Spatial and spectral selection methods combined with a multimodal fusing approach for the analysis of dataset from multimodal imaging setups

We present a new method that merges multimodal imaging dataset acquired via multimodal hyperspectral imaging (HSI) setups. These instruments simultaneously collect data from multiple techniques, ensuring aligned spatial resolutions and seamless data integration. We combine band and spatial selection methods with the fusion approach to address issues related to the high dimensionality of HSI datasets. The spectral merging of the datasets in an early stage of the analysis enables a more comprehensive understanding of material compositions, while computational demands are decreased without sacrificing important information by choosing informative wavelengths and locations of interest. Deeper understanding of material compositions and distributions is subsequently provided by using multivariate and clustering analysis on fused data. This framework has demonstrated efficacy with various dataset combinations, such as reflectance and photoluminescence (PL) or time-gated PL at varying delays.