Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans
Autor(a) principal: | |
---|---|
Data de Publicação: | 2023 |
Tipo de documento: | Dissertação |
Idioma: | eng |
Título da fonte: | Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
Texto Completo: | http://hdl.handle.net/10362/160558 |
Resumo: | This study aims to assess the feasibility of reducing the acquisition time of whole-body 18Flabelled fluorodeoxyglucose ([18F]FDG) positron emission tomography coupled with computed tomography (PET/CT) scans through deep-learning-based denoising. 112 whole-body [18F]FDG PET/CT scans of patients with cancer were included. 92 were employed in the training of three convolutional neural networks: 2D, 2.5D and 3D U-Nets. Mean squared error (MSE) was the appointed loss function. The remaining 20 scans were set aside for testing. The images were acquired on a Philips Vereos Digital PET/CT scanner. From the standard-duration (70 seconds per axial field of view (AFOV)) raw data, fast scans were simulated by cropping the data to 15, 20 and 30 s/AFOV. Reconstructionwas performed on-site using the manufacturer’s protocol and following EARL1 standards. MSE, structural similarity index measure (SSIM) and intraclass correlation coefficient (ICC) were used for a voxel-wise comparison between the deep-learning-denoised (DL-denoised) fast scans and the reference images (70 s/AFOV). Signal-to-noise ratio (SNR) was computed in regions with expected uptake uniformity (liver and lungs) through the quotient between the mean standardised uptake value (SUV) and the SUV standard deviation. On a tumour basis, quantification was performed in terms of maximum SUV, mean SUV, SUV standard deviation, peak SUV, total lesion glycolysis (TLG) and metabolic tumour volume (MTV) in both the lesions in the DLdenoised and reference images. For benchmarking, Gaussian filter (GF), the state-of-the-art denoising method, was implemented and its width optimised in the training set through MSE minimisation relatively to the reference images. The voxel-wise results revealed a strong agreement between the DL-denoised 15, 20 and 30-s/AFOV-based sets and the reference images, with an ICC equal or higher than 0.985. Quantification in the liver and lungs unveiled the DL-denoised images to have higher SNR compared to the original (fast), the GF-denoised and even the reference images. Tumour quantification exposed variations in the lesions’ features that are not expected to have clinical impact, particularly in the 20 and 30-s/AFOV-based sets. Deep-learning-based denoising outperformed optimised Gaussian filter in every instance. The deep-learning-based denoising models for fast whole-body [18F]FDG PET/CT scans developed in this study proved to have potential to achieve images with clinically-suitable quantitative parameters. The 20 s/AFOV scans with post-processing with the 2.5D U-Net or the 3D U-Net seemed to be the best compromise between scan duration and image quality, compared to the 15 and 30-s/AFOV-based scans. |
id |
RCAP_67efd1adbc07d1b58d0698b9cd064c06 |
---|---|
oai_identifier_str |
oai:run.unl.pt:10362/160558 |
network_acronym_str |
RCAP |
network_name_str |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
repository_id_str |
7160 |
spelling |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans[18F]FDG PET/CTdeep learningdenoisingmolecular imagingoncologyDomínio/Área Científica::Engenharia e Tecnologia::Outras Engenharias e TecnologiasThis study aims to assess the feasibility of reducing the acquisition time of whole-body 18Flabelled fluorodeoxyglucose ([18F]FDG) positron emission tomography coupled with computed tomography (PET/CT) scans through deep-learning-based denoising. 112 whole-body [18F]FDG PET/CT scans of patients with cancer were included. 92 were employed in the training of three convolutional neural networks: 2D, 2.5D and 3D U-Nets. Mean squared error (MSE) was the appointed loss function. The remaining 20 scans were set aside for testing. The images were acquired on a Philips Vereos Digital PET/CT scanner. From the standard-duration (70 seconds per axial field of view (AFOV)) raw data, fast scans were simulated by cropping the data to 15, 20 and 30 s/AFOV. Reconstructionwas performed on-site using the manufacturer’s protocol and following EARL1 standards. MSE, structural similarity index measure (SSIM) and intraclass correlation coefficient (ICC) were used for a voxel-wise comparison between the deep-learning-denoised (DL-denoised) fast scans and the reference images (70 s/AFOV). Signal-to-noise ratio (SNR) was computed in regions with expected uptake uniformity (liver and lungs) through the quotient between the mean standardised uptake value (SUV) and the SUV standard deviation. On a tumour basis, quantification was performed in terms of maximum SUV, mean SUV, SUV standard deviation, peak SUV, total lesion glycolysis (TLG) and metabolic tumour volume (MTV) in both the lesions in the DLdenoised and reference images. For benchmarking, Gaussian filter (GF), the state-of-the-art denoising method, was implemented and its width optimised in the training set through MSE minimisation relatively to the reference images. The voxel-wise results revealed a strong agreement between the DL-denoised 15, 20 and 30-s/AFOV-based sets and the reference images, with an ICC equal or higher than 0.985. Quantification in the liver and lungs unveiled the DL-denoised images to have higher SNR compared to the original (fast), the GF-denoised and even the reference images. Tumour quantification exposed variations in the lesions’ features that are not expected to have clinical impact, particularly in the 20 and 30-s/AFOV-based sets. Deep-learning-based denoising outperformed optimised Gaussian filter in every instance. The deep-learning-based denoising models for fast whole-body [18F]FDG PET/CT scans developed in this study proved to have potential to achieve images with clinically-suitable quantitative parameters. The 20 s/AFOV scans with post-processing with the 2.5D U-Net or the 3D U-Net seemed to be the best compromise between scan duration and image quality, compared to the 15 and 30-s/AFOV-based scans.O objetivo deste estudo é avaliar a exequibilidade da redução do tempo de aquisição de exames de tomografia por emissão de positrões aliada a tomografia computadorizada (PET/CT) com fluorodesoxiglicose marcada com flúor-18 ([18F]FDG) de corpo inteiro, através de um pósprocessamento com aprendizagem profunda (DL, do inglês deep learning). 112 exames PET/CT com [18F]FDG de corpo inteiro de pacientes com cancro foram incluídos. 92 foram utilizados no treino de três redes neuronais convolucionais: U-Nets 2D, 2.5D e 3D. O erro quadrático médio (MSE) foi utilizado como função-objetivo. Os restantes 20 exames foram reservados para teste. As imagens foram adquiridas num equipamento PET/CT Digital Philips Vereos. Das imagens padrão (70 segundos por campo de visão axial (AFOV)), imagens rápidas foram simuladas, cortando os dados para obter 15, 20 e 30 s/AFOV. A reconstrução foi feita localmente com o protocolo do fabricante e seguindo as normas EARL1. O MSE, a medida do índice de semelhança estrutural (SSIM) e o coeficiente de correlação intraclasse (ICC) foram utilizados para comparação vóxel-a-vóxel entre as imagens pós-processadas com DL e as de referência (70 s/AFOV). A razão sinal-ruído (SNR) foi calculada em regiões em que se espera captação uniforme (fígado e pulmões) através do quociente entre o valor médio de captação padrão (SUV) e o desvio padrão de SUV. Para a quantificação nos tumores, foram utilizados o SUV máximo, o SUV médio, o desvio padrão de SUV, o pico de SUV, a glicólise total da lesão (TLG) e o volume metabólico do tumor (MTV). Para análise comparativa, o filtro Gaussiano (GF) foi implementado e otimizado para o conjunto de treino por meio da minimização do MSE relativamente à referência. Os resultados da análise vóxel-a-vóxel revelaram forte concordância entre as imagens pósprocessadas com DL e as de referência, com um ICC igual ou superior a 0.985. A quantificação no fígado e nos pulmões mostrou que as imagens pós-processadas com DL apresentavam uma maior SNR do que as imagens rápidas originais, as pós-processadas com GF e mesmo do que as de referência. A quantificação nas lesões mostrou uma variação nas características que não se espera ter impacto clínico, em particular nas imagens resultantes das aquisições com 20 e 30 s/AFOV. O pós-processamento com DL superou o obtido com o GF em todas as ocasiões. A remoção de ruído de imagens PET/CT com [18F]FDG de corpo inteiro rápidas com base nos modelos de deep learning, desenvolvidos neste estudo, mostrou ter potencial para conseguir imagens com parâmetros quantitativos adequados para a clínica. As imagens adquiridas com 20 s/AFOV e pós-processamento, seja com a U-Net 2.5D seja com a 3D, pareceram ser o compromisso mais indicado entre a redução do tempo de aquisição e a qualidade de imagem.Oliveira, FranciscoVigário, RicardoRUNSilva, Luísa Castelbranco da Silveira Coelho2023-11-27T16:09:54Z2023-112023-11-01T00:00:00Zinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisapplication/pdfhttp://hdl.handle.net/10362/160558enginfo:eu-repo/semantics/openAccessreponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãoinstacron:RCAAP2024-03-11T05:43:17Zoai:run.unl.pt:10362/160558Portal AgregadorONGhttps://www.rcaap.pt/oai/openaireopendoar:71602024-03-20T03:58:06.333413Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãofalse |
dc.title.none.fl_str_mv |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans |
title |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans |
spellingShingle |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans Silva, Luísa Castelbranco da Silveira Coelho [18F]FDG PET/CT deep learning denoising molecular imaging oncology Domínio/Área Científica::Engenharia e Tecnologia::Outras Engenharias e Tecnologias |
title_short |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans |
title_full |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans |
title_fullStr |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans |
title_full_unstemmed |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans |
title_sort |
Development of deep-learning-based denoising algorithms for fast whole-body [18F]FDG PET/CT scans |
author |
Silva, Luísa Castelbranco da Silveira Coelho |
author_facet |
Silva, Luísa Castelbranco da Silveira Coelho |
author_role |
author |
dc.contributor.none.fl_str_mv |
Oliveira, Francisco Vigário, Ricardo RUN |
dc.contributor.author.fl_str_mv |
Silva, Luísa Castelbranco da Silveira Coelho |
dc.subject.por.fl_str_mv |
[18F]FDG PET/CT deep learning denoising molecular imaging oncology Domínio/Área Científica::Engenharia e Tecnologia::Outras Engenharias e Tecnologias |
topic |
[18F]FDG PET/CT deep learning denoising molecular imaging oncology Domínio/Área Científica::Engenharia e Tecnologia::Outras Engenharias e Tecnologias |
description |
This study aims to assess the feasibility of reducing the acquisition time of whole-body 18Flabelled fluorodeoxyglucose ([18F]FDG) positron emission tomography coupled with computed tomography (PET/CT) scans through deep-learning-based denoising. 112 whole-body [18F]FDG PET/CT scans of patients with cancer were included. 92 were employed in the training of three convolutional neural networks: 2D, 2.5D and 3D U-Nets. Mean squared error (MSE) was the appointed loss function. The remaining 20 scans were set aside for testing. The images were acquired on a Philips Vereos Digital PET/CT scanner. From the standard-duration (70 seconds per axial field of view (AFOV)) raw data, fast scans were simulated by cropping the data to 15, 20 and 30 s/AFOV. Reconstructionwas performed on-site using the manufacturer’s protocol and following EARL1 standards. MSE, structural similarity index measure (SSIM) and intraclass correlation coefficient (ICC) were used for a voxel-wise comparison between the deep-learning-denoised (DL-denoised) fast scans and the reference images (70 s/AFOV). Signal-to-noise ratio (SNR) was computed in regions with expected uptake uniformity (liver and lungs) through the quotient between the mean standardised uptake value (SUV) and the SUV standard deviation. On a tumour basis, quantification was performed in terms of maximum SUV, mean SUV, SUV standard deviation, peak SUV, total lesion glycolysis (TLG) and metabolic tumour volume (MTV) in both the lesions in the DLdenoised and reference images. For benchmarking, Gaussian filter (GF), the state-of-the-art denoising method, was implemented and its width optimised in the training set through MSE minimisation relatively to the reference images. The voxel-wise results revealed a strong agreement between the DL-denoised 15, 20 and 30-s/AFOV-based sets and the reference images, with an ICC equal or higher than 0.985. Quantification in the liver and lungs unveiled the DL-denoised images to have higher SNR compared to the original (fast), the GF-denoised and even the reference images. Tumour quantification exposed variations in the lesions’ features that are not expected to have clinical impact, particularly in the 20 and 30-s/AFOV-based sets. Deep-learning-based denoising outperformed optimised Gaussian filter in every instance. The deep-learning-based denoising models for fast whole-body [18F]FDG PET/CT scans developed in this study proved to have potential to achieve images with clinically-suitable quantitative parameters. The 20 s/AFOV scans with post-processing with the 2.5D U-Net or the 3D U-Net seemed to be the best compromise between scan duration and image quality, compared to the 15 and 30-s/AFOV-based scans. |
publishDate |
2023 |
dc.date.none.fl_str_mv |
2023-11-27T16:09:54Z 2023-11 2023-11-01T00:00:00Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://hdl.handle.net/10362/160558 |
url |
http://hdl.handle.net/10362/160558 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf |
dc.source.none.fl_str_mv |
reponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação instacron:RCAAP |
instname_str |
Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação |
instacron_str |
RCAAP |
institution |
RCAAP |
reponame_str |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
collection |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
repository.name.fl_str_mv |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação |
repository.mail.fl_str_mv |
|
_version_ |
1799138162964955136 |