Multi-prong multitask convolutional neural network for biomedical image inference
AssigneeThe Regents of the University of California
InventorsAlbert Hsiao, Evan Masutani
Life of this patent
Estimating the term…
Abstract
A neural network architecture and method for analysis of time series images from an image source employs a 3D-UNet convolutional neural network (CNN) configured to receive the time series images and generate spatiotemporal feature maps therefrom. Multiple sub-convolutional neural network output prongs based on an SRNet architecture receive the feature maps and simultaneously generate inferences for image segmentation, regression of values, and multi-landmark localization.
Claims18
A method for analysis of time series images from an image source, the method comprising:
- inputting the time series images into a computer processor configured to execute a 3D-UNet convolutional neural network (CNN) and multiple sub-convolutional neural network output prongs based on an SRNet architecture, wherein the computer processor:
- generates spatiotemporal feature maps from the time series images;
- and
- receives and distributes the feature maps to the multiple output prongs to simultaneously and in parallel generate a plurality of different inferences including image segmentation, regression of values, and multi-landmark localization.
A method for analysis of time series images, the method comprising:
- receiving the time series images in a computer processor configured for executing a trained 3D-UNet convolutional neural network (CNN) and generating spatiotemporal feature maps therefrom;
- and
- distributing the feature maps to multiple sub-convolutional neural network output prongs based on an SRNet architecture for simultaneously and in parallel generating a plurality of different inferences comprising a first prong configured to generate inferences for image segmentation, a second prong configured to generates inferences for regression of values, and a third prong configured to generate inferences for multi-landmark localization.
Dependencies are inferred from the claim text.