Cooperative AI training for cardiothoracic segmentation in computed tomography: An iterative multi-center annotation approach

Lassen-Schmidt B, Baessler B, Gutberlet M, Berger J, Brendel JM, Bucher AM, Emrich T, Fervers P, Kottlors J, Kuhl P, May M, Penzkofer T, Persigehl T, Renz D, Sähn MJ, Siegler L, Kohlmann P, Köhn A, Link F, Meine H, Thiemann MT, Hahn HK, Sieren MM (2024)

Publication Type: Journal article

Publication year: 2024

Journal

European Journal of Radiology Elsevier

Book Volume: 176

Article Number: 111534

DOI: 10.1016/j.ejrad.2024.111534

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 with CRIS profile

Matthias May Institute of Radiology Lisa Siegler Institute of Radiology

Additional Organisation(s)

FAU Profilzentrum Licht.Materie.Quantentechnologien (FAU LMQ)

Involved external institutions

Fraunhofer-Institut für Bildgestützte Medizin (MEVIS) DE Germany (DE) Universitätsklinikum Würzburg DE Germany (DE) Universität Leipzig DE Germany (DE) Eberhard Karls Universität Tübingen DE Germany (DE) Universitätsklinikum Frankfurt am Main (KGU) DE Germany (DE) Universitätsmedizin der Johannes Gutenberg-Universität Mainz DE Germany (DE) Universitätsklinikum Köln DE Germany (DE) Charité - Universitätsmedizin Berlin DE Germany (DE) Medizinische Hochschule Hannover (MHH) / Hannover Medical School DE Germany (DE) Universitätsklinikum Aachen (UKA) DE Germany (DE) Universitätsklinikum Schleswig-Holstein (UKSH) DE Germany (DE)

How to cite

APA:

Lassen-Schmidt, B., Baessler, B., Gutberlet, M., Berger, J., Brendel, J.M., Bucher, A.M.,... Sieren, M.M. (2024). Cooperative AI training for cardiothoracic segmentation in computed tomography: An iterative multi-center annotation approach. European Journal of Radiology, 176. https://doi.org/10.1016/j.ejrad.2024.111534

MLA:

Lassen-Schmidt, Bianca, et al. "Cooperative AI training for cardiothoracic segmentation in computed tomography: An iterative multi-center annotation approach." European Journal of Radiology 176 (2024).

BibTeX: Download