Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization: A multicenter study using COVLIAS 2.0

Mohit Agarwal; Sushant Agarwal; Luca Saba; Gian Luca Chabert; Suneet Gupta; Alessandro Carriero; Alessio Pasche; Pietro Danna; Armin Mehmedovic; Gavino Faa; Saurabh Shrivastava; Kanishka Jain; Harsh Jain; Tanay Jujaray; Inder M. Singh; Monika Turk; Paramjit S. Chadha; Amer M. Johri; Narendra N. Khanna; Sophie Mavrogeni; John R. Laird; David W. Sobel; Martin Miner; Antonella Balestrieri; Petros P. Sfikakis; George Tsoulfas; Durga Prasanna Misra; Vikas Agarwal; George D. Kitas; Jagjit S. Teji; Mustafa Al-Maini; Surinder K. Dhanjil; Andrew Nicolaides; Aditya Sharma; Vijay Rathore; Mostafa Fatemi; Azra Alizad; Pudukode R. Krishnan; Rajanikant R. Yadav; Frence Nagy; Zsigmond Tamás Kincses; Zoltan Ruzsa; Subbaram Naidu; Klaudija Viskovic; Manudeep K. Kalra; Jasjit S. Suri

doi:10.1016/j.compbiomed.2022.105571

Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization: A multicenter study using COVLIAS 2.0

Mohit Agarwal
, Sushant Agarwal
, Luca Saba
, Gian Luca Chabert
, Suneet Gupta
, Alessandro Carriero
, Alessio Pasche
, Pietro Danna
, Armin Mehmedovic
, Gavino Faa
, Saurabh Shrivastava
, Kanishka Jain
, Harsh Jain
, Tanay Jujaray
, Inder M. Singh
, Monika Turk
, Paramjit S. Chadha
, Amer M. Johri
, Narendra N. Khanna
, Sophie Mavrogeni

John R. Laird, David W. Sobel, Martin Miner, Antonella Balestrieri, Petros P. Sfikakis, George Tsoulfas, Durga Prasanna Misra, Vikas Agarwal, George D. Kitas, Jagjit S. Teji, Mustafa Al-Maini, Surinder K. Dhanjil, Andrew Nicolaides, Aditya Sharma, Vijay Rathore, Mostafa Fatemi, Azra Alizad, Pudukode R. Krishnan, Rajanikant R. Yadav, Frence Nagy, Zsigmond Tamás Kincses, Zoltan Ruzsa, Subbaram Naidu, Klaudija Viskovic, Manudeep K. Kalra, Jasjit S. Suri

Research output: Contribution to journal › Article › peer-review

50 Scopus citations

Abstract

BACKGROUND: COVLIAS 1.0: an automated lung segmentation was designed for COVID-19 diagnosis. It has issues related to storage space and speed. This study shows that COVLIAS 2.0 uses pruned AI (PAI) networks for improving both storage and speed, wiliest high performance on lung segmentation and lesion localization.

METHOD: ology: The proposed study uses multicenter ∼9,000 CT slices from two different nations, namely, CroMed from Croatia (80 patients, experimental data), and NovMed from Italy (72 patients, validation data). We hypothesize that by using pruning and evolutionary optimization algorithms, the size of the AI models can be reduced significantly, ensuring optimal performance. Eight different pruning techniques (i) differential evolution (DE), (ii) genetic algorithm (GA), (iii) particle swarm optimization algorithm (PSO), and (iv) whale optimization algorithm (WO) in two deep learning frameworks (i) Fully connected network (FCN) and (ii) SegNet were designed. COVLIAS 2.0 was validated using "Unseen NovMed" and benchmarked against MedSeg. Statistical tests for stability and reliability were also conducted.

RESULTS: Pruning algorithms (i) FCN-DE, (ii) FCN-GA, (iii) FCN-PSO, and (iv) FCN-WO showed improvement in storage by 92.4%, 95.3%, 98.7%, and 99.8% respectively when compared against solo FCN, and (v) SegNet-DE, (vi) SegNet-GA, (vii) SegNet-PSO, and (viii) SegNet-WO showed improvement by 97.1%, 97.9%, 98.8%, and 99.2% respectively when compared against solo SegNet. AUC > 0.94 (p < 0.0001) on CroMed and > 0.86 (p < 0.0001) on NovMed data set for all eight EA model. PAI <0.25 s per image. DenseNet-121-based Grad-CAM heatmaps showed validation on glass ground opacity lesions.

CONCLUSIONS: Eight PAI networks that were successfully validated are five times faster, storage efficient, and could be used in clinical settings.

Original language	English (US)
Article number	105571
Journal	Computers in Biology and Medicine
Volume	146
DOIs	https://doi.org/10.1016/j.compbiomed.2022.105571
State	Published - Jul 2022

Bibliographical note

Publisher Copyright:
© 2022

UN SDGs

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

SDG 3 Good Health and Well-being

Keywords

AI
COVID-19
Deep learning
Glass ground opacities
Hounsfield units
Lung CT
Lung segmentation
Pruning
Neural Networks, Computer
Reproducibility of Results
Lung/diagnostic imaging
Tomography, X-Ray Computed/methods
COVID-19 Testing
Humans
Deep Learning
COVID-19/diagnostic imaging
Image Processing, Computer-Assisted/methods

PubMed: MeSH publication types

Multicenter Study
Journal Article

Publisher link

10.1016/j.compbiomed.2022.105571

Find It

Find It at U of M Libraries

Cite this

Agarwal, M., Agarwal, S., Saba, L., Chabert, G. L., Gupta, S., Carriero, A., Pasche, A., Danna, P., Mehmedovic, A., Faa, G., Shrivastava, S., Jain, K., Jain, H., Jujaray, T., Singh, I. M., Turk, M., Chadha, P. S., Johri, A. M., Khanna, N. N., ... Suri, J. S. (2022). Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization: A multicenter study using COVLIAS 2.0. Computers in Biology and Medicine, 146, Article 105571. https://doi.org/10.1016/j.compbiomed.2022.105571

Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization: A multicenter study using COVLIAS 2.0. / Agarwal, Mohit; Agarwal, Sushant; Saba, Luca et al.
In: Computers in Biology and Medicine, Vol. 146, 105571, 07.2022.

Research output: Contribution to journal › Article › peer-review

Agarwal, M, Agarwal, S, Saba, L, Chabert, GL, Gupta, S, Carriero, A, Pasche, A, Danna, P, Mehmedovic, A, Faa, G, Shrivastava, S, Jain, K, Jain, H, Jujaray, T, Singh, IM, Turk, M, Chadha, PS, Johri, AM, Khanna, NN, Mavrogeni, S, Laird, JR, Sobel, DW, Miner, M, Balestrieri, A, Sfikakis, PP, Tsoulfas, G, Misra, DP, Agarwal, V, Kitas, GD, Teji, JS, Al-Maini, M, Dhanjil, SK, Nicolaides, A, Sharma, A, Rathore, V, Fatemi, M, Alizad, A, Krishnan, PR, Yadav, RR, Nagy, F, Kincses, ZT, Ruzsa, Z, Naidu, S, Viskovic, K, Kalra, MK & Suri, JS 2022, 'Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization: A multicenter study using COVLIAS 2.0', Computers in Biology and Medicine, vol. 146, 105571. https://doi.org/10.1016/j.compbiomed.2022.105571

@article{a736fcc4d21c43c1b788ffe569ec6cd6,

title = "Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization: A multicenter study using COVLIAS 2.0",

abstract = "BACKGROUND: COVLIAS 1.0: an automated lung segmentation was designed for COVID-19 diagnosis. It has issues related to storage space and speed. This study shows that COVLIAS 2.0 uses pruned AI (PAI) networks for improving both storage and speed, wiliest high performance on lung segmentation and lesion localization.METHOD: ology: The proposed study uses multicenter ∼9,000 CT slices from two different nations, namely, CroMed from Croatia (80 patients, experimental data), and NovMed from Italy (72 patients, validation data). We hypothesize that by using pruning and evolutionary optimization algorithms, the size of the AI models can be reduced significantly, ensuring optimal performance. Eight different pruning techniques (i) differential evolution (DE), (ii) genetic algorithm (GA), (iii) particle swarm optimization algorithm (PSO), and (iv) whale optimization algorithm (WO) in two deep learning frameworks (i) Fully connected network (FCN) and (ii) SegNet were designed. COVLIAS 2.0 was validated using {"}Unseen NovMed{"} and benchmarked against MedSeg. Statistical tests for stability and reliability were also conducted.RESULTS: Pruning algorithms (i) FCN-DE, (ii) FCN-GA, (iii) FCN-PSO, and (iv) FCN-WO showed improvement in storage by 92.4\%, 95.3\%, 98.7\%, and 99.8\% respectively when compared against solo FCN, and (v) SegNet-DE, (vi) SegNet-GA, (vii) SegNet-PSO, and (viii) SegNet-WO showed improvement by 97.1\%, 97.9\%, 98.8\%, and 99.2\% respectively when compared against solo SegNet. AUC > 0.94 (p < 0.0001) on CroMed and > 0.86 (p < 0.0001) on NovMed data set for all eight EA model. PAI <0.25 s per image. DenseNet-121-based Grad-CAM heatmaps showed validation on glass ground opacity lesions.CONCLUSIONS: Eight PAI networks that were successfully validated are five times faster, storage efficient, and could be used in clinical settings.",

keywords = "AI, COVID-19, Deep learning, Glass ground opacities, Hounsfield units, Lung CT, Lung segmentation, Pruning, Neural Networks, Computer, Reproducibility of Results, Lung/diagnostic imaging, Tomography, X-Ray Computed/methods, COVID-19 Testing, Humans, Deep Learning, COVID-19/diagnostic imaging, Image Processing, Computer-Assisted/methods",

author = "Mohit Agarwal and Sushant Agarwal and Luca Saba and Chabert, \{Gian Luca\} and Suneet Gupta and Alessandro Carriero and Alessio Pasche and Pietro Danna and Armin Mehmedovic and Gavino Faa and Saurabh Shrivastava and Kanishka Jain and Harsh Jain and Tanay Jujaray and Singh, \{Inder M.\} and Monika Turk and Chadha, \{Paramjit S.\} and Johri, \{Amer M.\} and Khanna, \{Narendra N.\} and Sophie Mavrogeni and Laird, \{John R.\} and Sobel, \{David W.\} and Martin Miner and Antonella Balestrieri and Sfikakis, \{Petros P.\} and George Tsoulfas and Misra, \{Durga Prasanna\} and Vikas Agarwal and Kitas, \{George D.\} and Teji, \{Jagjit S.\} and Mustafa Al-Maini and Dhanjil, \{Surinder K.\} and Andrew Nicolaides and Aditya Sharma and Vijay Rathore and Mostafa Fatemi and Azra Alizad and Krishnan, \{Pudukode R.\} and Yadav, \{Rajanikant R.\} and Frence Nagy and Kincses, \{Zsigmond Tam{\'a}s\} and Zoltan Ruzsa and Subbaram Naidu and Klaudija Viskovic and Kalra, \{Manudeep K.\} and Suri, \{Jasjit S.\}",

note = "Publisher Copyright: {\textcopyright} 2022",

year = "2022",

month = jul,

doi = "10.1016/j.compbiomed.2022.105571",

language = "English (US)",

volume = "146",

journal = "Computers in Biology and Medicine",

issn = "0010-4825",

publisher = "Elsevier Ltd",

}

TY - JOUR

T1 - Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization

T2 - A multicenter study using COVLIAS 2.0

AU - Agarwal, Mohit

AU - Agarwal, Sushant

AU - Saba, Luca

AU - Chabert, Gian Luca

AU - Gupta, Suneet

AU - Carriero, Alessandro

AU - Pasche, Alessio

AU - Danna, Pietro

AU - Mehmedovic, Armin

AU - Faa, Gavino

AU - Shrivastava, Saurabh

AU - Jain, Kanishka

AU - Jain, Harsh

AU - Jujaray, Tanay

AU - Singh, Inder M.

AU - Turk, Monika

AU - Chadha, Paramjit S.

AU - Johri, Amer M.

AU - Khanna, Narendra N.

AU - Mavrogeni, Sophie

AU - Laird, John R.

AU - Sobel, David W.

AU - Miner, Martin

AU - Balestrieri, Antonella

AU - Sfikakis, Petros P.

AU - Tsoulfas, George

AU - Misra, Durga Prasanna

AU - Agarwal, Vikas

AU - Kitas, George D.

AU - Teji, Jagjit S.

AU - Al-Maini, Mustafa

AU - Dhanjil, Surinder K.

AU - Nicolaides, Andrew

AU - Sharma, Aditya

AU - Rathore, Vijay

AU - Fatemi, Mostafa

AU - Alizad, Azra

AU - Krishnan, Pudukode R.

AU - Yadav, Rajanikant R.

AU - Nagy, Frence

AU - Kincses, Zsigmond Tamás

AU - Ruzsa, Zoltan

AU - Naidu, Subbaram

AU - Viskovic, Klaudija

AU - Kalra, Manudeep K.

AU - Suri, Jasjit S.

PY - 2022/7

Y1 - 2022/7

N2 - BACKGROUND: COVLIAS 1.0: an automated lung segmentation was designed for COVID-19 diagnosis. It has issues related to storage space and speed. This study shows that COVLIAS 2.0 uses pruned AI (PAI) networks for improving both storage and speed, wiliest high performance on lung segmentation and lesion localization.METHOD: ology: The proposed study uses multicenter ∼9,000 CT slices from two different nations, namely, CroMed from Croatia (80 patients, experimental data), and NovMed from Italy (72 patients, validation data). We hypothesize that by using pruning and evolutionary optimization algorithms, the size of the AI models can be reduced significantly, ensuring optimal performance. Eight different pruning techniques (i) differential evolution (DE), (ii) genetic algorithm (GA), (iii) particle swarm optimization algorithm (PSO), and (iv) whale optimization algorithm (WO) in two deep learning frameworks (i) Fully connected network (FCN) and (ii) SegNet were designed. COVLIAS 2.0 was validated using "Unseen NovMed" and benchmarked against MedSeg. Statistical tests for stability and reliability were also conducted.RESULTS: Pruning algorithms (i) FCN-DE, (ii) FCN-GA, (iii) FCN-PSO, and (iv) FCN-WO showed improvement in storage by 92.4%, 95.3%, 98.7%, and 99.8% respectively when compared against solo FCN, and (v) SegNet-DE, (vi) SegNet-GA, (vii) SegNet-PSO, and (viii) SegNet-WO showed improvement by 97.1%, 97.9%, 98.8%, and 99.2% respectively when compared against solo SegNet. AUC > 0.94 (p < 0.0001) on CroMed and > 0.86 (p < 0.0001) on NovMed data set for all eight EA model. PAI <0.25 s per image. DenseNet-121-based Grad-CAM heatmaps showed validation on glass ground opacity lesions.CONCLUSIONS: Eight PAI networks that were successfully validated are five times faster, storage efficient, and could be used in clinical settings.

AB - BACKGROUND: COVLIAS 1.0: an automated lung segmentation was designed for COVID-19 diagnosis. It has issues related to storage space and speed. This study shows that COVLIAS 2.0 uses pruned AI (PAI) networks for improving both storage and speed, wiliest high performance on lung segmentation and lesion localization.METHOD: ology: The proposed study uses multicenter ∼9,000 CT slices from two different nations, namely, CroMed from Croatia (80 patients, experimental data), and NovMed from Italy (72 patients, validation data). We hypothesize that by using pruning and evolutionary optimization algorithms, the size of the AI models can be reduced significantly, ensuring optimal performance. Eight different pruning techniques (i) differential evolution (DE), (ii) genetic algorithm (GA), (iii) particle swarm optimization algorithm (PSO), and (iv) whale optimization algorithm (WO) in two deep learning frameworks (i) Fully connected network (FCN) and (ii) SegNet were designed. COVLIAS 2.0 was validated using "Unseen NovMed" and benchmarked against MedSeg. Statistical tests for stability and reliability were also conducted.RESULTS: Pruning algorithms (i) FCN-DE, (ii) FCN-GA, (iii) FCN-PSO, and (iv) FCN-WO showed improvement in storage by 92.4%, 95.3%, 98.7%, and 99.8% respectively when compared against solo FCN, and (v) SegNet-DE, (vi) SegNet-GA, (vii) SegNet-PSO, and (viii) SegNet-WO showed improvement by 97.1%, 97.9%, 98.8%, and 99.2% respectively when compared against solo SegNet. AUC > 0.94 (p < 0.0001) on CroMed and > 0.86 (p < 0.0001) on NovMed data set for all eight EA model. PAI <0.25 s per image. DenseNet-121-based Grad-CAM heatmaps showed validation on glass ground opacity lesions.CONCLUSIONS: Eight PAI networks that were successfully validated are five times faster, storage efficient, and could be used in clinical settings.

KW - AI

KW - COVID-19

KW - Deep learning

KW - Glass ground opacities

KW - Hounsfield units

KW - Lung CT

KW - Lung segmentation

KW - Pruning

KW - Neural Networks, Computer

KW - Reproducibility of Results

KW - Lung/diagnostic imaging

KW - Tomography, X-Ray Computed/methods

KW - COVID-19 Testing

KW - Humans

KW - Deep Learning

KW - COVID-19/diagnostic imaging

KW - Image Processing, Computer-Assisted/methods

UR - https://www.scopus.com/pages/publications/85133101763

UR - https://www.scopus.com/pages/publications/85133101763#tab=citedBy

U2 - 10.1016/j.compbiomed.2022.105571

DO - 10.1016/j.compbiomed.2022.105571

M3 - Article

C2 - 35751196

AN - SCOPUS:85133101763

SN - 0010-4825

VL - 146

JO - Computers in Biology and Medicine

JF - Computers in Biology and Medicine

M1 - 105571

ER -

Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization: A multicenter study using COVLIAS 2.0

Abstract

Bibliographical note

UN SDGs

Keywords

PubMed: MeSH publication types

Publisher link

Find It

Other files and links

Fingerprint

Cite this