Improvement of the clinical utility of optical coherence tomography (OCT) retinal nerve fiber layer (RNFL) measurement by establishing data comparability across the OCT technology generations and models

摘要

Glaucoma is the second leading cause of blindness worldwide, which induces irreversible structural damage (retinal ganglion cell loss and retinal nerve fiber layer (RNFL) thinning) on the retina. Optical coherence tomography (OCT) provides RNFL thickness measurements, which have become an essential clinical measure for objective glaucoma assessment. RNFL thickness is measured on a cross-sectional retinal image sampled along a 3.4mm circle centered around the optic nerve head (ONH). With the conventional time-domain OCT (TD-OCT), its operator dependent scan registration is responsible for the majority of measurement variability. Recently, spectral domain OCT (SD-OCT) technology has been introduced. SD-OCT provides faster scanning (up to 100x) and finer axial resolution (up to 2x) compared to TD-OCT, allowing three-dimensional (3D) volume sampling. 3D SD-OCT data can be visualized as an en face image of the retina. This allows us to create a virtual OCT image along any sampling line (curved or straight), which permits virtually perfect scan registration. The objective of this study is to improve the clinical utility of OCT RNFL measurement by establishing data comparability across the multiple OCT generations and models. First, we developed an algorithm to match the TD-OCT scan location within the corresponding 3D SD-OCT volume. Scan location matching (SLM) enables computation of the calibration equation between TD-OCT and SD-OCT for direct comparison of measurements, bridging the old technology with new ones. Second, the performance of the SLM method was measured using various SD-OCT devices with different spatial sampling methods. By making TD-OCT measurements at one time point comparable to the most recent SD-OCT measurement using SLM, glaucoma progression can be assessed on one to one basis. However, due to the variable TD-OCT scan registration over multiple visits, one can still not analyze the trend of glaucoma progression because RNFL thickness measured at different locations is not directly comparable even after calibration. Therefore, we developed a mathematical model of the retinal nerve fiber bundle distribution pattern to normalize the off-centered TD-OCT RNFL thickness to a virtually centered one. The outcome of this study would facilitate more accurate and reliable glaucoma disease/progression detection in cross-sectional as well as longitudinal clinical settings.