Hubert Beaumont1, Antoine Iannessi2, Anne-Sophie Bertrand3, Jean Michel Cucchi3, Olivier Lucidarme4. 1. Median Technologies, 06560, Valbonne, France. Hubert.beaumont@mediantechnologies.com. 2. Centre Antoine Lacassagne, 06100, Nice, France. 3. Centre Hospitalier Princesse Grâce, Avenue Pasteur, 98000, Monaco, Monaco. 4. Hopital La Pitié Salpêtrière, 75000, Paris, France.
Abstract
OBJECTIVES: Following the craze for radiomic features (RF), their lack of reliability raised the question of the generalizability of classification models. Inter-site harmonization of images therefore becomes a central issue. We compared RF harmonization processing designed to detect liver diseases in CT images. METHODS: We retrospectively analyzed 76 multi-center portal CT series of non-diseased (NDL) and diseased liver (DL) patients. In each series, we positioned volumes of interest in spleen and liver, then extracted 9 RF (histogram and texture). We evaluated two RF harmonization approaches. First, in each series, we computed the Z-score of liver measurements based on those computed in the spleen. Second, we evaluated the ComBat method according to each imaging center; parameters were computed in the spleen and applied to the liver. We compared RF distributions and classification performances before/after harmonization. We classified NDL versus spleen and versus DL tissues. RESULTS: The RF distributions were all different between liver and spleen (p < 0.05). The Z-score harmonization outperformed for the detection of liver versus spleen: AUC = 93.1% (p < 0.001). For the detection of DL versus NDL, in a case/control setting, we found no differences between the harmonizations: mean AUC = 73.6% (p = 0.49). Using the whole datasets, the performances were improved using ComBat (p = 0.05) AUC = 82.4% and degraded with Z-score AUC = 67.4% (p = 0.008). CONCLUSIONS: Data harmonization requires to first focus on data structuring to not degrade the performances of subsequent classifications. Liver tissue classification after harmonization of spleen-based RF is a promising strategy for improving the detection of DL tissue. KEY POINTS: • Variability of acquisition parameter makes radiomics of CT features non-reproducible. • Data harmonization can help circumvent the inter-site variability of acquisition protocols. • Inter-site harmonization must be carefully implemented and requires designing consistent data sets.
OBJECTIVES: Following the craze for radiomic features (RF), their lack of reliability raised the question of the generalizability of classification models. Inter-site harmonization of images therefore becomes a central issue. We compared RF harmonization processing designed to detect liver diseases in CT images. METHODS: We retrospectively analyzed 76 multi-center portal CT series of non-diseased (NDL) and diseased liver (DL) patients. In each series, we positioned volumes of interest in spleen and liver, then extracted 9 RF (histogram and texture). We evaluated two RF harmonization approaches. First, in each series, we computed the Z-score of liver measurements based on those computed in the spleen. Second, we evaluated the ComBat method according to each imaging center; parameters were computed in the spleen and applied to the liver. We compared RF distributions and classification performances before/after harmonization. We classified NDL versus spleen and versus DL tissues. RESULTS: The RF distributions were all different between liver and spleen (p < 0.05). The Z-score harmonization outperformed for the detection of liver versus spleen: AUC = 93.1% (p < 0.001). For the detection of DL versus NDL, in a case/control setting, we found no differences between the harmonizations: mean AUC = 73.6% (p = 0.49). Using the whole datasets, the performances were improved using ComBat (p = 0.05) AUC = 82.4% and degraded with Z-score AUC = 67.4% (p = 0.008). CONCLUSIONS: Data harmonization requires to first focus on data structuring to not degrade the performances of subsequent classifications. Liver tissue classification after harmonization of spleen-based RF is a promising strategy for improving the detection of DL tissue. KEY POINTS: • Variability of acquisition parameter makes radiomics of CT features non-reproducible. • Data harmonization can help circumvent the inter-site variability of acquisition protocols. • Inter-site harmonization must be carefully implemented and requires designing consistent data sets.
Authors: Roberto Berenguer; María Del Rosario Pastor-Juan; Jesús Canales-Vázquez; Miguel Castro-García; María Victoria Villas; Francisco Mansilla Legorburo; Sebastià Sabater Journal: Radiology Date: 2018-04-24 Impact factor: 11.105
Authors: Virendra Kumar; Yuhua Gu; Satrajit Basu; Anders Berglund; Steven A Eschrich; Matthew B Schabath; Kenneth Forster; Hugo J W L Aerts; Andre Dekker; David Fenstermacher; Dmitry B Goldgof; Lawrence O Hall; Philippe Lambin; Yoganand Balagurunathan; Robert A Gatenby; Robert J Gillies Journal: Magn Reson Imaging Date: 2012-08-13 Impact factor: 2.546
Authors: James P B O'Connor; Eric O Aboagye; Judith E Adams; Hugo J W L Aerts; Sally F Barrington; Ambros J Beer; Ronald Boellaard; Sarah E Bohndiek; Michael Brady; Gina Brown; David L Buckley; Thomas L Chenevert; Laurence P Clarke; Sandra Collette; Gary J Cook; Nandita M deSouza; John C Dickson; Caroline Dive; Jeffrey L Evelhoch; Corinne Faivre-Finn; Ferdia A Gallagher; Fiona J Gilbert; Robert J Gillies; Vicky Goh; John R Griffiths; Ashley M Groves; Steve Halligan; Adrian L Harris; David J Hawkes; Otto S Hoekstra; Erich P Huang; Brian F Hutton; Edward F Jackson; Gordon C Jayson; Andrew Jones; Dow-Mu Koh; Denis Lacombe; Philippe Lambin; Nathalie Lassau; Martin O Leach; Ting-Yim Lee; Edward L Leen; Jason S Lewis; Yan Liu; Mark F Lythgoe; Prakash Manoharan; Ross J Maxwell; Kenneth A Miles; Bruno Morgan; Steve Morris; Tony Ng; Anwar R Padhani; Geoff J M Parker; Mike Partridge; Arvind P Pathak; Andrew C Peet; Shonit Punwani; Andrew R Reynolds; Simon P Robinson; Lalitha K Shankar; Ricky A Sharma; Dmitry Soloviev; Sigrid Stroobants; Daniel C Sullivan; Stuart A Taylor; Paul S Tofts; Gillian M Tozer; Marcel van Herk; Simon Walker-Samuel; James Wason; Kaye J Williams; Paul Workman; Thomas E Yankeelov; Kevin M Brindle; Lisa M McShane; Alan Jackson; John C Waterton Journal: Nat Rev Clin Oncol Date: 2016-10-11 Impact factor: 66.675
Authors: Petros Kalendralis; Alberto Traverso; Zhenwei Shi; Ivan Zhovannik; René Monshouwer; Martijn P A Starmans; Stefan Klein; Elisabeth Pfaehler; Ronald Boellaard; Andre Dekker; Leonard Wee Journal: Med Phys Date: 2019-01-29 Impact factor: 4.071
Authors: Yang Nan; Javier Del Ser; Simon Walsh; Carola Schönlieb; Michael Roberts; Ian Selby; Kit Howard; John Owen; Jon Neville; Julien Guiot; Benoit Ernst; Ana Pastor; Angel Alberich-Bayarri; Marion I Menzel; Sean Walsh; Wim Vos; Nina Flerin; Jean-Paul Charbonnier; Eva van Rikxoort; Avishek Chatterjee; Henry Woodruff; Philippe Lambin; Leonor Cerdá-Alberich; Luis Martí-Bonmatí; Francisco Herrera; Guang Yang Journal: Inf Fusion Date: 2022-06 Impact factor: 17.564