Time-dependent diffusion MRI for microstructural characterisation of rhabdomyosarcoma and extraosseous Ewing sarcoma xenografts: a proof-of-concept study with slice-by-slice radiologic-pathologic correlation

Abstract Background Differentiating rhabdomyosarcoma (RMS) from extraosseous Ewing sarcoma (EES) remains clinically challenging due to their highly overlapping features on multiparametric MRI. Therefore, this study aimed to investigate the feasibility of time-dependent diffusion MRI (T d -dMRI) for detecting microstructural differences between RD (RMS) and A673 (EES) xenograft models through slice-by-slice radiologic-pathologic correlation of matched MRI and histological sections. Methods Human-derived xenograft models (33 RMS, 30 EES) established in nude mice underwent T d -dMRI. Microstructural parameters, including the extracellular diffusivity ( D ex ), intracellular volume fraction ( V in ), cellularity, and cell diameter ( d ), were derived from whole-tumor volumes using the IMPULSED model by two independent readers. Crucially, H&E-stained sections were registered with MRI images slice-by-slice to enable radiologic-pathologic correlation of matched sections. Spearman’s correlation assessed the relationships between in vivo MRI parameters and matched ex vivo microscopic features. Diagnostic accuracies of T d -dMRI parameters and apparent diffusion coefficients (ADCs) were compared using receiver operating characteristic (ROC) analysis. Results Compared to EES, RMS exhibited significantly lower cellularity (2.88 vs. 4.34 μm −1 ) and V in (0.44 vs. 0.58), larger cell diameter (18.59 vs. 16.56 μm), higher D ex (0.82 vs. 0.69 µm 2 /ms), and a lower relative ADC change from PGSE to OGSE 50Hz (85.5% vs. 179.8%) (all P < 0.01). T d -dMRI cellularity (AUC = 0.92) and ADC 50Hz (AUC = 0.91) showed comparable diagnostic performance ( P = 0.67, DeLong test) for differentiating the two xenograft models. Radiologic-pathologic correlation of the matched sections revealed significant associations between all T d -dMRI parameters and histological features, with in vivo MRI cellularity exhibiting the strongest positive correlation with ex vivo pathological nuclear density (r = 0.784, P < 0.001). Conclusions T d -dMRI microstructural parameters, particularly cellularity, can detect microstructural differences between RD and A673 xenograft models and correlate with co-registered histopathology. These model-specific findings do not yet establish clinical differentiation of human RMS and EES. Validation across additional biological models and human cohorts is required before clinical translation.