Multi-Modal Registration of Embryonic Images for In Vitro Fertilization

This paper presents a robust and fully automated registration algorithm for alignment of embryonic images taken by optical quadrate microscopy (OQM) and differential interference contrast (DIC) microscopy. Registration of such images allows recognition of distinguishing features of viable embryos, such as growth rate, to reduce multiple births from in vitro fertilization. OQM is an optical interferometer whereas the DIC images gradient of the phase. Due to the very different nature of the image formation, the interior of the embryo appears substantially different if not totally uncorrelated between different modalities. We present an algorithm which extracts and aligns the silhouettes of the embryo-the most invariant features between different modalities. The algorithm is tested with 28 image pairs taken from different embryos at various development stages. The success rate is 85.7% if only affine transformation is adopted, and 96% if similarity transformation is allowed for embryos without sufficient outline complexity.