Impact of CT dose on AI performance: A comparison of radiomics, deep, and foundation models in a multicentric anthropomorphic phantom study
Medical Physics · 2026-03 · 53 (3)
DOI: 10.1002/mp.70374 ↗Kurz gefasst
Vergleichsstudie in der Zeitschrift Medical Physics, online erschienen am 18.03.2026. Verfasst von María Martín Asiain und 12 weiteren Autor:innen. Beteiligte Einrichtung: Institute of Informatics, School of Management, HES-SO Valais-Wallis University of Applied Sciences and Arts Western Switzerland, Sierre, Switzerland. Hauptthemen laut MEDLINE: Radiomics, Tomography, X-Ray Computed, Radiation Dosage, Phantoms, Imaging. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 17 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Swiss Personalized Health Network (SPHN), Swiss National Science Foundation (SNSF), Swiss Cancer Research Foundation.
Abstract (Originalsprache)
Background
Computed tomography (CT) is widely used in clinical practice due to its ability to provide detailed anatomical information. However, variations in radiation dose can affect image quality, potentially compromising the performance and reliability of artificial intelligence (AI) models applied to these images.
Purpose
To evaluate the robustness of radiomics‐based and deep learning‐based models to variations in CT dose levels using a standardized dataset obtained from a 3D‐printed anthropomorphic phantom simulating liver tissue with anomalies, as well as in the publicly available dataset CT‐ORG with real patient data for organ classification. This study is in an early experimental stage, tested only on retrospective data.
Methods
A total of 1378 image series from 649 scans were acquired across 13 scanners from four manufacturers at five dose levels. Features were extracted from six regions of interest (ROIs), representing four liver tissue types (normal, cyst, hemangioma, metastasis), using four methods: PyRadiomics, a shallow convolutional neural network (CNN), SwinUNETR, and a CT foundation model (CT‐FM). Feature stability was assessed using the Intraclass Correlation Coefficient (ICC), while Uniform Manifold Approximation and Projection (UMAP) was employed to evaluate tissue types separability and the influence of scanner variations. Generalizability was tested by training liver tissue classifiers on one dose level and testing on others, alongside a dose classification task (10‐fold cross‐validation) to determine the sensitivity of each method to dose variations. In addition, we compared the four methods in addressing the task of organ classification (10‐fold cross‐validation) with the CT‐ORG dataset containing 140 CT scans acquired with varying dose levels.
Results
Radiomic features showed limited robustness to dose variations, leading to reduced performance in liver tissue classification and the lowest ICC among methods (ICC: 0.8355 0.1705). SwinUNETR and CT‐FM exhibited the highest stability (SwinUNETR ICC: 0.9528 0.0272; CT‐FM ICC: 0.9347 0.0420), clearly above the Shallow CNN (ICC: 0.8416 0.2018). CT‐FM also showed strong generalization across dose levels: its features effectively distinguished between liver tissue types and dose levels simultaneously, without compromising performance in either task. Consistent with these trends in dose sensitivity, CT‐FM obtained the highest dose‐classification accuracy (0.6517 0.0179), whereas SwinUNETR showed the lowest (0.3796 0.0250). These trends were confirmed in the context of organ classification with real patient data on the CT‐ORG dataset, where CT‐FM achieved the highest accuracy (0.965).
Conclusions
The study highlights the limited robustness of traditional radiomics and deep models to CT dose variation and underscores the potential of foundation models like CT‐FM to enable robust clinical applications by mitigating dose‐related variability. This enhanced performance is likely due to the model's pretraining on large and diverse datasets, allowing it to learn robust and generalizable representations across varying acquisition conditions.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Institute of Informatics, School of Management, HES-SO Valais-Wallis University of Applied Sciences and Arts Western Switzerland, Sierre, Switzerland.
- Nuclear Medicine and Molecular Imaging Department, Lausanne University Hospital, Lausanne, Switzerland.
- Clinic of Radiology and Nuclear Medicine, University Hospital Basel, University of Basel, Basel, Switzerland.
- Faculty of Medicine, University of Geneva (UNIGE), Geneva, Switzerland.
- Computer Vision Lab, ETH Zurich, Zurich, Switzerland.
- Idiap Research Institute, Martigny, Switzerland.
Förderung
- Swiss Personalized Health Network (SPHN)
- Swiss National Science Foundation (SNSF)
- Swiss Cancer Research Foundation
- Lundin Family Brain Tumour Research Centre at CHUV
Zitieren
APA
Martín Asiain, M., Amirian, M., Jimenez Del Toro, O., Aberle, C., Schaer, R., Bach, M., Obmann, M., Flouris, K., Müller, H., Stieltjes, B., Konukoglu, E., Andrearczyk, V., & Depeursinge, A. (2026). Impact of CT dose on AI performance: A comparison of radiomics, deep, and foundation models in a multicentric anthropomorphic phantom study. Medical Physics, 53(3). https://doi.org/10.1002/mp.70374
Vancouver
Martín Asiain M, Amirian M, Jimenez Del Toro O, Aberle C, Schaer R, Bach M, et al. Impact of CT dose on AI performance: A comparison of radiomics, deep, and foundation models in a multicentric anthropomorphic phantom study. Medical Physics. 2026;53(3). doi:10.1002/mp.70374
BibTeX
@article{MartnAsiain2026,
author = {Martín Asiain, María and Amirian, Mohammadreza and Jimenez Del Toro, Oscar and Aberle, Christoph and Schaer, Roger and Bach, Michael and Obmann, Markus and Flouris, Kyriakos and Müller, Henning and Stieltjes, Bram and Konukoglu, Ender and Andrearczyk, Vincent and Depeursinge, Adrien},
title = {Impact of CT dose on AI performance: A comparison of radiomics, deep, and foundation models in a multicentric anthropomorphic phantom study},
journal = {Medical Physics},
year = {2026},
volume = {53},
number = {3},
doi = {10.1002/mp.70374}
}
EUVIMED