Med Phys. 2026 Jan;53(1):e70208. doi: 10.1002/mp.70208.
ABSTRACT
BACKGROUND: Sacral chordoma (SC) is a rare and locally aggressive tumor, characterized by a high risk of local recurrence (LR). Advanced imaging biomarkers capturing tumor heterogeneity and treatment response may help refine patient stratification and improve outcomes.
PURPOSE: This study aims to explore the potential of microstructural parameters and habitat analysis derived from diffusion-weighted MRI (DW-MRI) to characterize intratumoral heterogeneity, assess treatment-induced changes, following carbon ion radiotherapy (CIRT), and investigate their association with recurrence patterns in SC patients.
METHODS: A retrospective analysis was performed on 40 SC patients treated with CIRT. Microstructural parameters-cell radius (R), eccentricity (ecc), diffusion coefficient (D), cell volume fraction (vf), and apparent cellularity (ρ-app)-were estimated from DW-MRI using a computational model based on Monte Carlo simulations. Longitudinal changes were assessed from DW-MRI scans acquired at baseline, first follow-up (3 months after CIRT), and, for LR patients, at the time of recurrence. Habitat imaging analysis was performed by applying K-means clustering (k = 2) independently to each map of the six microstructural parameters. A joint clustering approach was then used to combine these binary maps, resulting in 64 unique microstructural configurations (26 habitats). The spatial distribution of these habitats was quantified within the gross tumor volume (GTV) and recurrence regions. Associations with clinical outcomes were explored by analyzing habitat fractions in patients with early (early-LR) vs. late recurrence (late-LR), and by comparing recurrences predominantly located within the high-dose clinical target volume irradiated during CIRT (in-field-LR) vs. those partially outside the target (marginal-LR). Statistical tests, multiregional spatial interaction (MSI) matrix analysis, Cox regression, Spearman correlation and Kaplan-Meier survival analyses were performed.
RESULTS: Treatment-induced changes in microstructural parameters were statistically significant in both local control (LC) and LR patients, characterized by an increase in cell radius, eccentricity, ADC, and diffusion, and a concomitant reduction in cell volume fraction and apparent cellularity. This pattern suggested a reduction in cell density and increased tissue disorganization, likely reflecting CIRT-induced cell damage and extracellular matrix remodeling. Habitat analysis revealed distinct microstructural configurations associated with clinical outcome. A favorable microstructural pattern-high diffusivity and low cellularity-was predominantly found in LC patients (HABITAT47_212221), whereas more aggressive configurations (low diffusivity and high cellularity) were enriched in LR patients (HABITAT26_122112 and HABITAT28_122122). Univariate Cox regression identified these high-risk habitats as significantly associated with recurrence (hazard ratio > 1, p-value < 0.05). Moreover, habitat fractions showed a progressive trend across recurrence risk subgroups (LC, late-LR, early-LR; LC, in-field-LR, marginal-LR), while spatial analyses revealed increased co-localization of high-risk habitats in LR patients based on MSI matrix analysis. Finally, Kaplan-Meier analysis based on hazard ratios predicted at first follow-up using Cox models trained on baseline habitat fractions revealed significant differences in progression-free survival.
CONCLUSION: This study demonstrates the potential of DW-MRI-derived microstructural parameters and habitat imaging to capture intratumoral heterogeneity, assess treatment-induced microstructural changes, and identify high-risk patterns associated with LR in SC patients treated with CIRT. These findings support the integration of microstructural and habitat-based imaging as a non-invasive tool for risk stratification and personalized treatment planning in CIRT for SC.
PMID:41455110 | DOI:10.1002/mp.70208