US Patent Application 17804224. GENERATING NEURAL NETWORKS TAILORED TO OPTIMIZE SPECIFIC MEDICAL IMAGE PROPERTIES USING NOVEL LOSS FUNCTIONS simplified abstract

GENERATING NEURAL NETWORKS TAILORED TO OPTIMIZE SPECIFIC MEDICAL IMAGE PROPERTIES USING NOVEL LOSS FUNCTIONS

Organization Name

GE Precision Healthcare LLC

Inventor(s)

Obaidullah Rahman of South Bend IN (US)

Madhuri Mahendra Nagare of Karmala IN (US)

Roman Melnyk of New Berlin WI (US)

Jie Tang of Merion Station PA (US)

Brian E. Nett of Wauwatosa WI (US)

Charles Addison Bouman of West Lafayette IN (US)

Ken Sauer of South Bend IN (US)

GENERATING NEURAL NETWORKS TAILORED TO OPTIMIZE SPECIFIC MEDICAL IMAGE PROPERTIES USING NOVEL LOSS FUNCTIONS - A simplified explanation of the abstract

This abstract first appeared for US patent application 17804224 titled 'GENERATING NEURAL NETWORKS TAILORED TO OPTIMIZE SPECIFIC MEDICAL IMAGE PROPERTIES USING NOVEL LOSS FUNCTIONS

Simplified Explanation

- This patent application describes techniques for generating neural networks (NNs) that are specifically designed to optimize certain properties of medical images. - The system includes a memory and a processor, which work together to execute computer executable components. - One of these components is a training component, which trains a NN to generate a modified version of computed tomography (CT) data with optimized properties. - The training process uses a novel loss function that is tailored to control the learning adaptation of the NN based on the error associated with specific components of the CT data. - These defined components can include frequency components or spatial feature components. - The result of the training process is a trained NN that can generate modified CT data with improved properties.

Original Abstract Submitted

Techniques are described that facilitate generating neural network (NNs) tailored to optimize specific properties of medical images using novel loss functions. According to an embodiment, a system is provided that comprises a memory that stores computer executable components, and a processor that executes the computer executable components stored in the memory. The computer executable components comprise a training component that trains a NN to generate a modified version of computed tomography (CT) data comprising one or more optimized properties relative to the CT data using a loss function tailored to control learning adaptation of the NN based on error attributed to one or more defined components associated with the CT data, resulting in a trained NN, wherein the one or more defined components comprise at least one of a frequency component or a spatial feature component.