COVID-19 Diagnosis in 3D Chest CT Scans with Attention-Based Models

Kathrin Hartmann; Enrique Hortal

doi:10.1007/978-3-031-34344-5_27

COVID-19 Diagnosis in 3D Chest CT Scans with Attention-Based Models

Kathrin Hartmann, Enrique Hortal^*

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Chapter › Academic

9 Downloads (Pure)

Abstract

The three-dimensional information in CT scans reveals notorious findings in the medical context, also for detecting symptoms of COVID-19 in chest CT scans. However, due to the lack of availability of large-scale datasets in 3D, the use of attention-based models in this field is proven to be difficult. With transfer learning, this work tackles this problem, investigating the performance of a pre-trained TimeSformer model, which was originally developed for video classification, on COVID-19 classification of three-dimensional chest CT scans. The attention-based model outperforms a DenseNet baseline. Furthermore, we propose three new attention schemes for TimeSformer improving the accuracy of the model by 1.5% and reducing runtime by almost 25% compared to the original attention scheme.

Original language	English
Title of host publication	Artificial Intelligence in Medicine
Subtitle of host publication	AIME 2023
Editors	J.M. Juarez, M. Marcos, G. Stiglic, A. Tucker
Publisher	Springer, Cham
Pages	229-238
Volume	13897
ISBN (Electronic)	978-3-031-34344-5
ISBN (Print)	978-3-031-34343-8
DOIs	https://doi.org/10.1007/978-3-031-34344-5_27
Publication status	Published - 2023
Event	International Conference on Artificial Intelligence in Medicine - Singapore, Singapore, Singapore Duration: 5 Aug 2023 → 7 Aug 2023 https://iaim2023.sg/

Publication series

Series	Lecture Notes in Computer Science
Volume	13897
ISSN	0302-9743

Conference

Conference	International Conference on Artificial Intelligence in Medicine
Abbreviated title	iAIM 2023
Country/Territory	Singapore
City	Singapore
Period	5/08/23 → 7/08/23
Internet address	https://iaim2023.sg/

Access to Document

10.1007/978-3-031-34344-5_27

Full TextFinal published version, 765 KBLicence: Taverne

https://link.springer.com/chapter/10.1007/978-3-031-34344-5_27#chapter-info

Cite this

@inbook{79aef22d2a3942f0a52bab35cc298d32,

title = "COVID-19 Diagnosis in 3D Chest CT Scans with Attention-Based Models",

abstract = "The three-dimensional information in CT scans reveals notorious findings in the medical context, also for detecting symptoms of COVID-19 in chest CT scans. However, due to the lack of availability of large-scale datasets in 3D, the use of attention-based models in this field is proven to be difficult. With transfer learning, this work tackles this problem, investigating the performance of a pre-trained TimeSformer model, which was originally developed for video classification, on COVID-19 classification of three-dimensional chest CT scans. The attention-based model outperforms a DenseNet baseline. Furthermore, we propose three new attention schemes for TimeSformer improving the accuracy of the model by 1.5% and reducing runtime by almost 25% compared to the original attention scheme.",

author = "Kathrin Hartmann and Enrique Hortal",

year = "2023",

doi = "10.1007/978-3-031-34344-5_27",

language = "English",

isbn = "978-3-031-34343-8",

volume = "13897",

series = "Lecture Notes in Computer Science",

publisher = "Springer, Cham",

pages = "229--238",

editor = "J.M. Juarez and M. Marcos and G. Stiglic and A. Tucker",

booktitle = "Artificial Intelligence in Medicine",

address = "Switzerland",

note = "International Conference on Artificial Intelligence in Medicine, iAIM 2023 ; Conference date: 05-08-2023 Through 07-08-2023",

url = "https://iaim2023.sg/",

}

Hartmann, K & Hortal, E 2023, COVID-19 Diagnosis in 3D Chest CT Scans with Attention-Based Models. in JM Juarez, M Marcos, G Stiglic & A Tucker (eds), Artificial Intelligence in Medicine: AIME 2023. vol. 13897, Springer, Cham, Lecture Notes in Computer Science, vol. 13897, pp. 229-238, International Conference on Artificial Intelligence in Medicine, Singapore, Singapore, 5/08/23. https://doi.org/10.1007/978-3-031-34344-5_27

COVID-19 Diagnosis in 3D Chest CT Scans with Attention-Based Models. / Hartmann, Kathrin; Hortal, Enrique.
Artificial Intelligence in Medicine: AIME 2023. ed. / J.M. Juarez; M. Marcos; G. Stiglic; A. Tucker. Vol. 13897 Springer, Cham, 2023. p. 229-238 (Lecture Notes in Computer Science, Vol. 13897).

Research output: Chapter in Book/Report/Conference proceeding › Chapter › Academic

TY - CHAP

T1 - COVID-19 Diagnosis in 3D Chest CT Scans with Attention-Based Models

AU - Hartmann, Kathrin

AU - Hortal, Enrique

PY - 2023

Y1 - 2023

N2 - The three-dimensional information in CT scans reveals notorious findings in the medical context, also for detecting symptoms of COVID-19 in chest CT scans. However, due to the lack of availability of large-scale datasets in 3D, the use of attention-based models in this field is proven to be difficult. With transfer learning, this work tackles this problem, investigating the performance of a pre-trained TimeSformer model, which was originally developed for video classification, on COVID-19 classification of three-dimensional chest CT scans. The attention-based model outperforms a DenseNet baseline. Furthermore, we propose three new attention schemes for TimeSformer improving the accuracy of the model by 1.5% and reducing runtime by almost 25% compared to the original attention scheme.

AB - The three-dimensional information in CT scans reveals notorious findings in the medical context, also for detecting symptoms of COVID-19 in chest CT scans. However, due to the lack of availability of large-scale datasets in 3D, the use of attention-based models in this field is proven to be difficult. With transfer learning, this work tackles this problem, investigating the performance of a pre-trained TimeSformer model, which was originally developed for video classification, on COVID-19 classification of three-dimensional chest CT scans. The attention-based model outperforms a DenseNet baseline. Furthermore, we propose three new attention schemes for TimeSformer improving the accuracy of the model by 1.5% and reducing runtime by almost 25% compared to the original attention scheme.

U2 - 10.1007/978-3-031-34344-5_27

DO - 10.1007/978-3-031-34344-5_27

M3 - Chapter

SN - 978-3-031-34343-8

VL - 13897

T3 - Lecture Notes in Computer Science

SP - 229

EP - 238

BT - Artificial Intelligence in Medicine

A2 - Juarez, J.M.

A2 - Marcos, M.

A2 - Stiglic, G.

A2 - Tucker, A.

PB - Springer, Cham

T2 - International Conference on Artificial Intelligence in Medicine

Y2 - 5 August 2023 through 7 August 2023

ER -