{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,17]],"date-time":"2026-01-17T22:15:10Z","timestamp":1768688110813,"version":"3.49.0"},"reference-count":24,"publisher":"Wiley","issue":"10","license":[{"start":{"date-parts":[[2017,8,31]],"date-time":"2017-08-31T00:00:00Z","timestamp":1504137600000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/onlinelibrary.wiley.com\/termsAndConditions#vor"}],"funder":[{"DOI":"10.13039\/501100000272","name":"National Institute of Health Research","doi-asserted-by":"crossref","id":[{"id":"10.13039\/501100000272","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Medical Physics"],"published-print":{"date-parts":[[2017,10]]},"abstract":"Purpose<\/jats:title>As part of a program to implement automatic lesion detection methods for whole body magnetic resonance imaging (MRI<\/jats:styled-content>) in oncology, we have developed, evaluated, and compared three algorithms for fully automatic, multiorgan segmentation in healthy volunteers.<\/jats:p><\/jats:sec>Methods<\/jats:title>The first algorithm is based on classification forests (CF<\/jats:styled-content>s), the second is based on 3D convolutional neural networks (CNN<\/jats:styled-content>s) and the third algorithm is based on a multi\u2010atlas (MA<\/jats:styled-content>) approach. We examined data from 51 healthy volunteers, scanned prospectively with a standardized, multiparametric whole body MRI<\/jats:styled-content> protocol at 1.5 T. The study was approved by the local ethics committee and written consent was obtained from the participants. MRI<\/jats:styled-content> data were used as input data to the algorithms, while training was based on manual annotation of the anatomies of interest by clinical MRI<\/jats:styled-content> experts. Fivefold cross\u2010validation experiments were run on 34 artifact\u2010free subjects. We report three overlap and three surface distance metrics to evaluate the agreement between the automatic and manual segmentations, namely the dice similarity coefficient (DSC<\/jats:styled-content>), recall (RE<\/jats:styled-content>), precision (PR<\/jats:styled-content>), average surface distance (ASD<\/jats:styled-content>), root\u2010mean\u2010square surface distance (RMSSD<\/jats:styled-content>), and Hausdorff distance (HD<\/jats:styled-content>). Analysis of variances was used to compare pooled label metrics between the three algorithms and the DSC<\/jats:styled-content> on a \u2018per\u2010organ\u2019 basis. A Mann\u2013Whitney U<\/jats:italic> test was used to compare the pooled metrics between CF<\/jats:styled-content>s and CNN<\/jats:styled-content>s and the DSC<\/jats:styled-content> on a \u2018per\u2010organ\u2019 basis, when using different imaging combinations as input for training.<\/jats:p><\/jats:sec>Results<\/jats:title>All three algorithms resulted in robust segmenters that were effectively trained using a relatively small number of datasets, an important consideration in the clinical setting. Mean overlap metrics for all the segmented structures were: CF<\/jats:styled-content>s: DSC<\/jats:styled-content> = 0.70 \u00b1 0.18, RE<\/jats:styled-content> = 0.73 \u00b1 0.18, PR<\/jats:styled-content> = 0.71 \u00b1 0.14, CNN<\/jats:styled-content>s: DSC<\/jats:styled-content> = 0.81 \u00b1 0.13, RE<\/jats:styled-content> = 0.83 \u00b1 0.14, PR<\/jats:styled-content> = 0.82 \u00b1 0.10, MA<\/jats:styled-content>: DSC<\/jats:styled-content> = 0.71 \u00b1 0.22, RE<\/jats:styled-content> = 0.70 \u00b1 0.34, PR<\/jats:styled-content> = 0.77 \u00b1 0.15. Mean surface distance metrics for all the segmented structures were: CF<\/jats:styled-content>s: ASD<\/jats:styled-content> = 13.5 \u00b1 11.3 mm, RMSSD<\/jats:styled-content> = 34.6 \u00b1 37.6 mm and HD<\/jats:styled-content> = 185.7 \u00b1 194.0 mm, CNN<\/jats:styled-content>s; ASD<\/jats:styled-content> = 5.48 \u00b1 4.84 mm, RMSSD<\/jats:styled-content> = 17.0 \u00b1 13.3 mm and HD<\/jats:styled-content> = 199.0 \u00b1 101.2 mm, MA<\/jats:styled-content>: ASD<\/jats:styled-content> = 4.22 \u00b1 2.42 mm, RMSSD<\/jats:styled-content> = 6.13 \u00b1 2.55 mm, and HD<\/jats:styled-content> = 38.9 \u00b1 28.9 mm. The pooled performance of CF<\/jats:styled-content>s improved when all imaging combinations (T2w + T1w + DWI<\/jats:styled-content>) were used as input, while the performance of CNN<\/jats:styled-content>s deteriorated, but in neither case, significantly. CNN<\/jats:styled-content>s with T2w images as input, performed significantly better than CF<\/jats:styled-content>s with all imaging combinations as input for all anatomical labels, except for the bladder.<\/jats:p><\/jats:sec>Conclusions<\/jats:title>Three state\u2010of\u2010the\u2010art algorithms were developed and used to automatically segment major organs and bones in whole body MRI<\/jats:styled-content>; good agreement to manual segmentations performed by clinical MRI<\/jats:styled-content> experts was observed. CNN<\/jats:styled-content>s perform favorably, when using T2w volumes as input. Using multimodal MRI<\/jats:styled-content> data as input to CNN<\/jats:styled-content>s did not improve the segmentation performance.<\/jats:p><\/jats:sec>","DOI":"10.1002\/mp.12492","type":"journal-article","created":{"date-parts":[[2017,7,30]],"date-time":"2017-07-30T11:16:37Z","timestamp":1501413397000},"page":"5210-5220","source":"Crossref","is-referenced-by-count":36,"title":["Fully automatic, multiorgan segmentation in normal whole body magnetic resonance imaging (MRI<\/scp>), using classification forests (CF<\/scp>s), convolutional neural networks (CNN<\/scp>s), and a multi\u2010atlas (MA<\/scp>) approach"],"prefix":"10.1002","volume":"44","author":[{"given":"Ioannis","family":"Lavdas","sequence":"first","affiliation":[{"name":"Imperial College Comprehensive Cancer Imaging Centre (C.C.I.C.) Hammersmith Campus, Commonwealth Building Main Office, Ground Floor, Du Cane Road London W12 0NN UK"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ben","family":"Glocker","sequence":"additional","affiliation":[{"name":"Biomedical Image Analysis Group Department of Computing Imperial College London Huxley Building, 180 Queen's Gate London SW7 2AZ UK"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Konstantinos","family":"Kamnitsas","sequence":"additional","affiliation":[{"name":"Biomedical Image Analysis Group Department of Computing Imperial College London Huxley Building, 180 Queen's Gate London SW7 2AZ UK"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Daniel","family":"Rueckert","sequence":"additional","affiliation":[{"name":"Biomedical Image Analysis Group Department of Computing Imperial College London Huxley Building, 180 Queen's Gate London SW7 2AZ UK"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Henrietta","family":"Mair","sequence":"additional","affiliation":[{"name":"Department of Imaging University College London Hospitals NHS Foundation Trust Euston Road London NW1 2BU UK"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Amandeep","family":"Sandhu","sequence":"additional","affiliation":[{"name":"Department of Radiology Hammersmith Hospital Imperial College Healthcare NHS Trust DuCane Road London W12 0NN UK"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Stuart A.","family":"Taylor","sequence":"additional","affiliation":[{"name":"Department of Imaging University College London Hospitals NHS Foundation Trust Euston Road London NW1 2BU UK"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Eric O.","family":"Aboagye","sequence":"additional","affiliation":[{"name":"Imperial College Comprehensive Cancer Imaging Centre (C.C.I.C.) Hammersmith Campus, Commonwealth Building Main Office, Ground Floor, Du Cane Road London W12 0NN UK"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Andrea G.","family":"Rockall","sequence":"additional","affiliation":[{"name":"Department of Radiology The Royal Marsden NHS Foundation Trust Fulham Road London SW3 6JJ UK"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"311","published-online":{"date-parts":[[2017,8,31]]},"reference":[{"key":"e_1_2_8_2_1","first-page":"275","article-title":"Diffusion weighted whole body imaging with background body signal suppression (DWIBS): technical improvement using free breathing, STIR and high resolution 3D display","volume":"22","author":"Takahara T","year":"2004","journal-title":"Radiat Med"},{"key":"e_1_2_8_3_1","doi-asserted-by":"publisher","DOI":"10.2214\/AJR.06.1403"},{"key":"e_1_2_8_4_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ejrad.2009.03.045"},{"key":"e_1_2_8_5_1","doi-asserted-by":"publisher","DOI":"10.1002\/jmri.22608"},{"key":"e_1_2_8_6_1","doi-asserted-by":"publisher","DOI":"10.1148\/radiol.11110474"},{"key":"e_1_2_8_7_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-540-39899-8_78"},{"key":"e_1_2_8_8_1","doi-asserted-by":"publisher","DOI":"10.1002\/jmri.24726"},{"key":"e_1_2_8_9_1","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0091779"},{"key":"e_1_2_8_10_1","first-page":"94","volume-title":"Medical Recognition, Segmentation and Parsing","author":"Glocker B","year":"2015"},{"key":"e_1_2_8_11_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-33765-9_62"},{"key":"e_1_2_8_12_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-33454-2_9"},{"key":"e_1_2_8_13_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4471-4929-3_17"},{"key":"e_1_2_8_14_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-23626-6_30"},{"key":"e_1_2_8_15_1","doi-asserted-by":"publisher","DOI":"10.1109\/TMI.2013.2256922"},{"key":"e_1_2_8_16_1","doi-asserted-by":"publisher","DOI":"10.1109\/TMI.2013.2265805"},{"key":"e_1_2_8_17_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-540-85988-8_60"},{"key":"e_1_2_8_18_1","doi-asserted-by":"publisher","DOI":"10.2214\/AJR.14.13964"},{"key":"e_1_2_8_19_1","doi-asserted-by":"publisher","DOI":"10.1023\/A:1010933404324"},{"key":"e_1_2_8_20_1","doi-asserted-by":"publisher","DOI":"10.1561\/0600000035"},{"key":"e_1_2_8_21_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2016.10.004"},{"key":"e_1_2_8_22_1","unstructured":"KamnitsasK.DeepMedic source code;2016.https:\/\/github.com\/Kamnitsask\/deepmedic. Accessed November 2016."},{"key":"e_1_2_8_23_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2008.03.006"},{"key":"e_1_2_8_24_1","doi-asserted-by":"publisher","DOI":"10.1109\/TMI.2009.2013851"},{"key":"e_1_2_8_25_1","doi-asserted-by":"publisher","DOI":"10.1016\/S0031-3203(98)00091-0"}],"container-title":["Medical Physics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/api.wiley.com\/onlinelibrary\/tdm\/v1\/articles\/10.1002%2Fmp.12492","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/mp.12492","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,9,24]],"date-time":"2023-09-24T15:57:45Z","timestamp":1695571065000},"score":1,"resource":{"primary":{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/10.1002\/mp.12492"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2017,8,31]]},"references-count":24,"journal-issue":{"issue":"10","published-print":{"date-parts":[[2017,10]]}},"alternative-id":["10.1002\/mp.12492"],"URL":"https:\/\/doi.org\/10.1002\/mp.12492","archive":["Portico"],"relation":{},"ISSN":["0094-2405","2473-4209"],"issn-type":[{"value":"0094-2405","type":"print"},{"value":"2473-4209","type":"electronic"}],"subject":[],"published":{"date-parts":[[2017,8,31]]}}}