Cooperative AI training for cardiothoracic segmentation in computed tomography: An iterative multi-center annotation approach
European Journal of Radiology, vol. 176, pp. 111534
Abstract
PURPOSE: Radiological reporting is transitioning to quantitative analysis, requiring large-scale multi-center validation of biomarkers. A major prerequisite and bottleneck for this task is the voxelwise annotation of image data, which is time-consuming for large cohorts. In this study, we propose an iterative training workflow to support and facilitate such segmentation tasks, specifically for high-resolution thoracic CT data. METHODS: Our study included 132 thoracic CT scans from clinical practice, annotated by 13 radiologists. In three iterative training experiments, we aimed to improve and accelerate segmentation of the heart and mediastinum. Each experiment started with manual segmentation of 5-25 CT scans, which served as training data for a nnU-Net. Further iterations incorporated AI pre-segmentation and human correction to improve accuracy, accelerate the annotation process, and reduce human involvement over time. RESULTS: Results showed consistent improvement in AI model quality with each iteration. Resampled datasets improved the Dice similarity coefficients for both the heart (DCS 0.91 [0.88; 0.92]) and the mediastinum (DCS 0.95 [0.94; 0.95]). Our AI models reduced human interaction time by 50 % for heart and 70 % for mediastinum segmentation in the most potent iteration. A model trained on only five datasets achieved satisfactory results (DCS > 0.90). CONCLUSIONS: The iterative training workflow provides an efficient method for training AI-based segmentation models in multi-center studies, improving accuracy over time and simultaneously reducing human intervention. Future work will explore the use of fewer initial datasets and additional pre-processing methods to enhance model quality.
Authors 23
-
Bianca C Lassen-Schmidt corresponding
Fraunhofer Institute for Digital Medicine
Affiliation as printed
Fraunhofer Institute for Digital Medicine MEVIS, Max-von-Laue-Str. 2 28359, Bremen, Germany
-
Bettina Baeßler corresponding
Affiliation as printed
Department of Diagnostic and Interventional Radiology, University Hospital Würzburg, Oberdürrbacher Str. 6 97080, Würzburg, Germany
-
Matthias Gutberlet corresponding
Affiliation as printed
Herzzentrum - University Leipzig, Strümpellstraße 39 04289, Leipzig, Germany
Herzzentrum – University Leipzig, Strümpellstraße 39 04289, Leipzig, Germany
-
Josephine Berger corresponding
Affiliation as printed
Department of Radiology, Diagnostic and Interventional Radiology, University of Tübingen, Hoppe-Seyler-Straße 3 72076, Tübingen, Germany
-
Jan Michael Brendel corresponding
Affiliation as printed
Department of Radiology, Diagnostic and Interventional Radiology, University of Tübingen, Hoppe-Seyler-Straße 3 72076, Tübingen, Germany
-
Andreas Michael Bucher corresponding
Affiliation as printed
Department of Diagnostic and Interventional Radiology, University Hospital Frankfurt, Theodor-Stern-Kai 7 60596, Frankfurt am Main, Germany
-
Tilman Emrich corresponding
Medical University of South Carolina · Johannes Gutenberg University Mainz · University Medical Center of the Johannes Gutenberg University Mainz · German Centre for Cardiovascular Research
Affiliation as printed
University Medical Center of Johannes-Gutenberg-University, Langenbeckstraße 1 55131, Mainz, Germany; Department of Diagnostic and Interventional Radiology, and Division of Cardiovascular Imaging, Department of Radiology and Radiological Science, Medical University of South Carolina, Charleston, USA; German Centre for Cardiovascular Research, Partner Site Rhine-Main, Mainz, Germany
Department of Diagnostic and Interventional Radiology, and Division of Cardiovascular Imaging, Department of Radiology and Radiological Science, Medical University of South Carolina, Charleston, USA
German Centre for Cardiovascular Research, Partner Site Rhine-Main, Mainz, Germany
University Medical Center of Johannes-Gutenberg-University, Langenbeckstraße 1 55131, Mainz, Germany
-
Philipp Fervers corresponding
Affiliation as printed
Department of Diagnostic and Interventional Radiology, University Hospital Cologne, Kerpener Str. 62 50937, Köln, Germany
-
Jonathan Kottlors corresponding
Affiliation as printed
Department of Diagnostic and Interventional Radiology, University Hospital Cologne, Kerpener Str. 62 50937, Köln, Germany
-
Philipp Josef Kuhl corresponding
Affiliation as printed
Department of Diagnostic and Interventional Radiology, University Hospital Würzburg, Oberdürrbacher Str. 6 97080, Würzburg, Germany
-
Matthias Stefan May corresponding
Affiliation as printed
Department of Radiology, University Hospital Erlangen, Maximilianspl. 1 91054, Erlangen, Germany
-
Tobias Penzkofer corresponding
Charité - Universitätsmedizin Berlin
Affiliation as printed
Department of Radiology, Charité, University Hospital Berlin, Augustenburger Pl. 1 13353, Berlin, Germany
-
Thorsten Persigehl corresponding
Affiliation as printed
Department of Diagnostic and Interventional Radiology, University Hospital Cologne, Kerpener Str. 62 50937, Köln, Germany
-
Diane Miriam Renz corresponding
Medizinische Hochschule Hannover
Affiliation as printed
Institute of Diagnostic and Interventional Radiology, Department of Pediatric Radiology, Hannover Medical School, Carl-Neuberg-Straße 1 30625, Hannover, Germany
-
Affiliation as printed
Department of Diagnostic and Interventional Radiology, University Hospital Aachen, Pauwelsstraße 30 52074, Aachen, Germany
-
Lisa Siegler corresponding
Affiliation as printed
Department of Radiology, University Hospital Erlangen, Maximilianspl. 1 91054, Erlangen, Germany
-
Peter Kohlmann corresponding
Fraunhofer Institute for Digital Medicine
Affiliation as printed
Fraunhofer Institute for Digital Medicine MEVIS, Max-von-Laue-Str. 2 28359, Bremen, Germany
-
Alexander Köhn corresponding
Fraunhofer Institute for Digital Medicine
Affiliation as printed
Fraunhofer Institute for Digital Medicine MEVIS, Max-von-Laue-Str. 2 28359, Bremen, Germany
-
Florian Link corresponding
Fraunhofer Institute for Digital Medicine
Affiliation as printed
Fraunhofer Institute for Digital Medicine MEVIS, Max-von-Laue-Str. 2 28359, Bremen, Germany
-
Hans Meine corresponding
Fraunhofer Institute for Digital Medicine
Affiliation as printed
Fraunhofer Institute for Digital Medicine MEVIS, Max-von-Laue-Str. 2 28359, Bremen, Germany
-
Marc T. Thiemann corresponding
Fraunhofer Institute for Digital Medicine
Affiliation as printed
Fraunhofer Institute for Digital Medicine MEVIS, Max-von-Laue-Str. 2 28359, Bremen, Germany
-
Horst Karl Hahn corresponding
University of Bremen · Fraunhofer Institute for Digital Medicine
Affiliation as printed
Fraunhofer Institute for Digital Medicine MEVIS, Max-von-Laue-Str. 2 28359, Bremen, Germany
University of Bremen, Department of Mathematics/Computer Science, Bibliothekstraße 5 28359, Bremen, Germany
-
Malte Maria Sieren corresponding
University Hospital Schleswig-Holstein
Affiliation as printed
Department of Radiology and Radiotherapy, University Hospital Schleswig-Holstein, Ratzeburger Allee 160 23562, Lübeck, Germany
Cited by 4 stored of 4
4 results
No patents citing this paper on Lens.org (checked 2026-10-06).