Related Experiment Video
Updated: Aug 14, 2026
Three-Dimensional Phase Resolved Functional Lung Magnetic Resonance Imaging
Published on: June 21, 2024
Quantitative assessment of fetal lung development using MRI-based radiomics: a comparative analysis of 3D and 2D
Yaoxi Jiang1,2, Pan Xiao3, Song Peng2
1Department of Radiology, The First Affiliated Hospital of Chongqing Medical University, Chongqing, China.
Objective:
Fetal lung development is a critical determinant of neonatal health. This study aimed to externally validate and compare the performance of 3D versus 2D MRI radiomics models in quantifying fetal lung development. Gestational age estimation was utilized as a quantitative surrogate marker for this assessment in a large multi-institutional cohort with an independent test set.
Methods:
To compare 2D versus 3D radiomics, we analyzed 292 fetal MRI scans using T2-weighted coronal images. The cohort comprised an original dataset (n=252) and an independent test set (n=40) from two institutions, with gestational ages ranging from 28 to 38+6 weeks. For each case, 102 radiomics features were extracted from manually segmented 2D (tracheal bifurcation plane) and 3D (entire lung) regions of interest (ROIs). Gestational age estimation models were developed using support vector regression (SVR) and partial least squares regression (PLSR) and validated via regression metrics. Furthermore, intraclass correlation coefficients (ICCs) were calculated to evaluate the correspondence between 2D and 3D radiomic measurements.
Results:
Features extracted from 3D ROIs exhibited superior performance for gestational age estimation, achieving Pearson correlation coefficients (r) of 0.904 (SVR) and 0.902 (PLSR) in the original dataset, and 0.846 (SVR) and 0.861 (PLSR) in the independent test set. In contrast, 2D ROI-based models showed significantly lower accuracy. The median ICCs for the correspondence between 2D and 3D radiomic measurements were 0.379 (IQR: 0.179-0.554; range: 0.010-0.880) in the original dataset and 0.247 (IQR: 0.060-0.384; range: -0.024-0.721) in the test set.
Conclusion:
This study demonstrates that 3D MRI radiomics features outperform 2D features for gestational age estimation, serving as a quantitative surrogate for fetal lung development. These findings underscore the clinical potential of 3D radiomics in non-invasive prenatal assessment. Future studies incorporating larger cohorts and standardized imaging protocols are encouraged to further enhance the model's robustness and clinical utility.
