Uncertainty-refined image segmentation under domain shift
| DWPI Title: Method for realizing digital image segmentation in e.g. medicine field, involves replacing segmentation label of image element with uncertainty value above predefined threshold with new-segmentation label corresponding to segmentation class |
| Abstract: Digital image segmentation is provided. The method comprises training a neural network for image segmentation with a labeled training dataset from a first domain, wherein a subset of nodes in the neural net are dropped out during training. The neural network receives image data from a second, different domain. A vector of N values that sum to 1 is calculated for each image element, wherein each value represents an image segmentation class. A label is assigned to each image element according to the class with the highest value in the vector. Multiple inferences are performed with active dropout layers for each image element, and an uncertainty value is generated for each image element. Uncertainty is resolved according to expected characteristics. The label of any image element with an uncertainty above a threshold is replaced with a new label corresponding to a segmentation class based on domain knowledge. |
| Use: Method for realizing digital image segmentation in medicine field, manufacturing field and materials science field. |
| Advantage: The method enables utilizing a cloud computing as a model of service delivery for enabling convenient, on-demand network access to a shared pool of configurable computing resources that can be rapidly provisioned and released with minimal management effort or interaction with a provider of the service. The method enables allowing a cloud consumer to unilaterally provision computing capabilities automatically without requiring human interaction with the service's provider. The method enables allowing cloud computing environment to offer infrastructure, platforms and/or software as services for cloud consumer not need to maintain resources on a local computing device, so that capabilities are rapidly and elastically provisioned to quickly scale out and quickly scale in. |
| Novelty: The method involves training a neural network for image segmentation with a labeled training dataset from a first domain. Image data is received by the neural network from a second different domain, where image data comprises a number of image elements. A vector of values for the image element is calculated by the neural network. A segmentation label is assigned by the neural network to the image element. Multiple inferences for the image element are performed by the neural network with active dropout layers. An uncertainty value for the image element is generated by the neural network according to the inferences. Uncertainty is resolved according to expected characteristics based on domain knowledge. The segmentation label of the image element is replaced with an uncertainty value above a predefined threshold with a new segmentation label corresponding to a segmentation class according to the domain knowledge for that image element. |
| Filed: 6/3/2022 |
| Application Number: US17832477A |
| Tech ID: SD 15006.1 |
| This invention was made with Government support under Contract No. DE-NA0003525 awarded by the United States Department of Energy/National Nuclear Security Administration. The Government has certain rights in the invention. |
| Data from Derwent World Patents Index, provided by Clarivate All rights reserved. Republication or redistribution of Clarivate content, including by framing or similar means, is prohibited without the prior written consent of Clarivate. Clarivate and its logo, as well as all other trademarks used herein are trademarks of their respective owners and used under license. |