Novel 3D ultrasound imaging and automated deep learning analysis in ophthalmology

Abstract

We present novel 3D ultrasound technologies for imaging of the eye. We computationally align and improve imaging volumes acquired by mechanical scanning of conventional ophthalmic ultrasound systems. We also present results of using emerging ultrahigh frequency array systems that increase volumetric image reconstruction speed by several orders of magnitude and allowing Doppler processing of blood flow. The easily interpretable 3D visualizations can greatly assist in automated analysis and reducing operator dependency. We present deep learning methods for volumetric segmentation and quantification. The 3D ophthalmic ultrasound combined with automated deep learning analysis allows non-expert users to easily use this technology for diagnosis, monitoring response to therapy and improving overall ocular disease management at low cost, versatility, and safety for repeated and timely scanning.