CROSS-REFERENCE TO PRIOR APPLICATIONS

This application claims the benefit of United States Provisional Application No. 63/054197, filed on 20 July 2020. This application is hereby incorporated by reference herein.

BACKGROUND OF THE INVENTION

1. Field of the Invention

[01] The present invention relates to systems and methods for providing positive airway pressure (PAP) therapy to a patient, and more particularly systems and methods for providing personalized PAP therapy recommendations to a patient. The present invention also relates to predictive artificial intelligence systems for use in such systems and carrying out such methods.

2. Description of the Related Art

[02] Sleep apnea is a common sleep disorder in which patients experience interruptions (apneas) in breathing. Apneas are caused when the airway tissues collapse during sleep. This may happen during parts or the entirety of the sleep duration. These interruptions have a negative impact on a patient’s overall sleep quality, energy during the day, and can also have long-term health consequences.

[03] Positive airway pressure (PAP) therapy is generally considered first-line therapy for the management of sleep-related breathing disorders and obstructive sleep apnea (OSA) syndrome. In PAP therapy, a constant flow of airway pressure provided by a patient interface and mask keeps the patient’s airway open during sleep. This method is highly effective in treating sleep apnea. However, poor adherence to PAP therapy remains the biggest barrier in treatment. Non-adherence rate for PAP therapy has been estimated as high as 46-83%.

[04] Common problems reported by patients undergoing PAP therapy include a leaky mask, trouble falling asleep, stuffy nose, red marks on the skin, sore spots, and dry mouth because therapy devices use a mask and hose to deliver air pressure (either continuous or bilevel PAP). Such side effects, among others, commonly result in PAP therapy not being well tolerated by patients, thus resulting in the aforementioned non adherence rates. Various approaches have been taken to try to improve adherence to PAP therapies by improving the comfort level of the patient interface, mask, and related components. While such approaches have improved comfort levels of PAP therapies, there is still room for improving adherence to such therapies.

SUMMARY OF THE INVENTION

[05] Embodiments of the present invention improve adherence to PAP therapy by providing personalized PAP therapy recommendations to the patient. By personalizing the recommended PAP therapy for a given night, the overall comfort of the device is improved (especially usage-fatigue related discomforts, such as skin red marks, etc.) and users are able to achieve maximum benefit from the PAP therapy while minimizing usage time. This leads to an increased adherence to the PAP therapy by saving the users from avoidable side effects and improved outcomes based on optimizing the times when the PAP therapy is used.

[06] As one aspect of the present invention a method of providing personalized

PAP therapy recommendations to a patient is provided. The method comprises: receiving in a controller patient information obtained from one or more electronic devices associated with the patient; analyzing the patient information with the controller; determining in the controller from the analyzing of the patient information personalized PAP therapy recommendations for the patient; and providing the personalized PAP therapy recommendations to the patient.

[07] The controller may be structured and configured to implement a predictive

AI system, and the analyzing and determining may be performed by the predictive AI system. The predictive AI system may be an artificial neural network trained using one or more of: previous patient information about the patient and/or patient information about a number of other patients. [08] The one or more electronic devices associated with the patient may comprise a wearable smart device. The wearable smart device may comprise a smartwatch.

[09] The one or more electronic devices associated with the patient may comprise a smartphone.

[10] The patient information may comprise subjective information provided by the patient. The subjective information may comprise information regarding alcohol intake by the patient. The subjective information may comprise planned bed time and wake time provided by the patient.

[11] The patient information may comprise objective information gathered passively by sensors in close proximity to the patient.

[12] The method may further comprise receiving additional patient information actively from the patient.

[13] The personalized PAP therapy recommendations may comprise a recommended start and duration time for the PAP therapy.

[14] The personalized PAP therapy recommendations may comprise an indication of an actual benefit that a prior PAP therapy treatment had on the patient’s sleep.

[15] As another aspect of the present invention a system for determining personalized PAP therapy treatment recommendations for a patient and providing such recommendations to the patient is provided. The system comprises: a number of electronic devices, each structured to passively capture patient information from the patient; a controller in communication with the number of electronic devices and structured to carry out an analysis of the patient information and determine from the analysis of the patient information personalized PAP therapy recommendations for the patient; and a user interface structured to convey the personalized PAP therapy recommendations to the patient.

[16] As yet a further aspect of the present invention, a predictive artificial intelligence system is provided that is trained to: receive patient information obtained passively from one or more electronic devices associated with a patient; analyze the patient information; and determine from analyzing the patient information a personalized PAP therapy recommendation.

[17] These and other objects, features, and characteristics of the present invention, as well as the methods of operation and functions of the related elements of structure and the combination of parts and economies of manufacture, will become more apparent upon consideration of the following description and the appended claims with reference to the accompanying drawings, all of which form a part of this specification, wherein like reference numerals designate corresponding parts in the various figures. It is to be expressly understood, however, that the drawings are provided for the purpose of illustration and description only and are not intended as a definition of the limits of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

[18] FIG. 1 is a method for determining and providing personalized PAP therapy recommendations to a patient in accordance with one example embodiment of the present invention;

[19] FIG. 2 is a schematic representation of a system in accordance with one example embodiment of the present invention that may be employed in carrying out one or more methods in accordance with one more example embodiments of the present invention;

[20] FIG. 3 is a schematic diagram showing inputs to, and outputs from, the controller of the system FIG. 2 in accordance with one or more example embodiments of the present invention; and

[21] FIG. 4 is a high-level architecture of back and forth data transmission between cloud servers, sensors and devices in accordance with one example embodiment of the present invention.

DETAILED DESCRIPTION OF THE EXEMPLARY EMBODIMENTS

[22] As used herein, the singular form of “a”, “an”, and “the” include plural references unless the context clearly dictates otherwise. [23] As used herein, the term “number” shall mean one or an integer greater than one (i.e., a plurality).

[24] As used herein, the terms “user”, “patient”, and “individual” are used interchangeably to refer to a person who is the receiver of a PAP therapy and being provided with personalized PAP therapy recommendations.

[25] As used herein, the term “controller” shall mean a number of programmable analog and/or digital devices (including an associated memory part or portion) that can store, retrieve, execute and process data (e.g., software routines and/or information used by such routines), including, without limitation, a field programmable gate array (FPGA), a complex programmable logic device (CPLD), a programmable system on a chip (PSOC), an application specific integrated circuit (ASIC), a microprocessor, a microcontroller, a programmable logic controller, or any other suitable processing device or apparatus. The memory portion can be any one or more of a variety of types of internal and/or external storage media such as, without limitation, RAM, ROM, EPROM(s), EEPROM(s), FLASH, and the like that provide a storage register, i.e., a non-transitory machine readable medium, for data and program code storage such as in the fashion of an internal storage area of a computer, and can be volatile memory or nonvolatile memory.

[26] As previously discussed, adherence to PAP therapy is a substantial problem. Embodiments of the present invention provide systems and methods that improve a patient’s adherence to PAP therapy by providing the patient with personalized PAP therapy recommendations as well as other feedback related to the patient’s PAP therapy and/or response thereto. Such personalized recommendations for the patient are determined through the use of a predictive model that utilizes information gathered from the patient before, during, and after PAP therapy has been provided to the patient (hereinafter “PAP therapy analysis”). As discussed further below, PAP therapy analysis can be achieved using a variety of techniques including, but not limited to, monitoring by means of symptoms, questionnaires, overnight polysomnography (PSG) tests examining the Apnea-Hypopnea Index (AHI), home sleep tests, and other portable sleep apnea monitoring instruments. Such portable sleep apnea monitoring systems include, but are not limited to, respiratory and effort sensors utilizing at least one channel (e.g. saturation of oxygen or airflow). Drawbacks of conventional portable monitoring systems (e.g., complicated sensor attachment, obtrusiveness, sensor detachment, and compliance issues) can be addressed through the use of wearable devices (e.g., a smartwatch) worn by the user. Personalizing the PAP therapy recommendations for a patient helps reduce the impact of the PAP therapy on the patient’s life and improves the overall comfort of the therapy as perceived by the patient. By recommending reduced therapy durations and specific nights the patient requires, or does not require, such therapy, the overall usage requirements and side effects of the therapy are reduced and thus the overall user experience of a patient is improved. By improving the overall user experience of a patient, the patient’s adherence to the PAP therapy is increased.

[27] An example method 10 for determining and providing personalized PAP therapy recommendations to a patient in accordance with one example embodiment of the present invention is shown in FIG. 1. A schematic representation of a system 100 in accordance with one example embodiment of the present invention that may be employed in carrying out method 10 is shown in FIG. 2.

[28] Referring to FIG. 2, system 100 includes a controller 102 that is structured to receive inputs 103 from one or more of a number of input devices 104. Controller 102 may be provided locally as a computing device or (portion thereof) or remotely as a cloud based arrangement. A memory portion of controller 102 has stored therein a number of routines that are executable by a processor portion of controller 102. One or more of the aforementioned routines implement (by way of computer/processor executable instructions) a software application that is configured (by way of one or more algorithms) to, among other things, receive inputs 103 from one or more of the number of input devices 104 and analyze such inputs 103 in order to determine personalized PAP therapy recommendations for a particular patient. In the example embodiment of FIG. 2, controller 102 is provided with a predictive AI system 108, such as a trained neural network or other supervised learning system(s), for this purpose.

[29] As shown in the example embodiment of FIG. 2, input devices 104 may include one or more of: wearable device(s) 110 (e.g., without limitation, smartwatch, smart ring, clip-on sensor(s), smart clothing, adhesive wearable sensor(s)), a smartphone 112, specialized monitors (e.g., without limitation, in-room sensor(s), in/under mattress sensor(s)), a PAP device 116 used in treating the patient, and/or other computing devices 118 (e.g., without limitation, tablet computers, laptops, etc.). Controller 102 is further structured to provide output 105 to a user interface 106, which may be provided on one of the aforementioned devices (e.g., a wearable smartwatch 110, a smartphone 112, PAP device 116, and/or another computing device(s) 118 (e.g., tablet computer, laptop, etc.)) or as a separate output device.

[30] FIG. 3 is a schematic diagram showing example inputs 103 that may be provided to controller 102/predictive AI system 108 of system 100 by one or more of the aforementioned input devices 104, as well as outputs 105 which may be output from controller 102/predictive AI system 108 of system 100 in accordance with one or more example embodiments of the present invention. Such inputs 103 provided to controller 102 may generally be classified as “patient information”, as generally shown in step 12 of method 10 shown in FIG. 1. It is to be appreciated that such patient information may be obtained passively or actively by one or more of input devices 104.

[31] Inputs 103 to controller 102/predictive AI system 108 may include objective inputs typically received passively from one or more sensors of input devices 104 and subjective inputs actively provided by the user or in response to prompts from controller 102. Objective inputs are data including, but not limited to, data received from:

• Wearable device(s) 110 (e.g., without limitation, a smartwatch) - data such as heart rate, respiratory rate, step counts, distance, calories burned, activity tracking, etc. - which may also provide history of sleep parameters (e.g., without limitation, sleep architecture, total sleep time - TST, sleep onset latency - SOL, time in bed - TIB, wake after sleep onset - WASO, apnea-hypopnea index - AHI, arousal index, oxygen desaturation index, oxygen saturation variability, etc.); and/or

• Specialized monitors 114 (e.g., without limitation, an under the mattress sensor) - may also be used to collect more accurate data. • PAP device 116 - data from the PAP device (e.g. AutoCPAP) provides information about the objective AHI as well as the pressures at which the patient was treated by the PAP device. This is important for the continued system learning, as you need to continue to generate severity data when the user is using the PAP device. For example, higher pressures on an AutoCPAP device means that the device needed to maintain a higher pressure in order to maintain airway patency (i.e. higher severity). If the PAP device was able to adequately treat the patient at a lower pressure level, then that indicates a lower severity of OSA for that night or period of time. Hour-by-hour pressure statistics can be used, as well as an aggregate statistic (e.g. average pressure or 90% Auto pressure - the pressure at which the device was at or below 90% of the night).

OSA severity may be measured/determined based on wearable data (or under-mattress sensor) during the night when PAP device was not used.

[32] Subjective inputs provided by the user (e.g., via smartphone 112 and/or other computing device(s) 118) are typically related to demographics (e.g., one-time input - age, gender, body-mass index, etc.), calendar sync options, and information on timing of meals. Some key subjective inputs that may be provided to, and considered by, predictive AI system 108 include (e.g., without limitation):

• Drinking alcohol or taking medications, including sedatives, opioids, or muscle relaxers;

• Prior sleep debt (subjective, based on daytime activity, including physical and cognitive stress, as well as the amount of sleep over the prior nights, if objective sleep and daytime tracking is not available in a given embodiment);

• Smoking or other environmental irritants, traveling to a higher altitude or any other location that could cause breathing problems, health issues related to sleep;

• Information about late meals (intake in unusual timings) and daytime salt intake for predicting fluid retention and rostral fluid shift;

• Bedroom temperature and sleep environment details such as light, air quality etc.;

• Bedpartner and snoring habits; • Projected dominant sleep position, e.g. supine sleep is likely to have high OSA severity of OSA without PAP (AHI during periods without PAP device);

• Daytime activity information such as sedentary, workout, outdoor activities (anything which cannot be tracked by wearable devices), etc.; and

• Target bedtime and wakeup time(s).

Certain other inputs may also be derived from components of system 100 and fed into the machine learning algorithm of predictive AI system 108, such as, sleep need/debt (from sleep tracking metrics), history of PAP benefit (from sleep/AHI data), etc..

[33] Data collected actively from users and passively from devices is migrated to cloud/servers in a secured streaming channel via internet and stored in cloud/servers. Computations for predictive AI system 108 are performed on the cloud, and generated output for users is then displayed on user interface 106 which, as previously discussed may be provided on a wearable device 110 (e.g., a display on a smartwatch), smartphone 112, PAP device 116 (e.g., on a display thereof), other computing device 118 (e.g., a tablet computer), or another suitable output device. FIG. 4 shows a high-level architecture of back and forth data transmission between cloud servers, sensors and devices. Data from the applications/web interfaces are migrated to cloud servers stored in a data lake. APIs handle the requests from backend and push the data accordingly. Tech stack comprises of APIs, listening services, machine learning libraries, databases to store data and cloud infrastructure for running machine learning algorithms. At the user’s end, many services such as notifications, scheduling system, logging framework, alert mechanism (including those for physicians), configuration services, and data auditing are developed. Building PAP Personalization engine is done in series of steps: from data ingestion to, data preparation such as pre-processing stage for data segregation, training the model, improving the model, candidate model evaluation, model deployment, and performance monitoring.

[34] In order to carry out the functionality described herein, predictive AI system 108 is “built” in a series of steps: from data ingestion to, data preparation such as pre-processing stage for data segregation, training the model, improving the model, candidate model evaluation, model deployment, and performance monitoring. [35] Predictive AI system 108 is developed to generate and provide recommendations to users, it is built and works as in the following steps:

1. Building a baseline model - in this stage a model is built and scaled to a production deployable model. This is done in phases to evaluate our model and determine if it is good enough by several factors.

• This phase involves collecting necessary and sufficient data from different users and their devices as described in the above system architecture.

• Data collected goes through a transformation process involving data pre processing, feature engineering, feature selection, filter methods, wrapper/embedded methods and used in training the ML model, evaluating the model.

• PAP Personalization engine model: Subjective inputs from the user are pre- processed using Natural Language Processing models, as necessary, and then combined with the sensor and device data to feed into the machine learning algorithms.

2. Integration and Continuous Learning - in this stage, the model is deployed and integrated into system 100. The model will be in an active learning environment. Predictive AI system 108 learns how users interact with system 100 and completes tasks, using user feedback to adjust the model and customize the person's future recommendations.

[36] As generally shown in steps 14, 16 and 18 of method 10, once built/trained, predictive AI system 108 will analyze the various information regarding the patient received as inputs 103 from one or more of input devices 104 and provide as output 105 personalized PAP therapy recommendations for the patient via user interface 106 to be carried out by the patient. Over time, such arrangement “learns” what dose and timing of PAP therapy works best for a particular patient in particular circumstances and thus will provide recommendations/instructions to the patient for optimum use of their PAP therapy.

[37] Outputs 105 from predictive AI system 108/controller 102 include recommendations on the need to use, or not use therapy on any given night. Recommendations can be binary (e.g. high need to use vs. low need to use) or on a scale (e.g. high benefit from using to low benefit to use). Also, recommendations can provide recommendations over period of times (e.g. low need to use from projected bedtime to 2am, high benefit to use from 2am-6am, etc.). Predictive AI system 108 has the ability to suggest the timing and duration of therapy that the user must undergo that night. Predictive AI System 108 also computes the benefit that PAP therapy received had on the user’s sleep, e.g., decrease in AHI from predicted, arousal index, desaturation index, cardiac risks (e.g. sleep apnea and ejection fraction, nocturnal blood pressure oscillation, etc.), sleep quality, etc. and may provide such information as output 105 which is displayed to the user (e.g., via user interface 106) so as to demonstrate to the patient the benefit(s) of a particular PAP therapy session.

[38] The user is able to get a holistic view of metrics from any sleep session and predicted metrics for an upcoming sleep session via an application provided on their smartphone/tablet/web browser/etc.. Based on inputs like calendar, meals, activity, alcohol intake, etc., predictive AI system 108 can compute a tiredness score and a predicted sleep position, all of which the user can view on a dashboard interface via which a user can receive output from, or provide input to controller 102 (and thus predictive AI system 108). For instance, if the algorithm has data that shows more than 3 or 4 alcoholic drinks that night and that the user has had a busy day, the user’s tiredness score would be high, and the user is more likely to have an elevated AHI. Such information regarding the aforementioned daytime activities increases the likelihood that PAP therapy will be recommended for that given night.

[39] Controller 102/predictive AI system 108 provides recommendations on the need to use therapy on any given night. Output is based on various metrics that are essential for a quality sleep. The model is designed to give various recommendations to different users for example the patient of OSA severity with and without a therapy for the entire night and on an hour-by-hour basis so predicting for the upcoming night OSA severity can be done on a full night and hour-by-hour basis. Output is generally focused on suggesting the necessity of PAP therapy for that night, PAP metric benefit, tiredness score, recommended position of sleep. [40] Based on the subjective inputs of the user, predictive AI system 108 can measure the tiredness score and the predicted sleep position, both of which the user would be able to view on user interface 106. For instance, if the algorithm has data that shows that more than 3 or 4 drinks were consumed and that the user has had a busy day, the user’s tiredness score would be high, and the user is more prone to sleeping on his/her back that night. This might cause significant airway obstruction and necessitates the use of PAP therapy.

[41] As previously discussed, system 100 has the ability to recommend the timing and duration of the therapy the user would have to undergo that night. In one example embodiment, predictive AI system 108 predicts the AHI for the user for the upcoming night on an hour-by-hour basis. Different factors affect the distribution of the sleep disordered breathing events throughout the night; for example, without limitation, alcohol intake may increase early night events, whereas significant daytime sodium intake and fluid loading combined with low activity levels will increase rostral fluid shifting and increase the propensity of later-night events. Increased OSA severity toward the end of the night is also common due to the predominance of REM sleep stage. In this embodiment, system 100 may recommend using the PAP device for a portion of the night in order to significantly decrease the effective AHI for the entire night without the user having to wear the PAP for the duration of the night (e.g. recommend wearing the PAP device from 2:30-6AM, based on the prediction). Optionally, this embodiment may include a smart alarm feature to remind the user to put on the PAP device at a time where the user is naturally awakened close to the targeted time.

[42] In another example, system 100 may recommend whether to use a PAP device for a given night or for certain hours throughout the night (e.g. put the PAP device on before 2AM) based on a total predicted severity indicator (e.g. full-night AHI <10). Other examples of thresholding may also be employed, e.g. for all periods of time where predicted AHI > 10, etc. In some embodiments, the patient will be able to get a comprehensive view of the metrics from each historical sleep sessions and make projections for future sessions. An example of the hour-by-hour predicted severity and recommended time is shown in the figure below. [43] Outputs 105 from controller 102 are used to provide highly targeted coaching to the users of the PAP therapy to improve adherence. For example, without limitation, such coaching may suggest lifestyle changes and/or mask-free alternative therapies to treat sleep apnea based on user preferences.

[44] The personalized dose-response and timing profile will be continuously updated as the user make use of system 100 and more data is available to build an accurate profile. Additionally, predictive AI system 108 provides for continuous adaptation as the user changes their health status (e.g. weight loss/gain, etc.)

[45] In another example embodiment, the projected body position over the course of the night is used to recommend positional therapy (e.g. sleep positional trainer) for usage for the whole night or for a portion of the night (i.e. when a significant portion of sleep is projected in the supine position or when high AHI is projected associated with high AHI in the supine position). In one example embodiment, the projected AHI is relayed to the electronic sleep positional trainer device and the sleep positional trainer device only provides sleep positional training therapy for a portion of the night (similar to the PAP embodiment), e.g. when a projected AHI threshold exceeds a certain level.

[46] In another example embodiment, system 100 recommends usage of a PAP device, no device, or sleep positional therapy device for a given night on the basis of projected AHI and body position. In such example embodiment, system 100 recommends the minimally invasive device required to achieve a goal in SDB severity, in this example AHI < 10. In such example embodiment, system 100 will recommend “no device” for a given night if possible, then “positional therapy” for a given night if it is projected to keep AHI <10, and finally PAP therapy, if necessary. In this example embodiment, the components of system 100 necessary to determine PAP personalization are repeated for sleep positional therapy personalization (e.g. determine predicted SPT Benefit Metric, etc.).

[47] From the foregoing it is thus to be appreciated that embodiments of the present invention provide system and methods that consider one or more of everyday sleep, activity, history of PAP benefit, nutrition, and/or other input from users, and provides guidance to users on the level of need to use PAP therapy for any given sleep session.

[48] Although the invention has been described in detail for the purpose of illustration based on what is currently considered to be the most practical and preferred embodiments, it is to be understood that such detail is solely for that purpose and that the invention is not limited to the disclosed embodiments, but, on the contrary, is intended to cover modifications and equivalent arrangements that are within the spirit and scope of the appended claims. For example, it is to be understood that the present invention contemplates that, to the extent possible, one or more features of any embodiment can be combined with one or more features of any other embodiment.

[49] In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word “comprising” or “including” does not exclude the presence of elements or steps other than those listed in a claim. In a device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The word “a” or “an” preceding an element does not exclude the presence of a plurality of such elements. In any device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The mere fact that certain elements are recited in mutually different dependent claims does not indicate that these elements cannot be used in combination.