Estimating Nonplanar Flow from 2D Motion-blurred Widefield Microscopy Images via Deep Learning

• Published in: arXiv.org
• Main Authors: Shajkofci, Adrian, Liebling, Michael
• Format: Paper
• Language: English
• Published: Ithaca Cornell University Library, arXiv.org 14.02.2021
• Subjects: Artificial neural networks; Blurring; Deep learning; Fields (mathematics); Flow measurement; Ground truth; High speed cameras; Luminous intensity; Microscopy; Optical flow (image analysis); Organelles; Organs; Parameter estimation; Regression coefficients; Three dimensional motion; Two dimensional flow; Velocity

Optical flow is a method aimed at predicting the movement velocity of any pixel in the image and is used in medicine and biology to estimate flow of particles in organs or organelles. However, a precise optical flow measurement requires images taken at high speed and low exposure time, which induces phototoxicity due to the increase in illumination power. We are looking here to estimate the three-dimensional movement vector field of moving out-of-plane particles using normal light conditions and a standard microscope camera. We present a method to predict, from a single textured wide-field microscopy image, the movement of out-of-plane particles using the local characteristics of the motion blur. We estimated the velocity vector field from the local estimation of the blur model parameters using an deep neural network and achieved a prediction with a regression coefficient of 0.92 between the ground truth simulated vector field and the output of the network. This method could enable microscopists to gain insights about the dynamic properties of samples without the need for high-speed cameras or high-intensity light exposure.