Summary
Clinical CT scans can't show the tiny structures inside bones clearly, but very high-resolution scans that can are not practical for living patients. The authors present a new method that doesn't try to make sharp images pixel-by-pixel but instead focuses on predicting the bone's structural features from lower-resolution scans. This approach uses a network that models both how detailed images become blurry and how to recover the detailed structure, making the predictions more trustworthy and consistent with high-resolution references. The method works well on different kinds of bone scan data and can help doctors get detailed bone information from clinical scans.
What this means in practice
- •For medical imaging teams: Enable accurate bone microstructure analysis from standard clinical UHRCT scans using structural dual super-resolution for better diagnosis.
- •For medical device manufacturers: Incorporate the structural dual network to enhance CT scanner software, improving bone morphology visualization without new hardware.$Commercial implications: The paper's method makes it possible to sell upgraded imaging software that adds micro-level bone analysis to standard CT devices.
Abstract
Clinical CT and UHRCT cannot resolve individual trabeculae, whereas synchrotron radiation microCT (SRμCT) provides 3.2μm high-resolution references but is not applicable for in vivo imaging. The two domains differ by 31.25x in resolution, are only coarsely paired, and have drastically different data volumes. Moreover, clinical UHRCT suffers from severe partial volume effects, strong noise, and beam hardening/scatter artifacts, while SRμCT is nearly free. Existing super-resolution networks and pretrained-prior methods underperform because they target pixel generation--diverse details and SSIM/PSNR--and do not explicitly model these physical differences. This indicates that 32x super-resolution via pixel generation is intrinsically ill-posed. We propose a paradigm shift from pixel generation to topological inference: deterministically predicting invariant microstructures from macro-scale low-resolution inputs, evaluated by morphological parameters. We realize this paradigm via structural dual super-resolution, coupling forward physical degradation (micro-to-macro) with inverse structural inference (macro-to-micro) through structural duality constraints. The method is an end-to-end, few-shot, compact structural dual network (SDN), comprising a bidirectional modeling network for forward degradation and inverse reconstruction, a pyramid structural consistency discriminator, and four structural duality constraints. On the testset, SDN achieves morphological parameters largely consistent with SRμCT across 7 metrics, enabling clinical UHRCT with micro-imaging-level morphological quantification, with SSIM reaching 0.8. Trained on 3.2μm SSRF data, the model generalizes well to 3.25μm BSRF data from an independent source, validating cross-source generalization and confirming that the designed network achieves trustworthy structural inference rather than pixel generation.