Description:FORM 2

The Patents Act 1970
(39 of 1970)
&
The Patent Rules, 2003

COMPLETE SPECIFICATION
(Section 10 and Rule 13)

Title: A BLUETOOTH AUSCULTATION SYSTEM FOR EARLY DETECTION OF CONGENITAL HEART DISEASE OF CHILDREN

Applicant:
IITM PRAVARTAK TECHNOLOGIES FOUNDATION
B5 - 01, B - Block, 5th Floor, IIT Madras Research Park, Kanagam Road, Taramani, Chennai – 600113, India

The following specification particularly describes the invention and the manner in which it is to be performed.
A BLUETOOTH AUSCULTATION SYSTEM FOR EARLY DETECTION OF CONGENITAL HEART DISEASE OF CHILDREN
TECHNICAL FIELD
The present invention relates to the field of health monitoring systems and its techniques. More specifically, the present invention related to a Bluetooth auscultation system for early detection of congenital heart diseases of children in minimum time duration without the need of skilled medical workers.
BACKGROUND
The Congenital Heart Disease (CHD) is a major public health concern, affecting approximately 150,000 children in India every year. The early detection is crucial but often delayed due to limited access to pediatric cardiologists and diagnostic tools, particularly in rural areas. The traditional methods such as echocardiography require trained professionals and are not widely available in primary healthcare settings. The auscultation using a stethoscope is a cost-effective screening method, but it relies on the clinical expertise of medical professionals to identify abnormal heart sounds. In rural and underserved areas, non-medical health workers are responsible for primary healthcare delivery, yet they lack the skills to accurately detect CHD using auscultation alone.
The existing solutions include electronic stethoscopes that record heart sounds, but they do not offer real-time, AI- powered diagnosis or an integrated ecosystem for remote triaging especially using Non-Medical Health Workers. Most AI-based auscultation models are not optimized for pediatric heart sounds, leading to high false-positive rates and low specificity.
Prior art: Eko “https://www.ekohealth.com/” and Littmann “https://www.littmann.com/en-us/home/” address the problem by providing stethoscopes that capture high-fidelity heart sounds with noise reduction and wireless connectivity. These devices record and amplify auscultation data, which can be transmitted to companion apps for visualization, storage, and basic analysis. Eko, in particular, incorporates some machine learning algorithms to flag abnormal sounds, but these features are generally intended for use by trained healthcare professionals. Littmann focuses on precise sound capture and clarity to aid in manual diagnosis by clinicians. Both systems enhance traditional auscultation but are not specifically tailored to pediatric CHD screening or integrated with an automated real-time referral system for non-medical health workers in resource- limited settings.
Thus, there is a need to introduce a Bluetooth auscultation system for early detection of congenital heart diseases of children in an accurate manner. The present invention will overcome the aforementioned problems, limitations and disadvantages in an effective manner.
OBJECTIVE OF THE INVENTION
The primary object of the present invention is to provide a Bluetooth auscultation system for early detection of congenital heart diseases of children using the non-medical professionals in rural areas.
Another object of the present invention is to provide a detection system in affordable price and maximum accurate level.
Another object of the present invention is to avoid the need of internet for early detection of congenital heart diseases of children.
Another object of the present invention is to provide a system specifically for children i.e in the age of 1-12.
Yet another object of the present invention is to avoid the need of skilled medical workers for checking the CHD in the emergency time or rural areas.
These and other objects and advantages of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings.
SUMMARY
The various embodiment of the present invention discloses a Bluetooth auscultation system for early detection of congenital heart disease in offline condition. The said system comprises a power source and a MEMS microphone. The MEMS microphone is configured to capture low frequency heart sounds typically 20-1000 Hz, for detecting the heart murmurs and other anomalies indicative of congenital heart disease of the user.
The signal conditioning unit houses the low noise amplifiers and analog filters to pre-process the raw heart sounds before digitalization. The said signal conditioning unit improves the signal to noise ratio. The analog to digital converter is configured to convert analog heart sound signals into digital data for data transmission and processing. The Bluetooth low energy module is configured to enable wireless transmission of digitized heart sound signals to one or more paired android devices.
The data acquisition module is configured to capture raw heart sound signals from the Bluetooth low energy module and transmit it to the processing platform. The data acquisition module is also configured to: stream high-fidelity digital audio in real-time or records a 10–15 second clip; perform buffering and error checking to prevent data loss during wireless transmission; and, provide synchronization and time stamping to ensure accurate temporal alignment of heart sounds. The pre-processing module is configured to prepare raw audio data for the classification by enhancing signal quality and extracting meaningful features. The classification module is configured to analyse the pediatric heart sound using the plurality of convolution layers, wherein the said plurality of convolution layers may comprise the one or combination of initial noise suppression, separable convolution, dilated convolution and an attention augmented self-attention layer.
The post processing module comprises a threshold calibration module and a confidence scoring module, wherein the threshold calibration module is configured to: set a decision threshold on the classification output probability to maximize sensitivity; and, allow the system to minimize false negatives, crucial for early screening.
The said confidence scoring module is configured to: calculate a confidence score or uncertainty estimate for each prediction; support decision support systems by indicating prediction reliability; and, help the users to decide on further medical referral or repeated testing.
The interface module is configured to serve as the user interaction and processing platform by means of receiving, analysing and displaying heart sound data. The said interface module can be accessed by the user handheld devices, wherein the said interface module also configured to: establish Bluetooth connection, receive live heart sound streams, and manage recording duration (typically 10–15 seconds); apply digital bandpass filters (20–1000 Hz), segment cardiac cycles, and convert raw audio into Mel-spectrograms and MFCCs suitable for classification input; run the custom classification model optimized for mobile deployment and produce a probability score indicating CHD risk; access with simple instructions and clear output messages by the non-medical health workers for minimizing the need for specialized workers; and, store or upload screening results securely, supporting integration into healthcare databases, telemedicine systems or handheld devices.
These and other aspects of the embodiments herein will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following descriptions, while indicating preferred embodiments and numerous specific details thereof, are given by way of illustration and not of limitation. Many changes and modifications may be made within the scope of the embodiments herein without departing from the spirit thereof, and the embodiments herein include all such modifications.
BRIEF DESCRIPTION OF THE DRAWINGS
The other objects, features and advantages will occur to those skilled in the art from the following description of the preferred embodiment and the accompanying drawings in which:
Fig 1 illustrates the schematic flowchart of the Bluetooth auscultation system for early detection of congenital heart disease, according to an embodiment of the present invention.
Fig 2 illustrates the photographic view of the Bluetooth auscultation system, according to an embodiment of the present invention.
Although the specific features of the present invention are shown in some drawings and not in others. This is done for convenience only as each feature may be combined with any or all of the other features in accordance with the present invention.
10- A Bluetooth Auscultation System/Device for Early Detection of Congenital Heart Disease, 11- Power Source, 12- MEMS Microphone, 13- Signal Conditioning Unit, 14- Analog to Digital Converter, 15- Bluetooth Low Energy Module, 16- Button, 17- Data Acquisition Module, 18- Pre-Processing Module, 19- Classification Module, 20- Post Processing Module & 21- Interface Module.
DETAILED DESCRIPTION
The various embodiments and the other advancements and features are illustrated with the reference to the non-limiting details in the following detailed description. Illustration of processing techniques of well-known components are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
The various embodiment of the present invention discloses a Bluetooth auscultation system (10) for early detection of congenital heart disease in offline condition. The said system (10) comprises a power source, MEMS microphone, signal conditioning unit, analog to digital converter, Bluetooth low energy module, button, data acquisition module, pre-processing module, classification module, post processing module and an interface module. The MEMS microphone (12) is configured to capture low frequency heart sounds typically 20-1000 Hz, for detecting the heart murmurs and other anomalies indicative of congenital heart disease of the user. The signal conditioning unit (13) houses the low noise amplifiers and analog filters to pre-process the raw heart sounds before digitalization. The said signal conditioning unit improves the signal to noise ratio.
Fig 1 illustrates the schematic flowchart of the Bluetooth auscultation system for early detection of congenital heart disease, according to an embodiment of the present invention. The analog to digital converter (14) is configured to convert analog heart sound signals into digital data for data transmission and processing. The Bluetooth low energy module (15) is configured to enable wireless transmission of digitized heart sound signals to one or more paired android devices. The data acquisition module is configured to capture raw heart sound signals from the Bluetooth low energy module and transmit it to the processing platform.
The data acquisition module (17) is also configured to: stream high-fidelity digital audio in real-time or records a 10–15 second clip; perform buffering and error checking to prevent data loss during wireless transmission; and, provide synchronization and time stamping to ensure accurate temporal alignment of heart sounds. The pre-processing module is configured to prepare raw audio data for the classification by enhancing signal quality and extracting meaningful features. The classification module (19) is configured to analyse the pediatric heart sound using the plurality of convolution layers, wherein the said plurality of convolution layers may comprise the one or combination of initial noise suppression, separable convolution, dilated convolution and an attention augmented self-attention layer.
The post processing module (20) comprises a threshold calibration module and a confidence scoring module, wherein the threshold calibration module is configured to: set a decision threshold on the classification output probability to maximize sensitivity; and, allow the system to minimize false negatives, crucial for early screening.
The said confidence scoring module is configured to: calculate a confidence score or uncertainty estimate for each prediction; support decision support systems by indicating prediction reliability; and, help the users to decide on further medical referral or repeated testing.
The interface module (21) is configured to serve as the user interaction and processing platform by means of receiving, analysing and displaying heart sound data. The said interface module can be accessed by the user handheld devices, wherein the said interface module also configured to: establish Bluetooth connection, receive live heart sound streams, and manage recording duration (typically 10–15 seconds); apply digital bandpass filters (20–1000 Hz), segment cardiac cycles, and convert raw audio into Mel-spectrograms and MFCCs suitable for classification input; run the custom classification model optimized for mobile deployment and produce a probability score indicating CHD risk; access with simple instructions and clear output messages by the non-medical health workers for minimizing the need for specialized workers; and, store or upload screening results securely, supporting integration into healthcare databases, telemedicine systems or handheld devices.
Fig 2 illustrates the photographic view of the Bluetooth auscultation system, according to an embodiment of the present invention. In an embodiment, the system (10) is operable in offline condition for analysing the congenital heart disease. The said system/device is having a button which helps to start and stop the operation by the user, wherein the operational status can be identified by the LED indication.
The said pre-processing module (18) comprises: a noise reduction module which applies a digital bandpass filter (20–1000 Hz) to remove frequencies outside the heart sound range, effectively filtering out ambient noise and irrelevant sounds; a segmentation into cardiac cycle conversion module is configured to: detect the boundaries of individual heartbeats (cardiac cycles) within the audio stream; and, use peak detection algorithms or envelope analysis to identify S1 and S2 heart sounds, enabling the model to focus on consistent temporal segments; and, a feature augmentation module is configured to: convert segmented audio into Mel-spectrograms, representing the time-frequency distribution of sound energy on a perceptually meaningful scale; extract Mel-Frequency Cepstral Coefficients (MFCCs), which are compact feature vectors representing the spectral envelope of the heart sounds; and, normalize feature inputs for consistent amplitude and dynamic range.
The initial noise suppression convolution is configured for batch normalization, wherein the separable convolution is configured for efficient temporal-spatial feature extraction, wherein the dilated convolutions is configured to capture wider context without increasing model size; and, wherein the attention-augmented self-attention layer is configured to refine feature maps by emphasizing important temporal-spatial cues.
The present invention (10) addresses critical challenges in early CHD detection by overcoming the delays caused by limited access to pediatric cardiologists and costly diagnostic tools. It empowers non-medical health workers to conduct reliable screenings using a Bluetooth-enabled device, reducing dependency on specialist evaluations. By incorporating a custom-built deep learning CNN model, the system eliminates the inconsistencies and subjectivity of manual auscultation, ensuring accurate heart sound analysis. It also provides real-time decision support, advising immediate referrals to a Public Health Center when abnormalities are detected. Overall, this integrated digital ecosystem offers a scalable, low-cost solution for timely and consistent pediatric CHD screening in resource-limited settings.
The high-sensitivity MEMS microphone (12) optimized for capturing low-frequency heart sounds (typically 20–1000 Hz), which is crucial for detecting heart murmurs and other anomalies indicative of congenital heart disease (CHD). The said microphone is designed with a noise-isolating diaphragm to reduce ambient noise and ensure signal clarity, important since the device must operate in noisy environments like rural clinics or homes. The signal conditioning unit includes low-noise amplifiers and analog filters to preprocess the raw heart sounds before digitization, improving signal-to-noise ratio. The ADC converts analog heart sound signals into digital data for transmission and processing. The BLE module (15) enables wireless transmission of digitized heart sound signals to a paired Android device or web interface in real-time. The BLE is selected for low power consumption, essential for portable, battery-powered devices. The power source is the rechargeable Lithium-Polymer battery for long use during fieldwork. The auscultation device (10) is a dome shaped device which is ergonomically designed for pediatric use, with a small, lightweight casing to comfortably fit on a child's chest. The durable and biocompatible materials ensure safety and ease of cleaning. The device/system Includes minimal controls (e.g., a single button to start/stop recording) and LED indicators for status feedback.
The data acquisition module (17) captures raw heart sound signals from the digital stethoscope and reliably transmits them to the processing platform via Bluetooth. The said module: establishes and manages a Bluetooth Low Energy (BLE) connection with the MEMS microphone; streams high-fidelity digital audio in real-time or records a 10–15 second clip; performs buffering and error checking to prevent data loss during wireless transmission; provides synchronization and time stamping to ensure accurate temporal alignment of heart sounds; and, designed to be lightweight and efficient to conserve mobile device battery and memory.
The pre-processing module (18) prepares raw audio data for the CNN classifier by enhancing signal quality and extracting meaningful features. Noise Reduction: Applies a digital band pass filter (20–1000 Hz) to remove frequencies outside the heart sound range, effectively filtering out ambient noise and irrelevant sounds (e.g., speech, movement noise). The Adaptive noise reduction techniques may be integrated to further suppress background sounds in real-world noisy environments. Segmentation into Cardiac Cycles: Detects the boundaries of individual heartbeats (cardiac cycles) within the audio stream. It uses peak detection algorithms or envelope analysis to identify S1 and S2 heart sounds, enabling the model to focus on consistent temporal segments. This segmentation is critical to standardize input data and improve CNN learning of temporal patterns. Feature Augmentation: Converts segmented audio into Mel-spectrograms, representing the time-frequency distribution of sound energy on a perceptually meaningful scale. It extracts Mel-Frequency Cepstral Coefficients (MFCCs), which are compact feature vectors representing the spectral envelope of the heart sounds. These features help the CNN focus on subtle audio characteristics relevant to CHD detection. It also normalizes feature inputs for consistent amplitude and dynamic range.
The user interface (21) can run on any Windows (tablet or laptops) or supports modern web browsers. The auscultation device is equipped with Bluetooth capabilities to connect seamlessly with the digital stethoscope. The sufficient processing power to run the lightweight CNN model locally or transmit data for cloud processing. The user interface comprises the following operations in a predefined sequence.
Data Acquisition Module: Establishes Bluetooth connection, receives live heart sound streams, and manages recording duration (typically 10–15 seconds).
Preprocessing: Applies digital bandpass filters (20–1000 Hz), segments cardiac cycles, and converts raw audio into Mel-spectrograms and MFCCs suitable for CNN input.
Inference Engine: Runs the custom CNN model optimized for mobile deployment (e.g., TensorFlow Lite), producing a probability score indicating CHD risk.
User Interface: Designed for non-medical health workers with simple instructions and clear output messages (“Refer to Hospital” or “No action”), minimizing the need for specialized training.
Data Management: Stores or uploads screening results securely, supporting integration into healthcare databases or telemedicine systems.
Usage of the system in different cases:
Example 1: Community Health Screening by ASHA Workers
An Accredited Social Health Activist (ASHA) may conduct routine health check-ups in a rural village. With no formal medical training, anyone may use the device to screen infants and young children for early signs of CHD.
Process:
Device Placement: ASHA places the digital stethoscope on the child’s chest (typically near the apex of the heart).
Data Acquisition: The web app or processing unit receives heart sound data via Bluetooth in real time.
Pre-processing:
Filters noise (20–1000 Hz)
Segments the audio into cardiac cycles
Converts it into Mel-spectrograms and MFCCs
Classification:
The lightweight CNN model runs directly on the mobile device, processing the features to predict the probability of CHD.
Decision Output:
Based on a sensitivity-optimized threshold, the app shows:
“Refer to Hospital” if CHD is suspected
“No action” if normal
Human-in-the-Loop:
The health worker simply follows the interface guidance, without the need of clinical interpretation.
Example 2: Primary Health Center (PHC)
A nurse or general physician in a PHC uses the device to screen all newborns and infants during scheduled immunization or wellness visits.
Routine Use: The device becomes part of the standard pediatric exam process.
User Interface: The health worker uses the web interface in any browser.
Screening:
Audio is preprocessed and analyzed in real time.
CHD probability is calculated instantly.
Confidence score can assist physician review.
Result Recording: Results are optionally stored in the child’s digital health record or sent to a central database.
Example 3: Pediatric Research and Surveillance Programs
Scenario:
A pediatric research team or NGO conducts a large-scale CHD prevalence survey across multiple districts.
Working:
Batch Screening:
Field workers use the device across hundreds of schools and Anganwadis.
Offline Capability:
App works offline; audio + inference results are stored and synced later.
Cloud Analytics (Optional):
Spectrograms and predictions are uploaded for post-hoc analysis, refinement of the model, or population-level CHD mapping.
Quality Control:
Confidence scoring and raw audio can be reviewed by expert pediatricians for validation.
Example 4: Home Monitoring for High-Risk Infants
Scenario:
A child previously diagnosed with a minor heart defect is under home monitoring by trained parents or caregivers.
Periodic Monitoring:
Parents use the device weekly or monthly to monitor for changes in heart sounds.
Interface Based Alerts:
CNN flags any abnormal progression or irregular patterns.
Remote Supervision:
Results are shared with a remote cardiologist through the web dashboard for clinical interpretation.
The present invention (10) solves these problems by integrating a Bluetooth-enabled chest auscultation device with a custom- built deep learning CNN model to provide accurate, real-time analysis of pediatric heart sounds. It enables non- medical health workers to capture and transmit heart sound data via Bluetooth to a handheld device, where the CNN model processes the audio to identify signs of CHD with high consistency and precision. This automated system eliminates the need for specialist interpretation and mitigates the inconsistencies of manual auscultation, while its built-in decision support promptly advises referrals to public health centers when abnormalities are detected.
The method of operation of the system/device (10) comprising the steps of:
- Pressing the power button (16) to switch ON the system;
- Monitoring the active status by the LED indication;
- Pairing of the auscultation system with the handheld devices’ user interface by means of Bluetooth;
- Placing the rubberized microphone gently over the apex area of the child’s chest (typically under the left nipple line);
- Ensuring the child is relatively still and breathing normally;
- Holding the device in place with steady pressure i.e no need to press hard;
- Recording, by a start recording button in the interface, the heart sound for 10-15 seconds;
- Transmitting audio in real-time using Bluetooth;
- Applying the noise filtering, segmentation, and feature extraction (Mel-spectrogram) in the background;
- Sending the spectrogram to the embedded lightweight classification model; Providing a probability score (e.g., 0.91 = high CHD risk);
- Applying a threshold (e.g., >0.85 triggers alert);
- Providing the outcome in the form of “no action needed” or “refer to hospital”;
- Providing an optional confidence score or signal quality indicator;
- Storing of Audio sample (for review), Spectrogram, CHD probability score and Decision outcome; and,
- Pressing of button to turn off the system.
The present invention (10) offers a low-cost, scalable solution that bridges the gap between early screening and specialist care, ensuring timely and effective intervention in resource-limited settings. The novelty of the proposed invention lies in its integrated, end-to-end solution for pediatric CHD screening, combining a Bluetooth-enabled chest auscultation device with a custom-built deep learning CNN model specifically optimized for pediatric heart sounds.
Unlike traditional methods that depend on manual auscultation by specialists or generic AI models not tailored to children, this invention captures high-quality heart sound data from children aged 1-12 and processes it in real-time on a handheld device. It provides immediate, actionable diagnostic feedback to non-medical health workers, enabling early intervention by advising referrals to Public Health Centers when necessary. Moreover, the invention's unique architecture and training methodology reduce false positives and inconsistencies, making it a scalable, low-cost solution for remote and resource-limited settings.
It is noted that the above-described examples of the present invention is for the purpose of illustration only. Although the present invention has been described in conjunction with a specific example thereof, numerous modifications may be possible without materially departing from the teachings and advantages of the subject matter described herein. Other substitutions, modifications and changes may be made without departing from the spirit of the present solution. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and/or all of the steps of any method or process so disclosed, may be combined in any combination, except combinations where at least some of such features and/or steps are mutually exclusive. Although the embodiments herein are described with various specific embodiments, it will be obvious for a person skilled in the art to practice the embodiments herein with modifications.
, Claims:CLAIMS
We claim,
1. A Bluetooth auscultation system (10) for early detection of congenital heart disease of children, comprising:
a power source (11);
a MEMS microphone (12) is configured to capture low frequency heart sounds typically 20-1000 Hz, for detecting the heart murmurs and other anomalies indicative of congenital heart disease of the user;
a signal conditioning unit (13) houses the low noise amplifiers and analog filters to pre-process the raw heart sounds before digitalization, wherein the said unit improves the signal to noise ratio;
an analog to digital converter (14) is configured to convert analog heart sound signals into digital data for data transmission and processing;
a Bluetooth low energy module (15) is configured to enable wireless transmission of digitized heart sound signals to one or more paired android devices;
a data acquisition module (17) is configured to capture raw heart sound signals from the Bluetooth low energy module and transmit it to the processing platform, wherein the data acquisition module is configured to: stream high-fidelity digital audio in real-time or records a 10–15 second clip; perform buffering and error checking to prevent data loss during wireless transmission; and, provide synchronization and time stamping to ensure accurate temporal alignment of heart sounds;
a pre-processing module (18) is configured to prepare raw audio data for the classification by enhancing signal quality and extracting meaningful features;
a classification module (19) is configured to analyse the pediatric heart sound using the plurality of convolution layers, wherein the said plurality of convolution layers may comprise the one or combination of initial noise suppression, separable convolution, dilated convolution and an attention augmented self-attention layer;
a post processing module (20) comprises a threshold calibration module and a confidence scoring module, wherein the threshold calibration module is configured to: set a decision threshold on the classification output probability to maximize sensitivity; and, allow the system to minimize false negatives, crucial for early screening;
wherein the confidence scoring module is configured to: calculate a confidence score or uncertainty estimate for each prediction; support decision support systems by indicating prediction reliability; and, help the users to decide on further medical referral or repeated testing; and,
an interface module (21) is configured to serve as the user interaction and processing platform by means of receiving, analysing and displaying heart sound data, wherein the said module can be accessed by the user handheld devices, wherein the said interface module also configured to: establish Bluetooth connection, receive live heart sound streams, and manage recording duration (typically 10–15 seconds); apply digital bandpass filters (20–1000 Hz), segment cardiac cycles, and convert raw audio into Mel-spectrograms and MFCCs suitable for classification input; run the custom classification model optimized for mobile deployment and produce a probability score indicating CHD risk; access with simple instructions and clear output messages by the non-medical health workers for minimizing the need for specialized workers; and, store or upload screening results securely, supporting integration into healthcare databases, telemedicine systems or handheld devices.
2. The Bluetooth auscultation system for early detection of congenital heart disease of children as claimed in claim 1, wherein the system (10) is operable in offline condition for analysing the congenital heart disease.
3. The Bluetooth auscultation system for early detection of congenital heart disease of children as claimed in claim 1, wherein the said system/device (10) is having a button which helps to start and stop the operation by the user, wherein the operational status can be identified by the LED indication.
4. The Bluetooth auscultation system for early detection of congenital heart disease of children as claimed in claim 1, wherein the pre-processing module (18) comprises:
a noise reduction module which applies a digital bandpass filter (20–1000 Hz) to remove frequencies outside the heart sound range, effectively filtering out ambient noise and irrelevant sounds;
a segmentation into cardiac cycle conversion module is configured to: detect the boundaries of individual heartbeats (cardiac cycles) within the audio stream; and, use peak detection algorithms or envelope analysis to identify S1 and S2 heart sounds, enabling the model to focus on consistent temporal segments; and,
a feature augmentation module is configured to: convert segmented audio into Mel-spectrograms, representing the time-frequency distribution of sound energy on a perceptually meaningful scale; extract Mel-Frequency Cepstral Coefficients (MFCCs), which are compact feature vectors representing the spectral envelope of the heart sounds; and, normalize feature inputs for consistent amplitude and dynamic range.
5. The Bluetooth auscultation system for early detection of congenital heart disease of children as claimed in claim 1, wherein the initial noise suppression convolution is configured for batch normalization, wherein the separable convolution is configured for efficient temporal-spatial feature extraction, wherein the dilated convolutions is configured to capture wider context without increasing model size; and, wherein the attention-augmented self-attention layer is configured to refine feature maps by emphasizing important temporal-spatial cues.
6. A method of operation of the Bluetooth auscultation system (10) for early detection of congenital heart disease of children, comprising:
- Pressing the power button (16) to switch ON the system;
- Monitoring the active status by the LED indication;
- Pairing of the auscultation system with the handheld devices’ user interface by means of Bluetooth;
- Placing the rubberized microphone gently over the apex area of the child’s chest (typically under the left nipple line);
- Ensuring the child is relatively still and breathing normally;
- Holding the device (10) in place with steady pressure i.e no need to press hard;
- Recording, by a start recording button in the interface, the heart sound for 10-15 seconds;
- Transmitting audio in real-time using Bluetooth;
- Applying the noise filtering, segmentation, and feature extraction (Mel-spectrogram) in the background;
- Sending the spectrogram to the embedded lightweight classification model; Providing a probability score (e.g., 0.91 = high CHD risk);
- Applying a threshold (e.g., >0.85 triggers alert);
- Providing the outcome in the form of “no action needed” or “refer to hospital”;
- Providing an optional confidence score or signal quality indicator;
- Storing of Audio sample (for review), Spectrogram, CHD probability score and Decision outcome; and,
- Pressing of button to turn off the system.