Abstract: Systems and methods for efficient implementation of a convolutional layer of a convolutional neural network are disclosed. In one aspect, weight values of kernels in a kernel stack of a convolutional layer can be reordered into a tile layout with tiles of runnels. Pixel values of input activation maps of the convolutional layer can be reordered into an interleaved layout comprising a plurality of clusters of input activation map pixels. The output activation maps can be determined using the clusters of the input activation map pixels and kernels tile by tile.

Abstract: A computer-implemented method, system, and instructions can include receiving a reconstructed volume of an object; determining an air density and a material density from the reconstructed volume of the object; subdividing the reconstructed volume into one or more blocks, each block comprising one or more voxels; determining one or more contrast blocks within the reconstructed volume; and determining a reconstruction quality based on the number of contrast blocks in the reconstructed volume.

Abstract: A method for extracting data from lineless tables includes storing an image including a table in a memory. A processor operably coupled to the memory identifies a plurality of text-based characters in the image, and defines multiple bounding boxes based on the characters. Each of the bounding boxes is uniquely associated with at least one of the text-based characters. A graph including multiple nodes and multiple edges is generated based on the bounding boxes, using a graph construction algorithm. At least one of the edges is identified for removal from the graph, and removed from the graph to produce a reduced graph. The reduced graph can be sent to a neural network to predict row labels and column labels for the table.

Abstract: A method and apparatus for processing image data is provided. The method includes the steps of employing a main processing network for classifying one or more features of the image data, employing a monitor processing network for determining one or more confusing classifications of the image data, and spawning a specialist processing network to process image data associated with the one or more confusing classifications.

Abstract: Systems and methods for evaluating ND are described herein. An example method can include constructing a non-sequential (NSC) ray-tracing model of an eye with an ophthalmic lens, and modelling a light source and a detector. The detector can be configured to mimic a retina of the eye. The method can also include computing irradiance data using the light source, the NSC ray-tracing model, and the detector. Irradiance data can be computed for each of a plurality of pupil sizes. The method can further include evaluating ND by analyzing the respective irradiance data for each of the pupil sizes. Also described herein are methods for designing an ophthalmic lens edge that reduces the incidence of ND for a given ophthalmic lens by adjusting the edge thickness and/or the scatter.

Abstract: Some implementations of the disclosure involve, at a device having one or more processors, one or more image sensors, and an illumination source, detecting a first attribute of an eye based on pixel differences associated with different wavelengths of light in a first image of the eye. These implementations next determine a first location associated with the first attribute in a three dimensional (3D) coordinate system based on depth information from a depth sensor. Various implementations detect a second attribute of the eye based on a glint resulting from light of the illumination source reflecting off a cornea of the eye. These implementations next determine a second location associated with the second attribute in the 3D coordinate system based on the depth information from the depth sensor, and determine a gaze direction in the 3D coordinate system based on the first location and the second location.

Abstract: A method and system using face tracking and object tracking is disclosed. The method and system use face tracking, location, and/or recognition to enhance object tracking, and use object tracking and/or location to enhance face tracking.

Abstract: A vertebra localization and identification method includes: receiving one or more images of vertebrae of a spine; applying a machine learning model on the one or more images to generate three-dimensional (3-D) vertebra activation maps of detected vertebra centers; performing a spine rectification process on the 3-D vertebra activation maps to convert each 3-D vertebra activation map into a corresponding one-dimensional (1-D) vertebra activation signal; performing an anatomically-constrained optimization process on each 1-D vertebra activation signal to localize and identify each vertebra center in the one or more images; and outputting the one or more images, wherein on each of the one or more outputted images, a location and an identification of each vertebra center are specified.

Abstract: With a diagnostic imaging support apparatus, a diagnostic imaging support method, and a diagnostic imaging support program, an optimum image for extracting a contour can be automatically selected from a superimposed image obtained by superimposing a plurality of images of different types. A diagnostic imaging support apparatus 1 includes: an accepting unit 22 that accepts a specification of the position of a predetermined defined region R on a superimposed image G obtained by superimposing a plurality of images of different types including a target image G0 that is a target on which a contour is created; a selection unit 23 that selects an image for extracting a contour on the basis of image information about regions R0, R1, and R2, in the plurality of images of different types, each corresponding to the accepted defined region R; and a contour extraction unit 24 that extracts the contour from the selected image.

Abstract: A method for segmenting images is provided including tessellating an image obtained from one of an image database and an imaging system into a plurality of sectors; classifying each of the plurality of sectors by applying one or more pre-defined labels to each of the plurality of sectors, wherein the pre-defined labels indicate at least one of an image quality metric (IQM) and a metric of structure; assigning each of the plurality of classified sectors an Image Quality Classification (IQC); identifying anchor sectors among the plurality of classified sectors, applying filtering and edge detection to identify target boundaries; applying contouring across contiguous sectors and using the assigned IQC as a guide to complete segmentation of an edge between any two identified anchor sectors; and smoothing across segmented regions to increase parametric second-order continuity.

Abstract: A method for providing a foraminal stenosis ratio using 3-dimensional CT includes (a) transmitting a spine image of a patient to an information extracting unit by an image capturing unit in response to an input signal transmitted from an input unit; (b) extracting spine boundary information and neural foramen area information based on a pixel value of the spine image by the information extracting unit; (c) storing the spine boundary information and the neural foramen area information by an information storing unit; (d) calculating the foraminal stenosis ratio by using the spine image, the spine boundary information, and the neural foramen area information by an information calculating unit; and (e) outputting maximum neural foramen area information of the neural foramen area information, a neural foramen angle for the maximum neural foramen area information, and the foraminal stenosis ratio by an output unit.

Abstract: There is provided a computer implemented method of automatic segmentation of three dimensional (3D) anatomical region of interest(s) (ROI) that includes predefined anatomical structure(s) of a target individual, comprising: receiving 3D images of a target individual, each including the predefined anatomical structure(s), each 3D image is based on a different respective imaging modality. In one implementation, each respective 3D image is inputted into a respective processing component of a multi-modal neural network, wherein each processing component independently computes a respective intermediate, and the intermediate outputs are inputted into a common last convolutional layer(s) for computing the indication of segmented 3D ROI(s). In another implementation, each respective 3D image is inputted into a respective encoding-contracting component a multi-modal neural network, wherein each encoding-contracting component independently computes a respective intermediate output.

Abstract: Systems, methods, and devices for detecting brake lights are disclosed herein. A system includes a mode component, a vehicle region component, and a classification component. The mode component is configured to select a night mode or day mode based on a pixel brightness in an image frame. The vehicle region component is configured to detect a region corresponding to a vehicle based on data from a range sensor when in a night mode or based on camera image data when in the day mode. The classification component is configured to classify a brake light of the vehicle as on or off based on image data in the region corresponding to the vehicle.

Abstract: Various approaches are disclosed to temporally and spatially filter noisy image data—generated using one or more ray-tracing effects—in a graphically rendered image. Rather than fully sampling data values using spatial filters, the data values may be sparsely sampled using filter taps within the spatial filters. To account for the sparse sampling, locations of filter taps may be jittered spatially and/or temporally. For filtering efficiency, a size of a spatial filter may be reduced when historical data values are used to temporally filter pixels. Further, data values filtered using a temporal filter may be clamped to avoid ghosting. For further filtering efficiency, a spatial filter may be applied as a separable filter in which the filtering for a filter direction may be performed over multiple iterations using reducing filter widths, decreasing the chance of visual artifacts when the spatial filter does not follow a true Gaussian distribution.

Abstract: In various examples, frames of a video may include a first visual object that may appear relative to a second visual object within a region of the frames. Once a relationship between the first visual object and the region is known, one or more operations may be performed on the relative region. For example, optical character recognition may be performed on the relative region where the relative region is known to contain textual information. As a result, the identification of the first visual object may serve as an anchor for determining the location of the relative region including the second visual object—thereby increasing accuracy and efficiency of the system while reducing run-time.

Abstract: An embodiment in accordance with the present invention provides a method for 3D-2D registration (for example, registration of a 3D CT image to a 2D radiograph) that permits deformable motion between structures defined in the 3D image based on a series of locally rigid transformations. This invention utilizes predefined annotations in 3D images (e.g., the location of anatomical features of interest) to perform multiple locally rigid registrations that yield improved accuracy in aligning structures that have undergone deformation between the acquisition of the 3D and 2D images (e.g., a preoperative CT compared to an intraoperative radiograph). The 3D image is divided into subregions that are masked according to the annotations, and the registration is computed simultaneously for each divided region by incorporating a volumetric masking method within the 3D-2D registration process.

Abstract: A device, apparatus, and method provide logic for processing information. In one implementation, a device may include an image acquisition unit configured to acquire an image, and a transmission unit configured to transmit information associated with the image to an information processing apparatus, such as a server. The server may be associated with a first feature quantity dictionary. The device also may include a receiving unit configured to receive a second feature quantity dictionary from the server in response to the transmission. The second feature quantity dictionary may include less information than the first feature quantity dictionary, and the server may generate the second feature quantity dictionary based on the image information and the first feature quantity dictionary. The device may include an identification unit configured to identify an object within the image using the second feature quantity dictionary.

Abstract: Techniques are provided for generating enhanced image representations from original X-ray images using deep learning techniques. In one embodiment, a system is provided that includes a memory storing computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can include a reception component, an analysis component, and an artificial intelligence component. The analysis component analyzes the original X-ray image using an AI-based model with respect to a set of features of interest. The AI component generates a plurality of enhanced image representations. Each enhanced image representation highlights a subset of the features of interest and suppresses remaining features of interest in the set that are external to the subset.

Abstract: The present disclosure generally relates to processing visual data of a road surface that includes a vertical deviation with a lateral slope. In some embodiments, a system determines a path expected to be traversed by at least one wheel of the vehicle on a road surface. In some embodiments, a system determines, using at least two images captured by one or more cameras, a height of the road surface for at least one point along the path to be traversed by the wheel. In some embodiments, a system computes an indication of a lateral slope of the road surface at the at least one point along the path. In some embodiments, a system outputs, on a vehicle interface bus, an indication of the height of the point and an indication of the lateral slope at the at least one point along the path.

Abstract: A method by which a left atrial wall thickness measurement device for the heart measures the left atrial wall thickness of the heart, according to one embodiment of the present invention, can simply, quickly, and accurately measure the thickness of the left atrium of the heart with only a simple input of a user so as to also derive the optimum effect of a radiofrequency catheter ablation procedure, and measures the thickness of the left atrium of the heart by using, without a separate additional inspection, computed tomography, which is relatively inexpensive and is performed for most patients with arrhythmia, thereby minimizing the financial burden on a patient.