality of probabilities corresponding to a respective predicted future breathing state of the user, within a predicted time period; and a processor that is electronically coupled to the memory unit, the processor being configured to generate a control signal to control the breathing assistance device for a current monitoring time period by: receiving sensor data obtained by one or more sensors,The sensor data, including measurements of at least one airflow parameter of the user's airflow during the current monitoring time period when the user is wearing the breathing assistance device; apply the trained predictive model to generate the real-time estimate and forecast; identify one or more false-negative predictions when the sensor data corresponding to the current monitoring time period indicate a breathing event and the real-time estimate and / or forecast corresponding to the current monitoring time period indicate a normal breathing state; extract false-negative data comprising, for each of the one or more false-negative predictions, the real-time estimate, the forecast, and a portion of Petition 870250083544, dated 09 / 17 / 2025,Page 75 / 96 10 / 29 sensor data extending from an initial point in time before the start of a false negative prediction to a second point in time after a shift in the false negative prediction; generate a custom predictive model by retraining the trained predictive model using the false negative data, the custom predictive model being customized for the user; and save the custom predictive model to the memory unit.

24. Controller according to claim 23, the controller being characterized in that it is additionally configured to perform the method as defined in any one of claims 2 to 22.

25. Method for generating a customized predictive model to adjust airflow provided by a breathing assistance device for a user, the method being characterized in that it comprises: implementing a predictive model trained on a processor of a breathing assistance device controller, the trained predictive model being able to generate, based on sensor data, a real-time estimate of the user's current breathing state by determining a first plurality of probabilities, each of the first plurality of probabilities corresponding to a Petition 870250083544, dated 09 / 17 / 2025, p.76 / 96 11 / 29 respective current breathing state of the user, and a prediction of the user's future breathing state by determining a second plurality of probabilities, each of the second plurality of probabilities corresponding to a respective predicted future breathing state of the user, within a predicted time period; receive the sensor data measured from one or more sensors; operate the processor of the respiratory assistance device controller to apply the trained predictive model to generate the real-time estimate and prediction; generate a summary representation of the user, the summary representation comprising user data; generate the custom predictive model by conditioning the trained predictive model using the summary representation, the custom predictive model being customized for the user; and implement the custom predictive model in the processor of the respiratory assistance device controller.

26. Method according to claim 25, characterized in that the user data comprises one or more of the user's weight, height, gender, sex, age, body mass index, apnea-hypopnea index, SpO2, type of mask of the breathing assistance device, prescribed pressure to be provided by the breathing assistance device, type of location, or elevation of the user's location.

27. A method according to claim 25 or 26, characterized in that the user data comprises one or more statistical representations of the user's breathing based on sensor data, the one or more statistical representations of the user's breathing comprising an average waveform of a user's breath, a variance for each time point of the sample in the average waveform, or one or more minimums, maximums, averages, medians, or variances of one or more of the user's airflow, air pressure, tidal volume, respiratory rate, SpO2, heart rate, sound, or movement.

28. A method according to any one of claims 25 to 27, characterized in that the user data comprises one or more statistical representations of the user's environment based on sensor data, the one or more statistical representations of the user's environment comprising one or more minimum, maximum, mean, median, or variance of one or more ambient temperature, CO2, or ambient O2.

29. Method, according to any one of claims 25 to 28, characterized in that the summary representation is generated by the predictive model Petition 870250083544, dated 09 / 17 / 2025, page 78 / 96 13 / 29 trained using one or more embedding layers.

30. A method according to any one of claims 25 to 29, characterized in that conditioning the predictive model trained using the summary representation comprises: providing the summary representation as input to the trained predictive model; adapting a feature representation based on cross-attention to the summary representation, the feature representation being based on sensor data; or providing the summary representation as input to a normalization block distinguished by a shift factor and a scale factor for each feature corresponding to the normalization block being determined by a machine learning model conditioned on the summary representation.

31. A controller for controlling the operation of a respiratory assistance device that provides respiratory assistance to a user, the controller being characterized in that it comprises: a memory unit comprising software instructions and parameters for at least one trained predictive model, the trained predictive model being able to generate, based on sensor data, a real-time estimate of the user's current respiratory state by determining a first plurality of probabilities, each of the first plurality of probabilities corresponding to a respective current respiratory state of the user, and a prediction of the user's future respiratory state by determining a second plurality of probabilities, each of the second plurality of probabilities corresponding to a respective predicted future respiratory state of the user, within a predicted time period;and a processor that is electronically coupled to the memory unit, the processor being configured to generate a control signal to control the respiratory assistance device for a current monitoring time period by: receiving sensor data obtained by one or more sensors, the sensor data corresponding to measurements of at least one airflow parameter of the user's airflow during the current monitoring time period when the user is using the respiratory assistance device; applying the trained predictive model to generate the real-time estimate and prediction; generating a summary representation of the user, the summary representation comprising user data; generating the customized predictive model per Petition 870250083544, dated 09 / 17 / 2025, p. 80 / 96 15 / 29 conditioning the trained predictive model using the summary representation, the customized predictive model being customized for the user;and implement the custom predictive model in the processor of the respiratory assistance device controller.

32. Controller according to claim 31, characterized in that the process is additionally configured to perform the method defined as set forth in any one of claims 26 to 30.

33. A method for simulating one or more operations of a respiratory assistance device and a health status of a user receiving assistance from the respiratory assistance device, the method being characterized in that it comprises: receiving a user model from the user, the user model being adapted to model one or more physiological systems of the user; receiving a device model from the respiratory assistance device, the device model being adapted to model one or more components of the respiratory assistance device, one or more subsystems of the respiratory assistance device, or one or more functions of the respiratory assistance device, or any operable combination thereof;Petition 870250083544, dated 09 / 17 / 2025, page 81 / 96 16 / 29 to receive a customized predictive model to adjust an airflow provided by the breathing assistance device, the customized predictive model being adapted to generate, based on sensor data, a real-time estimate of the user's current breathing state by determining a first plurality of probabilities, each of the first plurality of probabilities corresponding to a respective current breathing state of the user, and a prediction of the user's future breathing state by determining a second plurality of probabilities, each of the second plurality of probabilities corresponding to a respective predicted future breathing state of the user, within a predicted time period; to receive the sensor data from one or more sensors;and determine an expected state of the respiratory assistance device and / or an expected health state of the user based on the application of sensor data to the device model, user model, and / or custom predictive model; determine a device correction factor for the respiratory assistance device to improve the user's health state, based on the user's expected health state; and automatically adjust the operation of the respiratory assistance device according to the device correction factor.

34. Method according to claim 33, characterized in that the one or more sensors include any combination of: (a) one or more of the user sensors placed on the user, (b) device sensors that measure the properties of the breathing assistance device, and (c) environmental sensors.

35. Method according to claim 34, characterized in that the user sensors include a device for measuring the user's blood parameters.

36. A method according to any one of claims 33 to 35, characterized in that the user model is generated, in part, based on one or more personal characteristics of the user, the one or more personal characteristics of the user being received from one or more of: the user, a medical professional, a pharmacy system, a payer system, a health system, other home medical equipment, and an electronic medical records system.

37. Method, according to claim 36, characterized in that one or more personal characteristics comprise the user's alcohol consumption, the user's drug consumption, medication taken by the user, the user's height, the user's weight, the user's age, a blood test result, or any operable combination thereof.

38. A method according to any one of claims 33 to 37, characterized in that the user model is generated based on a model of the user's respiratory system, a model of the user's cardiovascular system, a model of the user's nervous system, or any operable combination thereof.

39. A method according to any one of claims 33 to 38, characterized in that it further comprises: determining a user's current health status by applying sensor data to the user model; comparing the user's current health status with the user's expected health status; and generating a recommendation based on the comparison.

40. A method according to claim 39, characterized in that the user model is associated with a physical model that models one or more of the user's physiological systems, and in that determining the user's current health status comprises applying sensor data to the user's physical model.

41. A method, according to any one of claims 33 to 40, wherein the user's health status comprises sleep health and the method is characterized by the fact that it further comprises: determining one or more expected sleep parameter values ​​for the user by performing a simulation by applying sensor data to the user model and, optionally, to the device; determining one or more actual sleep parameter values ​​for the user based on received sensor data; comparing the one or more expected sleep parameter values ​​and the one or more actual sleep parameter values; and generating a recommendation based on the comparison.

42. A method according to claim 41, characterized in that the one or more sleep parameter values ​​comprise a sleep duration, a length of sleep stages, a sleep depth, a heart rate during sleep, or any operable combination thereof.

43. Method, according to claim 41 or 42, characterized in that the recommendation is a health recommendation, a recommendation to consult a medical professional, a diagnosis, a mask change, a tubing change, and / or an adjustment in the operation of the respiratory assistance device.

44. Method, according to any one of claims 33 to 43, characterized in that the expected state is a future state.

45. A method according to any one of claims 33 to 44, characterized in that it further comprises: simulating the operation of the respiratory assistance device when the device correction factor is applied using the device model; and adjusting the operation of the respiratory assistance device according to the device correction factor when the simulation indicates an improvement in the user's health status when the correction factor is applied.

46. ​​A method according to any one of claims 33 to 45, characterized in that the correction factor of the device is based, in part, on the sensor data received.

47. A method according to any one of claims 33 to 46, characterized in that it further comprises: subsequent to adjusting the operation of the respiratory assistance device, determining an intervention index that distinguishes a probability of the user experiencing a respiratory event at a given time; and when the intervention index exceeds a predetermined threshold, readjusting the operation of the respiratory assistance device by one of: Petition 870250083544, dated 09 / 17 / 2025, pp. 86 / 96 21 / 29 reverting the operation of the respiratory assistance device to an operation prior to the adjustment of the respiratory assistance device operation or modifying the operation of the respiratory assistance device to provide reactive therapy, wherein the probability of the user experiencing the respiratory event at the given time is determined based on the user model.

48. A method, according to any one of claims 33 to 47, characterized in that it further comprises: subsequent to adjusting the operation of the respiratory assistance device, determining a respiratory event index that distinguishes a number of respiratory events experienced by the user in one hour; and determining an average historical number of respiratory events per hour experienced by the user; and when the respiratory event index exceeds the average historical number of respiratory events by a predetermined respiratory event index threshold, readjusting the operation of the respiratory assistance device by one of: reverting the operation of the respiratory assistance device to an operation prior to the adjustment of the respiratory assistance device operation or modifying the operation of the respiratory assistance device to provide reactive therapy.

49. A method according to any one of claims 33 to 48, characterized in that adjusting the operation of the respiratory assistance device comprises adjusting an operating mode of the respiratory assistance device, wherein the operating mode is adjusted by selecting a device profile from: a continuous positive airway pressure (CPAP) profile, an automatic positive airway pressure (APAP) profile, a bilevel positive airway pressure (BiPAP) profile, an adaptive servo-ventilation (ASV) profile, and a non-invasive ventilator (NIV) profile.

50. Method, according to any one of claims 33 to 49, the method being characterized in that it comprises: determining a current state of the respiratory assistance device based on sensor data; comparing the current state of the respiratory assistance device and an expected state of the respiratory assistance device; in response to identifying a difference in the current state from the expected state, determining that one or more components of the respiratory assistance device are defective; and Petition 870250083544, dated 09 / 17 / 2025, pp. 88 / 96 23 / 29 in response to determining that one or more components are defective, generating a device recommendation.

51. Method according to claim 50, characterized in that the device recommendation is to replace one or more components or to perform maintenance on one or more components.

52. A method, according to any one of claims 33 to 51, characterized in that, when the expected state of the respiratory assistance device and / or the expected health status of the user are acceptable, the method comprises implementing at least one model, calibration data, and / or an operational adjustment of the respiratory assistance device that was used during the simulation to update the future operation of the respiratory assistance device.

53. A method according to claim 52, characterized in that the received user model used during the simulation also includes a customized predictive model for the user, and the implementation includes sending the customized predictive model to update the future operation of the respiratory assistance device.

54. Method, according to any one of claims 33 to 53, the method being characterized in that it further comprises performing authentication by receiving a breath signature for the user model and / or sensor data, obtaining a stored breath signature for the user, and performing the simulation when the received breath signature for the user is equal to the stored breath signature for the user.

55. A method according to any one of claims 33 to 54, the method being characterized in that it comprises providing an outlet for interrupting or adjusting therapy in situations where simulation determines that a quantity of air pressure is unsafe for a user who has undergone surgery.

56. Method, according to any one of claims 33 to 55, the method being characterized in that it comprises obtaining an increase in a quantity of recorded data and analyzing the increased data to investigate whether sleep therapy has caused and / or aggravated a medical condition for the user.

57. System for simulating one or more operations of a respiratory assistance device and a health status of a user receiving assistance from the respiratory assistance device, the system characterized in that it comprises: one or more sensors for measuring sensor data; a database storing a user model of the user, the user model being adapted to model one or more internal systems of the user; a device model of the respiratory assistance device, the device model being adapted to model one or more components of the respiratory assistance device, one or more subsystems of the respiratory assistance device, or one or more functions of the respiratory assistance device, or any operable combination thereof.and a customized predictive model to adjust an airflow provided by the breathing assistance device, the customized predictive model being adapted to generate, based on sensor data, a real-time estimate of the user's current breathing state by determining a first plurality of probabilities, each of the first plurality of probabilities corresponding to a respective current breathing state of the user, and a prediction of the user's future breathing state by determining a second plurality of probabilities, each of the second plurality of probabilities corresponding to a respective predicted future breathing state of the user, within a predicted time period; a controller communicating with one or more sensors and the database, the controller comprising at least one processor configured to: i) receive sensor data from one or more sensors; Petition 870250083544, dated 09 / 17 / 2025,pg. 91 / 96 26 / 29 ii) determine one or more expected states of the respiratory assistance device and an expected health state of the user based on one or more of the respiratory assistance device model, the user model, the custom predictive model, and sensor data; iii) determine a device correction factor for the respiratory assistance device to improve the user's health state, based on the user's expected health state; and iv) automatically adjust the operation of the respiratory assistance device according to the device correction factor.

58. System according to claim 57, characterized in that at least one processor is additionally configured to perform the method that is defined as set forth in any one of claims 33 to 56.

59. A method for adjusting an airflow provided by a breathing assistance device for a user, the method being characterized in that it comprises: determining a user-specific predictive model by retraining the trained predictive model using user-specific false negative data associated with false negative predictions of the user's breathing state, the user-specific predictive model being adapted to control the operation of the user-specific breathing assistance device, the user-specific predictive model being adapted to generate, based on sensor data, a real-time estimate of the user's current breathing state by determining a first plurality of probabilities, each of the first plurality of probabilities corresponding to a respective current breathing state of the user,and a prediction of the user's future breathing state by determining a second plurality of probabilities, each of the second plurality of probabilities corresponding to a respective predicted future breathing state of the user, within a predicted time period; perform simulation of the user and the respiratory assistance device when the respiratory assistance device is controlled by the custom predictive model; determine whether the simulation indicates that the use of the custom predictive model has a beneficial effect on the user's health; and when there is a beneficial effect on the user's health, automatically implement the custom predictive model to adjust the future operation of the respiratory assistance device. Petition 870250083544, dated 09 / 17 / 2025, pp. 93 / 96 28 / 29, 60. Method according to claim 59, characterized in that the customized predictive model is determined as defined in any one of claims 1 to 30.

61. Method according to claim 59 or 60, characterized in that the simulation is performed as defined in any one of claims 33 to 56.

62. A method, according to any one of claims 59 to 61, characterized in that false negative data associated with the custom predictive model are used during the simulation.

63. System for adjusting an airflow provided by a breathing assistance device for a user, characterized in that the system comprises a memory that stores program instructions for a method for determining a customized predictive model and a realization of a digital simulation and a processor that is coupled to the memory to receive the program instructions, the processor being configured, during the execution of the program instructions, to realize the method defined as set forth in any one of claims 59 to 62.

64. Non-transient computer-readable media, characterized in that it stores program instructions that are executable by a processor to perform a method that is defined as set forth in any one of claims 1 to 22, 25 to 30, 33 to 56 or 59 to 62.

Images