Method, system and apparatus for identifying a potential pathology using biological sounds

Abstract

A phonocardiogram apparatus and related method for diagnosis of cardiac pathologies through advanced sound recording and analysis. This apparatus utilizes a grid of microphones embedded in a fabric and positioned over the patient's chest and back to capture heart, lung, and neck sounds from multiple locations. An artificial intelligence (AI) based algorithm is employed to determine the most relevant microphones based on the patient's unique anatomy, thereby optimizing sound collection for diagnostic purposes. The collected sounds undergo a series of pre-processing and machine learning analyses to diagnose cardiac abnormalities such as aortic stenosis, murmurs, and extra systoles, create cardiac auscultatory signatures and allow longitudinal monitoring by clinicians or patients themselves for comparison over time.

Description

METHOD, SYSTEM AND APPARATUS FOR IDENTIFYING A POTENTIAL PATHOLOGY USING BIOLOGICAL SOUNDSCROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims priority of U.S. Provisional Patent Application No. 63 / 575,980 filed on April 8, 2024, the content of which is hereby incorporated by reference.TECHNICAL FIELD

[0002] This disclosure relates to the field of monitoring and diagnosis. More particularly, the disclosure relates to a method and apparatus for identifying a health condition such as cardiac abnormalities or other physical conditions of a subject using advanced sound recording and analysis.BACKGROUND

[0003] In the vast and ever-evolving field of medical care, practitioners have traditionally relied on a wide array of specialized instruments that serve as invaluable tools for assessing and evaluating various biological characteristics of their patients. Each of these instruments has been meticulously designed and tailored to target specific biometrics or classes of biometrics, allowing for a more focused and specialized approach to patient care. However, while this approach has proven to be immensely beneficial in many ways, it is not without its challenges and limitations.

[0004] One of the key challenges that arise from the reliance on multiple instruments for a comprehensive patient assessment is the increased complexity this introduces. Medical professionals are required to familiarize themselves with the intricacies and specificities of each instrument, which can often result in a steep learning curve. Moreover, the diverse nature of these instruments means that practitioners must possess a broad knowledge base and skill set in order to use them effectively, further contributing to the complexity of the process.

[0005] In addition to the complexity, there is also the issue of cost. The acquisition and maintenance of a diverse range of specialized instruments can be a substantial financialburden for medical care practitioners and healthcare facilities. Furthermore, the need for multiple instruments also leads to logistical challenges related to portability and data interoperability. Carrying and managing an assortment of instruments becomes cumbersome and impractical, hindering the ease and efficiency of patient care.

[0006] Moreover, despite their specialized nature, some of these conventional instruments exhibit limited functionality. For example, the tried-and-true stethoscope has long been a staple in medical practice for auscultation, allowing practitioners to listen to audible body sounds, such as those produced by the heart, lungs, and gastrointestinal system. However, even this revered instrument has its limitations. Conventional stethoscopes fail to capture crucial information encoded in low-frequency, low-amplitude inaudible signals that could provide vital insights into cardiac, respiratory, and digestive health. This untapped potential represents a significant gap in our understanding and assessment of patient wellbeing.

[0007] The analysis of these low-frequency, low-amplitude signals is an area that necessitates further exploration. The medical community is still striving to unravel the mysteries and potential benefits hidden within these signals. Regrettably, the current technological landscape does not adequately address this challenge, leaving healthcare professionals with a diagnostic picture that is incomplete and lacking in comprehensiveness.

[0008] Moreover, auscultation through standard methods is typically a snapshot in time in terms of what is normal for a given patient. Repeated auscultation over time may not capture subtle but important changes in a given patient’s ongoing health and fitness. Accordingly, the limitations of conventional stethoscopes extend to their need for periodic and momentary placement on the patient's skin for optimal signal detection. While this requirement has proven effective in many instances, there are situations where circumstances make it necessary to detect signals with minimally detectable variances over time. In such cases, the conventional stethoscope's effectiveness is compromised, highlighting the need for alternative solutions that can bypass these limitations and provide accurate insights regardless of minimally perceptible signal variations. Moreover, traditional phonocardiogram devices are limited by their reliance on single-point sound collection, which can miss nuances in heart sounds that vary spatially across the chest. This limitation necessitates a solution capable of capturing a broader acoustic profde, offering a more detailed and accurate diagnostic tool.

[0009] Given the myriad challenges and limitations associated with the current state of medical instruments, it becomes clear that a new and improved sensor platform is urgently needed. Such a platform would revolutionize the field of medical care by offering comprehensive characterizations of one or more health conditions such as cardiac abnormalities as well as other physical aspects of a subject. By overcoming the limitations of existing instruments, this innovative platform would provide medical practitioners with a powerful and versatile toolset that ensures more accurate and holistic assessments of their patients' well-being. This would ultimately lead to improved diagnoses, better treatment plans, and enhanced overall patient care. The development of such a diagnostic innovation is an exciting prospect that holds tremendous potential for the future of healthcare.SUMMARY

[0010] It is an object of the present technology to ameliorate at least some of the inconveniences present in the prior art. One or more embodiments of the present technology may provide and / or broaden the scope of approaches to and / or methods of achieving the aims and objects of the present technology.

[0011] The present disclosure provides a phonocardiogram apparatus designed to enhance the diagnosis of cardiac pathologies through advanced sound recording and analysis. This apparatus utilizes a grid of microphones embedded in a fabric and positioned over the patient's chest and back to capture heart and lung sounds from multiple locations. Alternative embodiments may include a similar separate grid for location and for similar sound collection on a patient’s neck. An artificial intelligence (Al) based algorithm is employed to determine the optimal microphone selection based on the patient's unique anatomy, thereby optimizing sound collection for diagnostic purposes. The collected sounds undergo a series of pre-processing and machine learning analyses to identify cardiac and lung abnormalities such as aortic stenosis, murmurs, extra systoles and pulmonary edema.

[0012] In the context of the present specification, a “server” is a computer program that is running on appropriate hardware and is capable of receiving requests (e.g., from computing devices) over a network (e.g., a communication network), and carrying out those requests, or causing those requests to be carried out. The hardware may be one physical computer or one physical computer system, but neither is required to be the case with respect to the present technology. In the present context, the use of the expression a “server” is notintended to mean that every task (e.g., received instructions or requests) or any particular task will have been received, carried out, or caused to be carried out, by the same server (i.e., the same software and / or hardware); it is intended to mean that any number of software elements or hardware devices may be involved in receiving / sending, carrying out or causing to be carried out any task or request, or the consequences of any task or request; and all of this software and hardware may be one server or multiple servers, both of which are included within the expressions “at least one server” and “a server”.

[0013] In the context of the present specification, “computing device” or “processor” is any computing apparatus or computer hardware that is capable of running software appropriate to the relevant task at hand. Thus, some (non-limiting) examples of computing devices include general purpose personal computers (desktops, laptops, netbooks, etc.), mobile computing devices, smartphones, and tablets, and network equipment such as routers, switches, and gateways. It should be noted that a computing device in the present context is not precluded from acting as a server to other computing devices. The use of the expression “a computing device” does not preclude multiple computing devices being used in receiving / sending, carrying out or causing to be carried out any task or request, or the consequences of any task or request, or steps of any method described herein. In the context of the present specification, a “client device” refers to any of a range of end-user client computing devices, associated with a user, such as personal computers, tablets, smartphones, and the like.

[0014] In the context of the present specification, the expression "computer readable storage medium" (also referred to as "storage medium” and “storage”) is intended to include non-transitory media of any nature and kind whatsoever, including without limitation RAM, ROM, disks (CD-ROMs, DVDs, floppy disks, hard drivers, etc.), USB keys, solid statedrives, tape drives, etc. A plurality of components may be combined to form the computer information storage media, including two or more media components of a same type and / or two or more media components of different types.

[0015] In the context of the present specification, a "database" or “storage” is any structured collection of data, irrespective of its particular structure, the database management software, or the computer hardware on which the data is stored, implemented or otherwise rendered available for use. A database may reside on the same hardware as theprocess that stores or makes use of the information stored in the database or it may reside on separate hardware, such as a dedicated server or plurality of servers.

[0016] In the context of the present specification, the expression “communication network” is intended to include a telecommunications network such as a computer network, the Internet, a telephone network, a Telex network, a TCP / IP data network (e.g., a WAN network, a LAN network, etc.), and the like. The term “communication network” includes a wired network or direct-wired connection, and wireless media such as acoustic, radio frequency (RF), infrared and other wireless media, as well as combinations of any of the above.

[0017] In the context of the present specification, the words “first”, “second”, “third”, etc. have been used as adjectives only for the purpose of allowing for the distinction between the nouns that they modify from one another, and not to describe any particular relationship between those nouns. Thus, for example, it should be understood that the use of the terms “server” and “third server” is not intended to imply any particular order, type, chronology, hierarchy or ranking (for example) of / between the server, nor is their use (by itself) intended to imply that any “second server” must necessarily exist in any given situation. Further, as is discussed herein in other contexts, reference to a “first” element and a “second” element does not preclude the two elements from being the same actual real-world element. Thus, for example, in some instances, a “first” server and a “second” server may be the same software and / or hardware, in other cases they may be different software and / or hardware.

[0018] In one or more embodiments, the present disclosure provides for telemonitoring of patients with known cardiovascular or lung disease at a distance. This is particularly advantageous in situations where patients who live in remote areas far from major centers are diagnosed with valvular heart disease, heart failure, obstructive lung disease or are on a ventricular assist device (LVAD).

[0019] In one or more embodiments, the present disclosure enables a global health program for cardiovascular diagnostics. This provides access to cardiovascular diagnostics globally with a safe, inexpensive and non-invasive point-of-care test. This is particularly advantageous for populations in areas with limited access to medical doctors or cardiovascular specialists. This includes small clinics in developing countries. An advantageous clinical focus for the present inventive apparatus is to diagnose treatablecardiac conditions such as valvular disease that may be intervened upon percutaneously and heart failure that may be treated medically.

[0020] In one or more embodiments, the present disclosure provides for precision cardiovascular care which enables precise and reproducible physical exam findings at the time of a primary care encounter with a physician or nurse. This is particularly advantageous for patients going for an evaluation with a primary care physician or advanced nurse practitioner and who are the gatekeepers for access to cardiovascular care. This applies to patients who need longitudinal monitoring as well because the inventive apparatus enables comparison from year-to-year to monitor for progression of cardiac abnormalities and the like.

[0021] In one or more embodiments, the present disclosure provides for predictive aspects such as, but not limited to, creating cardiovascular personalized signatures by simultaneously using multiple auscultatory locations (i.e., on the torso and neck) that predict cardiovascular outcomes. This is particularly advantageous, for example, for insurance companies that want quick point-of-care tests to predict cardiovascular health before issuing life insurance.

[0022] According to a first broad aspect, there is provided a phonocardiogram apparatus for determining cardiac pathologies, the apparatus comprising: a wearable portion; a plurality of domaphones formed within the wearable portion; a plurality of sensing units, each sensing unit affixed within a corresponding one of the plurality of domaphones; and a digital processing unit operably connected to the plurality of sensing units, the digital processing unit including a non-transitory storage medium storing instructions thereon, and at least one processor operatively connected to the non-transitory storage medium, the processor, upon executing the instructions, being configured for: filtering an output of the sensing units to produce a preconditioned signal, boosting the preconditioned signal to produce a conditioned signal, inferring at least one cardiac pathology from the conditioned signal, and displaying the at least one cardiac pathology to a user.

[0023] In one or more embodiments, the displaying of at least one cardiac pathology to a user may occur near to the wearable portion or remote from the wearable portion.

[0024] In one or more embodiments, the displaying of at least one cardiac pathology to a user occurs in real-time relative to the output of the sensing units.

[0025] In one or more embodiments, the displaying of at least one cardiac pathology to a user occurs in a delayed manner relative to the output of the sensing units.

[0026] In one or more embodiments, each of the plurality of sensing units are removably affixed to each corresponding one of the plurality of domaphones.

[0027] In one or more embodiments, the at least one cardiac pathology displayed to the user is in the form of an auscultation signature.

[0028] In one or more embodiments, the inferring of the at least one cardiac pathology is accomplished by comparison to a labeled dataset of heart sounds.

[0029] In one or more embodiments, the at least one cardiac pathology displayed to the user includes a confidence score corresponding to the at least one cardiac pathology.

[0030] In one or more embodiments, the at least one cardiac pathology displayed to the user includes a visualization of heart sound origin points most indicative of the at least one cardiac pathology.

[0031] In one or more embodiments, the wearable portion is fabricated from a stretchable fabric conformable to a torso of a wearer of the apparatus.

[0032] In one or more embodiments, the plurality of domaphones are arranged in a grid pattern to optimize placement on the torso of the wearer of the apparatus.

[0033] In one or more embodiments, the plurality of domaphones each further includes a secondary microphone located therein, each secondary microphone used by the digital processing unit for adaptive noise cancellation.

[0034] In one or more embodiments, the plurality of sensing units are wirelessly connected to the digital processing unit.

[0035] In one or more embodiments, the plurality of sensing units are hardwired to the digital processing unit.

[0036] According to a second broad aspect, there is provided a computer-implemented method for determining cardiac pathologies, the method being executed by at least one processor operatively connected to a non-transitory storage medium, the computer-implemented method comprising: filtering an output of the sensing units to produce a preconditioned signal; boosting the preconditioned signal to produce a conditioned signal; inferring at least one cardiac pathology from the conditioned signal, and displaying the at least one cardiac pathology to a user.

[0037] In one or more embodiments, the filtering includes sound pre-processing using adaptive noise cancellation to eliminate ambient and physiological noise obscuring cardiac sounds.

[0038] In one or more embodiments, the filtering includes microphone salience detection to determine the relative importance of each microphone provided within a plurality of sensing units.

[0039] In one or more embodiments, the boosting includes automated biasing and selection using gradient boosting machines to dynamically adjust the input weights of selected ones of each microphone.

[0040] In one or more embodiments, the inferring includes at least one cardiac pathology using machine learning analysis to decompose heart sound signals, capturing both frequency and location information of cardiac events, and recognizing patterns within the heart sound signals to identify specific sound patterns associated with normal and pathological conditions.

[0041] According to a third broad aspect, there is provided a phonocardiogram apparatus for determining cardiac pathologies, the apparatus comprising: a wearable portion configured to conform to a user’s chest and back; a grid of domaphones embedded within the wearable portion, each domaphone housing a sensing unit; and a digital processing unit operatively connected to the sensing units, the digital processing unit configured to: filter outputs from the sensing units to produce preconditioned signals, boost the preconditioned signals to generate conditioned signals, infer at least one cardiac pathology from the conditioned signals using an Al-based algorithm, and display inferred cardiac pathologies to an end-user, including a visualization of heart sound origin points and a confidence score for each pathology; wherein the wearable portion is made from a stretchable fabric, and the domaphones are arranged to ensure optimal placement for sound capture across the user’s chest and back, enabling comprehensive and nuanced cardiac and pulmonary assessment.

[0042] According to a fourth broad aspect, there is provided a method for determining cardiac pathologies using a phonocardiogram apparatus, the method comprising: fdtering outputs from sensing units embedded within a grid of domaphones wearable by a user to remove ambient noise and enhance signal clarity; employing adaptive noise cancellation and microphone salience detection to refine sound signal processing; using an Al-based algorithm to analyze preconditioned sound signals, infer cardiac pathologies, and generate a diagnostic output; and displaying the diagnostic output near or remotely from the user in real-time or delayed fashion, including visualizations of heart sound origin points on the user’s chest and back, and providing a longitudinal monitoring capability for comparison over time; wherein the method leverages gradient boosting machines and machine learning analyses, including discrete wavelet transforms and convolutional neural networks, for precise pathology identification.

[0043] According to another broad aspect, there is provided an acoustic apparatus for operating a sensor about a skin surface, the acoustic apparatus comprising: a first body made at least partially of a first material, having an inner surface defining a dome-shaped cavity opening at least at a lower end, and comprising sensor-mounting means operably connected to the dome-shaped cavity adapted to mount a first sensor, and a second body made at least partially of a second material, having a main hollow wall adapted to securely engage a portion of the first body, the main hollow wall having an upper end defining a body-opening and an opposite and connected skin-facing end adapted to engage the skin surface and defining a skin-facing opening, wherein the second material of the second body comprises a resilient material.

[0044] In one or more embodiments, the first material has a greater stiffness than the second material.

[0045] In one or more embodiments, the first body and the second body are configurable into a sealed engagement at an interface therebetween.

[0046] In one or more embodiments, the first body is detachably securable to the second body.

[0047] In one or more embodiments, the first body defines a body-retaining groove disposed along an outer perimeter thereof, and wherein the second body comprises a body-retaining rib complementarity configured to securely engage the body-retaining groove of the first body in a snap-fit connection.

[0048] In one or more embodiments, the sensor mounting means comprises a sensor mounting -aperture extending from the dome-shaped cavity to an upper end of the first body, thus further opening the dome-shaped cavity.

[0049] In one or more embodiments, the first body further comprises a handle portion having a neck and a knob, the neck extending from a first main wall of the first body, and the knob further extending from the neck to define the upper end of the first body, and wherein the sensor-mounting aperture extends in the neck and knob of the handle portion.

[0050] In one or more embodiments, the sensor mounting -aperture is configured to accept an acoustic sensor configured to capture a phonocardiogram signal.

[0051] In one or more embodiments, the dome-shaped cavity of the first body has an internal radius between 5 mm and 45 mm.

[0052] In one or more embodiments, the dome-shaped cavity of the first body has an internal radius between 11 mm and 31 mm.

[0053] In one or more embodiments, dome-shaped cavity has a nominal spline radius of 17.25 mm and a height of 10.5 mm.

[0054] In one or more embodiments, the inner surface of the first body is adapted such that the dome-shaped cavity comprises a paraboloid dome cavity.

[0055] In one or more embodiments, the first material comprises an elastic 50A resin.

[0056] In one or more embodiments, the second body is adapted to operably mount a garment positionable in proximity to the skin surface.

[0057] In one or more embodiments, the skin-facing end of the main hollow wall of the second body forms a planar ring surface to interface with the skin surface.

[0058] In one or more embodiments, the main hollow wall has an annular resilient portion (450) surrounding the body-opening and defining the upper end of the second body, wherein the annular resilient portion is configured to resiliently deform when the first bodyis securely engaging the second body, thus resisting an upward disengagement of the first body.

[0059] In one or more embodiments, the second material comprises a platinum -cured silicone rubber material.

[0060] In one or more embodiments, the first body and the second body are monolithic, thus enabling the secured engagement therebetween.

[0061] According to a further broad aspect, there is provided a garment for capturing a series of signals from the body, the garment comprising: a fabric portion comprising a plurality of socket openings, and configured for securely receiving a socket portion of an acoustic apparatus for operating a sensor about a skin surface.

[0062] In one or more embodiments, the garment is configured such that the plurality of sockets openings are positioned proximate to a plurality of regions of the body to capture the body data.

[0063] According to still another broad aspect, there is provided a method for identifying a potential pathology, the method being executed by a processor, the method comprising: receiving a plurality of biological sound signals each detected at a respective position on a body of a subject; selecting given signals amongst the plurality of biological sound signals; identifying a potential pathology based on the given signals; and outputting the potential pathology.

[0064] In one or more embodiments, the step of receiving the plurality of biological sound signals comprises receiving a plurality of heart sound signals and said selecting the given signals comprising selecting given ones of the heart sound signals amongst the plurality of heart sound signals.

[0065] In one or more embodiments, the step of selecting the given ones of the heart sound signals comprises: assigning an initial weight value to each one of the plurality of heart sound signals, the initial weight value being indicative of a relative importance of each one of the plurality of heart sound signals; adjusting the initial weight value to obtain a final weight value for each one of the plurality of heart sound signals; and identifying the given ones of the heart sound signals based on the final weight value.

[0066] In one or more embodiments, the step of assigning the initial weight value is performed using a K-means clustering method.

[0067] In one or more embodiments, the step of adjusting the initial weight value is performed using a gradient boosting method.

[0068] In one or more embodiments, the step of identifying the potential pathology comprises: extracting features from the given ones of the heart sound signals; and analyzing the features to identify the pathology.

[0069] In one or more embodiments, the step of extracting the features is performed using a discrete wavelet transform method.

[0070] In one or more embodiments, the step of analyzing the features is performed using one of: at least one convolutional neural network (CNN); at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

[0071] In one or more embodiments, the method further comprises denoising the heart sound signals, thereby obtaining denoised sound signals, said selecting the given ones of the heart sound signals being performed amongst the denoised sound signals.

[0072] In one or more embodiments, the method further comprises receiving at least one reference sound signal indicative of ambient noise, said denoising being performed using the at least one reference sound signal.

[0073] In one or more embodiments, the method further comprises: receiving at least one electrocardiogram signal being synchronized with the heart sound signals; identifying at least one relevant temporal portion of the at least one electrocardiogram (ECG) signal; and identifying, for each one of the plurality of heart sound signals, at least one portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of heart sound signal portions, said selecting the given ones of the heart sound signals being performed amongst the plurality of the heart sound signal portions.

[0074] In one or more embodiments, the step of identifying the potential pathology is performed further based on the at least one ECG signal.

[0075] In one or more embodiments, the method further comprises receiving a plurality of acceleration signals each detected at a respective location of the body of the subject, said selecting the given ones of the heart sound signals comprising selecting particular signals amongst the plurality of heart sound signals and the plurality of acceleration signals.

[0076] In one or more embodiments, the step of selecting the particular signals comprises: assigning an initial weight value to each one of the plurality of heart sound signals and the plurality of acceleration signals, the initial weight value being indicative of a relative importance of each one of the plurality of heart sound signals and the plurality of acceleration signals; adjusting the initial weight value to obtain a final weight value for each one of the plurality of heart sound signals and the plurality of acceleration signals; and identifying the particular signals based on the final weight value.

[0077] In one or more embodiments, the step of assigning the initial weight value is performed using a K-means clustering method.

[0078] In one or more embodiments, the step of adjusting the initial weight value is performed using a gradient boosting method.

[0079] In one or more embodiments, the step of identifying the potential pathology comprises: extracting features from the particular signals; and analyzing the features to identify the pathology.

[0080] In one or more embodiments, the step of extracting the features is performed using a discrete wavelet transform method.

[0081] In one or more embodiments, the step of analyzing the features is performed using one of: at least one convolutional neural network (CNN); at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

[0082] In one or more embodiments, the method further comprises denoising the heart sound signals, thereby obtaining denoised sound signals, said selecting the particular signals being performed using the denoised sound signals.

[0083] In one or more embodiments, the method further comprises receiving at least one reference sound signal indicative of ambient noise, said denoising being performed using the at least one reference sound signal.

[0084] In one or more embodiments, the method further comprises: receiving at least one electrocardiogram signal being synchronized with the heart sound signals and the acceleration signals; identifying at least one relevant temporal portion of the at least one electrocardiogram (ECG) signal; and identifying, for each one of the plurality of heart sound signals, at least one first portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of heart sound signal portions; identifying, for each one of the plurality of acceleration signals, at least one second portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of acceleration signal portions, said selecting the particular signals being performed amongst the plurality of the heart sound signal portions and the plurality of acceleration signal portions.

[0085] In one or more embodiments, the step of identifying the potential pathology is performed further based on the at least one ECG signal.

[0086] In one or more embodiments, the step of receiving the plurality of biological sound signals comprises receiving a plurality of lung sound signals and said selecting the given signals comprising selecting given ones of the lung sound signals amongst the plurality of lung sound signals.

[0087] In one or more embodiments, the step of said outputting the potential pathology comprises providing the potential pathology for display.

[0088] According still a further broad aspect, there is provided a computer program product comprising a computer readable memory storing computer executable instructions thereon that when executed by at least one processor perform the method steps of the above method.

[0089] According to still another broad aspect, there is provided a system for identifying a potential pathology, the system comprising: a processor; a non-transitory storage medium operatively connected to the processor, the non-transitory storage medium comprising computer-readable instructions; the processor, upon executing the instructions, beingconfigured for: receiving a plurality of biological sound signals each detected at a respective position on a body of a subject; selecting given signals amongst the plurality of biological sound signals; identifying a potential pathology based on the given signals; and outputting the potential pathology.

[0090] In one or more embodiments, said receiving the plurality of biological sound signals comprises receiving a plurality of heart sound signals and said selecting the given signals comprising selecting given ones of the heart sound signals amongst the plurality of heart sound signals.

[0091] In one or more embodiments, said selecting the given ones of the heart sound signals comprises: assigning an initial weight value to each one of the plurality of heart sound signals, the initial weight value being indicative of a relative importance of each one of the plurality of heart sound signals; adjusting the initial weight value to obtain a final weight value for each one of the plurality of heart sound signals; and identifying the given ones of the heart sound signals based on the final weight value.

[0092] In one or more embodiments, said assigning the initial weight value is performed using a K-means clustering method.

[0093] In one or more embodiments, said adjusting the initial weight value is performed using a gradient boosting method.

[0094] In one or more embodiments, said identifying the potential pathology comprises: extracting features from the given ones of the heart sound signals; and analyzing the features to identify the pathology.

[0095] In one or more embodiments, said extracting the features is performed using a discrete wavelet transform method.

[0096] In one or more embodiments, said analyzing the features is performed using one of: at least one convolutional neural network (CNN); at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

[0097] In one or more embodiments, the processor is further configured for denoising the heart sound signals, thereby obtaining denoised sound signals, said selecting the given ones of the heart sound signals being performed amongst the denoised sound signals.

[0098] In one or more embodiments, the processor is further configured for receiving at least one reference sound signal indicative of ambient noise, said denoising being performed using the at least one reference sound signal.

[0099] In one or more embodiments, the processor is further configured for: receiving at least one electrocardiogram signal being synchronized with the heart sound signals; identifying at least one relevant temporal portion of the at least one electrocardiogram (ECG) signal; and identifying, for each one of the plurality of heart sound signals, at least one portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of heart sound signal portions, said selecting the given ones of the heart sound signals being performed amongst the plurality of the heart sound signal portions.

[0100] In one or more embodiments, said identifying the potential pathology is performed further based on the at least one ECG signal.

[0101] In one or more embodiments, the processor is further configured for receiving a plurality of acceleration signals each detected at a respective location of the body of the subject, said selecting the given ones of the heart sound signals comprising selecting particular signals amongst the plurality of heart sound signals and the plurality of acceleration signals.

[0102] In one or more embodiments, said selecting the particular signals comprises: assigning an initial weight value to each one of the plurality of heart sound signals and the plurality of acceleration signals, the initial weight value being indicative of a relative importance of each one of the plurality of heart sound signals and the plurality of acceleration signals; adjusting the initial weight value to obtain a final weight value for each one of the plurality of heart sound signals and the plurality of acceleration signals; and identifying the particular signals based on the final weight value.

[0103] In one or more embodiments, said assigning the initial weight value is performed using a K-means clustering method.

[0104] In one or more embodiments, said adjusting the initial weight value is performed using a gradient boosting method.

[0105] In one or more embodiments, said identifying the potential pathology comprises: extracting features from the particular signals; and analyzing the features to identify the pathology.

[0106] In one or more embodiments, said extracting the features is performed using a discrete wavelet transform method.

[0107] In one or more embodiments, said analyzing the features is performed using one of: at least one convolutional neural network (CNN); at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

[0108] In one or more embodiments, the processor is further configured for denoising the heart sound signals, thereby obtaining denoised sound signals, said selecting the particular signals being performed using the denoised sound signals.

[0109] In one or more embodiments, the processor is further configured for receiving at least one reference sound signal indicative of ambient noise, said denoising being performed using the at least one reference sound signal.

[0110] In one or more embodiments, the processor is further configured for: receiving at least one electrocardiogram signal being synchronized with the heart sound signals and the acceleration signals; identifying at least one relevant temporal portion of the at least one electrocardiogram (ECG) signal; and identifying, for each one of the plurality of heart sound signals, at least one first portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of heart sound signal portions; identifying, for each one of the plurality of acceleration signals, at least one second portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of acceleration signal portions, said selecting the particular signals being performed amongst the plurality of the heart sound signal portions and the plurality of acceleration signal portions.

[0111] In one or more embodiments, said identifying the potential pathology is performed further based on the at least one ECG signal.

[0112] In one or more embodiments, receiving the plurality of biological sound signals comprises receiving a plurality of lung sound signals and said selecting the given signals comprising selecting given ones of the lung sound signals amongst the plurality of lung sound signals.

[0113] In one or more embodiments, said outputting the potential pathology comprises providing the potential pathology for display.

[0114] According to an additional broad aspect, there is provided a system for identifying a pathology, the system comprising: the above-described garment wearable by a subject, a plurality of above-described acoustic apparatuses each mountable to the garment; a plurality of microphones each mountable to the sensor-mounting means of a respective one of the acoustic apparatuses; and the above-described system comprising the processor, wherein each one of the microphones is configured to output a respective one of the biological sound signals.

[0115] Implementations of the present technology each have at least one of the above- mentioned features and / or aspects, but do not necessarily have all of them. It should be understood that some aspects of the present technology that have resulted from attempting to attain the above-mentioned object may not satisfy this object and / or may satisfy other objects not specifically recited herein.

[0116] Additional and / or alternative features, aspects and advantages of implementations of the present technology will become apparent from the following description, the accompanying drawings and the appended claims.BRIEF DESCRIPTION OF THE DRAWINGS

[0117] Having thus generally described the nature of the invention, reference will now be made to the accompanying drawings, showing by way of illustration example embodiments thereof and in which:

[0118] FIGURE 1 is a schematic representing the system architecture of one embodiment in accordance with the present invention;

[0119] FIGURE 2 is a schematic of a cloud-based, telehealth embodiment in accordance with the present invention;

[0120] FIGURE 3 is an image of one embodiment of the present phonocardiogram apparatus taken from a front torso perspective;

[0121] FIGURE 4 is an image of the embodiment shown in FIGURE 3 taken from a rear torso perspective; and

[0122] FIGURES 5A through 5D are images from varying views of another embodiment of the present phonocardiogram apparatus.

[0123] FIGURE 6 is an image of a variety of domaphones of varying sizes for use in the phonocardiogram apparatus.

[0124] FIGURE 7 is an image showing an alternative embodiments including the phonocardiogram apparatus in position on a patient' s neck.

[0125] FIGURE 8 is a top perspective view of an acoustic apparatus or domaphone, showing a first body and a second body of the acoustic apparatus in an assembled configuration, in accordance with an embodiment;

[0126] FIGURE 9 is a bottom perspective view of the acoustic apparatus of FIG. 8;

[0127] FIGURE 10 is a side elevation view of the acoustic apparatus of FIG. 8;

[0128] FIGURE 11 is a cross-section view of the acoustic apparatus of FIG. 8, taken along a center cross-section line;

[0129] FIGURE 12 is a top perspective view of the first body of the acoustic apparatus of FIG. 8;

[0130] FIGURE 13 is a bottom perspective view of the first body of the acoustic apparatus of FIG. 8;

[0131] FIGURE 14 is a side elevation view of the first body of the acoustic apparatus of FIG. 8;

[0132] FIGURE 15 is a cross-section view of the first body of the acoustic apparatus of FIG. 8, taken along the center cross-section line;

[0133] FIGURE 16 is a top perspective view of the second body of the acoustic apparatus of FIG. 8;

[0134] FIGURE 17 is a bottom perspective view of the second body of the acoustic apparatus of FIG. 8;

[0135] FIGURE 18 is a cross-section view of the second body of the acoustic apparatus of FIG. 8, taken along the center cross-section line;

[0136] FIGURE 19 is a top perspective view of an acoustic apparatus, showing a first body and a second body of the acoustic apparatus in an assembled configuration, in accordance with another embodiment;

[0137] FIGURE 20 is a bottom plan view of the acoustic apparatus of FIG. 19;

[0138] FIGURE 21 is a side elevation view of the acoustic apparatus of FIG. 19;

[0139] FIGURE 22 is a cross-section view of the acoustic apparatus of FIG. 19, taken along a center cross-section line;

[0140] FIGURE 23 is a top perspective view of the second body of the acoustic apparatus of FIG. 19;

[0141] FIGURE 24 is a bottom perspective view of the second body of the acoustic apparatus of FIG. 19;

[0142] FIGURE 25 is a cross-section view of the second body of the acoustic apparatus of FIG. 19, taken along the center cross-section line;

[0143] FIGURE 26 is atop perspective view of a second body of an acoustic apparatus, in accordance with yet another embodiment.

[0144] FIGURE 27 is a schematic of the circuit architecture present phonocardiogram apparatus.

[0145] FIGURE 28 depicts a schematic diagram of a computing device in accordance with one or more non-limiting implementations of the present technology.

[0146] FIGURE 29 is a flow chart illustrating a method for identifying a pathology from biological sound signals in accordance with one or more non-limiting implementations of the present technology.

[0147] FIGURES 30A and 30B each illustrate a respective exemplary auscultation signature, in accordance with one or more non-limiting implementations of the present technology.DETAILED DESCRIPTION

[0148] The examples and conditional language recited herein are principally intended to aid the reader in understanding the principles of the present technology and not to limit its scope to such specifically recited examples and conditions. It will be appreciated that those skilled in the art may devise various arrangements which, although not explicitly described or shown herein, nonetheless embody the principles of the present technology and are included within its spirit and scope.

[0149] Furthermore, as an aid to understanding, the following description may describe relatively simplified implementations of the present technology. As persons skilled in the art would understand, various implementations of the present technology may be of a greater complexity.

[0150] In some cases, what are believed to be helpful examples of modifications to the present technology may also be set forth. This is done merely as an aid to understanding, and, again, not to define the scope or set forth the bounds of the present technology. These modifications are not an exhaustive list, and a person skilled in the art may make other modifications while nonetheless remaining within the scope of the present technology. Further, where no examples of modifications have been set forth, it should not be interpreted that no modifications are possible and / or that what is described is the sole manner of implementing that element of the present technology.

[0151] Moreover, all statements herein reciting principles, aspects, and implementations of the present technology, as well as specific examples thereof, areintended to encompass both structural and functional equivalents thereof, whether they are currently known or developed in the future. Thus, for example, it will be appreciated by those skilled in the art that any block diagrams herein represent conceptual views of illustrative circuitry embodying the principles of the present technology. Similarly, it will be appreciated that any flowcharts, flow diagrams, state transition diagrams, pseudo-code, and the like represent various processes which may be substantially represented in computer- readable media and so executed by a computer or processor, whether or not such computer or processor is explicitly shown.

[0152] The functions of the various elements shown in the figures, including any functional block labeled as a "processor" or a “graphics processing unit”, may be provided through the use of dedicated hardware as well as hardware capable of executing software in association with the appropriate software. When provided by a processor, the functions may be provided by a single dedicated processor, by a single shared processor, or by a plurality of individual processors, some of which may be shared. In one or more non-limiting implementations of the present technology, the processor may be a general-purpose processor, such as a central processing unit (CPU) or a processor dedicated to a specific purpose, such as a graphics processing unit (GPU). Moreover, explicit use of the term "processor" or "controller" should not be construed to refer exclusively to hardware capable of executing software, and may implicitly include, without limitation, digital signal processor (DSP) hardware, network processor, application-specific integrated circuit (ASIC), field programmable gate array (FPGA), read-only memory (ROM) for storing software, random access memory (RAM), and non-volatile storage. Other hardware, conventional and / or custom, may also be included.

[0153] With these fundamentals in place, we will now consider some non-limiting examples to illustrate various implementations of aspects of the present technology.

[0154] Referring to Fig. 1, there is shown a schematic representing an exemplary system architecture. Here, a representative image of a patient 101 is provided upon which a phonocardiogram apparatus 102 is placed. The phonocardiogram apparatus 102 may include a flexible pad embedded with an array of microphones each embedded within a dome-like structure - i.e., a “domaphone” - configured and suitably dimensioned in order to focus sound from internal areas of the patient through the patient’s skin surface to a given microphone centrally placed within each domaphone.

[0155] The dome-like structure that forms each domaphone may be formed in several ways. For example, they may include a parabolic radial cross-sectional design of the domaphones, with a thickness not exceeding 20% of their opening diameter, thus optimizing acoustic properties and signal clarity. They may include a precise mechanical configuration within the domaphones to position a microphone at the focal point of the parabolic curve, enhancing sound capture and signal fidelity for accurate cardiac diagnostics. They may include a soft edge ring and a dome-shaped body with one or multiple radial holes designed to accept electrical sensors, such as microphones, wherein the dome's opening diameter is within 25% to 100% of the dome's overall diameter. They may include a dome opening diameter ranging from 5mm to 70mm. They may include radial, circumferential, or a combination of both hollow passageways within the given domaphone to conduct cardiophonic acoustic signals (or heart sound signals).

[0156] Each domaphone may include an additional anchorage platform for securing a secondary microphone sensor, intended for the active noise cancellation of peripheral motion noise and artifacts. They may include a soft rim infused with ionic materials, enabling the domaphones to function as electrical electrodes for capturing ECG signals or measuring skin impedance. They may include embedded printed flexible or rigid circuit boards within the domaphones for enhanced signal acquisition and processing. They may also include mechanical features facilitating snap-fixation of the domaphones to the wearable component, ensuring secure attachment and optimal positioning on the patient's body. They may, as an alternative mechanical feature, allow for magnetic -fixation of the domaphones to the wearable component, offering a non-invasive and easily adjustable method of attachment. Still further, the domaphones may be formed in an individualized manner for different auscultatory areas such as provided in FIG. 6 which shows an image of a variety of domaphones of varying sizes for use in the phonocardiogram apparatus.

[0157] The plurality of domaphone elements may be fabricated from an elastic resin or suitable polymeric material and arranged in a grid pattern in a stretchable fabric (e.g., Lycra™, Spandex™, etc.) worn by the patient 101. Accordingly, it should be readily apparent that the phonocardiogram apparatus 102 is configured to conform to the patient's chest, ensuring optimal placement of each microphone for comprehensively capturing sound emitted from the patient. In the context of the present disclosure, cardiac sounds will be contemplated. The microphones are operatively connected by a wired or wireless connection103 to a digital processing unit 104 that provides filtering, analyzing, and recording of sound data emitted from the patient.

[0158] It should be understood that the digital processing unit 104 comprises a computing apparatus suitable for use with some implementations of the present technology. The digital processing unit 104 may be formed by various hardware components including one or more single or multi-core processors that may include a processor, a graphics processing unit (GPU), a solid-state drive, a random-access memory, a display interface, and an input / output interface. The digital processing unit 104 provide three functions as shown including audio processing in terms of signal filtering and boosting, Al inference in terms of pathology identification (e.g., determination of a murmur), and display of the given result (e.g., presentation of a finding of a murmur).

[0159] The digital processing unit 104 may be implemented as a server, a desktop computer, a laptop computer, a tablet, a smartphone, a personal digital assistant or any device that may be configured to implement the present technology, as it may be understood by a person skilled in the art. Likewise, the present technology may be implemented in an alternative embodiment suitable for telemedicine such as the architecture illustrated in FIG. 2.

[0160] Referring to FIG. 2, there is shown a schematic of a cloud-based, telehealth embodiment 200 in accordance with the present invention. By telehealth, it should be readily apparent that this pertains to digital and virtual care services provided over distance between a patient located remotely from an end user, typically a healthcare provider. Here, a patient 101 is shown remote from an end user 206 (e.g., doctor, nurse, technician, etc.). As before, the digital processing unit 104 again provides filtering, analyzing, and recording of sound data emitted from the patient. However, in this embodiment, additional processing and / or data storage may occur in a remote server 201 operably connected to the digital processing unit 104. Likewise, the end user 206 is provided access to a remote terminal 205 (e.g., a server, a desktop computer, a laptop computer, a tablet, a smartphone, a personal digital assistant or any device that may be configured to implement the present technology) whereupon data obtained from the phonocardiogram apparatus 102 may be presented in realtime or subsequent to patient engagement.

[0161] The digital processing unit 104, remote server 201, and remote terminal 205 are each interconnected via data paths 202, 204, and 207 via the Internet cloud 203. While interoperability may be implemented in an Internet-based embodiment as shown, it should be understood that architecture of the present invention may be implemented in a private network or virtual private network (VPN) using on-premises equipment and / or public networks without straying from the intended scope of the present invention. While data flow as indicated by data paths 202, 204, and 207 may be bi-directional, it should be understood that one or more data paths may be uni -directional. For example, the remote terminal 205 may be configured to solely receive incoming data for a “view only” session provided to the end user 206.

[0162] With regard to the cloud-based, telehealth embodiment 200 of FIG. 2, and equally applicable to all embodiments, the incoming data and related session may be realtime or provided later as a playback session depending upon the given requirements and constraints of the situation. For example, time zone differentials between patients and end users may necessitate playback sessions. As well, stored and replayed sessions may be forwarded to additional health care providers such as, but not limited to, situations requiring subsequent expert review (i.e., “second opinions”). For example, a stored session may capture, fdter, and analyze patient data including a display of Al generated pathological findings. It should be understood that such findings may be digitally packaged and conveyed to an additional party in a suitable manner - e.g., as a self-executing program sent via email or suitable similar technology.

[0163] With reference to FIG. 3, there is shown an image of one embodiment 250 of the wearable portion 251 of the phonocardiogram apparatus taken from a front torso perspective and, as shown in FIG. 4, taken from a rear torso perspective. Here, the wearable portion 251 is formed as a vest-like structure that includes adjustable straps to resize the wearable portion 251 for individuals of varying body sizes. Likewise, the wearable portion 251 may be formed in a range of fixed sizes (e.g., small, medium, large, extra large, etc.) and also be modified for female versus male anatomies without straying from the intended scope of the present invention. Common to any configuration of the wearable portion 251 are a plurality of domaphones 252.

[0164] The plurality of domaphones 252 are located on the front and rear of the wearable portion 251 and arranged in a pattern so as to obtain suitable sound readings fromthe patient. The specific placement, number, and size of each of the domaphones 252 may be varied in accordance with the given implementation. For example, an implementation for assessment of cardiac abnormalities would of course require domaphones 252 in the chest areas nearer to the patient’s heart. Likewise, an implementation for assessment of lung abnormalities would require domaphones 252 in chest areas nearer to the patient’s lungs. Alternatively, as shown in FIG. 7, an alternative embodiments may be provided for the phonocardiogram apparatus configured for positioning on a patient' s neck. Here, a wearable neck portion 260 is shown and useful, for example, in terms sensing carotid bruit. Accordingly, it should be readily apparent that the wearable portion may be provided in any suitable configuration without straying from the intended scope of the invention.

[0165] As may be seen from the embodiment of FIGS. 3 and 4, domaphones 252 are provided symmetrically along the upper chest area as well as offset to lower diaphragm areas and rear lower back areas. In such a configuration, both heart and lung assessment may be implemented in the same manner as further described hereinbelow. One alternative embodiment of the domaphones is shown by way of FIGS. 5 A through 5D which include are images from varying views of another embodiment of the present phonocardiogram apparatus.

[0166] As mentioned, the wearable portion 251 may be formed by a flexible pad embedded with an array of microphones each embedded in a corresponding domaphone 252 that are arranged in a grid pattern in a stretchable fabric worn like a garment by the patient. The wearable portion 251 thus is designed to conform to the patient's chest thereby ensuring optimal placement of each microphone for comprehensive sound capture. The microphone array may then be connected to a computing device capable of filtering, analyzing, and recording sound data as will become apparent below.

[0167] Turning now to FIGS. 8 to 26, there are provided alternative embodiments of the domaphone 302 previously described. In view of FIGS. 8 to 26, the domaphone may be described more generally as an acoustic apparatus 300, 300', 300" adapted to engage and interface with a skin surface of a patient.

[0168] With reference to the exemplary embodiment of FIGS. 8 to 18, the acoustic apparatus 300 may be used to receive and operate a body sensor (not shown) onto or in proximity with the skin surface. It will be appreciated that proximity between a body sensorand the skin surface enhances the quality of a signal captured by the sensor as it reduces a likelihood of signal artifacts.

[0169] In some applications, the body sensor is an acoustic sensor configured to capture a phonocardiogram signal for recordings of a heart sound and murmur during the cardiac cycle. An acoustic sensor configured for phonospirometry is also envisioned herein.

[0170] Referring to FIGS. 8 to 11, there is shown an embodiment of the acoustic apparatus 300 that includes a first body 310 and a second body 410 in a secured engagement with one another. More specifically, the first body 310 is resiliently embedded in the second body 410. The illustrated assembled configuration of the first and second bodies 310, 410 illustrates the manner by which the acoustic apparatus 300 may be used in operation. It should be noted that the acoustic apparatus 300 is illustrated exempt of a body sensor, which would be needed to capture a signal from a region of the body. When the acoustic apparatus 300 is operably positioned on the skin surface (not shown), the apparatus 300 may be considered in the operating configuration.

[0171] As better represented by FIG. 11, in the assembled-operating configuration, the first and second bodies 310, 410 form a chamber within their respective inner surfaces. The chamber is an encased space surrounded by: the skin surface, below; an inner surface of the second body 410, on the lateral sides; and an inner surface of the first body 310, above (provided that a body sensor is installed therewith, otherwise the chamber has one opening).

[0172] When the body signal of interest includes an acoustic signal, the chamber may concentrate sounds towards the sensor, whether it is a microphone or any other type of transducer. For example, the acoustic apparatus 300 may focus heart sounds towards a mounted microphone. The acoustic apparatus 300 may amplify a sound, thus providing an improved acoustic signal to the sensor.

[0173] The acoustic apparatus 300 may be used with a garment, such as a vest similar to the one illustrated in FIGS. 3, 4, 5A to 5D and 7. As such, the second body 410 may be configured and used as a gasket or a gasket portion to interconnect and join the first body 310 and a fabric portion of the garment.

[0174] The garment has a fabric portion (e.g., the wearable portion 301) that may include a plurality of socket holes (not visible) configured to securely fix a gasket portion410 of the acoustic apparatus 300. The gasket portion 410 of the acoustic apparatus 300 may be attached and / or made integral with the fabric with any suitable mechanical means, including glueing or molding. It is understood that the connection between the fabric portion and the gasket portion should not interfere or otherwise block a path between the skin surface and the sensor.

[0175] In one embodiment, the garment is configured such that the plurality of sockets are positioned proximate to a plurality of regions of the body to target specific body data. Evidently, the acoustic apparatus 300 would gain from being positioned on a region of the skin surface in proximity with the organ of interest to reduce a distance between the body sensor and the source of the signal.

[0176] Referring now to FIGS. 12 to 15, there is shown an embodiment of the first body 310. In the embodiment, the first body 310 has a circular bell shape which should not be understood as limitative.

[0177] The first body 310 includes a first main wall 320 that presents an inner surface. The inner surface is better appreciated jointly by FIGS. 13 and 15. The first main wall 320 and more particularly the inner surface thereof are shaped to provide a dome-shaped cavity 370. The dome-shaped cavity 370 is delimited by a lower end 330 of the first body 310 on one side and the inner surface thereof on the other side. In other words, the dome-shaped cavity 370 opens at least at the lower end 330. When assembled with the second body 410, the dome-shaped cavity 370 faces the skin surface.

[0178] The geometry of the cavity may amplify acoustic signals emanating from the patient's organs and provides a degree of auditory selectivity to the body sensor.

[0179] The term "dome-shaped" as used herein to characterize the first body 310 cavity should be construed broadly to encompass any concave shape conducive to amplifying sound and should not be limited to the specific hemispherical dome shape illustrated in the drawings. For instance, the dome shape of the cavity may be implemented by a paraboloid dome cavity having an axis of symmetry, as is the case of the embodiment of FIGS. 8 to 15 wherein the axis of symmetry is concentric with an aperture 380 to receive the body sensor. Other dome shapes (not illustrated) are envisioned herein, such as, and without being limited to a geodesic dome, an onion dome, an oval dome, or any combinations thereof. It isappreciated that other curvature profiles may be used as needed to achieve similar acoustic benefits.

[0180] In the exemplary embodiment shown, the dome-shaped cavity 370 has a cross- sectional profile with a curvature radius of about 21 mm, defined by a spline geometry with a nominal spline radius of about 17.25 mm and a height of about 10.5 mm. Various acoustic trials and tests conducted on the acoustic apparatus 300 with a dome-shaped cavity 370 having the exemplary dimensions yielded optimal results in terms of sound conduction quality and sound isolation.

[0181] The dimensions provided are exemplary and not limitative. A curvature radius of the cavity as low as about 5 mm or as high as about 45 mm may be used without departing from the scope of the present disclosure. In some other embodiments, the cavity of the first body 310 has an internal radius between 5 mm and 45 mm.

[0182] The first main wall 320 has a concave portion (i.e., that does not include the handle portion described below). The concave portion of the first main wall 320 may have an average radial thickness between about 6 mm and about 15 mm. The thickness of the first body 310 at least partially isolates the cavity 370 and shields the targeted internal sounds from ambient noises or adjacent sound sources.

[0183] The first body 310 is made at least partially of a first material. In some embodiments, the first material comprises an elastic 50A resin. In one exemplary embodiment, the first material is particularly made of Formlabs™ Elastic 50A resin. Testing has revealed that this exemplary resin offers a suitable balance of printability and compliance, since the acoustic apparatus 300 may be manufactured via 3D printing. Molding is an alternative manufacturing method for the acoustic apparatus 300 considered herein.

[0184] It will be appreciated that the materials of the acoustic apparatus 300 are selected based on five criteria: acoustic performance, patient comfort, safety, maintenance, and usability. Similar comments apply to the wall of the second body 410, as described in more details below.

[0185] The acoustic apparatus 300 includes sensor-mounting means 380 operably connected to the dome-shaped cavity 370 to mount a first sensor configured to capture a body signal. In this context, the operable connection refers to an acoustic connection. In theembodiment shown in FIGS. 8 to 18, the sensor-mounting means 380 is implemented by a sensor mounting-aperture to receive (and secure) the body sensor therein. As better understood from FIG. 11, the sensor mounting-aperture extends from the dome-shaped cavity 370 to an upper end of the first body 310. The sensor mounting-aperture has a diameter of about 6.51 mm, which is adapted to snugly receive a corresponding acoustic sensor, such as a microphone. If the microphone snugly fits in the sensor mounting -aperture, it will limit foreign sounds from reaching the sensory tip of the microphone positioned in or about the dome-shaped cavity 370. In case of a significant deviation between the aperture diameter and a width of the microphone, sealing agents or means may be used to hold the microphone into place, if needed.

[0186] Accordingly, the acoustic apparatus 300 is generally device-agnostic and brandagnostic with regard to the body sensor.

[0187] The concave portion of the first body 310 may define a body-retaining groove 332 disposed along an outer perimeter (i.e., an outer circumference) thereof. As better shown in FIGS. 12 to 15, the body-retaining groove 332 extends along a rim of the concave portion. The body-retaining groove 332 may engage a corresponding feature of the second body 410 to securely engage the body-retaining groove 332 of the first body 310 in a snap-fit connection (FIG. 11). The first body 310 is detachably securable to the second body 410.

[0188] Still referring to FIGS. 12 to 15, the first body 310 further includes a handle portion 390 having a neck 392 and a knob 394. The neck 392 extends upwardly from an apex of the upper surface of the concave portion. The knob 394 then further extends radially and outwardly from the neck 392 to define the upper end of the first body 310. As better shown in FIG. 11, the sensor-mounting aperture extends in the neck 392 and knob 394 of the handle portion 390.

[0189] With reference to FIGS. 16 to 18, the second body 410 includes a second main hollow wall 420 adapted to securely engage a portion of the first body 310. The second main hollow wall 420 has a (lower) hollow cylindrical portion and a (higher) annular resilient portion 450. The hollow cylindrical portion and the annular resilient portion 450 are made integral to one another in this embodiment.

[0190] An apex of the annular resilient portion 450 defines the upper end 430 of the second body 410. The upper end 430 defines a body-opening 422. An opposite skin-facingend 440 of the second body 410 defines a skin-facing opening 442 adapted to engage the skin surface.

[0191] Defining the upper end 430 of the second body 410, the main hollow wall 420 has the annular resilient portion 450. As shown in FIG. 11, the annular resilient portion 450 is configured to resiliently deform when the first body 310 is securely engaging the second body 410, thus resisting an upward disengagement of the first body 310. FIG. 11 shows the annular resilient portion 450 in a deformed state. FIG. 18 shows the annular resilient portion 450 in a non-deformed or unbent state, free of a pressure exerted by the first body 310.

[0192] As better shown in FIG. 17, the skin-facing end 440 of the second body 410 forms a planar ring surface. It will be appreciated that the planar ring surface offers a relatively large skin-apparatus interface that may create sufficient friction with the skin and resist an undesirable displacement of the second body 410, and by extension the acoustic apparatus 300 as a whole.

[0193] As shown in FIG. 11 , the inner surface of the second body 410 defines an annular shoulder portion adapted to receive the lower end of the first body 310 in abutment, for instance in the assembled configuration.

[0194] It will be appreciated that the first and second bodies 310, 410 are adapted so that in the assembled configuration, the respective inner surfaces align to provide a relatively smooth combined inner surface that effectively extends the dome-shaped cavity 370 downwardly, thereby improving on acoustic performance.

[0195] The second body 410 may include a body-retaining rib 432 complementarily configured to securely engage the body-retaining groove 332 of the first body 310 in a snap- fit connection. The first body 310 and the second body 410 are thus configurable into a sealed engagement at an interface therebetween. The term "sealed" as used herein refers to an airtight condition to relatively isolate the cavity 370 from exterior sounds. The interface refers to the contact area between the first and second bodies 310, 410 in the assembled configuration. Referring to FIG. 18, the body-engagement rib 432 is annular and extend along an inner perimeter (i.e., inner circumference in this embodiment) of the second main wall 420 of the second body 410.

[0196] According to an alternative embodiment (not shown), the person of ordinary skills in the art would understand that the body-retaining groove 332 and the body-retaining rib 432 may be substituted between the second body 410 and the first body 310.

[0197] The second body 410 is made at least partially of a second material. In some embodiments, the second material includes a relatively soft silicone, such as a Shore 30A hardness silicone, for a balance of regulatory compliance and biocompatibility. In one embodiment, the second material comprises a platinum-cured silicone rubber material.

[0198] With regard to the skin-facing end 440 of the second body 410, the Shore 30A silicone offers a relatively low stiffness which allows the second body 410 to deform readily under even slight pressure, thus molding to the shape (i.e., a contour of the patient's chest) and with the texture of the patient's skin surface. The exemplary material provides a sealing engagement and comfort, even with regard to prominences or curves, without causing discomfort.

[0199] With regard to the upper end 430 of the second body 410, the second material's inherent grip and friction against the first body 310 also contributes to the secure attachment, thus supplementing the mechanical engagement.

[0200] In one embodiment, the second material includes a Smooth-On Dragon Skin™ 30 silicone. Testing has revealed that this silicone offers a suitable balance of printability and compliance, since the apparatus 300 may be manufactured via 3D printing. Alternatively, Shore 00-10A Silicone to a Shore 50A silicone are envisioned by this disclosure.

[0201] Turning to FIGS. 19 to 26, there are shown alternative embodiments of the acoustic apparatus 300', 300" wherein the features are numbered with reference numerals annotated with a prime symbol which correspond to the reference numerals of the previous embodiment 300.

[0202] The acoustic apparatus 300' of FIGS. 19 to 25 is similar to the acoustic apparatus 300 shown in FIGS. 8 to 18, except that the second body 410' further includes an accessory socket 402' configured to receive another body sensor, such as an ECG sensor. The accessory socket 402' is defined in a protrusion 404' extending from a side of the second main wall420. As beter shown in FIG. 24, the accessory socket 402' is recessed from a skin-facing end 440' of said protrusion 404'.

[0203]

[0001] FIG. 26 shows yet another embodiment of the second body 410" of an acoustic apparatus 300" that includes an elongated helicoidal channel extending upwardly along the exterior of the cylindrical portion from the accessory socket. The helicoidal channel may be used to snugly receive a wire associated with an accessory fited in the accessory socket.

[0204] Referring to FIG. 27, there is shown a schematic diagram of an exemplary overall circuit architecture 500 being suitable for implementing one or more non-limiting implementations of the present technology. It is to be expressly understood that the circuit architecture 500 as shown is merely an illustrative implementation ofthe present technology. Thus, the description thereof that follows is intended to be only a description of illustrative examples of the present technology. This description is not intended to define the scope or set forth the bounds of the present technology. In some cases, what are believed to be helpful examples of modifications to the circuit architecture 500 may also be set forth below. This is done merely as an aid to understanding, and, again, not to define the scope or set forth the bounds of the present technology. These modifications are not an exhaustive list, and, as a person skilled in the art would understand, other modifications are likely possible. Further, where this has not been done (i.e., where no examples of modifications have been set forth), it should not be interpreted that no modifications are possible and / or that what is described is the sole manner of implementing that element of the present technology. As a person skilled in the art would understand, this is likely not the case. In addition, it is to be understood that the circuit architecture 500 may provide in certain instances simple implementations of the present technology, and that where such is the case they have been presented in this manner as an aid to understanding. As persons skilled in the art would understand, various implementations of the present technology may be of a greater complexity.

[0205] With further reference to FIG. 27, there is shown the circuit architecture 500 in accordance with one embodiment of the present phonocardiogram apparatus. Here, a domaphone 502 is provided for placement within a wearable portion 506. As may be seen by doted line, the domaphone 502 embeds within the thickness of the wearable portion 506 in an integrated manner. This ensures proper placement of each domaphone 502 atop theskin surface (not shown) of a patient against wearable portion 506 is placed. For purposes of illustrative clarity, only one domaphone 502 of the plurality of domaphones is shown. Each domaphone 502 includes a sensing unit 501 attached thereto via a snap feature 507 thereby rendering the sensing unit removable from the domaphone 502 and thus detachable from the wearable portion 506. It should be noted that the dome-like structure of the domaphone 502 increases via acoustic amplification the signal to noise ratio (SNR) of the sound signals emitted from the patient thereby resulting in sound data of exceptional quality and clarity.

[0206] The sensing unit 501 may be provisioned for and capable of obtaining a variety of data ranging from, but not limited to, electrocardiogram (ECG) data, sound data, acceleration data, and temperature data. Details of the sensing unit and more particularly the specific sensors used are known to those in the sensor art and will not be further described herein. Each sensor 501 collects the aforementioned data and provides such collected data in packet form to a first bus for example. As there may be 1 through N sensor units 501, there are likewise 1 to N data outputs from the respective sensor units to the 1 to N data inputs of the bus. Although a single bus is shown, communication may be enabled by one or more internal and / or external buses (e.g., a PCI bus, universal serial bus, IEEE 1394 “Firewire” bus, SCSI bus, Serial-ATA bus, etc.), to which the various hardware components are electronically coupled.

[0207] As the bus obtains aggregate signals from the array of (1 through N) sensing units, the circuit architecture 500 further includes 1 through N demultiplexers to separate each aggregate signal into discrete signal channels (SCI through SC4) representing, for example, ECG data, sound data, acceleration data, and temperature data, respectively, from each aggregate signal. It should be noted that the first bus and the demultiplexers may be provided in the form of an FPGA, ARM, DSP or any suitable system-on-chip solution for enabling data conditioning software given the packetized data.

[0208] This multi-channel approach obtaining multiple data streams from multiple locations (e.g., chest, back, neck) within the body of the patient and fusion of such data may be advantageous in that simultaneous data is obtained from multiple cardiac areas (or lung or digestive areas depending on the given implementation) which differs greatly from known medical methods such as standard stethoscope auscultations. In some embodiments, the multi-channel approach in accordance with the present technology provides benefits such ascapturing time differentials between the S 1 and S2 heart sounds imperceptible to the human ear. Such intervals between heart sounds may be captured and analyzed over time and usefully provide pathology trends.

[0209] In some embodiments, sixteen (16) channels (i.e., discrete locations from which data is gathered) may be present. However, the Al augmented implementation may automatically analyze the real-time data to determine (i.e., automatically calibrate) the best channels in order to adjust the diagnostics obtained and processed. A byproduct of the digital output produced by way of the present invention may include an “auscultation signature” which is reproducible over time and provides a patient-specific spectrum of observations. Such signature will of course change overtime as any given pathology improves or worsens. Thus, longitudinal monitoring overtime a given patient’s auscultation signature enabled by the present phonocardiogram apparatus represents an improvement over existing technologies.

[0210] The auscultation signature will be the unique patterns and characteristics of sounds heard. FIGS. 30A and 30B each illustrate a respective exemplary auscultation signature.

[0211] Each demultiplexer (Demux 1 though Demux N) then weaves and compresses the multi-channel data to provide the discrete signals (SCI through SC4) to a second bus. At the second bus, each discrete signal (SCI through SC4) from the respective sensing unit (1 through N) are then combined by a corresponding multiplexer (Mux 1 through Mux 4). The second bus may be a peripheral component interconnect (PCI) or I2C bus or any suitable short distance, intra-board communication mechanism to provide a local bus standardization to connect to a computing device (here, a PC is shown at 505, though any suitable processing device may be utilized). The computing device will then incur processing at 504 of the multiplexed data in conjunction with Al enhancements. In some embodiments, processing 504 may occur locally such as shown in FIG. 1. In alternative embodiments such as the cloud-based model of FIG. 2, processing 504 may occur remotely (i.e., via remote server 201). It should be readily apparent that processing may be a combination of local edge processing or remote cloud processing without straying from the intended scope of the present technology.

[0212] It will be understood that the number of types of data measured by a sensing unit 501 may vary and that not all sensing units 501 may measure the same type of data. While in the above description, each sensing unit 501 is provided with a microphone to generate sound data, an ECG sensor to measure ECG data, an accelerometer to measure acceleration data and a temperature sensor to measure temperature data, it will be understood that the sensing unit 501 may be configured to measure only three types of data, such as sound data, ECG data and acceleration data, or only two types of data such as sound data and acceleration data. In some embodiments, the sensing units 501 may be configured to only measured a single type of data such as sound data. The person skilled in the art will understand that, in this case, the demultiplexers are adapted to the number of data types measured by the sensing units 501 and the number of multiplexers is adapted to correspond to the number of data types.

[0213] It will also be understood that not all sensing units 501 may be configured to measure the same types of data. For example, a first sensing unit may be configured to measure sound data and acceleration data while a second sensing unit 501 may be configured to measure sound data, acceleration data and ECG data while a third sensing unit 501 may be configured to measure sound data only.

[0214] In some embodiments, with continued reference to FIG. 27, the circuit architecture 500 provides processing 504 that includes four primary categories of function. These include: data preparation (Data Prep) which may include sound pre-processing, microphone salience detection, and automated biasing and selection; machine learning analysis via a deep neural network (Deep Net); storage of data both obtained and processed overtime (Storage); and provisioning of a diagnostic output (Display).

[0215] As previously discussed, the processing of data is accomplished in conjunction with the assistance of Al in terms of applied Machine Learning Algorithms (MLA). It is well understood that MLAs build a mathematical model based on sample data, known as "training data", in order to make predictions or decisions without being explicitly programmed to do so. The present phonocardiogram apparatus provides a computer implemented process embodied in software to accomplish the following detailed processing steps.

[0216] Initial processing of data (Data Prep) obtained from the aforementioned multichannel approach will first include sound pre-processing using Adaptive Noise Cancellation (ANC) in order to eliminate ambient and physiological noise that could obscure cardiac sounds. This is accomplished and implemented using a secondary microphone to capture ambient noise, which is then subtracted from the primary heart sound signal, enhancing the clarity of cardiac events. Placement of the secondary microphone may vary depending on the given implementation. Indeed, one or more such secondary microphones may be provided, each being included in a corresponding domaphone.

[0217] Further, initial processing of data (Data Prep) will include microphone salience detection to determine the relative importance of each microphone within the given sensing units. This is accomplished via feature-based clustering, specifically K-means clustering which is a well-known MLA that is an unsupervised method because that starts without labels and then forms and labels groups itself. This step in processing serves to analyze the initial sound data from all microphones and identify those that capture the heart sounds with the highest fidelity and least noise. This is determined by evaluating the SNR, frequency content, and temporal characteristics of the sounds. The K-means MLA clusters microphones based on their sound characteristics, identifying clusters that represent the most informative acoustic signals for the patient's heart sounds. It will be appreciated that alternative supervised and non-supervised clustering techniques may be used in the context of the present technology.

[0218] Still further, initial processing of data (Data Prep) will include automated biasing and selection using gradient boosting machines (GBM). GBM is a known technique in machine learning that combines the predictions from several models to improve overall predictive accuracy. Here, GBM is used to dynamically adjust the input weights of the selected microphones, ensuring the machine learning model receives the best possible input signal. This is accomplished whereby the GBM adjusts the sensitivity and selection of microphones in real-time, based on their contribution to accurate model predictions, effectively learning which microphones' data most improve diagnostic accuracy. It will be appreciated that additionality or alternatively, techniques including adaptive filtering, beamforming, noise cancellation algorithms, automatic gain control, signal fusion techniques, and time-delay estimation, and machine learning may be used alone or in combination with machine learning to perform initial processing.

[0219] Once initial processing of data (Data Prep) is accomplished, machine learning analysis via the deep neural network (Deep Net) provides feature extraction using discrete wavelet transform (DWT) to decompose heart sounds into wavelet coefficients, capturing both frequency and location information of cardiac events. It should be understood that alternative forms of machine learning analysis other than DWT may be used for signal decomposition such as, but not limited to, spectral analysis, convolutional techniques, or the like.

[0220] Once decomposed, the data is processed by way of a classification algorithm using convolutional neural networks (CNN) combined with long short-term memory (LSTM) networks in order to recognize patterns within the heart sound signals, while the LSTMs analyze the temporal progression of these sounds. This may be advantageous for identifying patterns associated with different cardiac pathologies. The CNN-LSTM model is implemented and trained on a labeled dataset of heart sounds held a data repository (Storage), learning to identify specific sound patterns associated with normal and pathological conditions. The labeled dataset of heart sounds may be sourced from existing web-based repositories of known heart sounds and / or may be obtained and built via clinical use of the present phonocardiograph apparatus over time. The CNN-LSTM model outputs a diagnosis with a confidence score, indicating the likelihood of various cardiac pathologies. In one or more alternative implementations, one or more of recurrent neural networks (RNN), decision trees, random forests, principal component analysis (PCA) may be used to output diagnosis.

[0221] In one or more implementations adjustment in real time may be performed by using one or more of reinforcement learning algorithms, online learning algorithms, adaptive neural fuzzy inference system (ANFIS), real-time decision trees, and incremental learning algorithms.

[0222] Once initial processing and machine learning analysis is accomplished, a diagnostic output is rendered typically in a visual manner (Display). Here, there is provided a comprehensive report detailing the detected cardiac condition(s), including a visualization of heart sound origin points on the chest that were most indicative of pathology. A visual representation of the “lub dub” sounds is shown, along with automatically labeled salient segments including S 1 and S2 portions with corresponding durations and features that are related to the diagnosis.

[0223] The digital processing unit may additionally provide for a fine-tuned large language model (LLM) for auto-generating reports based on cardiophonic signal observations obtained by way of the phonocardiogram apparatus. This may include the LLM being specifically trained to interpret cardiophonic signal data, including heart and lung sounds captured by the phonocardiogram apparatus, to identify patterns indicative of cardiac pathologies. This may also include the capability of the LLM to analyze complex datasets derived from the cardiophonic signals, employing advanced algorithms to detect subtle nuances and correlations within the data that may signify health conditions. This may further include the automatic generation of comprehensive diagnostic reports by the LLM, which include detailed descriptions of observed cardiophonic signal patterns, inferred cardiac pathologies, and potential health implications, all presented in a clear and understandable format for end users (e.g., healthcare providers). Still further, this may include the integration of the LLM to work in conjunction with the phonocardiogram apparatus’s AI- based algorithm for sound analysis and pathology identification, ensuring a seamless workflow from signal capture to diagnostic reporting. Yet still further, this may include the LLM's ability to continuously learn and improve its reporting accuracy and detail through machine learning, with each new dataset processed contributing to the model's expanding knowledge base, thereby enhancing its diagnostic capabilities over time.

[0224] FIG. 28 illustrates one embodiment of a computing device 600 that may be used to execute the processing 504 described above or the below described method 700.

[0225] The computing device 600 suitable for use with some implementations of the present technology, the computing device 600 comprising various hardware components including one or more single or multi -core processors collectively represented by processor 610, agraphics processing unit (GPU) 611, a solid-state drive 620, a random-access memory 630, a display interface 640, and an input / output interface 650.

[0226] Communication between the various components of the computing device 600 may be enabled by one or more internal and / or external buses 660 (e.g., a PCI bus, universal serial bus, IEEE 1394 "Firewire" bus, SCSI bus, Serial-ATA bus, etc.), to which the various hardware components are electronically coupled.

[0227] The input / output interface 650 may be coupled to a touchscreen 690 and / or to the one or more internal and / or external buses 660. The touchscreen 690 may be part of thedisplay. In one or more implementations, the touchscreen 690 is the display. The touchscreen 690 may equally be referred to as a screen 690. In the implementations illustrated in FIG. 9, the touchscreen 690 comprises touch hardware 694 (e.g., pressure-sensitive cells embedded in a layer of a display allowing detection of a physical interaction between a user and the display) and a touch input / output controller 692 allowing communication with the display interface 640 and / or the one or more internal and / or external buses 660. In one or more implementations, the input / output interface 650 may be connected to a keyboard (not shown), a mouse (not shown) or a trackpad (not shown) allowing the user to interact with the computing device 600 in addition or in replacement of the touchscreen 690.

[0228] According to implementations of the present technology, the solid-state drive 620 stores program instructions suitable for being loaded into the random-access memory 630 and executed by the processor 610 and / or the GPU 611 for executing the steps of the method 700 for example. For example, the program instructions may be part of a library or an application.

[0229] The computing device 600 may be implemented as a server, a desktop computer, a laptop computer, a tablet, a smartphone, a personal digital assistant or any device that may be configured to implement the present technology, as it may be understood by a person skilled in the art.

[0230] In the following and with reference to FIG. 29, there is described a computer- implemented method for identifying a potential pathology such as a potential cardiac pathology in accordance with an embodiment.

[0231] FIG. 10 depicts a flowchart of a method 700 for identifying a potential pathology in accordance with one or more non-limiting implementations of the present technology.

[0232] In one or more implementations, the computing device 600 or the server 201 comprising at least the processor 610 and / or the GPU 611 which is operatively connected to a non-transitory computer readable storage medium such as the solid-state drive 620 and / or the random-access memory 630 storing computer-readable instructions. The processing device or the server upon executing the computer-readable instructions, is configured to or operable to execute the method 700.

[0233] According to processing step 702, the processing device 600 receives a plurality of biological sound signal or data each having been measured by a respective microphone located at a respective and different location or position on the body of a subject. It will be understood that the sound signals may have been collected using the above-described domaphones or sensing units 501 each positioned at a respective location on the body of the subject. However, any other adequate device provided with a microphone to detect a biological sound may be used.

[0234] It should be understood that each biological signal is associated with a respective position on the body of the subject from which the biological signal has been measured or determined. In some embodiments, step 702 further comprises receiving, for each biological signal, the corresponding position on the body or an identification of the microphone that measured the biological signal.

[0235] In some embodiments, the biological sounds correspond to heart sounds. In other embodiments, the biological sounds correspond to lung sounds.

[0236] According to processing step 704, the processing device 600 selects some or given signals amongst the received biological sound signals, as described in greater detail below. For example, the selected signals may correspond to the given biological signals that offer the best quality or the most information to improve the identification of the potential pathology.

[0237] According to processing step 706, the processing device 600 identifies a potential pathology based on the selected biological signals, as described in greater detail below.

[0238] In the following, the method 700 will further be described for biological signals corresponding to heart sound signals. Therefore, step 702 then consists in receiving a plurality of heart sound signals.

[0239] In some embodiments, the given heart sound signals selected amongst the received heart sound signals at step 704 are identified as follow. First, an initial weight value is assigned to each heart sound signal received at step 702. The initial weight value is indicative of a relative importance of each one of the received heart sound signals. Then, the initial weight value of each heart sound signal is adjusted to obtain a final weight value, andthe identification of the selected heart sound signals is performed based on the final weight values assigned to the heart sound signals. In some embodiments, only the heart sound signals having assigned thereto a final weight value being greater than a predefined threshold are selected at step 704. In other embodiments, a predefined number of heart sound signals having the greatest final weight values are selected at step 704. For example, the heart sound signals may be ranked based on their assigned final weight value (i.e., the heart sound signal having the greatest final weight value assigned thereto is ranked first while the heart sound signal having the lowest final weight value assigned thereto is ranked last) and only the top five heart sound signals having the highest final weight values are selected at step 704.

[0240] In some embodiments, the initial weight value is determined using a K-means clustering method, as described above with reference to FIG. 27.

[0241] In some embodiments, the adjustment of the initial weight values is performed using a gradient boosting method, as described above with reference to FIG. 27.

[0242] In some embodiments, the identification of the potential pathology at step 706 is performed in two steps. First, features are extracted from each selected heart sound signal, and then the extracted features are analyzed to determine whether a potential pathology is detected.

[0243] It should be understood that any adequate method for extracting features from audio signals may be used at step 706. In some embodiments, the features are extracted using a discrete wavelet transform method, as described above with reference to FIG. 27.

[0244] It should also be understood that any adequate method for analyzing the features to identify a potential pathology may be used at step 706. In some embodiments, a machine learning classifier is used to identify a potential pathology based on the features extracted from the selected heart sound signals. Examples of classifiers that may be used include: at least one convolutional neural network (CNN), at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

[0245] In some embodiments, the feature extractor and the classifier are trained using labeled data, such as labeled sound signals. Cross-entropy optimizers may be used loss functions. The loss functions of the trained models may represent signal saliency topathology, power spectrum density and entropy. An optimization method such as Adam optimization may be used. The feature extractor and the classifier may be trained jointly and in multiple phases.

[0246] In some embodiments, the method 700 further comprises a step of denoising the heart sound signals received at step 702, thereby obtaining denoised heart sound signals. In this case, the denoised heart sound signals are used as input of step 704, which then consists in selecting at least one denoised heart sound signal.

[0247] It should be understood that any adequate method for denoising sound or audio signals may be used. In some embodiments, the system further comprises at least one reference microphone for measuring ambient noise and generating at least one reference sound signal and the denoising of the received heart sound signals is performed based on the reference sound signal(s).

[0248] In some embodiments, the method further comprises a step of receiving at least one ECG signal measured on the body of the subject. As known in the art, an ECG signal may be obtained by fixing an electrode at an adequate position on the body of the subject and up to 12 electrodes are usually used. However, the person skilled in the art will understand that at least one electrode may be used.

[0249] It will be understood that the ECG signal(s) and the heart audio signals are obtained concurrently and are therefore synchronized so that a given heart event occurs at the same type within the ECG signal and within the heart sound signals.

[0250] In some embodiments, the method 700 further comprises a step of identifying at least one temporal portion or window of the ECG signal during which a heart event of interest occurs, i.e., the start and end times of the heart event of interest within the ECG signal. For example, the period(s) of time during which atrial systole occurs within the ECG signal may be identified. Ventricular systole and diastole will also be identified by the ECG signal. It will be understood that if the ECG signal comprises more than one cardiac cycle, more than one temporal portion may be identified. Then the portion(s) within each heart sound signal received at step 702 corresponding to the identified ECG portion is(are) identified and extracted to obtain at least one heart sound signal portion for each heart sound signal. Each heart sound signal portion has then same start and end times as those of the identified ECG portion. The extracted heart sound signal portion(s) is(are) then used as inputfor step 704, i.e., the selection step 704 consists in selecting the heart sound signal portion(s) of at least one received heart sound signals.

[0251] In some embodiments, the method 700 further comprises receiving at least one acceleration signal indicative of an acceleration measured by an accelerometer installed on the body of the subject at an adequate position. An accelerometer may identify a grade IV heart murmur which would be accompanied by a vibratory sensation over the chest. These accelerometers can be positioned on the chest adjacent to the domaphones. In this case, the selection of given signals at step 704 is performed amongst the received heart sound signals and the received acceleration signals. This would use the same selection method as described above for heart sound signals. The selected signals then comprise at least one heart sound signal and / or at least one acceleration signal.

[0252] In some embodiments, the selection method described above applies when at least one signal is to be selected amongst a group of signals comprising the heart sound signals and the acceleration signals, i.e. an initial weight value is assigned to each received heart sound signal and each received acceleration signal and each determined initial weight value is then adjusted so as all heart rate signals and all acceleration signals are part of a same ranking. For example, the signals having a final weight value greater than a threshold value may be selected independently of whether they are heart sound signals or acceleration signals. In another example, a predefined number of signals having the greatest final weight values is selected independently of whether they are heart sound signals or acceleration signals.

[0253] The identification of the potential cardiac pathology is performed based on these selected signals which comprise at least one heart sound signal and / or at least one acceleration signal.

[0254] In some embodiments in which the selected signals comprise at least one heart sound signal and at least one acceleration signal, the same feature extractor and the same classifier as described above when the selected signals only comprise heart sound signals are used for extracting features from the selected heart sound signal(s) and selected acceleration signal(s) and analyzing the extracted features to identify a potential cardiac pathology.

[0255] In other embodiments in which the selected signals comprise at least one heart sound signal and at least one acceleration signal, two distinct feature extractors and two distinct classifiers are used. A first feature extractor is used for extracting the features from the selected heart sound signal(s) and a second and different feature extractor is use for extracting features from the selected acceleration signal(s). A first classifier is then used for determining if the subject potentially suffers from a cardiac pathology based on the features extracted from the selected hear sound signal(s) and a second and different classifier is used for determining if the subject potentially suffers from a cardiac pathology based on the features extracted from the acceleration signals.

[0256] The accelerometers will be used to disambiguate motion artifacts in the ECG and the sound signals. These motion artifacts include but are not limited to a heart murmur of grade IV or higher, and right ventricular hypertrophy associated with an RV heave.

[0257] In some embodiments and as described above, the method 700 may further comprise a step of selecting and extracting portions of the heart sound signals and portions of the acceleration signals prior to the selection step 704 based on identified portions of at least one ECG signal. It will be understood that the above-described method for extracting portions of heart sound signals can be used for extracting corresponding portions of acceleration signals.

[0258] In some embodiments in which at least one ECG signal is received, the identification of a potential cardiac pathology is further based on the ECG signal(s). Am ECG signal is treated as analogous to heart sound signals and acceleration signals for the purpose of initial grouping before the extraction of features. This allows for the usage of the same feature extractor and classifier as in use for heart sounds and accelerometer signals.

[0259] In some embodiments, the biological signals received at step 702 comprise lung sound signals. Sound signals such as loud P2 could identify the pulmonary condition of pulmonary hypertension.

[0260] In some embodiments, the classifier, when identifying a potential pathology is configured to determine a confidence score indicative of the likelihood of the pathology. In this case, the step 708 further comprises outputting the confidence score.

[0261] The phonocardiogram apparatus in accordance with the present invention embodies a significant leap in cardiac technology, leveraging advanced algorithms to offer a nuanced and accurate tool for cardiac diagnosis and affording a more holistic assessment of cardiac health.

[0262] Having described the embodiments of the present invention, it should therefore be readily apparent that the present invention improves upon the state of the art to thereby provide a multi-channel diagnostic tool able to identify acoustic time differentials between different parts of the heart. Such information is fundamental the identification of certain pathologies such as aortic stenosis. Further, because of the multi-channel characteristics of the phonocardiogram apparatus, this enables the novel feature of self-selecting preferential channels (i.e., automatic calibration) to optimize sound quality. Moreover, a patient’s selfdetection of cardiac murmurs suggestive of valvular heart disease is also possible. Without limiting the intended scope of the present invention and by way of example only, various additional features and advantageous diagnoses possible by way of the present phonocardiogram apparatus include objective detection, interpretation, and longitudinal monitoring and remote telemonitoring by a clinician of the following exemplary cardiac pathologies and cardiac sounds:• Aortic stenosis;• Aortic regurgitation;• Mitral stenosis;• Mitral regurgitation;• Opening snap• S2 to opening snap interval to grade severity of mitral stenosis;• Pulmonic stenosis;• Pulmonic regurgitation;• Tricuspid stenosis;• Tricuspid regurgitation;• Opening snap consistent with mitral stenosis;• Loud SI consistent with mitral stenosis;• Pericardial knock consistent with pericarditis;• Systolic flow murmur;• Pericardial rub (pericarditis);• S4 patern consistent with left ventricular hypertrophy;• S3 patern consistent with left ventricular dilatation;• Continuous murmur consistent with left ventricular septal defect;• Paradoxical S2 consistent with aortic stenosis or hypertrophic cardiomyopathy;• Fix-split S2 consistent with left atrial septal defect;• Loud P2 consistent with pulmonary hypertension;• Lung crepitation consistent with congestive heart failure;• Lung crepitation consistent with pulmonary fibrosis;• Bronchial breathing consistent with pneumonia;• S2 to S 1 interval to assess atrial filling pressures;• Carotid bruit (using neck domaphones);• Typical or atypical sound of a ventricular assist device; and• Typical or atypical sound of a mechanical or bioprosthetic valve.

[0263] The embodiments described above are intended to be exemplary only. The scope of the invention is therefore intended to be limited solely by the appended claims.

Claims

CLAIMS1. An acoustic apparatus (300) for operating a sensor about a skin surface, the acoustic apparatus (300) comprising: a first body (310) made at least partially of a first material, having an inner surface (312) defining a dome-shaped cavity (370) opening at least at a lower end (330), and comprising sensor-mounting means (380) operably connected to the dome-shaped cavity (370) adapted to mount a first sensor, and a second body (410) made at least partially of a second material, having a main hollow wall (420) adapted to securely engage a portion of the first body (310), the main hollow wall (420) having an upper end (430) defining a body-opening (422) and an opposite and connected skinfacing end (440) adapted to engage the skin surface and defining a skin-facing opening (442); wherein the second material of the second body comprises a resilient material.

2. The acoustic apparatus of claim 1, wherein the first material has a greater stiffness than the second material.

3. The acoustic apparatus of claim 1 or 2, wherein the first body and the second body are configurable into a sealed engagement at an interface therebetween.

4. The acoustic apparatus of any one of claims 1 to 3, wherein the first body is detachably securable to the second body.

5. The acoustic apparatus of claim 4, wherein the first body defines a body-retaining groove (332) disposed along an outer perimeter thereof, and wherein the second body comprises a body-retaining rib (432) complementarily configured to securely engage the body-retaining groove of the first body in a snap-fit connection.

6. The acoustic apparatus of any one of claims 1 to 5, wherein the sensor mounting means(380) comprises a sensor mounting-aperture extending from the dome-shaped cavity to an upper end of the first body, thus further opening the dome-shaped cavity.

7. The acoustic apparatus of claim 6, wherein the first body further comprises a handle portion (390) having a neck (392) and a knob (394), the neck extending from a first main wall (320) of the first body, and the knob further extending from the neck to define the upper end of the first body, and wherein the sensor-mounting aperture extends in the neck and knob of the handle portion.

8. The acoustic apparatus of claim 6 or 7, wherein the sensor mounting-aperture is configured to accept an acoustic sensor configured to capture a phonocardiogram signal.

9. The acoustic apparatus of any one of claims 1 to 8, wherein the dome-shaped cavity of the first body has an internal radius between 5 mm and 45 mm.

10. The acoustic apparatus of claim 9, wherein the dome-shaped cavity of the first body has an internal radius between 11 mm and 31 mm.

11. The acoustic apparatus of any one of claims 1 to 10, wherein dome-shaped cavity has a nominal spline radius of 17.25 mm and a height of 10.5 mm.

12. The acoustic apparatus of any one of claims 1 to 11, wherein the inner surface of the first body is adapted such that the dome-shaped cavity comprises a paraboloid dome cavity.

13. The acoustic apparatus of any one of claims 1 to 12, wherein the first material comprises an elastic 50A resin.

14. The acoustic apparatus of any one of claims 1 to 13, wherein the second body is adapted to operably mount a garment positionable in proximity to the skin surface.

15. The acoustic apparatus of any one of claims 1 to 14, wherein the skin- facing end of the main hollow wall of the second body forms a planar ring surface to interface with the skin surface.

16. The acoustic apparatus of any one of claims 1 to 15, wherein the main hollow wall has an annular resilient portion (450) surrounding the body-opening and defining the upper end of the second body, wherein the annular resilient portion is configured to resiliently deform when the first body is securely engaging the second body, thus resisting an upward disengagement of the first body.

17. The acoustic apparatus of any one of claims 1 to 16, wherein the second material comprises a platinum-cured silicone rubber material.

18. The acoustic apparatus of claim 1, wherein the first body and the second body are monolithic, thus enabling the secured engagement therebetween.

19. A garment for capturing a body signal of a body, the garment comprising: a fabric portion comprising a plurality of socket openings each located at a respective position thereon, and configured for securely receiving a socket portion of an acoustic apparatus for operating a sensor about a skin surface.

20. The garment of claim 19, wherein the garment is configured such that the plurality of sockets openings are positioned proximate to a plurality of regions of the body to capture the body data.

21. A method for identifying a potential pathology, the method being executed by a processor, the method comprising: receiving a plurality of biological sound signals each detected at a respective position on a body of a subject; selecting given signals amongst the plurality of biological sound signals; identifying a potential pathology based on the given signals; andoutputting the potential pathology.

22. The method of claim 21, wherein said receiving the plurality of biological sound signals comprises receiving a plurality of heart sound signals and said selecting the given signals comprising selecting given ones of the heart sound signals amongst the plurality of heart sound signals.

23. The method of claim 22, wherein said selecting the given ones of the heart sound signals comprises: assigning an initial weight value to each one of the plurality of heart sound signals, the initial weight value being indicative of a relative importance of each one of the plurality of heart sound signals; adjusting the initial weight value to obtain a final weight value for each one of the plurality of heart sound signals; and identifying the given ones of the heart sound signals based on the final weight value.

24. The method of claim 23, wherein said assigning the initial weight value is performed using a K-means clustering method.

25. The method of claim 23 or 24, wherein said adjusting the initial weight value is performed using a gradient boosting method.

26. The method of any one of claims 22 to 25, wherein said identifying the potential pathology comprises: extracting features from the given ones of the heart sound signals; and analyzing the features to identify the pathology.

27. The method of claim 26, wherein said extracting the features is performed using a discrete wavelet transform method.

28. The method of claim 26 or 27, wherein said analyzing the features is performed using one of: at least one convolutional neural network (CNN); at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

29. The method of any one of claims 22 to 28, further comprising denoising the heart sound signals, thereby obtaining denoised sound signals, said selecting the given ones of the heart sound signals being performed amongst the denoised sound signals.

30. The method of claim 29, further comprising receiving at least one reference sound signal indicative of ambient noise, said denoising being performed using the at least one reference sound signal.

31. The method of any one of claims 22 to 30, further comprising: receiving at least one electrocardiogram signal being synchronized with the heart sound signals; identifying at least one relevant temporal portion of the at least one electrocardiogram (ECG) signal; and identifying, for each one of the plurality of heart sound signals, at least one portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of heart sound signal portions, said selecting the given ones of the heart sound signals being performed amongst the plurality of the heart sound signal portions.

32. The method of claim 31 , wherein said identifying the potential pathology is performed further based on the at least one ECG signal.

33. The method of any one of claims 22 to 30, further comprising receiving a plurality of acceleration signals each detected at a respective location of the body of the subject, said selecting the given ones of the heart sound signals comprising selecting particular signals amongst the plurality of heart sound signals and the plurality of acceleration signals.

34. The method of claim 33, wherein said selecting the particular signals comprises: assigning an initial weight value to each one of the plurality of heart sound signals and the plurality of acceleration signals, the initial weight value being indicative of a relative importance of each one of the plurality of heart sound signals and the plurality of acceleration signals; adjusting the initial weight value to obtain a final weight value for each one of the plurality of heart sound signals and the plurality of acceleration signals; and identifying the particular signals based on the final weight value.

35. The method of claim 34, wherein said assigning the initial weight value is performed using a K-means clustering method.

36. The method of claim 34 or 35, wherein said adjusting the initial weight value is performed using a gradient boosting method.

37. The method of any one of claims 33 to 36, wherein said identifying the potential pathology comprises: extracting features from the particular signals; and analyzing the features to identify the pathology.

38. The method of claim 37, wherein said extracting the features is performed using a discrete wavelet transform method.

39. The method of claim 37 or 38, wherein said analyzing the features is performed using one of: at least one convolutional neural network (CNN);at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

40. The method of any one of claims 33 to 39, further comprising denoising the heart sound signals, thereby obtaining denoised sound signals, said selecting the particular signals being performed using the denoised sound signals.

41. The method of claim 40, further comprising receiving at least one reference sound signal indicative of ambient noise, said denoising being performed using the at least one reference sound signal.

42. The method of any one of claims 33 to 41, further comprising: receiving at least one electrocardiogram signal being synchronized with the heart sound signals and the acceleration signals; identifying at least one relevant temporal portion of the at least one electrocardiogram (ECG) signal; and identifying, for each one of the plurality of heart sound signals, at least one first portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of heart sound signal portions; identifying, for each one of the plurality of acceleration signals, at least one second portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of acceleration signal portions, said selecting the particular signals being performed amongst the plurality of the heart sound signal portions and the plurality of acceleration signal portions.

43. The method of claim 42, wherein said identifying the potential pathology is performed further based on the at least one ECG signal.

44. The apparatus of claim 21, wherein said receiving the plurality of biological sound signals comprises receiving a plurality of lung sound signals and said selecting the given signals comprising selecting given ones of the lung sound signals amongst the plurality of lung sound signals.

45. The method of any one of claims 21 to 44, wherein said outputting the potential pathology comprises providing the potential pathology for display.

46. A computer program product comprising a computer readable memory storing computer executable instructions thereon that when executed by at least one processor perform the method steps of any one of claims 21 to 45.

47. A system for identifying a potential pathology, the system comprising: a processor; a non-transitory storage medium operatively connected to the processor, the non- transitory storage medium comprising computer-readable instructions; the processor, upon executing the instructions, being configured for: receiving a plurality of biological sound signals each detected at a respective position on a body of a subject; selecting given signals amongst the plurality of biological sound signals; identifying a potential pathology based on the given signals; and outputting the potential pathology.

48. The system of claim 47, wherein said receiving the plurality of biological sound signals comprises receiving a plurality of heart sound signals and said selecting the given signals comprising selecting given ones of the heart sound signals amongst the plurality of heart sound signals.

49. The system of claim 48, wherein said selecting the given ones of the heart sound signals comprises:assigning an initial weight value to each one of the plurality of heart sound signals, the initial weight value being indicative of a relative importance of each one of the plurality of heart sound signals; adjusting the initial weight value to obtain a final weight value for each one of the plurality of heart sound signals; and identifying the given ones of the heart sound signals based on the final weight value.

50. The system of claim 49, wherein said assigning the initial weight value is performed using a K-means clustering method.

51. The system of claim 49 or 50, wherein said adjusting the initial weight value is performed using a gradient boosting method.

52. The system of any one of claims 48 to 51, wherein said identifying the potential pathology comprises: extracting features from the given ones of the heart sound signals; and analyzing the features to identify the pathology.

53. The system of claim 52, wherein said extracting the features is performed using a discrete wavelet transform method.

54. The system of claim 52 or 53, wherein said analyzing the features is performed using one of: at least one convolutional neural network (CNN); at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

55. The system of any one of claims 48 to 54, wherein the processor is further configured for denoising the heart sound signals, thereby obtaining denoised sound signals, said selecting the given ones of the heart sound signals being performed amongst the denoised sound signals.

56. The system of claim 55, wherein the processor is further configured for receiving at least one reference sound signal indicative of ambient noise, said denoising being performed using the at least one reference sound signal.

57. The system of any one of claims 48 to 56, wherein the processor is further configured for: receiving at least one electrocardiogram signal being synchronized with the heart sound signals; identifying at least one relevant temporal portion of the at least one electrocardiogram (ECG) signal; and identifying, for each one of the plurality of heart sound signals, at least one portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of heart sound signal portions, said selecting the given ones of the heart sound signals being performed amongst the plurality of the heart sound signal portions.

58. The system of claim 57, wherein said identifying the potential pathology is performed further based on the at least one ECG signal.

59. The system of any one of claims 48 to 56, wherein the processor is further configured for receiving a plurality of acceleration signals each detected at a respective location of the body of the subject, said selecting the given ones of the heart sound signals comprising selecting particular signals amongst the plurality of heart sound signals and the plurality of acceleration signals.

60. The system of claim 59, wherein said selecting the particular signals comprises:assigning an initial weight value to each one of the plurality of heart sound signals and the plurality of acceleration signals, the initial weight value being indicative of a relative importance of each one of the plurality of heart sound signals and the plurality of acceleration signals; adjusting the initial weight value to obtain a final weight value for each one of the plurality of heart sound signals and the plurality of acceleration signals; and identifying the particular signals based on the final weight value.

61. The system of claim 60, wherein said assigning the initial weight value is performed using a K-means clustering method.

62. The system of claim 60 or 61, wherein said adjusting the initial weight value is performed using a gradient boosting method.

63. The system of any one of claims 59 to 62, wherein said identifying the potential pathology comprises: extracting features from the particular signals; and analyzing the features to identify the pathology.

64. The system of claim 63, wherein said extracting the features is performed using a discrete wavelet transform method.

65. The system of claim 63 or 64, wherein said analyzing the features is performed using one of: at least one convolutional neural network (CNN); at least one CNN combined with at least one long short-term memory (LSTM) network; at least one recurrent neural network (RNN); and at least one fully connected neural network.

66. The system of any one of claims 59 to 65, wherein the processor is further configured for denoising the heart sound signals, thereby obtaining denoised sound signals, said selecting the particular signals being performed using the denoised sound signals.

67. The system of claim 66, wherein the processor is further configured for receiving at least one reference sound signal indicative of ambient noise, said denoising being performed using the at least one reference sound signal.

68. The system of any one of claims 59 to 67, wherein the processor is further configured for: receiving at least one electrocardiogram signal being synchronized with the heart sound signals and the acceleration signals; identifying at least one relevant temporal portion of the at least one electrocardiogram (ECG) signal; and identifying, for each one of the plurality of heart sound signals, at least one first portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of heart sound signal portions; identifying, for each one of the plurality of acceleration signals, at least one second portion of interest temporally corresponding to the at least one relevant temporal portion of the ECG signal, thereby obtaining a plurality of acceleration signal portions, said selecting the particular signals being performed amongst the plurality of the heart sound signal portions and the plurality of acceleration signal portions.

69. The system of claim 68, wherein said identifying the potential pathology is performed further based on the at least one ECG signal.

70. The system of claim 47, wherein said receiving the plurality of biological sound signals comprises receiving a plurality of lung sound signals and said selecting the given signals comprising selecting given ones of the lung sound signals amongst the plurality of lung sound signals.

71. The system of any one of claims 47 to 70, wherein said outputting the potential pathology comprises providing the potential pathology for display.

72. A system for identifying a pathology, the system comprising: the garment of claim 19 or 20; a plurality of the acoustic apparatuses of any one of claims 1 to 18 each mountable to the garment; a plurality of microphones each mountable to the sensor-mounting means of a respective one of the acoustic apparatuses; and the system of any one of claims 47 to 71, wherein each one of the microphones is configured to output a respective one of the biological sound signals.

Images