Axial Length Estimation Using Metadata From Swept-Source Optical Coherence Tomography Angiography

Area of Science:

• Ophthalmology
• Medical Imaging
• Biometry

Background:

• Accurate axial length (AL) measurement is crucial for ophthalmic diagnostics and surgical planning.
• Traditional biometry methods may have limitations; exploring novel data sources is essential.
• Optical coherence tomography angiography (OCTA) provides detailed retinal vasculature and structural information.

Purpose of the Study:

• To investigate the feasibility of utilizing metadata derived from OCTA imaging for axial length (AL) estimation.
• To develop and validate a predictive model for AL based on OCTA metadata.

Main Methods:

• Retrospective analysis of 198 eyes from 130 patients with available OCTA and biometry data (2019-2024).
• Univariate and multivariate regression analyses were employed to build the AL estimation model.
• Key OCTA metadata parameters (y_chinrest, z_chinrest, z_lens, z_motor) were identified as predictors.

Main Results:

• A multivariate regression model was established: Axial_Length = 24.79611 - 0.25510 × y_chinrest - 0.23443 × z_chinrest - 1.41824 × z_lens + 0.71947 × z_motor.
• The model achieved a good fit (R-squared = 0.5586), with 68.7% of predictions within 1 mm and 92.9% within 2 mm of true AL.
• Incorporating estimated AL into foveal avascular zone (FAZ) analysis reduced mean absolute error by 31%.

Conclusions:

• Metadata from OCTA imaging presents a viable and reliable method for estimating axial length (AL).
• This approach offers a potential non-invasive technique for AL assessment in clinical practice.
• The developed model demonstrates the utility of OCTA data beyond its primary diagnostic capabilities.