Powerful Artificial Neural Network for Planar Chromatographic Image Evaluation, Shown for Denoising and Feature Extraction

Abstract

An artificial neural network (ANN) is presented as a new and superior technique for processing planar chromatography images. Though several algorithms are available for image processing in planar chromatography, the use of ANN has not been explored so far. It simulates how the human brain interprets images, and the intrinsic features of the image were captured on patches of pixels and successfully reconstructed afterward. The obtained high number of observations was a perfect basis for using ANN. As examples, three quite different data sets were processed with this new algorithm to demonstrate its versatility and benefits. Powerful features, which the ANN learned from the image data set, improved the quality of the analytical data. Thus, noise or inhomogeneous background of bioautograms was removed as demonstrated for salvia extracts, improving their bioquantifications. On colorful fluorescence chromatograms of further botanical extracts, the power and benefit of the feature extraction were demonstrated. Using ANN, videodensitometric results were improved. If compared to conventional digital processing, the resolution between two adjacent blue fluorescent bands increased from 0.95 to 1.18 or between two orange fluorescent bands from 0.77 to 1.57. The trueness of the new ANN was successfully verified by comparison with conventional densitometric results of the absorbance of separated tea extracts. The correlation coefficients of epigallocatechin gallate therein improved from 0.9889 with median filter to 0.9959 using this new ANN algorithm. The code was released open-source to the scientific community as a ready-to-use tool to exploit this potential, spread its usage, and boost improvements in planar chromatographic image evaluation.

Comparison of spectral and spatial denoising techniques in the context of High Definition FT-IR imaging hyperspectral data

“…Median, Mean and Gauss filters 1 are commonly used in data processing of imaging data, while wavelets have been suggested as very promising techniques for IR as well 24 . Finally, Artificial Neural Networks can be used for denoising 25 and recently the first Deep Neural Network (DNN) denoising approach was enabled in Matlab packages and is applied here to IR data for the first time.…”

Open‐source add‐on kit for automation of zone elution in planar chromatography

“…It may provide the platform for easy mass spectra readout and processing in a facilitated workflow to automatically define target zones and to combine all possible datasets from HPTLC plates for an intelligent evaluation. For example, quanTLC for peak detection and videodensitometric quantification or superior image evaluation by artificial neural networks or powerful cluster analysis of the recorded HPTLC‐MS spectra were recently developed modern tools that are helpful for analysts in the field of TLC/HPTLC.…”

Study on Waste Type Identification Method Based on Bird Flock Neural Network

“…Backpropagation neural network (BPNN) is a classic neural network; at present, BPNN has been applied in many fields [15][16][17]. ere are problems of local optimum [18], accuracy [19], and hidden layer neuron (nodes) adjustment [20] in a typical BPNN and other improved algorithms. According to the above problems, a bird flock neural network (BFNN) was proposed, and then a waste type identification method based on the BFNN was proposed.…”