Abstract: Systems and methods for determining the health status of a retina of a subject. An optical coherence tomography (OCT) volume image of a retina of a subject may be received. A health indication output is generated, via a deep learning model, using the OCT volume image. The health indication output indicates a level of association between the OCT volume image and a selected health status category for the retina. A map output for the deep learning model is generated using a saliency mapping algorithm, generating a map output for the deep learning model using a saliency mapping algorithm. The map output indicates a level of contribution of a set of regions in the OCT volume image to the health indication output generated by the deep learning model.

Abstract: A method and system for detecting nascent geographic atrophy. An optical coherence tomography (OCT) volume image of a retina of a subject is received. Using a deep learning model, an output is generated using the OCT volume image in which the output indicates whether nascent geographic atrophy is detected. A map output is generated for the deep learning model using a saliency mapping algorithm, wherein the map output indicates a level of contribution of a set of regions in the OCT volume image to the output generated by the deep learning model.

Abstract: Systems and methods for determining the health status of a retina of a subject. An optical coherence tomography (OCT) volume image of a retina of a subject may be received. A health indication output is generated, via a deep learning model, using the OCT volume image. The health indication output indicates a level of association between the OCT volume image and a selected health status category for the retina. A map output for the deep learning model is generated using a saliency mapping algorithm, generating a map output for the deep learning model using a saliency mapping algorithm. The map output indicates a level of contribution of a set of regions in the OCT volume image to the health indication output generated by the deep learning model.

Abstract: A method for denoising magnetic resonance images and data is disclosed herein. An example method includes receiving a series of MRF images from a scanning device; identifying one or more subsets of voxels for the series of MRF images; generating one or more sets of eigenvectors, each set of the one or more sets of eigenvectors corresponding to one of the one or more subsets of voxels, and each eigenvector of the one or more sets of eigenvectors having a corresponding eigenvalue; applying a noise distribution model to each of the eigenvalues; identifying a subset of the eigenvalues as corresponding to noise based on the noise distribution model; and reconstructing the series of MRF images without the subset of eigenvalues identified as corresponding to noise.

Abstract: A system for automatically analyzing a video recording of a colonoscopy includes a processor and memory storing instructions, which when executed by the processor, cause the processor to receive the video recording of the colonoscopy performed on the colon and detect informative frames in the video recording. A frame is informative if the clarity of the frame is above a threshold or if the frame includes clinically relevant information about the colon. The instructions cause the processor to generate scores indicating severity levels of a disease for a plurality of the informative frames, estimate locations of the plurality of the informative frames in the colon, and generate an output indicating a distribution of the scores over one or more segments of the colon by combining the scores generated for the plurality of the informative frames and the estimated locations of the plurality of the informative frames in the colon.

Abstract: An automated pruning technique is proposed for reducing the size of a convolutional neural network. A large-sized network is trained and then connections between layers are explored to remove redundant parameters. Specifically, a scaling neural subnetwork is connected to the neural network and designed to infer importance of the filters in the neural network during training of the neural network. Output from the scaling neural subnetwork can then be used to remove filters from the neural network, thereby reducing the size of the convolutional neural network.

Abstract: A system for automatically analyzing a video recording of a colonoscopy includes a processor and memory storing instructions, which when executed by the processor, cause the processor to receive the video recording of the colonoscopy performed on the colon and detect informative frames in the video recording. A frame is informative if the clarity of the frame is above a threshold or if the frame includes clinically relevant information about the colon. The instructions cause the processor to generate scores indicating severity levels of a disease for a plurality of the informative frames, estimate locations of the plurality of the informative frames in the colon, and generate an output indicating a distribution of the scores over one or more segments of the colon by combining the scores generated for the plurality of the informative frames and the estimated locations of the plurality of the informative frames in the colon.

Abstract: An automated pruning technique is proposed for reducing the size of a convolutional neural network. A large-sized network is trained and then connections between layers are explored to remove redundant parameters. Specifically, a scaling neural subnetwork is connected to the neural network and designed to infer importance of the filters in the neural network during training of the neural network. Output from the scaling neural subnetwork can then be used to remove filters from the neural network, thereby reducing the size of the convolutional neural network.