Loading…
TMTV-Net: fully automated total metabolic tumor volume segmentation in lymphoma PET/CT images — a multi-center generalizability analysis
Purpose Total metabolic tumor volume (TMTV) segmentation has significant value enabling quantitative imaging biomarkers for lymphoma management. In this work, we tackle the challenging task of automated tumor delineation in lymphoma from PET/CT scans using a cascaded approach. Methods Our study incl...
Saved in:
| Published in: | European journal of nuclear medicine and molecular imaging 2024-06, Vol.51 (7), p.1937-1954 |
|---|---|
| Main Authors: | , , , , , , , , , , , , , , , , |
| Format: | Article |
| Language: | English |
| Subjects: | |
| Citations: | Items that this one cites Items that cite this one |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| cited_by | cdi_FETCH-LOGICAL-c375t-c5667678ebd0d7eee6ee9f94754bad85c340bf38e876432060153467ed1951c83 |
|---|---|
| cites | cdi_FETCH-LOGICAL-c375t-c5667678ebd0d7eee6ee9f94754bad85c340bf38e876432060153467ed1951c83 |
| container_end_page | 1954 |
| container_issue | 7 |
| container_start_page | 1937 |
| container_title | European journal of nuclear medicine and molecular imaging |
| container_volume | 51 |
| creator | Yousefirizi, Fereshteh Klyuzhin, Ivan S. O, Joo Hyun Harsini, Sara Tie, Xin Shiri, Isaac Shin, Muheon Lee, Changhee Cho, Steve Y. Bradshaw, Tyler J. Zaidi, Habib Bénard, François Sehn, Laurie H. Savage, Kerry J. Steidl, Christian Uribe, Carlos F. Rahmim, Arman |
| description | Purpose
Total metabolic tumor volume (TMTV) segmentation has significant value enabling quantitative imaging biomarkers for lymphoma management. In this work, we tackle the challenging task of automated tumor delineation in lymphoma from PET/CT scans using a cascaded approach.
Methods
Our study included 1418 2-[
18
F]FDG PET/CT scans from four different centers. The dataset was divided into 900 scans for development/validation/testing phases and 518 for multi-center external testing. The former consisted of 450 lymphoma, lung cancer, and melanoma scans, along with 450 negative scans, while the latter consisted of lymphoma patients from different centers with diffuse large B cell, primary mediastinal large B cell, and classic Hodgkin lymphoma cases. Our approach involves resampling PET/CT images into different voxel sizes in the first step, followed by training multi-resolution 3D U-Nets on each resampled dataset using a fivefold cross-validation scheme. The models trained on different data splits were ensemble. After applying soft voting to the predicted masks, in the second step, we input the probability-averaged predictions, along with the input imaging data, into another 3D U-Net. Models were trained with semi-supervised loss. We additionally considered the effectiveness of using test time augmentation (TTA) to improve the segmentation performance after training. In addition to quantitative analysis including Dice score (DSC) and TMTV comparisons, the qualitative evaluation was also conducted by nuclear medicine physicians.
Results
Our cascaded soft-voting guided approach resulted in performance with an average DSC of 0.68 ± 0.12 for the internal test data from developmental dataset, and an average DSC of 0.66 ± 0.18 on the multi-site external data (
n
= 518), significantly outperforming (
p
|
| doi_str_mv | 10.1007/s00259-024-06616-x |
| format | article |
| fullrecord | <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_miscellaneous_2923909278</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2923909278</sourcerecordid><originalsourceid>FETCH-LOGICAL-c375t-c5667678ebd0d7eee6ee9f94754bad85c340bf38e876432060153467ed1951c83</originalsourceid><addsrcrecordid>eNp9kbtuFDEUhkcIRELgBSiQJRoaE19mfKFDq3CRwqUYaC3PzJnFkT1ebA_KUlFT84Q8CQ4bgkRBZUv-_v8c62uah5Q8pYTI00wI6zQmrMVECCrw5a3mmAqqsSRK3765S3LU3Mv5ghCqmNJ3myOuOBOi646b7_2b_iN-C-UZmlfv98iuJQZbYEIlFutRgGKH6N2IyhpiQl-iXwOgDNsAS7HFxQW5Bfl92H2qQfT-rD_d9MgFu4WMfn77gSwKqy8Oj5WHhLawQLLefbWD867UiYv1--zy_ebObH2GB9fnSfPhxVm_eYXP3718vXl-jkcuu4LHTggppIJhIpMEAAGgZ93Krh3spLqRt2SYuQIlRcsZEYR2vBUSJqo7Oip-0jw59O5S_LxCLia4PIL3doG4ZsM045poJq_Qx_-gF3FNdd9sOBGMKakUrxQ7UGOKOSeYzS7V_6e9ocRcmTIHU6aaMr9NmcsaenRdvQ4BppvIHzUV4Acg16dlC-nv7P_U_gJeMqEU</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>3062287883</pqid></control><display><type>article</type><title>TMTV-Net: fully automated total metabolic tumor volume segmentation in lymphoma PET/CT images — a multi-center generalizability analysis</title><source>Springer Nature</source><creator>Yousefirizi, Fereshteh ; Klyuzhin, Ivan S. ; O, Joo Hyun ; Harsini, Sara ; Tie, Xin ; Shiri, Isaac ; Shin, Muheon ; Lee, Changhee ; Cho, Steve Y. ; Bradshaw, Tyler J. ; Zaidi, Habib ; Bénard, François ; Sehn, Laurie H. ; Savage, Kerry J. ; Steidl, Christian ; Uribe, Carlos F. ; Rahmim, Arman</creator><creatorcontrib>Yousefirizi, Fereshteh ; Klyuzhin, Ivan S. ; O, Joo Hyun ; Harsini, Sara ; Tie, Xin ; Shiri, Isaac ; Shin, Muheon ; Lee, Changhee ; Cho, Steve Y. ; Bradshaw, Tyler J. ; Zaidi, Habib ; Bénard, François ; Sehn, Laurie H. ; Savage, Kerry J. ; Steidl, Christian ; Uribe, Carlos F. ; Rahmim, Arman</creatorcontrib><description>Purpose
Total metabolic tumor volume (TMTV) segmentation has significant value enabling quantitative imaging biomarkers for lymphoma management. In this work, we tackle the challenging task of automated tumor delineation in lymphoma from PET/CT scans using a cascaded approach.
Methods
Our study included 1418 2-[
18
F]FDG PET/CT scans from four different centers. The dataset was divided into 900 scans for development/validation/testing phases and 518 for multi-center external testing. The former consisted of 450 lymphoma, lung cancer, and melanoma scans, along with 450 negative scans, while the latter consisted of lymphoma patients from different centers with diffuse large B cell, primary mediastinal large B cell, and classic Hodgkin lymphoma cases. Our approach involves resampling PET/CT images into different voxel sizes in the first step, followed by training multi-resolution 3D U-Nets on each resampled dataset using a fivefold cross-validation scheme. The models trained on different data splits were ensemble. After applying soft voting to the predicted masks, in the second step, we input the probability-averaged predictions, along with the input imaging data, into another 3D U-Net. Models were trained with semi-supervised loss. We additionally considered the effectiveness of using test time augmentation (TTA) to improve the segmentation performance after training. In addition to quantitative analysis including Dice score (DSC) and TMTV comparisons, the qualitative evaluation was also conducted by nuclear medicine physicians.
Results
Our cascaded soft-voting guided approach resulted in performance with an average DSC of 0.68 ± 0.12 for the internal test data from developmental dataset, and an average DSC of 0.66 ± 0.18 on the multi-site external data (
n
= 518), significantly outperforming (
p
< 0.001) state-of-the-art (SOTA) approaches including nnU-Net and SWIN UNETR. While TTA yielded enhanced performance gains for some of the comparator methods, its impact on our cascaded approach was found to be negligible (DSC: 0.66 ± 0.16). Our approach reliably quantified TMTV, with a correlation of 0.89 with the ground truth (
p
< 0.001). Furthermore, in terms of visual assessment, concordance between quantitative evaluations and clinician feedback was observed in the majority of cases. The average relative error (ARE) and the absolute error (AE) in TMTV prediction on external multi-centric dataset were ARE = 0.43 ± 0.54 and AE = 157.32 ± 378.12 (mL) for all the external test data (
n
= 518), and ARE = 0.30 ± 0.22 and AE = 82.05 ± 99.78 (mL) when the 10% outliers (
n
= 53) were excluded.
Conclusion
TMTV-Net demonstrates strong performance and generalizability in TMTV segmentation across multi-site external datasets, encompassing various lymphoma subtypes. A negligible reduction of 2% in overall performance during testing on external data highlights robust model generalizability across different centers and cancer types, likely attributable to its training with resampled inputs. Our model is publicly available, allowing easy multi-site evaluation and generalizability analysis on datasets from different institutions.</description><identifier>ISSN: 1619-7070</identifier><identifier>ISSN: 1619-7089</identifier><identifier>EISSN: 1619-7089</identifier><identifier>DOI: 10.1007/s00259-024-06616-x</identifier><identifier>PMID: 38326655</identifier><language>eng</language><publisher>Berlin/Heidelberg: Springer Berlin Heidelberg</publisher><subject>Automation ; Biomarkers ; Cardiology ; Computed tomography ; Datasets ; Female ; Fluorodeoxyglucose F18 ; Hodgkin's lymphoma ; Humans ; Image processing ; Image Processing, Computer-Assisted - methods ; Image segmentation ; Imaging ; Lung cancer ; Lymphoma ; Lymphoma - diagnostic imaging ; Male ; Medical imaging ; Medicine ; Medicine & Public Health ; Melanoma ; Metabolism ; Nuclear Medicine ; Oncology ; Original Article ; Orthopedics ; Outliers (statistics) ; Performance enhancement ; Positron emission ; Positron Emission Tomography Computed Tomography - methods ; Predictions ; Qualitative analysis ; Quantitative analysis ; Radiology ; Resampling ; Segmentation ; Testing time ; Tumor Burden ; Tumors ; Visual observation</subject><ispartof>European journal of nuclear medicine and molecular imaging, 2024-06, Vol.51 (7), p.1937-1954</ispartof><rights>The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.</rights><rights>2024. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c375t-c5667678ebd0d7eee6ee9f94754bad85c340bf38e876432060153467ed1951c83</citedby><cites>FETCH-LOGICAL-c375t-c5667678ebd0d7eee6ee9f94754bad85c340bf38e876432060153467ed1951c83</cites><orcidid>0000-0003-0141-7628 ; 0000-0001-7995-3581 ; 0000-0003-3127-7478 ; 0000-0001-9549-7002 ; 0000-0001-5261-6163 ; 0000-0002-5835-9863 ; 0000-0001-9842-9750 ; 0000-0003-2724-7778 ; 0000-0001-6196-6982 ; 0000-0001-7559-5297 ; 0000-0002-9980-2403 ; 0000-0003-1860-9765 ; 0000-0002-5735-0736</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>314,777,781,27905,27906</link.rule.ids><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/38326655$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>Yousefirizi, Fereshteh</creatorcontrib><creatorcontrib>Klyuzhin, Ivan S.</creatorcontrib><creatorcontrib>O, Joo Hyun</creatorcontrib><creatorcontrib>Harsini, Sara</creatorcontrib><creatorcontrib>Tie, Xin</creatorcontrib><creatorcontrib>Shiri, Isaac</creatorcontrib><creatorcontrib>Shin, Muheon</creatorcontrib><creatorcontrib>Lee, Changhee</creatorcontrib><creatorcontrib>Cho, Steve Y.</creatorcontrib><creatorcontrib>Bradshaw, Tyler J.</creatorcontrib><creatorcontrib>Zaidi, Habib</creatorcontrib><creatorcontrib>Bénard, François</creatorcontrib><creatorcontrib>Sehn, Laurie H.</creatorcontrib><creatorcontrib>Savage, Kerry J.</creatorcontrib><creatorcontrib>Steidl, Christian</creatorcontrib><creatorcontrib>Uribe, Carlos F.</creatorcontrib><creatorcontrib>Rahmim, Arman</creatorcontrib><title>TMTV-Net: fully automated total metabolic tumor volume segmentation in lymphoma PET/CT images — a multi-center generalizability analysis</title><title>European journal of nuclear medicine and molecular imaging</title><addtitle>Eur J Nucl Med Mol Imaging</addtitle><addtitle>Eur J Nucl Med Mol Imaging</addtitle><description>Purpose
Total metabolic tumor volume (TMTV) segmentation has significant value enabling quantitative imaging biomarkers for lymphoma management. In this work, we tackle the challenging task of automated tumor delineation in lymphoma from PET/CT scans using a cascaded approach.
Methods
Our study included 1418 2-[
18
F]FDG PET/CT scans from four different centers. The dataset was divided into 900 scans for development/validation/testing phases and 518 for multi-center external testing. The former consisted of 450 lymphoma, lung cancer, and melanoma scans, along with 450 negative scans, while the latter consisted of lymphoma patients from different centers with diffuse large B cell, primary mediastinal large B cell, and classic Hodgkin lymphoma cases. Our approach involves resampling PET/CT images into different voxel sizes in the first step, followed by training multi-resolution 3D U-Nets on each resampled dataset using a fivefold cross-validation scheme. The models trained on different data splits were ensemble. After applying soft voting to the predicted masks, in the second step, we input the probability-averaged predictions, along with the input imaging data, into another 3D U-Net. Models were trained with semi-supervised loss. We additionally considered the effectiveness of using test time augmentation (TTA) to improve the segmentation performance after training. In addition to quantitative analysis including Dice score (DSC) and TMTV comparisons, the qualitative evaluation was also conducted by nuclear medicine physicians.
Results
Our cascaded soft-voting guided approach resulted in performance with an average DSC of 0.68 ± 0.12 for the internal test data from developmental dataset, and an average DSC of 0.66 ± 0.18 on the multi-site external data (
n
= 518), significantly outperforming (
p
< 0.001) state-of-the-art (SOTA) approaches including nnU-Net and SWIN UNETR. While TTA yielded enhanced performance gains for some of the comparator methods, its impact on our cascaded approach was found to be negligible (DSC: 0.66 ± 0.16). Our approach reliably quantified TMTV, with a correlation of 0.89 with the ground truth (
p
< 0.001). Furthermore, in terms of visual assessment, concordance between quantitative evaluations and clinician feedback was observed in the majority of cases. The average relative error (ARE) and the absolute error (AE) in TMTV prediction on external multi-centric dataset were ARE = 0.43 ± 0.54 and AE = 157.32 ± 378.12 (mL) for all the external test data (
n
= 518), and ARE = 0.30 ± 0.22 and AE = 82.05 ± 99.78 (mL) when the 10% outliers (
n
= 53) were excluded.
Conclusion
TMTV-Net demonstrates strong performance and generalizability in TMTV segmentation across multi-site external datasets, encompassing various lymphoma subtypes. A negligible reduction of 2% in overall performance during testing on external data highlights robust model generalizability across different centers and cancer types, likely attributable to its training with resampled inputs. Our model is publicly available, allowing easy multi-site evaluation and generalizability analysis on datasets from different institutions.</description><subject>Automation</subject><subject>Biomarkers</subject><subject>Cardiology</subject><subject>Computed tomography</subject><subject>Datasets</subject><subject>Female</subject><subject>Fluorodeoxyglucose F18</subject><subject>Hodgkin's lymphoma</subject><subject>Humans</subject><subject>Image processing</subject><subject>Image Processing, Computer-Assisted - methods</subject><subject>Image segmentation</subject><subject>Imaging</subject><subject>Lung cancer</subject><subject>Lymphoma</subject><subject>Lymphoma - diagnostic imaging</subject><subject>Male</subject><subject>Medical imaging</subject><subject>Medicine</subject><subject>Medicine & Public Health</subject><subject>Melanoma</subject><subject>Metabolism</subject><subject>Nuclear Medicine</subject><subject>Oncology</subject><subject>Original Article</subject><subject>Orthopedics</subject><subject>Outliers (statistics)</subject><subject>Performance enhancement</subject><subject>Positron emission</subject><subject>Positron Emission Tomography Computed Tomography - methods</subject><subject>Predictions</subject><subject>Qualitative analysis</subject><subject>Quantitative analysis</subject><subject>Radiology</subject><subject>Resampling</subject><subject>Segmentation</subject><subject>Testing time</subject><subject>Tumor Burden</subject><subject>Tumors</subject><subject>Visual observation</subject><issn>1619-7070</issn><issn>1619-7089</issn><issn>1619-7089</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><recordid>eNp9kbtuFDEUhkcIRELgBSiQJRoaE19mfKFDq3CRwqUYaC3PzJnFkT1ebA_KUlFT84Q8CQ4bgkRBZUv-_v8c62uah5Q8pYTI00wI6zQmrMVECCrw5a3mmAqqsSRK3765S3LU3Mv5ghCqmNJ3myOuOBOi646b7_2b_iN-C-UZmlfv98iuJQZbYEIlFutRgGKH6N2IyhpiQl-iXwOgDNsAS7HFxQW5Bfl92H2qQfT-rD_d9MgFu4WMfn77gSwKqy8Oj5WHhLawQLLefbWD867UiYv1--zy_ebObH2GB9fnSfPhxVm_eYXP3718vXl-jkcuu4LHTggppIJhIpMEAAGgZ93Krh3spLqRt2SYuQIlRcsZEYR2vBUSJqo7Oip-0jw59O5S_LxCLia4PIL3doG4ZsM045poJq_Qx_-gF3FNdd9sOBGMKakUrxQ7UGOKOSeYzS7V_6e9ocRcmTIHU6aaMr9NmcsaenRdvQ4BppvIHzUV4Acg16dlC-nv7P_U_gJeMqEU</recordid><startdate>20240601</startdate><enddate>20240601</enddate><creator>Yousefirizi, Fereshteh</creator><creator>Klyuzhin, Ivan S.</creator><creator>O, Joo Hyun</creator><creator>Harsini, Sara</creator><creator>Tie, Xin</creator><creator>Shiri, Isaac</creator><creator>Shin, Muheon</creator><creator>Lee, Changhee</creator><creator>Cho, Steve Y.</creator><creator>Bradshaw, Tyler J.</creator><creator>Zaidi, Habib</creator><creator>Bénard, François</creator><creator>Sehn, Laurie H.</creator><creator>Savage, Kerry J.</creator><creator>Steidl, Christian</creator><creator>Uribe, Carlos F.</creator><creator>Rahmim, Arman</creator><general>Springer Berlin Heidelberg</general><general>Springer Nature B.V</general><scope>CGR</scope><scope>CUY</scope><scope>CVF</scope><scope>ECM</scope><scope>EIF</scope><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7TK</scope><scope>K9.</scope><scope>NAPCQ</scope><scope>7X8</scope><orcidid>https://orcid.org/0000-0003-0141-7628</orcidid><orcidid>https://orcid.org/0000-0001-7995-3581</orcidid><orcidid>https://orcid.org/0000-0003-3127-7478</orcidid><orcidid>https://orcid.org/0000-0001-9549-7002</orcidid><orcidid>https://orcid.org/0000-0001-5261-6163</orcidid><orcidid>https://orcid.org/0000-0002-5835-9863</orcidid><orcidid>https://orcid.org/0000-0001-9842-9750</orcidid><orcidid>https://orcid.org/0000-0003-2724-7778</orcidid><orcidid>https://orcid.org/0000-0001-6196-6982</orcidid><orcidid>https://orcid.org/0000-0001-7559-5297</orcidid><orcidid>https://orcid.org/0000-0002-9980-2403</orcidid><orcidid>https://orcid.org/0000-0003-1860-9765</orcidid><orcidid>https://orcid.org/0000-0002-5735-0736</orcidid></search><sort><creationdate>20240601</creationdate><title>TMTV-Net: fully automated total metabolic tumor volume segmentation in lymphoma PET/CT images — a multi-center generalizability analysis</title><author>Yousefirizi, Fereshteh ; Klyuzhin, Ivan S. ; O, Joo Hyun ; Harsini, Sara ; Tie, Xin ; Shiri, Isaac ; Shin, Muheon ; Lee, Changhee ; Cho, Steve Y. ; Bradshaw, Tyler J. ; Zaidi, Habib ; Bénard, François ; Sehn, Laurie H. ; Savage, Kerry J. ; Steidl, Christian ; Uribe, Carlos F. ; Rahmim, Arman</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c375t-c5667678ebd0d7eee6ee9f94754bad85c340bf38e876432060153467ed1951c83</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Automation</topic><topic>Biomarkers</topic><topic>Cardiology</topic><topic>Computed tomography</topic><topic>Datasets</topic><topic>Female</topic><topic>Fluorodeoxyglucose F18</topic><topic>Hodgkin's lymphoma</topic><topic>Humans</topic><topic>Image processing</topic><topic>Image Processing, Computer-Assisted - methods</topic><topic>Image segmentation</topic><topic>Imaging</topic><topic>Lung cancer</topic><topic>Lymphoma</topic><topic>Lymphoma - diagnostic imaging</topic><topic>Male</topic><topic>Medical imaging</topic><topic>Medicine</topic><topic>Medicine & Public Health</topic><topic>Melanoma</topic><topic>Metabolism</topic><topic>Nuclear Medicine</topic><topic>Oncology</topic><topic>Original Article</topic><topic>Orthopedics</topic><topic>Outliers (statistics)</topic><topic>Performance enhancement</topic><topic>Positron emission</topic><topic>Positron Emission Tomography Computed Tomography - methods</topic><topic>Predictions</topic><topic>Qualitative analysis</topic><topic>Quantitative analysis</topic><topic>Radiology</topic><topic>Resampling</topic><topic>Segmentation</topic><topic>Testing time</topic><topic>Tumor Burden</topic><topic>Tumors</topic><topic>Visual observation</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Yousefirizi, Fereshteh</creatorcontrib><creatorcontrib>Klyuzhin, Ivan S.</creatorcontrib><creatorcontrib>O, Joo Hyun</creatorcontrib><creatorcontrib>Harsini, Sara</creatorcontrib><creatorcontrib>Tie, Xin</creatorcontrib><creatorcontrib>Shiri, Isaac</creatorcontrib><creatorcontrib>Shin, Muheon</creatorcontrib><creatorcontrib>Lee, Changhee</creatorcontrib><creatorcontrib>Cho, Steve Y.</creatorcontrib><creatorcontrib>Bradshaw, Tyler J.</creatorcontrib><creatorcontrib>Zaidi, Habib</creatorcontrib><creatorcontrib>Bénard, François</creatorcontrib><creatorcontrib>Sehn, Laurie H.</creatorcontrib><creatorcontrib>Savage, Kerry J.</creatorcontrib><creatorcontrib>Steidl, Christian</creatorcontrib><creatorcontrib>Uribe, Carlos F.</creatorcontrib><creatorcontrib>Rahmim, Arman</creatorcontrib><collection>Medline</collection><collection>MEDLINE</collection><collection>MEDLINE (Ovid)</collection><collection>MEDLINE</collection><collection>MEDLINE</collection><collection>PubMed</collection><collection>CrossRef</collection><collection>Neurosciences Abstracts</collection><collection>ProQuest Health & Medical Complete (Alumni)</collection><collection>Nursing & Allied Health Premium</collection><collection>MEDLINE - Academic</collection><jtitle>European journal of nuclear medicine and molecular imaging</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Yousefirizi, Fereshteh</au><au>Klyuzhin, Ivan S.</au><au>O, Joo Hyun</au><au>Harsini, Sara</au><au>Tie, Xin</au><au>Shiri, Isaac</au><au>Shin, Muheon</au><au>Lee, Changhee</au><au>Cho, Steve Y.</au><au>Bradshaw, Tyler J.</au><au>Zaidi, Habib</au><au>Bénard, François</au><au>Sehn, Laurie H.</au><au>Savage, Kerry J.</au><au>Steidl, Christian</au><au>Uribe, Carlos F.</au><au>Rahmim, Arman</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>TMTV-Net: fully automated total metabolic tumor volume segmentation in lymphoma PET/CT images — a multi-center generalizability analysis</atitle><jtitle>European journal of nuclear medicine and molecular imaging</jtitle><stitle>Eur J Nucl Med Mol Imaging</stitle><addtitle>Eur J Nucl Med Mol Imaging</addtitle><date>2024-06-01</date><risdate>2024</risdate><volume>51</volume><issue>7</issue><spage>1937</spage><epage>1954</epage><pages>1937-1954</pages><issn>1619-7070</issn><issn>1619-7089</issn><eissn>1619-7089</eissn><abstract>Purpose
Total metabolic tumor volume (TMTV) segmentation has significant value enabling quantitative imaging biomarkers for lymphoma management. In this work, we tackle the challenging task of automated tumor delineation in lymphoma from PET/CT scans using a cascaded approach.
Methods
Our study included 1418 2-[
18
F]FDG PET/CT scans from four different centers. The dataset was divided into 900 scans for development/validation/testing phases and 518 for multi-center external testing. The former consisted of 450 lymphoma, lung cancer, and melanoma scans, along with 450 negative scans, while the latter consisted of lymphoma patients from different centers with diffuse large B cell, primary mediastinal large B cell, and classic Hodgkin lymphoma cases. Our approach involves resampling PET/CT images into different voxel sizes in the first step, followed by training multi-resolution 3D U-Nets on each resampled dataset using a fivefold cross-validation scheme. The models trained on different data splits were ensemble. After applying soft voting to the predicted masks, in the second step, we input the probability-averaged predictions, along with the input imaging data, into another 3D U-Net. Models were trained with semi-supervised loss. We additionally considered the effectiveness of using test time augmentation (TTA) to improve the segmentation performance after training. In addition to quantitative analysis including Dice score (DSC) and TMTV comparisons, the qualitative evaluation was also conducted by nuclear medicine physicians.
Results
Our cascaded soft-voting guided approach resulted in performance with an average DSC of 0.68 ± 0.12 for the internal test data from developmental dataset, and an average DSC of 0.66 ± 0.18 on the multi-site external data (
n
= 518), significantly outperforming (
p
< 0.001) state-of-the-art (SOTA) approaches including nnU-Net and SWIN UNETR. While TTA yielded enhanced performance gains for some of the comparator methods, its impact on our cascaded approach was found to be negligible (DSC: 0.66 ± 0.16). Our approach reliably quantified TMTV, with a correlation of 0.89 with the ground truth (
p
< 0.001). Furthermore, in terms of visual assessment, concordance between quantitative evaluations and clinician feedback was observed in the majority of cases. The average relative error (ARE) and the absolute error (AE) in TMTV prediction on external multi-centric dataset were ARE = 0.43 ± 0.54 and AE = 157.32 ± 378.12 (mL) for all the external test data (
n
= 518), and ARE = 0.30 ± 0.22 and AE = 82.05 ± 99.78 (mL) when the 10% outliers (
n
= 53) were excluded.
Conclusion
TMTV-Net demonstrates strong performance and generalizability in TMTV segmentation across multi-site external datasets, encompassing various lymphoma subtypes. A negligible reduction of 2% in overall performance during testing on external data highlights robust model generalizability across different centers and cancer types, likely attributable to its training with resampled inputs. Our model is publicly available, allowing easy multi-site evaluation and generalizability analysis on datasets from different institutions.</abstract><cop>Berlin/Heidelberg</cop><pub>Springer Berlin Heidelberg</pub><pmid>38326655</pmid><doi>10.1007/s00259-024-06616-x</doi><tpages>18</tpages><orcidid>https://orcid.org/0000-0003-0141-7628</orcidid><orcidid>https://orcid.org/0000-0001-7995-3581</orcidid><orcidid>https://orcid.org/0000-0003-3127-7478</orcidid><orcidid>https://orcid.org/0000-0001-9549-7002</orcidid><orcidid>https://orcid.org/0000-0001-5261-6163</orcidid><orcidid>https://orcid.org/0000-0002-5835-9863</orcidid><orcidid>https://orcid.org/0000-0001-9842-9750</orcidid><orcidid>https://orcid.org/0000-0003-2724-7778</orcidid><orcidid>https://orcid.org/0000-0001-6196-6982</orcidid><orcidid>https://orcid.org/0000-0001-7559-5297</orcidid><orcidid>https://orcid.org/0000-0002-9980-2403</orcidid><orcidid>https://orcid.org/0000-0003-1860-9765</orcidid><orcidid>https://orcid.org/0000-0002-5735-0736</orcidid></addata></record> |
| fulltext | fulltext |
| identifier | ISSN: 1619-7070 |
| ispartof | European journal of nuclear medicine and molecular imaging, 2024-06, Vol.51 (7), p.1937-1954 |
| issn | 1619-7070 1619-7089 1619-7089 |
| language | eng |
| recordid | cdi_proquest_miscellaneous_2923909278 |
| source | Springer Nature |
| subjects | Automation Biomarkers Cardiology Computed tomography Datasets Female Fluorodeoxyglucose F18 Hodgkin's lymphoma Humans Image processing Image Processing, Computer-Assisted - methods Image segmentation Imaging Lung cancer Lymphoma Lymphoma - diagnostic imaging Male Medical imaging Medicine Medicine & Public Health Melanoma Metabolism Nuclear Medicine Oncology Original Article Orthopedics Outliers (statistics) Performance enhancement Positron emission Positron Emission Tomography Computed Tomography - methods Predictions Qualitative analysis Quantitative analysis Radiology Resampling Segmentation Testing time Tumor Burden Tumors Visual observation |
| title | TMTV-Net: fully automated total metabolic tumor volume segmentation in lymphoma PET/CT images — a multi-center generalizability analysis |
| url | http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-19T14%3A56%3A47IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=TMTV-Net:%20fully%20automated%20total%20metabolic%20tumor%20volume%20segmentation%20in%20lymphoma%20PET/CT%20images%20%E2%80%94%20a%20multi-center%20generalizability%20analysis&rft.jtitle=European%20journal%20of%20nuclear%20medicine%20and%20molecular%20imaging&rft.au=Yousefirizi,%20Fereshteh&rft.date=2024-06-01&rft.volume=51&rft.issue=7&rft.spage=1937&rft.epage=1954&rft.pages=1937-1954&rft.issn=1619-7070&rft.eissn=1619-7089&rft_id=info:doi/10.1007/s00259-024-06616-x&rft_dat=%3Cproquest_cross%3E2923909278%3C/proquest_cross%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c375t-c5667678ebd0d7eee6ee9f94754bad85c340bf38e876432060153467ed1951c83%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=3062287883&rft_id=info:pmid/38326655&rfr_iscdi=true |