3D Reconstruction of Medical Images Using UNet and Edge Loss for Enhanced Diagnostic Visibility

Abstract

In medical diagnosis, enhancing the visibility of 3D CT images is critical for improving diagnostic utility. In this study, we propose a method aimed at enhancing image visibility by incorporating UNet architecture and an edge loss function into existing 3D reconstruction models, thereby clarifying image boundaries and local structures. Specifically, the use of UNet enhances the extraction of local features, while the edge loss function accentuates anatomical boundaries, collectively improving the visual clarity of the reconstructed images. The efficacy of the proposed approach is evaluated through quantitative metrics and visual assessments in comparison with existing 3D reconstruction models. Experimental results confirm that our method not only improves visibility and structural clarity but also enhances the diagnostic usefulness of reconstructed medical CT images.