Pulminary acoustic sensor telemetry array for remote monitoring and diagnosis

Abstract

Various examples are provided related to pulmonary acoustic sensor telemetry arrays and their use. In one example, a pulmonary acoustic sensor telemetry array (PASTA) system for measuring audio characteristics of a body including an array of acoustic sensors; an edge computing device; and an audio interface that can receive asynchronous audio signals from the array of acoustic sensors and provide synchronized audio signal data to the edge computing device in near-real time for data collection and processing.

Description

Docket: 320903-2560PULMINARY ACOUSTIC SENSOR TELEMETRY ARRAY FOR REMOTE MONITORING AND DIAGNOSISCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to, and the benefit of, U.S. provisional application entitled “PASTA: PULMINARY ACOUSTIC SENSOR TELEMETRY ARRAY FOR REMOTE MONITORING AND DIAGNOSIS” having serial no. 63 / 695,299, filed September 16, 2024, which is hereby incorporated by reference in its entirety.BACKGROUND

[0002] Respiratory disease is the 4thleading cause of death worldwide. More than one in eight people have a respiratory condition, which is over 1 billion people. Among the pediatric population, lower respiratory infection is the leading cause of death in children under 5 years of age.

[0003] Driven by the COVID-19 pandemic and recent advancements in Internet-of- Things (loT) technologies, the healthcare industry has experienced a heightened demand for remote and contactless methods of clinical care. While cardiac telemetry is commonplace, especially in hospital care settings, pulmonary evaluation typically occurs via manual auscultation, with various imaging modalities (e.g., X-ray, ultrasound, and computerized tomography) employed as subsequent evaluation steps. While oxygen saturation monitoring and capnography are common, continuous non-invasive pulmonary monitoring is not commonly available for inpatient clinical care, despite respiratory conditions remaining the number one cause of hospitalization in children nationwide.SUMMARY

[0004] Aspects of the present disclosure are related to pulmonary acoustic sensor telemetry arrays and their use for remote monitoring and diagnosis. To modernize the pulmonary auscultation process, the present disclosure discloses a hardware prototype capable of recording data from an array of acoustic pulmonary sensors simultaneously. These acoustic sensors can be placed at various locations on the chest and back to collect and monitor lung sounds in a comfortable and non-invasive manner. The disclosed pulmonary acoustic sensor telemetry array (PASTA) technology can provide clinicians with a portable and lightweight device which enables them to evaluate patients via remote continuous pulmonary auscultation.Docket: 320903-2560

[0005] An aspect of the disclosure provides a method and apparatus for obtaining sound data from one or more sensing areas and processing the synched audio data to be manipulated in a remote or direct synchronous or asynchronous manner by using a low-cost portable wearable device. These acoustic sensors can be placed at various locations on the check or back to collect and monitor lung sounds in a comfortable and noninvasive manner. This has thus outlined, rather broadly, the base features of the principals of present disclosure in order that the present contribution may be better appreciated. There are additional features of the principals of present disclosure that will be described hereinafter and that will form the subject matter of the claims appended hereto.

[0006] In this respect, before explaining at least one embodiment of the principals of present disclosure in detail, it is understood that the principals of present disclosure are not limited in its application to the details of construction and to the arrangements of the components set forth in the following description or illustrated in the following drawings. The principals of present disclosure are capable of other embodiments and of being practiced and carried out in various ways. Also, it is to be understood that the phraseology and terminology employed herein are for the purpose of the description and should not be regarded as limiting.

[0007] In one aspect, among others, a pulmonary acoustic sensor telemetry array (PASTA) system for measuring audio characteristics of a body, comprising: an array of acoustic sensors; an edge computing device; and an audio interface configured to receive asynchronous audio signals from the array of acoustic sensors and provide synchronized audio signal data to the edge computing device in near-real time for data collection and processing. In one or more aspects, the array of acoustic sensors can comprise a plurality of microphones configured to be attached to front and back surfaces of the body. The plurality of microphones can be attached at defined auscultation points on the body. The array of acoustic sensors can comprise eight microphones with four attached to the front surface and four attached to the back surface. In various aspects, the audio signals from the array of acoustic sensors can be analog signals and the audio interface converts the analog signals to digital signals for processing by the edge computing device. The analog signals can be converted to 1 -bit digital data through the exploitation of pulse density modulation. The analog signals can be filtered prior to digital conversion.

[0008] In one or more aspects, the edge computing device can comprise a processor and memory that stores executable instructions that can be executed by the processor. The edge computing device can comprise a wireless communications interface configured to link to a wireless network, where the edge computing device communicates the audio signal data to a remote computing device via the wireless network. The edge computing device can compile the audio signal data from the audio interface and continuously or non-continuouslyDocket: 320903-2560 processes the data for communication or diagnosis. The edge computing device can communicate the audio signal data to the remote computing device for storage of the audio signal data. The edge computing device can communicate the audio signal data to the remote computing device for remote processing of the audio signal data. The remote processing can comprise comparing a performance of an output sample to a baseline performance sample to determine a difference in user health. The baseline performance sample can be received from an input sample. The audio signal data can be stored and analyzed over time for abnormalities, deviations, or both from a baseline performance. The edge computing device can communicate the audio signal data to the remote computing device for access by an authorized user. The authorized user can be a physician or nurse practitioner, and access is in near-real time. The user can manipulate the audio signal data. The PASTA system can be wearable. Each sensor of the array of acoustic sensors can be encased in silicon.

[0009] In one aspect, the system can contains a processor system with a second processor operably connected directly or indirectly to the processor and / or PASTA interface, and a memory unit or storage device that stores executable instructions that can be executed by the processing system.

[0010] In another aspect, the memory unit or storage device facilitating performance of operations can contain the following steps: obtaining the analog acoustic signals from the PASTA interface device, generally filtering, including for external noise filtering; and converted from analog to 1 -bit digital data through the exploitation of pulse density modulation or other similar algorithms, transmitting information associated with the refined output to the processor system, and comparing the performance of the output sample to a baseline performance sample to determine a difference in user health.

[0011] In another aspect, the system can contain a baseline performance sample received from an input sample and / or a baseline performance sample captured when the user is determined to be healthy to determine health of sensing area.

[0012] In another aspect, the system can contain eight sensing areas on the patient, with two sensing areas on the chest, two sensing areas on the sides of the torso, and four sensing areas on the back.

[0013] In another aspect, the system can contain a processing system compiling the data from the PASTA interface and continuously or non-continuously processing the data for diagnosis; and / or the processing system compiling the data from the PASTA interface and in real time transmits the data to an operably connected processor or internet where the data can be transmitted and manipulated by a non-patient to a remote operably connected computer; and / or the processing system compiling the data from the PASTA interface stored and analyzed over time for abnormalities and / or deviations from baseline performance;Docket: 320903-2560 and / or the processing system compiling the data from the PASTA interface operably connected to a database to perform analysis and is adjustable to a specific condition, age, or other known demographic.

[0014] Other systems, methods, features, and advantages of the present disclosure will be or become apparent to one with skill in the art upon examination of the following drawings and detailed description. It is intended that all such additional systems, methods, features, and advantages be included within this description, be within the scope of the present disclosure, and be protected by the accompanying claims. In addition, all optional and preferred features and modifications of the described embodiments are usable in all aspects of the disclosure taught herein. Furthermore, the individual features of the dependent claims, as well as all optional and preferred features and modifications of the described embodiments are combinable and interchangeable with one another.BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Many aspects of the present disclosure can be better understood with reference to the following drawings. The components in the drawings are not necessarily to scale, emphasis instead being placed upon clearly illustrating the principles of the present disclosure. Moreover, in the drawings, like reference numerals designate corresponding parts throughout the several views.

[0016] FIG. 1A is a system-level operational diagram that illustrates an example of the proposed pulmonary acoustic sensor telemetry array (PASTA) technology, in accordance with various embodiments of the present disclosure.

[0017] FIG. 1 B is a flow diagram illustrating an example of the flow of data across various PASTA components, in accordance with various embodiments of the present disclosure.

[0018] FIG. 1C illustrates an example of integration of PASTA into a hospital Internet of Things (loT) network, in accordance with various embodiments of the present disclosure.

[0019] FIG. 2 illustrates auscultation points for lung sounds on a human body, in accordance with various embodiments of the present disclosure.

[0020] FIG. 3 illustrates an example of different types of acoustic signals, in accordance with various embodiments of the present disclosure.

[0021] FIG. 4 is an image of a conventional Eko Core 500TM digital stethoscope, in accordance with various embodiments of the present disclosure.

[0022] FIG. 5 is a block diagram illustrating an example of the PASTA audio interface architecture, in accordance with various embodiments of the present disclosure.Docket: 320903-2560

[0023] FIG. 6 is an image of I2S MEMS telemetry sensors (microphone) array used in an implemented PASTA device, in accordance with various embodiments of the present disclosure.

[0024] FIG. 7 is an image of an audio interface (MCHStreamer multi-channel) board used in the implemented PASTA device, in accordance with various embodiments of the present disclosure.

[0025] FIGS. 8A and 8B are images of edge computing devices, Raspberry Pi 4 (8A) and NVIDIA Jetson Orin (8B) used for data collection in the PASTA device, in accordance with various embodiments of the present disclosure.

[0026] FIG. 9 is a table of hardware and parameters used in PASTA device, in accordance with various embodiments of the present disclosure.

[0027] FIG. 10 is a flow diagram illustrating an example of data flow with low-level acoustic (microphone) sensors array (operating at high bandwidth) and high-level software applications (operating at low bandwidth), in accordance with various embodiments of the present disclosure.

[0028] FIG. 11 illustrates an example of a validation process, in accordance with various embodiments of the present disclosure.

[0029] FIG. 12 illustrates an example of an algorithm for frequency spectrum adjustment for similarity measurement, in accordance with various embodiments of the present disclosure.

[0030] FIGS. 13A and 13B are images of the PASTA device mounted on a simulation manikin and data collection with the PASTA device, in accordance with various embodiments of the present disclosure.

[0031] FIG. 14 illustrates the four different hardware configurations for sound data collection from the simulation manikin during testing, in accordance with various embodiments of the present disclosure.

[0032] FIG. 15A illustrates examples of experimental results collected from the simulation manikin generating nine different sound profiles on the PASTA device, in accordance with various embodiments of the present disclosure.

[0033] FIG. 15B is a table illustrating a cosine similarity comparison of the PASTA technology with different pulmonary devices, in accordance with various embodiments of the present disclosure.

[0034] FIG. 16 illustrates examples of experimental data vs time collected from manikin generating normal breathing sound and pneumonia sound, in accordance with various embodiments of the present disclosure.Docket: 320903-2560

[0035] FIG. 17 is a table illustrating a comparison of specifications and pricing between PASTA and two commercial digital stethoscopes, in accordance with various embodiments of the present disclosure.

[0036] FIG. 18 illustrates an example of a PASTA device encased in a housing, in accordance with various embodiments of the present disclosure.

[0037] FIG. 19 is a flow diagram illustrating an example of the high-level process flow from the collection of data to form a sound sample or data package, to sample comparison that can be done prior to or after the off-site analysis done by an off-site processor prior to diagnosis, in accordance with various embodiments of the present disclosure.DETAILED DESCRIPTION

[0038] Disclosed herein are various examples related to pulmonary acoustic sensor telemetry arrays and their use for remote monitoring and diagnosis. Reference will now be made in detail to the description of the embodiments as illustrated in the drawings, wherein like reference numbers indicate like parts throughout the several views.

[0039] The advantages and features of the present disclosure and the manner of achieving them will become apparent with reference to the embodiments described in detail below. The present disclosure may, however, be embodied in many different forms and is not limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0040] It will be understood that each block of the flowchart illustrations, and combinations of blocks in the flowchart illustrations, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate means for implementing the functions specified in the flowchart block or blocks. These computer program instructions may also be stored in a computer usable or computer- readable memory that may direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer usable or computer-readable memory produce an article of manufacture including instruction means that implement the function specified in the flowchart block or blocks. The computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented processDocket: 320903-2560 such that the instructions that are executed on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0041] Each block of the flowchart illustrations may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks may occur out of order. For example, two blocks shown in succession may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.

[0042] Auscultation refers to the medical practice of listening to sounds produced from a patient’s heart, lungs, or other organs, and is a crucial aspect of standard physical examinations. As a core medical instrument, conventional stethoscopes are used by placing a diaphragm on a patient’s body; the sound is acoustically amplified, then conducted into the earpieces worn in the ears. However, this requires physical contact between patient and practitioner, which is not always convenient.

[0043] Continuous monitoring of physiological sounds is particularly crucial in the management of illnesses that require ongoing assessment of cardiac, respiratory, and / or vascular function. Conditions such as heart diseases, lung disorders, vascular abnormalities, and postoperative complications often necessitate frequent auscultation to detect changes in sound patterns indicative of disease progression, treatment response, or potential complications.

[0044] Fortunately, with the recent advancement in technology, digital stethoscopes were developed. Studies showed digital stethoscopes are not only equal in quality to conventional stethoscopes, but also helpful in identifying false negative diagnoses during in- person cardiology visits when used as a tool of telecardiology. The telecardiology (a part of telehealth systems) method has proven to decrease hospitalizations by 10% in nursing homes. A modern telehealth system can be established to deliver secure and timely diagnosis to the public.

[0045] Past studies presented preliminary proof of concept of similar devices with low- cost sound sensors, such as piezoelectric and MEMS devices, together with embedded microcontrollers. However, the result, including number of channels, sampling rate, and accuracy, is far from sufficient to compete with commercial digital stethoscopes on the market. A study performed in 2011 demonstrated the possibility that open-platform digital stethoscopes can be integrated into a point-of-care device to deliver high-quality healthcare to rural populations.

[0046] Finally, leveraging the high data throughput and authenticity of digital auscultation, a cloud-based machine learning model could be employed such that aDocket: 320903-2560 preliminary diagnosis can be provided as an early detection method almost instantaneously upon data collection, possibly preventing serious conditions from developing. For example, another 2011 study employed joint time-frequency domain analysis on pulmonary sounds and performed k-nearest-neighbor method analysis to classify wheezes and stridor from normal breathing sounds.

[0047] To detect abnormal patterns in cardiopulmonary data, traditional signal processing in time and frequency is effective, but advanced methods may result in better feature extraction, such as LungBRN, a machine learning algorithm that can diagnose respiratory disease using short-term Fourier transform and wavelet feature extraction, and the bi-ResNet deep learning architecture. Using transfer learning, it has been proven that tuning a pre-trained ResNet model is effective enough to outperform many lung sound classification systems at the time.

[0048] Though it is evident that using digital stethoscopes for modern cardiology is feasible, from an application standpoint, such devices still have limitations yet to overcome, including lack of validation, skepticism from patients, relatively high cost, and lack of full integration into healthcare systems. In addition, they are usually made for a closed platform where they require auxiliary devices and do not allow for auscultation data to be sent to physicians over the internet, and usually do not allow for continuous monitoring from multiple points on the body, either.

[0049] Therefore, there is a clear need for an innovative device and method that enables remote auscultation, allowing healthcare providers to listen to internal body sounds from a distance in real time. Such a device can facilitate continuous monitoring of physiological sounds without the need for direct physical contact between the patient and the clinician, while also being a cost-effective open-platform digital device for monitoring adventitious sounds, while also keeping the platform low-profile and modular for extended usage.

[0050] This disclosure presents a pulmonary acoustic sensor telemetry array (PASTA), a novel hardware prototype developed for remote pulmonary monitoring, diagnosis, and prognostication of pulmonary sounds that can be utilized in, e.g., clinical settings. The proposed hardware prototype employs an array of acoustic sensors that can be strategically positioned at standard auscultation sites on the patient’s chest and back, enabling real-time acquisition and analysis of pulmonary sounds with an edge computing device. This task is beyond the capabilities of existing commercially available digital stethoscopes. The hardware prototype presented herein addresses critical challenges in pulmonary care by providing continuous, high-fidelity, multi-site acoustic data collection. This functionality is pivotal for enhancing diagnostic accuracy and enabling early detection of respiratory abnormalities, potentially reducing treatment delays in acute respiratory conditions.Docket: 320903-2560

[0051] The PASTA integrates a compact design, multi-channel monitoring, and real-time telemetry, distinguishing itself from existing commercial medical devices. A novel hardware prototype was designed to revolutionize the pulmonary auscultation process. The prototype leverages an array of acoustic pulmonary sensors strategically placed on the patient’s chest and back. These sensors were engineered to collect and monitor lung sounds simultaneously, offering a comfortable and non-invasive solution for continuous pulmonary evaluation. This system can enhance the accuracy and efficiency of respiratory assessments and meets the growing need for innovative healthcare solutions that prioritize patient safety and convenience.

[0052] FIG. 1A highlights a system-level operational diagram that illustrates an example of the proposed pulmonary acoustic sensor telemetry array (PASTA) technology. It illustrates the comprehensive architecture for remote monitoring and diagnosis of lung sounds through auscultation. FIG. 1 B illustrates the flow of data across various components, highlighting the integration of microphone arrays, data processing units, and medical professional monitoring interfaces. The block diagram illustrates the different components and data / information flows. FIG. 10 shows an example of how PASTA can be integrated into a hospital Internet of Things (loT) network. The proposed pulmonary sensor, along with body temperature and other biometric sensors, can continuously capture data from patients. These health metrics can become a real-time representation of the patient’s health status and can be uploaded into the hospital loT Wide Area Network (WAN). There are several uses of such data. For example, (1) it can provide real-time monitoring of the patient’s health, alerting staff in time when necessary before patients realize a problem exists; (2) the metrics can be stored in a database available for healthcare providers to conduct in-depth diagnosis; and (3) because the captured biometrics are synchronous and continuous, the system can provide a multidimensional view of the data that is not currently available for research in the fields of Pulmonology, Hospital Medicine, and Critical Care.

[0053] PASTA collects a patient’s lung acoustics with an array of microphone sensors, which are then passed on to the medical provider in near-real-time, supporting accurate and timely diagnosis. Its real-time multi-channel functionality enables clinicians to perform auscultation analysis from remote locations, something that is impossible with current digital stethoscopes. In addition, the device gives healthcare providers the ability to listen to one or more audio channels simultaneously, allowing them to directly compare the homogeneity of sounds bilaterally, which is a key component of diagnostic analysis. These data can be selectively saved. The device can also enable annotation as well. In addition to remote auscultation, PASTA can localize lung sounds and enabling review using a three- dimensional sound format (i.e., ambisonics). This format, when coupled with motion tracking audio devices, can allow providers to listen to lung sounds as if they were inside the patient’sDocket: 320903-2560 lungs, moving their head to hear certain areas better, and to discern where certain sounds are being produced in the lungs. This capability has immense potential to increase the speed and effectiveness of identified focal lung findings (such as, e.g., bacterial pneumonia), which in turn may result in more accurate treatment plans for a variety of respiratory conditions. PASTA’s low cost encourages its use in resource-limited areas, while its low profile, non- invasive design, and use of standard data connections and transfer protocols ensures its potential for scalability within even the largest healthcare systems. PASTA’s novel combination of features enclosed within a small form factor has the potential to set a new standard of care for respiratory system monitoring in both adult and pediatric medicine.

[0054] With the final goal of building a mature product in the consumer market, the development of PASTA was divided into several phases. This disclosure presents the results, with the objective of simultaneously recording eight channels of audio at a frequency of 48 kHz and depth of 24 bits. The number of channels satisfies a minimum requirement of simultaneously capturing lung sounds at the most common auscultation locations, labeled as L1 , L4, L5, L6, R1 , R4, R5, and R6 as shown in FIG. 2. Usually, auscultation points labeled L2, L3, R2 and R3 (FIG. 2) are ignored since they are close to L1 , L4, R1 , R4, respectively. The audio is captured at a 24-bit depth and a 48 kHz sampling rate, which is the maximum frequency supported by the MP3 audio coding format, also known as the ISO / MPEG Audio Layer III. To conform to this widely used digital audio standard, the PASTA prototype was implemented as an loT device.

[0055] The PASTA technology presented herein contributes to the development of a cutting-edge loT device designed to provide real-time, multi-channel monitoring of different pulmonary sounds, surpassing the capabilities of existing digital stethoscopes. FIG. 3 illustrates examples of different types of acoustic signals. Unlike traditional stethoscopes, which rely on manual, sequential auscultation, PASTA offers simultaneous data acquisition from multiple auscultation sites, enabling spatially localized and high-fidelity sound analysis. Integrated with edge computing, PASTA ensures efficient data processing with low latency for timely diagnoses. Additionally, its modular design, remote connectivity, and support for telemedicine applications enhance its functionality and versatility, effectively bridging a crucial gap in contemporary pulmonary diagnostics.

[0056] Digital stethoscopes transform analog acoustics into electronic signals, enabling post-processing with techniques like filtering and amplification, while also allowing for high- fidelity sound archiving. FIG. 4 shows an example of a popular digital stethoscope, Eko Core 500™. This stethoscope provides up to 40 times acoustic amplification and active noise cancellation. It also detects the patient’s heart rate and displays this information on the top panel of the chestpiece like an electrocardiogram (ECG) device. As another example, the 3M™ Littmann CORE Digital Stethoscope also features sound amplification and ambientDocket: 320903-2560 sound reduction. Its “tunable diaphragm technology” allows user to focus on high-frequency sounds by applying gentle pressure on the chestpiece, and low-frequency sounds with lighter pressure, offering a convenient way to adjust the frequency window. Thinklabs One is another digital stethoscope capable of 100 times amplification. Its users can apply one of the five filtering options, which include cardiac filter, pulmonary filter, wideband filter and others. The ViScope® stethoscope has an embedded display in addition to the check piece, which allows the user to see the waveform over time. Most modern off-the-shelf digital stethoscope products aim to gain a competitive edge through enhanced sound quality and added quality- of-life features. However, they largely retain the traditional form factor of analog stethoscopes. This bulky, single-sensor design hinders integration with remote health monitoring systems and medical loT networks. To address this, PASTA breaks free from the traditional form factor. Furthermore, unlike existing digital stethoscopes that offer mono or stereo sound, the design highlights the device’s ability to perform concurrent auscultation at multiple sites, enhancing the spatial dimension of the acoustic data.

[0057] Studies have demonstrated that digital stethoscopes not only match the quality of conventional devices, but also can provide value by preventing false negative diagnoses during in-person cardiology visits when employed in telecardiology. Telecardiology, as part of telehealth systems, has been shown to reduce hospitalizations by 10 percent in nursing homes. Integrated with healthcare frameworks, these systems can deliver secure and timely diagnoses to the public.

[0058] Early research explored devices leveraging low-cost acoustic sensors, including piezoelectric sensors and MEMS-based microphones. While these sensors, integrated with embedded microcontrollers, demonstrated some potential for pulmonary sound monitoring, they were limited in terms of channel count, sampling rate, and signal fidelity compared to commercial digital stethoscopes. To address these limitations, a silicon-based acoustic sensor array was developed that significantly enhanced spatial resolution and signal quality. However, the higher manufacturing costs of such systems remained a major barrier to widespread adoption when compared to conventional stethoscope technology.

[0059] The potential of integrating open-platform digital stethoscopes into point-of-care systems was explored to provide high-quality healthcare in under-served rural areas. Given the high data throughput and fidelity of digital auscultation, cloud-based machine learning models could be employed for early detection of serious conditions in near-real-time. For instance, joint time-frequency domain analysis of pulmonary sounds and applied k-nearest neighbor classification was utilized to distinguish between wheezes, stridor, and normal breathing. While traditional time-domain and frequency-domain signal processing remains effective for detecting abnormal cardiopulmonary patterns, LungBRN, a machine learning algorithm that employs short-term Fourier transform, wavelet feature extraction, and a biDocket: 320903-2560ResNet deep learning architecture was developed to diagnose respiratory diseases. Moreover, using transfer learning, it has been shown that fine-tuning pre-trained ResNet models outperformed most lung sound classification systems available at the time.

[0060] Despite the demonstrated feasibility of digital stethoscopes in modern cardiology practice, certain limitations remain. These include a lack of clinical validation, patient skepticism, relatively high costs, and limited integration into healthcare systems. Furthermore, most current devices operate on closed platforms, requiring auxiliary equipment for data transmission and lacking the capability for continuous, multi-point body monitoring. To address these challenges, the development of an affordable, open-platform digital stethoscope, PASTA, capable of monitoring adventitious lung sounds as an loT device is proposed. This device can feature multi-channel, simultaneous monitoring capabilities while maintaining a modular and low-profile design for extended use in a variety of healthcare settings.Pulmonary Care Technology

[0061] The proposed pulmonary care technology, PASTA, overcomes the limitations associated with contemporary digital stethoscopes and traditional pulmonary auscultation methods. PASTA can comprise off-the-shelf, low-profile components characterized by their compact size, broad availability, and cost effectiveness. Along with current trends in remote healthcare systems, this technology is designed primarily for remote pulmonary care that facilitates remote monitoring, diagnosis, and prognostication of lung sounds for patients in rural or under-served regions. The PASTA hardware prototype presented herein comprises several components, which are outlined below.• Telemetry sensor array: This array captures acoustic signals from the lungs. Its design can ensure high fidelity sound acquisition, which provides for accurate subsequent analysis.• Audio interface: The interface processes the audio signals from the sensors of the sensor array, converting them into a digital format suitable for analysis. It can be optimized for minimal signal loss and high data integrity.• Data collection via edge computing: An edge computer collects and processes data locally, reducing latency and facilitating real-time data analysis and decision-making. This component can ensure timely interventions in remote settings.

[0062] The integration of these components in the PASTA hardware prototype enables a seamless pipeline for capturing, processing, and analyzing pulmonary sounds with high accuracy. By leveraging real-time data acquisition and edge computing, the system ensures minimal latency while maintaining signal integrity, making it well-suited for remote monitoring applications. This prototype serves as a foundational step towards enhancing pulmonaryDocket: 320903-2560 diagnostics by providing a scalable and efficient solution for continuous and reliable lung sound assessment.

[0063] Telemetry Sensors (Microphone) Array. The core of the PASTA device is its array of acoustic sensors, specifically microphones, tailored for use with patients who may be bedridden or have restricted mobility, affecting common auscultation points. A high-level block diagram of the PASTA audio interface architecture is shown in FIG. 5. The proposed system employs an array of microphones 503 strategically positioned at multiple locations on the patient’s body to capture high-fidelity pulmonary sounds. The acquired signals are subsequently routed through a signal conditioning stage 506, where they undergo amplification and noise filtering to enhance signal quality and ensure optimal input for subsequent analysis 509. The conditioned analog signals are then digitized, facilitating realtime digital processing, analysis, and storage utilizing edge computing capabilities to support advanced diagnostic applications.

[0064] A compact design is important for these microphones to minimize pressure- related injury and enhance patient comfort, which also influences the choice of the audio interface. The I2S (Inter-IC Sound) interface protocol was selected for its streamlined standards and minimal hardware demands, such as the elimination of external noise filtering requirements. The PASTA prototype employed surface-mounted I2S micro-electro- mechanical systems (MEMS) microphones, which are well-suited for integration with other circuit components on a printed circuit board (PCB). FIG. 6 is an image of an I2S MEMS telemetry sensor (microphone) array used in the PASTA prototype. As shown in FIG. 6, the prototype features an array of eight l2S-compliant MEMS microphones mounted on a 6-pin breakout board. These microphones produce electrical signals from the on-board MEMS transducer, which are converted to a digital stream of data through quantization and low- pass filtering. The digital audio signals are then relayed to the audio interface board for further processing.

[0065] Audio Interface. Following the collection of audio signals by the microphone array, these signals are transmitted to an audio interface device. This device integrates audio data from multiple microphone sources in real time and is intended to be non-inferior with respect to the sensitivity and signal quality of conventional systems. To achieve this, a multi-channel audio interface digital signal processing board was used, specifically the MCHStreamer. FIG. 7 shows an audio interface (MCHStreamer multi-channel) board used in the PASTA prototype, showing connections 703 to I2S ports via red and black wires and the central processing role of the main CPU 706, with 8-channel audio data output through the USB port 709 to the right. This pre-programmed microcontroller board is adept at managing complex audio data integration. Equipped with an XMOS Xcore 200 CPU, renowned for its capability in audio DSP and high-efficiency I / O processing suitable for loT applications, thisDocket: 320903-2560 board handles the asynchronous transmission of multi-channel digital audio via USB with exceptional precision. In the PASTA prototype, the MCHStreamer Lite board serves as an important link between the acoustic sensors and the main processing unit, which is an edge computing device. This combination of devices ensures that the audio data is processed efficiently and transmitted to the edge computing device without delay or loss of fidelity. The data from the sensors wired to chip can be processed and synced by the CPU 706 before the data is output 709 for post collection and processing.

[0066] Data Collection and Processing. One objective is to prove the capability of the PASTA device to collect pulmonary data and immediately receive feedback from a healthcare provider. To achieve this, an edge computing device is integrated into PASTA. Furthermore, with the small form factor requirement of PASTA, this edge computing device can be narrowed down to a mid-to-high-range embedded single board computer (SBC). Fortunately, the audio interface selected for the PASTA device is USB audio class 2.0 (UAC2) compliant, so any SBC that supports a modern Linux operating system can meet the technical specification. For the prototype in particular, two types of single-board computers were chosen as a solution to the edge computing (main processing) unit of the PASTA system for collecting pulmonary data as shown in FIGS. 8A and 8B. The first type of edge computer was the Raspberry Pi 4 Model B (shown in FIG. 8A), which is a 56 x 85 mm single-board computer powered by a quad-core ARM Cortex-A72 CPU. The official supported Raspberry Pi OS is a variant of the Debian Linux distribution. The second type of edge computer used in the PASTA device was the NVIDIA Jetson Orin (shown in FIG. 8B), which features an NVIDIA Ampere architecture GPU compatible with CUDA, making it suitable for machine learning needs. The NVIDIA Jetson Orin is a Linux-based (Ubuntu) embedded single-board computer supported with additional drivers, libraries, and tools to customize the hardware. As shown in FIG. 8B, with a slightly larger form factor (100 x 87 mm), the NVIDIA Jetson Orin highlights better computational performance due to its exceptional technical specifications. Wireless communication capabilities (e.g., dual-band Wi-Fi) allows for data communications and remote access.

[0067] The aforementioned parts and their specifications are summarized in the table of FIG. 9. These parameters satisfy the requirements for PASTA. Consequently, the listed set of components were chosen as a mature design to conduct thorough experiments. FIG. 10 demonstrates the flow of the audio data from the lowest level (operating at high bandwidth) to the highest level (operating at low bandwidth). High-bandwidth electrical signals collected from the microphone sensor array 1003 are sent to the audio interface 1006, which provides a standardized serial connection for seamless integration with various edge computing devices 1009. Given the raw audio data and depending on the need, there is a wide range of software applications that can be conducted at a high level (operating at low bandwidth). ForDocket: 320903-2560 a tele-health system, an automatic diagnostic system can be supported by a machine learning algorithm 1012 that extracts and classifies hidden features in the audio data. Alternatively, for an experienced healthcare provider who has been trained on traditional stethoscope auscultation techniques, a 3D spatial audio format can support the identification of respiratory health problems more accurately using enhanced audio 1015 an / or data visualization 1018. Within the framework of developing the proposed pulmonary device, programs to validate the prototype were developed from the perspective of an loT device.Validation Methods

[0068] The validation of the PASTA prototype addresses two primary objectives: (1 ) to ascertain if the prototype’s capability to capture pulmonary sounds is comparable to existing devices in the market, and (2) to verify whether it fulfills the desired design specifications of the real-time, multi-channel, low-cost, modular, and open-platform. A comprehensive validation strategy was implemented to achieve these objectives, including on-site data collection and off-site data analysis. The prototype was evaluated alongside two commercially available digital stethoscopes. A detailed specification comparison of the evaluated devices was also conducted. FIG. 11 depicts the three stages of this validation process, each of which is elaborated below.

[0069] On-Site Acoustic Data Collection. The PASTA prototype and two conventional digital stethoscopes were set up in a simulated inpatient hospital care setting at the Jump Trading Simulation and Education Center to collect data samples of lungs sound. For the commercial stethoscopes, the acquired sound samples were exported by following their respective guidelines. For PASTA, a program was implemented to run on the edge computing device, recording eight channels of audio data over ten seconds, and storing them in the local storage for analysis. This step provided concrete data across all three devices for the next two analysis steps to reference.

[0070] Off-Site Acoustic Data Analysis. Audio files stored by the data collection program were post-processed and analyzed numerically. The initial stage in the postprocessing sequence involves normalizing the raw acoustic data. This data, represented as scaled voltage levels captured by the microphones, can introduce biases in subsequent analyses due to variable scaling. Mathematically, the normalization process is pursued using x - — , - 6 [-1,11, where x denotes the raw data, ii represents the mean of the data set, max(| - |) and x is the normalized data constrained within the range [-1 , 1], This normalization not only standardizes the acoustic range of the data but also mitigates the likelihood of high DC offsets by centering the data around a near-zero mean.

[0071] ACOUSTIC ANALYSIS. For the subsequent stage of acoustic analysis, a frequency spectrum analysis was conducted employing the Discrete Fourier TransformDocket: 320903-2560(DFT), which was adapted for digital signals as described by “Numerical Recipes 3rd Edition: The Art of Scientific Computing" by W.H. Press et al. (Cambridge Univ. Press, 2007). The DFT extracts the frequency components from time-based data, utilized for understanding the audio signals’ characteristics. Mathematically, the transformation can be expressed as:where xfn) represents the discrete time-based data array,denotes the Fourier coefficients, N is the length of the data, and Wn- e~i2n Nsignifies the fundamental DFT kernel. Given that x n) are real numbers in the audio data, negative frequencies in X k) become redundant and are typically omitted in the spectral analysis. The amplitude spectrum ||X(fc)|| of pulmonary sounds captured by PASTA and other devices were analyzed. The amplitude spectrum provides important insights into the dominant frequency components of the recorded signals, which are directly related to PASTA’s sensitivity in terms of detecting pulmonary abnormalities.

[0072] COSINE SIMILARITY FOR SPECTRAL COMPARISON. To compare the frequency spectra of audio signals captured by different devices, cosine similarity was applied. This metric evaluates the similarity between two vectors, which in the context are the amplitude spectra of different devices. To quantify spectral similarities between devices, the cosine similarity metric can be computed using:where vAand vBrepresent the frequency spectra vectors from devices A and B, respectively. A cosine similarity close to 1 indicates a high resemblance between the spectra.

[0073] To ensure a fair comparison across devices with potentially different sampling frequencies, the frequency spectra was standardized to a common set of frequencies. Cosine similarity was then computed between the spectra to quantitatively measure the resemblance of the frequency content captured by each device. The detailed process is outlined in the Algorithm of FIG. 12, which ensures that the vectors are comparable.

[0074] This standardized approach ensures the robustness of the analysis and allows for an accurate comparison of device performance regarding the capture and reproduction of audio signals. The results confirm the efficacy of PASTA in replicating and analyzing pulmonary sounds.Experiments and Results

[0075] A series of experiments of the PASTA device were conducted in a simulated inpatient environment at the Jump Trading Simulation and Education Center, and the resultsDocket: 320903-2560 are presented using methods explained in the previous sections. FIG. 13A shows the PASTA device mounted on a Laerdal SimMan 3G simulation manikin, a standard setting for patient care simulation. All eight acoustic sensors were attached at different points as shown in FIG. 2. The data collected by the PASTA device was observed at a remote station as shown in FIG. 13B. The patient model, Laerdal SimMan 3G, can be programmed to generate a variety of lung diagnostic sounds across various auscultation points while expanding the chest to mimic realistic breathing activities of a human being. In particular, the sounds used to collect were classified as (1) normal breathing, (2) coarse crackles, (3) fine crackles, (4) pleural rub, (5) pneumonia, (6) gurgling rhonchi, (7) rhonchi, (8) stridor, and (9) wheezes. With the exception of normal breathing, all other sound profiles correspond to sounds produced by abnormal respiratory conditions. The sounds generated by the manikin were collected with four different configurations of hardware as shown in FIG. 14. The first two configurations shown in FIG. 14 are the proposed PASTA prototype with the NVIDIA Jetson Orin (Config A, which is called PASTA 1) and the Raspberry Pi 4 (Config B, which is called PASTA 2) as data collection devices. The last two configurations are commercially available electronic stethoscopes that were selected as competitors: namely the ThinkLabs One (Config C), and Eko CORE 500 (Config D). The remote station at the end of FIG. 14 is a laptop computer used for controlling the devices and storing the sound samples.

[0076] After normalization and FFT transform as previously explained, FIG. 15A shows a frequency spectrum between 0 to 5000 Hz of one channel from each configuration and each sound profile. FIG. 15A illustrates examples of experimental results collected from manikin generating nine different sound profiles on the PASTA using config A (first column) and config B (second column), ThinkLabs One (third column) and Eko Core 500 (fourth column). Data is visualized in log-scale frequency spectrum over 0 - 5000 Hz. The frequency domain data can be compared with frequency-adjusted cosine similarity method using the Algorithm of FIG. 12 and Eq. (2), resulting in the numbers shown in the table of FIG. 15B. Note that PASTA1 and PASTA2 represent two versions of the PASTA prototype, each utilizing a different edge computing platform: Jetson Orin and Raspberry Pi, respectively. Comparing the left two columns, which are the proof-of-concept PASTA device configurations, the frequency profile between configurations on the same sound profile (such as plot (u) vs plot (v) of FIG. 15A, or plot (ac) vs plot (ad) of FIG. 15A) shows high similarity. This result is expected because these configurations use the same set of microphone devices, as shown with 0.9451 cosine similarity between the two. Comparing the PASTA prototype with the ThinkLabs stethoscope, only a few sound profiles demonstrate similar signal spectrograms (gurgling rhonchi, stridor, for example); however, the two devices achieve a similarity score of 0.759 - 0.8032. This has validated the first question regarding whether or not the prototype is comparable with other competitors in capturing pulmonary sounds. The signalsDocket: 320903-2560 obtained from the second commercial digital stethoscope competitor, the Eko Core 500, differ greatly from the other devices that were tested, as depicted by its spectrogram. It achieves similarity scores lower than 0.4 in all comparisons. It is believed this was caused by data loss incurred during the export process, which is discussed further later in the paper.

[0077] Furthermore, FIG. 16 depicts time-based normalized audio across all eight channels. FIG. 16 illustrates examples of two instances of experimental data vs time collected from manikin generating normal breathing sound (left) and pneumonia sound (right). The left column shows one set of data collected with the NVIDIA Jetson Orin, where two inhalation phases were captured at about 1 second and 6 seconds during the recording. The plots on the right column, as another set of data from Raspberry Pi 4, pick up faster breathing, but the waveform is still synchronized. This verifies that, to support real-time 8- channel 48 kHz data throughput, the specification of the data collection machine can go as low as a mid-tier SBC.Technical Analysis and Impact of PASTA

[0078] The competitive features of PASTA, relative to existing commercially available digital stethoscopes, are presented in the table of FIG. 17. This comparison underscores the technical and economic advantages of the proposed PASTA technology. The design of PASTA emphasizes affordability and flexibility, supporting a wide range of machines without inflating the estimated price. In contrast, Eko Core 500 maintains a highly proprietary platform, limiting interface options to proprietary earpieces and mobile apps. Despite the user-friendly interfaces of these apps, rated highly on digital distribution services, their operational processes involve cumbersome steps that raise concerns regarding practicality and efficiency.

[0079] Conversely, ThinkLabs One offers a simpler, more direct connection through a 3.5 mm AUX port, facilitating both real-time auscultation and straightforward audio recording. PASTA, leveraging its USB audio interface, avoids the limitations of a single AUX port, which would otherwise restrict multi-channel audio capabilities. This feature exemplifies its adaptability, accommodating up to eight audio channels and enabling easier integration with various data collection devices and applications, such as VLC (https: / / www.videolan.org / ) or custom in-house solutions.

[0080] The enhancements provided by PASTA extend beyond feature parity with existing devices; it excels in delivering real-time, multi-channel, cost-effective, modular, and open-platform capabilities. The expected impact of such a device is profound across several domains. In diagnostic care, the ability to monitor multiple locations simultaneously promises more accurate identification of focal lung findings (distinguishing them from the more common diffuse processes, which are often viral), while its cost-effectiveness enhances accessibility. Another significant potential impact of this technology is the ability to monitorDocket: 320903-2560 multiple patients in real time. This feature can lead to reduced treatment delays as the current standard of care for inpatient auscultation requires a healthcare provider to physically visit the patient several times throughout the day. Due to the unpredictable and sometimes precipitous nature of many lung pathologies, these auscultations may not be frequent and timely enough to prevent clinical deterioration. A system such as PASTA that enables continuous monitoring and detection of adventitious lung sounds has the potential to decrease the time needed to detect and diagnose these sounds. This technological advancement could set a new standard in pulmonary care, fostering a new era in pulmonology and inpatient medicine that harnesses cutting-edge technology to offer superior patient care.

[0081] Moreover, the data collected by PASTA are invaluable for medical research, particularly for applications in machine learning and Al-based diagnostics, improving both the timeliness and accuracy of healthcare services. Its configuration as an loT device also highlights its potential in real-world applications, simplifying data collection for consumer use across edge computing, cloud, and mobile platforms.

[0082] The PASTA prototype can be integrated into an enclosure that meets medical standards, ensuring safety and reliability for clinical use. FIG. 18 graphically illustrates an example of a PASTA device encased in a housing including a carrying handle and power cord for connection to a power source to facilitate portability and clinical use. In some implementations, the PASTA device can be a wearable device with a rechargeable power supply. A touchscreen interface can facilitate user access and interaction with the device. In the example of FIG. 18, the sensor array includes 8 microphones for placement on the anterior and posterior chest cavity walls of the user. The microphones can be connected to the PASTA device through a cable hub and trunk cable (e.g., PVC cable), which can be detachably connected to the processing unit of the PASTA via a cable connector. The microphones can be encased in silicone to avoid direct contact with the skin and attached to the user using microphone bandages (e.g., acrylic adhesive with fabric backing) or other appropriate attachment means.

[0083] During use in a clinical setting, a nurse practitioner can deploy the PASTA device at, e.g., a patient's bedside. The power cable of the PASTA device can be plugged into a wall outlet if needed. The PASTA device can be turned on and connected to the wireless communication network (e.g., the hospital WAN), which can allow the unit to pair with the patient’s electronic medical record. The array of acoustic pulmonary sensors (microphone assembly) can be coupled to the PASTA device and a test can be performed to ensure that all microphones are working properly. The nurse practitioner can then attach the sensors at the appropriate locations on the front and back of the patient using, e.g., adhesive patchesDocket: 320903-2560 or microphone bandages. Another test can be performed to ensure proper recording. Once attached, data can be gathered in real time or near-real time.

[0084] FIG. 19 illustrates a flowchart of an example of a high-level process flow from the collection of data to form a sound sample or data package, to sample comparison that can be done prior to or after the off-site analysis done by an off-site processor prior to diagnosis. Referring to FIG. 19, the diagram illustrates the process of data collection when the system is attached to a patient 1910 comprising: turning on the microphone array 1910A, recording for a predetermined amount of time 191 OB, saving the record in memory 1910C, and writing the data to file 1910D. This process makes the sound sample or data package 1920 for analysis or diagnosis by comparing the sound to a baseline sound comparison 1940. The sound sample 1920 can also be processed by the off-site data analysis 1930 comprising: initializing the interface 1930A, conducting a Fast Fourier transformation (FFT) 1930B, and then conducting a cross comparison with cosine sine 1930C. The data can also be compared to the baseline sound comparison 1940 after the transformation of the sound sample through off-site data analysis 1930 to perform a diagnostic 1950.

[0085] It should be emphasized that the above-described embodiments of the present disclosure are merely possible examples of implementations set forth for a clear understanding of the principles of the disclosure. Many variations and modifications may be made to the above-described embodiment(s) without departing substantially from the spirit and principles of the disclosure. All such modifications and variations are intended to be included herein within the scope of this disclosure and protected by the following claims.

[0086] The term "substantially" is meant to permit deviations from the descriptive term that don't negatively impact the intended purpose. Descriptive terms are implicitly understood to be modified by the word substantially, even if the term is not explicitly modified by the word substantially.

[0087] It should be noted that ratios, concentrations, amounts, and other numerical data may be expressed herein in a range format. It is to be understood that such a range format is used for convenience and brevity, and thus, should be interpreted in a flexible manner to include not only the numerical values explicitly recited as the limits of the range, but also to include all the individual numerical values or sub-ranges encompassed within that range as if each numerical value and sub-range is explicitly recited. To illustrate, a concentration range of “about 0.1% to about 5%” should be interpreted to include not only the explicitly recited concentration of about 0.1 wt% to about 5 wt%, but also include individual concentrations (e.g., 1%, 2%, 3%, and 4%) and the sub-ranges (e.g., 0.5%, 1.1%, 2.2%, 3.3%, and 4.4%) within the indicated range. The term “about” can include traditional rounding according to significant figures of numerical values. In addition, the phrase “about ‘x’ to ‘y’” includes “about ‘x’ to about y

Claims

Docket: 320903-2560CLAIMSTherefore, at least the following is claimed:1 . A pulmonary acoustic sensor telemetry array (PASTA) system for measuring audio characteristics of a body, comprising: an array of acoustic sensors; an edge computing device; and an audio interface configured to receive asynchronous audio signals from the array of acoustic sensors and provide synchronized audio signal data to the edge computing device in near-real time for data collection and processing.

2. The PASTA system of claim 1 , wherein the array of acoustic sensors comprises a plurality of microphones configured to be attached to front and back surfaces of the body.

3. The PASTA system of claim 2, wherein the plurality of microphones are attached at defined auscultation points on the body.

4. The PASTA system of claim 2, wherein the array of acoustic sensors comprises eight microphones with four attached to the front surface and four attached to the back surface.

5. The PASTA system of any one of claims 1 -4, wherein the audio signals from the array of acoustic sensors are analog signals and the audio interface converts the analog signals to digital signals for processing by the edge computing device.

6. The PASTA system of claim 5, wherein the analog signals are converted to 1 -bit digital data through the exploitation of pulse density modulation.

7. The PASTA system of any of claims 5 and 6, wherein the analog signals are filtered prior to digital conversion.

8. The PASTA system of any one of claims 1 -7, wherein the edge computing device comprises a processor and memory that stores executable instructions that can be executed by the processor.Docket: 320903-25609. The PASTA system of any one of claims 1 -8, wherein the edge computing device comprises a wireless communications interface configured to link to a wireless network, where the edge computing device communicates the audio signal data to a remote computing device via the wireless network.

10. The PASTA system of claim 9, wherein the edge computing device compiles the audio signal data from the audio interface and continuously or non-continuously processes the data for communication or diagnosis.11 . The PASTA system of any of claims 9 and 10, wherein the edge computing device communicates the audio signal data to the remote computing device for storage of the audio signal data.

12. The PASTA system of any of claims 9 and 10, wherein the edge computing device communicates the audio signal data to the remote computing device for remote processing of the audio signal data.

13. The PASTA system of claim 12, wherein the remote processing comprises comparing a performance of an output sample to a baseline performance sample to determine a difference in user health.

14. The PASTA system of claim 13, wherein the baseline performance sample is received from an input sample.

15. The PASTA system of claim 12, wherein the audio signal data is stored and analyzed over time for abnormalities, deviations, or both from a baseline performance.

16. The PASTA system of any of claims 9 and 10, wherein the edge computing device communicates the audio signal data to the remote computing device for access by an authorized user.

17. The PASTA system of claim 16, wherein the authorized user is a physician or nurse practitioner, and access is in near-real time.

18. The PASTA system of claim 17, wherein the authorized user can manipulate the audio signal data.Docket: 320903-256019. The PASTA system of any of claims 1-18, wherein the PASTA system is wearable.

20. The PASTA system of any of claims 1-19, wherein each sensor of the array of acoustic sensors is encased in silicon.

Images