Browse Prior Art Database

Method and System for Providing Hashing-Based Atlas Ranking and Selection for Improving Multi-Atlas Segmentation

IP.com Disclosure Number: IPCOM000252455D
Publication Date: 2018-Jan-12
Document File: 2 page(s) / 94K

Publishing Venue

The IP.com Prior Art Database

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

1

Method and System for Providing Hashing-Based Atlas Ranking and Selection for Improving Multi-Atlas Segmentation

Abstract

A method and system is disclosed for providing hashing-based atlas ranking and selection for improving multi-atlas segmentation.

Description

In existing anatomy segmentation mechanisms, training images (atlases) that are annotated with labels are obtained and deformable registration is performed between each of the atlases and a target image in order to perform multi-atlas segmentation. This process is computationally intensive. Atlas ranking and selection mechanisms may be a solution to the computational requirements of multi-atlas segmentation, however all existing atlas ranking and selection techniques are dependent on an affine registration as a pre-processing operation. Moreover, existing mechanisms rely on global registration of multiple atlas images with a target image. This degrades accuracy, increases computational time, and is not an optimum solution for organ-specific segmentation.

Disclosed is a method and system for providing hashing-based atlas ranking and selection for improving multi-atlas segmentation. The method and system utilizes a multi-atlas segmentation engine to train a hashing forest based model using image volumes and corresponding labels. The training may be performed at both an organ or individual anatomical structure level and a whole image level.

In accordance with the method and system, the multi-atlas segmentation engine trains a hashing forest model based on a set of training images and corresponding label annotations to thereby generate a trained hashing forest model. The trained hashing forest model is then us...