Los Alamos Tech Takes Imagery Analysis to the Next Level

LosAlamos Imagery

Los Alamos National Laboratory’s (LANL) GeniePro software is a general purpose, interactive, adaptive tool for automatically finding and labeling areas of interest in image data.

Rather than having to teach the program how to find areas of interest, the user has merely to provide examples of these areas.

GeniePro then designs, tests, and perfects its algorithms to produce consistent and accurate image recognition.

The software uses techniques from statistical machine learning theory and evolutionary computation theory to perform robust and customized automatic feature extraction in multispectral, hyperspectral, panchromatic, and multi-instrument fused imagery.

Genie was first used during the Cerro Grande fire to track the path of the wildfire for the Burned Area Emergency Rehabilitation team.

It was used again just after the September 11 terrorist attacks in New York City to track the cloud of ash and soot from the collapse of the World Trade Center towers.

Originally developed to analyze multispectral satellite data, GeniePro works with satellite data, aerial imagery, and medical images.

It has been applied to a variety

of tasks, including crop and terrain mapping; road, railroad, river, and communications network mapping; broad-area search for vehicles and buildings; and cancer cell identification in histological images.

The GeniePro system incorporates many tools for manually optimizing results that can be easily exported in a variety of raster or vector formats for use by other software packages, such as commonly used image formats (NITF, GeoTIFF, ERDAS HFA, etc.), and it runs on both Windows and Linux systems.

GeniePro represents a significant technical advancement beyond the simple image classification algorithms found in popular geographic information system (GIS) packages.

For example, GeniePro is able to distinguish between features that share similar spectral characteristics (asphalt roads vs. asphalt roofs) because it processes the spatial context of each pixel, enabling it to produce precise results for tasks that would defeat traditional classification algorithms.

GeniePro, the most recent version of the software, was transferred to two firms through exclusive, fieldof-use licenses.

Observera, Inc., licensed the technology from LANL for remote sensing and geographic information system applications. Observera’s release of a commercial version of the software, GeniePro 2.0, marks the beginning of a line of new geospatial products.

Aperio Technologies, Inc., a global leader in digital pathology for the healthcare and life sciences industry, obtained an exclusive license to GeniePro for all digital pathology applications and has incorporated the code into its Spectrum digital pathology information management software, enabling it to be used as a pre-processing engine for various tissue-scoring algorithms such as finding tumor regions in digital immunohistochemistry slides.

In addition, whereas traditional analysis techniques require considerable hands-on time by the investigator to manually identify regions for analysis, the GeniePro software allows entire studies to be analyzed objectively and consistently without manual intervention.

Genie Pro may be applied to oncology, ophthalmology, diabetes, neurology, hematology, nephrology, cardiology, toxicology, and microbiology.