Pith. sign in

REVIEW 4 major objections 6 minor 44 references

Anchor intervals give lung-sound detection direct event timing

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A GNN plus anchor-interval framework detects abnormal respiratory sound events, reaching F1 of 22.3 percent on SPRSound, with anchor scales tuned to the observed event durations.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Novel GNN+anchor framework for respiratory event detection, but the evaluation as written doesn't prove the headline improvement—the main baseline comparison is confounded and the anchor priors are fitted to the test set. the 4 major comments →

arxiv 2509.01153 v2 pith:ALUZ4TGO submitted 2025-09-01 cs.SD cs.AIeess.AS

EZhouNet:A framework based on graph neural network and anchor interval for the respiratory sound event detection

Yun Chu , Qiuhao Wang , Enze Zhou , Qian Liu , Gang Zheng This is my paper
classification cs.SD cs.AIeess.AS
keywords respiratory sound event detectiongraph neural networkanchor intervalvariable-length audioSPRSoundHF Lung V1auscultationsound event detection
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 claims that respiratory sound event detection can be cast as an end-to-end interval-refinement problem instead of frame-level classification plus post-processing. It builds a graph from a multi-channel spectrogram, where each node is a short chunk of audio and edges connect neighboring chunks in time, then refines predefined anchor intervals to output event boundaries directly. Because the graph can contain any number of nodes, the framework accepts variable-length recordings without resampling or padding. On the SPRSound 2024 and HF Lung V1 datasets, the method reports event-level F1 scores of 22.3% and 16.8%, outperforming the compared CNN+VAD baseline on SPRSound, and the authors show that adding respiratory position information raises recall across abnormal sound classes. A sympathetic reader would care because direct boundary learning and variable-length input address two known limitations in respiratory sound event detection.

Core claim

The paper's central claim is that separating sound-event detection into interval confidence, interval category, and interval-boundary regression over predefined anchor intervals lets a network learn event timing directly, rather than inferring boundaries from frame-level outputs. The anchor intervals are temporal windows at three fixed durations (0.5, 0.8, and 1.5 seconds normalized to clip length), each labeled by its best IoU with a ground-truth event; the network predicts an offset, a confidence, and a class per anchor and decodes refined intervals. The graph representation is the enabler of variable-length handling: every five spectrogram frames form one node, nodes are linked in a forwa

What carries the argument

Anchor interval refinement is the load-bearing mechanism: a fixed set of candidate intervals at three scales, with soft IoU-based labels for confidence, class, and target interval, lets the network regress start/end offsets instead of predicting event boundaries from scratch. The supporting machinery is the spectrogram-to-graph construction, in which each group of five spectrogram frames becomes a node, edges form a directed temporal chain, and GAT layers update node features while preserving variable-length structure.

Load-bearing premise

The framework depends on the three fixed anchor durations (0.5, 0.8, and 1.5 seconds) matching the durations of abnormal events in the target data; if a new dataset's events fall mostly outside these ranges, the offset regressor must compensate on its own.

What would settle it

Run the same framework on a dataset of respiratory recordings whose abnormal events are predominantly shorter than 0.3 s or longer than 2.5 s, with all other training details unchanged. If event-level F1 drops sharply relative to a version with anchor scales matched to those durations, the reliance on fixed anchor priors is confirmed.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Event-level respiratory sound detection can be trained end-to-end, since interval confidence, category, and localization are learned jointly from anchor labels.
  • Variable-length recordings can be processed without fixed-length constraints, because the input graph naturally grows with the number of chunks.
  • Adding respiratory position information improves recall of abnormal events, so location cues are worth including in clinical audio models.
  • Five-frame grouping favors discontinuous sounds such as crackle; continuous sounds such as wheeze, rhonchi, and stridor need larger groups or longer context.
  • The reported event F1 remains modest in absolute terms, indicating that anchor priors and frame grouping are not yet sufficient for clinical use alone.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Because the anchor scales are fixed to the duration distribution observed in SPRSound and HF Lung v1, the framework should transfer best to settings with similar event durations; new populations with systematically longer or shorter abnormal events would likely require re-tuning the scales.
  • The same anchor-interval idea could be applied to other periodic or event-based bioacoustic tasks, such as cough or snore detection, where event durations are variable and frame-level post-processing is the norm.
  • A testable extension is to replace fixed anchor scales with learned or adaptive anchors, such as a small set of learnable durations per clip, which would remove the manual prior without abandoning the regression formulation.
  • The chain-graph connectivity limits long-range context; using higher-order edges or a second graph over longer spans might improve detection of sustained wheeze and rhonchi.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes EZhouNet, an end-to-end framework for abnormal respiratory sound event detection that combines multichannel spectrograms (Mel, Gamma, CQT), a graph construction module that converts spectrogram chunks into variable-length graph nodes, two GAT layers for node feature update, and an anchor-interval refinement module with three predefined duration scales. The anchor intervals are refined via GRU-based local feature extraction, soft offset regression over discrete bins, and separate confidence/classification heads. Experiments are reported on SPRSound 2024 and HF Lung V1, including a comparison against a CNN+VAD baseline [42], a node-level ablation for respiratory position information, and a discussion of limitations. The central claim is that the framework improves temporal localization and event-level detection of abnormal respiratory sounds while natively handling variable-length audio.

Significance. If the central claim were adequately supported, the paper would make a useful contribution by transferring anchor-based object detection ideas to respiratory sound event detection, avoiding frame-level post-processing, and proposing a graph-based representation that can in principle handle variable-length recordings. The availability of reference code is a positive feature, and the task itself is clinically relevant and relatively understudied. However, the reported empirical evidence is not yet sufficient to establish the claimed improvement: the comparison with [42] is not controlled, the anchor priors appear to be fitted to the evaluation datasets, key hyperparameters are missing, and no uncertainty estimates or significance tests are provided. These issues are fixable, but they are load-bearing for the paper's main conclusion.

major comments (4)
  1. [3.3.2, Table 3, Eq. (47)] The comparison with CNN+VAD [42] on SPRSound is not apples-to-apples. The text states that [42] also includes normal sounds in the detection process, while 'our experiments exclude them.' In the event-based metrics of Eq. (47), N is the number of reference events, and FP/FN are counted with respect to the label set. A system that never predicts 'normal' cannot be charged with deletions or insertions for normal events, whereas a baseline that must detect them is penalized for every normal-event error. The reported F1 gap (22.3 vs 6.81) and ER gap (1.08 vs 1.26) may therefore be a metric artifact, not an architectural improvement. To support the central claim, the baseline must be re-evaluated under the identical label protocol, or both systems must be reported under both protocols (including class-wise matched labels).
  2. [3.3.3, Eq. (25)-(26)] The anchor interval priors appear to be fitted to the evaluation datasets. The paper observes that event durations in SPRSound and HF Lung V1 fall at 0.5s, 0.75s, 1s, and 1.25s, and states that these 'align with the designed prior anchor intervals of 0.5s, 0.75s, and 1.5s' (Eq. (25)). Since the anchor durations and densities are chosen after inspecting the test duration distribution, the reported detection performance partly reflects this fitting. This weakens the generalization claim and makes the framework's interval-detection behavior dependent on the evaluation datasets. I recommend either evaluating on a held-out dataset with a different event-duration distribution, or performing a sensitivity analysis over anchor scales/configurations and reporting results for a prior chosen independently of the test data.
  3. [3.3.1-3.3.2, Eq. (31)-(32), Eq. (38)-(39), Eq. (46)] Several load-bearing experimental details are missing. The loss weights w_n_conf, w_n_cls, w_i_conf, w_i_cls, w_i_loc in Eq. (46) are never given numerical values. The IoU threshold theta_iou in Eq. (31)-(32) is defined but its value is never reported. The bin counts B_k and their ranges in Eq. (38)-(39) are also unspecified, and the temporal encoding scale gamma in Eq. (24) is only described as 'e.g. 0.05.' In addition, Tables 1-3 report single runs without standard deviations, confidence intervals, or number of seeds. Given the moderate F1 values and class imbalance, the reported differences (e.g., integrated vs separate head) cannot be assessed for statistical significance. These details are needed both for reproducibility and for interpreting the central empirical claim.
  4. [3.3.2, Tables 4-5] The ablation for respiratory position information is performed at the node level, not at the event level. The text claims that position information improves recall for all four abnormal classes, and indeed Table 5 shows recall increases for rhonchi, wheeze, stridor, and crackle. However, F1 for crackle decreases (0.7905 to 0.7804), and there are no confidence intervals or significance tests. More importantly, the central claim is about event-level detection (Table 3); a node-level ablation does not directly establish that position information improves the event-level detection of abnormal respiratory sounds. An event-level ablation with and without position information is needed.
minor comments (6)
  1. [3.3.2 (title)] Typo in section title: 'stduy' should be 'study.'
  2. [Eq. (9)] The argmax notation contains stray vertical bars ('|||{...}|||') and the tie-breaking rule could be stated more cleanly. Please clarify whether ties are resolved by the first-appearing class in chronological order.
  3. [Figures 1-3] The text refers to 'Fig. 2' when describing the network architecture, but the first mentioned figure is 'Fig. 1'; please check figure numbering. Also, Fig. 4 in the text corresponds to 'The convergence of interval classification losses' but is placed after the evaluation metrics section; renumbering may improve readability.
  4. [References] The reference list appears to be duplicated: the same entries [1]-[9] and then a second full 'References' list with the same numbered papers are printed. Please remove the duplicate block.
  5. [Table 3] For the HF Lung V1 row, no baseline value is reported. The table should clearly state 'not comparable / not reported' rather than a dash, to avoid implying an empty or zero baseline.
  6. [3.4] The discussion admits that variable-length audio could not be batched in practice due to the PyTorch Lightning collate_fn, and training used sequential loading. This is an honest limitation, but it should also be noted in Section 1 and the abstract contribution statement, since 'handling variable-length audio' is presented as a native capability.

Circularity Check

0 steps flagged

No significant circularity: the interval predictions are anchor-plus-learned-offset outputs, not predetermined by the anchor priors, and no load-bearing self-citation is present.

full rationale

The paper's central derivation is an end-to-end architecture: multichannel spectrogram -> graph construction -> GAT node updates -> anchor-interval refinement. The final interval predictions are computed in Eq. (40) as anchor + learned offsets, where the offsets come from a trained MLP/GRU head over discrete bins (Eqs. 33-39). The anchor scales in Eq. (25) are manually specified hyperparameters, and the statement in Sec. 3.3.3 that observed event durations 'align with the designed prior anchor intervals' is a post-hoc hyperparameter justification, not an equation-level reduction: the offsets are learned and, for the best configuration in Table 1, can range over [-20,20], so the anchor prior does not by itself determine the output intervals. No load-bearing self-citation appears: [40], [42], and [43] are external datasets/baselines, and the paper invokes no uniqueness theorem or prior work by the same authors to force its design. The limitations acknowledged in Sec. 3.4 (fixed anchor sizes; inability to batch variable-length audio in the current implementation) are practical weaknesses and do not constitute circular derivation. The Table 3 comparison with [42] is confounded by a different label protocol (normal events included vs. excluded), but that is an evaluation-validity concern, not circularity. Under the required standard of exhibiting a specific reduction where a fitted parameter is renamed as a prediction or an equation reduces to its own input, no circular step can be identified.

Axiom & Free-Parameter Ledger

7 free parameters · 5 axioms · 0 invented entities

The central claim depends on several hand-set numbers: anchor scales matched to test durations, a fixed 5-frame node size, and unreported loss weights and IoU thresholds. These are the main free parameters. The graph and anchor machinery add no new physical entities.

free parameters (7)
  • anchor scale durations d = [0.5, 0.8, 1.5] s
    Hand-chosen and explicitly matched to the observed event durations in SPRSound and HF Lung V1 (Sec 3.3.3), making them fitted to the test distribution.
  • anchor density weights w_k and base centers N0 = [0.75, 2.0, 0.75], N0=20
    Selected without a systematic search; control anchor density per scale.
  • frames per node (group size) = 5 frames (~80 ms)
    Chosen to capture discontinuous sounds; paper notes it is less suited for continuous sounds (Sec 3.3.3).
  • offset bin count B_k and ranges = swept ranges like [-0.5, 0.5] to [-20, 20]
    Selected per experiment (Tables 1, 2) with no rule; best configuration reported.
  • loss weights w_n_conf, w_n_cls, w_i_conf, w_i_cls, w_i_loc = not reported
    The five-term loss (Eq. 46) depends on these weights but they are never specified.
  • IoU threshold theta_iou = not reported
    Controls which anchors are treated as foreground in Eq. (32); value omitted.
  • Gaussian scaling gamma = 0.05
    Chosen to avoid overwhelming features (Eq. 24) without evidence.
axioms (5)
  • standard math FFT, Mel, gammatone and CQT filterbank definitions
    Used as input features in Sec 2.1; standard DSP background.
  • domain assumption Each 5-frame chunk maps to a graph node that preserves event boundary information
    Underlies the graph construction in Sec 2.2; if boundaries fall inside chunks, node labels are noisy.
  • ad hoc to paper Fixed 3-scale anchor intervals cover the event durations of interest
    Scales in Sec 2.4.1 selected after observing duration distributions of the evaluation sets (Sec 3.3.3).
  • domain assumption Event-based F1 with 200 ms onset and offset collar is the appropriate evaluation
    Used in Sec 3.2.3; different collars would change rankings.
  • ad hoc to paper Soft confidence labels improve training stability
    Claimed in Sec 2.2.2 with no supporting comparison.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of EZhouNet:A framework based on graph neural network and anchor interval for the respiratory sound event detection." pith.science (2026). https://pith.science/paper/ALUZ4TGO

@misc{pith2026250901153,
  author       = {Pith},
  title        = {Pith review of: EZhouNet:A framework based on graph neural network and anchor interval for the respiratory sound event detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ALUZ4TGO}},
  note         = {Machine review of arXiv:2509.01153}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Auscultation is a key method for early diagnosis of respiratory and pulmonary diseases, relying on skilled healthcare professionals. However, the process is often subjective, with variability between experts. As a result, numerous deep learning-based automatic classification methods have emerged, most of which focus on respiratory sound classification. In contrast, research on respiratory sound event detection remains limited. Existing sound event detection methods typically rely on frame-level predictions followed by post-processing to generate event-level outputs, making interval boundaries challenging to learn directly. Furthermore, many approaches can only handle fixed-length audio, limiting their applicability to variable-length respiratory sounds. Additionally, the impact of respiratory sound location information on detection performance has not been extensively explored. To address these issues, we propose a graph neural network-based framework with anchor intervals, capable of handling variable-length audio and providing more precise temporal localization for abnormal respiratory sound events. Our method improves both the flexibility and applicability of respiratory sound detection. Experiments on the SPRSound 2024 and HF Lung V1 datasets demonstrate the effectiveness of the proposed approach, and incorporating respiratory position information enhances the discrimination between abnormal sounds. The reference implementation is available at https://github.com/chumingqian/EzhouNet.

Figures

Figures reproduced from arXiv: 2509.01153 by Enze Zhou, Gang Zheng, Qian Liu, Qiuhao Wang, Yun Chu.

Figure 1
Figure 1. Figure 1: The proposed framework based on Graph Neural Network and anchor interval for Respiratory Sound Event Detection Finally, the refined start and end points of each interval are then given by (40): start𝑘,𝑖 = 𝛼𝑘,𝑖𝐿 + Δ𝑠 𝑘,𝑖, end𝑘,𝑖 = 𝛽𝑘,𝑖𝐿 + Δ𝑒 𝑘,𝑖 (40) These predicted intervals are optionally clamped to [0, 𝐿]. 2.5. The proposed Framework In this section, we will introduce the respiratory sound network archit… view at source ↗
Figure 2
Figure 2. Figure 2: Event-based evaluation metrics with fixed 200ms collar on onsets and 200ms/10 % of the event’s length collar on offsets [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Calculation of the event-based evaluation metrics. Fig.4 illustrates the evaluation principle using a 15- second audio file containing multiple events. By comparing system outputs with ground truth, we identify false negatives (FN) , false positives (FP) , and true positives (TP) . An event is considered a true positive if it matches both the timing and event type in the system output and ground truth; oth… view at source ↗
Figure 4
Figure 4. Figure 4: The convergence of interval classification losses under different offset configurations [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: In the SPR Sound dataset, the distribution ratio of different abnormal respiratory sounds in various duration periods. The upper one is the training set, and the lower one is the validation set [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: In the HF Lung v1 dataset, the distribution ratio of different abnormal respiratory sounds in various duration periods. The upper one is the training set, and the lower one is the validation set. training time. Unfortunately, we have not yet found a solution to reduce the training time caused by this issue. We hope to address this in future work and explore potential optimiza￾tions. 4. Conclusions In this … view at source ↗
Figure 7
Figure 7. Figure 7: The training process of the proposed method on different datasets. The top one is on the SPR sound dataset, and the bottom one is on the HF lung v1 dataset. on the spectrogram and construct a graph data structure for respiratory sounds, enabling the processing of variable￾length audio inputs through GNNs. Secondly, inspired by object detection in computer vision, we propose the use of anchor intervals in s… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 42 canonical work pages

  1. [1]

    World health statistics 2022: monitoring health for the sdgs, sustainable development goals,

    W. H. Organizationet al., “World health statistics 2022: monitoring health for the sdgs, sustainable development goals,” 2022

    work page 2022

  2. [2]

    Projections of global mortality and burden of disease from 2002 to 2030,

    C. D. Mathers and D. Loncar, “Projections of global mortality and burden of disease from 2002 to 2030,”PLoS medicine, vol. 3, no. 11, p. e442, 2006

    work page 2002

  3. [3]

    Whoglobalprogressreportontuberculosiselimination,

    E.Harding,“Whoglobalprogressreportontuberculosiselimination,” The Lancet Respiratory Medicine, vol. 8, no. 1, p. 19, 2020

    work page 2020

  4. [4]

    The global impact of respiratory disease–second edition. sheffield, european respiratory society,

    F. of International Respiratory Societies, “The global impact of respiratory disease–second edition. sheffield, european respiratory society,” 2017

    work page 2017

  5. [5]

    Fundamentals of lung auscultation,

    A. Bohadana, G. Izbicki, and S. S. Kraman, “Fundamentals of lung auscultation,” New England Journal of Medicine, vol. 370, no. 8, pp. 744–751, 2014

    work page 2014

  6. [6]

    New parameters for respiratory sound classification,

    M. Bahoura and C. Pelletier, “New parameters for respiratory sound classification,” inCCECE 2003-Canadian Conference on Electrical and Computer Engineering. Toward a Caring and Humane Technol- ogy (Cat. No. 03CH37436), vol. 3, pp. 1457–1460, IEEE, 2003

    work page 2003

  7. [7]

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

    J. Li, J. Yuan, H. Wang, S. Liu, Q. Guo, Y. Ma, Y. Li, L. Zhao, and G. Wang, “Lungattn: advanced lung sound classification using attention mechanism with dual tqwt and triple stft spectrogram,” Physiological Measurement, vol. 42, no. 10, p. 105006, 2021

    work page 2021

  8. [8]

    A lightweight cnn model for detecting respiratory diseases from lung auscultation sounds using emd-cwt-based hybrid scalogram,

    S.B.Shuvo,S.N.Ali,S.I.Swapnil,T.Hasan,andM.I.H.Bhuiyan, “A lightweight cnn model for detecting respiratory diseases from lung auscultation sounds using emd-cwt-based hybrid scalogram,” IEEE Journal of Biomedical and Health Informatics, vol. 25, no. 7, pp. 2595–2603, 2020

    work page 2020

  9. [9]

    Classificationofauscultation sounds into objective spirometry findings using mvmd and 3d cnn,

    S.Gupta,M.Agrawal,andD.Deepak,“Classificationofauscultation sounds into objective spirometry findings using mvmd and 3d cnn,” in2022 National Conference on Communications (NCC), pp. 42–47, IEEE, 2022

    work page 2022

  10. [10]

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

    L. Xu, J. Cheng, J. Liu, H. Kuang, F. Wu, and J. Wang, “Arsc-net: Adventitious respiratory sound classification network using parallel paths with channel-spatial attention,” in2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 1125– 1130, IEEE, 2021

    work page 2021

  11. [11]

    Lung sounds classifica- tion using convolutional neural networks,

    D. Bardou, K. Zhang, and S. M. Ahmad, “Lung sounds classifica- tion using convolutional neural networks,”Artificial intelligence in medicine, vol. 88, pp. 58–69, 2018

    work page 2018

  12. [12]

    Triple-classification of respiratory sounds using optimized s-transform and deep residual networks,

    H. Chen, X. Yuan, Z. Pei, M. Li, and J. Li, “Triple-classification of respiratory sounds using optimized s-transform and deep residual networks,”IEEE Access, vol. 7, pp. 32845–32852, 2019

    work page 2019

  13. [13]

    Auto- matic classification of large-scale respiratory sound dataset based on convolutionalneuralnetwork,

    K. Minami, H. Lu, H. Kim, S. Mabu, Y. Hirano, and S. Kido, “Auto- matic classification of large-scale respiratory sound dataset based on convolutionalneuralnetwork,”in 201919thInternationalConference on Control, Automation and Systems (ICCAS), pp. 804–807, IEEE, 2019

    work page 2019

  14. [14]

    Lung- brn: A smart digital stethoscope for detecting respiratory disease using bi-resnet deep learning algorithm,

    Y. Ma, X. Xu, Q. Yu, Y. Zhang, Y. Li, J. Zhao, and G. Wang, “Lung- brn: A smart digital stethoscope for detecting respiratory disease using bi-resnet deep learning algorithm,” in2019 IEEE Biomedical Circuits and Systems Conference (BioCAS), pp. 1–4, IEEE, 2019

    work page 2019

  15. [15]

    Gammatonegrambasedtriple classificationoflungsoundsusingdeepconvolutionalneuralnetwork with transfer learning,

    S.Gupta,M.Agrawal,andD.Deepak,“Gammatonegrambasedtriple classificationoflungsoundsusingdeepconvolutionalneuralnetwork with transfer learning,”Biomedical Signal Processing and Control, vol. 70, p. 102947, 2021

    work page 2021

  16. [16]

    Feature extraction of lung sounds based on bispectrum analysis,

    S. Li and Y. Liu, “Feature extraction of lung sounds based on bispectrum analysis,” in 2010 Third International Symposium on Information Processing, pp. 393–397, IEEE, 2010

    work page 2010

  17. [17]

    A comparison of svm and gmm-based classifier configurations for diagnostic classification of pulmonary sounds,

    I. Sen, M. Saraclar, and Y. P. Kahya, “A comparison of svm and gmm-based classifier configurations for diagnostic classification of pulmonary sounds,”IEEE Transactions on Biomedical Engineering, vol. 62, no. 7, pp. 1768–1776, 2015

    work page 2015

  18. [18]

    Incorporating support vector machine to the classification of respi- ratory sounds by convolutional neural network,

    F. Cinyol, U. Baysal, D. Köksal, E. Babaoğlu, and S. S. Ulaşlı, “Incorporating support vector machine to the classification of respi- ratory sounds by convolutional neural network,”Biomedical Signal Processing and Control, vol. 79, p. 104093, 2023

    work page 2023

  19. [19]

    Deep learning and feature fusion-based lung sound recognition model to diagnoses the respiratory diseases,

    S. A. Shehab, K. K. Mohammed, A. Darwish, and A. E. Hassanien, “Deep learning and feature fusion-based lung sound recognition model to diagnoses the respiratory diseases,”Soft Computing, pp. 1– 17, 2024. Chu Yun et al.:Preprint submitted to Elsevier Page 18 of 19 EZhouNet

    work page 2024

  20. [20]

    Super-cough: A super learner-based ensemble machine learning method for detecting disease on cough acousticsignals,

    E. K. Topuz and Y. Kaya, “Super-cough: A super learner-based ensemble machine learning method for detecting disease on cough acousticsignals,” BiomedicalSignalProcessingandControl ,vol.93, p. 106165, 2024

    work page 2024

  21. [21]

    Patch-level contrastive embedding learning for respiratory sound classification,

    W. Song and J. Han, “Patch-level contrastive embedding learning for respiratory sound classification,”Biomedical Signal Processing and Control, vol. 80, p. 104338, 2023

    work page 2023

  22. [22]

    Contrastive embeddind learning method for respiratory sound classification,

    W. Song, J. Han, and H. Song, “Contrastive embeddind learning method for respiratory sound classification,” inICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 1275–1279, IEEE, 2021

    work page 2021

  23. [23]

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

    S. Bae, J. Kim, W. Cho, H. Baek, S. Son, B. Lee, C. Ha, K. Tae, S. Kim, and S. Yun, “Patch-mix contrastive learning with audio spectrogram transformer on respiratory sound classification,” 2023

    work page 2023

  24. [24]

    Pretraining respiratory sound repre- sentations using metadata and contrastive learning,

    I. Moummad and N. Farrugia, “Pretraining respiratory sound repre- sentations using metadata and contrastive learning,” in2023 IEEE Workshop on Applications of Signal Processing to Audio and Acous- tics (WASPAA), pp. 1–5, IEEE, 2023

    work page 2023

  25. [25]

    Audioevent- relational graph representation learning for acoustic scene classifica- tion,

    Y.Hou,S.Song,C.Yu,W.Wang,andD.Botteldooren,“Audioevent- relational graph representation learning for acoustic scene classifica- tion,”IEEE signal processing letters, 2023

    work page 2023

  26. [26]

    Graph-based audio classifica- tion using pre-trained models and graph neural networks,

    A. E. Castro-Ospina, M. A. Solarte-Sanchez, L. S. Vega-Escobar, C. Isaza, and J. D. Martínez-Vargas, “Graph-based audio classifica- tion using pre-trained models and graph neural networks,”Sensors, vol. 24, no. 7, p. 2106, 2024

    work page 2024

  27. [27]

    Soundeventdetectionintrafficscenesbasedongraphconvolutional network to obtain multi-modal information,

    Y. Jiang, D. Guo, L. Wang, H. Zhang, H. Dong, Y. Qiu, and H. Zou, “Soundeventdetectionintrafficscenesbasedongraphconvolutional network to obtain multi-modal information,”Complex & Intelligent Systems, pp. 1–16, 2024

    work page 2024

  28. [28]

    Sound event detection utilizing graph laplacianregularizationwitheventco-occurrence,

    K. Imoto and S. Kyochi, “Sound event detection utilizing graph laplacianregularizationwitheventco-occurrence,” IEICETRANSAC- TIONS on Information and Systems, vol. 103, no. 9, pp. 1971–1977, 2020

    work page 1971

  29. [29]

    Musicy- olo: A vision-based framework for automatic singing transcription,

    X. Wang, B. Tian, W. Yang, W. Xu, and W. Cheng, “Musicy- olo: A vision-based framework for automatic singing transcription,” IEEE/ACM Transactions on Audio, Speech, and Language Process- ing, vol. 31, pp. 229–241, 2022

    work page 2022

  30. [30]

    Musicyolo: A sight- singingonset/offsetdetectionframeworkbasedonobjectdetectionin- steadofspectrumframes,

    X. Wang, W. Xu, W. Yang, and W. Cheng, “Musicyolo: A sight- singingonset/offsetdetectionframeworkbasedonobjectdetectionin- steadofspectrumframes,”in ICASSP2022-2022IEEEInternational Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 396–400, IEEE, 2022

    work page 2022

  31. [31]

    You only hear once: a yolo-like algorithm for audio segmentation and sound event detec- tion,

    S. Venkatesh, D. Moffat, and E. R. Miranda, “You only hear once: a yolo-like algorithm for audio segmentation and sound event detec- tion,”Applied Sciences, vol. 12, no. 7, p. 3293, 2022

    work page 2022

  32. [32]

    R-crnn: Region-based convolutional recurrent neural network for audio event detection,

    C.-C. Kao, W. Wang, M. Sun, and C. Wang, “R-crnn: Region-based convolutional recurrent neural network for audio event detection,” in Proc. Interspeech 2018, pp. 1358–1362, 2018

    work page 2018

  33. [33]

    Sound event detection transformer: An event-based end-to-end model for sound event detection,

    Z.Ye,X.Wang,H.Liu,Y.Qian,R.Tao,L.Yan,andK.Ouchi,“Sound event detection transformer: An event-based end-to-end model for sound event detection,”arXiv preprint arXiv:2110.02011, 2021

    Pith/arXiv arXiv 2021

  34. [34]

    Metrics for polyphonic sound event detection,

    A. Mesaros, T. Heittola, and T. Virtanen, “Metrics for polyphonic sound event detection,”Applied Sciences, vol. 6, no. 6, p. 162, 2016

    2016

  35. [35]

    Improvingsoundeventdetectionmetrics:insights from dcase 2020,

    G. Ferroni, N. Turpault, J. Azcarreta, F. Tuveri, R. Serizel, Ç. Bilen, andS.Krstulović,“Improvingsoundeventdetectionmetrics:insights from dcase 2020,” inICASSP 2021-2021 IEEE international confer- ence on acoustics, speech and signal processing (ICASSP), pp. 631– 635, IEEE, 2021

    work page 2020

  36. [36]

    A framework for the robust evaluation of sound event detection,

    Ç. Bilen, G. Ferroni, F. Tuveri, J. Azcarreta, and S. Krstulović, “A framework for the robust evaluation of sound event detection,” in ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 61–65, IEEE, 2020

    work page 2020

  37. [37]

    Graph-based semi- supervised one class support vector machine for detecting abnor- mal lung sounds,

    R. Lang, R. Lu, C. Zhao, H. Qin, and G. Liu, “Graph-based semi- supervised one class support vector machine for detecting abnor- mal lung sounds,”Applied Mathematics and Computation, vol. 364, p. 124487, 2020

    work page 2020

  38. [38]

    Analysisofunlabeledlungsound samples using semi-supervised convolutional neural networks,

    R.Lang,Y.Fan,G.Liu,andG.Liu,“Analysisofunlabeledlungsound samples using semi-supervised convolutional neural networks,”Ap- plied Mathematics and Computation, vol. 411, p. 126511, 2021

    work page 2021

  39. [39]

    Convolutional neural network for breathing phase detection in lung sounds,

    C. Jácome, J. Ravn, E. Holsbø, J. C. Aviles-Solis, H. Melbye, and L. Ailo Bongo, “Convolutional neural network for breathing phase detection in lung sounds,”Sensors, vol. 19, no. 8, p. 1798, 2019

    work page 2019

  40. [40]

    Benchmarking of eight recurrent neural network variants for breath phase and adventitious sound detection on a self-developed open- access lung sound database—hf_lung_v1,

    F.-S. Hsu, S.-R. Huang, C.-W. Huang, C.-J. Huang, Y.-R. Cheng, C.-C. Chen, J. Hsiao, C.-W. Chen, L.-C. Chen, Y.-C. Lai,et al., “Benchmarking of eight recurrent neural network variants for breath phase and adventitious sound detection on a self-developed open- access lung sound database—hf_lung_v1,”PLoS One, vol. 16, no. 7, p. e0254134, 2021

    work page 2021

  41. [41]

    A dual- purpose deep learning model for auscultated lung and tracheal sound analysis based on mixed set training,

    F.-S. Hsu, S.-R. Huang, C.-F. Su, C.-W. Huang, Y.-R. Cheng, C.-C. Chen, C.-Y. Wu, C.-W. Chen, Y.-C. Lai, T.-W. Cheng,et al., “A dual- purpose deep learning model for auscultated lung and tracheal sound analysis based on mixed set training,”Biomedical Signal Processing and Control, vol. 86, p. 105222, 2023

    work page 2023

  42. [42]

    A novel ai- inspired method and system implementation for detecting and clas- sifyingpediatricrespiratorysoundevents,

    C.-Y. Yeh, S.-A. Chiu, X.-Y. Deng, and W.-C. Fang, “A novel ai- inspired method and system implementation for detecting and clas- sifyingpediatricrespiratorysoundevents,”in 2024IEEEBiomedical Circuits and Systems Conference (BioCAS), pp. 1–5, IEEE, 2024

    work page 2024

  43. [43]

    Meta: Data compression and event detection grand challenge 2024 with sprsound dataset,

    Q. Zhang, C. Chen, S. Yuan, J. Zhang, J. Yuan, H. Huang, Y. Zhang, R. Pan, X. Jiang, J. Zhao,et al., “Meta: Data compression and event detection grand challenge 2024 with sprsound dataset,”IEEE Data Descriptions, 2024

    work page 2024

  44. [44]

    Automaticdiagnosisofglaucoma using two-dimensional fourier-bessel series expansion based empiri- cal wavelet transform,

    P.K.ChaudharyandR.B.Pachori,“Automaticdiagnosisofglaucoma using two-dimensional fourier-bessel series expansion based empiri- cal wavelet transform,”Biomedical Signal Processing and Control, vol. 64, p. 102237, 2021. Chu Yun et al.:Preprint submitted to Elsevier Page 19 of 19

    work page 2021

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.