Research > Software > LONI-ICE

LONI It's Close Enuf

LONI ICE is a Java application that generates seed points for other image processing applications. ICE stands for "It's Close Enuf" and implies that the seed points it generates are "close enough" to be useful.

Features

  • Learns how to find user-selected anatomical point landmarks in medical images.
  • Given examples of manually-determined points in a set of test images, LONI ICE can be trained to locate the points in other images.
  • No assumptions on structural or intensity characteristics, and no run-time parameters to tune. LONI ICE contains no implicit modeling and does not make assumptions about intensity distributions.
  • Images acquired from different protocols can be mixed together for training. LONI ICE can be trained to generate seed points in any predictable images for any image modality (MRI, CT, PET, CR, etc.).
  • The graphical user interface provides a visual environment for viewing image data, selecting training points, and training the algorithm.
  • Support for many common image file formats. LONI ICE provides file format conversions to support the ANALYZE, DICOM, GE, MINC, ECAT, HRRT Interfile, and NIFTI file formats.

Description

LONI ICE is a Java application that generates seed points for other image processing applications. ICE stands for "It\'s Close Enuf" and implies that the seed points it generates are "close enough" to be useful. Figuratively, it\'s similar to taking a thick magic marker, making a splotch on an image, and saying that a point of interest is somewhere inside that spot.

Usage

<br />

System Requirements

  • OS: Windows/Linux/Mac
  • Processor: any
  • Memory: 500 Mb or more
  • Other: Java 1.4.2 or higher

Installation

Installation Instructions
Unzip the download file into a directory of your choice
In this directory, type on the command line: java -Xmx800m -jar ice.jar

Purpose

Medical images of human body parts normally look quite similar. For example, MRI brain scans of adult subjects are commonly centered and contain similarly-shaped structures in predictable locations. Many image segmentation applications (e.g., active contours) are sensitive to how they are initialized; the structures they segment are dependent upon where in image space they are started from. LONI ICE allows users to train it to locate user-selected points in medical images within well-defined precisions. These seed points can then be used to initialize other image processing applications.