BLUE RIBBON POSTER IMAGING: Development and Validation of a Multimodal Imaging Framework to Assess Mandibular Osteoradionecrosis In a Rat Model
Abstract
Purpose
Mandibular osteoradionecrosis (ORN) is a severe late complication of head-and-neck radiotherapy. Vascular compromise is thought to play a key role in ORN pathogenesis, yet early progression preceding bone failure remains poorly characterized. While computed tomography (CT) identifies advanced bone changes, recent clinical dynamic contrast-enhanced MRI (DCE-MRI) studies report increased vascular permeability in ORN, suggesting earlier sensitivity. This study aimed to develop and validate a multimodal imaging framework combining DCE-MRI, µCT, and cryo-fluorescence tomography (CFT) to link early perfusion and metabolic changes to structural damage in a rat model of mandibular ORN.
Methods
A multimodal imaging framework was designed incorporating µCT, DCE-MRI and CFT with cross-modality co-registration. µCT quantified bone volume and alveolar bone height. DCE-MRI data were denoised using higher-order total variation (HOTV) spatiotemporal filtering prior to pharmacokinetic modeling (Tofts) to extract transfer constant (Ktrans), extravascular extracellular space volume fraction (ve), and plasma volume fraction (vp), along with semi-quantitative metrics including area under curve, peak enhancement, and wash-in/out slopes. Denoising performance was evaluated using signal-to-noise ratio (SNR), edge preservation metrics, and noise power spectrum (NPS). Model performance was assessed using coefficient of determination (R²) and root mean square error (RMSE). Registration accuracy was quantified using target registration error (TRE).
Results
HOTV denoising reduced noise by 29% in the mandible, improving SNR from 14.7 dB to 17.8 dB while preserving 70% of edge information based on gradient magnitude ratios. In an adjacent 2D muscle region, NPS analysis demonstrated 9.7 dB noise suppression. Pharmacokinetic modeling produced stable fits within mandibular tissue (median R² = 0.88, RMSE = 0.006 mM). Representative TREs were 168 µm (µCT-to-CFT) and 217 µm (MRI-to-CFT).
Conclusion
This multimodal framework enables quantitative ORN assessment through integrated perfusion, metabolic, and structural imaging. HOTV denoising improves DCE-MRI data quality, enabling robust perfusion modeling to detect vascular compromise preceding morphological changes.