Improving the Generalizability of Deep Learning for T2-Lesion Segmentation of Gliomas in the Post-Treatment Setting
Bioengineering · 16.05.2024 · 11 (5) · 497
DOI: 10.3390/bioengineering11050497 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bioengineering, online erschienen am 16.05.2024. Verfasst von Jacob Ellison und 11 weiteren Autor:innen. Beteiligte Einrichtung: Department of Radiology and Biomedical Imaging, UCSF, San Francisco, CA 94143, USA. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 80 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NCI NIH HHS, Department of Defense, UCSF Helen Diller Family Cancer Center Cancer.
Abstract (Originalsprache)
Although fully automated volumetric approaches for monitoring brain tumor response have many advantages, most available deep learning models are optimized for highly curated, multi-contrast MRI from newly diagnosed gliomas, which are not representative of post-treatment cases in the clinic. Improving segmentation for treated patients is critical to accurately tracking changes in response to therapy. We investigated mixing data from newly diagnosed (n = 208) and treated (n = 221) gliomas in training, applying transfer learning (TL) from pre- to post-treatment imaging domains, and incorporating spatial regularization for T2-lesion segmentation using only T2 FLAIR images as input to improve generalization post-treatment. These approaches were evaluated on 24 patients suspected of progression who had received prior treatment. Including 26% of treated patients in training improved performance by 13.9%, and including more treated and untreated patients resulted in minimal changes. Fine-tuning with treated glioma improved sensitivity compared to data mixing by 2.5% (p < 0.05), and spatial regularization further improved performance when used with TL by 95th HD, Dice, and sensitivity (6.8%, 0.8%, 2.2%; p < 0.05). While training with ≥60 treated patients yielded the majority of performance gain, TL and spatial regularization further improved T2-lesion segmentation to treated gliomas using a single MR contrast and minimal processing, demonstrating clinical utility in response assessment.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Radiology and Biomedical Imaging, UCSF, San Francisco, CA 94143, USA.
- Center for Intelligent Imaging, UCSF, San Francisco, CA 94143, USA.
- UCSF/UC Berkeley Graduate Program in Bioengineering, San Francisco, CA 94143, USA.
- Department of Neurological Surgery, UCSF, San Francisco, CA 94143, USA.
ORCID-Profile: Jacob Ellison, Francesco Caliva, Pablo Damasceno, Marisa LaFontaine, Yan Li, Janine M Lupo
Förderung
- NCI NIH HHS
- Department of Defense
- UCSF Helen Diller Family Cancer Center Cancer
- NIH HHS
- National Institutes of Health
- UCSF Helen Diller Family Cancer Center Cancer Imaging Resources Pilot Grant
Zitieren
APA
Ellison, J., Caliva, F., Damasceno, P., Luks, T. L., LaFontaine, M., Cluceru, J., Kemisetti, A., Li, Y., Molinaro, A. M., Pedoia, V., Villanueva-Meyer, J. E., & Lupo, J. M. (2024). Improving the Generalizability of Deep Learning for T2-Lesion Segmentation of Gliomas in the Post-Treatment Setting. Bioengineering, 11(5), 497. https://doi.org/10.3390/bioengineering11050497
Vancouver
Ellison J, Caliva F, Damasceno P, Luks TL, LaFontaine M, Cluceru J, et al. Improving the Generalizability of Deep Learning for T2-Lesion Segmentation of Gliomas in the Post-Treatment Setting. Bioengineering. 2024;11(5):497. doi:10.3390/bioengineering11050497
BibTeX
@article{Ellison2024,
author = {Ellison, Jacob and Caliva, Francesco and Damasceno, Pablo and Luks, Tracy L and LaFontaine, Marisa and Cluceru, Julia and Kemisetti, Anil and Li, Yan and Molinaro, Annette M and Pedoia, Valentina and Villanueva-Meyer, Javier E and Lupo, Janine M},
title = {Improving the Generalizability of Deep Learning for T2-Lesion Segmentation of Gliomas in the Post-Treatment Setting},
journal = {Bioengineering},
year = {2024},
volume = {11},
number = {5},
pages = {497},
doi = {10.3390/bioengineering11050497}
}
EUVIMED