Granted Patent B2

US 12,653,488 · App. 17/074,749 · Granted Jun 16, 2026

Inventor: Satish Somayya Jeevannavar (Bangalore, IN)

Assignee: Ai Health Highway India Private Limited

A61B7/045A61B5/0006A61B5/002A61B5/316A61B5/339A61B5/6823A61B5/7203A61B5/7267A61B5/7275A61B5/7405

Abstract

Claims (24)

1 . An electronic stethoscope system comprising:

a) a chest piece comprising a diaphragm, a bell, a micro-electro-mechanical (MEMS) microphone, an electronic amplifier, and a plurality of vibration transducers and acoustic transducers forming a 3-dimensional array, the chest piece is configured to:

detect contact of the chest piece with an upper torso region of a human subject using one or more touch-based sensors integrated into the chest piece to dynamically trigger capturing of chest sounds and bodily vibrations of the human subject;

capture sound signals corresponding to the chest sounds and vibration signals corresponding to the bodily vibrations via a sound capture component associated with the chest piece;

transduce the sound signals and the vibration signals into electrical signals via a microphone array within the chest piece, wherein the microphone array comprises a set of vibration transducers and acoustic transducers, wherein the vibration transducers convert the vibration signals into acoustic signals, and wherein the electrical signals are amplified via an electronic amplifier communicatively connected to the microphone array to generate amplified electrical signals; and

perform signal pre-processing of the amplified electrical signals via a digital signal processing (DSP) unit within a processing unit, wherein the signal preprocessing comprises at least one of a noise filtration, beamforming, noise cancellation, a signal normalization, and conversion of the amplified electrical signals into electronic signal waveforms;

b) the processing unit communicatively coupled to the chest piece, comprising:

a data interpretation module to:

analyze the electronic signal waveforms received from the DSP unit, by comparing the electronic signal waveforms with pre-defined waveforms based on standardized data;

classify, using one or more artificial intelligence (AI) and machine learning (ML) classification models within the data interpretation module, one or more cardiac conditions comprising at least one of an aortic stenosis, a mitral regurgitation, and normal heart sounds, using pre-validated clinical datasets, wherein the classification comprises:

analyzing input signal features corresponding to the electronic signal waveforms, the input signal features comprising wavelet transforms, time-domain, frequency-domain, and time-frequency features for clinical relevance;

determining at least one of the normal heart sounds and abnormal heart sounds based on at least one of the wavelet features and frequency band ranges in the electronic signal waveforms, wherein the AI and ML classification models comprise an Artificial Neural Network (ANN) based classification model configured to perform clinical diagnostic decision-making using the input signal features, wherein determining at least one of the normal heart sounds and the abnormal heart sounds comprises:

 analyzing the input signal features to classify the input signal features as at least one of a normal pattern, an abnormal pattern, and a specific clinical condition-based pattern, wherein the ANN based classification model uses discrete wavelet transform, fuzzy logic, and fast Fourier transform (FFT) techniques to differentiate patterns in the input signal features;

 classifying at least one of the normal heart sounds and the abnormal heart sounds, using supervised machine learning techniques comprising a support vector machine (SVM)-based classification model, based on the at least one of the wavelet features and the frequency band ranges in the signal waveforms, for the wavelet transforms features and the frequency-domain features in the input signal features;

 analyzing one or more heart sounds comprising at least one of time features and frequency features extracted by linear decomposition and tiling partition of the time-frequency plane, time interval durations between systole and diastole, and frequency-based features and patterns with logistic regression and Gaussian distribution-based probabilities, via a hybrid classification model combining supervised learning and a hidden Markov model (HMM) for the time-frequency features;

 classifying the one or more heart sounds as the normal heart sound or abnormal heart sound based on the analyzed one or more heart sounds comprising at least one of time features and frequency features, time interval durations between systole and diastole, and frequency-based features and patterns, using an unsupervised clustering-based classification model comprising k-Nearest Neighbors (kNN), based on the extracted time features and the frequency features by linear decomposition and tiling partition of the time-frequency plane; and

 classifying clinical diagnostic data using convolutional neural networks (CNNs) for two-dimensional (2D) time-frequency graphs and analyzing a time-domain signal using a deep learning framework comprising Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) models for time-domain features; and

generating at least one of one or more electronic wave forms and time parameters, frequency parameters and time-frequency parameters, based on determining the normal heart sounds, the abnormal heart sounds, and differentiated and classified the normal heart sounds, and the abnormal heart sounds, wherein the generated at least one of one or more electronic wave forms and time parameters, frequency parameters and time-frequency parameters is used to train the one or more AI and ML classification models; and

an integrated diagnostic module to determine a diagnosis corresponding to the one or more cardiac conditions using the trained one or more AI and ML classification models, by comparing the electronic signal waveforms with information from electronic medical recordkeeping (EMR); and

c) an output unit communicatively coupled to the processing unit configured to:

output at least one of a visual diagnostic output and a textual diagnostic output corresponding to the diagnosis based on the generated at least one of one or more electronic wave forms and time parameters, frequency parameters and time-frequency parameters, via at least one of a display, a display device, one or more cloud servers and one or more remote facilities, wherein the at least one of a visual diagnostic output and a textual diagnostic output comprises at least one of textual indication, waveforms, acoustics, augmented reality, virtual reality and mixed reality.

2 . The electronic stethoscope system of claim 1 , wherein the noise cancellation comprises filtering unwanted sound waves from at least one of the bell and the diaphragm moving across the skin, hair, or clothing of the human subject, and reducing ambient noise.

3 . The electronic stethoscope system of claim 1 , further comprising an additional microphone to record the voice of a user with voice-to-text conversion, for Electronic Medical Recordkeeping (EMR).

4 . The electronic stethoscope system of claim 1 , further comprising a payment gateway interface for transactions and virtual services.

Continuity (3)

References Cited (23)