Abstract
Purpose :
Optical coherence tomography (OCT) may help differentiate anterior segment inflammation by characterizing blood cells found in the anterior chamber. Mie theory fitting is useful for estimating particle sizes in order to classify cells based on differences in size (e.g., smaller red blood cells compared to white blood cells (WBCs) and differentiation of WBC subtypes). Previous reports of Mie fitting of cells in the anterior chamber were limited by small field-of-view (FOV) in cell tracking and diameter-only optimization. We propose a method to address these limitations by utilizing adaptive scanning OCT and multi-variable Mie theory optimization.
Methods :
In preliminary experiments using cell phantoms, OCT volumes were taken of 10.08 um diameter 0.001% polystyrene microbeads in solution in a 5mm path length glass cuvette with FOV of 2.5 × 2.5 × 3.8 mm3 (C × B × A scan directions). Adaptive scanning was used which only updates areas containing new information in the next volume, allowing increases in OCT FOV while maintaining high sampling density (Fig 1A). We multiplied raw A-scan spectra with different Hamming windows centered at different wavelengths to calculate the short-term Fourier transform (STFT) (Fig 1B). Then we generated spectroscopic data based on the particle location voted by STFTs (Fig 1C). Individual tracked particles over multiple volumes were averaged to achieve higher SNR for Mie theory size estimation (Fig 1D). For multi-variable Mie theory optimization, we sampled 1000 random seeds of initial particle size, particle refractive index, and environmental refractive index as variables to optimize the backscattering at different wavelengths minimizing root mean square error (RMSE).
Results :
In a sample imaging session, 72 OCT volumes were captured, and 17 particles were captured and tracked in at least 15 sequential OCT volumes. Multi-variable Mie theory fitting for all three fit variables resulted in a best-fit sphere diameter of 9.10 ± 1.90 um, with RMSE 0.23 ± 0.03 (an example is shown in Fig. 2). Limiting the fit variable to particle size only, the Mie theory best-fit particle size was 8.74 ± 2.38 um, with RMSE 0.30 ± 0.02.
Conclusions :
With the adaptive scan algorithm and the multi-variable Mie theory optimization, our system estimated the sphere particle size with less than 1um error.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.