PET image enhancement using artificial intelligence for better characterization of epilepsy lesions

Anthime Flaus; Tahya Deddah; Anthonin Reilhac; Nicolas De Leiris; Marc Janier; Ines Merida; Thomas Grenier; Colm J. McGinnity; Alexander Hammers; Carole Lartizien; Nicolas Costes

doi:10.3389/fmed.2022.1042706

PET image enhancement using artificial intelligence for better characterization of epilepsy lesions

Anthime Flaus^*, Tahya Deddah, Anthonin Reilhac, Nicolas De Leiris, Marc Janier, Ines Merida, Thomas Grenier, Colm J. McGinnity, Alexander Hammers, Carole Lartizien, Nicolas Costes

^*Corresponding author for this work

PET Imaging Centre Facility

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

9 Citations (Scopus)

Abstract

Introduction: [¹⁸F]fluorodeoxyglucose ([¹⁸F]FDG) brain PET is used clinically to detect small areas of decreased uptake associated with epileptogenic lesions, e.g., Focal Cortical Dysplasias (FCD) but its performance is limited due to spatial resolution and low contrast. We aimed to develop a deep learning-based PET image enhancement method using simulated PET to improve lesion visualization. Methods: We created 210 numerical brain phantoms (MRI segmented into 9 regions) and assigned 10 different plausible activity values (e.g., GM/WM ratios) resulting in 2100 ground truth high quality (GT-HQ) PET phantoms. With a validated Monte-Carlo PET simulator, we then created 2100 simulated standard quality (S-SQ) [¹⁸F]FDG scans. We trained a ResNet on 80% of this dataset (10% used for validation) to learn the mapping between S-SQ and GT-HQ PET, outputting a predicted HQ (P-HQ) PET. For the remaining 10%, we assessed Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Root Mean Squared Error (RMSE) against GT-HQ PET. For GM and WM, we computed recovery coefficients (RC) and coefficient of variation (COV). We also created lesioned GT-HQ phantoms, S-SQ PET and P-HQ PET with simulated small hypometabolic lesions characteristic of FCDs. We evaluated lesion detectability on S-SQ and P-HQ PET both visually and measuring the Relative Lesion Activity (RLA, measured activity in the reduced-activity ROI over the standard-activity ROI). Lastly, we applied our previously trained ResNet on 10 clinical epilepsy PETs to predict the corresponding HQ-PET and assessed image quality and confidence metrics. Results: Compared to S-SQ PET, P-HQ PET improved PNSR, SSIM and RMSE; significatively improved GM RCs (from 0.29 ± 0.03 to 0.79 ± 0.04) and WM RCs (from 0.49 ± 0.03 to 1 ± 0.05); mean COVs were not statistically different. Visual lesion detection improved from 38 to 75%, with average RLA decreasing from 0.83 ± 0.08 to 0.67 ± 0.14. Visual quality of P-HQ clinical PET improved as well as reader confidence. Conclusion: P-HQ PET showed improved image quality compared to S-SQ PET across several objective quantitative metrics and increased detectability of simulated lesions. In addition, the model generalized to clinical data. Further evaluation is required to study generalization of our method and to assess clinical performance in larger cohorts.

Original language	English
Article number	1042706
Journal	Frontiers in Medicine
Volume	9
DOIs	https://doi.org/10.3389/fmed.2022.1042706
Publication status	Published - 16 Nov 2022

Keywords

brain
clinical application
deblurring
deep learning
focal cortical dysplasia (FCD)
Monte-Carlo simulation
residual network
super resolution (SR)

Access to Document

10.3389/fmed.2022.1042706

Cite this

@article{05f4e6a820fe4c4d9a848d6e9d4950fa,

title = "PET image enhancement using artificial intelligence for better characterization of epilepsy lesions",

abstract = "Introduction: [18F]fluorodeoxyglucose ([18F]FDG) brain PET is used clinically to detect small areas of decreased uptake associated with epileptogenic lesions, e.g., Focal Cortical Dysplasias (FCD) but its performance is limited due to spatial resolution and low contrast. We aimed to develop a deep learning-based PET image enhancement method using simulated PET to improve lesion visualization. Methods: We created 210 numerical brain phantoms (MRI segmented into 9 regions) and assigned 10 different plausible activity values (e.g., GM/WM ratios) resulting in 2100 ground truth high quality (GT-HQ) PET phantoms. With a validated Monte-Carlo PET simulator, we then created 2100 simulated standard quality (S-SQ) [18F]FDG scans. We trained a ResNet on 80% of this dataset (10% used for validation) to learn the mapping between S-SQ and GT-HQ PET, outputting a predicted HQ (P-HQ) PET. For the remaining 10%, we assessed Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Root Mean Squared Error (RMSE) against GT-HQ PET. For GM and WM, we computed recovery coefficients (RC) and coefficient of variation (COV). We also created lesioned GT-HQ phantoms, S-SQ PET and P-HQ PET with simulated small hypometabolic lesions characteristic of FCDs. We evaluated lesion detectability on S-SQ and P-HQ PET both visually and measuring the Relative Lesion Activity (RLA, measured activity in the reduced-activity ROI over the standard-activity ROI). Lastly, we applied our previously trained ResNet on 10 clinical epilepsy PETs to predict the corresponding HQ-PET and assessed image quality and confidence metrics. Results: Compared to S-SQ PET, P-HQ PET improved PNSR, SSIM and RMSE; significatively improved GM RCs (from 0.29 ± 0.03 to 0.79 ± 0.04) and WM RCs (from 0.49 ± 0.03 to 1 ± 0.05); mean COVs were not statistically different. Visual lesion detection improved from 38 to 75%, with average RLA decreasing from 0.83 ± 0.08 to 0.67 ± 0.14. Visual quality of P-HQ clinical PET improved as well as reader confidence. Conclusion: P-HQ PET showed improved image quality compared to S-SQ PET across several objective quantitative metrics and increased detectability of simulated lesions. In addition, the model generalized to clinical data. Further evaluation is required to study generalization of our method and to assess clinical performance in larger cohorts.",

keywords = "brain, clinical application, deblurring, deep learning, focal cortical dysplasia (FCD), Monte-Carlo simulation, residual network, super resolution (SR)",

author = "Anthime Flaus and Tahya Deddah and Anthonin Reilhac and Leiris, {Nicolas De} and Marc Janier and Ines Merida and Thomas Grenier and McGinnity, {Colm J.} and Alexander Hammers and Carole Lartizien and Nicolas Costes",

note = "Funding Information: This research was funded in whole, or in part, by the Wellcome Trust [WT 203148/Z/16/Z]. For the purpose of open access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. AF received funding from the French branch of the International League Against Epilepsy (ILAE), Ligue Fran7aise contre l'Epilepsie (LFCE), the Labex Primes from Lyon University, Lyon and Hospices Civils de Lyon. Funding Information: The School of Biomedical Engineering and Imaging Sciences is supported by the Wellcome EPSRC Centre for Medical Engineering at King's College London [WT 203148/Z/16/Z] and the Department of Health via the National Institute for Health Research (NIHR) comprehensive Biomedical Research Centre award to Guy's & St Thomas' NHS Foundation Trust in partnership with King's College London and King's College Hospital NHS Foundation Trust. This work was supported by the LABEX PRIMES (ANR-11-LABX-0063) of Universit{\'e} de Lyon, within the program “Investissements d'Avenir” operated by the French National Research Agency (ANR). Publisher Copyright: Copyright {\textcopyright} 2022 Flaus, Deddah, Reilhac, Leiris, Janier, Merida, Grenier, McGinnity, Hammers, Lartizien and Costes.",

year = "2022",

month = nov,

day = "16",

doi = "10.3389/fmed.2022.1042706",

language = "English",

volume = "9",

journal = "Frontiers in Medicine",

issn = "2296-858X",

publisher = "Frontiers Media S.A.",

}

TY - JOUR

T1 - PET image enhancement using artificial intelligence for better characterization of epilepsy lesions

AU - Flaus, Anthime

AU - Deddah, Tahya

AU - Reilhac, Anthonin

AU - Leiris, Nicolas De

AU - Janier, Marc

AU - Merida, Ines

AU - Grenier, Thomas

AU - McGinnity, Colm J.

AU - Hammers, Alexander

AU - Lartizien, Carole

AU - Costes, Nicolas

N1 - Funding Information: This research was funded in whole, or in part, by the Wellcome Trust [WT 203148/Z/16/Z]. For the purpose of open access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. AF received funding from the French branch of the International League Against Epilepsy (ILAE), Ligue Fran7aise contre l'Epilepsie (LFCE), the Labex Primes from Lyon University, Lyon and Hospices Civils de Lyon. Funding Information: The School of Biomedical Engineering and Imaging Sciences is supported by the Wellcome EPSRC Centre for Medical Engineering at King's College London [WT 203148/Z/16/Z] and the Department of Health via the National Institute for Health Research (NIHR) comprehensive Biomedical Research Centre award to Guy's & St Thomas' NHS Foundation Trust in partnership with King's College London and King's College Hospital NHS Foundation Trust. This work was supported by the LABEX PRIMES (ANR-11-LABX-0063) of Université de Lyon, within the program “Investissements d'Avenir” operated by the French National Research Agency (ANR). Publisher Copyright: Copyright © 2022 Flaus, Deddah, Reilhac, Leiris, Janier, Merida, Grenier, McGinnity, Hammers, Lartizien and Costes.

PY - 2022/11/16

Y1 - 2022/11/16

N2 - Introduction: [18F]fluorodeoxyglucose ([18F]FDG) brain PET is used clinically to detect small areas of decreased uptake associated with epileptogenic lesions, e.g., Focal Cortical Dysplasias (FCD) but its performance is limited due to spatial resolution and low contrast. We aimed to develop a deep learning-based PET image enhancement method using simulated PET to improve lesion visualization. Methods: We created 210 numerical brain phantoms (MRI segmented into 9 regions) and assigned 10 different plausible activity values (e.g., GM/WM ratios) resulting in 2100 ground truth high quality (GT-HQ) PET phantoms. With a validated Monte-Carlo PET simulator, we then created 2100 simulated standard quality (S-SQ) [18F]FDG scans. We trained a ResNet on 80% of this dataset (10% used for validation) to learn the mapping between S-SQ and GT-HQ PET, outputting a predicted HQ (P-HQ) PET. For the remaining 10%, we assessed Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Root Mean Squared Error (RMSE) against GT-HQ PET. For GM and WM, we computed recovery coefficients (RC) and coefficient of variation (COV). We also created lesioned GT-HQ phantoms, S-SQ PET and P-HQ PET with simulated small hypometabolic lesions characteristic of FCDs. We evaluated lesion detectability on S-SQ and P-HQ PET both visually and measuring the Relative Lesion Activity (RLA, measured activity in the reduced-activity ROI over the standard-activity ROI). Lastly, we applied our previously trained ResNet on 10 clinical epilepsy PETs to predict the corresponding HQ-PET and assessed image quality and confidence metrics. Results: Compared to S-SQ PET, P-HQ PET improved PNSR, SSIM and RMSE; significatively improved GM RCs (from 0.29 ± 0.03 to 0.79 ± 0.04) and WM RCs (from 0.49 ± 0.03 to 1 ± 0.05); mean COVs were not statistically different. Visual lesion detection improved from 38 to 75%, with average RLA decreasing from 0.83 ± 0.08 to 0.67 ± 0.14. Visual quality of P-HQ clinical PET improved as well as reader confidence. Conclusion: P-HQ PET showed improved image quality compared to S-SQ PET across several objective quantitative metrics and increased detectability of simulated lesions. In addition, the model generalized to clinical data. Further evaluation is required to study generalization of our method and to assess clinical performance in larger cohorts.

AB - Introduction: [18F]fluorodeoxyglucose ([18F]FDG) brain PET is used clinically to detect small areas of decreased uptake associated with epileptogenic lesions, e.g., Focal Cortical Dysplasias (FCD) but its performance is limited due to spatial resolution and low contrast. We aimed to develop a deep learning-based PET image enhancement method using simulated PET to improve lesion visualization. Methods: We created 210 numerical brain phantoms (MRI segmented into 9 regions) and assigned 10 different plausible activity values (e.g., GM/WM ratios) resulting in 2100 ground truth high quality (GT-HQ) PET phantoms. With a validated Monte-Carlo PET simulator, we then created 2100 simulated standard quality (S-SQ) [18F]FDG scans. We trained a ResNet on 80% of this dataset (10% used for validation) to learn the mapping between S-SQ and GT-HQ PET, outputting a predicted HQ (P-HQ) PET. For the remaining 10%, we assessed Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Root Mean Squared Error (RMSE) against GT-HQ PET. For GM and WM, we computed recovery coefficients (RC) and coefficient of variation (COV). We also created lesioned GT-HQ phantoms, S-SQ PET and P-HQ PET with simulated small hypometabolic lesions characteristic of FCDs. We evaluated lesion detectability on S-SQ and P-HQ PET both visually and measuring the Relative Lesion Activity (RLA, measured activity in the reduced-activity ROI over the standard-activity ROI). Lastly, we applied our previously trained ResNet on 10 clinical epilepsy PETs to predict the corresponding HQ-PET and assessed image quality and confidence metrics. Results: Compared to S-SQ PET, P-HQ PET improved PNSR, SSIM and RMSE; significatively improved GM RCs (from 0.29 ± 0.03 to 0.79 ± 0.04) and WM RCs (from 0.49 ± 0.03 to 1 ± 0.05); mean COVs were not statistically different. Visual lesion detection improved from 38 to 75%, with average RLA decreasing from 0.83 ± 0.08 to 0.67 ± 0.14. Visual quality of P-HQ clinical PET improved as well as reader confidence. Conclusion: P-HQ PET showed improved image quality compared to S-SQ PET across several objective quantitative metrics and increased detectability of simulated lesions. In addition, the model generalized to clinical data. Further evaluation is required to study generalization of our method and to assess clinical performance in larger cohorts.

KW - brain

KW - clinical application

KW - deblurring

KW - deep learning

KW - focal cortical dysplasia (FCD)

KW - Monte-Carlo simulation

KW - residual network

KW - super resolution (SR)

UR - http://www.scopus.com/inward/record.url?scp=85143215015&partnerID=8YFLogxK

U2 - 10.3389/fmed.2022.1042706

DO - 10.3389/fmed.2022.1042706

M3 - Article

AN - SCOPUS:85143215015

SN - 2296-858X

VL - 9

JO - Frontiers in Medicine

JF - Frontiers in Medicine

M1 - 1042706

ER -

PET image enhancement using artificial intelligence for better characterization of epilepsy lesions

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this