Respiratory sound classification utilizing human auditory-based feature extraction.

Physica Scripta, Journal Year: 2025, Volume and Issue: 100(4), P. 046003 - 046003

Published: Feb. 19, 2025

Abstract A major worldwide health concern is chronic respiratory diseases (CRDs), which include disorders including asthma, pulmonary hypertension, occupational lung diseases, and obstructive disease (COPD). Improving clinical results treatment efficacy requires an early precise diagnosis. In order to classify sounds, this study presents a novel framework that incorporates auditory-inspired characteristics, such as Mel-Frequency Cepstral Coefficients (MFCCs), Mel Spectrograms, Cochleograms, into CNN-LSTM architecture. The uses sophisticated feature extraction techniques in conjunction with strong data augmentation approaches address the issue of class imbalance guarantee thorough representation variety sound patterns. Using Respiratory Sound Database, suggested model was assessed showed remarkable performance, obtaining F1 score 98.94%, accuracy 98.90%, specificity 99.80%, sensitivity ICBHI 99.40%. These findings demonstrate model’s potential reliable efficient tool for identification evaluation CRDs, would significantly improve patient care management illnesses. outstanding performance further emphasizes importance settings, enabling improved conditions.

Language: Английский

---

Skin cancer identification utilizing deep learning: A survey

IET Image Processing, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 2, 2024

Abstract Melanoma, a highly prevalent and lethal form of skin cancer, has significant impact globally. The chances recovery for melanoma patients substantially improve with early detection. Currently, deep learning (DL) methods are gaining popularity in assisting the identification melanoma. Despite their high performance, relying solely on an image classifier undermines credibility application makes it difficult to understand rationale behind model's predictions highlighting need Explainable AI (XAI). This study provides survey cancer using DL techniques utilized studies from 2017 2024. Compared existing studies, authors address latest related covering several public datasets focusing segmentation, classification based convolutional neural networks vision transformers, explainability. analysis comparisons will be beneficial researchers developers this area, identify suitable used automated classification. Thereby, findings can implement support applications advancing diagnosis process.

Language: Английский

---

Autism spectrum disorder identification using multi‐model deep ensemble classifier with transfer learning

Expert Systems, Journal Year: 2024, Volume and Issue: unknown

Published: May 16, 2024

Abstract Identifying autism spectrum disorder (ASD) symptoms accurately is a challenging task. The traditional subjective diagnostic process of ASD relies on time‐consuming behavioural and psychological observations. In this study, we introduce an ensemble learning‐based classification model using open‐access database focusing functional magnetic resonance imaging (fMRI). We propose novel multi‐model classifier (MMEC) multisite (MSEC) with transfer learning (TL) for to improve the prediction accuracy. MMEC utilizes four base classifiers, Inception V3, ResNet50, MobileNet, DenseNet boost performance individual convolutional neural network (CNN) models. MSEC combined classifiers trained from different data sites. evaluate two models averaging, weighted stacking methods. proposed shows state art compared MSEC, improving accuracy by 3.25%. obtained results have shown 97.82%, 97.78% methods, respectively, multi‐site datasets. performed better than single dataset. opens new paradigm design universal framework.

Language: Английский

---

Auscultation-Based Pulmonary Disease Detection through Parallel Transformation and Deep Learning

Bioengineering, Journal Year: 2024, Volume and Issue: 11(6), P. 586 - 586

Published: June 8, 2024

Respiratory diseases are among the leading causes of death, with many individuals in a population frequently affected by various types pulmonary disorders. Early diagnosis and patient monitoring (traditionally involving lung auscultation) essential for effective management respiratory diseases. However, interpretation sounds is subjective labor-intensive process that demands considerable medical expertise, there good chance misclassification. To address this problem, we propose hybrid deep learning technique incorporates signal processing techniques. Parallel transformation applied to adventitious sounds, transforming sound signals into two distinct time-frequency scalograms: continuous wavelet transform mel spectrogram. Furthermore, parallel convolutional autoencoders employed extract features from scalograms, resulting latent space fused feature pool. Finally, leveraging long short-term memory model, used as input classifying Our work evaluated using ICBHI-2017 dataset. The experimental findings indicate our proposed method achieves promising predictive performance, average values accuracy, sensitivity, specificity, F1-score 94.16%, 89.56%, 99.10%, respectively, eight-class diseases; 79.61%, 78.55%, 92.49%, 78.67%, four-class 85.61%, 83.44%, 84.21%, binary-class (normal vs. abnormal) sounds.

Language: Английский

---

Navigating the Challenges and Opportunities of Tiny Deep Learning and Tiny Machine Learning in Lung Cancer Identification

Al-Nahrain Journal for Engineering Sciences, Journal Year: 2025, Volume and Issue: 28(1), P. 97 - 120

Published: April 7, 2025

Lung cancer is the most common dangerous disease that, if treated late, can lead to death. It more likely be successfully discovered at an early stage before it worsens. Distinguishing size, shape, and location of lymphatic nodes identify spread around these nodes. Thus, identifying lung remarkably helpful for doctors. diagnosed by expert doctors; however, their limited experience may misdiagnosis cause medical issues in patients. In line computer-assisted systems, many methods strategies used predict malignancy level that plays a significant role provide precise abnormality detection. this paper, use modern learning machine-based approaches was explored. More than 70 state-of-the-art articles (from 2019 2024) were extensively explored highlight different machine deep (DL) techniques models detection, classification, prediction cancerous tumors. The efficient model Tiny DL must built assist physicians who are working rural centers swift rapid diagnosis cancer. combination lightweight Convolutional Neural Networks resources could produce portable with low computational cost has ability substitute skill doctors needed urgent cases.

Language: Английский

---

Non-invasive diagnosis of lung diseases via multimodal feature extraction from breathing audio and chest dynamics

Computers in Biology and Medicine, Journal Year: 2025, Volume and Issue: 191, P. 110182 - 110182

Published: April 10, 2025

Language: Английский

---

Artificial Intelligence Models for Pediatric Lung Sound Analysis: Systematic Review and Meta-Analysis

Journal of Medical Internet Research, Journal Year: 2025, Volume and Issue: 27, P. e66491 - e66491

Published: April 18, 2025

Background Pediatric respiratory diseases, including asthma and pneumonia, are major causes of morbidity mortality in children. Auscultation lung sounds is a key diagnostic tool but prone to subjective variability. The integration artificial intelligence (AI) machine learning (ML) with electronic stethoscopes offers promising approach for automated objective sound. Objective This systematic review meta-analysis assess the performance ML models pediatric sound analysis. study evaluates methodologies, model performance, database characteristics while identifying limitations future directions clinical implementation. Methods A search was conducted Medline via PubMed, Embase, Web Science, OVID, IEEE Xplore studies published between January 1, 1990, December 16, 2024. Inclusion criteria as follows: developing classification defined database, physician-labeled reference standard, reported metrics. Exclusion focusing on adults, cardiac auscultation, validation existing models, or lacking Risk bias assessed using modified Quality Assessment Diagnostic Accuracy Studies (version 2) framework. Data were extracted design, dataset, methods, feature extraction, tasks. Bivariate performed binary tasks, wheezing abnormal detection. Results total 41 met inclusion criteria. most common task detection sounds, particularly wheezing. Pooled sensitivity specificity wheeze 0.902 (95% CI 0.726-0.970) 0.955 0.762-0.993), respectively. For detection, pooled 0.907 0.816-0.956) 0.877 0.813-0.921). frequently used extraction methods Mel-spectrogram, Mel-frequency cepstral coefficients, short-time Fourier transform. Convolutional neural networks predominant model, often combined recurrent residual network architectures. However, high heterogeneity dataset size, annotation evaluation observed. Most relied small, single-center datasets, limiting generalizability. Conclusions show accuracy analysis, face due heterogeneity, lack standard guidelines, limited external validation. Future research should focus standardized protocols development large-scale, multicenter datasets improve robustness

Language: Английский

---

Acumm: An Effective Framework for Tackling Domain Mismatch and Class Imbalance in Respiratory Sound Classification

Published: Jan. 1, 2025

Language: Английский

---

Vdd: voice deepfake detection with three-channel acoustic representations and advanced split-attention networks

Signal Image and Video Processing, Journal Year: 2025, Volume and Issue: 19(7)

Published: May 12, 2025

Language: Английский

---

Artificial Intelligence in Respiratory Health: A Review of AI-Driven Analysis of Oral and Nasal Breathing Sounds for Pulmonary Assessment

Electronics, Journal Year: 2025, Volume and Issue: 14(10), P. 1994 - 1994

Published: May 14, 2025

Continuous monitoring of pulmonary function is crucial for effective respiratory disease management. The COVID-19 pandemic has also underscored the need accessible and convenient diagnostic tools health assessment. While traditional lung sound auscultation been primary method evaluating function, emerging research highlights potential nasal oral breathing sounds. These sounds, shaped by upper airway, serve as valuable non-invasive biomarkers detection. Recent advancements in artificial intelligence (AI) have significantly enhanced analysis enabling automated feature extraction pattern recognition from spectral temporal characteristics or even raw acoustic signals. AI-driven models demonstrated promising accuracy detecting conditions, paving way real-time, smartphone-based monitoring. This review examines AI-enhanced analysis, discussing methodologies, available datasets, future directions toward scalable solutions.

Language: Английский

---