Abstract

Optical coherence tomography (OCT) A-scan backscattering signals provide depth-resolved textural information about internal structures. However, conventional OCT imaging is limited by refraction-induced distortion and speckle noise, hindering fine detail resolution. While multi-angle imaging systems alleviate these issues through incoherent compounding of backscattering signals, in vivo applications face challenges: limited angular coverage during surface scanning degrades backscatter intensity compounding quality, and the absence of angular information introduces artifacts in multi-view position-intensity alignment. Furthermore, excessive smoothing during speckle suppression obscures fine textures. Consequently, reconstructing ultra-fine structures from limited-angle, sparse-view measurements remains a critical challenge. To address this, we present Backscattering-Corrected Implicit Representation Tomography (BCIRT), a framework for reconstructing multi-angle low-coherence signals. We also develop a dedicated limited-angle imaging system for intraoperative BCIRT deployment. BCIRT formulates cross-view backscattering signals as a continuous function of spatial position, utilizing implicit neural representation (INR) for fitting. A physics-informed iterative mechanism inversely models ray propagation to determine corrected ray paths, enhancing the neural representation’s robustness against distortions. Leveraging these corrected paths, we introduce a dual dynamic line mixer and a contrastive-guided discriminative deblurring module to achieve high-resolution microstructure reconstruction with reduced speckle noise. Extensive experiments on biological samples and surgical resected samples demonstrate that our method achieves state-of-the-art performance, highlighting its potential for clinical applications and biomedical research.

Fig. 1. Overall pipeline of BCIRT. The multi-angle A-scan rays are first aligned through ray-tracing strategy and then mapped to measurements via INR model. Due to the angular-dependent refractive distortion, coordinates from different angles corresponding to the same physical location are displaced to x1 and x2. The mismatched coordinates are subsequently corrected to the physical position through ray optimization. The dashed box indicates the A-scan corresponding to the line position of B-scans.

Fig. 2. In stage I, the initial ray positions are optimized through the ray-tracing process to obtain the corrected ray positions φi. In stage II, these ray positions are fed into the multi-branch INR architecture and encoded into a latent feature space to reconstruct the backscattering signals in the B-scans. Specifically, we introduce the spatial mask learning, which extracts the corresponding frequency features by applying masks with DLM module and SIREN.

https://doi.org/10.1016/j.media.2026.104000