Optical coherence tomography and deep learning for automated oocyte detection in opaque follicular fluid
Alejandro Chavez-Badiola, Daniel Farfán, Vladimir Ocegueda-Hernández, Giuseppe Silvestri, Víctor-Manuel Rico-Botero, Yessenia Flores-Esparza, Edman Mendoza-Leyva, Celina Mendez-Zavala, Gerardo Mendizabal-RuizIn in vitro fertilization (IVF), oocytes enclosed in cumulus–oocyte complexes (COCs) must be found in opaque follicular fluid, where transmitted brightfield microscopy (TBM) can miss them. We evaluated whether spectral-domain optical coherence tomography (OCT; 900 nm central wavelength, 3.0 µ m axial resolution in air) combined with a real-time detection transformer (RT-DETR) detects oocyte presence within intact bovine COCs. On an independent test set of 141 wells, with post-denudation confirmation as the reference standard, six embryologists assessing static TBM images detected 65.6–87.0% of oocytes; OCT–RT-DETR detected 99.2% (95% CI 95.8–100%), with 90.0% specificity. OCT-based detection could provide an operator-independent input to automated oocyte search.