Impact of Field Strength and Contrast Agents on Cardiac MRI Radiomic Features of the Myocardium and Blood Pool
Abstract The aim of this study was to investigate the effects of gadolinium-based contrast agents on cardiac MRI radiomic features across different magnetic field strengths. With Institutional Review Board (IRB) approval, eight healthy volunteers (38.5 ± 15.4 years; 4 males) underwent four consecutive cardiac MRI examinations using a Modified Look-Locker Inversion Recovery (MOLLI) sequence at 1.5T and 3T, with and without two contrast agents, gadopentetate dimeglumine (Gd-DTPA), and gadofosveset trisodium. Radiomic features were extracted from the myocardium and blood pool at the mid-ventricular level of the left ventricle using the first raw MOLLI image, corresponding to the shortest inversion time. Before contrast administration, switching from 1.5 to 3 T resulted in significant changes in 52% of myocardial and 56% of blood pool radiomic features. Contrast administration further altered radiomic feature values. At 1.5T, Gd-DTPA altered 42% of myocardial and 63% of blood pool features, whereas gadofosveset altered 73% and 85%, respectively. Similar patterns were observed at 3 T, with gadofosveset producing a greater proportion of altered features than Gd-DTPA ( p < 0.05). Blood pool features were generally more sensitive to contrast administration than myocardial features. At 10 min after contrast administration, significant differences between the two contrast agents were observed in 48% of myocardial and 23% of blood pool features at 1.5T and in 72% and 60%, respectively, at 3 T. In conclusion, the administration and type of gadolinium-based contrast agents significantly alter cardiac MRI radiomic features across different magnetic field strengths. Future research should prioritize normalization strategies and evaluate feature robustness across larger patient populations to enhance the generalizability of cardiac radiomics.
Authors
- Kai Lin (ORCID: https://orcid.org/0000-0002-0540-4077)
- James Carr (ORCID: https://orcid.org/0000-0003-1380-8095)
- Daniel Z. Gordon (ORCID: https://orcid.org/0000-0001-8059-3608)
- Roberto Sarnari
Publication Details
- Journal
- Journal of Imaging Informatics in Medicine
- Published
- 2026-10-19
- DOI
- https://doi.org/10.1007/s10278-026-02406-z
- Primary Topic
- Radiomics and Machine Learning in Medical Imaging
- Type
- article
- Field-Weighted Citation Impact
- 0.00