Abstract: Disclosed techniques for image processing three-dimensional image data include: obtaining three-dimensional image data representing contiguous slices parallel to a plane, constructing training data from the image data by, for each of a plurality of angles: rotating the image data in the plane to produce rotated image data, blurring the rotated image data in a dimension parallel to the plane to produce low resolution rotated image data, and introducing aliasing into the low resolution rotated image data in the dimension parallel to the plane to produce aliased low resolution rotated image data, training an anti-aliasing neural network with the aliased low resolution image data and the low resolution image data, training a super-resolution neural network with the aliased low resolution image data and the rotated image data, and processing the image data using the trained anti-aliasing neural network and the trained super-resolution neural network to produce processed image data.

Abstract: Techniques for retinal layer segmentation are presented. The techniques include obtaining current optical coherence tomography (OCT) data for a retina; generating an estimated current en face image based on the current OCT data and an estimated current retinal layer segmentation; determining a registered previous retinal layer segmentation based on a previous retinal layer segmentation, a previous en face image, the estimated current en face image, and the estimated current retinal layer segmentation; updating the estimated current retinal layer segmentation using a deep neural network and based on the current OCT data and the registered previous retinal layer segmentation; repeating the generating, the determining, and the updating to obtain a current retinal layer segmentation as the estimated current retinal layer segmentation; and outputting a property of the retina determined at least in part from the current retinal layer segmentation.

Abstract: Disclosed techniques for image processing three-dimensional image data include: obtaining three-dimensional image data representing contiguous slices parallel to a plane, constructing training data from the image data by, for each of a plurality of angles: rotating the image data in the plane to produce rotated image data, blurring the rotated image data in a dimension parallel to the plane to produce low resolution rotated image data, and introducing aliasing into the low resolution rotated image data in the dimension parallel to the plane to produce aliased low resolution rotated image data, training an anti-aliasing neural network with the aliased low resolution image data and the low resolution image data, training a super-resolution neural network with the aliased low resolution image data and the rotated image data, and processing the image data using the trained anti-aliasing neural network and the trained super-resolution neural network to produce processed image data.

Abstract: A device receives a two-dimensional (2-D) image that depicts a cross-sectional view of a macula comprised of layers and boundaries to segment the layers, and determines spatial coordinates of the 2-D image that include x-coordinates and y-coordinates. The device uses a data model, that has been trained using a deep learning technique, to process the 2-D image and the spatial coordinates to generate boundary maps that indicate likelihoods of voxels of the 2-D image being in positions that are part of particular boundaries. The device determines, by analyzing the boundary maps, an initial set of boundary positions, and determines a final set of boundary positions by using a topological order identification technique to refine the initial set of boundary positions. The device determines the thickness levels of the layers of the macula based on the final set of boundary positions, and performs one or more actions based on the thickness levels.

Abstract: Disclosed techniques for image processing three-dimensional image data include: obtaining three-dimensional image data representing contiguous slices parallel to a plane, constructing training data from the image data by, for each of a plurality of angles: rotating the image data in the plane to produce rotated image data, blurring the rotated image data in a dimension parallel to the plane to produce low resolution rotated image data, and introducing aliasing into the low resolution rotated image data in the dimension parallel to the plane to produce aliased low resolution rotated image data, training an anti-aliasing neural network with the aliased low resolution image data and the low resolution image data, training a super-resolution neural network with the aliased low resolution image data and the rotated image data, and processing the image data using the trained anti-aliasing neural network and the trained super-resolution neural network to produce processed image data.

Abstract: A device receives a two-dimensional (2-D) image that depicts a cross-sectional view of a retina that includes a macula comprised of layers and boundaries used to segment the layers. The device converts the 2-D image to a standardized format, determines features for voxels included in the 2-D image, and generates, by using a data model to process the features, probability maps that indicate likelihoods of the voxels being in positions within particular boundaries. The device analyzes the probability maps to determine an initial set of boundary positions and to generate directional vectors that point in directions based on values included in the set of probability maps, determines a final set of boundary positions by performing a layer boundary evolution technique using the directional vectors to refine the initial set of boundary positions, and provides data that identifies the final set of boundary positions for display via an interface.

Abstract: According to one or more of the embodiments herein, a subject image of biological tissue is acquired from a pulse sequence of a magnetic resonance imaging (MRI) device, and one or more pulse sequence parameters used to acquire the subject image may be estimated based on a relationship between the subject image and the biological tissue. A new atlas image may then be synthesized using the pulse sequence and the estimated pulse sequence parameters of the subject image, and an intensity transformation between the new atlas image and a desired reference atlas image may be learned. As such, a desired subject image may be synthesized by applying the intensity transformation to the subject image.