Prior to his retirement, Dr. Mike Keenan was a Distinguished Member of the Technical Staff at Sandia National Laboratories. He joined Sandia after being awarded a PhD in physical chemistry by the University of Illinois at Urbana-Champaign. Dr. Keenan spent most of his career in the Materials Science Center where his early research interests included the physical properties of polymers and packaging of electronic components. Subsequently, Dr. Keenan managed Sandia’s Analytical Chemistry Department; following that, he took advantage of an opportunity to pursue long-standing interests in computing and statistics by applying them to the multivariate analysis of hyperspectral images. The
analysis of such images, where a full spectrum is acquired at each point in a 2D or 3D sample, poses significant challenges ranging from the massive sizes the data sets generated by current imaging systems to the low signal-to-noise typical of the individual spectra.

Dr. Keenan’s contributions included developing efficient algorithms to extract chemical information from spectral images in an optimal and unbiased manner, and providing
approaches to deal with the critically important task of accounting for the noise characteristic of counting measurements. These accomplishments were recognized by a 2002 R&D 100 Award shared with Paul Kotula, also at Sandia, for Component Analysis Software. This development enabled the routine multivariate statistical analysis of large spectral images, given the modest computing resources generally available in the lab. Dr. Keenan was also member of the Sandia team that was awarded an R&D 100 Award in 2009 for the
Hyperspectral Confocal Fluorescence Microscope System.

Since his retirement from Sandia, Dr. Keenan has continued to pursue research in this area as an independent scientist. His interests include developing and applying efficient numerical algorithms, general noise models, and new analysis approaches that accentuate selectable aspects of the multivariate models with the goal of improving interpretability.