抄録
Image data fusion of time-of-flight secondary ion mass spectrometry (TOF-SIMS) and scanning electron microscopy (SEM) data was investigated. The fused data were analyzed with multivariate analysis, deep learning, and sparse modeling to study what types of information could be extracted from data analysis. Image fusion with a method having higher spatial resolution provides not only higher resolution images but also more detailed chemical information. Microscope images could be useful guides to classify the samples and extract region of interest even if it is unknown. Model samples having micro meter patterns were analyzed by TOF-SIMS and SEM or optical microscopy. The same sample areas were trimmed from image data obtained with different methods and then fused by adding the other one’s intensity at each pixel to TOF-SIMS peak intensities. The patterns in the sample observed with SEM are helpful to extract important secondary ion information with reduced influence of noises. LASSO (least absolute shrinkage and selection operator) and Autoencoder, one of the deep learning techniques, extracted main information of the model samples as well as principal component analysis (PCA).
