Article révisé par les pairs
Résumé : Abstract Prolonged microgravity exposure causes significant cardiac deconditioning through adaptive physiological changes [1]. Head-Down Tilt (HDT) bed rest reproduces these effects on Earth, demonstrating progressive decreases in cardiac volumes and masses during unloading, with subsequent rapid readjustment upon gravity restoration in the head-to-foot direction [2]. This suggests apparent myocardial mass loss may result from transient extracellular matrix fluid shifts rather than true atrophy. However, quantitative studies investigating these mechanisms remain limited. We investigate myocardial extracellular matrix hydration changes during HDT using T2 mapping-based radiomics analysis. T2 mapping quantifies tissue liquid content [3], while radiomics extracts quantitative features from CMR images, enabling detailed myocardial tissue characterization. Radiomic features were extracted from T2-maps by four operators segmenting the myocardium using LIFEX software. Images were acquired at baseline (BDC), HDT day 56 (HDT56), and 4 days post-HDT (R) in 8 control subjects from the 60-day Artificial Gravity Bed Rest European Space Agency (AGBRESA) campaign. To facilitate the comparison, features were normalized using median subtraction and mean absolute deviation division. Feature selection involved identifying highly correlated pairs (r>0.8), retaining the feature among them with higher variance explanation, in this way removing redundant and less informative features. Statistical analysis used Friedman and post-hoc Wilcoxon tests (p<0.05, Bonferroni-corrected p<0.0167) to identify significant radiomic features reflecting hydration changes between epochs. Inter-operator agreement was assessed using coefficient of variation (CV) and intra-class correlation coefficient (ICC). RIM-IntensitySum (ICC 0.67, CV 13.53%), reflecting T2 signal summation in the myocardial rim (or periphery), decreased (p=0.0156) by 18.72% from BDC to HDT56, with 12.40% recovery (p=0.2500) from HDT56 to R. MeanIntensity (ICC 0.77, CV 9.72%), quantifying average myocardial T2 signal, showed similar trends: 11.48% reduction (p=0.0781) and 4.49% recovery (p=0.7422), respectively. RIM-CountingVoxels (ICC 0.96, CV 4.35%), representing voxels count in the rim, decreased (p=0.0078) by 8.22% (BDC to HDT56) and increased (p=0.0156) by 4.48% (HDT56 to R). This study confirmed significant radiomic feature changes across HDT epochs in T2 maps. Intensity-based features reflected myocardial hydration shifts, mirroring trends in myocardial mass. ICC values (0.67–0.972) showed moderate to excellent reproducibility. While RIM-IntensitySum had limited interpretability due to high variability relative to the effect size, other features showed better precision. These findings support T2-based radiomics as a promising tool for detecting myocardial tissue changes during simulated microgravity, suggesting a link to fluid shifts rather than structural atrophy.

Documents en relation

DI-fusion