Loading…

Multiple Kernel Synthesis of Head CT Using a Task-Based Loss Function

In CT imaging of the head, multiple image series are routinely reconstructed with different kernels and slice thicknesses. Reviewing the redundant information is an inefficient process for radiologists. We address this issue with a convolutional neural network (CNN)-based technique, synthesiZed Impr...

Full description

Saved in:
Bibliographic Details
Published in:Journal of digital imaging 2024-04, Vol.37 (2), p.864-872
Main Authors: Nelson, Brandon J., Gomez-Cardona, Daniel G., Thorne, Jamison E., Huber, Nathan R., Yu, Lifeng, Leng, Shuai, McCollough, Cynthia H., Missert, Andrew D.
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites cdi_FETCH-LOGICAL-c356t-8f3a4df6c4047e541e8678c943a64267bda53abc8fd3dd2b690e54f639ea0d353
container_end_page 872
container_issue 2
container_start_page 864
container_title Journal of digital imaging
container_volume 37
creator Nelson, Brandon J.
Gomez-Cardona, Daniel G.
Thorne, Jamison E.
Huber, Nathan R.
Yu, Lifeng
Leng, Shuai
McCollough, Cynthia H.
Missert, Andrew D.
description In CT imaging of the head, multiple image series are routinely reconstructed with different kernels and slice thicknesses. Reviewing the redundant information is an inefficient process for radiologists. We address this issue with a convolutional neural network (CNN)-based technique, synthesiZed Improved Resolution and Concurrent nOise reductioN (ZIRCON), that creates a single, thin, low-noise series that combines the favorable features from smooth and sharp head kernels. ZIRCON uses a CNN model with an autoencoder U-Net architecture that accepts two input channels (smooth- and sharp-kernel CT images) and combines their salient features to produce a single CT image. Image quality requirements are built into a task-based loss function with a smooth and sharp loss terms specific to anatomical regions. The model is trained using supervised learning with paired routine-dose clinical non-contrast head CT images as training targets and simulated low-dose (25%) images as training inputs. One hundred unique de-identified clinical exams were used for training, ten for validation, and ten for testing. Visual comparisons and contrast measurements of ZIRCON revealed that thinner slices and the smooth-kernel loss function improved gray-white matter contrast. Combined with lower noise, this increased visibility of small soft-tissue features that would be otherwise impaired by partial volume averaging or noise. Line profile analysis showed that ZIRCON images largely retained sharpness compared to the sharp-kernel input images. ZIRCON combined desirable image quality properties of both smooth and sharp input kernels into a single, thin, low-noise series suitable for both brain and skull imaging.
doi_str_mv 10.1007/s10278-023-00959-x
format article
fullrecord <record><control><sourceid>proquest_pubme</sourceid><recordid>TN_cdi_pubmedcentral_primary_oai_pubmedcentral_nih_gov_11589061</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2926076205</sourcerecordid><originalsourceid>FETCH-LOGICAL-c356t-8f3a4df6c4047e541e8678c943a64267bda53abc8fd3dd2b690e54f639ea0d353</originalsourceid><addsrcrecordid>eNp9kcFu1DAQhi0EolXpC3BAlrhwCYw9jmOfEKxaWrGIA9uz5Y0n25Sss9gJat8et1tK4cDJluabfzz-GHsp4K0AaN5lAbIxFUisAGxtq-sn7FBaZSppEZ8-uh-w45yvAABRIGp4zg7QoEJZy0N28mUepn43EP9MKdLAv93E6ZJyn_nY8TPygS9W_CL3ccM9X_n8vfroMwW-HHPmp3Nsp36ML9izzg-Zju_PI3ZxerJanFXLr5_OFx-WVYu1nirToVeh060C1VCtBBndmNYq9FpJ3ayDr9GvW9MFDEGutYVCdRoteQhY4xF7v8_dzesthZbilPzgdqnf-nTjRt-7vyuxv3Sb8acTojYWtCgJb-4T0vhjpjy5bZ9bGgYfaZyzk1ZqaLSE22Gv_0GvxjnFsp9DUEIbaYwplNxTbSo_kqh7eI0Ad2vK7U25YsrdmXLXpenV4z0eWn57KQDugVxKcUPpz-z_xP4Curmd4Q</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>3041682888</pqid></control><display><type>article</type><title>Multiple Kernel Synthesis of Head CT Using a Task-Based Loss Function</title><source>PubMed (Medline)</source><source>Springer Nature</source><creator>Nelson, Brandon J. ; Gomez-Cardona, Daniel G. ; Thorne, Jamison E. ; Huber, Nathan R. ; Yu, Lifeng ; Leng, Shuai ; McCollough, Cynthia H. ; Missert, Andrew D.</creator><creatorcontrib>Nelson, Brandon J. ; Gomez-Cardona, Daniel G. ; Thorne, Jamison E. ; Huber, Nathan R. ; Yu, Lifeng ; Leng, Shuai ; McCollough, Cynthia H. ; Missert, Andrew D.</creatorcontrib><description>In CT imaging of the head, multiple image series are routinely reconstructed with different kernels and slice thicknesses. Reviewing the redundant information is an inefficient process for radiologists. We address this issue with a convolutional neural network (CNN)-based technique, synthesiZed Improved Resolution and Concurrent nOise reductioN (ZIRCON), that creates a single, thin, low-noise series that combines the favorable features from smooth and sharp head kernels. ZIRCON uses a CNN model with an autoencoder U-Net architecture that accepts two input channels (smooth- and sharp-kernel CT images) and combines their salient features to produce a single CT image. Image quality requirements are built into a task-based loss function with a smooth and sharp loss terms specific to anatomical regions. The model is trained using supervised learning with paired routine-dose clinical non-contrast head CT images as training targets and simulated low-dose (25%) images as training inputs. One hundred unique de-identified clinical exams were used for training, ten for validation, and ten for testing. Visual comparisons and contrast measurements of ZIRCON revealed that thinner slices and the smooth-kernel loss function improved gray-white matter contrast. Combined with lower noise, this increased visibility of small soft-tissue features that would be otherwise impaired by partial volume averaging or noise. Line profile analysis showed that ZIRCON images largely retained sharpness compared to the sharp-kernel input images. ZIRCON combined desirable image quality properties of both smooth and sharp input kernels into a single, thin, low-noise series suitable for both brain and skull imaging.</description><identifier>ISSN: 2948-2933</identifier><identifier>ISSN: 0897-1889</identifier><identifier>ISSN: 2948-2925</identifier><identifier>EISSN: 2948-2933</identifier><identifier>EISSN: 1618-727X</identifier><identifier>DOI: 10.1007/s10278-023-00959-x</identifier><identifier>PMID: 38343252</identifier><language>eng</language><publisher>Cham: Springer International Publishing</publisher><subject>Artificial neural networks ; Computed tomography ; Drug dosages ; Fractures ; Head ; Head injuries ; Image contrast ; Image quality ; Imaging ; Informatics ; Information overload ; Machine learning ; Medical imaging ; Medicine ; Medicine &amp; Public Health ; Neural networks ; Neuroimaging ; Noise reduction ; Patients ; Quality management ; Radiology ; Review boards ; Substantia alba ; Supervised learning ; Training ; Trauma ; Zircon</subject><ispartof>Journal of digital imaging, 2024-04, Vol.37 (2), p.864-872</ispartof><rights>The Author(s) 2024</rights><rights>2024. The Author(s).</rights><rights>The Author(s) 2024. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><rights>The Author(s) 2024 2024</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c356t-8f3a4df6c4047e541e8678c943a64267bda53abc8fd3dd2b690e54f639ea0d353</cites><orcidid>0000-0002-2825-696X</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://www.ncbi.nlm.nih.gov/pmc/articles/PMC11589061/pdf/$$EPDF$$P50$$Gpubmedcentral$$Hfree_for_read</linktopdf><linktohtml>$$Uhttps://www.ncbi.nlm.nih.gov/pmc/articles/PMC11589061/$$EHTML$$P50$$Gpubmedcentral$$Hfree_for_read</linktohtml><link.rule.ids>230,314,727,780,784,885,27924,27925,53791,53793</link.rule.ids><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/38343252$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>Nelson, Brandon J.</creatorcontrib><creatorcontrib>Gomez-Cardona, Daniel G.</creatorcontrib><creatorcontrib>Thorne, Jamison E.</creatorcontrib><creatorcontrib>Huber, Nathan R.</creatorcontrib><creatorcontrib>Yu, Lifeng</creatorcontrib><creatorcontrib>Leng, Shuai</creatorcontrib><creatorcontrib>McCollough, Cynthia H.</creatorcontrib><creatorcontrib>Missert, Andrew D.</creatorcontrib><title>Multiple Kernel Synthesis of Head CT Using a Task-Based Loss Function</title><title>Journal of digital imaging</title><addtitle>j Imaging. Inform. med</addtitle><addtitle>J Imaging Inform Med</addtitle><description>In CT imaging of the head, multiple image series are routinely reconstructed with different kernels and slice thicknesses. Reviewing the redundant information is an inefficient process for radiologists. We address this issue with a convolutional neural network (CNN)-based technique, synthesiZed Improved Resolution and Concurrent nOise reductioN (ZIRCON), that creates a single, thin, low-noise series that combines the favorable features from smooth and sharp head kernels. ZIRCON uses a CNN model with an autoencoder U-Net architecture that accepts two input channels (smooth- and sharp-kernel CT images) and combines their salient features to produce a single CT image. Image quality requirements are built into a task-based loss function with a smooth and sharp loss terms specific to anatomical regions. The model is trained using supervised learning with paired routine-dose clinical non-contrast head CT images as training targets and simulated low-dose (25%) images as training inputs. One hundred unique de-identified clinical exams were used for training, ten for validation, and ten for testing. Visual comparisons and contrast measurements of ZIRCON revealed that thinner slices and the smooth-kernel loss function improved gray-white matter contrast. Combined with lower noise, this increased visibility of small soft-tissue features that would be otherwise impaired by partial volume averaging or noise. Line profile analysis showed that ZIRCON images largely retained sharpness compared to the sharp-kernel input images. ZIRCON combined desirable image quality properties of both smooth and sharp input kernels into a single, thin, low-noise series suitable for both brain and skull imaging.</description><subject>Artificial neural networks</subject><subject>Computed tomography</subject><subject>Drug dosages</subject><subject>Fractures</subject><subject>Head</subject><subject>Head injuries</subject><subject>Image contrast</subject><subject>Image quality</subject><subject>Imaging</subject><subject>Informatics</subject><subject>Information overload</subject><subject>Machine learning</subject><subject>Medical imaging</subject><subject>Medicine</subject><subject>Medicine &amp; Public Health</subject><subject>Neural networks</subject><subject>Neuroimaging</subject><subject>Noise reduction</subject><subject>Patients</subject><subject>Quality management</subject><subject>Radiology</subject><subject>Review boards</subject><subject>Substantia alba</subject><subject>Supervised learning</subject><subject>Training</subject><subject>Trauma</subject><subject>Zircon</subject><issn>2948-2933</issn><issn>0897-1889</issn><issn>2948-2925</issn><issn>2948-2933</issn><issn>1618-727X</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><recordid>eNp9kcFu1DAQhi0EolXpC3BAlrhwCYw9jmOfEKxaWrGIA9uz5Y0n25Sss9gJat8et1tK4cDJluabfzz-GHsp4K0AaN5lAbIxFUisAGxtq-sn7FBaZSppEZ8-uh-w45yvAABRIGp4zg7QoEJZy0N28mUepn43EP9MKdLAv93E6ZJyn_nY8TPygS9W_CL3ccM9X_n8vfroMwW-HHPmp3Nsp36ML9izzg-Zju_PI3ZxerJanFXLr5_OFx-WVYu1nirToVeh060C1VCtBBndmNYq9FpJ3ayDr9GvW9MFDEGutYVCdRoteQhY4xF7v8_dzesthZbilPzgdqnf-nTjRt-7vyuxv3Sb8acTojYWtCgJb-4T0vhjpjy5bZ9bGgYfaZyzk1ZqaLSE22Gv_0GvxjnFsp9DUEIbaYwplNxTbSo_kqh7eI0Ad2vK7U25YsrdmXLXpenV4z0eWn57KQDugVxKcUPpz-z_xP4Curmd4Q</recordid><startdate>20240401</startdate><enddate>20240401</enddate><creator>Nelson, Brandon J.</creator><creator>Gomez-Cardona, Daniel G.</creator><creator>Thorne, Jamison E.</creator><creator>Huber, Nathan R.</creator><creator>Yu, Lifeng</creator><creator>Leng, Shuai</creator><creator>McCollough, Cynthia H.</creator><creator>Missert, Andrew D.</creator><general>Springer International Publishing</general><general>Springer Nature B.V</general><scope>C6C</scope><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7QO</scope><scope>7SC</scope><scope>7TK</scope><scope>8FD</scope><scope>FR3</scope><scope>JQ2</scope><scope>K9.</scope><scope>L7M</scope><scope>L~C</scope><scope>L~D</scope><scope>NAPCQ</scope><scope>P64</scope><scope>7X8</scope><scope>5PM</scope><orcidid>https://orcid.org/0000-0002-2825-696X</orcidid></search><sort><creationdate>20240401</creationdate><title>Multiple Kernel Synthesis of Head CT Using a Task-Based Loss Function</title><author>Nelson, Brandon J. ; Gomez-Cardona, Daniel G. ; Thorne, Jamison E. ; Huber, Nathan R. ; Yu, Lifeng ; Leng, Shuai ; McCollough, Cynthia H. ; Missert, Andrew D.</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c356t-8f3a4df6c4047e541e8678c943a64267bda53abc8fd3dd2b690e54f639ea0d353</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Artificial neural networks</topic><topic>Computed tomography</topic><topic>Drug dosages</topic><topic>Fractures</topic><topic>Head</topic><topic>Head injuries</topic><topic>Image contrast</topic><topic>Image quality</topic><topic>Imaging</topic><topic>Informatics</topic><topic>Information overload</topic><topic>Machine learning</topic><topic>Medical imaging</topic><topic>Medicine</topic><topic>Medicine &amp; Public Health</topic><topic>Neural networks</topic><topic>Neuroimaging</topic><topic>Noise reduction</topic><topic>Patients</topic><topic>Quality management</topic><topic>Radiology</topic><topic>Review boards</topic><topic>Substantia alba</topic><topic>Supervised learning</topic><topic>Training</topic><topic>Trauma</topic><topic>Zircon</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Nelson, Brandon J.</creatorcontrib><creatorcontrib>Gomez-Cardona, Daniel G.</creatorcontrib><creatorcontrib>Thorne, Jamison E.</creatorcontrib><creatorcontrib>Huber, Nathan R.</creatorcontrib><creatorcontrib>Yu, Lifeng</creatorcontrib><creatorcontrib>Leng, Shuai</creatorcontrib><creatorcontrib>McCollough, Cynthia H.</creatorcontrib><creatorcontrib>Missert, Andrew D.</creatorcontrib><collection>SpringerOpen</collection><collection>PubMed</collection><collection>CrossRef</collection><collection>Biotechnology Research Abstracts</collection><collection>Computer and Information Systems Abstracts</collection><collection>Neurosciences Abstracts</collection><collection>Technology Research Database</collection><collection>Engineering Research Database</collection><collection>ProQuest Computer Science Collection</collection><collection>ProQuest Health &amp; Medical Complete (Alumni)</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>Computer and Information Systems Abstracts – Academic</collection><collection>Computer and Information Systems Abstracts Professional</collection><collection>Nursing &amp; Allied Health Premium</collection><collection>Biotechnology and BioEngineering Abstracts</collection><collection>MEDLINE - Academic</collection><collection>PubMed Central (Full Participant titles)</collection><jtitle>Journal of digital imaging</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Nelson, Brandon J.</au><au>Gomez-Cardona, Daniel G.</au><au>Thorne, Jamison E.</au><au>Huber, Nathan R.</au><au>Yu, Lifeng</au><au>Leng, Shuai</au><au>McCollough, Cynthia H.</au><au>Missert, Andrew D.</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Multiple Kernel Synthesis of Head CT Using a Task-Based Loss Function</atitle><jtitle>Journal of digital imaging</jtitle><stitle>j Imaging. Inform. med</stitle><addtitle>J Imaging Inform Med</addtitle><date>2024-04-01</date><risdate>2024</risdate><volume>37</volume><issue>2</issue><spage>864</spage><epage>872</epage><pages>864-872</pages><issn>2948-2933</issn><issn>0897-1889</issn><issn>2948-2925</issn><eissn>2948-2933</eissn><eissn>1618-727X</eissn><abstract>In CT imaging of the head, multiple image series are routinely reconstructed with different kernels and slice thicknesses. Reviewing the redundant information is an inefficient process for radiologists. We address this issue with a convolutional neural network (CNN)-based technique, synthesiZed Improved Resolution and Concurrent nOise reductioN (ZIRCON), that creates a single, thin, low-noise series that combines the favorable features from smooth and sharp head kernels. ZIRCON uses a CNN model with an autoencoder U-Net architecture that accepts two input channels (smooth- and sharp-kernel CT images) and combines their salient features to produce a single CT image. Image quality requirements are built into a task-based loss function with a smooth and sharp loss terms specific to anatomical regions. The model is trained using supervised learning with paired routine-dose clinical non-contrast head CT images as training targets and simulated low-dose (25%) images as training inputs. One hundred unique de-identified clinical exams were used for training, ten for validation, and ten for testing. Visual comparisons and contrast measurements of ZIRCON revealed that thinner slices and the smooth-kernel loss function improved gray-white matter contrast. Combined with lower noise, this increased visibility of small soft-tissue features that would be otherwise impaired by partial volume averaging or noise. Line profile analysis showed that ZIRCON images largely retained sharpness compared to the sharp-kernel input images. ZIRCON combined desirable image quality properties of both smooth and sharp input kernels into a single, thin, low-noise series suitable for both brain and skull imaging.</abstract><cop>Cham</cop><pub>Springer International Publishing</pub><pmid>38343252</pmid><doi>10.1007/s10278-023-00959-x</doi><tpages>9</tpages><orcidid>https://orcid.org/0000-0002-2825-696X</orcidid><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 2948-2933
ispartof Journal of digital imaging, 2024-04, Vol.37 (2), p.864-872
issn 2948-2933
0897-1889
2948-2925
2948-2933
1618-727X
language eng
recordid cdi_pubmedcentral_primary_oai_pubmedcentral_nih_gov_11589061
source PubMed (Medline); Springer Nature
subjects Artificial neural networks
Computed tomography
Drug dosages
Fractures
Head
Head injuries
Image contrast
Image quality
Imaging
Informatics
Information overload
Machine learning
Medical imaging
Medicine
Medicine & Public Health
Neural networks
Neuroimaging
Noise reduction
Patients
Quality management
Radiology
Review boards
Substantia alba
Supervised learning
Training
Trauma
Zircon
title Multiple Kernel Synthesis of Head CT Using a Task-Based Loss Function
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-04T23%3A54%3A38IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_pubme&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Multiple%20Kernel%20Synthesis%20of%20Head%20CT%20Using%20a%20Task-Based%20Loss%20Function&rft.jtitle=Journal%20of%20digital%20imaging&rft.au=Nelson,%20Brandon%20J.&rft.date=2024-04-01&rft.volume=37&rft.issue=2&rft.spage=864&rft.epage=872&rft.pages=864-872&rft.issn=2948-2933&rft.eissn=2948-2933&rft_id=info:doi/10.1007/s10278-023-00959-x&rft_dat=%3Cproquest_pubme%3E2926076205%3C/proquest_pubme%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c356t-8f3a4df6c4047e541e8678c943a64267bda53abc8fd3dd2b690e54f639ea0d353%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=3041682888&rft_id=info:pmid/38343252&rfr_iscdi=true