The Johns Hopkins University

Other

Jerry L Prince

Nael Fakhry Osman

Aaron Carass

Matthias Stuber

Khaled Zakarya Abd-Elmoniem

Can Zhao

Yufan He

Russell H Taylor

Everette C Burdette

Emad M Boctor

Gabor Fichtinger

Michael Choti

Junghoon Lee

Christian Labat

Yihao Liu

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.

Segmenting retinal OCT images with inter-B-scan and longitudinal information

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.

Machine learning processing of contiguous slice image data

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.

Neural-network-driven topology for optical coherence tomography segmentation

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.

Layer boundary evolution for macular optical coherence tomography segmentation

A method of processing image data from an imaging system for locating a plurality N of objects embedded in a body includes receiving data for a first two-dimensional image of a region of interest of the body containing the plurality N of objects, the first two-dimensional image being obtained from a first imaging setting of the imaging system relative to the region of interest; receiving data for a second two-dimensional image of a region of interest of the body containing the plurality N of objects, the second two-dimensional image being obtained from a second imaging setting of the imaging system relative to the region of interest; and receiving data for a third two-dimensional image of a region of interest of the body containing the plurality N of objects, the third two-dimensional image being obtained from a third imaging setting of said imaging system relative to said region of interest.

Methods and systems for locating objects embedded in a body

A robotic 5D ultrasound system and method, for use in a computer integrated surgical system, wherein 3D ultrasonic image data is integrated over time with strain (i.e., elasticity) image data. By integrating the ultrasound image data and the strain image data, the present invention is capable of accurately identifying a target tissue in surrounding tissue; segmenting, monitoring and tracking the target tissue during the surgical procedure; and facilitating proper planning and execution of the surgical procedure, even where the surgical environment is noisy and the target tissue is isoechoic.

Robotic 5-dimensional ultrasound

Three-dimensional MR motion estimation on a single image plane based on tagged MRI and HARP processing. Tagged magnetic resonance imaging technique encodes and automatically tracks displacement of spatially modulated object in three dimensions, encoding both in plane and through-plane motion in a single image plane without affecting acquisition speed. Post-processing unravels encoding in order to directly track 3-D displacement of points within the image plane throughout image sequence. The invention is particularly suited to use on a heart for tracking and determining myocardial displacement. In one embodiment, an MR pulse sequence extends a slice following complementary spatial modulation of magnetization (CSPAMM) pulse sequence with two small z-encoding gradients immediately before the readouts in successive CSPAMM acquisitions, thereby adding a through-plane encoding from which through-plane motion can be computed from acquired images. HARP processing is used to determine in-plane motion, after which through-plane position can be determined using phase encodings. Use of balanced encodings and horizontal and vertical tags permits cancellation of systematic phase artifacts present in CSPAMM acquisitions.

Three dimensional magnetic resonance motion estimation on a single image plane

The present invention relates to a method of measuring motion of an object such as a heart by magnetic resonance imaging. A pulse sequence is applied to spatially modulate a region of interest of the object and at least one first spectral peak is acquired from the Fourier domain of the spatially modulated object. The inverse Fourier transform information of the acquired first spectral-peaks is computed and a computed first harmonic phase image is determined from each spectral peak. The process is repeated to create a second harmonic phase image from each second spectral peak and the strain is determined from the first and second harmonic phase images. In a preferred embodiment, the method is employed to determine strain within the myocardium and to determine change in position of a point at two different times which may result in an increased distance or reduced distance. The method may be employed to determine the path of motion of a point through a sequence of tag images depicting movement of the heart. The method may be employed to determine circumferential strain and radial strain.

Cardiac motion tracking using cine harmonic phase (HARP) magnetic resonance imaging

The present invention provides methods for real-time measurement of motion of an object such as a portion of a patient in real-time through the use of harmonic phase (HARP) magnetic resonance imaging. This is accomplished by employing certain tagging protocols and imaging protocols. The imaging may be accomplished in two-dimension or three-dimension. In one embodiment, first and second tag pulse sequences are employed to provide two-dimensional pulse strain images. In another embodiment, a first tag pulse sequence is employed to determine a first harmonic phase image and a second tag pulse sequence is employed to determine a second harmonic phase image which is combined with the first image to create tagged images of circumferential and radial strains with third and fourth tag pulse sequences being employed to create images which are combined to establish longitudinal strain and thereby provide a three-dimensional strain image.

Method for harmonic phase magnetic resonance imaging

A method of measuring motion of an object by magnetic resonance imaging including applying a pulse sequence to spatially modulate a region of interest of said object. At least one spectral peak is acquired from the Fourier domain of the spatially modulated object. The inverse Fourier transform information of the acquired spectral peaks is computed. The angle images are computed from the spectral peak. The angle images employed to measure motion of the object. The method may employ a SPAMM pulse sequence as the pulse sequence. The angle images may be employed to compute directly and automatically, planar strain in two dimensions or a full strain tensor in three dimension. The data may be useful in detection and quantification of myocardial ischemia and infarction. The angle images may also be employed to generate data equivalent to planar tag data automatically and can be employed to generate any desired tag separations. The angle images may also be employed to compute displacement, synthesize tag patterns and compute optical flow without requiring the use of regularization.

Growing community of inventors

Lutherville, MD, United States of America

Jerry L Prince