Pith. sign in

REVIEW 4 major objections 6 minor 69 references

Improving Deep Learning-based Respiratory Sound Analysis with Frequency Selection and Attention Mechanism

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A frequency-selection module for lung-sound AI boosts accuracy on two benchmark datasets while halving computation.

desk verdict Useful FBS trick with honest held-out evaluation, but the SOTA claims overreach on three comparison rows. read the letter →

arxiv 2507.20052 v1 pith:TOKTHQ4H submitted 2025-07-26 cs.SD cs.LG

Nouhaila Fraihi , Ouassim Karrakchou , Mounir Ghogho This is my paper
classification cs.SDcs.LG
keywords respiratorysoundclassificationfrequencybandselectiontemporalself-attentionCNNaudiospectrogramtransformerGrad-CAMSPRSoundICBHI
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that respiratory sound classification can be made both more accurate and far cheaper by teaching the model to ignore useless frequency bands. It proposes a compact CNN with a temporal self-attention module (CNN-TSA) and an importance-based frequency band selection (FBS) routine that removes noisy Mel-spectrogram bands before classification. The authors claim this combination sets new state-of-the-art results on the SPRSound-2022/2023 pediatric datasets and competitive or best results on ICBHI-2017, while reducing FLOPs by up to 50%. If true, the FBS module is a practical, architecture-agnostic drop-in that helps bring automated lung-sound analysis to real-time, resource-limited settings.

What carries the argument

The load-bearing mechanism is the importance-based frequency band selection (FBS) module, which scores each of the 64 Mel-frequency bands with $I_f = \mathrm{Mean}[f] - \lambda\,\mathrm{MaxDiff}[f]$, where $\mathrm{Mean}[f]$ is the band's average Grad-CAM attribution across classes and $\mathrm{MaxDiff}[f]$ is the largest inter-class attribution gap. The module iteratively removes the four lowest-scoring bands, retrains with 5-fold cross-validation, and stops when performance degrades, yielding a binary frequency mask. This mask is applied to the input spectrogram, so the downstream CNN-TSA network processes only the informative bands; the temporal self-attention operates on a frequency-aggregated representation to capture long-range temporal dependencies at low cost.

What would settle it

Retrain the strongest compared baselines (e.g., Patch-Mix CL and the proposed CNN-TSA with all 64 bands) on the exact training splits and preprocessing used in this paper and check whether the FBS-masked models still outperform them. If the baselines recover their original scores but the masked models do not, the benchmark claim fails. A complementary test would be to apply the frequency masks learned on one dataset to an unseen respiratory sound dataset and see whether the accuracy gain transfers.

Watch

Extended reading notes

Core claim

The paper's central claim is that a model-driven importance score for each frequency band, computed from Grad-CAM attributions and penalized by inter-class inconsistency, can identify a smaller subset of Mel-spectrogram bands that improves classification accuracy while cutting computation roughly in half. Applied to the authors' CNN-TSA network—CNN6 backbone with frequency aggregation and temporal self-attention—this frequency band selection yields new benchmarks on SPRSound-2022 and SPRSound-2023 and state-of-the-art results on ICBHI-2017, using only 1.1M–4.6M parameters. The same FBS procedure, using Integrated Gradients instead of Grad-CAM, is shown to improve a transformer baseline (Patch-Mix with Audio Spectrogram Transformer) to a new ICBHI record. The authors also report that training separate models for adult and pediatric groups improves accuracy on ICBHI.

Load-bearing premise

The claim of new benchmarks assumes that the compared prior results were obtained under the same training data, preprocessing, and evaluation protocol; if protocol differences drive the gains, the headline improvement would not be attributable to the method.

Editorial extensions

If this is right

  • Respiratory sound classification can run in real time on embedded hardware: the best models use 1.1M–4.6M parameters and about half the FLOPs of their unfiltered counterparts.
  • Frequency band selection acts as a form of input regularization, so other audio-classification tasks with noisy spectra may benefit from attribution-guided band pruning.
  • The FBS module improves both CNN and transformer backbones, making it a drop-in enhancement for existing respiratory sound classifiers.
  • Age-specific training offers a simple accuracy gain on datasets spanning children and adults, without changing the architecture.
  • Retaining only 50–75% of Mel bands is consistently enough for top performance, indicating that much of the spectrogram is redundant or harmful.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper does not specify which training data was used for the SPRSound-2023 evaluation (the dataset is test-only in the challenge), so the reported 'new benchmark' may depend on an unstated training-set choice; a reader should treat that comparison as provisional.
  • FBS's dependence on Grad-CAM (or Integrated Gradients) means the selected mask inherits the attribution method's biases; comparing masks from different attribution methods on the same data could reveal how sensitive the gains are to that choice.
  • The same pruning idea could be tested on other physiological sounds, such as heart or bowel sounds, which also have noisy spectra, though the paper does not evaluate this.
  • The iterative elimination procedure retrains the model many times; a cheaper variant that predicts band importance from a single pass could make the method more practical for large datasets.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript proposes a compact CNN-Temporal Self-Attention (CNN-TSA) architecture for respiratory sound classification, which combines a CNN6 backbone with temporal self-attention applied after frequency aggregation, and an Importance-Based Frequency Band Selection (FBS) module that uses Grad-CAM (or Integrated Gradients for transformer backbones) to iteratively remove low-importance mel bands. It also introduces age-specific adult/child models for the ICBHI dataset. The system is evaluated on ICBHI 2017, SPRSound 2022, and SPRSound 2023 for binary and 4-class tasks, reporting state-of-the-art or benchmark results and up to 50% FLOPs reduction.

Significance. If the claims hold, FBS is a practical, model-agnostic input-regularization technique that improves both accuracy and efficiency, and the compact CNN-TSA would be attractive for deployment in resource-constrained settings. The paper has several strengths: held-out test evaluation, 5-fold cross-validation for band selection, means over five runs with standard deviations, and detailed ablations of TSA placement, number of retained bands, and the lambda hyperparameter. However, the headline claims of new benchmarks and state-of-the-art performance depend on comparisons that are not yet like-for-like, which weakens the central contribution until those comparisons are controlled.

major comments (4)
  1. [Section IV.A.2 and IV.D] The SPRSound 2023 training set is never specified. Section IV.A.2 describes SPRSound 2023 as '871 test-only recordings' with no training split, while Section IV.D states that the authors 'used the official train-test splits provided with each dataset.' Since SPRSound 2023 has no official training split, it is unclear what data the SPRSound 2023 rows in Table IV were trained on. If the models were trained on SPRSound 2022 splits or any other set, the reported improvements from 66.66 to 71.20 TS (Task 1) and 80.97 to 83.20 TS (Task 2) are not controlled against the cited baselines, which may have used different training data. The authors must specify the exact training set used for SPRSound 2023 and, if necessary, re-run the cited baselines under the same training protocol to support the 'new benchmark' claim.
  2. [Section IV.B and Table III] For age-specific models, Section IV.B states that 'the reported performance metric is the average of both models.' The age-specific Task-1 AS of 62.68% in Table III is exactly the mean of the adult 58.03% and child 67.32% scores from Table II, not a score computed on the full ICBHI test set. All comparator rows in Table III report pooled test-set AS values, so this row is not directly comparable, and the statement that the age-specific model 'matches' Patch-Mix CL (62.37%) is an artifact of the unweighted subgroup average. The authors should report a pooled AS computed on the full ICBHI test set for the age-specific models, or perform per-subgroup comparisons against comparators split by the same age threshold.
  3. [Section V.B and Table III] The 'new record' of 64.01% for Patch-Mix CL + FBS(IS) is compared against the published Patch-Mix CL score of 62.37% from Bae et al. [17] with no in-house reproduction under the authors' preprocessing (16 kHz resampling, 64 mel bands, 50–2000 Hz range, SpecAugment). Because FBS modifies the input representation and the baseline score is taken from the original paper, the improvement could be due to preprocessing or evaluation-protocol differences rather than to FBS. Provide a same-protocol Patch-Mix CL baseline without FBS and report the FBS gain relative to that reproduced baseline.
  4. [Section III.B.1 and Figure 3] The hyperparameter λ in Equation (7) is central to the FBS importance score, but the paper does not state how λ was selected for the final models in Tables III and IV, nor whether the values shown in Figure 3 were computed on the test sets. The ablation shows that multiclass performance peaks at λ = 0.5–0.6 and binary performance improves up to λ = 1.0. If λ was chosen based on test-set performance, the reported gains over the full-spectrum baseline are inflated by selection on the test set. Specify the λ values used in each final model and describe the selection procedure (e.g., inner cross-validation on the training folds).
minor comments (6)
  1. [Section IV.B] The age threshold used to split the ICBHI dataset into adult and child models is not reported anywhere; without this threshold the age-specific experiments cannot be reproduced.
  2. [Table IV] In the SPRSound 2023 Task 2 block, there are two rows labeled 'CNN-TSA +FBS(BS) 50%f req (Ours)'; the second row almost certainly denotes 75% frequency retention and should be relabeled.
  3. [Equation (3)] The Grad-CAM formula omits the ReLU activation, and the text says this is deliberate, but the authors should justify why the omission is appropriate and whether the negative attributions have a clear interpretation in the frequency-selection context.
  4. [Algorithm 1] The stopping criterion 'no degradation in CV performance' is vague; specify which metric is monitored, what tolerance is allowed, and how many bands were ultimately retained for each dataset in the final models.
  5. [Section IV.C] Preprocessing details for the transformer adaptation are not given; the paper only says the pipeline from [17] was adopted. Provide the number of mel bands, normalization, and any augmentation used for the Patch-Mix CL + FBS experiments so that they can be reproduced.
  6. [Section III.A] The description of the three-layer CNN6 variant for SPRSound lacks the channel sizes, kernel sizes, and pooling configurations; include a layer-by-layer table or a precise architectural specification.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: FBS masks are chosen on training folds and evaluated on held-out official test splits; the only self-citation is a non-load-bearing earlier-version footnote.

full rationale

The paper's derivation chain is self-contained. Frequency Band Selection computes importance scores from Grad-CAM attributions (Eq. 7), but the selection is performed inside 5-fold cross-validation on the training set (Section IV.D and Algorithm 1), and the final model is retrained on the full training set and evaluated on the official test splits (Algorithm 1 final step). No reported test metric is an input to the selection algorithm, so the improved AS/TS numbers are not forced by construction. The transformer adaptation uses Integrated Gradients rather than Grad-CAM, and the Patch-Mix CL baseline is an external published system, not a self-citation. The only self-referential element, the EMBC earlier-version footnote, is descriptive and is never used as evidence for the benchmarks. Comparability caveats about SPRSound 2023 training data and pooled versus subgroup-averaged ICBHI AS are evaluation-protocol concerns, not circularity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claims rest on a few assumptions: AudioSet pretrained weights transfer to respiratory audio, Grad-CAM importance scores are informative for band selection, and prior published results are directly comparable. Free hyperparameters include lambda, the number of retained bands, the removal step r, and the adult/child age cutoff. No new entities are introduced.

free parameters (4)
  • lambda (importance penalty weight) = 0.5-1.0 (validated)
    Controls the penalty for class-wise variability in the importance score Eq. 7. Figure 3 shows the best lambda depends on task and dataset, and the final experiments do not specify which values were used.
  • Number of retained frequency bands = 48 (75%) or 40/32 depending on dataset
    Selected by 5-fold CV on the training set; the stopping criterion is a drop in CV performance. This is a model selection choice that directly affects reported test accuracy.
  • r (bands removed per iteration) = 4
    Heuristic step size for iterative elimination in Algorithm 1; not extensively tuned, but affects the greedy search path.
  • Age split threshold for adult/child models = Not stated
    The paper defines age-specific models but does not specify the exact age cutoff used to split ICBHI into adult and pediatric groups, which affects the reported age-specific AS.
assumptions (4)
  • domain assumption AudioSet pretrained weights transfer to respiratory sound classification.
    The CNN backbone is initialized with weights pretrained on AudioSet (Section III-A-a). No experiment is shown for training from scratch, so the reported performance depends on this transfer.
  • domain assumption Grad-CAM attributions computed from the full-frequency model are reliable indicators of which frequency bands are informative for a retrained model.
    The FBS procedure in Section III-B.1 uses Grad-CAM importance scores to eliminate bands. The validity of this assumption is tested indirectly via the ablation, but it is not guaranteed.
  • domain assumption Prior published results are directly comparable to the authors' results.
    Table III and IV compare against previously published numbers without re-implementing prior methods under the same preprocessing and evaluation protocol, which is a load-bearing assumption for the SOTA claim.
  • domain assumption Mel filter bank with 50-2000 Hz range captures relevant respiratory sound features.
    Preprocessing in Section IV-C fixes the frequency range based on prior work; this range is not experimentally justified in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Deep Learning-based Respiratory Sound Analysis with Frequency Selection and Attention Mechanism." pith.science (2026). https://pith.science/paper/TOKTHQ4H

@misc{pith2026250720052,
  author       = {Pith},
  title        = {Pith review of: Improving Deep Learning-based Respiratory Sound Analysis with Frequency Selection and Attention Mechanism},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TOKTHQ4H}},
  note         = {Machine review of arXiv:2507.20052}
}
read the original abstract

Accurate classification of respiratory sounds requires deep learning models that effectively capture fine-grained acoustic features and long-range temporal dependencies. Convolutional Neural Networks (CNNs) are well-suited for extracting local time-frequency patterns but are limited in modeling global context. In contrast, transformer-based models can capture long-range dependencies, albeit with higher computational demands. To address these limitations, we propose a compact CNN-Temporal Self-Attention (CNN-TSA) network that integrates lightweight self-attention into an efficient CNN backbone. Central to our approach is a Frequency Band Selection (FBS) module that suppresses noisy and non-informative frequency regions, substantially improving accuracy and reducing FLOPs by up to 50%. We also introduce age-specific models to enhance robustness across diverse patient groups. Evaluated on the SPRSound-2022/2023 and ICBHI-2017 lung sound datasets, CNN-TSA with FBS sets new benchmarks on SPRSound and achieves state-of-the-art performance on ICBHI, all with a significantly smaller computational footprint. Furthermore, integrating FBS into an existing transformer baseline yields a new record on ICBHI, confirming FBS as an effective drop-in enhancement. These results demonstrate that our framework enables reliable, real-time respiratory sound analysis suitable for deployment in resource-constrained settings.

Figures

Figures reproduced from arXiv: 2507.20052 by the authors.

Figure 1
Figure 1. Overview of the proposed framework with three components: (1) Frequency Band Selection (FBS), transforming audio into Mel spectrograms, generating a frequency mask, and refining the input; (2) CNN-TSA, consisting of a CNN backbone, spectral aggregation, and temporal self-attention, followed by classification; and (3) Age-Specific Models, adapting classification for children and adults. to the temporal self-attention… view at source ↗
Figure 2
Figure 2. Average Scores (AS) for Binary (left) and Multiclass (right) Classification across ICBHI 2017, SPRSound 2022, and SPRSound 2023 datasets using different numbers of retained frequency bands. 3) Impact of the Hyperparameter λ on Model Performance [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. illustrates how the weighting hyperparameter λ affects classification performance across the ICBHI 2017, SPRSound 2022, and SPRSound 2023 datasets using our importance-based frequency band selection (FBS (IS)) with 50% of bands retained. The importance score is defined as If = Mean[f] − λ · MaxDiff[f], where λ ∈ [0, 1] controls the trade-off between overall contribution and class-specific variability. In Binary Clas… view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Visualization of feature maps at different stages of the CNN backbone [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

69 extracted references · 67 canonical work pages

  1. [17]

    Patch-mix contrastive learning with audio spectrogram transformer on respiratory sound classification

    S. Bae et al. , “Patch-mix contrastive learning with audio spectrogram transformer on respiratory sound classification.” International Speech Communication Association, 2023, pp. 5436–5440

    work page 2023

  2. [1]

    Global burden of chronic respiratory diseases and risk factors, 1990–2019: an update from the global burden of disease study 2019,

    S. Momtazmanesh et al., “Global burden of chronic respiratory diseases and risk factors, 1990–2019: an update from the global burden of disease study 2019,” EClinicalMedicine, vol. 59, 2023

    work page 1990

  3. [2]

    Respiratory sounds: advances beyond the stetho- scope,

    H. Pasterkamp et al., “Respiratory sounds: advances beyond the stetho- scope,” American Journal of Respiratory Critical Care Medicine , vol. 156, no. 3, pp. 974–987, 1997

    work page 1997

  4. [3]

    Auscultation of the respiratory system,

    M. Sarkar et al. , “Auscultation of the respiratory system,” Annals of thoracic medicine, vol. 10, no. 3, pp. 158–168, 2015

    work page 2015

  5. [4]

    J. N. Siebert et al., “Deep learning diagnostic and severity-stratification for interstitial lung diseases and chronic obstructive pulmonary disease in digital lung auscultations and ultrasonography: clinical protocol for an observational case–control study,” BMC Pulmonary Medicine , vol. 23, no. 1, p. 191, 2023

    work page 2023

  6. [5]

    Machine learning for automated classi- fication of abnormal lung sounds obtained from public databases: A systematic review,

    J. P. Garcia-Mendez et al. , “Machine learning for automated classi- fication of abnormal lung sounds obtained from public databases: A systematic review,” Bioengineering, vol. 10, no. 10, p. 1155, 2023

    work page 2023

  7. [6]

    Lung disease recognition methods using audio-based analysis with machine learning,

    A. H. Sabry et al., “Lung disease recognition methods using audio-based analysis with machine learning,” Heliyon, 2024

    work page 2024

  8. [7]

    Short term spectral analysis, synthesis, and modification by discrete fourier transform,

    J. Allen, “Short term spectral analysis, synthesis, and modification by discrete fourier transform,” IEEE Transactions on Acoustics, Speech, and Signal Processing , vol. 25, no. 3, pp. 235–238, 1977

    work page 1977

Show all 69 references
  1. [8]

    Pretraining respiratory sound representations using metadata and contrastive learning

    I. Moummad et al., “Pretraining respiratory sound representations using metadata and contrastive learning.” IEEE, 2023, pp. 1–5

    2023

  2. [9]

    Lung sound classification using co-tuning and stochastic normalization,

    T. Nguyen et al. , “Lung sound classification using co-tuning and stochastic normalization,” IEEE Transactions on Biomedical Engineer- ing, vol. 69, no. 9, pp. 2872–2882, 2022

    2022

  3. [10]

    Explainable cnn with fuzzy tree regularization for respi- ratory sound analysis,

    J. Li et al. , “Explainable cnn with fuzzy tree regularization for respi- ratory sound analysis,” IEEE Transactions on Fuzzy Systems , vol. 30, no. 6, pp. 1516–1528, 2022

    2022

  4. [11]

    Deep learning for audio signal processing,

    H. Purwins et al. , “Deep learning for audio signal processing,” IEEE Journal of Selected Topics in Signal Processing , vol. 13, no. 2, pp. 206–219, 2019

    2019

  5. [12]

    Classify respiratory abnormality in lung sounds using stft a fine-tuned resnet18 network,

    Z. Chen et al. , “Classify respiratory abnormality in lung sounds using stft a fine-tuned resnet18 network,” in 2022 IEEE BioCAS. IEEE, 2022, pp. 233–237

    2022

  6. [13]

    Deep learning,

    Y . LeCun et al., “Deep learning,” Nature, vol. 521, no. 7553, pp. 436– 444, 2015

    2015

  7. [14]

    A tutorial on deep learning for music information retrieval,

    K. Choi et al. , “A tutorial on deep learning for music information retrieval,” arXiv preprint arXiv:1709.04396 , 2017. [Online]. Available: https://arxiv.org/abs/1709.04396

    2017 arXiv

  8. [15]

    Normal versus adventitious respiratory sounds,

    A. Marques et al. , “Normal versus adventitious respiratory sounds,” Breath Sounds: From Basic Science to Clinical Practice , pp. 181–206, 2018

    2018

  9. [16]

    Classification of adventitious sounds combining cochleogram and vision transformers,

    L. D. Mang et al. , “Classification of adventitious sounds combining cochleogram and vision transformers,” Sensors, vol. 24, no. 2, p. 682, 2024

    2024

  10. [18]

    Multi-view spectrogram transformer for respiratory sound classification,

    W. He et al., “Multi-view spectrogram transformer for respiratory sound classification,” in ICASSP. IEEE, 2024, pp. 8626–8630. 12

    2024

  11. [19]

    Ast: Audio spectrogram transformer,

    Y . Gong et al. , “Ast: Audio spectrogram transformer,” in Proc. Inter- speech, 2021, pp. 571–575

    2021

  12. [20]

    Transformers in vision: A survey,

    S. Khan et al. , “Transformers in vision: A survey,” ACM Computing Surveys (CSUR), vol. 54, no. 10s, pp. 1–41, 2022

    2022

  13. [21]

    Computerized lung sound screening for pediatric auscultation in noisy field environments,

    D. Emmanouilidou et al. , “Computerized lung sound screening for pediatric auscultation in noisy field environments,” IEEE Transactions on Biomedical Engineering , vol. 65, no. 7, pp. 1564–1574, 2017

    2017

  14. [22]

    Respiratory sound classification for crackles, wheezes, and rhonchi in the clinical field using deep learning,

    Y . Kim et al. , “Respiratory sound classification for crackles, wheezes, and rhonchi in the clinical field using deep learning,” Scientific Reports, vol. 11, no. 1, pp. 1–11, 2021

    2021

  15. [23]

    Noise masking recurrent neural network for respiratory sound classification,

    K. Kochetov et al. , “Noise masking recurrent neural network for respiratory sound classification,” in Artificial Neural Networks Machine Learning–ICANN. Springer, 2018, pp. 208–217

    2018

  16. [24]

    Survey on explainable ai: From approaches to lim- itations and applications aspects,

    W. Yang et al. , “Survey on explainable ai: From approaches to lim- itations and applications aspects,” Human-Centric Intelligent Systems , vol. 3, no. 3, pp. 161–188, 2023

    2023

  17. [25]

    Measuring feature importance of convolutional neural networks,

    X. Zhang et al., “Measuring feature importance of convolutional neural networks,” IEEE Access, vol. 8, pp. 196 062–196 074, 2020

    2020

  18. [26]

    Feature selection: A perspective on inter- attribute cooperation,

    G. Sosa-Cabrera et al. , “Feature selection: A perspective on inter- attribute cooperation,” International Journal of Data Science and An- alytics, vol. 17, no. 2, pp. 139–151, 2024

    2024

  19. [27]

    The relationship between normal lung sounds, age, and gender,

    V . Gross et al., “The relationship between normal lung sounds, age, and gender,” American Journal of Respiratory and Critical Care Medicine , vol. 162, no. 3, pp. 905–909, 2000

    2000

  20. [28]

    An open access database for the evaluation of respiratory sound classification algorithms,

    B. M. Rocha et al. , “An open access database for the evaluation of respiratory sound classification algorithms,” Physiological measurement, vol. 40, no. 3, p. 035001, 2019

    2019

  21. [29]

    Sprsound: Open-source sjtu paediatric respiratory sound database,

    Q. Zhang et al. , “Sprsound: Open-source sjtu paediatric respiratory sound database,” IEEE Transactions on Biomedical Circuits and Sys- tems, vol. 16, no. 5, pp. 867–881, 2022

    2022

  22. [30]

    Grand challenge on respiratory sound classification for sprsound dataset,

    Q. ZHang et al. , “Grand challenge on respiratory sound classification for sprsound dataset,” in 2022 IEEE BioCAS . IEEE, pp. 213–217

    2022

  23. [31]

    Respirenet: A deep neural network for accurately detecting abnormal lung sounds in limited data setting,

    S. Gairola et al. , “Respirenet: A deep neural network for accurately detecting abnormal lung sounds in limited data setting,” in 2021 IEEE EMBC. IEEE, 2021, pp. 527–530

    2021

  24. [32]

    Multiclass categorisation of respiratory sound signals using neural network,

    N. Babu et al. , “Multiclass categorisation of respiratory sound signals using neural network,” in 2022 IEEE BioCAS . IEEE, 2022, pp. 228– 232

    2022

  25. [33]

    Cnn-moe based framework for classification of respira- tory anomalies and lung disease detection,

    L. Pham et al., “Cnn-moe based framework for classification of respira- tory anomalies and lung disease detection,” IEEE Journal of Biomedical Health Informatics, vol. 25, no. 8, pp. 2938–2947, 2021

    2021

  26. [34]

    Pediatric respiratory sound classification using a dual input deep learning architecture,

    D. Pessoa et al., “Pediatric respiratory sound classification using a dual input deep learning architecture,” in 2023 IEEE BioCAS . IEEE, 2023, pp. 1–5

    2023

  27. [35]

    Trespnet: A dual-route exploratory cnn model for pediatric adventitious respiratory sound identification,

    B. TaghiBeyglou et al., “Trespnet: A dual-route exploratory cnn model for pediatric adventitious respiratory sound identification,” Biomedical Signal Processing , Control , vol. 93, p. 106170, 2024

    2024

  28. [36]

    Adventitious respiratory classification using attentive residual neural networks,

    Z. Yang et al. , “Adventitious respiratory classification using attentive residual neural networks,” Interspeech 2020, pp. 2912–2916

    2020

  29. [37]

    Interpretation of lung disease classification with light attention connected module,

    Y . Choi et al. , “Interpretation of lung disease classification with light attention connected module,” Biomedical Signal Processing and Control, vol. 84, p. 104695, 2023

    2023

  30. [38]

    A deep learning architecture with spatio-temporal focusing for detecting respiratory anomalies,

    D. Ngo et al. , “A deep learning architecture with spatio-temporal focusing for detecting respiratory anomalies,” in 2023 IEEE BioCAS . IEEE, 2023, pp. 1–5

    2023

  31. [39]

    Masked modeling duo: Towards a universal audio pre-training framework,

    D. Niizumi et al. , “Masked modeling duo: Towards a universal audio pre-training framework,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2024

    2024

  32. [40]

    Bts: Bridging text and sound modalities for metadata- aided respiratory sound classification,

    J.-W. Kim et al., “Bts: Bridging text and sound modalities for metadata- aided respiratory sound classification,” Interspeech, 2024

    2024

  33. [41]

    Classifying imbalanced data: The relevance of accuracy and feature importance,

    T. Widmann, “Classifying imbalanced data: The relevance of accuracy and feature importance,” 2024

    2024

  34. [42]

    Selecting critical features for data classification based on machine learning methods,

    R.-C. Chen et al., “Selecting critical features for data classification based on machine learning methods,” Journal of Big Data, vol. 7, no. 1, p. 52, 2020

    2020

  35. [43]

    A meticulous presaging of heart disease optimized by boruta feature selection and rfe over gradient boosting,

    R. Aggarwal et al., “A meticulous presaging of heart disease optimized by boruta feature selection and rfe over gradient boosting,” in 2022 OPJU International Technology Conference on Emerging Technologies for Sustainable Development (OTCON) . IEEE, 2023, pp. 1–6

    2022

  36. [44]

    Traditional machine learning algorithms for breast cancer image classification with optimized deep features,

    F. Atban et al. , “Traditional machine learning algorithms for breast cancer image classification with optimized deep features,” Biomedical Signal Processing and Control , vol. 81, p. 104534, 2023

    2023

  37. [45]

    Fusion of statistical importance for feature selection in deep neural network-based intrusion detection system,

    A. Thakkar et al., “Fusion of statistical importance for feature selection in deep neural network-based intrusion detection system,” Information Fusion, vol. 90, pp. 353–363, 2023

    2023

  38. [46]

    Comparison of parametric representations for mono- syllabic word recognition in continuously spoken sentences,

    S. Davis et al. , “Comparison of parametric representations for mono- syllabic word recognition in continuously spoken sentences,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 28, no. 4, pp. 357–366, 1980

    1980

  39. [47]

    Panns: Large-scale pretrained audio neural networks for audio pattern recognition,

    Q. Kong et al., “Panns: Large-scale pretrained audio neural networks for audio pattern recognition,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 28, pp. 2880–2894, 2020

    2020

  40. [48]

    Crackles: recording, analysis and clinical significance,

    P. P et al. , “Crackles: recording, analysis and clinical significance,” European Respiratory Journal, vol. 8, no. 12, pp. 2139–2148, 1995

    1995

  41. [49]

    Attention is all you need,

    A. Vaswani, et al. , “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

    2017

  42. [50]

    Network in network,

    M. Lin et al. , “Network in network,” 2014. [Online]. Available: https://arxiv.org/abs/1312.4400

    2014 arXiv

  43. [51]

    Grad-cam: Visual explanations from deep networks via gradient-based localization,

    R. R. Selvaraju et al. , “Grad-cam: Visual explanations from deep networks via gradient-based localization,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 618–626

    2017

  44. [52]

    Axiomatic attribution for deep networks,

    M. Sundararajan et al. , “Axiomatic attribution for deep networks,” in International conference on machine learning. PMLR, 2017, pp. 3319– 3328

    2017

  45. [53]

    Grand challenge on respiratory sound classification for sprsound dataset,

    Q. Zhang et al., “Grand challenge on respiratory sound classification for sprsound dataset,” in 2023 IEEE BioCAS . IEEE, pp. 1–5

    2023

  46. [54]

    Contrastive embedding learning method for respiratory sound classification,

    S. Wenjie et al., “Contrastive embedding learning method for respiratory sound classification,” in ICASSP IEEE International Conference on Acoustics Speech Signal Processing . IEEE, 2021, pp. 1275–1279

    2021

  47. [55]

    Detection of patients considering observation frequency of continuous and discontinuous adventitious sounds in lung sounds,

    N. Nakamura et al. , “Detection of patients considering observation frequency of continuous and discontinuous adventitious sounds in lung sounds,” in 2016 38th Annual International Conference of the IEEE EMBC. IEEE, 2016, pp. 3457–3460

    2016

  48. [56]

    Arsc-net: Adventitious respiratory sound classification network using parallel paths with channel-spatial attention,

    L. Xu et al. , “Arsc-net: Adventitious respiratory sound classification network using parallel paths with channel-spatial attention,” in 2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE, 2021, pp. 1125–1130

    2021

  49. [57]

    Hidden markov model based respiratory sound classification,

    N. Jakovljevi ´c et al. , “Hidden markov model based respiratory sound classification,” in Precision Medicine Powered by pHealth Connected Health ICBHI. Springer, 2018, pp. 39–43

    2018

  50. [58]

    Specaugment: A simple data augmentation method for automatic speech recognition,

    D. S. Park et al. , “Specaugment: A simple data augmentation method for automatic speech recognition,” Interspeech 2019, p. 2613, 2019

    2019

  51. [59]

    The real-world-weight cross-entropy loss function: Mod- eling the costs of mislabeling,

    Y . Ho et al., “The real-world-weight cross-entropy loss function: Mod- eling the costs of mislabeling,” IEEE Access , vol. 8, pp. 4806–4813, 2019

    2019

  52. [60]

    Lungattn: advanced lung sound classification using attention mechanism with dual tqwt and triple stft spectrogram,

    J. Li et al., “Lungattn: advanced lung sound classification using attention mechanism with dual tqwt and triple stft spectrogram,” Physiological Measurement, vol. 42, no. 10, p. 105006, 2021

    2021

  53. [61]

    Prototype learning for interpretable respiratory sound analysis,

    Z. Ren et al. , “Prototype learning for interpretable respiratory sound analysis,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2022, pp. 9087–9091

    2022

  54. [62]

    Example-based Explanations with Adversarial Attacks for Respiratory Sound Analysis,

    C. Yi et al. , “Example-based Explanations with Adversarial Attacks for Respiratory Sound Analysis,” in Proc. Interspeech 2022 , 2022, pp. 4003–4007

    2022

  55. [63]

    A domain transfer based data augmentation method for automated respiratory classification,

    Z. Wang et al. , “A domain transfer based data augmentation method for automated respiratory classification,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022, pp. 9017–9021

    2022

  56. [64]

    An ensemble of deep learning frameworks for predicting respiratory anomalies,

    L. Pham et al., “An ensemble of deep learning frameworks for predicting respiratory anomalies,” in 2022 44th Annual International Conference of the IEEE EMBC . IEEE, 2022, pp. 4595–4598

    2022

  57. [65]

    Stethoscope-guided supervised contrastive learning for cross-domain adaptation on respiratory sound classification,

    J.-W. Kim et al., “Stethoscope-guided supervised contrastive learning for cross-domain adaptation on respiratory sound classification,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 1431–1435

    2024

  58. [66]

    An effective lung sound classification system for respiratory disease diagnosis using densenet cnn model with sound pre- processing engine,

    W.-B. Ma et al. , “An effective lung sound classification system for respiratory disease diagnosis using densenet cnn model with sound pre- processing engine,” in 2022 IEEE BioCAS . IEEE, 2022, pp. 218–222

    2022

  59. [67]

    Improving the resnet-based respiratory sound classification systems with focal loss,

    J. Li et al., “Improving the resnet-based respiratory sound classification systems with focal loss,” in 2022 IEEE BioCAS . IEEE, 2022, pp. 223–227

    2022

  60. [68]

    A feature polymerized based two-level ensemble model for respiratory sound classification,

    L. Zhang et al., “A feature polymerized based two-level ensemble model for respiratory sound classification,” in2022 IEEE BioCAS. IEEE, 2022, pp. 238–242

    2022

  61. [69]

    Supervised contrastive pretrained resnet with mixup to enhance respiratory sound classification on imbalanced and limited dataset,

    J. Hu et al. , “Supervised contrastive pretrained resnet with mixup to enhance respiratory sound classification on imbalanced and limited dataset,” in 2023 IEEE BioCAS . IEEE, 2023, pp. 1–5

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.