James C. Tilton, a scientist at NASA’s Goddard Space Flight Center, for some years now has been working on new image enhancement software to help automatically analyze satellite data of the Earth. Hierarchical Segmentation Software (HSEG), as the tool is called, identifies relatively homogeneous areas of an image and highlights them.
Our eyes and brains are pretty good at image analysis, but large dense maps can be quite a challenge. Although originally designed for aerial cartography, the first commercial use of the software came in the form of a mammogram enhancement and analysis system.
The lakes of northern Wisconsin (top) are very much like dense breast tissue (side) to a NASA scientist it turns out, and porting over the code and optimizing it led to the MED-SEG™ system from Bartron Medical Imaging (New Haven, Connecticut). Having received FDA approval, plans are now being made to conduct clinical trials evaluating how MED-SEG can benefit radiologists in cancer diagnosis.
From NASA news:
Like a single piece of a jigsaw puzzle, a pixel often does not provide enough information about where it fits into the overall scene. To overcome the deficiency, Tilton focused on an approach called image segmentation, which organizes and groups an image’s pixels together at different levels of detail. But Tilton’s approach to image segmentation was different than other approaches in that it not only finds region objects, but also groups spatially separated region objects together into region classes.
For example, an Earth satellite image may contain several lakes of different depths. Deep lakes appear dark blue, while shallow lakes are a lighter shade of blue. HSEG first finds each individual lake; then it groups together all shallow lakes into one class and the deeper lakes into another. Because lakes are more similar than they are to trees, grass, roads, buildings, and other objects, the software then groups all lakes together, regardless of their varying colors. As a result, HSEG allows the user to distinguish important features in the scene accurately and quickly.
With the MED-SEG system, medical centers will be able to send images via a secure Internet connection to a Bartron data center for processing by the company’s imaging application. The data are then sent back to the medical center for use by medical personnel during diagnosis. Bartron has installed the system at the University of Connecticut Health Center, with the possibility of installing evaluation systems at New York University Medical Center, Yale-New Haven Medical Center, and the University of Maryland Medical Center.
Product page: Bartron MED-SEG
*This blog post was originally published at Medgadget*