Early ADC changes in risk-model-defined brain regions precede contrast-enhancing lesions after proton therapy for low-grade glioma

Background Contrast-enhancing brain lesions (CEBLs) are a challenging late effect of proton therapy (PT) in low-grade glioma (LGG), typically detected after blood–brain barrier disruption. Diffusion-weighted imaging (DWI) and derived apparent diffusion coefficient (ADC) metrics can reveal early microstructural alterations preceding this disruption. Their diagnostic value can be enhanced by combining it with voxel-wise risk-modelling, such as the Probability of Lesion Origin (POLO) model, which maps spatial susceptibility to radiation-induced damage. This pilot study investigates the feasibility of combining longitudinal DWI with POLO-based risk stratification to characterize regional tissue response trajectories after PT in patients with and without CEBL. Methods Twenty LGG-patients treated with PT were retrospectively analyzed: ten developed CEBLs (CEBL⁺) and ten remained lesion-free. Longitudinal ADC-volume-histogram metrics (ADC 20% , ADC mean , ADC median ) were extracted from POLO-defined high-risk (HR) and low-risk (LR) volumes. Temporal evolution was quantified using coefficients of variation and slopes. Results CEBL⁺ patients exhibited higher ADC values, greater temporal variability, and steeper increases over time, especially within HR regions. For ADC 20% , HR values exceeded LR in CEBL⁺, while no HR–LR difference was observed in CEBL⁻. The HR–LR ADC 20% difference distinguished CEBL⁺ from CEBL⁻ patients as early as six months post-treatment, yielding AUCs of 0.70–0.79 across quarterly timepoints. Conclusions Risk-model-informed interpretation of DWI, considering localized information on anticipated biological damage, demonstrates the potential for quantitative assessment of post-treatment tissue dynamics. Early HR–LR ADC divergence, particularly in ADC 20% , may serve as an imaging biomarker for monitoring radiation-induced tissue changes after PT, supporting post-treatment imaging surveillance.