An ai-powered detachable inhaler monitoring device and method

Abstract

A detachable inhaler monitoring device is disclosed herein. The monitoring device is configured to removably attach to an inhaler. The monitoring device comprises a processor / microcontroller coupled to a memory, a plurality of sensors configured to monitor at least one of movement of the inhaler and / or parameters of a user using the inhaler and to provide sensor signals to the processor / microcontroller, visual indicators, and a power source for powering the processor / microcontroller, sensors, and memory. The memory comprises program instructions, which when executed on the processor / microcontroller, cause the processor / microcontroller to run an algorithm or an AI / machine learning program / model to analyse the sensor signals and provide an evaluation of the inhaler technique and / or prediction of the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring based on the analysed sensor signals.

Description

[0001] An Al-powered detachable inhaler monitoring device and method

[0002] Field of the invention

[0003] The present disclosure relates to an Al-powered detachable inhaler monitoring device and method, in particular for monitoring parameters relating to the user and inhaler use.

[0004] Background

[0005] Exacerbations of asthma and chronic obstructive pulmonary disease (COPD) are a major cause of acute hospitalizations. Prompt intervention with short-acting bronchodilators, corticosteroids, and / or antibiotics may prevent admissions and improve quality of life, but difficulties in recognizing early symptoms of deterioration often result in delays in accessing care and starting treatment.

[0006] Summary of the invention

[0007] Aspects of the invention are as set out in the independent claims and optional features are set out in the dependent claims. Aspects of the invention may be provided in conjunction with each other and features of one aspect may be applied to other aspects.

[0008] In a first aspect there is provided a detachable inhaler monitoring device, the monitoring device being configured to removably attach to an inhaler. The monitoring device comprises a processor or microcontroller coupled to a memory, a plurality of sensors configured to monitor at least one of movement of the inhaler and / or parameters of a user using the inhaler and to provide sensor signals to the processor or microcontroller, a power source for powering the processor / microcontroller and memory. The memory comprises program instructions, which when executed on the processor / microcontroller, cause the processor / microcontroller to run an algorithm or machine learning or artificial intelligence (Al) program / model to analyse the sensor signals and provide a prediction of the risk of a chronic obstructive pulmonary disease exacerbation occurring based on the analysed sensor signals. The prediction of the risk of a chronic obstructive pulmonary disease exacerbation occurring may be a prediction of the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring within a selected time window of the use of the inhaler, for example within the next 72 hours or over other selected time windows.

[0009] Advantageously, this may allow immediate feedback to be provided using an embedded algorithm or AI / ML coded in the “edge” on the device’s ROM or firmware. By providing the algorithm or machine learning program / model on the device itself, this enables the model to quickly process signals from the sensors in real time, avoiding the need for data to be sent to a remote device and / or avoiding the need for other devices to process the data. Additionally, by computing the predictions on the edge, the predictions will be accessible to the user immediately after using the inhaler device, without the need to open a mobile app or access the Internet.

[0010] While the memory comprises program instructions, which when executed on the processor / microcontroller, cause the processor / microcontroller to run a machine learning program / model to analyse the sensor signals and provide a prediction of the risk of a chronic obstructive pulmonary disease exacerbation occurring based on the analysed sensor signals, it will be appreciated that the machine learning program / model may be configured to support the detection and identification of the risk of other diseases and health conditions. For example, the needs of people with other conditions, including asthma, may be addressed. As such, while the machine learning or Al model may be trained to detect and identify the risk of a chronic obstructive pulmonary disease exacerbation occurring, in other examples the machine learning model may be trained to identify the risk of an asthma attack occurring.

[0011] While the program instructions will implement a machine learning or Al model or algorithm trained from user data, it is important to note that in other examples, the program may utilize other fixed or "pre-trained" models or algorithms. These models or algorithms may involve counting the number of biophysiological variables that fall outside intervals specified by certain fixed thresholds. For instance, these pre-trained models or algorithms may incorporate rules outlined in clinical guidelines for diagnosing and managing acute exacerbations of Chronic Obstructive Pulmonary Disease or Asthma.

[0012] The detachable inhaler monitoring device of embodiments of the disclosure is more advanced than other smart inhalers and will support all of: Adherence tracking, Basic error detection, Complex error detection (such as incorrect orientation of the inhaler device or loose lips, which may be detected by gyro and flow sensors and AI / ML models), Direct on- device feedback (with algorithmic support immediately available at the edge, without necessitating the opening of an app or awaiting feedback from healthcare professionals), and Exacerbation prediction. The detachable inhaler monitoring device of embodiments of the disclosure are capable of Measuring biomarkers (such as respiratory flow, for example using SensirionO’s SDP3x, although it will be understood that other pressure sensors may be used) and support both pressurised metered dose inhalers, pMDI, and Sustainable dry powder inhaler, DPI, devices. It may also be the only inhaler with embedded AI / ML, and the only inhaler displaying warnings derived from biological or physiological signals when the inhaler is used.

[0013] At the high level, the predictions may be carried out by evaluating the probability of an exacerbation occurring within a specified time window, such as 0-72 or 0-168 hours, by using time series or biophysiological readouts as input data. As a special case, the approach includes symptom counting using thresholds specified in clinical guidelines for diagnosis and management of acute exacerbations.

[0014] The machine learning program may implement a machine learning or Al model, such as the ones described in Orchard P, Agakova A, Pinnock H, Burton CD, Sarran C, Agakov F, McKinstry B. Improving Prediction of Risk of Hospital Admission in Chronic Obstructive Pulmonary Disease: Application of Machine Learning to Telemonitoring Data. J Med Internet Res. 2018 Sep 21 ;20(9):e263. doi: 10.2196 / jmir.9227. PMID: 30249589; PMCID: PMC6231768.

[0015] Of course, various other machine learning models can be employed. These models will utilize time series of the readouts of biophysiological variables collected during users' inhaler usage as inputs. Their purpose will be to assess the probability of a mild, moderate, or severe exacerbations based on existing definitions and computed over the pre-specified time windows (the exact choice of the definition and time window may need to stay flexible, as different providers may prefer different definitions). The predictive models may incorporate parametric or non-parametric model types suitable for structured stationary and time series data. These models can include neural networks, regularized classifiers / regression models, maximum margin classifiers (such as support vector machines), ensembles (including random forests), boosted classifiers (such as XGBoost, LightGBM), and other similar approaches or their combinations. In some examples the device further comprises at least one of a visual indicator and / or a vibration element coupled to the processor / microcontroller, and wherein the processor / microcontroller is configured to operate at least one of the visual indicator and / or vibration element to provide a warning, such as an indication of the risk of a chronic obstructive pulmonary disease exacerbation or asthma exacerbation occurring. It will be understood that the device may comprise one or more visual indicators, such as a plurality of visual indicators of respective colours. In some examples, the device may comprise at least one visual indicator, or at least one vibration element, or a combination of both. The processor / microcontroller may be configured to operate at least one of the visual indicator and / or vibration element if the determined risk of an exacerbation occurring is greater than a selected threshold level of risk. Additionally, or alternatively, the processor / microcontroller may be configured to operate at least one of the visual indicator and / or vibration element if the determined risk of an exacerbation occurring is within a selected time window.

[0016] The visual indicator may, for example, be a set of coloured red / amber / green LED indicators or could be a small LED display capable of displaying characters / numerals. The vibration element may be the same type of vibration element commonly used in mobile phones / smartphone, and advantageously may be useful for people with low vision. Advantageously, providing a visual indicator and / or vibration element on the device itself may provide real-time feedback to users without the need for a second device, such as a paired local / mobile device. It may also avoid the need for any wider connectivity - for example, the device may provide immediate real-time feedback to users who only have the device and are in a communications blackspot. It may advantageously be particularly suitable for users who do not engage with, or find it difficult to engage with, smartphones or apps, such as the very young or elderly (see, for example, Cooper R, Giangreco A, Duffy M, Finlayson E, Hamilton S, Swanson M, Colligan J, Gilliatt J, Mclvor M, Sage EK. Evaluation of myCOPD Digital Self-management Technology in a Remote and Rural Population: Real-world Feasibility Study. JMIR Mhealth Uhealth. 2022 Feb 7;10(2):e30782. doi: 10.2196 / 30782. PMID: 35129453; PMCID: PMC8861861 ; and Baumel A, Muench F, Edan S, Kane JM. Objective User Engagement With Mental Health Apps: Systematic Search and Panel-Based Usage Analysis. J Med Internet Res. 2019 Sep 25;21(9):e14567. doi: 10.2196 / 14567. PMID: 31573916; PMCID: PMC6785720).

[0017] The machine learning program / model may additionally or alternatively be configured to determine whether the inhaler has been used incorrectly by the user based on the sensor signals. Examples of incorrect use of the inhaler may include, for example, forgetting to shake (pMDI); wrong orientation; wrong coordination of inhalation and actuation; wrong inhalation duration; loose lips; not holding breath after inhalation; forgetting to actuate at the beginning of inhalation; not waiting long enough between repeat actuations, and others (Usmani OS, Lavorini F, Marshall J, Dunlop WON, Heron L, Farrington E, Dekhuijzen R. Critical inhaler errors in asthma and COPD: a systematic review of impact on health outcomes. Respir Res. 2018 Jan 16;19(1):10. doi: 10.1186 / s12931 -017-0710-y. PMID: 29338792; PMCID: PMC5771074).

[0018] The sensors may comprise a gyroscope, a microphone, and an airflow / pressure sensor, and wherein the machine learning model is configured to determine whether the inhaler has been used incorrectly by the user based on the signals from the gyroscope, microphone, and airflow / pressure sensor. Poor inhaler technique is known to have an impact on health outcomes (see, for example, Usmani OS, Lavorini F, Marshall J, Dunlop WCN, Heron L, Farrington E, Dekhuijzen R. Critical inhaler errors in asthma and COPD: a systematic review of impact on health outcomes. Respir Res. 2018 Jan 16; 19(1 ): 10. doi: 10.1186 / s12931-017-0710-y. PMID: 29338792; PMCID: PMC5771074.). One technique for monitoring inhaler technique may involve the use of a user recording a video of themselves via a smartphone inhaling with the inhaler. However, this can be cumbersome or awkward for the user to hold both the inhaler and the camera / smartphone and suffers from problems in low light conditions. Embedding a gyroscope, microphone, and airflow / pressure sensor in the inhaler casing may reduce these problems. Providing immediate feedback about the inhaler technique on the inhaler device may improve user engagement.

[0019] The device may comprise a communications interface, and wherein the processor / microcontroller is configured to receive an update to the algorithm and / or machine learning program / model via the communications interface. The communications interface, for example, may be configured to communicate over a short-range wireless network, such as Bluetooth® (Including Bluetooth® Low Energy), Wi-Fi® or Zigbee®. In some examples, the processor / microcontroller is configured to send information indicative of the sensor signals to another device, for example via the communications interface.

[0020] Advantageously, this may enable detailed information about exacerbation risks and techniques to be made available via a multidevice app (with user and clinician frontends). However, it is envisaged that essential information will also be available immediately to the user without the apps or the Internet via the at least one visual indicator and / or vibration element.

[0021] The information sent via the communications interface may include, for example, raw data indicative of the sensor signals, or may include pre-processed or filtered data. For example, the data may be filtered to only include changes in parameters indicated by the sensor signals. The data transmitted via the communications interface may include biophysiological readouts collected by the sensors during inhaler use.

[0022] The processor / microcontroller may be configured to run a machine learning model to analyse the sensor signals and provide a prediction of the risk of a chronic obstructive pulmonary disease or asthma exacerbation based on the trajectory of the sensor signals over time. For example, a rolling average of the sensor signals, or other features extracted from the time series, may be recorded, and the trajectory or rate of change of the rolling average or other features may be used to determine or predict the risk of a chronic obstructive pulmonary disease exacerbation. Advantageously, this may predict COPD exacerbations 0-72 hours in advance, for example, although predictions over different time windows may also be possible. This may be done using a recalibration or refinement of a neural network, such as the one used in Orchard P, Agakova A, Pinnock H, Burton CD, Sarran C, Agakov F, McKinstry B. Improving Prediction of Risk of Hospital Admission in Chronic Obstructive Pulmonary Disease: Application of Machine Learning to Telemonitoring Data. J Med Internet Res. 2018 Sep 21 ;20(9):e263. doi: 10.2196 / jmir.9227. PMID: 30249589; PMCID: PMC6231768.

[0023] The subsequent processing may involve any combination of the following:

[0024] (1) Monitoring of symptoms, exacerbations, and inhaler techniques by healthcare providers or community caretakers;

[0025] (2) Integration with health records;

[0026] (3) Training more advanced models using the data stored on a local or remote device (only anonymized data will be required for this purpose);

[0027] (4) Generating predictions using enhanced models that combine the inhaler data with additional user information, such as questionnaires and health records.

[0028] Users will have the option to choose whether they want to share their data or not by setting their preferences in a paired app / device. Clear instructions will be provided to users on how to opt in or opt out of data sharing. Users will be able to change their preferences at any time.

[0029] The sensor signals may provide an indication of physiological parameters, such as respiratory flow, heart rate, oxygen saturation, and body temperature (for example, passively measured during medication routine. Accordingly, in some examples the sensors are configured to provide an indication of respiratory flow / air pressure, heart rate, oxygen saturation, and body temperature to the processor / microcontroller. In some examples the device comprises a reflective sensor for monitoring heart rate and / or oxygen saturation of the user. In some examples the device is configured to obtain an indication of heart rate, oxygen saturation, and body temperature. For example, body temperature may be obtained from a contactless infrared sensor measuring the user’s forehead. For example, the device may comprise an infra-red sensor, for example on a button on the inhaler or on the housing (enclosure, casing) of the inhaler. The sensor for measuring oxygen saturation and heart rate may require the light source and detector to be on the same side of the body part (that is, the reflective sensor).

[0030] In some examples the device further comprises an accelerometer, and wherein the device is configured to operate in a sleep mode and an active mode, and wherein movement of the device as indicated by the accelerometer is configured to trigger the device to switch between the sleep mode and the active mode. Advantageously this may help with power management, to help conserve power so that the power source of the device (which may be a batter) can last longer between charges. For example, in the active mode the processor / microcontroller is configured to receive sensor signals from the plurality of sensors more frequently than in the sleep mode. For example, the accelerometer may be configured to go back to the sleep mode when the inhaler is not used for a specified duration of time. This may benefit users with dementia or Alzheimer's or younger asthma users, among others, who may forget to turn the inhaler on or off.

[0031] In another aspect there is provided a detachable inhaler monitoring device. The monitoring device is configured to removably attach to an inhaler. The monitoring device comprises a processor / microcontroller coupled to a memory; a plurality of sensors configured to monitor at least one of movement of the inhaler and / or parameters of a user using the inhaler and to provide sensor signals to the processor / microcontroller; at least one of a visual indicator and / or a vibration element coupled to the processor / microcontroller; and a power source for powering the processor / microcontroller, the memory, and the visual indicator and / or vibration element. The memory comprises program instructions, which when executed on the processor / microcontroller, cause the processor / microcontroller to run a machine learning program / model to analyse the sensor signals and operate at least one of the visual indicator and / or vibration element to provide an indication of whether the inhaler has been used incorrectly by the user based on the analysed sensor signals.

[0032] The sensors may comprise a gyroscope, a microphone, and an airflow or air pressure sensor, and wherein the machine learning model / program is configured to determine whether the inhaler has been used incorrectly by the user based on the signals from the gyroscope, microphone, and airflow / pressure sensor. Poor inhaler technique is known to have an impact on health outcomes (see, for example, Usmani OS, Lavorini F, Marshall J, Dunlop WON, Heron L, Farrington E, Dekhuijzen R. Critical inhaler errors in asthma and COPD: a systematic review of impact on health outcomes. Respir Res. 2018 Jan 16;19(1):10. doi: 10.1186 / s12931 -017-0710-y. PMID: 29338792; PMCID: PMC5771074.). One technique for monitoring inhaler technique may involve the use of a user recording a video of themselves via a smartphone inhaling with the inhaler. However, this can be cumbersome or awkward for the user to hold both the inhaler and the camera / smartphone and suffers from problems in low light conditions. Embedding a gyroscope, microphone, and airflow / pressure sensor in the inhaler casing may these problems.

[0033] The device may comprise a communications interface, and wherein the processor / microcontroller is configured to receive an update to the machine learning program / model via the communications interface. The communications interface, for example, may be configured to communicate over a short-range wireless network, such as Bluetooth® (including Bluetooth® Low Energy), Wi-Fi® or Zigbee®. In some examples, the processor / microcontroller is configured to send information indicative of the sensor signals to another device, for example via the communications interface. Advantageously, this may enable detailed information about exacerbation risks and techniques to be made available via a multidevice app (with user and clinician frontends). However, it is envisaged that essential information will also be available immediately to the user without the apps or the Internet via the at least one visual indicator and / or vibration element.

[0034] The information sent via the communications interface may include, for example, raw data indicative of the sensor signals, or may include pre-processed or filtered data. For example, the data may be filtered to only include changes in parameters indicated by the sensor signals.

[0035] The processor / microcontroller may be configured to run a machine learning program / model to analyse the sensor signals and provide a prediction of the risk of a chronic obstructive pulmonary disease exacerbation based on the trajectory of the sensor signals over time. For example, a rolling average of the sensor signals or other features may be recorded, and the trajectory or rate of change of the rolling average or other features may be used to determine or predict the risk of a chronic obstructive pulmonary disease exacerbation. Advantageously, this may predict COPD exacerbations 0-72 hours in advance, for example. This may be done using a recalibration or refinement of a multitask recurring neural network, such as the one used in Orchard P, Agakova A, Pinnock H, Burton CD, Sarran C, Agakov F, McKinstry B. Improving Prediction of Risk of Hospital Admission in Chronic Obstructive Pulmonary Disease: Application of Machine Learning to Telemonitoring Data. J Med Internet Res. 2018 Sep 21 ;20(9):e263. doi: 10.2196 / jmir.9227. PMID: 30249589; PMCID: PMC6231768.

[0036] The sensor signals may provide an indication of physiological parameters, such as respiratory flow / air pressure, heart rate, oxygen saturation, and body temperature (for example, passively measured during medication routine). Accordingly, in some examples the sensors are configured to provide an indication of respiratory flow / pressure, heart rate, oxygen saturation, and body temperature to the processor / microcontroller. In some examples the device comprises a reflective sensor for monitoring heart rate and / or oxygen saturation of the user. In some examples the device is configured to obtain an indication of heart rate, oxygen saturation, and body temperature for example from a user’s forehead using, for example using an infra-red sensor. For example, the device may comprise an infra-red sensor, for example on a button on the inhaler or on the housing (enclosure, casing) of the inhaler.

[0037] In some examples the device further comprises an accelerometer, and wherein the device is configured to operate in a sleep mode and an active mode, and wherein movement of the device as indicated by the accelerometer is configured to trigger the device to switch between the sleep mode and the active mode. Advantageously this may help with power management, to help conserve power so that the power source of the device (which may be a batter) can last longer between charges. For example, in the active mode the processor / microcontroller is configured to receive sensor signals from the plurality of sensors more frequently than in the sleep mode.

[0038] In some examples, the detachable inhaler monitoring device comprises a clock module for precise time measurement. The clock module may be configured to record the time duration for each instance of inhaler usage. The processor / microcontroller is configured to use the time measurement in providing a prediction of the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring. Additionally, or alternatively, the processor / microcontroller is configured to use the time measurement in providing an indication of whether the inhaler has been used incorrectly by the user.

[0039] The clock module may be triggered, for example, based on sensor signals from other sensors, such as for example a flow meter configured to measure or record inhalation, and / or from accelerometer measurements. For example, the processor / microcontroller may be configured to control the clock module to record a time duration of inhalation based on a specific combination of sensor signals being received.

[0040] While the examples above describe the inhaler having some on-board processing and machine learning capability, in some examples no analysis using a machine learning program or model may be performed on the device - this may be because there is no machine learning program / model embedded locally on the device, and / or the machine learning program / model is determined to be old or no longer supported (for example, it may not have been updated within a selected time period, and / or it may have an unsupported version number), or because better predictions are available. Such predictions may be based on more comprehensive data than the data collected by the inhaler and may include feedback from healthcare professionals . Instead, or in some examples in addition to, data may be sent to another device for processing, and an indication of, for example, risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring, or incorrect inhaler usage may be provided to the inhaler device, for example for display via the visual indicator or for haptic feedback via the vibration element.

[0041] Additionally, or alternatively, a degree of processing, for example, using an algorithm or machine learning model / program, may be performed locally on the device, but for reasons of processing efficiency and power, this analysis may be less extensive than analysis performed on another device in parallel with the processing performed locally. In such instances, a risk of a chronic obstructive pulmonary disease exacerbation occurring and / or indication of incorrect inhaler usage may be identified by the other device but not by the inhaler device. By processing the data at another device as well, this can help improve the accuracy of the device and feedback to the user and provides a good balance between on-board processing power on the device, and a realistic compromise between cost and power draw if extensive processing is to be performed locally.

[0042] In other examples, predictions produced by external vendors or using external sources of data or information may be displayed on the inhaler device. These predictions could include direct feedback from healthcare providers or AI / ML models utilizing more comprehensive data than what is collected by the inhaler.

[0043] For example, if the inhaler is not connected to the smartphone and is not connected to the cloud, the prediction from the simpler model embedded in the inhaler will be shown. If the inhaler is connected to the smartphone (e.g., via Bluetooth®) or the cloud (e.g., via WiFi®) and if the user agrees to share their data, the prediction of a refined model combining the inhaler data with external data may be shown.

[0044] In such examples, the detachable inhaler monitoring device may be configured to removably attach to an inhaler, the monitoring device comprising a processor / microcontroller coupled to a memory; a communications interface; at least one of a visual indicator and / or a vibration element coupled to the processor / microcontroller; a plurality of sensors configured to monitor at least one of movement of the inhaler and / or parameters of a user using the inhaler and to provide sensor signals to the processor / microcontroller; and a power source for powering the processor / microcontroller, memory, visual indicator and / or vibration element and communications interface. The memory comprises program instructions, which when executed on the processor / microcontroller, cause the processor / microcontroller to receive sensor signals from the plurality of sensors, send the sensor signals to another device (such as the local device or remote device) via the communications interface, and to operate at least one of the visual indicator and / or vibration element in response to receiving signals from the other device indicating that (i) the inhaler has been used incorrectly, and / or (ii) there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring.

[0045] It will be understood that the other device may comprise an app on a mobile / smart phone or may be a computer system running in the “cloud”. Advantageously this does not require the user to use a smartphone or to open the app to still receive feedback indicating that the inhaler was used incorrectly and / or there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring. The cloud may receive both raw and processed data.

[0046] It will be understood that in some examples the inhaler monitoring device may be configured to communicate with a remote device such as a cloud. It will be understood that in some examples the monitoring device may be configured to communicate with a local device instead of, or in addition to a remote device. The local device may comprise an app on a mobile / smart phone. The local device may in turn be configured to communicate with a remote device. The remote device may be a computer system running in the “cloud”. The remote device may communicate directly with the monitoring device, and / or via the local device. The local and / or remote device may require pairing with the inhaler monitoring device, for example an app running on the local device may be paired with the inhaler monitoring device. The pairing may be carried out by a user, for example upon initial set up.

[0047] As noted above, in some examples a degree of pre-processing or filtering may be performed by the processor / microcontroller before data is sent to another device. For example, the processor / microcontroller may be configured to process the sensor signals by filtering the sensor signals prior to sending to the other device via the communications interface. This may reduce the bandwidth needed for communication between the inhaler device and the other device.

[0048] Based on whether the user opts in / out from sharing the data with the app and / or the cloud, the data collected by the inhaler may be less comprehensive than the data collected by the app, which in turn may be less comprehensive than the data collected by the cloud. (For example, the inhaler will collect sensor data, the app may additionally collect symptom questionnaire data, the cloud may additionally link health records and clinical information, including feedback from doctors or nurses). The inhaler, mobile device, and cloud may run different models.

[0049] Predictions of the inhaler model may be available regardless of whether users use apps or the Internet. The predictions of the app-based model may available if the user pairs the inhaler with the app or local / remote device when the inhaler is used and engages with the app (for example, by providing additional user-reported outcome measurements or symptom questionnaires). The predictions of the cloud-based model may be available if the user agrees to send inhaler data to the cloud, pairs the inhaler with the app, and connects to the Internet when the inhaler is used; this cloud-based model may provide even better predictions when health records are linked to the inhaler data and app in the future.

[0050] In some examples, the device further comprises an accelerometer, and the device is configured to operate in a sleep mode and an active mode. For example, movement of the device as indicated by the accelerometer may be configured to trigger the device to switch between the sleep mode and the active mode. Advantageously this may help with power management and reduce power consumption. For example, in the active mode the processor / microcontroller may be configured to receive sensor signals from the plurality of sensors more frequently than in the sleep mode.

[0051] The sensors may be configured to provide an indication of a physiological parameter, such as respiratory flow / pressure, heart rate, oxygen saturation, and body temperature, to the processor / microcontroller. For example, the device may comprise a reflective sensor for monitoring heart rate and / or oxygen saturation of the user. For example, the device may be configured to obtain an indication of heart rate, oxygen saturation, and body temperature from a user’s forehead.

[0052] It will be understood that the device described above may be configured to detachable fasten to a variety of different inhalers and in a variety of different places. In some examples, the device may be configured to fit over the mouthpiece of a conventional inhaler. In other examples, the device may be configured to be mounted over the canister of a pressurised metered dose inhaler, pMDI. In other examples, the device may be configured to be mounted between the canister and the housing of a pressurised metered dose inhaler, pMDI. In other examples, the device is configured to be mounted to a dry powder inhaler, DPI.

[0053] In another aspect there is provided a method of providing an indication of incorrect inhaler use. The method comprises obtaining sensor signals from an inhaler device; processing the sensor signals using a machine learning model; determining that the inhaler has been used incorrectly based on the processing by the machine learning model; and instructing at least one of a visual indicator and a vibration element on the inhaler to operate to provide feedback to the user that the inhaler was used incorrectly.

[0054] In another aspect there is provided a method of predicting the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring. The method comprising obtaining sensor signals from an inhaler device; processing the sensor signals using a machine learning model; determining that there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring based on the processing by the machine learning model; and instructing at least one of a visual indicator and a vibration element on the inhaler to operate to provide feedback to the user that there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring.

[0055] The method may further comprise determining the degree of risk of a chronic obstructive pulmonary or asthma disease exacerbation occurring based on the processing by the machine learning model. Instructing at least one of a visual indicator and a vibration element on the inhaler to operate to provide feedback to the user that there is a risk of a chronic obstructive pulmonary disease exacerbation occurring comprises instructing the visual indicator and / or vibration element to operate based on the degree of risk of a chronic obstructive pulmonary disease exacerbation occurring, such that a different visual indication or vibration is provided to the user for a greater determined risk compared to a lower determined risk.

[0056] In another aspect computer readable non-transitory storage medium comprising a program for a computer configured to cause a processor / microcontroller to perform the method of the aspects described above.

[0057] In another aspect there is provided a computer-implemented method for use in providing a personalised self-management plan to an inhaler device user, the method comprising: identifying, via a machine learning model on the edge on the inhaler device, an elevated risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation for the inhaler device user; obtaining, via a data extraction module, data on population, intervention, comparison, and outcome (PICO) from repositories of clinical trials on COPD and asthma exacerbation; converting, via a data representation module, the obtained PICO data into structured data, wherein the structured data comprises continuous vector embeddings, and wherein the data is structured using established classifications of selfmanagement interventions and population characteristics; training, via a model training module, a meta regression machine learning model to predict the outcomes of the clinical trials based on the structured data, and to predict the required intervention to bring about a wanted outcome for a given inhaler device user based on the characteristics of the inhaler device user; determining, via an intervention derivation module, the intervention required for the inhaler device user with a risk of COPD and asthma exacerbation based on the characteristics of the inhaler device user and the meta regression machine learning model; obtaining, via a plan generation module, a self-management plan based on the determined intervention required; and outputting, via a communication module, the selfmanagement plan to a mobile device associated with the inhaler device user.

[0058] The characteristics of the inhaler device user may comprise historical trajectories of heart rate, oxygen saturation, body temperature, airflow, and predictions of exacerbation risks by the machine learning model.

[0059] In examples, the step of obtaining the PICO data may comprise utilising natural language processing techniques, and wherein the natural language processing techniques used to obtain the PICO data comprise named entity recognition, semantic role labelling, contextual embeddings, summarisation algorithms, and regular expression-based extraction and parsing.

[0060] The population characteristics used to convert the obtained PICO data into structured data may comprise demographic data, medical history, and biomarker levels, and the selfmanagement intervention characteristics used to convert the obtained PICO data into structured data comprise physical activity, lifestyle, inhaler technique, medication adherence, symptom monitoring, disease education and supervision levels, and components of pulmonary rehabilitation.

[0061] The step of determining the intervention required for an inhaler device user may comprise using Bayesian inference.

[0062] The characteristics of the inhaler device user may be obtained from metrics received by the inhaler device and / or other connected or wearable devices used by the inhaler device user.

[0063] The characteristics of the inhaler device user may update via real-time data streams, and wherein the obtaining and outputting of the self-management plan updates in real time.

[0064] In another aspect there is provided a cloud-based device configured to provide a personalised self-management plan to an inhaler device user, the cloud-based device comprising: a machine learning model configured to identify an elevated risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation of the inhaler device user; a data extraction module configured to obtain data on population, intervention, comparison, and outcome (PICO) from repositories of clinical trials on COPD and asthma exacerbation; a data representation module configured to convert the obtained PICO data into structured data, wherein the structured data comprises continuous vector embeddings, and wherein the data is structured using established classifications of selfmanagement interventions and population characteristics; a model training module configured to train a meta regression machine learning model to predict the outcomes of the clinical trials based on the structured data, and to predict the required intervention to bring about a wanted outcome for a given inhaler device user based on the characteristics of the inhaler device user; an intervention derivation module configured to determine the intervention required for the inhaler device user with a risk of COPD and asthma exacerbation based on the characteristics of the inhaler device user and the meta regression machine learning model; a plan generation module configured to obtain a selfmanagement plan based on the determined intervention required; and a communication module configured to output the self-management plan to a mobile device associated with the inhaler device user.

[0065] The characteristics of the inhaler device user may comprise historical trajectories of heart rate, oxygen saturation, body temperature, airflow, and predictions of exacerbation risks by the machine learning model.

[0066] The data extraction module may utilise natural language processing techniques to obtain obtaining the PICO data, and wherein the natural language processing techniques used to obtain the PICO data comprise named entity recognition, semantic role labelling, contextual embeddings, summarisation algorithms, and regular expression-based extraction and parsing.

[0067] The population characteristics used to convert the obtained PICO data into structured data may comprise demographic data, medical history, and biomarker levels, and the selfmanagement intervention characteristics used to convert the obtained PICO data into structured data comprise physical activity, lifestyle, inhaler technique, medication adherence, symptom monitoring, disease education and supervision levels, and components of pulmonary rehabilitation.

[0068] In examples, determining the intervention required for the inhaler device user may comprise using Bayesian inference.

[0069] The characteristics of the inhaler device user may be obtained from metrics received by the inhaler device and / or other connected or wearable devices used by the inhaler device user.

[0070] The characteristics of the inhaler device user may update via real-time data streams, and wherein the obtaining and outputting of the self-management plan updates in real time.

[0071] In another aspect there is provided a method of using an inhaler device for reducing risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation, the method comprising: administering medication via the inhaler device on an as-needed basis; obtaining, via a visual indicator and / or vibration element, feedback on the inhaler device indicating a risk of COPD or asthma exacerbation and the need for a self-management plan; and obtaining, via a mobile device, a self-management plan associated with the inhaler device user.

[0072] The inhaler device may obtain sensor signals when the smart inhaler device is used, wherein the sensor signals correspond to real-time data on the characteristics of the inhaler device user.

[0073] Drawings

[0074] Embodiments of the disclosure will now be described, by way of example only, with reference to the accompanying drawings, in which:

[0075] Fig. 1 shows a schematic diagram of an example inhaler monitoring system, comprising an inhaler with a detachable inhaler monitoring device, a local device and a cloud;

[0076] Fig. 2 shows a functional schematic diagram of an example Printed Circuit Board (PCB) structure of an example inhaler monitoring device, such as the example inhaler monitoring device of Fig. 1 ;

[0077] Fig. 3 shows a cross-section view of a conventional pulse metered dose inhaler, pMDI;

[0078] Fig. 4 shows a cross-section view of the pulse metered dose inhaler of Fig. 3 comprising a detachable inhaler monitoring device of embodiments of the disclosure mounted over the top of the canister;

[0079] Fig. 5 shows a cross-section view of the pulse metered dose inhaler of Fig. 3 comprising a detachable inhaler monitoring device of embodiments of the disclosure mounted on the mouthpiece;

[0080] Fig. 6 shows a cross-section view of the pulse metered dose inhaler of Fig. 3 comprising a detachable inhaler monitoring device of embodiments of the disclosure mounted between the inhaler housing and the canister;

[0081] Fig. 7 shows a cross-section view of a conventional dry powder inhaler, DPI, comprising a detachable inhaler monitoring device of embodiments of the disclosure;

[0082] Fig. 8 shows a flow chart of an example method of providing an indication of incorrect inhaler use;

[0083] Fig. 9 shows a flow chart of an example method of predicting the risk of a chronic obstructive pulmonary disease exacerbation occurring;

[0084] Fig. 10 shows a flow chart of an example computer-implemented method for providing a personalised self-management plan to a smart inhaler device user; and

[0085] Fig. 11 shows a schematic diagram of an example system used to perform the computer- implemented method of Fig. 10.

[0086] Fig. 1 shows a schematic diagram of an example inhaler monitoring system 100. The inhaler monitoring system comprises an inhaler 110 (which in this example is a pMDI) comprising a detachable inhaler monitoring device 200, in communication with a local device 120 (such as a smartphone, laptop, tablet or other computing device). The local device 120 in turn is in communication with a remote device 130 which may be a cloudbased solution accessible via the internet. However, it will be understood that in other examples there may not be a local device and the detachable inhaler monitoring device 200 may be configured to communicate directly with the remote device 130.

[0087] The communication with each of the inhaler monitoring device 200 of the inhaler 110 and local device 120, and local device and remote device 130, may be two-way; that is, the detachable inhaler monitoring device 200 may be operable to both transmit and receive data to / from the local device 120, and the local device 120 may be operable to both transmit and receive data to / from the remote device 130. Although a local device 120 is shown in Fig. 1 , it will be appreciated that in some examples the detachable inhaler monitoring device 200 may be operable to communicate directly with the remote device 130.

[0088] The detachable inhaler monitoring device 200 is configured to detachable fasten to a variety of different inhalers and in a variety of different places. As will be described in more detail below, in some examples the device may be configured to fit over the mouthpiece of a conventional inhaler. In other examples, the device may be configured to be mounted over the canister of a pressurised metered dose inhaler, pMDI. In other examples, the device may be configured to be mounted between the canister and the housing of a pressurised metered dose inhaler, pMDI. In other examples, the device is configured to be mounted to a dry powder inhaler, DPI.

[0089] At a high level, the detachable inhaler monitoring device 200 is configured to wireless connect to a local 120 or remote 130 device via a wireless communications interface, to obtain sensor signals indicative of physiological parameters of a user using the inhaler 110 to which the detachable inhaler monitoring device 200 is attached, and to perform an onboard analysis of the sensor signals to provide immediate feedback to the user on their inhalation technique and / or a predicted risk of an exacerbation of chronic obstructive pulmonary disease or asthma occurring.

[0090] Fig. 2 shows a functional schematic diagram of an example PCB structure of an example inhaler monitoring device 200, such as the example inhaler monitoring device of Fig. 1. It should be understood that the PCB structure is an example. The actual layout of the sensors may be different to what is shown on the figure. To provide the ability to perform on-board analysis as described above, the detachable inhaler monitoring device 200 comprises a processor (CPU) or microcontroller (MCU) 201 coupled to a memory 203, a plurality of sensors 205, 207 configured to monitor at least one of movement of the inhaler and / or parameters of a user using the inhaler and to provide sensor signals to the processor / microcontroller 201 , a power source 209 for powering the processor / microcontroller 201 and memory 203. The memory 203 comprises program instructions, which when executed on the processor / microcontroller 201 , cause the processor / microcontroller 201 to run a machine learning program / model to analyse the sensor signals and provide a prediction of the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring based on the analysed sensor signals.

[0091] The sensors comprise exacerbation prediction sensors 205, and technique evaluation sensors 207. The exacerbation prediction sensors 205 may be configured to provide an indication of a physiological parameter, such as respiratory flow / pressure, heart rate, oxygen saturation, and body temperature, to the processor / microcontroller. For example, the detachable inhaler monitoring device 200 may comprise a heart rate and / or oxygen saturation sensor 205a for monitoring the heart rate and / or oxygen saturation of the user (for example, via their mouth or hand holding the inhaler 110) - this may be a Max30112 sensor, a temperature sensor 205b for measuring the body temperature of the user (for example, via their mouth or hand holding the inhaler 110, or via the user’s forehead using a contactless infrared sensor), and a respiratory flow / pressure sensor 205c - this may be a Sensirion® SDP3x sensor, although of course other sensors may be used.

[0092] The technique evaluation sensors 207 may be used for monitoring the inhalation technique of the user and may include for example a respiratory flow / pressure sensor 205c as described above, a 6-axis gyroscope 207a (for determining the orientation of the inhaler 110) and a microphone 207b. The microphone 207b may be used to provide information regarding inhalation technique and timing of inhalation.

[0093] In the example shown, the detachable monitoring device further comprises an optional clock module 214 which may be a Real-Time Clock (RTC) or a Real-Time Watchdog clock (RWD) module for precise time measurement. The clock module may be triggered, for example, based on sensor signals from other sensors, such as for example a flow meter configured to measure or record inhalation, and / or from accelerometer measurements. For example, the processor / microcontroller may be configured to control the clock module to record a time duration of inhalation based on a specific combination of sensor signals being received.

[0094] This enables the device to accurately track the time duration for each instance of inhaler usage, facilitating more comprehensive monitoring and analysis of inhalation patterns by the user. Moreover, the generated time stamps can be utilized to establish a chronological record of inhaler usage events, enabling the seamless linkage of these events with other user data stored on a local or remote device. This linkage enhances the capability to derive valuable insights and correlations by examining the timing of inhaler usage in relation to other data, including various contextual factors, sensor readouts, or other health indicators.

[0095] In another scenario, the duration of inhaler use can be implemented by utilizing a real-time working counter functionality in software executed by the CPU / MCU 201. By incorporating a dedicated real-time counter into the system, the device can more precisely measure the duration of each inhaler session. This real-time counter starts counting when the inhaler is activated and stops when the inhaler session concludes.

[0096] The recorded duration data can then be leveraged for various purposes, such as monitoring medication adherence, providing dosage recommendations, improving inhalation technique, improving predictions of clinical outcomes such as exacerbations of chronic obstructive pulmonary disease or asthma, or generating usage reports for healthcare professionals or patients themselves. The real-time counter functionality adds an additional dimension to the device's capabilities, enabling more accurate measurement and analysis of inhaler usage durations, thereby enhancing the overall effectiveness and monitoring of respiratory therapy.

[0097] The detachable inhaler monitoring device 200 further comprises a number of I / O elements for interaction with a user or other device. These may include a visual indicator 211 b which may a series of coloured LEDs (e.g., red / amber / green) or a small display (such as an LED, OLED or LCD display) capable of displaying characters or numbers), a vibration element 211d, a button 211a and a communications interface 211c which may be a Bluetooth® Low Energy (BLE) interface. These I / O elements are coupled to the processor / microcontroller 201.

[0098] A power source 209, which may take the form of a battery, provides power to the components of the detachable inhaler monitoring device 200. The detachable inhaler monitoring device 200 further comprises power management features 213. The power management features include an accelerometer 213a, a power switch 213b and an optional power management module 213c.

[0099] The machine learning program / model run by the processor / microcontroller 201 is configured to determine whether the inhaler has been used incorrectly by the user based on the received sensor signals from the technique evaluation sensors 207. The machine learning program / model may be configured to determine technique errors including forgetting to shake (pMDI); wrong orientation; wrong coordination of inhalation and actuation; wrong inhalation duration; loose lips; not holding breath after inhalation; forgetting to actuate at the beginning of inhalation; not waiting long enough between repeat actuations, and others (Usmani OS, Lavorini F, Marshall J, Dunlop WON, Heron L, Farrington E, Dekhuijzen R. Critical inhaler errors in asthma and COPD: a systematic review of impact on health outcomes. Respir Res. 2018 Jan 16;19(1):10. doi: 10.1186 / s12931-017-0710-y. PMID: 29338792; PMCID: PMC5771074).

[0100] Additionally, or alternatively, the machine learning program / model is configured to determine whether the inhaler has been used incorrectly by the user based on the signals from the exacerbation prediction sensors, in particular, the respiratory flow / pressure sensor 205. Because the exacerbation prediction sensors 205 are housed in the detachable inhaler monitoring device 200 which is coupled to or mounted on the inhaler 110, this means that these sensor signals are obtained more easily and naturally from the user, which may address problems of using e.g., a smartphone to capture a video to check inhalation technique - both in terms of the model not working well in low light and also the awkwardness of holding in both hands.

[0101] In particular, the inventors have surprisingly found that a prediction of the risk of a chronic obstructive pulmonary disease exacerbation can be improved if it is based on the trajectory of the sensor signals over time. In other words, the processor / microcontroller 201 is configured to run a machine learning program / model to analyse the sensor signals and provide a prediction of the risk of a chronic obstructive pulmonary disease exacerbation based on the trajectory of the sensor signals over time. This may enable a prediction of chronic obstructive pulmonary disease exacerbations 0-72 hours in advance (for example), using trajectories of respiratory flow, heart rate, oxygen saturation, body temperature passively measured during medication routine. In particular, it has been found that trajectories of these four variables can accurately predict mild, moderate, and severe COPD exacerbations, reducing false alerts 3-fold (at 80% specificity) over conventional symptom-counting scores.

[0102] The processor / microcontroller is configured to operate the I / O elements 211 , for example, at least one of the visual indicator 211 b and / or vibration element 211d to provide an indication of the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring, and / or an indication of incorrect use of the inhaler.

[0103] For example, if the visual indicator 211 b comprises a series of red, amber and greed LED indicators, the processor or microcontroller 201 may control the visual indicator 211 b to illuminate a green indicator may indicate a risk less than a first selected threshold level of risk, the processor / microcontroller 201 may control the visual indicator 211 b to illuminate an amber indicator if the risk is between a first selected threshold level of risk and a second selected threshold level of risk, and the processor / microcontroller 201 may control the visual indicator 211b to illuminate a red indicator if the risk is above the second selected threshold level of risk. Additionally or alternatively, the vibration element 211d may be activated to provide a series of vibrations according to the determined level of risk, for example a single vibration may indicate a risk less than a first selected threshold level of risk, two discrete vibrations may indicate a risk between a first selected threshold level of risk and a second selected threshold level of risk, and three discrete vibrations may indicate a risk is above the second selected threshold level of risk. Advantageously, the vibration element 211 d may be useful for people with low vision.

[0104] Additionally, or alternatively, if the visual indicator 211 b comprises a series of red, amber and greed LED indicators, the processor / microcontroller 201 may control the visual indicator 211b to illuminate a green indicator may indicate correct inhalation technique, the processor / microcontroller 201 may control the visual indicator 211 b to illuminate an amber indicator if there are minor faults in inhalation technique, and the processor / microcontroller 201 may control the visual indicator 211 b to illuminate a red indicator if there are a plurality of inhalation technique errors, or an inhalation technique error that is from a selected list of major inhalation technique errors. Additionally, or alternatively, the vibration element 211d may be activated to provide a series of vibrations according to the determined level of risk, for example a single vibration may indicate correct inhalation technique, two discrete vibrations may indicate minor faults in inhalation technique, and three discrete vibrations may indicate a plurality of inhalation technique errors, or an inhalation technique error that is from a selected list of major inhalation technique errors. Advantageously, the vibration element 211 d may be useful for people with low vision.

[0105] Advantageously, this provides immediate feedback using the embedded algorithm and / or AI / ML coded in the device’s memory 203, with predictions accessible on the detachable inhaler monitoring device 200 itself in real time even for people not engaging with smartphones or apps.

[0106] The processor / microcontroller 201 may also be configured to operate the communications interface 211c to send information indicative of the sensor signals to another device such as the local device 120 or remote device 130. This may enable detailed information about exacerbation risks and techniques to be made available via a multidevice app (with user and clinician frontends) which may be operating on the local device 120 and / or remote device 130. In some examples, the processor / microcontroller 201 is configured to process the sensor signals by filtering the sensor signals prior to sending to the remote device via the communications interface 211c. Advantageously this may reduce the bandwidth of data that the detachable inhaler monitoring device 200 sends via the communications interface 211c. Additionally, the communication with the local or remote device may enable predictions using more accurate and precise models using more comprehensive sources of data and information external to the inhaler.

[0107] Sending data to another device such as the local device 120 or remote device 130 may enable additional detailed information about exacerbation risks, inhaler mistakes, historical data, personalised self-management to be made available via connected user- and clinician-facing multi-device apps with the FHIR (fast healthcare interoperability resources) interoperability layer. In some examples, users will receive push notifications when unusual events, such as exacerbations or technique errors, are detected.

[0108] In some examples, immediate feedback provided using the embedded algorithm and / or AI / ML coded in the device’s memory 203, with predictions accessible on the detachable inhaler monitoring device 200 itself in real time, may prompt the user to pair the device with an app / remote device. For example, an elevated risk detected by the embedded algorithm and / or AI / ML model displayed to the user during the inhaler use may encourage the user to pair the inhaler device with the app and answer questions or provide additional data on the app to get improved predictions or seek medical advice from health or community care providers.

[0109] The other device such as the local device 120 or remote device 130 may be configured to use anonymised user data to improve the machine learning program / model. The improved machine learning model / program may in turn be sent back to the inhaler 110 / detachable inhaler monitoring device 200. In such circumstances, the processor / microcontroller 201 is configured to receive an update to the machine learning program / model via the communications interface 211c.

[0110] In examples where the detachable inhaler monitoring device 200 further comprises power management features 213, the detachable inhaler monitoring device 200 may be configured to operate in a sleep mode and an active mode, and wherein movement of the device as indicated by the accelerometer 213a may be configured to trigger the device to switch between the sleep mode and the active mode. In some examples, the processor / microcontroller 201 is configured to switch between the sleep mode and the active mode, however additionally or alternatively this functionality may be performed by the power management module 213c.

[0111] In the active mode the processor / microcontroller 201 may, for example, be configured to receive or obtain (for example, by polling the sensors for a measurement) sensor signals from the plurality of sensors 205, 207 more frequently than in the sleep mode. While in the examples described above the detachable inhaler monitoring device 200 is configured to perform on-board processing of the sensor signals using a machine learning model / program, in some examples the detachable inhaler monitoring device 200 may be configured to additionally or alternatively send data to another device, such as local device 120 or remote device 130, for processing, for example, in combination with other sources of data and information not available on the inhaler. In such examples, the processor / microcontroller 201 may be configured to receive sensor signals from the plurality of sensors 205, 207, send the sensor signals to another device such as the local device 120 or remote device 130 of Fig. 1 via the communications interface 211c, and to operate at least one of the visual indicator 211 b and / or vibration element 211 d in response to receiving signals from the other device indicating that (i) the inhaler has been used incorrectly, and / or (ii) there is a risk of a chronic obstructive pulmonary disease exacerbation occurring. As noted above, the other device may be the local device 120 or remote device 130 of Fig. 1. Advantageously, this may still not require the user to use a smartphone or to open the app to get feedback on the detachable inhaler monitoring device 200 itself.

[0112] In some examples, the processor / microcontroller 201 is configured to process the sensor signals by filtering the sensor signals prior to sending to the remote device via the communications interface. For example, a high or low band pass filter may be applied, and / or a Fast Fourier T ransform or wavelet transform. This may advantageously reduce the bandwidth required.

[0113] The detachable inhaler monitoring device 200 may also be configured to support over-the- air firmware updates of AI / ML models or algorithms, which can be downloaded to the detachable inhaler monitoring device 200 after validations. The detachable inhaler monitoring device 200 will also be able to support third-party models; for example, pharmaceuticals will be able to recalibrate the warnings or use their own validated algorithms for specific indications and user populations. Anonymised data may also be stored, for example on the local device 120 and / or remote device 130 such as the cloud to enable performance monitoring and model improvement; users will be able to opt in or out of the data sharing. T rained Al models will be programmatically accessible to providers via an API for improved connectivity and future integration with health IT platforms / ecosystems.

[0114] As an example, the Al model may be trained using symptom counting risk scores from clinical guidelines, where certain threshold values are used to detect abnormally low oxygen saturation or abnormally high temperature or resting heart rate. Another example may include AI / ML models learned from data, which may include nonparametric models, regularized parametric models, boosted tree models, ensembles, or neural network models trained from data, for example described in Orchard P, Agakova A, Pinnock H, Burton CD, Sarran C, Agakov F, McKinstry B. Improving Prediction of Risk of Hospital Admission in Chronic Obstructive Pulmonary Disease: Application of Machine Learning to Telemonitoring Data. J Med Internet Res. 2018 Sep 21 ;20(9):e263. doi: 10.2196 / jmir.9227. PMID: 30249589; PMCID: PMC6231768. The training process involves optimization of a predefined supervised, self-supervised, or unsupervised objective function to enhance model performance and generalize to new data. For example, it may involve minimizing a predefined loss function that measures the discrepancy between the model predictions and the ground truth output labels of the training examples. The machine learning or Al model is trained using an iterative optimization algorithm, such as gradient descent, to update the model parameters in each iteration based on the gradients of the loss function. The training process continues until a convergence condition is satisfied, indicating that the model has reached an optimal state with respect to the training data.

[0115] As described above, the detachable inhaler monitoring device 200 is configured to detachable fasten to a variety of different inhalers and in a variety of different places. As will be described in more detail below, in some examples the device may be configured to fit over the mouthpiece of a conventional inhaler. In other examples, the device may be configured to be mounted over the canister of a pressurised metered dose inhaler, pMDI. In other examples, the device may be configured to be mounted between the canister and the housing of a pressurised metered dose inhaler, pMDI. In other examples, the device is configured to be mounted to a dry powder inhaler, DPI. When medication runs out, detachable inhaler monitoring device 200 can be reattached to a new inhaler.

[0116] Fig. 3 shows a cross-section view of a conventional pulse metered dose inhaler, pMDI

[0117] 300. The pMDI 300 comprises a canister 301 mounted in a housing 302 having a mouthpiece 309. The canister 301 comprises propellant and formulation 303. The canister 301 also comprise a metering valve 305 coupled to a nozzle-type actuator 313 via a stem 315. The nozzle-type actuator 313 comprise a spray nozzle 311. Depression of the canister 301 into the housing 302 causes the stem to travel into the metering valve to cause propellant and formulation to be sprayed out of the spray nozzle 311 and through the mouthpiece 309 as an aerosol spray 307.

[0118] Fig. 4 shows a cross-section view of the pulse metered dose inhaler 300 of Fig. 3 comprising a detachable inhaler monitoring device 200 of embodiments of the disclosure mounted over the top of the canister 301 . In this example it can be effectively seen how the detachable inhaler monitoring device 200 may effectively be an extension of the housing 302. In this example, the detachable inhaler monitoring device 200 may have a button or be otherwise depressible to still enable a user to depress the canister 301. In this way, the user will have to physically touch the detachable inhaler monitoring device 200 and this may be how physiological parameters may be obtained from the user via the detachable inhaler monitoring device 200.

[0119] Fig. 5 shows a cross-section view of the pulse metered dose inhaler 300 of Fig. 3 comprising a detachable inhaler monitoring device 200 of embodiments of the disclosure mounted on the mouthpiece 309. In this way, when the user places their mouth around the inhaler 300 to inhale, they will place their mouth around the detachable inhaler monitoring device 200 and this may be how physiological parameters may be obtained from the user via the detachable inhaler monitoring device 200.

[0120] Fig. 6 shows a cross-section view of the pulse metered dose inhaler 300 of Fig. 3 comprising a detachable inhaler monitoring device 200 of embodiments of the disclosure mounted between the inhaler housing 302 and the canister 301 .

[0121] It will be appreciated from the above that detachable inhaler monitoring device 200 of embodiments of the disclosure may be detachably mounted to a range of different inhalers.

[0122] Fig. 7 shows a cross-section view of a conventional dry powder inhaler, DPI, 700. The DPI 700 comprises a body 702 holding a blister strip of medicament. The inhaler 700 is configured such that movement of the mouthpiece cover 730 advances the blister strip such that inhalation on the mouthpiece (not shown, covered under mouthpiece cover 730) inhales a dose of medicament. The inhaler monitoring device 300 of embodiments of the disclosure may also be configured to be used with a DPI such as the DPI 700 shown in Fig. 7. For example, the inhaler monitoring device 300 may be coupled to the mouthpiece, the mouthpiece cover 730 and / or the inhaler body 702.

[0123] Fig. 8 shows a flow chart of an example method 1000 of providing an indication of incorrect inhaler use. The method comprises the steps of obtaining 1010 sensor signals from an inhaler device such as the detachable inhaler monitoring device 200 described above. The method then comprises processing the sensor signals. The sensor signals may be processed 1020a locally using a local algorithm or machine learning model on the inhaler monitoring device itself; the sensor signals may be processed 1020b locally using a local algorithm or machine learning model on the inhaler monitoring device itself in combination with a degree of processing on a local device (such as a paired mobile phone running an app); and / or the sensor signals may be processed 1020c locally using a local algorithm or machine learning model on the inhaler monitoring device itself in combination with processing on a remote device (such as the cloud) with optional feedback from a healthcare professional. In other words, the sensor signals may be processed in any combination of locally using a local algorithm or machine learning model on the inhaler monitoring device itself, processed locally using a local algorithm or machine learning model on the inhaler monitoring device itself in combination with a degree of processing on a local device (such as a paired mobile phone running an app); and processed locally using a local algorithm or machine learning model on the inhaler monitoring device itself in combination with processing on a remote device (such as the cloud) with optional feedback from a healthcare professional. That is, the sensor signals may be processed on the inhaler device or on a local device or on a remote (e.g., cloud) device or on all three devices or on any subset of these three devices in some combination.

[0124] The method then comprises determining 1030 that the inhaler has been used incorrectly based on the processing; and instructing 1040 at least one of a visual indicator 211b and a vibration element 211d on the inhaler or detachable inhaler monitoring device 200 to operate to provide feedback to the user that the inhaler was used incorrectly. Fig. 9 shows a flow chart of an example method 2000 of predicting the risk of a chronic obstructive pulmonary disease exacerbation occurring. The method comprises the steps of obtaining 2010 sensor signals from an inhaler device such as the detachable inhaler monitoring device 200 described above. The method then comprises processing the sensor signals. The sensor signals may be processed 2020a locally using a local algorithm or machine learning model on the inhaler monitoring device itself; the sensor signals may be processed 2020b locally using a local algorithm or machine learning model on the inhaler monitoring device itself in combination with a degree of processing on a local device (such as a paired mobile phone running an app); and / or the sensor signals may be processed 2020c locally using a local algorithm or machine learning model on the inhaler monitoring device itself in combination with processing on a remote device (such as the cloud) with optional feedback from a healthcare professional. In other words, the sensor signals may be processed in any combination of locally using a local algorithm or machine learning model on the inhaler monitoring device itself, processed locally using a local algorithm or machine learning model on the inhaler monitoring device itself in combination with a degree of processing on a local device (such as a paired mobile phone running an app); and processed locally using a local algorithm or machine learning model on the inhaler monitoring device itself in combination with processing on a remote device (such as the cloud) with optional feedback from a healthcare professional. That is, the sensor signals may be processed on the inhaler device or on a local device or on a remote (e.g., cloud) device or on all three devices or on any subset of these three devices in some combination.

[0125] The method then comprises determining 2030 that there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring based on the processing; and instructing 2040 at least one of a visual indicator 211 b and a vibration element 211 d on the inhaler or detachable inhaler monitoring device 200 to operate to provide feedback to the user that there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring.

[0126] In some examples, instructing at least one of the visual indicator 211 b and a vibration element 211 d on the inhaler device or detachable inhaler monitoring device 200 to operate to provide feedback to the user that there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring comprises instructing the visual indicator and / or vibration element to operate based on the degree of risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring, such that a different visual indication or vibration is provided to the user for a greater determined risk compared to a lower determined risk.

[0127] For example, a green indicator may be illuminated to indicate a risk less than a first selected threshold level of risk, an amber indicator may be illuminated if the risk is between a first selected threshold level of risk and a second selected threshold level of risk, a red indicator may be illuminated if the risk is above the second selected threshold level of risk. Additionally or alternatively, the vibration element 211d may be activated to provide a series of vibrations according to the determined level of risk, for example a single vibration may indicate a risk less than a first selected threshold level of risk, two discrete vibrations may indicate a risk between a first selected threshold level of risk and a second selected threshold level of risk, and three discrete vibrations may indicate a risk is above the second selected threshold level of risk. Advantageously, the vibration element 211 d may be useful for people with low vision.

[0128] Additionally, or alternatively, if the visual indicator 211 b comprises a series of red, amber and greed LED indicators, a green indicator may be illuminated to indicate correct inhalation technique, an amber indicator illuminated if there are minor faults in inhalation technique, and a red indicator illuminated if there are a plurality of inhalation technique errors, or an inhalation technique error that is from a selected list of major inhalation technique errors. Additionally, or alternatively, the vibration element 211d may be activated to provide a series of vibrations according to the determined level of risk, for example a single vibration may indicate correct inhalation technique, two discrete vibrations may indicate minor faults in inhalation technique, and three discrete vibrations may indicate a plurality of inhalation technique errors, or an inhalation technique error that is from a selected list of major inhalation technique errors. Advantageously, the vibration element 211d may be useful for people with low vision.

[0129] There is a lack of a gold standard definition for what constitutes an exacerbation, which is a challenge to research in this area. Many mild to moderate exacerbations may be defined by medication use. In some examples, moderate to severe COPD may be defined by forced expiratory volume in the first second of expiration [FEV] and forced vital capacity both <70%. Note that this is the definition of COPD, which is different from the definition of COPD exacerbation.

[0130] Alternatively, an exacerbation may be defined using one of at least five different definitions based on major symptoms and minor symptoms. Major symptoms may include changes in users’ self-reported breath-lessness, sputum colour, and sputum amount, and minor symptoms include cold, wheeze, sore throat, cough, and fever.

[0131] Fig. 10 shows a flow chart of an example computer-implemented method 3000 for providing a personalised self-management plan to an inhaler device 110 user. The method comprises the steps of identifying, via a machine learning model on the edge on the inhaler device 110, an elevated risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation for the inhaler device 110 user 3010. The machine learning model is located on the edge on the smart inhaler device as it is located in the detachable inhaler monitoring device 200 which wirelessly connects to a local 120 or remote 130 device via a wireless communication interface, as discussed with reference to Fig. 1. Further details of the onboard inhaler monitoring device 200 and its ability to perform on-board analysis from sensor signals is described above with reference to Fig. 2.

[0132] The method then comprises obtaining data, via a data extraction module on the populations, interventions, comparisons, and outcomes (PICO) from repositories of clinical trials on COPD and asthma exacerbation 3020. The data extraction module is located in the remote device 120 which may be a cloud-based solution accessible via the internet. In examples, the data is taken from external databases and utilises natural language processing techniques such as named entity recognition, semantic role labelling, contextual embeddings, summarisation algorithms, and regular expression-based extraction and parsing. In other examples, the data extraction module is connected to a data storage module which holds previous smart inhaler device user PICO data.

[0133] The method then comprises converting, via a data representation module, the obtained PICO data into structured data, wherein the structured data comprises continuous vector embeddings, and wherein the data is structured using established classifications of selfmanagement interventions and population characteristics 3030. The data representation module is also located on the remote device 130. In examples, the population characteristics used to convert the data into structured data comprises demographic data, medical history, and biomarker levels, and the self-management intervention characteristics used to convert the obtained PICO data into structured data comprise physical activity, lifestyle, inhaler technique, medication adherence, symptom monitoring, disease education and supervision levels, and components of pulmonary rehabilitation.

[0134] The method then comprises training, via a model training module, a meta regression machine learning model to predict the outcomes of the clinical trials based on the structured data, and to predict the required intervention to bring about a wanted outcome for a given smart inhaler device user based on the characteristics of the inhaler device 110 user 3040. The training module is located on the remote device 130 and is in communication with the local device 120 and detachable inhaler monitoring device 200 in order to receive the characteristics of the inhaler device 110 user. In examples, the characteristics of the inhaler device 110 user comprises historical trajectories of heart rate, oxygen saturation, body temperature, airflow, and predictions of exacerbation risks by the machine learning model. Characteristics of the inhaler device 110 user are optionally provided by previously mentioned sensors on board the example inhaler monitoring device 200 of Fig. 1 and by external software containing data of the inhaler device 110 user. The external software may be located on the local device 120 associated with the inhaler device 110 user. As such, the characteristics of the inhaler device 110 user are obtained from metrics received by the onboard monitoring device 200 onboard the inhaler device 110 and / or other connected or wearable devices used by the inhaler device 110 user. The sensors provide real-time data of the characteristics of the inhaler device 110 user to the processor / microcontroller onboard the inhaler device 110 and local 120 and remote 130 devices, and the external software provides additional data, which is not detected by the sensors, to the processor / microcontroller onboard the inhaler device 110 and remote devices 130. For example, additional data may comprise the height and weight of the smart inhaler device 110 user. The sensors onboard the inhaler device 110 comprise exacerbation prediction sensors 205 and technique evaluation sensors 207 wherein the exacerbation prediction sensors 205 may be configured to provide an indication of a physiological parameter, such as respiratory flow / pressure, heart rate, oxygen saturation, and body temperature. The detachable inhaler monitoring device 200 may comprise a heart rate and / or oxygen saturation sensor 205a for monitoring the heart rate and / or oxygen saturation of the inhaler device 110 user, a temperature sensor 205b for measuring the body temperature of the user, and a respiratory flow / pressure sensor 205c.

[0135] The method then comprises determining, via an intervention derivation module, the intervention required for the inhaler device 110 user with a risk of COPD and asthma exacerbation based on the characteristics of the inhaler device 110 user and the meta regression machine learning model 3050. The intervention derivation module is also located on the remote device 130. In examples, the intervention required for an inhaler device 110 user is determined using uses Bayesian inference. The interventions determined for an inhaler device 110 user comprises, for example, medications, oxygen therapy, pulmonary rehabilitation, vaccinations, lifestyle changes, and surgery.

[0136] The method then comprises obtaining, via a plan generation module, a self-management plan based on the determined intervention required 3060 and outputting, via a communication module, the self-management plan to a mobile device associated with the inhaler device 110 user 3070. The plan generation module and communication module are located on the remote device 130 and the mobile device makes up at least one of the local devices 120 which are in communication with the inhaler device 110 and remote device 130. In examples, the characteristics of the inhaler device 110 user update via realtime data streams so that obtaining and outputting the self-management plan updates in real time.

[0137] The computer-implemented method 3000 for providing a personalised self-management plan to the inhaler device 110 user is configured to predict the outcomes of patients from clinical trials with COPD or asthma exacerbation based on the interventions used. The method 3000 is then configured to use the predictions and the characteristics of the inhaler device 110 user, provided by sensors onboard the inhaler device 110 and from external data sources, to suggest interventions required for the inhaler device 110 user if they are deemed to have a risk of COPD or asthma exacerbation. The interventions are provided in the form of a self-management plan and are configured to either reduce or nullify their symptoms of COPD or asthma exacerbation.

[0138] In use, the computer-implemented method 3000 for providing a personalised selfmanagement plan to an inhaler device 110 user identifies if an inhaler device 110 user has a risk of COPD and asthma exacerbation and provides a self-management plan to reduce or nullify the symptoms of the user.

[0139] The step of identifying, via a machine learning model on the edge on the smart inhaler device, an elevated risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation for the smart inhaler device user 3010 utilises a method for predicting the risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation. This method comprises obtaining, via a sensor interface module, sensor signals from an inhaler device that has been used by an inhaler device user and processing, via a processing module, the sensor signals. The method then comprises determining, via a risk assessment module, the risk of a COPD or asthma exacerbation occurring based on the processed sensor signals and outputting, via a feedback module, feedback if the smart inhaler device user has a risk of COPD or asthma exacerbation.

[0140] As previously discussed, the sensor signals are determined by the sensors onboard the smart inhaler device and feedback is outputted through the smart inhaler device by way of the visual indicator 211 b and a vibration element 211d on the inhaler device or detachable inhaler monitoring device 200.

[0141] The inhaler device 110 user follows a method of using an inhaler device 110 for reducing risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation. The method comprises utilising an inhaler device 110 as required, such as for maintenance therapy, quick relief, or before exercise. The method then comprises obtaining, via a visual indicator 211 b and / or vibration element 211 d, feedback on the inhaler device 110 indicating a risk of COPD or asthma exacerbation and the need for a self-management plan. The visual indicator 211 b and / or vibration element 211d are described in further detail with reference to Fig. 2 and feedback indicating a risk of COPD or asthma exacerbation is determined using the previously mentioned machine learning model on the edge on the onboard inhaler monitoring device 200. Lastly, the method comprises obtaining, via a mobile device, a self-management plan associated with the inhaler device 110 user.

[0142] The method of using an inhaler device 110 for reducing risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation analyses the regular use of an inhaler device 110 by an inhaler device 110 user and provides a self-management plan in the case that the inhaler device 110 user is at risk of COPD or asthma exacerbation. The selfmanagement plan shows information to the user which, when used or followed correctly, reduces the risk of COPD or asthma exacerbation of the user.

[0143] Fig. 11 shows a schematic diagram of an example system 4000 used to perform the computer-implemented method of Fig. 10. The system 4000 comprises a cloud-based device comprising a machine learning model 4010, a data extraction module 4030, a data representation module 4040, a model training module 4050, an intervention module 4060, a plan generation module 4070, and a communication module 4080. The cloud-based device interacts with an inhaler device 110, repositories of clinical trials 4020, and a mobile device 4090. In examples, the repositories of clinical trials 4020 are external databases. In other examples, the repositories of clinical trials 4020 may instead be previous inhaler device data.

[0144] The inhaler device 110 is configured to provide data on the use of the inhaler from an onboard monitoring device 200 to the machine learning model 4010. The machine learning device is configured to identify an elevated risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation for the inhaler device user 3010 wherein the risk of COPD or asthma exacerbation is identified by a predetermined threshold. The predetermined threshold may comprise certain levels of the characteristics of the inhaler user detected by the inhaler monitoring device, as previously mentioned. These characteristics may comprise heart rate, oxygen saturation, body temperature, and the respiratory flow / pressure of the inhaler device user. The data extraction module 4030 is configured to obtain data on population, intervention, comparison, and outcome (PICO) from the repositories of clinical trials 4020 on COPD and asthma exacerbation 3020. The data representation module 4040 is configured to convert the obtained PICO data into structured data, wherein the structured data comprises continuous vector embeddings, and wherein the data is structured using established classifications of self-management interventions and population characteristics 3030.

[0145] The model training module 4050 is configured to receive the structured data from the data representation module 4050 and the characteristics of the inhaler user and identified risk from the machine learning model 4010. This model training module 4050 is configured to train a meta regression machine learning model to predict the outcomes of the clinical trials based on the structured data and predict the required intervention to bring about a wanted outcome for the given inhaler device user based on the characteristics of the inhaler device user 3040. The intervention derivation module 4060 is configured to determine the intervention required for the inhaler device user with a risk of COPD and asthma exacerbation based on the characteristics of the inhaler device user and the meta regression machine learning model 3050. The plan generation module 4070 and communication module 4080 are configured to obtain a self-management plan based on the determined intervention required 3060 and output the self-management plan to a mobile device 4090, 3070.

[0146] In use the system 4000, using the cloud-based device, identifies if an inhaler device user has a risk of COPD or asthma exacerbation, determines an appropriate intervention to reduce or negate the risk using example external data, and produces and outputs a personalised self-management plan for the inhaler device user.

[0147] A user using the inhaler monitoring device may also follow a method to reduce a risk of COPD or asthma exacerbation if indicated to do so. The method comprises selfadministering medication via the inhaler device on an as-needed basis. In the case where the inhaler device user does have a risk of COPD or asthma exacerbation, as identified by the machine learning model 4010, the method comprises obtaining, via a visual indicator and / or vibration element, feedback on the inhaler device indicating a risk of COPD or asthma exacerbation and the need for a self-management plan. Lastly, the method comprises obtaining, via a mobile device, a self-management plan associated with the inhaler device user.

[0148] Major symptoms Breathlessness, sputum colour, and sputum amount.

[0149] Minor symptoms

[0150] Cold, wheeze, sore throat, cough, and fever.

[0151] Symptom counts a. nMajor = number of major symptoms present on day t, b. nMinor = number of minor symptoms present on day t, c. nAII = nMajor + nMinor.

[0152] Definitions a. Definition 1 : nMajor>2. b. Definition 2: nAII>5. c. Definition 3: define a 'bad day' as one where (nMajor>2) or ([nMajor=1] and [nMinor>1]). An exacerbation is said to occur on day t if days t and t +1 are bad, but days t -1and t -2 are not bad. d. Definition 4: Like Definition 3, but a bad day is defined as one where (nMajor>1) and (nAII>3). e. Definition 5: An exacerbation is said to occur on day t if: i. (nAII>5) on day t, or ii. (nAII=4) on day t and (nAII>4) on day t +1.

[0153] In another example, an exacerbation may be defined as a hospitalization due to a rapid worsening of symptoms of chronic obstructive pulmonary disease or asthma, or a change in treatment as prescribed in a treatment plan, or as defined in clinical guidelines or best practices for diagnosing acute exacerbations of chronic obstructive pulmonary disease or asthma.

[0154] In some examples the processor / microcontroller201 may be configured to operate at least one of the visual indicator 211 b and / or vibration element 211 d if the determined risk of an exacerbation occurring comprises a determination of an exacerbation according to one of the five definitions above occurring.

[0155] It will also be appreciated that while the examples above describe the prediction of a chronic obstructive pulmonary disease exacerbation occurring, embodiments of the disclosure can also be applied to the needs of people with other conditions, including asthma.

[0156] As utilized herein, terms “component,” “system,” “interface,” “unit” and the like are intended to refer to a computer-related entity, hardware, software (e.g., in execution), and / or firmware. For example, a component can be a processor / microcontroller, a process running on a processor / microcontroller, an object, an executable, a program, a storage device, and / or a computer. By way of illustration, an application running on a server and the server can be a component. One or more components can reside within a process, and a component can be localized on one computer and / or distributed between two or more computers.

[0157] Further, these components can execute from various computer readable media having various data structures stored thereon. The components can communicate via local and / or remote processes such as in accordance with a signal having one or more data packets (e.g., data from one component interacting with another component in a local system, distributed system, and / or across a network, e.g., the Internet, a local area network, a wide area network, etc. with other systems via the signal).

[0158] As another example, a component can be an apparatus with specific functionality provided by mechanical parts operated by electric or electronic circuitry; the electric or electronic circuitry can be operated by a software application or a firmware application executed by one or more processors / microcontrollers; the one or more processors / microcontrollers can be internal or external to the apparatus and can execute at least a part of the software or firmware application. As yet another example, a component can be an apparatus that provides specific functionality through electronic components without mechanical parts; the electronic components can include one or more processors / microcontrollers therein to execute software and / or firmware that confer(s), at least in part, the functionality of the electronic components. In some cases, a component can emulate an electronic component via a virtual machine, e.g., within a cloud computing system.

[0159] The local device 120 may be a computing device configured to perform one or more operations consistent with the disclosed embodiments. Examples of local devices 120 may include, but are not limited to, mobile devices, smartphones / cellphones, wearable device (e.g., smartwatches), tablets, personal digital assistants (PDAs), laptop or notebook computers, desktop computers, media content players, television sets, video gaming station / system, virtual reality systems, augmented reality systems, microphones, or any electronic device capable of analyzing, receiving (e.g., receiving user input in one or more fields in a form, receiving definition of rules associated with a page, etc.), providing or displaying certain types of data (e.g., system generated personalization parameter options, original document, derivative document, etc.) to a user. The local device 120 may be a handheld object. The local device 120 may be portable. The local device 120 may be carried by a human user. In some cases, the local device 120 may be located remotely from a human user, and the user can control the user device using wireless and / or wired communications. The remote device 120 can be any electronic device with a display.

[0160] The processor / microcontroller 201 of the detachable inhaler monitoring device 110 may include one or more processors / microcontrollers 201 that are capable of executing non- transitory computer readable media that may provide instructions for one or more operations consistent with the disclosed embodiments. The detachable inhaler monitoring device 110 may include one or more memory storage devices 203 comprising non- transitory computer readable media including code, logic, or instructions for performing the one or more operations. The user device may include software applications that allow the user device to communicate with and transfer data between the local device 120 and / or remote device 130.

[0161] In some cases, the remote device 130, which may be in the form of a server or cloud, may also be configured to store, search, retrieve, and / or analyze data and information stored in one or more of the databases. The data and information may include reader interaction data collected from the user / customer device as well as brand style data, mapping relationship for brand style management, content or document data, ruleset associated with a content item, data about a predictive model (e.g., parameters, model architecture, training dataset, performance metrics, threshold, etc.), data generated by a predictive model such as personalization options or extracted insight, reader feedback survey, and the like. While FIG. 1 illustrates the server as a single server, in some embodiments, multiple devices may implement the functionality associated with a server. The remote device 130 may include a web server, an enterprise server, or any other type of computer server, and can be computer programmed to accept requests (e.g., HTTP, HTTPS, or other protocols that can initiate data transmission) from a computing device (e.g., user device and / or meeting capturing device) and to serve the computing device with requested data. In addition, a server can be a broadcasting facility, such as free-to-air, cable, satellite, and other broadcasting facility, for distributing data. The remote device 130 server may also be a server in a data network (e.g., a cloud computing network).

[0162] The remote device 130 and local device 120 may include known computing components, such as one or more processors, one or more memory devices storing software instructions executed by the processor(s), and data. A server can have one or more processors and at least one memory for storing program instructions. The processor(s) can be a single or multiple microprocessors, graphics processing units (GPUs), field programmable gate arrays (FPGAs), or digital signal processors (DSPs) capable of executing particular sets of instructions. Computer-readable instructions can be stored on a tangible non-transitory computer-readable medium, such as a hard disk, a CD-ROM (compact disk-read only memory), and MO (magneto-optical), a DVD-ROM (digital versatile disk-read only memory), a DVD RAM (digital versatile disk-random access memory), or a semiconductor memory. Alternatively, the methods can be implemented in hardware components, combinations of hardware and software, or software that is embedded in a hardware, such as, for example, firmware, ASICs, special purpose computers, or general-purpose computers.

[0163] The detachable inhaler monitoring device 200, local device 120 and remote device 130 may be connected by a network that is configured to provide communication between the various components illustrated in Fig. 1. The network may be implemented, in some embodiments, as one or more networks that connect devices and / or components in the network layout for allowing communication between them. For example, the detachable inhaler monitoring device 200, local device 120 and remote device 130 may be in operable communication with one another over the network. Direct communications may be provided between them. The direct communications may occur without requiring any intermediary device or network. Indirect communications may be provided between two or more of the above components. The indirect communications may occur with aid of one or more intermediary device or network. For instance, indirect communications may utilize a telecommunications network. Indirect communications may be performed with aid of one or more router, communication tower, satellite, or any other intermediary device or network. Examples of types of communications may include, but are not limited to: communications via the Internet, Local Area Networks (LANs), Wide Area Networks (WANs), Bluetooth® and Bluetooth® LE, Near Field Communication (NFC) technologies, networks based on mobile data protocols such as General Packet Radio Services (GPRS), GSM, Enhanced Data GSM Environment (EDGE), 3G, 4G, 5G or Long Term Evolution (LTE) protocols, Infra-Red (IR) communication technologies, and / or Wi-Fi, and may be wireless, wired, or a combination thereof. In some embodiments, the network may be implemented using cell and / or pager networks, satellite, licensed radio, or a combination of licensed and unlicensed radio. The network may be wireless, wired, or a combination thereof.

[0164] In some cases, the memory 203 may store data related to machine learning-based models. For example, the memory 203 may store data about a trained predictive model (e.g., parameters, hyper-parameters, model architecture, performance metrics, threshold, rules, etc.), data generated by a predictive model (e.g., intermediary results, output of a model, latent features, input and output of a component of the model system, etc.), training datasets (e.g., labeled data, insight provided by expert, user feedback data, etc.), predictive models, algorithms, and the like. The memory 203 can store algorithms or ruleset utilized by one or more methods disclosed herein.

[0165] In some cases, data stored in the memory 203 or external databases can be utilized or accessed by a variety of applications through application programming interfaces (APIs). Access to the database may be authorized at per API level, per data level (e.g., type of data), per application level or according to other authorization policies.

[0166] Although particular computing devices are illustrated and networks described, it is to be appreciated and understood that other computing devices and networks can be utilized. In addition, one or more components of the network layout may be interconnected in a variety of ways, and may in some embodiments be directly connected to, co-located with, or remote from one another, as one of ordinary skill will appreciate.

[0167] Logic may be encoded in a computer readable medium, which may refer to any medium that participates in providing instructions to the processor / microcontroller 201 for execution. Such a medium may take many forms, including but not limited to, non-volatile media, volatile media, and transmission media. In various implementations, non-volatile media includes optical or magnetic disks, volatile media includes dynamic memory, such as the memory 203, and transmission media includes coaxial cables, copper wire, and fiber optics, including wires. In one embodiment, the logic is encoded in non-transitory computer readable medium. In one example, transmission media may take the form of acoustic or light waves, such as those generated during radio wave, optical, and infrared data communications.

[0168] Some common forms of computer readable media includes, for example, floppy disk, flexible disk, hard disk, magnetic tape, any other magnetic medium, CD-ROM, any other optical medium, punch cards, paper tape, any other physical medium with patterns of holes, RAM, PROM, EPROM, FLASH-EPROM, any other memory chip or cartridge, or any other medium from which a computer is adapted to read.

[0169] In various embodiments of the present disclosure, execution of instruction sequences to practice the present disclosure may be performed by the processor / microcontroller 201.

[0170] Where applicable, various embodiments provided by the present disclosure may be implemented using hardware, software, or combinations of hardware and software. Also, where applicable, the various hardware components and / or software components set forth herein may be combined into composite components comprising software, hardware, and / or both without departing from the spirit of the present disclosure. Where applicable, the various hardware components and / or software components set forth herein may be separated into sub-components comprising software, hardware, or both without departing from the scope of the present disclosure. In addition, where applicable, it is contemplated that software components may be implemented as hardware components and vice-versa.

[0171] Software in accordance with the present disclosure, such as program code and / or data, may be stored on one or more computer readable mediums such as memory 203. It is also contemplated that software identified herein may be implemented using one or more general purpose or specific purpose computers and / or computer systems, networked and / or otherwise. Where applicable, the ordering of various steps described herein may be changed, combined into composite steps, and / or separated into sub-steps to provide features described herein.

[0172] The machine learning model may comprise a neural network. The neural network may comprise at least one of a deep residual network, attention-based networks, a highway network, a densely connected network, a recurrent network, and a capsule network, for example.

[0173] For any such type of network, the network may comprise a plurality of different neurons, which are organised into different layers. Each neuron (unit) is configured to receive input data, process this input data and provide output data. Each neuron may be configured to perform a specific operation on its input, e.g. this may involve mathematically processing the input data. The input data for each neuron may comprise an output from a plurality of other preceding neurons. As part of a neuron’s operation on input data, each stream of input data (e.g. one stream of input data for each preceding neuron which provides its output to the neuron) is assigned a weighting. That way, processing of input data by a neuron comprises applying weightings to the different streams of input data so that different items of input data will contribute more or less to the overall output of a neuron. Adjustments to the value of the inputs for a neuron, e.g. as a consequence of the input weightings changing, may result in a change to the value of the output for that neuron. The output data from each neuron may be sent to a plurality of subsequent neurons.

[0174] The neurons are organised in layers. Each layer comprises a plurality of neurons which operate on data provided to them from the output of neurons in preceding layers. Within each layer there may be a large number of different neurons, each of which applies a different weighting to its input data and performs a different operation on its input data. The input data for all of the neurons in a layer may be the same, and the output from the neurons will be passed to neurons in subsequent layers. The exact routing between neurons in different layers forms a major difference between capsule networks and deep residual networks (including variants such as highway networks and densely connected networks).

[0175] For a residual network, layers may be organised into blocks, such that the network comprises a plurality of blocks, each of which comprises at least one layer. For a residual network, output data from one layer of neurons may follow more than one different path. For conventional neural networks (e.g. convolutional neural networks), output data from one layer is passed into the next layer, and this continues until the end of the network so that each layer receives input from the layer immediately preceding it and provides output to the layer immediately after it. However, for a residual network, a different routing between layers may occur. For example, the output from one layer may be passed on to multiple different subsequent layers, and the input for one layer may be received from multiple different preceding layers.

[0176] In a residual network, layers of neurons may be organised into different blocks, wherein each block comprises at least one layer of neurons. Blocks may be arranged with layers stacked together so that the output of a preceding layer (or layers) feeds into the input of the next block of layers. The structure of the residual network may be such that the output from one block (or layer) is passed into both the block (or layer) immediately after it and at least one other later subsequent block (or layer). Shortcuts may be introduced into the neural network which pass data from one layer (or block) to another whilst bypassing other layers (or blocks) in between the two. This may enable more efficient training of the network, e.g. when dealing with very deep networks, as it may enable problems associated with degradation to be addressed when training the network (which is discussed in more detail below). The arrangement of a residual neural network may enable branches to occur such that the same input provided to one layer, or block of layers, is provided to at least one other layer, or block of layers (e.g. so that the other layer may operate on both the input data and the output data from the one layer, or block of layers). This arrangement may enable a deeper penetration into the network when using back propagation algorithms to train the network. For example, this is because during learning, layers, or blocks of layers, may be able to take as an input, the input of a previous layer / block and the output of the previous layer / block, and shortcuts may be used to provide deeper penetration when updating weightings for the network.

[0177] In a Recurrent Neural Network (RNN), recurrent connections are used to allow the networks to maintain an internal state or hidden state while processing each input element in a sequence. This hidden state serves as a memory of the previous inputs seen so far and influences the processing of subsequent inputs. RNNs capture the sequential nature of data by sharing weights across different time steps, enabling them to model dependencies and patterns over time, making them suitable for time series predictions. However, RNNs can suffer from the challenges of vanishing and exploding gradients, which can make it difficult for the network to learn long-term dependencies. To address this issue, multiple variations of RNNs have been introduced, such as long short-term memory (LSTM) and gated recurrent units (GRU). These variations use specialized gates to control the flow of information through the network, which helps to prevent the vanishing and exploding gradients problem.

[0178] In an attention-based network, an attention mechanism is used to allow the network to capture dependencies between different elements of the input sequence, enabling the model to understand long-term relationships in sequential data. Each attention layer in the transformer computes a weighted sum of the hidden states from the previous layer. These weights are determined by how much a hidden state attends to other hidden states in the previous layer, regardless of their position in the sequence. This is achieved by computing attention weights using a similarity function, such as dot product, cosine similarity, or scaled dot product, between pairs of hidden states known as keys and queries. The attention weights are then used to compute a weighted sum of the hidden states called values, such as by using a softmax function or a normalized linear combination. The resulting values are passed to the next layer for further processing. The attention mechanism can be used in a variety of architectures including the transformer, BERT, GPT or sequence-to-sequence models.

[0179] For a capsule network, layers may be nested inside of other layers to provide ‘capsules’. Different capsules may be adapted so that they are more proficient at performing different tasks than other capsules. A capsule network may provide dynamic routing between capsules so that for a given task, the task is allocated to the most competent capsule for processing that task. For example, a capsule network may avoid routing the output from every neuron in a layer to every neuron in the next layer. A lower level capsule is configured to send its input to a higher level (subsequent) capsule which is determined to be the most likely capsule to deal with that input. Capsules may predict the activity of higher layer capsules. For example, a capsule may output a vector, for which the orientation represents properties of an object in question. In response, each subsequent capsule may provide, as an output, a probability that the object that capsule is trained to identify is present in the input data. This information (e.g. the probabilities) can be fed back to the capsule, which can then dynamically determine routing weights, and forward the input data to the subsequent capsule most likely to be the relevant capsule for processing that data.

[0180] For either type of neural network, there may be included a plurality of different layers which have different functions. The neural network may include at least one convolutional layer configured to convolve input data across its height and width. The neural network may also have a plurality of filtering layers, each of which comprises a plurality of neurons configured to focus on and apply filters to different portions of the input data. Other layers may be included for processing the input data such as pooling layers (to introduce non-linearity) such as maximum pooling and global average pooling, Rectified Linear Units layer (ReLU) and its variants, for example, including Leaky ReLU, PReLU, ELU, SELU, GELU, and loss layers, e.g. some of which may include regularization functions. The final block of layers may receive input from the last output layer (or more layers if there are branches present). The final block may comprise at least one fully connected layer.

[0181] The final output layer may comprise a classifier, such as a softmax, sigmoid or tanh classifier; or a regression model, such as linear, exponential, gamma, Cox, Weibull, lognormal, log-logistic regression, for example. Different classifiers may be suitable for different types of output; for example, a sigmoid classifier may be suitable where the output is a binary classifier. The neural network of the present disclosure may be configured to evaluate the probability of a chronic obstructive pulmonary disease or asthma exacerbation occurring in a specified time window or the expected time to the next exacerbation. The post-processed output of the neural network may provide an indication of the risk of a chronic obstructive pulmonary disease exacerbation or asthma occurring. The various features and steps described herein may be implemented as systems comprising one or more memories storing various information described herein and one or more processors / microcontrollers coupled to the one or more memories and a network, wherein the one or more processors / microcontroller are operable to perform steps as described herein, as non-transitory machine-readable medium comprising a plurality of machine-readable instructions which, when executed by one or more processors / microcontrollers, are adapted to cause the one or more processors / microcontrollers to perform a method comprising steps described herein, and methods performed by one or more devices, such as a hardware processor / microcontroller, user device, server, and other devices described herein.

[0182] As used herein, “or” is inclusive and not exclusive, unless expressly indicated otherwise by context. Therefore, “A or B” means “A, B, or both,” unless expressly indicated otherwise or indicated otherwise by context. Moreover, “and” is both joint and several, unless expressly indicated otherwise or indicated otherwise by context.

[0183] While preferred embodiments of the present invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. It is not intended that the invention be limited by the specific examples provided within the specification. While the invention has been described with reference to the aforementioned specification, the descriptions and illustrations of the embodiments herein are not meant to be construed in a limiting sense. Numerous variations, changes, and substitutions will now occur to those skilled in the art without departing from the invention. Furthermore, it shall be understood that all aspects of the invention are not limited to the specific depictions, configurations or relative proportions set forth herein which depend upon a variety of conditions and variables. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention. It is therefore contemplated that the invention shall also cover any such alternatives, modifications, variations or equivalents. It is intended that the following claims define the scope of the invention and that methods and structures within the scope of these claims and their equivalents be covered thereby.

Claims

CLAIMS:

1. A detachable inhaler monitoring device, the monitoring device being configured to removably attach to an inhaler, the monitoring device comprising: a processor / microcontroller coupled to a memory; a plurality of sensors configured to monitor at least one of movement of the inhaler and / or parameters of a user using the inhaler and to provide sensor signals to the processor / microcontroller; a power source for powering the processor / microcontroller, the sensors, and memory; wherein the memory comprises program instructions, which when executed on the processor / microcontroller, cause the processor / microcontroller to run an algorithm, a set of rules, or an Al or machine learning program / model to analyse the sensor signals and provide a prediction of the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring based on the analysed sensor signals.

2. The detachable inhaler monitoring device of any of the previous claims, wherein the device further comprises at least one of a visual indicator and / or a vibration element coupled to the processor / microcontroller, and wherein the processor / microcontroller is configured to operate at least one of the visual indicator and / or vibration element to provide an indication of the risk of a chronic obstructive pulmonary disease exacerbation occurring.

3. The detachable inhaler monitoring device of claim 1 or 2 wherein the Al or machine learning program / model is configured to determine whether the inhaler has been used incorrectly by the user based on the sensor signals.

4. The detachable inhaler monitoring device of claim 3, wherein the device further comprises at least one of a visual indicator and / or a vibration element coupled to the processor / microcontroller, and wherein the processor / microcontroller is configured to operate at least one of the visual indicator and / or vibration element to provide an indication of incorrect use of the inhaler.

5. A detachable inhaler monitoring device, the monitoring device being configured toremovably attach to an inhaler, the monitoring device comprising: a processor / microcontroller coupled to a memory; a plurality of sensors configured to monitor at least one of movement of the inhaler and / or parameters of a user using the inhaler and to provide sensor signals to the processor / microcontroller; at least one of a visual indicator and / or a vibration element coupled to the processor / microcontroller; a power source for powering the processor, the sensors, the memory and the visual indicator and / or vibration element; wherein the memory comprises program instructions, which when executed on the processor / microcontroller, cause the processor / microcontroller to run an Al or machine learning program / model to analyse the sensor signals and operate at least one of the visual indicator and / or vibration element to provide an indication of whether the inhaler has been used incorrectly by the user based on the analysed sensor signals.

6. The detachable inhaler monitoring device of any of claims 3 to 5, wherein the sensors comprise a gyroscope, a microphone, and an airflow / pressure sensor, and wherein the Al or machine learning program / model is configured to determine whether the inhaler has been used incorrectly by the user based on the signals from the gyroscope, microphone and airflow / pressure sensor.

7. The detachable inhaler monitoring device of any of the previous claims, wherein the device comprises a communications interface, and wherein the processor / microcontroller is configured to receive an update to the Al or machine learning program / model via the communications interface.

8. The detachable inhaler monitoring device of any of the previous claims, wherein the processor / microcontroller is configured to send information indicative of the sensor signals to another device.

9. The detachable inhaler monitoring device of any of the previous claims, wherein the processor / microcontroller is configured to run an Al or machine learning program / model to analyse the sensor signals and provide a prediction of the risk of achronic obstructive pulmonary disease or asthma exacerbation based on the trajectory of the sensor signals over time.

10. The detachable inhaler monitoring device of any of the previous claims, wherein the sensors are configured to provide an indication of respiratory flow / pressure, heart rate, oxygen saturation, and body temperature to the processor / microcontroller.

11. The detachable inhaler monitoring device of claim 10, wherein the device comprises a reflective sensor for monitoring heart rate and / or oxygen saturation of the user.

12. The detachable inhaler monitoring device of claim 10 or 11 , wherein the device is configured to obtain an indication of heart rate, oxygen saturation, and body temperature from a user holding the inhaler.

13. The detachable inhaler monitoring device of any of the previous claims wherein the device further comprises an accelerometer, and wherein the device is configured to operate in a sleep mode and an active mode, and wherein movement of the device as indicated by the accelerometer is configured to trigger the device to switch between the sleep mode and the active mode.

14. The detachable inhaler monitoring device of claim 13, wherein in the active mode the processor / microcontroller is configured to receive sensor signals from the plurality of sensors more frequently than in the sleep mode.

15. The detachable inhaler monitoring device of claim 1 or any claim as dependent thereon, further comprising a clock module for precise time measurement, wherein the clock module is configured to record the time duration for each instance of inhaler usage, and wherein the processor / microcontroller is configured to use the time measurement in providing a prediction of the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring.

16. The detachable inhaler monitoring device of claim 5 or any claim as dependent thereon, further comprising a clock module for precise time measurement, wherein the clock module is configured to record the time duration for each instance of inhaler usage, and wherein the processor / microcontroller is configured to use the time measurement in providing an indication of whether the inhaler has been used incorrectly by the user.

17. A detachable inhaler monitoring device, the monitoring device being configured to removably attach to an inhaler, the monitoring device comprising: a processor / microcontroller coupled to a memory; a communications interface; at least one of a visual indicator and / or a vibration element coupled to the processor / microcontroller; a plurality of sensors configured to monitor at least one of movement of the inhaler and / or parameters of a user using the inhaler and to provide sensor signals to the processor / microcontroller; a power source for powering the processor / microcontroller, memory, sensors, visual indicator and / or vibration element and communications interface; wherein the memory comprises program instructions, which when executed on the processor / microcontroller, cause the processor / microcontroller to receive sensor signals from the plurality of sensors, send the sensor signals to another device via the communications interface, and to operate at least one of the visual indicator and / or vibration element in response to receiving signals from the other device indicating that (i) the inhaler has been used incorrectly, and / or (ii) there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring.

18. The detachable inhaler monitoring device of claim 17 wherein the processor / microcontroller is configured to process the sensor signals by filtering the sensor signals prior to sending to the other device via the communications interface.

19. The detachable inhaler monitoring device of claim 17 or 18, wherein the device further comprises an accelerometer, and wherein the device is configured to operate in a sleep mode and an active mode, and wherein movement of the device as indicated by the accelerometer is configured to trigger the device to switch between the sleep mode andthe active mode.

20. The detachable inhaler monitoring device of claim 19, wherein in the active mode the processor / microcontroller is configured to receive sensor signals from the plurality of sensors more frequently than in the sleep mode.

21. The detachable inhaler monitoring device of any of claims 17 to 20, wherein the sensors are configured to provide an indication of respiratory flow, heart rate, oxygen saturation, and body temperature to the processor / microcontroller.

22. The detachable inhaler monitoring device of claim 21 , wherein the device comprises a reflective sensor for monitoring heart rate and / or oxygen saturation of the user.

23. The detachable inhaler monitoring device of claim 21 or 22, wherein the device is configured to obtain an indication of heart rate, oxygen saturation, inhalation duration, and body temperature from a user holding the inhaler.

24. The detachable inhaler monitoring device of any of the previous claims wherein the device is configured to fit over the mouthpiece of a conventional inhaler.

25. The detachable inhaler monitoring device of any of the previous claims wherein the device is configured to be mounted over the canister of a pressurised metered dose inhaler, pMDI.

26. The detachable inhaler monitoring device of any of the previous claims wherein the device is configured to be mounted between the canister and the housing of a pressurised metered dose inhaler, pMDI.

27. The detachable inhaler monitoring device of any of the previous claims wherein the device is configured to be mounted to a dry powder inhaler, DPI.

28. A method of providing an indication of incorrect inhaler use, the method comprising:obtaining sensor signals from an inhaler device; processing the sensor signals using an Al or machine learning model; determining that the inhaler has been used incorrectly based on the processing by the Al or machine learning model; and instructing at least one of a visual indicator and a vibration element on the inhaler to operate to provide feedback to the user that the inhaler was used incorrectly.

29. A computer-implemented method of predicting the risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring, the method comprising: obtaining sensor signals from an inhaler device; processing the sensor signals using an Al or machine learning model; determining that there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring based on the processing by the Al or machine learning model; and instructing at least one of a visual indicator and a vibration element on the inhaler to operate to provide feedback to the user that there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring.

30. The computer-implemented method of claim 29 further comprising determining the degree of risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring based on the processing by the Al or machine learning model; and wherein instructing at least one of a visual indicator and a vibration element on the inhaler to operate to provide feedback to the user that there is a risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring comprises instructing the visual indicator and / or vibration element to operate based on the degree of risk of a chronic obstructive pulmonary disease or asthma exacerbation occurring, such that a different visual indication or vibration is provided to the user for a greater determined risk compared to a lower determined risk.

31. A computer readable non-transitory storage medium comprising a program for a computer configured to cause a processor / microcontroller to perform the method of any of claims 28 to 30.

32. A computer-implemented method for use in providing a personalised selfmanagement plan to an inhaler device user, the method comprising: identifying, via a machine learning model on the edge on the inhaler device, an elevated risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation for the inhaler device user; obtaining, via a data extraction module, data on population, intervention, comparison, and outcome (PICO) from repositories of clinical trials on COPD and asthma exacerbation; converting, via a data representation module, the obtained PICO data into structured data, wherein the structured data comprises continuous vector embeddings, and wherein the data is structured using established classifications of self-management interventions and population characteristics; training, via a model training module, a meta regression machine learning model to predict the outcomes of the clinical trials based on the structured data, and to predict the required intervention to bring about a wanted outcome for a given inhaler device user based on the characteristics of the inhaler device user; determining, via an intervention derivation module, the intervention required for the inhaler device user with a risk of COPD and asthma exacerbation based on the characteristics of the inhaler device user and the meta regression machine learning model; obtaining, via a plan generation module, a self-management plan based on the determined intervention required; and outputting, via a communication module, the self-management plan to a mobile device associated with the inhaler device user.

33. The computer-implemented method of claim 32 wherein the characteristics of the inhaler device user comprise historical trajectories of heart rate, oxygen saturation, body temperature, airflow, and predictions of exacerbation risks by the machine learning model.

34. The computer-implemented method of claim 32 wherein the step of obtaining the PICO data comprises utilising natural language processing techniques, and wherein the natural language processing techniques used to obtain the PICO data comprise at least one of a large language model, named entity recognition, semantic role labelling,contextual embeddings, summarisation algorithms, and / or regular expression-based extraction and parsing.

35. The computer-implemented method of claim 32 wherein the population characteristics used to convert the obtained PICO data into structured data comprises demographic data, medical history, and biomarker levels, and the self-management intervention characteristics used to convert the obtained PICO data into structured data comprise physical activity, lifestyle, inhaler technique, medication adherence, symptom monitoring, disease education and supervision levels, and components of pulmonary rehabilitation.

36. The computer-implemented method of claims 32 wherein the step of determining the intervention required for an inhaler device user comprises using Bayesian inference.

37. The computer-implemented method of claim 32 wherein the characteristics of the inhaler device user are obtained from metrics received by the inhaler device and / or other connected or wearable devices used by the inhaler device user.

38. The computer-implemented method of claim 32 wherein the characteristics of the inhaler device user update via real-time data streams, and wherein the obtaining and outputting of the self-management plan updates in real time.

39. A cloud-based device configured to provide a personalised self-management plan to an inhaler device user, the cloud-based device comprising: a machine learning model configured to identify an elevated risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation of the inhaler device user; a data extraction module configured to obtain data on population, intervention, comparison, and outcome (PICO) from repositories of clinical trials on COPD and asthma exacerbation; a data representation module configured to convert the obtained PICO data into structured data, wherein the structured data comprises continuous vector embeddings, and wherein the data is structured using established classifications of self-management interventions and population characteristics;a model training module configured to train a meta regression machine learning model to predict the outcomes of the clinical trials based on the structured data, and to predict the required intervention to bring about a wanted outcome for a given inhaler device user based on the characteristics of the inhaler device user; an intervention derivation module configured to determine the intervention required for the inhaler device user with a risk of COPD and asthma exacerbation based on the characteristics of the inhaler device user and the meta regression machine learning model; a plan generation module configured to obtain a self-management plan based on the determined intervention required; and a communication module configured to output the self-management plan to a mobile device associated with the inhaler device user.

40. The cloud-based device of claim 39 wherein the characteristics of the inhaler device user comprise historical trajectories of heart rate, oxygen saturation, body temperature, airflow, and predictions of exacerbation risks by the machine learning model.

41. The cloud-based device of claim 39 wherein the data extraction module utilises natural language processing techniques to obtain obtaining the PICO data, and wherein the natural language processing techniques used to obtain the PICO data comprise at least one of a large language model, named entity recognition, semantic role labelling, contextual embeddings, summarisation algorithms, and / or regular expression-based extraction and parsing.

42. The cloud-based device of claim 39 wherein the population characteristics used to convert the obtained PICO data into structured data comprises demographic data, medical history, and biomarker levels, and the self-management intervention characteristics used to convert the obtained PICO data into structured data comprise physical activity, lifestyle, inhaler technique, medication adherence, symptom monitoring, disease education and supervision levels, and components of pulmonary rehabilitation.

43. The cloud-based device of claims 39 wherein determining the intervention required for the inhaler device user comprises using Bayesian inference.

44. The cloud-based device of claim 39 wherein the characteristics of the inhaler device user are obtained from metrics received by the inhaler device and / or other connected or wearable devices used by the inhaler device user.

45. The cloud-based device of claim 39 wherein the characteristics of the inhaler device user update via real-time data streams, and wherein the obtaining and outputting of the self-management plan updates in real time.

46. A method of using an inhaler device for reducing risk of chronic obstructive pulmonary disease (COPD) or asthma exacerbation, the method comprising: administering medication via the inhaler device on an as-needed basis; obtaining, via a visual indicator and / or vibration element, feedback on the inhaler device indicating a risk of COPD or asthma exacerbation and the need for a selfmanagement plan; and obtaining, via a mobile device, a self-management plan associated with the inhaler device user.

47. The method of claim 46 wherein the inhaler device obtains sensor signals when the smart inhaler device is used, wherein the sensor signals correspond to real-time data on the characteristics of the inhaler device user.

Images