Application of neural networks for classification of biological compounds from the characteristics of whispering-gallery-mode optical resonance

Abstract

We have analyzed the feasibility and expediency of using a probabilistic neural network as a classifier for processing and interpreting data obtained using a biosensor we have developed that is based on whisperinggallery-mode optical resonance. We have established that the probabilistic neural network makes it possible to correctly classify the biological compounds under study with average 97.3% probability.Introduction. Development of intelligent systems based on artificial neural networks to solve problems in classification of biological compounds is a promising approach in biomedical data processing [1][2][3][4][5][6][7][8]. Results of studies on development of a biosensor based on optical resonance of whispering-gallery modes (WGMs) in dielectric microspheres [9][10][11][12] have shown that measurements of the spectral shifts of the WGM resonance spectra are insufficient for classification and identification of biological compounds because the shifts are not individual, and for a broad class of compounds practically coincide not only in absolute value but also in direction. Therefore for more adequate identification, which is one of the key requirements for a biomedical sensor, we need to use not only other spectral characteristics of the optical resonance but also additional tools for processing the measured data. In classification problems in biomedicine, there has been widespread use of probabilistic neural networks (PNNs) [1-4], multilayer perceptron networks (MLPs) [5][6][7], and the error back-propagation algorithm [8]. In particular, serological type A influenza virus could be correctly determined in 100% of the cases using a PNN [1]. When all the required information was available, a PNN correctly classified posterior fossa tumors in 85.7% of the cases, and in 72.7% of the cases when there was insufficient data [2], which is better than when using other methods. A PNN was used for diagnosis and classification of osteoarthritis according to severity [3] and for classification of protein groups [4]. The problem of protein group classification (from a given unknown protein sequence) is very important for a number of reasons: new drug discovery, prediction of molecular weight, and also refinement of medical diagnosis. For the enzymes esterase, lipase, and cytochrome oxidase, the classification probability was respectively 95. 2-98.2%, 95.3-98.7%, and 96.7-99.4% [4]. By varying the hidden layers of the neural net using the multilayer perceptron (MLP) algorithm, a plant virus identification probability of 92.96-93.61% was obtained [5]. The MLP architecture used in [6] provided a correct diagnosis of diabetic retinopathy in 95% of the cases. A PNN and MLP were used by the authors of [7] to determine the cardiac arrhythmia type. The input vectors were formed using electrocardiogram (ECG) data and some additional data on the structure of the heart. The ECG was run through preliminary processing, including low-frequency filtering. In the final configuration, the MLP correctly determined the type of arrhythmia i...

Plasmonic improvement of microcavity biomedical sensor spectroscopic characteristics

“…We have used several schemes of experimental realization of the method [6][7][8][9][10]. Here the newer versions of the scheme have been developed (Fig.1).…”

“…Improved methods of nanoparticle synthesis based in particular on ultrafast laser ablations directly in biocompatible liquids and different materials under laser-induced synthesis can frame conditions for functionalizing and fragmentation of nanoparticles within a single process step and combine it with aggregate formation. On the other hand recently developed tools for data processing can realize real-time identification of biological agents [6][7][8][9]. So combining advantages of plasmon enhancing optical microcavity resonance with identification tools can give a new platform for ulta sensitive label-free biomedical sensor, which can be used for cellular and tissue imaging for diagnostics, blood components monitoring, drug monitoring and targeted drug delivery, remote control for detection, treatment and destruction of cancer, nanolaser for nanosurgery etc.…”

Diagnostics of biomedical agents by whispering gallery mode optical resonance based sensor

“…Introducing controlled amount of nano particles (50 nm in diameter glass gel solution) into microsphere surrounding was accompanied also by correlative resonance shift. The most attention was concentrated on development of a technique for recognition and identification of nanoparticles and biomolecules in liquid fluids [8][9][10][11][12]. We demonstrated that the only spectral shift is not sufficient for identification of biological agents by developed approach.…”

“…We demonstrated that the only spectral shift is not sufficient for identification of biological agents by developed approach. So classifier based on probabilistic neural network for biological agents and micro/nano particles classification has been developed [8][9][10][11][12]. It was tested on solutions of different biological agents.…”

“…A novel emerging method for the label-free analysis of nanoparticles and biomolecules in liquid fluids using optical micro cavity resonance of whispering-gallery-type modes is being developed [9][10][11][12]. Several schemes of experimental realization of the method have been tested [13][14][15][16].…”