Application of Deep Learning to Strain Retrieval in Optical Coherence Elastography

Optical coherence elastography (OCE) is an image formation tool used to retrieve the mechanical properties of biological tissues. OCE images are formed from the acquisition of two optical coherence tomography (OCT) images of the sample subjected to different states of mechanical loading. Conventional models for strain retrieval in OCE estimate the strain from the phase difference between the two OCT B-scans, the sensitivity of which are limited by approximations and phase wrapping. Furthermore, their performance decreases as strain increases. We present a deep learning model for OCE which overcomes these problems, and give examples to compare it with existing methods.