Random sampling for deep learning segmentation of acute ischemic stroke on non-contrast CT
Inventors
Ostmeier, Anna Sophie Gertrud • Axelrod, Brian M. • Heit, Jeremy J.
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Assignees
Leland Stanford Junior University
MemberStanford UniversityStanford UniversityTodd Coleman is an associate professor of bioengineering at Stanford University. His research combines applied probability, physiology, and bioelectronics to develop methods and devices for monitoring and modulating nervous-system and other physiological activity, including brain and visceral-organ signals.
Todd Coleman is an associate professor of bioengineering at Stanford University. His research combines applied probability, physiology, and bioelectronics to develop methods and devices for monitoring and modulating nervous-system and other physiological activity, including brain and visceral-organ signals.
Publication Number
US-12751682-B2
Publication Date
2026-10-06
Expiration Date
Abstract
A method is described for generating segmentation masks to assist in identification of acute ischemic stroke. The method includes performing by a non-contrast computed tomography scan to produce a computed tomography image; inputting the computed tomography image to an input layer of a deep learning neural network; and outputting a segmentation mask of acute ischemic stroke from an output layer of the deep learning neural network, wherein the segmentation mask of acute ischemic stroke is generated in response to the computed tomography image input to the deep learning neural network. The deep learning neural network is trained with ground truth non-contrast computed tomography images and corresponding segmentation masks of acute ischemic stroke, wherein multiple segmentation masks of acute ischemic stroke for each of the non-contrast computed tomography images are randomly sampled for training.
Core Innovation
The disclosed method generates a segmentation mask of acute ischemic stroke from a non-contrast computed tomography image using a deep learning neural network. The image is input to the network, which outputs the mask in response. The stated purpose is to assist in identification of acute ischemic stroke. The supplied content does not include the background section, so it does not explicitly describe a separate problem statement.
The network is trained using a set of ground truth non-contrast computed tomography images, with multiple corresponding segmentation masks for each image. One of those masks is randomly and uniformly sampled for training. The described modified nnUNet embodiment uses multiple stages, two 3D convolutions per stage, and leaky ReLU. The reported evaluation finds better key segmentation metrics than inter-expert agreement and majority-vote training, and reports correlation between predicted volumes and 90-day clinical outcomes.
Claims Coverage
The provided claims identify one independent claim. It covers generating an acute ischemic stroke segmentation mask from a non-contrast computed tomography image using a deep learning neural network trained with multiple corresponding masks and random, uniform sampling of one mask.
Random and uniform sampling of segmentation masks during training
The independent claim covers performing a non-contrast computed tomography scan, inputting the resulting image to a deep learning neural network, and outputting an acute ischemic stroke segmentation mask. The network is trained with ground truth images and corresponding masks, with multiple masks for each image and one mask randomly and uniformly sampled from those masks for training.
The independent claim centers on using randomly and uniformly sampled masks from multiple corresponding segmentation masks to train the network that generates an acute ischemic stroke segmentation mask.
Stated Advantages
The reported evaluation finds that random-expert-sampling training outperforms inter-expert agreement and majority-vote training on key segmentation metrics.
Predicted segmentation volumes correlate with 90-day clinical outcomes.
Documented Applications
Generating acute ischemic stroke segmentation masks from non-contrast computed tomography images to assist in identification of acute ischemic stroke.
Interested in licensing this patent?