Vis-OCT Explorer: an open-source software for visible-light optical coherence tomography data processing.

BACKGROUND AND

OBJECTIVES

Visible-light optical coherence tomography (vis-OCT) has enabled the visualization of retinal structures and functions beyond the capabilities of conventional OCTs. However, to reconstruct high-quality images, vis-OCT requires special post-processing, including balanced detection. An open-source, standardized vis-OCT data processing software is essential for clinical applications and translation of vis-OCT.

METHODS

We developed Vis-OCT Explorer, an open-source, modular Python-based software for processing vis-OCT images. In addition to the standard spectral-domain OCT processing pipeline - including k-space resampling, dispersion compensation, and fast Fourier transformation - Vis-OCT Explorer offers unique dual-spectrometer balanced detection, short-time-Fourier transformation (STFT) based dispersion compensation coefficient optimization, and GPU-accelerated processing. We evaluated the reconstruction performance by quantifying a quality index extracted from individual B-scan images. We also assessed the repeatability of retinal thickness measurements by five operators on images acquired from different testing sites using the intraclass correlation coefficient (ICC) analysis.

RESULTS

Balanced detection and STFT-based dispersion compensation significantly increased the quality index of reconstructed B-scan images. ICC values of the retinal nerve fiber layer (RNFL) and ganglion cell-inner plexiform layer (GCIPL) thickness measurements from four testing sites exceeded 0.8 in 87.5% of the macular-centered images. The ICC of RNFL thickness measurements on all optic nerve head-centered images is above 0.8, showing strong repeatability across users.

CONCLUSIONS

Vis-OCT Explorer provides high-quality image processing and enables highly repeatable measurements on vis-OCT human retinal images. It facilitates future multicenter clinical tests to validate vis-OCT's clinical efficacy.