SpecstatOR: speckle statistics-based iOCT segmentation network for ophthalmic surgery

This paper introduces an approach to intraoperative optical coherence tomography (iOCT) segmentation, utilizing speckle patterns from tissue and tool scattering properties, defined by refractive index and structural composition, to differentiate retinal layers and instruments. Unlike classical deep learning approaches, our model trains on tissue-specific characteristics, enhancing robustness across different devices and anatomical variations and eliminating retraining. Consequently, our approach reduces the dependency on shape and intensity, addressing the limitations of state-of-the-art iOCT segmentation techniques used during surgical procedures.