A Fully Automated, Faster Noise Reduction Approach to Increasing the Analytical Capability of Chemical Imaging for Digital Histopathology

High dimensional data, for example from infrared spectral imaging, involves an inherent trade-off in the acquisition time and quality of spatial-spectral data. Minimum Noise Fraction (MNF) developed by Green et al. [1] has been extensively studied as an algorithm for noise removal in HSI (Hyper-Spectral Imaging) data. However, there is a speed-accuracy trade-off in the process of manually deciding the relevant bands in the MNF space, which by current methods could become a person month time for analyzing an entire TMA (Tissue Micro Array). We propose three approaches termed Fast MNF, Approx MNF and Rand MNF where the computational time of the algorithm is reduced, as well as the entire process of band selection is fully automated. This automated approach is shown to perform at the same level of reconstruction accuracy as MNF with large speedup factors, resulting in the same task to be accomplished in hours. The different approximations of the algorithm, show the reconstruction accuracy vs storage (50x) and runtime speed (60x) trade-off. We apply the approach for automating the denoising of different tissue histology samples, in which the accuracy of classification (differentiating between the different histologic and pathologic classes) strongly depends on the SNR (signal to noise ratio) of recovered data. Therefore, we also compare the effect of the proposed denoising algorithms on classification accuracy. Since denoising HSI data is done without any ground truth, we also use a metric that assesses the quality of denoising in the image domain between the noisy and denoised image in absence of ground truth.