A new pre-processing method for scanning X-ray microdiffraction patterns

Scanning X-ray microdiffraction (SXMD) is a novel technique to study the macromolecular architecture of tissues, such as cellulose in biomass. SXMD can generate huge amount of scattering patterns corresponding to different positions on a sample. In this paper, 190 images in a 38 x 5 grid are collected from SXMD experiment done at APS in Argonne National Lab to study nanoscale architecture in plant cell wall. A pattern-partition strategy utilizing image entropy, similarity coefficient analysis and k-means based clustering was carried out to study these diffraction patterns. Both similarity coefficient analysis and k-means clustering provide informative results in regard of the nanoscale architecture of Arabidopsis stem. This strategy is shown to reduce the amount of pre-processing work needed to analyze SXMD data.