Automated Lung Sound Classification Without Flowmeter Reference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting pulmonary abnormalities using lung sounds are manual, unreliable due to aperiodic nature of signals, and often fail to identify abnormalities consistently, especially in developing countries with skewed doctor-to-patient ratios, and lack physically interpretable features for medical screening.
Innovation Solution
A processor-based method that preprocesses auscultation sound signals by resampling and removing heart sounds, then extracts spectral, spectrogram, and wavelet features, selecting discriminative features for classification, allowing for automated identification of healthy vs. abnormal lung sounds without manual labeling of respiratory cycles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual auscultation is used to screen lung sounds, then doctor expertise can identify abnormalities, but the process is time-consuming and unreliable due to aperiodic signals
Solution Approach 1:
The patent replaces manual mechanical auscultation with an automated computational system that processes lung sound signals using digital signal processing techniques, spectral analysis, and machine learning algorithms to detect abnormalities automatically
Solution Approach 2:
The system enables self-diagnosis capability by automating the analysis process, allowing the computer system to independently identify pulmonary abnormalities without requiring continuous doctor intervention for each case
2Extent of automation
If deep learning is used to identify features for lung sound analysis, then automation is achieved, but features are not physically interpretable and not relevant to medical fraternity
Solution Approach 1:
The patent transforms the feature extraction approach by changing from abstract deep learning features to physically interpretable parameters such as spectral power distribution, frequency band energies, and temporal patterns that are meaningful to medical professionals
Solution Approach 2:
The system introduces an intermediary layer of physically meaningful acoustic features that bridge the gap between raw signals and clinical interpretation, making the automated analysis results relevant and understandable to the medical fraternity
3Measurement precision
If reference flowmeter signals are used to label respiratory phases, then respiratory phases can be identified, but the method complexity increases and requires additional equipment
Solution Approach 1:
The patent extracts respiratory phase information directly from the lung sound signals themselves through signal processing techniques, eliminating the need for external reference flowmeter equipment while maintaining accurate phase identification
Solution Approach 2:
The system achieves multi-functionality by using the same audio recording device to capture both lung sounds and derive respiratory phase information, eliminating the need for separate specialized equipment
4Measurement precision
If manual separation of respiratory phases is performed, then accurate labeling is achieved, but the productivity is low due to skewed doctor-to-patient ratio
Solution Approach 1:
The patent replaces manual separation and labeling processes with automated computational algorithms that process lung sound signals to identify and label respiratory phases, dramatically increasing diagnosis throughput
Solution Approach 2:
The system performs preliminary automated separation and labeling of respiratory phases before clinical analysis, preparing the data in advance and enabling doctors to focus on interpretation rather than manual processing
Data Source
AI summary
Identification of pulmonary diseases involves accurate auscultation as well as elaborate and expensive pulmonary function tests. Also, there is a dependency on a reference signal from a flowmeter or need for labelled respiratory phases. The present disclosure provides extraction of frequency and time-frequency domain lung sound features such as spectral and spectrogram features respectively that enable classification of healthy and abnormal lung sounds without the dependencies of prior art. Furthermore extraction of wavelet and cepstral features improves accuracy of classification. The lung sound signals are pre-processed prior to feature extraction to eliminate heart sounds and reduce computational requirements while ensuring that information providing adequate discrimination between healthy and abnormal lung sounds is not lost.