Sequential Patch Analysis Framework for Lung Disease Classification

Kim Ngoc T. Le; Duc Tai Le; Hyunseung Choo

doi:10.1007/978-981-96-0437-1_24

Sequential Patch Analysis Framework for Lung Disease Classification

Kim Ngoc T. Le
, Duc Tai Le
, Hyunseung Choo

Department of Computer Science and Engineering

Sungkyunkwan University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Deep neural network (DNN) models have shown promise in lung disease classification using spectrogram transformations of respiratory sounds. However, these models typically process entire images, which can be computationally expensive, and they require large datasets for effective training. With the largest available respiratory dataset, ICBHI, containing only 920 recordings, training a robust DNN becomes challenging. To address this issue, this paper introduces a simplified DNN framework that incorporates novel techniques such as patch splitting and transfer learning. Specifically, the proposed framework splits the spectrogram image into a sequence of fixed-size patches, effectively reducing computational demands. These patches are processed with the position embeddings by DNN encoders to capture acoustic characteristics, leveraging a pre-trained ImageNet model to make efficient use of the limited dataset. Evaluation experiments on the ICBHI dataset show that the proposed method improves the state-of-the-art 8-class classification accuracy by 1.91%.

Original language	English
Title of host publication	Future Data and Security Engineering. Big Data, Security and Privacy, Smart City and Industry 4.0 Applications - 11th International Conference on Future Data and Security Engineering, FDSE 2024, Proceedings
Editors	Tran Khanh Dang, Josef Küng, Tai M. Chung
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	331-338
Number of pages	8
ISBN (Print)	9789819604364
DOIs	https://doi.org/10.1007/978-981-96-0437-1_24
State	Published - 2024
Event	11th International Conference on Future Data and Security Engineering, FDSE 2024 - Binh Duong, Viet Nam Duration: 27 Nov 2024 → 29 Nov 2024

Publication series

Name	Communications in Computer and Information Science
Volume	2310 CCIS
ISSN (Print)	1865-0929
ISSN (Electronic)	1865-0937

Conference

Conference	11th International Conference on Future Data and Security Engineering, FDSE 2024
Country/Territory	Viet Nam
City	Binh Duong
Period	27/11/24 → 29/11/24

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

ICBHI dataset
Lung disease
Multi-class classification
Pre-trained
Spectrogram

Access to Document

10.1007/978-981-96-0437-1_24

Cite this

Le, K. N. T., Le, D. T., & Choo, H. (2024). Sequential Patch Analysis Framework for Lung Disease Classification. In T. K. Dang, J. Küng, & T. M. Chung (Eds.), Future Data and Security Engineering. Big Data, Security and Privacy, Smart City and Industry 4.0 Applications - 11th International Conference on Future Data and Security Engineering, FDSE 2024, Proceedings (pp. 331-338). (Communications in Computer and Information Science; Vol. 2310 CCIS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-96-0437-1_24

Le, Kim Ngoc T. ; Le, Duc Tai ; Choo, Hyunseung. / Sequential Patch Analysis Framework for Lung Disease Classification. Future Data and Security Engineering. Big Data, Security and Privacy, Smart City and Industry 4.0 Applications - 11th International Conference on Future Data and Security Engineering, FDSE 2024, Proceedings. editor / Tran Khanh Dang ; Josef Küng ; Tai M. Chung. Springer Science and Business Media Deutschland GmbH, 2024. pp. 331-338 (Communications in Computer and Information Science).

@inproceedings{a033f4ad1ec544bda62e4349a7ad7c9f,

title = "Sequential Patch Analysis Framework for Lung Disease Classification",

abstract = "Deep neural network (DNN) models have shown promise in lung disease classification using spectrogram transformations of respiratory sounds. However, these models typically process entire images, which can be computationally expensive, and they require large datasets for effective training. With the largest available respiratory dataset, ICBHI, containing only 920 recordings, training a robust DNN becomes challenging. To address this issue, this paper introduces a simplified DNN framework that incorporates novel techniques such as patch splitting and transfer learning. Specifically, the proposed framework splits the spectrogram image into a sequence of fixed-size patches, effectively reducing computational demands. These patches are processed with the position embeddings by DNN encoders to capture acoustic characteristics, leveraging a pre-trained ImageNet model to make efficient use of the limited dataset. Evaluation experiments on the ICBHI dataset show that the proposed method improves the state-of-the-art 8-class classification accuracy by 1.91\%.",

keywords = "ICBHI dataset, Lung disease, Multi-class classification, Pre-trained, Spectrogram",

author = "Le, \{Kim Ngoc T.\} and Le, \{Duc Tai\} and Hyunseung Choo",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.; 11th International Conference on Future Data and Security Engineering, FDSE 2024 ; Conference date: 27-11-2024 Through 29-11-2024",

year = "2024",

doi = "10.1007/978-981-96-0437-1\_24",

language = "English",

isbn = "9789819604364",

series = "Communications in Computer and Information Science",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "331--338",

editor = "Dang, \{Tran Khanh\} and Josef K{\"u}ng and Chung, \{Tai M.\}",

booktitle = "Future Data and Security Engineering. Big Data, Security and Privacy, Smart City and Industry 4.0 Applications - 11th International Conference on Future Data and Security Engineering, FDSE 2024, Proceedings",

}

Le, KNT, Le, DT & Choo, H 2024, Sequential Patch Analysis Framework for Lung Disease Classification. in TK Dang, J Küng & TM Chung (eds), Future Data and Security Engineering. Big Data, Security and Privacy, Smart City and Industry 4.0 Applications - 11th International Conference on Future Data and Security Engineering, FDSE 2024, Proceedings. Communications in Computer and Information Science, vol. 2310 CCIS, Springer Science and Business Media Deutschland GmbH, pp. 331-338, 11th International Conference on Future Data and Security Engineering, FDSE 2024, Binh Duong, Viet Nam, 27/11/24. https://doi.org/10.1007/978-981-96-0437-1_24

Sequential Patch Analysis Framework for Lung Disease Classification. / Le, Kim Ngoc T.; Le, Duc Tai; Choo, Hyunseung.
Future Data and Security Engineering. Big Data, Security and Privacy, Smart City and Industry 4.0 Applications - 11th International Conference on Future Data and Security Engineering, FDSE 2024, Proceedings. ed. / Tran Khanh Dang; Josef Küng; Tai M. Chung. Springer Science and Business Media Deutschland GmbH, 2024. p. 331-338 (Communications in Computer and Information Science; Vol. 2310 CCIS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Sequential Patch Analysis Framework for Lung Disease Classification

AU - Le, Kim Ngoc T.

AU - Le, Duc Tai

AU - Choo, Hyunseung

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

PY - 2024

Y1 - 2024

N2 - Deep neural network (DNN) models have shown promise in lung disease classification using spectrogram transformations of respiratory sounds. However, these models typically process entire images, which can be computationally expensive, and they require large datasets for effective training. With the largest available respiratory dataset, ICBHI, containing only 920 recordings, training a robust DNN becomes challenging. To address this issue, this paper introduces a simplified DNN framework that incorporates novel techniques such as patch splitting and transfer learning. Specifically, the proposed framework splits the spectrogram image into a sequence of fixed-size patches, effectively reducing computational demands. These patches are processed with the position embeddings by DNN encoders to capture acoustic characteristics, leveraging a pre-trained ImageNet model to make efficient use of the limited dataset. Evaluation experiments on the ICBHI dataset show that the proposed method improves the state-of-the-art 8-class classification accuracy by 1.91%.

AB - Deep neural network (DNN) models have shown promise in lung disease classification using spectrogram transformations of respiratory sounds. However, these models typically process entire images, which can be computationally expensive, and they require large datasets for effective training. With the largest available respiratory dataset, ICBHI, containing only 920 recordings, training a robust DNN becomes challenging. To address this issue, this paper introduces a simplified DNN framework that incorporates novel techniques such as patch splitting and transfer learning. Specifically, the proposed framework splits the spectrogram image into a sequence of fixed-size patches, effectively reducing computational demands. These patches are processed with the position embeddings by DNN encoders to capture acoustic characteristics, leveraging a pre-trained ImageNet model to make efficient use of the limited dataset. Evaluation experiments on the ICBHI dataset show that the proposed method improves the state-of-the-art 8-class classification accuracy by 1.91%.

KW - ICBHI dataset

KW - Lung disease

KW - Multi-class classification

KW - Pre-trained

KW - Spectrogram

UR - https://www.scopus.com/pages/publications/85211235162

U2 - 10.1007/978-981-96-0437-1_24

DO - 10.1007/978-981-96-0437-1_24

M3 - Conference contribution

AN - SCOPUS:85211235162

SN - 9789819604364

T3 - Communications in Computer and Information Science

SP - 331

EP - 338

BT - Future Data and Security Engineering. Big Data, Security and Privacy, Smart City and Industry 4.0 Applications - 11th International Conference on Future Data and Security Engineering, FDSE 2024, Proceedings

A2 - Dang, Tran Khanh

A2 - Küng, Josef

A2 - Chung, Tai M.

PB - Springer Science and Business Media Deutschland GmbH

T2 - 11th International Conference on Future Data and Security Engineering, FDSE 2024

Y2 - 27 November 2024 through 29 November 2024

ER -

Le KNT, Le DT, Choo H. Sequential Patch Analysis Framework for Lung Disease Classification. In Dang TK, Küng J, Chung TM, editors, Future Data and Security Engineering. Big Data, Security and Privacy, Smart City and Industry 4.0 Applications - 11th International Conference on Future Data and Security Engineering, FDSE 2024, Proceedings. Springer Science and Business Media Deutschland GmbH. 2024. p. 331-338. (Communications in Computer and Information Science). doi: 10.1007/978-981-96-0437-1_24