Application of smoothing splines for spectroscopic analysis in hyperspectral images
Journal article, Peer reviewed
MetadataShow full item record
Original versionProceedings of SPIE. 2019, 10873 https://doi.org/10.1117/12.2506618
The spectral and spatial resolution of hyperspectral imaging is useful for investigation of tissue autofluorescence. The low-light, noisy conditions in fluorescence imaging usually necessitates noise removal for extraction of precise spectral signatures and peak shifts. However, noise removal techniques like low-pass filtering or the Maximum Noise Fraction transform might discard information or distort spectral features. In this study, smoothing splines is proposed as an alternative technique to avoid spectral distortion in analysis of hyperspectral fluorescence images in the wavelength range 400-1000 nm. Continuous tuning parameters and use of natural cubic splines makes the method advantageous for unbiased peak extraction. The method was tested on ex vivo images of atherosclerosis lesions and simulations. The method was used to estimate autofluorescence peak shifts, and found to perform well in comparison with MNF.