Abstract

Dynamic-contrast optical coherence tomography (OCT) using en face digital holography senses intracellular dynamics in living tumor tissue. Intracellular motions produce ultra-low-frequency Doppler shifts for speeds down to nanometers per second (10 mHz) and up to microns per second (10 Hz). Cancer drugs applied to human tumor biopsies induce changes in these dynamics and produce specific Doppler signatures of therapeutic efficacy. We have developed a new type of deep neural network that performs as a differential autoencoder with high common-mode rejection that isolates Doppler signatures associated with drug response and patient outcomes. The differential autoencoder is applied to Doppler signals from a clinical trial of esophageal cancer patients.

© 2022 The Author(s)