When unmanned or drone planes fly over land and collect billions of pieces of information, statistical modeling can help interpret the data to distinguish a tank from a residential home. Similar types of models are applicable to finding cancerous tissues in mammography, mapping galaxy star streams, medical imaging and even detecting landmines.

Ramani Pilla, assistant professor of statistics at Case Western Reserve University, is working on mixture distributions to model and analyze heterogeneous or highly variable data. Mixture models have been used to find clusters in multi-spectral images originating from such diverse sources as mammography, remote sensing and mine-field detection.

Pilla is one of the two statisticians nationwide to receive a five-year, $400,000 National Science Foundation Early Career Development Award to expand her ground-breaking research on mixture models and to integrate it into educational activities.

Her collaborators include scientists at the CWRU medical school, at the Naval Surface Warfare Center and in Italy and South Korea.

Pilla already has done work in this area, while a Fellow at the National Institutes of Health in 1997-2000. She also developed a special kind of random effects model to analyze complex survey data collected by the World Health Organization to measure bullying behaviors among youths in the United States. The research findings should aid in developing intervention programs to curb bullying behaviors.

"This statistical model was the first of its kind to be used in the fields of psychology and epidemiology," Pilla said.

The research findings were published in the Journal of American Medical Association and this article was featured in the national press including The New York Times, Los Angeles Times, CNN and NPR.

Much of her NSF-funded work involves using mixture models, a method used by statisticians for more than 100 years. Due to the explosion in the size and complexity of the data sets, current mixture modeling methods are hampered by inefficient estimation and limited inferential capabilities. With her NSF funds, Pilla said she plans to take "a three-pronged attack" to overcome these limitations and promises "new inferential tools for fitting mixture models."

"In the visual detection of cancer tumors and land mines, the underlying spatial process is a mixture of local processes-mixing over different neighborhoods," she added. "The idea behind the application of mixture models to spatial scan analysis is to contrast two or more regions in an image by comparing their relative levels of heterogeneity, which is captured by mixture distributions."

Her ultimate goal is to develop efficient pattern recognition algorithms for detecting tumors and locating land mines.

"One of my goals is to make these methods available to practitioners in other fields by simplifying the computations," she said. Pilla said that she also hopes to put some of the debates about mammography for younger women to rest.

"New imaging techniques coupled with appropriate statistical detection tools, such as the semi-parametric mixture models, might tilt the balance in favor of earlier and more accurate detection of abnormal tissues," she said.

In addition to the prestigious NSF award given to junior faculty members with creative projects, Pilla is the solo investigator on a $230,000 grant from the Office of Naval Research to develop mixture models to analyze the scanned information for use in automatic target recognition systems used by the U.S. Navy.

Pilla's work won the Best of the Journal of Computational and Graphical Statistics award. The editor of the journal has invited her to make a presentation on her work described in the article "An Adaptive Spatial Scan Density Estimation Method" in March at a meeting in Salt Lake City, Utah.

She also served as a statistical consultant for The National Academies and in 2002 contributed to a volume entitled "Reducing Suicide: A National Imperative."

"This award is really important to myself and the department since not only the methods I am developing have important application areas but also because this research will bring new strength to the department in the area of mixture models and increase visibility and recognition in attracting top international scholars and visitors to the department," Pilla said.

Pilla joined the CWRU statistics faculty in 2002. From 2000-2002, she was an assistant professor at the University of Illinois, Chicago. She received her doctorate in statistics from the Pennsylvania State University and her dissertation received an award from the American Statistical Association in 1997.