Reliability-Aware Hybrid Fusion of CT and MRI with Evidence-Consistent Edge Preservation

Multimodal fusion of Computed Tomography (CT) scans and Magnetic Resonance Imaging (MRI) scans aims to provide high-density anatomical structural information and soft-tissue contrast within a single image. Current pixel-wise approaches to this fusion attempt to retain structural detail from CT scans or imply rigid weighting or singleton cues with data intensive learn models, thereby undermining generalisability with registration errors. In this paper, we propose a train-free hybrid approach to MRI-CT scan fusion, combining contrast-guided structural averaging with Global Principal Component Analysis representation via a decision-making process guided by local reliability cues. The performance of this hybrid system was tested with 184 co-registered CT-MRI pairs with modality-scaled metrics for fidelity, edge coherence, MRI-PSNR with 16.59 dB, MRI edge coherence with 39.42% similarity, with CT-scaled trends remaining highly competitive with Baseline-PCA methods, being models with structural coherence.