Wearable spirometry utilizing a non-invasive motion sensor for measuring a volume of inspiration

Abstract

In some instances, a method and system for performing spirometry to diagnose one or more medical conditions based on using one or more machine learning – artificial intelligence (ML – AI) models is provided. For instance, the method comprises obtaining, from a plurality of wearable devices, training sensor data associated with a plurality of individuals, wherein the plurality of wearable devices are adhered to the plurality of individuals and configured to detect chest wall expansions of the plurality of individuals; training the one or more ML – AI models using the training sensor data; subsequent to training the one or more ML – AI models, obtaining new sensor data associated with an individual; and processing the new sensor data using the one or more ML – AI models to detect whether the individual has the one or more medical conditions.

Description

WEARABLE SPIROMETRY UTILIZING A NON-INVASIVE MOTION SENSOR FOR MEASURING A VOLUME OF INSPIRATIONCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 710,394, filed October 22, 2024, which is herein incorporated by reference in its entirety.STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] This invention was made with Government support under project numbers 1ZID BC011242 and Z1A CL040015 by the National Institutes of Health. The Government has certain rights in the invention.FIELD

[0003] This application is related to the diagnosing medical conditions such as postoperative pulmonary complications based on performing spirometry to measure a volume of inspiration in postoperative pulmonary patients using utilizing non-invasive, wearable motion sensors and machine learning - artificial intelligence (ML - Al) algorithms and / or models.BACKGROUND OF THE INVENTION

[0004] PPCs (postoperative pulmonary complications and / or potentially preventable complications) encompass a broad and heterogeneous spectrum of conditions that serve as a leading cause of suboptimal surgical outcomes globally. Approximately 10-20% of the 310 million patients undergoing surgery each year will develop a PPC, and up to 25% of all deaths occurring within one week post-operatively attributed to PPCs. These pulmonary complications, which range from minor respiratory alterations to severe conditions such as pneumonia or respiratory failure, are a culmination of changes to the respiratory system that begin immediately after induction of general anesthesia.

[0005] Incentive spirometry (IS) has traditionally served as a cornerstone in the prophylaxis and management of PPCs. The incentive spirometer works by emulating a natural deep breath,serving simply as a visual tool for both patients and providers to gauge the quality and depth of lung expansion. By promoting voluntary deep breathing, the device helps to improve pulmonary function, maintain alveolar inflation, and enhance the clearance of secretions.

[0006] However, despite its widespread acceptance in clinical practice, it has been well known that there may be various limitations associated with conventional incentive spirometry. Some of these drawbacks include patient compliance, nursing burden, and the lack of data storage for longitudinal tracking. Without a means to quantitatively assess patient effort and progress, it becomes difficult to assess the true impact of incentive spirometry on patient outcomes. But, the continual advancements of digital health technologies and artificial intelligence present a compelling opportunity to address these gaps.SUMMARY

[0007] In an exemplary embodiment, a system for performing spirometry to diagnose one or more medical conditions based on using one or more machine learning - artificial intelligence (ML - Al) models is provided. The system comprises a first wearable device configured to: obtain motion data associated with a first individual, wherein the first wearable device is adhered to the first individual and configured to detect chest wall expansions of the first individual; and transmit the motion data to a server. The system further comprises a server configured to: receive, from a plurality of wearable devices that includes the first wearable device, training sensor data associated with a plurality of individuals, wherein the training sensor data comprises the motion data associated with the first individual; train the one or more ML - Al models using the training sensor data; subsequent to training the one or more ML - Al models, obtain new sensor data associated with a second individual; and process the new sensor data using the one or more ML - Al models to detect whether the second individual has the one or more medical conditions.

[0008] In some instances, the first wearable device is a motion sensor, wherein the motion sensor is adhered to a level of a 9thand 10thrib at a mid-axillary line of the first individual.

[0009] In some examples, training the one or more ML - Al models comprises: categorizing the training sensor data into a plurality of different lung volume categories, wherein the plurality of different lung volume categories comprises a first lung volume category, a second lung volumecategory, and a third lung volume category; and training the one or more ML - Al models based on categorizing the training sensor data into the plurality of different lung volume categories.

[0010] In some variations, training the one or more ML - Al models further comprises: prior to categorizing the training sensor data into the plurality of different lung volume categories, preprocessing the training sensor data to remove corrupted waveforms from the training sensor data and / or to remove motion artifacts from the training sensor data.

[0011] In some instances, categorizing the training sensor data into the plurality of different lung volume categories comprises sorting the training sensor data into a first set of training data associated with the first lung volume category, a second set of training data associated with the second lung volume category, and a third set of training data associated with the third lung volume category, and wherein training the one or more ML - Al models comprises: training a first ML -Al model, from the one or more ML - Al models, using the first set of training data; training a second ML - Al model, from the one or more ML - Al models, using the second set of training data; and training a third ML - Al model, from the one or more ML - Al models, using the third set of training data.

[0012] In some examples, the training sensor data indicates motion data in three axes, wherein movement in an x-axis indicates chest wall movement in a cranial -caudal axis, wherein movement in a y-axis indicates movement that is perpendicular to a chest wall, and wherein movement in a z-axis indicates movement in an upwards and outwards direction from the plurality of individuals.

[0013] In some variations, training the one or more ML - Al models is based on using a leave-one-out validation strategy that splits the training sensor data into a training set and a testing set, wherein in each training epoch, a different portion of the training sensor data from a different individual of the plurality of individuals is included within the testing set and remaining training sensor data is included within the training set.

[0014] In some instances, processing the new sensor data using the one or more ML - Al models to detect whether the second individual has the one or more medical conditions comprises assessing whether the second individual has postoperative pulmonary complications (PPCs).

[0015] In another exemplary embodiment, a method for performing spirometry to diagnose one or more medical conditions based on using one or more machine learning - artificial intelligence (ML - Al) models is provided. The method comprises obtaining, from a plurality ofwearable devices, training sensor data associated with a plurality of individuals, wherein the plurality of wearable devices are adhered to the plurality of individuals and configured to detect chest wall expansions of the plurality of individuals; training the one or more ML - Al models using the training sensor data; subsequent to training the one or more ML - Al models, obtaining new sensor data associated with an individual; and processing the new sensor data using the one or more ML - Al models to detect whether the individual has the one or more medical conditions.

[0016] In some instances, the plurality of wearable devices comprises one or more motion sensors, wherein a first motion sensor, of the one or more motion sensors, is adhered to a level of a 9thand 10thrib at a mid-axillary line of a first individual of the plurality of individuals.

[0017] In some examples, obtaining the training sensor data comprises obtaining, by a secure server, the training sensor data from the plurality of wearable devices.

[0018] In some variations, training the one or more ML - Al models comprises: categorizing the training sensor data into a plurality of different lung volume categories, wherein the plurality of different lung volume categories comprises a first lung volume category, a second lung volume category, and a third lung volume category; and training the one or more ML - Al models based on categorizing the training sensor data into the plurality of different lung volume categories.

[0019] In some instances, training the one or more ML - Al models further comprises: prior to categorizing the training sensor data into the plurality of different lung volume categories, preprocessing the training sensor data to remove corrupted waveforms from the training sensor data and / or to remove motion artifacts from the training sensor data.

[0020] In some examples, categorizing the training sensor data into the plurality of different lung volume categories comprises sorting the training sensor data into a first set of training data associated with the first lung volume category, a second set of training data associated with the second lung volume category, and a third set of training data associated with the third lung volume category, and wherein training the one or more ML - Al models comprises: training a first ML - Al model, from the one or more ML - Al models, using the first set of training data; training a second ML - Al model, from the one or more ML - Al models, using the second set of training data; and training a third ML - Al model, from the one or more ML - Al models, using the third set of training data.

[0021] In some variations, the training sensor data indicates motion data in three axes, wherein movement in an x-axis indicates chest wall movement in a cranial -caudal axis, wherein movement in a y-axis indicates movement that is perpendicular to a chest wall, and wherein movement in a z-axis indicates movement in an upwards and outwards direction from the plurality of individuals.

[0022] In some instances, training the one or more ML - Al models is based on using a leave-one-out validation strategy that splits the training sensor data into a training set and a testing set, wherein in each training epoch, a different portion of the training sensor data from a different individual of the plurality of individuals is included within the testing set and remaining training sensor data is included within the training set.

[0023] In some examples, processing the new sensor data using the one or more ML - Al models to detect whether the individual has the one or more medical conditions comprises assessing whether the individual has postoperative pulmonary complications (PPCs).

[0024] In some variations, the one or more ML - Al models comprises a first ML - Al model, a second ML - Al model, and a third ML - Al model, and wherein processing the new sensor data using the one or more ML - Al models further comprises: sorting the new sensor data into a first set of sensor data associated with a first lung volume category, a second set of sensor data associated with a second lung volume category, and a third set of sensor data associated with a third lung volume category; processing the first set of sensor data using the first ML - Al model to generate a first ML - Al output; processing the second set of sensor data using the second ML - Al model to generate a second ML - Al output; and processing the third set of sensor data using the third ML - Al model to generate a third ML - Al output, wherein assessing whether the individual has PPCs is based on the first ML - Al output, the second ML - Al output, and the third ML - Al output.

[0025] In some instances, obtaining the new sensor data associated with the individual is in response to a user device providing a notification to the individual, wherein the notification comprises an automatic reminder to the individual to complete breathing exercises.

[0026] In yet another exemplary embodiment, a non-transitory computer-readable medium having processor-executable instructions stored thereon for performing spirometry to diagnose one or more medical conditions based on using one or more machine learning - artificial intelligence (ML - Al) models is provided. The processor-executable instructions, when executed, facilitate:obtaining, from a plurality of wearable devices, training sensor data associated with a plurality of individuals, wherein the plurality of wearable devices are adhered to the plurality of individuals and configured to detect chest wall expansions of the plurality of individuals; training the one or more ML - Al models using the training sensor data; subsequent to training the one or more ML - Al models, obtaining new sensor data associated with an individual; and processing the new sensor data using the one or more ML - Al models to detect whether the individual has the one or more medical conditions.

[0027] All examples and features mentioned above may be combined in any technically possible way.BRIEF DESCRIPTION OF DRAWINGS

[0028] The present application will be described in even greater detail below based on the exemplary figures. The application is not limited to the examples described below. All features described and / or illustrated herein can be used alone or combined in different combinations in examples of the application. The features and advantages of various examples of the present application will become apparent by reading the following detailed description with reference to the attached drawings which illustrate the following:

[0029] FIG. 1A shows an environment for obtaining raw wearable waveform data according to one or more exemplary embodiments of the present application.

[0030] FIGs. IB and 1C show representative plots of the raw wearable waveform data by volume of inspiration and axis labeling according to one or more exemplary embodiments of the present application.

[0031] FIG. 2 is a simplified block diagram depicting an exemplary computing environment in accordance with one or more exemplary embodiments of the present application.

[0032] FIG. 3 is a simplified block diagram of one or more devices or systems within the exemplary environment of FIG. 2.

[0033] FIG. 4A shows an exemplary motion sensor and user device in accordance with one or more exemplary embodiments of the present application.

[0034] FIG. 4B shows an exemplary graphical user interface (GUI) in accordance with one or more exemplary embodiments of the present application.

[0035] FIG. 5 is an exemplary process for performing spirometry to measure a volume of inspiration in postoperative pulmonary patients in accordance with one or more exemplary embodiments of the present application.DETAILED DESCRIPTION

[0036] Examples of the presented application will now be described more fully hereinafter with reference to the accompanying exemplary figures, in which some, but not all, examples of the application are shown. Indeed, the application may be embodied in any different forms and should not be construed as limited to the examples set forth herein; rather, these examples are provided so that the disclosure will satisfy applicable legal requirements. Where possible, any terms expressed in the singular form herein are meant to also include the plural form and vice versa, unless explicitly stated otherwise. Also, as used herein, the term “a” and / or “an” shall mean “one or more” even though the phrase “one or more” is also used herein. Furthermore, when it is said herein that something is “based on” something else, it may be based on one or more other things as well. In other words, unless expressly indicated otherwise, as used herein “based on” means “based at least in part on” or “based at least partially on”.

[0037] In some embodiments, systems and methods are described that perform incentive spirometry (IS) to measure volume of inspiration in postoperative pulmonary patients based on utilizing non-invasive, wearable motion sensors and machine learning - artificial intelligence (ML - Al) algorithms. For instance, IS has served as a cornerstone in the prophylaxis of postoperative pulmonary complications; however, IS uses may be limited by poor compliance, nursing burden, and lack of longitudinal tracking. As will be described below, examples of the present disclosure may adhere a wearable device (e.g., amotion sensor) to the right thorax of individuals (e.g., healthy volunteers) while taking measured breaths (e.g., 250-2500 milliliters (mL)) through IS. Examples of the present disclosure may train a model (e.g., an ML - Al model) on these waveforms to learn to recognize patterns in chest wall motion associated with low, medium, and high volumes. A leave-one-out validation strategy may be used to assess performance.

[0038] The results are now described. For instance, all subjects (e.g., total 6 subjects with 3 males and 3 females) were healthy volunteers. The average area under the curve (AUC) across allparticipants was 0.80 (range 0.68-0.92). The model performed best at identifying low-volume (AUC 0.85) and high-volume (AUC 0.80) breaths.

[0039] As will be described in further detail below, wearable technology may determine volume of inspiration from chest wall motion alone. In some examples, these models (e.g., ML -Al models) may be applied to improve compliance and even clinical outcomes by offering more standardized monitoring, documentation, and / or analysis.

[0040] PPCs (e.g., postoperative pulmonary complications and / or potentially preventable complications) encompass a broad and heterogeneous spectrum of conditions which serve as a leading cause of suboptimal surgical outcomes globally. Approximately 10-20% of the 310 million patients undergoing surgery each year may develop a PPC, and up to 25% of all deaths occurring within one week post-operatively may be attributed to PPCs. These pulmonary complications, which range from minor respiratory alterations to severe conditions such as pneumonia or respiratory failure, are a culmination of changes to the respiratory system that begin immediately after induction of general anesthesia. While under general anesthesia, patients experience decreased respiratory drive, reduced lung volumes, and altered ventilation perfusion relationship. The addition of a neuromuscular blocking drug for general anesthesia further increases risk for PPC with up to 75% of patients developing some degree of atelectasis post-operatively. The pain and fatigue many patients encounter post-operatively amplify the risk of PPCs, as patients become more hesitant to take deep, lung-expanding breaths. Considering these elements, it stands to reason that PPCs substantially influence prolonged hospital stays, elevated healthcare expenditures, and a deterioration in holistic patient outcomes.

[0041] IS has traditionally served as a cornerstone in the prophylaxis and management of PPCs. The incentive spirometer operates by emulating a natural deep breath, serving simply as a visual tool for both patients and providers to gauge the quality and depth of lung expansion. The device does not offer any added resistance to assist in alveoli opening. By promoting voluntary deep breathing, the device helps to improve pulmonary function, maintain alveolar inflation, and enhance the clearance of secretions. Despite its widespread acceptance in clinical practice, recent literature has shed light on various limitations associated with conventional incentive spirometry. Some of these drawbacks include patient compliance, nursing burden, and the lack of data storagefor longitudinal tracking. Without a means to quantitatively assess patient effort and progress, it becomes difficult to assess the true impact of incentive spirometry on patient outcomes.

[0042] The continual advancements of digital health technologies and artificial intelligence present a compelling opportunity to address these gaps. Previous efforts have been made to correlate chest wall motion with lung function, particularly in the asthma and apnea monitoring spaces. For instance, one effort was successful in measuring respiration rate and lung volume using wearable strain sensors on a small cohort of volunteers. Others have detected respiratory rate via acoustic adhesive sensors on the neck or detected changes in respiratory effort during sleep via an elastic band around the thorax and corresponding sensor measuring stretch patterns.

[0043] As such, examples of the present disclosure explored the possibility that unique patterns in chest wall motion may be identified via motion sensors adhered to the patient’s thorax to easily measure and track volume of inspiration and respiratory effort post-operatively. The technologies (e.g., systems and methods) of the present disclosure may open a door to improved IS compliance and may result in decreased nursing burden and improved post-operative management. For instance, as will be described below, a wearable motion sensor may serve as a substitute for incentive spirometry with features to improve patient compliance, facilitate longitudinal IS data collection, and reduce burden on nursing staff who continuously have to provide incentive spirometers and remind patients to use them.

[0044] In some examples, a motion sensor may be adhered to the right thorax of individuals (e g., healthy volunteers) using hypoallergenic medical tape at approximately the level of the 9thand 10thribs at the mid-axillary line. While seated upright with the motion sensor in place, participants may take three breaths through a traditional incentive spirometer at each of the following volumes: 250 milliliters (mL), 500 mb, 750 mL, 1000 mL, 1250 mL, 1500 mL, 1750 mL, 2000 mL, 2250 mL, and 2500 mL. Throughout this period, the wearable device detects chest wall expansion and transmits the data to a secure server (e.g., via a wireless protocol such as BLUETOOTH) for analysis and model creation.

[0045] Motion data may be quality checked, and corrupted waveforms may be removed. Waveforms corresponding to activity (motion artifacts) may also be removed, per standard signal processing techniques. The waveforms may then be parsed out by corresponding volume of inspiration and grouped into low-volume (e.g., 250 mL, 500 mL, 750 mL), medium-volume (e.g.,1000 mL, 1250 mL, 1500mL), and high-volume (e.g., 2000 mL, 2250 mL, 2500 mL) inspiration. A machine learning model may be trained on these sequences, learning to recognize patterns in chest wall motion associated with each incentive lung volume category (e.g., low, medium, or high). A leave-one-out validation strategy may be used to assess the model performance.

[0046] The results are described below. For instance, all subjects (e.g., n=6 subjects), with three males and three females, included in a study (e.g., an initial pilot study) were healthy volunteers with no known respiratory conditions. The x-axis was defined as chest wall movement in the cranial-caudal axis (e.g., parallel to the thorax). Movement in the y-axis corresponded to movement perpendicular to the chest wall, outward motion, and z-axis signal corresponded to movement in the upwards and outwards, or diagonal, direction. There were visible differences in waveform amplitude in the y and z axes as participants took higher-volume breaths. This is shown in FIGs 1A-1C.

[0047] For instance, FIG. 1 A shows an environment 10 for obtaining raw wearable waveform data according to one or more exemplary embodiments of the present application. For example, the environment 10 shows a user 15 (e.g., a patient, subject, or participant) and a sensor 20. The sensor 20 may be a motion sensor as mentioned above. However, in other embodiments, the sensor 20 may be any other type of sensor that is configured to obtain sensor data associated with the user 15 (e.g., volume inspiration of the user 15). The sensor 20 is shown to be positioned on a right thorax of the user 15 (e.g., at approximately the level of the 9thand 10thribs at the mid-axillary line). However, in other embodiments, the sensor 20 may be positioned at another location or position on the user 15. In addition, the x-axis 25, the y-axis 30, and the z-axis 35 are also shown. The movement in the x-axis 25 may be defined as chest wall movement in the cranial-caudal axis, the movement in the y-axis 30 may be defined as movement perpendicular to the chest wall (e.g., outward motion), and the movement in the z-axis 35 may be defined as movement in the upwards and outwards, or diagonal, direction.

[0048] FIGs. IB and 1C show representative plots 50 and 60 of the raw wearable waveform data by volume of inspiration and axis labeling according to one or more exemplary embodiments of the present application. For instance, referring to FIGs. IB and 1C, the raw wearable waveform data may be obtained by the sensor 20 for all three axes 25-35. The representative plot 50 may show the raw wearable waveform data from the y-axis 30 and the representative plot 60 may showthe raw wearable waveform data from the z-axis 35. For instance, the representative plots 50 and 60 may show on the y-axis the measurement values or amplitudes (e.g., relative distances). Furthermore, the representative plot 50 shows the waveforms for all of the breaths of the user 15 such as at 250 mL, 500 mb, 750 mL, 1000 mL, 1250 mL, 1500 mL, 1750 mL, 2000 mL, 2250 mL, and 2500 mL. The visible differences in waveform amplitude in the y-axis 30 and z-axis 35 are clearly shown in the representative plots 50 and 60.

[0049] In addition, as mentioned previously, when the model (e.g., the ML - Al model) was tested against traditional incentive spirometry volumes of inspiration, the average AUC across all 6 participants was 0.80 (range 0.68-0.92). This is shown in Table 1 below. The model performed best at identifying low-volume breaths (average AUC 0.85) and high-volume breaths (average AUC 0.80).

[0050] Table 1. Model PerformanceAUC Low AUC Med. AUC High Average Participant Volume Volume Volume AUCInspiration Inspiration Inspiration Score1 0.87 0.75 0.87 0.832 0.99 0.93 0.85 0.923 0.84 0.74 0.76 0.784 0.75 0.87 0.83 0.825 0.84 0.48 0.68 0.666 0.83 0.63 0.81 0.76Average 0.85 0.73 0.80 0.80

[0051] Despite the astonishing prevalence of PPCs, traditional incentive spirometry is highly underutilized, misused, poorly documented, and untracked. The burden of reminders and documentation associated with ideal incentive spirometry use often falls to overburdened nursing staff, leading to inconsistent application. As artificial intelligence (Al) and wearables continue to augment and streamline various facets of healthcare from diagnosis to patient management, such an approach described by examples of the present disclosure presents a promising opportunity to enhance the use and compliance of incentive spirometry.

[0052] Examples of the present disclosure demonstrate the potential of wearable motion sensors to assess volume of inspiration under sedentary conditions. These findings suggest that movement of the thorax and abdomen during inspiration may predict or correlate with the volume of air inhaled, which shows a promising application for large language models in this context. As such, examples of the present disclosure highlight an opportunity to leverage simple and cost-effective wearable technology with advanced computational models, to enhance the imperfect practice of incentive spirometry.

[0053] In some instances, the wearable technology described by embodiments of the present disclosure offers the potential to facilitate the postoperative recovery period from the hospital and even from home. In some examples, a device may automatically remind patients to complete breathing exercises via timed vibrations, interface with an application, or even interactive patient gaming strategies. Additionally, it may offer the ability for providers to interact with, document, and analyze longitudinal spirometry data in a low-cost way even outside the bounds of the hospital.

[0054] Studies were performed and demonstrated a unique opportunity to leverage Al to improve post-operative monitoring and characterize its potential to reduce the incidence PPCs. Ultimately, these results suggest that wearable technology and corresponding custom models are capable of determining volume of inspiration from chest wall motion alone and / or may be supplemented with further sensor data. As such, this technology may be used to improve compliance and potentially even clinical outcomes by offering more standardized and consistent monitoring, documentation, and data analysis.

[0055] In other words, examples of the present disclosure may relate to a system and method for measuring volume of inspiration in postoperative pulmonary patients. The system may include a sensor (e.g., a motion sensor) that is adhered to an individual’s right thorax, an ML - Al model and / or algorithm, and / or a computing device (e.g., a user device such as a smartphone). The motion sensor may be adhered to the right thorax of a patient to collect data on chest wall motion. The motion data is then transmitted to a secure server for analysis and model creation. A trained machine learning model analyzes the data to recognize patterns in the motion data associated with each incentive lung volume category (e.g., low, medium, or high).

[0056] For instance, examples of the present disclosure may describe methods and systems for performing wearable spirometry utilizing a non-invasive motion sensor. For instance, PPCs mayarise in 10-20% of the over 300 million patients undergoing surgery annually and may contribute to 25% of mortality occurring within one week of surgery. Measuring breathing dynamics such as volume of inspiration and respiratory effort may be done with an invective spirometer, but its use is limited by poor patient compliance and is perceived as a burden for nurses. Examples of the present disclosure have discovered that it is possible to correlate chest wall motion to volume of inspiration using a basic motion sensor. This technique has the potential to quantify breathing dynamics with high patient compliance without burdening nurses and to enable longitudinal data collection.

[0057] For instance, examples of the present disclosure may have the potential to reduce PPCs by increasing patient compliance with spirometry exercises. It may also serve to reduce the existing nursing burden of recording volumes of inspiration or reminding patients to use the incentive spirometer, which often is forgotten. Further, enabling longitudinal data collection may facilitate the study of impact of post-operative pulmonary function and compliance with breathing exercises on patient outcomes. In some instances, additional aspects may be performed such as strengthening the algorithm with more data points, developing the application, and / or designing a custom sensor configured for this specific use.

[0058] In some examples, spirometry may reduce PPCs and / or breathing exercise games for mindfulness exercises. Spirometry may be performed using a plurality of different devices including the devices described below. For example, in some instances, invective spirometer may be performed by a plastic non-digital device that measures volume of inspiration as an individual takes a deep breath. In some variations, a CAPMEDIC Smart Incentive Spirometer may be utilized, which may include and / or use a handheld home mouthpiece device that measures pulmonary function and transmits data to an application. This may be intended for use in patients with asthma or chronic obstructive pulmonary disease (COPD) who use an inhaler. In some examples, SPIROLINK smart spirometer at home may be utilized, which may include a handheld digital spirometer measures peak expiratory flow (PEF), forced expiratory volume in one second (FEV1), and more. This device might not be intended to measure volume of inspiration or remind patients to complete breathing exercises. In some instances, the MIR Smart ONE personal pocket spirometer may be used, which may include an expiratory spirometry handheld BLUETOOTH device that conveys PEF, FEV1, and / or other breathing metrics. In some examples, LePulse maybe used, which may include a BLUETOOTH handheld breathing exercise device for respiratory muscle training and lung recovery. The device measures volume of inspiration, peak inspiratory flow, and lung vital capacity. In some examples, Airofit may be used, which may include a BLUETOOTH handheld device designed to measure respiratory function and engage the user in exercises to improve respiratory function. In some variations, SleepSense may be used, which may include a sensor attached to a band around the chest that measures respiratory effort during sleep to alert patients of apneic events.

[0059] FIG. 2 is a simplified block diagram depicting an exemplary computing environment 100 in accordance with one or more exemplary embodiments of the present application. The environment 100 includes an individual 102, a user device (e.g., mobile devices) 106 associated with the individual 102, a motion sensor 104, and a computing system (e.g., server) 108. Although the entities within environment 100 may be described below and / or depicted in the FIGs. as being singular entities, it will be appreciated that the entities and functionalities discussed herein may be implemented by and / or include one or more entities.

[0060] The entities within the environment 100 such as the user device 106, the motion sensor 104, and / or the computing system 108 may be operatively coupled to (e g., in communication with) other systems within the environment 100 via the network 110. The network 110 may be a global area network (GAN) such as the Internet, a wide area network (WAN), a local area network (LAN), or any other type of network or combination of networks. The network 110 may provide a wireline, wireless, or a combination of wireline and wireless communication between the entities within the environment 100 such as a BLUETOOTH connection between the motion sensor 104, the computing system 108, and / or the user device 106.

[0061] Individual 102 (e.g., the user 15) may be associated with and / or operate a user device 106. For instance, the user device 106 may be a mobile phone such as a smartphone that is owned by the individual 102. The user device 106 may be and / or include, but is not limited to, a desktop, laptop, tablet, mobile device (e.g., smartphone device, or other mobile device), smart watch, an internet of things (IOT) device, or any other type of computing device that generally comprises one or more communication components, one or more processing components, and one or more memory components. The user device 106 may be able to execute software applications.

[0062] The motion sensor 104 (e.g., a wearable device) monitors motions of the individual 102. For instance, the motion sensor 104 may be attached to the individual 102 such as a body part of the individual 102. For example, as mentioned above, the motion sensor 104 may be adhered to the right thorax of the individual 102 while the individual 102 is taking measured breaths (e.g., 250-2500 milliliters (mL)) through IS. In some instances, the motion sensor 104 may detect chest wall expansion and may transmit the chest wall expansion to a secure computing system 108 via the network 110. The motion sensor 104 may be any type of motion sensor or other type of sensor that is configured to capture and / or obtain sensor data associated with the individual 102. In some examples, as an alternative to the motion sensor 104, other types of sensors may be used to detect chest wall expansion and / or the inspiration of the individual 102.

[0063] The computing system 108 is a computing system that is configured to perform one or more functionalities. For instance, the computing system 108 may be used to train one or more ML - Al models and / or algorithms. For example, the computing system 108 may obtain sensor data from the motion sensor 104 and / or other motion sensors 104. The sensor data may be associated with the individual 102 and / or additional individuals. The computing system 108 may train the one or more ML - Al models based on the sensor data. Furthermore, during an inference phase, the computing system 108 may utilize sensor data to perform one or more functionalities such as measuring the volume of inspiration in postoperative pulmonary patients.

[0064] The computing system 108 includes one or more computing devices, computing platforms, systems, servers, and / or other apparatuses capable of performing tasks, functions, and / or other actions. In some variations, the computing system 108 may be implemented as engines, software functions, and / or applications. In other words, the functionalities of the computing system 108 and / or the computing system 108 may be implemented as software instructions stored in a storage (e.g., memory) and executed by one or more processors.

[0065] It will be appreciated that the exemplary environment depicted in FIG. 2 is merely an example, and that the principles discussed herein may also be applicable to other situations — for example, including other types of institutions, organizations, devices, systems, and network configurations.

[0066] FIG. 3 is a simplified block diagram of one or more devices or systems within the exemplary environment of FIG. 2. The device / system 200 includes a processor 204, such as acentral processing unit (CPU), controller, and / or logic, that executes computer executable instructions for performing the functions, processes, and / or methods described herein. In some examples, the computer executable instructions are locally stored and accessed from a non-transitory computer readable medium, such as storage 210, which may be a hard drive, flash drive, or network drive. Read Only Memory (ROM) 206 includes computer executable instructions for initializing the processor 204, while the random-access memory (RAM) 208 is the main memory for loading and processing instructions executed by the processor 204. The network interface 212 may connect to a wired network or cellular network and to a local area network or wide area network, such as the network 110. The device / system 200 may also include a bus 202 that connects the processor 204, ROM 206, RAM 208, storage 210, and / or the network interface 212. The components within the device / system 200 may use the bus 202 to communicate with each other. The components within the device / system 200 are merely exemplary and might not be inclusive of every component, device, computing platform, and / or computing apparatus within the device / system 200.

[0067] FIG. 4A shows an exemplary motion sensor and user device in accordance with one or more exemplary embodiments of the present application. For instance, FIG. 4 shows an image 300 of the motion sensor 104 described in FIG. 2 and the user device 106 described in FIG. 2 (e.g., a smartphone). FIG. 4B shows a graphical user interface (GUI) 305 of the user device 106 shown in FIG. 4. For instance, the GUI 305 shows four tabs on top and the battery life. The individual 102 may select any of the four tabs, which may represent different measuring capabilities of the motion sensor 104 (e.g., the motion sensor 104 may obtain the magnitudes and / or angles). Furthermore, based on selecting one of the tabs, the different measurements may be obtained and displayed within the graphical representation of measurements. For example, the Angles X, Angle Y, Angle Z, and “T” may be obtained. In addition, one or more user selection buttons may also be displayed in the GUI 306. For instance, by selecting “Record”, the measurements may begin recording and showing in the graphical representation of the measurements.

[0068] FIG. 5 is an exemplary process 400 for performing spirometry to measure a volume of inspiration in postoperative pulmonary patients in accordance with one or more exemplary embodiments of the present application. The process 400 may be performed by the entities depicted in environment 100 of FIG. 2. However, it will be recognized that the process 400 maybe performed in any suitable environment and that any of the following blocks may be performed in any suitable order.

[0069] At block 402, the computing system 108 obtains sensor data from wearable devices (e.g., motion sensors such as the motion sensor 104). For instance, the wearable devices may be motion sensors such as motion sensor 104 that may be adhered to individuals such as individual 102 using medical tape (e.g., hypoallergenic medical tape). The wearable devices may be adhered at approximately the level of the 9thand 10thribs at the mid-axillary line (e.g., as shown in FIG. 1, the motion sensor 20 may be adhered to the right thorax of the user 15). The individuals may take a plurality of breaths (e.g., three breaths) through an incentive spirometer device at a plurality of different volumes (e.g., 250 mL, 500 mL, 750 mL, 1000 mL, 1250 mL, 1500 mL, 1750 mL, 2000 mL, 2250 mL, and 2500 mL). The wearable devices may detect chest wall expansion, and may transmit the detected data (e.g., sensor data such as motion data) to the computing system 108 (e.g., a secure server that receives the data from the wearable device using a communication protocol such as BLUETOOTH). The computing system 108 may obtain the sensor data / motion data from the wearable devices.

[0070] For example, as mentioned above, the motion sensor 104 may be adhered to an individual 102. The individual 102 may take a plurality of breaths at a plurality of different volumes (e.g., 250 mL, 500 mL, 750 mL, 1000 mL, 1250 mL, 1500 mL, 1750 mL, 2000 mL, 2250 mL, and 2500 mL) and the motion sensor 104 may obtain sensor data (e.g., raw sensor measurements) based on the individual 102 taking the breaths. For instance, the sensor data may be the raw wearable waveform data that is shown in FIGs. IB and 1C. For example, the motion sensor 104 may obtain sensor data indicating motion measurements (e.g., distance and / or displacement measurements) of the individual 102 as the individual 102 is taking the plurality of breaths at the different volumes. In addition, the motion sensor 104 may obtain the sensor data for a plurality of axes such as the x-axis, y-axis, and the z-axis, which are shown in FIG. 1A. The motion sensor 104 may provide the sensor data to another entity in environment 100 such as the user device 106 and / or the computing system 108. For instance, the computing system 108 may obtain the sensor data directly from the motion sensor 104. Alternatively, the user device 106 may initially receive the sensor data, and then provide the sensor data and / or other data to the computing system 108. In some examples, in addition to or as an alternative to the motion measurements, themotion sensor 104 may obtain other sensor data that may be used by the computing system 108 for training the one or more ML - Al models. For example, the other sensor data may include, but is not limited to, velocity measurements, acceleration measurements, and / or other types of sensor measurements and / or data.

[0071] At block 404, the computing system 108 trains one or more ML - Al models. For example, using the motion data and / or other data, the computing system 108 may train one or more ML - Al models, which may be trained to recognize patterns in chest wall motion associated with one or more lung volume categories (e.g., a low, medium, or high category). For instance, the computing system 108 may parse out the motion data from the wearable devices into different volume of inspiration categories. For example, as mentioned above, the individuals may take breaths through an IS device at a plurality of different volumes. Certain volumes may be placed in certain categories such as the 250 mL, 500 mL, and 750 mL being placed in a low-volume category (e.g., a low lung volume category), the 1000 mL, 1250 mL, and 1500 mL being placed in a medium -volume category (e.g., a medium lung volume category), and the 2000 mL, 2250 mL, and the 2500 mL being placed in a high-volume category (e.g., a high lung volume category). Afterwards, the computing system 108 may use the motion data associated with the different categories to train the one or more ML - Al models. For instance, the ML - Al models may be trained on these sequences to learn to recognize patterns in the chest wall motion associated with teach incentive lung volume category. Additionally, and / or alternatively, the computing system 108 may use a leave-one-out validation strategy to assess the performance of the ML - Al models.

[0072] Additionally, and / or alternatively, the computing system 108 may pre-process the motion data from the wearable devices such as quality checking the motion data, removing corrupted waveforms from the motion data, removing motion artifacts (e.g., waveforms corresponding to activity) from the motion data, and / or perform other signal processing techniques.

[0073] In some examples, the motion data may include data in three axes - an x-axis, a y-axis, and a z-axis. The movement in the x-axis may be defined as chest wall movement in the cranial-caudal axis (e.g., parallel to the thorax). The movement in the y-axis may correspond to movement perpendicular to the chest wall. The movement in the z-axis (e.g., the z-axis signal) may correspond to the movement in the upwards and outwards, or in the diagonal, direction. This isshown in FIG. 1. Furthermore, there may be visible differences in waveform amplitude in the y and z axes as participants took higher-volume breaths (e.g., in the medium and / or high categories).

[0074] For example, the computing system 108 may obtain sensor data from the motion sensor 104 as well as other motion sensors and / or other sensors. The computing system 108 may generate training data for the one or more ML - Al models based on the sensor data. For instance, after obtaining sensor data from a plurality of different sensors (e.g., different motion sensors), the computing system 108 may pre-process the sensor data to generate the training data. The computing system 108 may pre-process the sensor data using any type of pre-processing technique, process, and / or algorithm. For instance, the computing system 108 may perform quality checking of the sensor data such as by comparing the sensor data with one or more thresholds. The computing system 108 may then discard or remove portions of the sensor data (e.g., the corrupted waveforms). Additionally, and / or alternatively, the computing system 108 may perform one or more signal processing techniques to remove outliers and / or otherwise filter the sensor data such as removing motion artifacts. For instance, while the motion sensor 104 is obtaining sensor measurements for the individual 102, the individual 102 may have moved, which may cause a motion artifact. The computing system 108 may remove the motion artifacts (e.g., smooth out the waveform) from the sensor data to generate the training data.

[0075] After performing the pre-processing, the computing system 108 may obtain the training data and use the training data to train one or more ML - Al models. For instance, the one or more ML - Al models may be and / or include any type of ML - Al models and the computing system 108 may use one or more training techniques to train the ML - Al models using the training data. In some examples, the computing system 108 may train a plurality of ML - Al models. For example, the computing system 108 may categorize (e.g., sort) the training data into a plurality of categories such as a first lung volume category (e.g., a low volume category), a second lung volume category (e.g., a medium volume category), and a third lung volume category (e.g., a high volume category). For example, the first lung volume category may include a first set of training data associated with low volumes of inspiration such as 250 mL, 500 mL, and / or 750 mL. The second lung volume category may include a second set of training data associated with medium volumes of inspiration such as 1000 mL, 1250 mL, and / or 1500 mL. The third lung volume category may include a third set of training data associated with high volumes of inspiration such as 2000 mL,2250 mL, and / or 2500 mL. The computing system 108 may train a first ML - Al model using the first set of training data, a second ML - Al model using the second set of training data, and a third ML - Al model using the third set of training data. As such, each of the plurality of ML - Al models may be associated with different volumes of inspiration from the sensor data of the individuals including individual 102.

[0076] In some examples, based on the training, the ML - Al models may be configured to recognize patterns in chest wall motion associated with each incentive lung volume category. For example, the ML - Al models may be trained to detect anomalies, which may indicate one or more medical conditions such as PPCs. In other words, during inference, the individual 102 or another individual may attach a motion sensor such as the motion sensor 104 to themselves. The individual 102 may take deep breaths at a plurality of volumes (e.g., the volumes of inspiration described above). The motion sensor 104 may detect and / or record the motion (e.g., displacement, velocity, acceleration, and / or other types of motion in one or more axes such as the x-axis, y-axis, and z-axis). Subsequently, the sensor data indicating the detected motion may be processed by the ML - Al models to determine an anomaly. For instance, the ML - Al models may process the sensor data to generate an ML - Al output such as an indication (e.g., a value) indicating a likelihood or probability of an anomaly. Based on the ML - Al output, one or more medical conditions may be detected for the individual. In some variations, more than one ML - Al model may process the sensor data. For instance, the first ML - Al model, which may be trained on the first set of training data described above, may process a first portion of the sensor data from the individual (e.g., the sensor data associated with the low volumes of inspiration), the second ML - Al model may process a second portion of the sensor data (e.g., the sensor data associated with the medium volumes of inspiration), and the third ML - Al model may process a third portion of the sensor data (e.g., the sensor data associated with the high volumes of inspiration). Each of the ML - Al models may generate an output. Subsequently, the medical condition may be determined based on the outputs from the ML - Al models (e.g., based on an average, a majority, weighted values, and / or other processes and / or methods).

[0077] In some variations, a leave-one-out validation strategy may be used to assess the ML -Al models performance. For example, the leave-one-out validation strategy may include splitting the training data into a training set and a testing set where the training set includes all but oneobservation (e.g., if the sensor data of six individuals were obtained, the training set would include five of the six individual’s sensor data) and the testing set would include the last observation (e.g., the sensor data associated with the sixth individual). In each training epoch, the training data may be split such that a different observation is used as the testing set.

[0078] At block 406, subsequent to training the one or more ML - Al models, the computing system 108 uses the one or more ML - Al models to perform one or more tasks. For instance, after training the one or more ML - Al models, the computing system 108 may obtain sensor data from the motion sensor 104 associated with the individual 102. The computing system 108 may process the sensor data using the one or more trained ML - Al models to perform one or more tasks and / or functions such as assessing for PPCs. For example, pulmonary complications, which range from minor respiratory alterations to severe conditions such as pneumonia or respiratory failure, are a culmination of changes to the respiratory system that begin immediately after induction of general anesthesia. As mentioned above, IS has traditionally served as a cornerstone in the prophylaxis and management of PPCs, but it has been recently documented that conventional IS may have various limitations. As such, the computing system 108 may utilize sensor data from a wearable device (e g., the motion sensor 104) and the one or more ML - Al models to detect and / or identify unique patterns in chest wall motion via motion sensors adhered to the patient’s thorax to easily measure and track volume of inspiration and respiratory effort post-operatively. For instance, the wearable motion sensor 104 may serve as a replacement or substitute for IS, which may include advantages including, but not limited to, improved patient compliance, facilitating longitudinal IS data collection, and reducing the burden on nursing staff who continuously have to provide incentive spirometers and remind patients to use them.

[0079] In some examples, the computing system 108 may communicate with a user device 106. For instance, the computing system 108 and the user device 106 may help facilitate the postoperative recovery period from the hospital and even from the home of the individual 102. For instance, the user device 106 may automatically remind individuals to complete breathing exercises via timed vibrations, interface with an application, and / or even perform interactive patient gaming strategies. Additionally, and / or alternatively, it may offer the ability for providers to interact with, document, and analyze longitudinal spirometry data in a low-cost way even outside the bounds of the hospital.

[0080] In other words, in some variations, after training the ML - Al models, the ML - Al models may be used to diagnose one or more medical conditions such as PPCs. For example, the computing system 108 may obtain new sensor data (e.g., inference sensor data) from the motion sensor 104 and / or the individual 102. In some examples, the new sensor data may be from the same motion sensor 104 and / or the individual 102 as the training data (e.g., the sensor data obtained at block 404). In other examples, the new sensor data may be from a different motion sensor 104 and / or a different individual 102 as the training data. The computing system 108 may process the new sensor data using the one or more ML - Al models to generate one or more ML - Al outputs indicating identified patterns. For instance, in some variations, the computing system 108 may detect anomalies within the new sensor data processing the new sensor data using the one or more ML - Al models, which is described above. Based on detecting the anomalies, the computing system 108 may diagnose the individual 102 associated with the new sensor data (e.g., assessing whether the individual 102 has PPCs).

[0081] In some instances, instead of the computing system 108 diagnosing the individual 102, the user device 106 may be used to diagnose the individual 102. For instance, after training, the computing system 108 may provide the trained ML - Al models to the user device 106. The user device 106 may obtain the new sensor data from the motion sensor 104, and may process the new sensor data using the trained ML - Al models to detect anomalies indicating the one or more medical conditions such as PPCs.

[0082] In some examples, the user device 106 may be configured to provide notifications to the individual 102 indicating for the individual 102 to adhere the motion sensor 104 to themselves and record the new sensor data. For example, the user device 106 may automatically provide notifications reminding individuals such as the individual 102 to complete breathing exercises via timed vibrations, interface with an application, and / or even perform interactive patient gaming strategies. Based on the notifications, the individual 102 may adhere the motion sensor 104, and the motion sensor 104 may obtain the new sensor data. The new sensor data may be used by the trained ML - Al models to detect medical conditions such as PPCs.

[0083] All references, including publications, patent applications, and patents, cited herein are hereby incorporated by reference to the same extent as if each reference were individually and specifically indicated to be incorporated by reference and were set forth in its entirety herein.

[0084] The use of the terms “a” and “an” and “the” and “at least one” and similar referents in the context of describing the invention (especially in the context of the following claims) are to be construed to cover both the singular and the plural, unless otherwise indicated herein or clearly contradicted by context. The use of the term “at least one” followed by a list of one or more items (for example, “at least one of A and B”) is to be construed to mean one item selected from the listed items (A or B) or any combination of two or more of the listed items (A and B), unless otherwise indicated herein or clearly contradicted by context. The terms “comprising,” “having,” “including,” and “containing” are to be construed as open-ended terms (i.e., meaning “including, but not limited to,”) unless otherwise noted. Recitation of ranges of values herein are merely intended to serve as a shorthand method of referring individually to each separate value falling within the range, unless otherwise indicated herein, and each separate value is incorporated into the specification as if it were individually recited herein. All methods described herein can be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The use of any and all examples, or exemplary language (e.g., “such as”) provided herein, is intended merely to better illuminate the invention and does not pose a limitation on the scope of the invention unless otherwise claimed. No language in the specification should be construed as indicating any non-claimed element as essential to the practice of the invention.

[0085] Preferred embodiments of this invention are described herein, including the best mode known to the inventors for carrying out the invention. Variations of those preferred embodiments may become apparent to those of ordinary skill in the art upon reading the foregoing description. The inventors expect skilled artisans to employ such variations as appropriate, and the inventors intend for the invention to be practiced otherwise than as specifically described herein. Accordingly, this invention includes all modifications and equivalents of the subject matter recited in the claims appended hereto as permitted by applicable law. Moreover, any combination of the above-described elements in all possible variations thereof is encompassed by the invention unless otherwise indicated herein or otherwise clearly contradicted by context.

Claims

1. CLAIM(S):

1. A system for performing spirometry to diagnose one or more medical conditions based on using one or more machine learning - artificial intelligence (ML - Al) models, comprising: a first wearable device configured to:3.obtain motion data associated with a first individual, wherein the first wearable device is adhered to the first individual and configured to detect chest wall expansions of the first individual; and4.transmit the motion data to a server; and5.the server configured to:6.receive, from a plurality of wearable devices that includes the first wearable device, training sensor data associated with a plurality of individuals, wherein the training sensor data comprises the motion data associated with the first individual;7.train the one or more ML - Al models using the training sensor data; subsequent to training the one or more ML - Al models, obtain new sensor data associated with a second individual; and8.process the new sensor data using the one or more ML - Al models to detect whether the second individual has the one or more medical conditions.

2. The system of claim 1, wherein the first wearable device is a motion sensor, wherein the motion sensor is adhered to a level of a 9thand 10thrib at a mid-axillary line of the first individual.

3. The system of claim 1, wherein training the one or more ML - Al models comprises: categorizing the training sensor data into a plurality of different lung volume categories, wherein the plurality of different lung volume categories comprises a first lung volume category, a second lung volume category, and a third lung volume category; and11.training the one or more ML - Al models based on categorizing the training sensor data into the plurality of different lung volume categories.

4. The system of claim 3, wherein training the one or more ML - Al models further comprises:prior to categorizing the training sensor data into the plurality of different lung volume categories, pre-processing the training sensor data to remove corrupted waveforms from the training sensor data and / or to remove motion artifacts from the training sensor data.

5. The system of claim 3, wherein categorizing the training sensor data into the plurality of different lung volume categories comprises sorting the training sensor data into a first set of training data associated with the first lung volume category, a second set of training data associated with the second lung volume category, and a third set of training data associated with the third lung volume category, and wherein training the one or more ML - Al models comprises:14.training a first ML - Al model, from the one or more ML - Al models, using the first set of training data;15.training a second ML - Al model, from the one or more ML - Al models, using the second set of training data; and16.training a third ML - Al model, from the one or more ML - Al models, using the third set of training data.

6. The system of claim 1, wherein the training sensor data indicates motion data in three axes, wherein movement in an x-axis indicates chest wall movement in a cranial -caudal axis, wherein movement in a y-axis indicates movement that is perpendicular to a chest wall, and wherein movement in a z-axis indicates movement in an upwards and outwards direction from the plurality of individuals.

7. The system of claim 1, wherein training the one or more ML - Al models is based on using a leave-one-out validation strategy that splits the training sensor data into a training set and a testing set, wherein in each training epoch, a different portion of the training sensor data from a different individual of the plurality of individuals is included within the testing set and remaining training sensor data is included within the training set.

8. The system of claim 1, wherein processing the new sensor data using the one or more ML - Al models to detect whether the second individual has the one or more medical conditionscomprises assessing whether the second individual has postoperative pulmonary complications (PPCs).

9. A method for performing spirometry to diagnose one or more medical conditions based on using one or more machine learning - artificial intelligence (ML - Al) models, comprising: obtaining, from a plurality of wearable devices, training sensor data associated with a plurality of individuals, wherein the plurality of wearable devices are adhered to the plurality of individuals and configured to detect chest wall expansions of the plurality of individuals;21.training the one or more ML - Al models using the training sensor data;22.subsequent to training the one or more ML - Al models, obtaining new sensor data associated with an individual; and23.processing the new sensor data using the one or more ML - Al models to detect whether the individual has the one or more medical conditions.

10. The method of claim 9, wherein the plurality of wearable devices comprises one or more motion sensors, wherein a first motion sensor, of the one or more motion sensors, is adhered to a level of a 9thand 10thrib at a mid-axillary line of a first individual of the plurality of individuals.

11. The method of claim 9, wherein obtaining the training sensor data comprises obtaining, by a secure server, the training sensor data from the plurality of wearable devices.

12. The method of claim 9, wherein training the one or more ML - Al models comprises: categorizing the training sensor data into a plurality of different lung volume categories, wherein the plurality of different lung volume categories comprises a first lung volume category, a second lung volume category, and a third lung volume category; and27.training the one or more ML - Al models based on categorizing the training sensor data into the plurality of different lung volume categories.

13. The method of claim 12, wherein training the one or more ML - Al models further comprises:prior to categorizing the training sensor data into the plurality of different lung volume categories, pre-processing the training sensor data to remove corrupted waveforms from the training sensor data and / or to remove motion artifacts from the training sensor data.

14. The method of claim 12, wherein categorizing the training sensor data into the plurality of different lung volume categories comprises sorting the training sensor data into a first set of training data associated with the first lung volume category, a second set of training data associated with the second lung volume category, and a third set of training data associated with the third lung volume category, and wherein training the one or more ML - Al models comprises:30.training a first ML - Al model, from the one or more ML - Al models, using the first set of training data;31.training a second ML - Al model, from the one or more ML - Al models, using the second set of training data; and32.training a third ML - Al model, from the one or more ML - Al models, using the third set of training data.

15. The method of claim 9, wherein the training sensor data indicates motion data in three axes, wherein movement in an x-axis indicates chest wall movement in a cranial -caudal axis, wherein movement in a y-axis indicates movement that is perpendicular to a chest wall, and wherein movement in a z-axis indicates movement in an upwards and outwards direction from the plurality of individuals.

16. The method of claim 9, wherein training the one or more ML - Al models is based on using a leave-one-out validation strategy that splits the training sensor data into a training set and a testing set, wherein in each training epoch, a different portion of the training sensor data from a different individual of the plurality of individuals is included within the testing set and remaining training sensor data is included within the training set.

17. The method of claim 9, wherein processing the new sensor data using the one or more ML - Al models to detect whether the individual has the one or more medical conditions comprises assessing whether the individual has postoperative pulmonary complications (PPCs).

18. The method of claim 17, wherein the one or more ML - Al models comprises a first ML -Al model, a second ML - Al model, and a third ML - Al model, and wherein processing the new sensor data using the one or more ML - Al models further comprises:36.sorting the new sensor data into a first set of sensor data associated with a first lung volume category, a second set of sensor data associated with a second lung volume category, and a third set of sensor data associated with a third lung volume category;37.processing the first set of sensor data using the first ML - Al model to generate a first ML - Al output;38.processing the second set of sensor data using the second ML - Al model to generate a second ML - Al output; and39.processing the third set of sensor data using the third ML - Al model to generate a third ML - Al output, wherein assessing whether the individual has PPCs is based on the first ML - Al output, the second ML - Al output, and the third ML - Al output.

19. The method of claim 9, wherein obtaining the new sensor data associated with the individual is in response to a user device providing a notification to the individual, wherein the notification comprises an automatic reminder to the individual to complete breathing exercises.

20. A non-transitory computer-readable medium having processor-executable instructions stored thereon for performing spirometry to diagnose one or more medical conditions based on using one or more machine learning - artificial intelligence (ML - Al) models, wherein the processor-executable instructions, when executed, facilitate:42.obtaining, from a plurality of wearable devices, training sensor data associated with a plurality of individuals, wherein the plurality of wearable devices are adhered to the plurality of individuals and configured to detect chest wall expansions of the plurality of individuals;43.training the one or more ML - Al models using the training sensor data; subsequent to training the one or more ML - Al models, obtaining new sensor data associated with an individual; and44.processing the new sensor data using the one or more ML - Al models to detect whether the individual has the one or more medical conditions.

Images