Документ взят из кэша поисковой машины. Адрес оригинального документа : http://rswater.phys.msu.ru/en/articles/2014/JBO_19_11_117007.pdf Дата изменения: Wed Dec 17 13:01:08 2014 Дата индексирования: Sat Apr 9 23:58:26 2016 Кодировка: IBM-866

Optical imaging of fluorescent carbon biomarkers using artificial neural networks
Tatiana A. Dolenko Sergey A. Burikov Alexey M. Vervald Igor I. Vlasov Sergey A. Dolenko Kirill A. Laptinskiy Jessica M. Rosenholm Olga A. Shenderova

Downloaded From: http://biomedicaloptics.spiedigitallibrary.org/ on 12/17/2014 Terms of Use: http://spiedl.org/terms


Journal of Biomedical Optics 19(11), 117007 (November 2014)

Optical imaging of fluorescent carbon biomarkers using artificial neural networks
Tatiana A. Dolenko,a,* Sergey A. Burikov,a Alexey M. Vervald,a Igor I. Vlasov, Kirill A. Laptinskiy,a Jessica M. Rosenholm,e and Olga A. Shenderovag
a b c

b,c

Sergey A. Dolenko,

d

M. V. Lomonosov Moscow State University, Department of Physics, Leninskie Gory 1/2, Moscow 119991, Russia A. M. Prokhorov General Physics Institute, RAS, Vavilova Street 38, Moscow 119991, Russia National Research Nuclear University MEPhI, Kashirskoe Avenue 31, Moscow 115409, Russia d M. V. Lomonosov Moscow State University, D. V. Skobeltsyn Institute of Nuclear Physics, Leninskie Gory 1/2, Moscow 119991, Russia e Abo Akademi University, Laboratory for Physical Chemistry, Porthansgatan 3, 20500 Turku, Finland g Adсmas Nanotechnologies, Inc., Raleigh, North Carolina 27617, United States

Abstract. The principle possibility of extraction of fluorescence of nanoparticles in the presence of background autofluorescence of a biological environment using neural network algorithms is demonstrated. It is shown that the methods used allow detection of carbon nanoparticles fluorescence against the background of the autofluorescence of egg white with a sufficiently low concentration detection threshold (not more than 2 gml for carbon dots and 3 gml for nanodiamonds). It was also shown that the use of the input data compression can further improve the accuracy of solving the inverse problem by 1.5 times. й 2014 Society of Photo-Optical Instrumentation Engineers
(SPIE) [DOI: 10.1117/1.JBO.19.11.117007]

Keywords: carbon nanoparticles; autofluorescence; optical imaging; neural networks. Paper 140462R received Jul. 18, 2014; revised manuscript received Oct. 4, 2014; accepted for publication Oct. 15, 2014; published online Nov. 14, 2014.

1

Introduction

One of the main focuses of modern biomedicine is the development of supersensitive and high-throughput methods for rapid detection of proteins and genes as well as visualization of biomarkers and drug carriers in biological material.1н3 The most widespread method of recognition of biomarkers and study of cellular interactions is visualization by means of fluorescence. However, a variety of complications exist in the tracking of fluorescent markers in the biological environment. Most of the fluorescent markers are excited by ultraviolet or visible radiation and emit in the visible spectral range. However, in this spectral range, biological tissue absorbs and emits light strongly.1н3 Therefore, illumination of the biological material in this range induces fluorescence of the living tissue (autofluorescence), which is caused by natural fluorophores (tryptophan, phenylalanine, tyrosine, collagen, flavins and flavoproteins, beta-carotene, porphyrins, nucleic acids, coenzymes, vitamins, pigments and so on4). The spectrum of the autofluorescence is the result of superposition of fluorescence bands of fluorophores, vastly complicating the observation of cellular processes and the tracking of the fluorescent markers. Currently, there are two major approaches to overcome the problem of background fluorescence: (1) synthesis of new biomarkers with minimum overlap of their emission spectra with the background fluorescence2н25 and (2) development of advanced experimental techniques permitting one to decrease the background signal.26н34 Intense work is being carried out on the search for and synthesis of nanoparticles with optimal properties for biological applications as biomarkers, absorbents, and drug carriers.2н5

The existing biomarkers have several drawbacks. Molecules of organic dyes have a high intensity of luminescence per mass unit but they are not suitable for long-term in vitro and in vivo control because of fast photobleaching and cellular toxicity.6 Even dyes with improved properties, created, for example, by combinatorial organic synthesis,7 are not photostable. In comparison with organic probes, semi-conducting quantum dots have a few major advantages:8,9 high quantum yield and photostability, broad excitation range, and narrow emission band with the emission maximum strongly depending on the size of the quantum dot. The drawbacks include cytotoxicity, which is still debated, and luminescence blinking.8,9 Promising materials for optical imaging are fluorescent carbon nanoparticles such as nanodiamonds (NDs),10н16 carbon dots/graphene oxides (CD/GO),17н19 and nanodiamonds decorated with carbon dots (CD-D-ND).20 Apart from outstanding photostability and high quantum efficiency, CDs and NDs have poly-functional surfaces which can be modified to address specific needs. Methods of coating carbon nanoparticles with polymers or mesoporous silica shells have been developed. These coatings improve biocompatibility and decrease the toxicity of the particles while preserving their fluorescence properties.21н25 A drawback of the carbon nanoparticles is a broad emission band overlapping with the autofluorescence, which does not provide the required contrast in particles imaging. Thus, despite significant progress in the synthesis of new nanoparticles for biological applications, a big challenge remains in the development of nontoxic imaging probes providing a highly reliable visualization in combination with a pronounced therapeutic effect. Most of the methods of optical imaging are based on optical microscopy.26н30 Confocal laser scanning microscopy (CLSM)
0091-3286/2014/$25.00 й 2014 SPIE

*Address all correspondence to: Tatiana A. Dolenko, E-mail: tdolenko@lid.phys .msu.ru or tdolenko@mail.ru

Journal of Biomedical Optics

117007-1

November 2014

§

Vol. 19(11)

Downloaded From: http://biomedicaloptics.spiedigitallibrary.org/ on 12/17/2014 Terms of Use: http://spiedl.org/terms


Dolenko et al.: Optical imaging of fluorescent carbon biomarkers using artificial neural networks

is a commonly used tool that provides morphological and functional information within cells and tissues.28,29 It was shown that the unique capabilities of CLSM such as the ability to detect fluorescence, three-dimensional (3-D) image reconstruction, and use of the reflection mode in tandem with other methods, provide great promise with broad applications for the visualization of various nanoparticles such as quantum dots, fullerenes, dendrimers, for example, at their penetration into the surface of the skin. It was shown that multiphoton excitation microscopy has advantages over conventional one-photon imaging microscopy and can be used as a modern noncontact tool for 3-D fluorescence imaging and optical diagnostics.31,32 For visualization of markers within tissue, the capability of metal nanoparticles to generate plasmons while interacting with an electromagnetic field is widely used.33,34 Thus, a dark-field microspectroscopy technique for molecular imaging of live cells using antibody-labeled plasmonic nanoparticles was demonstrated.33,34 Experiments have shown that the accuracy of the visualization is influenced both by interactions of the nanoparticles with biomacromolecules and by the interactions of interparticle plasmons. Rapidly developing methods of visualization of nanoparticles in biological tissues include photoacoustic high-resolution imaging35,36 and high-sensitivity real-time radiometric imaging of nanoparticles with surface-enhanced Raman scattering.37,38 One of the approaches to solve the problem of autofluorescence suppression is the synthesis of new nanoparticles called upconversion nanoparticles (UCNPs).39н41 These particles have improved luminescence properties (the emission/excitation power ratio39). UCNPs properties allow strong suppression of background signals (excitation light back-scattering and biological tissue autofluorescence). Currently, UCNPs particles are among the most promising for bionanomedicine. Together with the microscopy technical capabilities, methods of image processing are also being developed. Recent progress in the 3-D single particle tracking is summarized in the review by Dupont and Lamb.42 A modified version of the original orbital tracking in which the intensities from two z-planes are simultaneously measured is demonstrated. The precision of single particle tracking is 5 nm in the x-y plane and 7 nm in the axial direction. Now, methods of background-free fluorescence based on fluorescence modulation by electromagnetic or magnetic fields combined with image subtraction are actively being developed. Igarashi et al.43 improved the image contrast of fluorescent NDs (FNDs) in vitro as well as in vivo based on the spin property of the NVн center. The authors acquired wide-field fluorescence images with and without microwave irradiation in resonance with the crystal-field splitting (2.87 GHz) of the ground-state spin, and then performed subtraction between these two images pixel by pixel. As the alternative microwave irradiation modulated only the fluorescence intensity of the NVн center, the operation effectively removed background autofluorescence signals and significantly improved the image contrast. Similarly, Hegyi and Yablonovitch44 applied the optically detected magnetic resonance technique to image FNDs in tissue in the field-free region using an amplitude-modulated microwave source. More recently, Sarkar et al.45 utlized a modulated external magnetic field to achieve contrast enhancement of FNDs in vivo. The magnetic field mixes the spin levels at the ground state, resulting in modulation of the FND fluorescence. As a demonstration of the potential translational relevance of the work, where unique
Journal of Biomedical Optics

improvements to imaging efficiency were observed, this technique improved the image contrast by nearly two orders of magnitude, allowing for wide-field imaging of FNDs in sentinel lymph nodes of mice. Thus, the problem of optical imaging of nanoparticles or cells in biological tissues is currently being solved by synthesis of novel nanoparticles or nanocomposites and their optimization and through improvement of advanced equipment and methods of image processing. At the same time, in all the methods under development, there remains a problem of accounting for interactions between nanoparticles and biomaterial, requiring intensive study. In this paper, a new approach to solve the inverse problem of separation of the fluorescent signal of nanoparticles from the background autofluorescence is proposed based on artificial neural networks (ANNs).46 ANNs represent powerful data analysis algorithms which provide an efficient solution of inverse problems and problems of pattern recognition, including those in optical spectroscopy.46н50 Artificial intelligence techniques are practically used in all fields of bioinformatics,47 since there are no universal methods of data processing comparable to the efficiency of human intellectual potential. It is very promising now to use ANN for the solution of such problems as a method for classification of proteins, selection of genome fragments, recognition of signal peptides and transmembrane helices, and so on.51 In Refs. 52 and 53, the method of breast cancer diagnosis based on ANN classification was proposed. The inverse problem of autofluorescence recognition of cell culture and cancer cells was solved. Total synchronous fluorescence spectra of normal skin, nevus, and melanoma samples were used as input for training the ANNs. Two different types of ANNs were trained, the self-organizing map and the feed-forward neural network. Histopathology results of investigated skin samples were used as the important standard for network output. Based on the obtained classification success rate of neural networks, Dramic ‡anin et al.52 concluded that both networks provided high sensitivity with classification errors between 2% and 4%. Despite the extremely wide application of pattern recognition methods in biomedicine, to the authors' knowledge, this paper is the first application of these methods for detection of the fluorescence of nanoparticles in the presence of background autofluorescence. In the current paper, the method was elaborated for selected fluorescence nanoparticlesннCDs17 and nanodiamonds and using a particular biological object--chicken egg white. Proof of concept of fluorescence biomarkers visualization in biological media using neural network algorithms was demonstrated. The minimum concentration of nanoparticles that can be reliably detected in the presence of autofluorescence background using ANN was determined. Ways to improve the sensitivity of the ANN method at solving the above inverse problem are also discussed.

2
2.1

Experimental
Carbon Nanoparticles and Their Characterization

In this study, nano-graphene oxide (NGO) particles17 and detonation nanodiamond G01 (PlasmaChem, Germany) were used. NGO nanoparticles were synthesized by oxidizing nanographite in a 31 sulfuric to nitric acids mixture at a temperature of 130ђC. The reaction supernatant was collected by removing the
November 2014
§

117007-2

Vol. 19(11)

Downloaded From: http://biomedicaloptics.spiedigitallibrary.org/ on 12/17/2014 Terms of Use: http://spiedl.org/terms


Dolenko et al.: Optical imaging of fluorescent carbon biomarkers using artificial neural networks

graphitic carbon residue by centrifugation. The supernatant was neutralized by the addition of sodium hydroxide. The neutralized solution was dialyzed for 2 days against water.17 Below, they are called CD. Nanodiamonds G01 were prepared by controlled dry detonation synthesis followed by purification. The bulk powder density of the sample was 0.69 gcm3 , the primary particle size was about 4 nm. Bidistilled deionized water was used for the preparation of aqueous suspensions of carbon nanoparticles. Aqueous suspensions of CD with a concentration of 0.1 gl and aqueous suspension of ND with a concentration of 1 gl were prepared. Suspensions were treated for 2 h in an ultrasonic bath (Bandelin Sonorex rk 31). Measurements of the sizes of CD and ND in water suspensions were carried out by the method of dynamic light scattering using correlator-goniometer system ALV-CGS-5000/6010 (Langen, Germany), equipped with a He-Ne laser (at a wavelength of 633 nm, output power 20 mW). For numerical processing of the correlation functions, the software package CONTIN was used. Measurements showed that the majority of the CD particles in water had dimensions of 6 - 9 nm, and 88% of the ND particles in the suspension had a size of 35 nm.

Fig. 1 Dependence of the integral fluorescence intensity of carbon dots (CDs), nanodiamonds (NDs) and dye Alexa Fluor 488 as a function of time of the laser irradiation of the suspensions.

2.2

Biological Media

fluorescent nanoparticles with a much lower quantum efficiency was elaborated (the quantum yield of detonation nanodiamonds is 6.5 times lower than the quantum yield of CD).
2.3.2 Photostability of aqueous suspensions of carbon particles

As a biological media, chicken egg white was chosen. This choice is justified by the fact that delivery of nanoparticles directly into the cell is provided, as the whole egg white is a single cell. Also the choice is associated with representative natural fluorophores in the egg white. In subsequent experiments, in order to improve the reliability of the trained neural network toward changes in biological objects, egg whites of different stages of development and from different vendors were used.

2.3

Quantum Efficiency and Photostability of Carbon Nanoparticles

For the use of fluorescence nanoparticles in biological applications, high quantum yield, photostability and low toxicity are required.
2.3.1 Quantum yield of fluorescence of carbon nanoparticles

The fluorescence quantum yield is a critical parameter in selecting a material as a biomarker. In this paper, the fluorescence quantum yields of CD and ND were measured by a standard method relative to a reference dye quinine sulfate. The absorption spectra of the aqueous suspensions of CD and ND were recorded by spectrophotometer Perkin Elmer Lambda 25 in the range 200 to 900 nm with a resolution of 0.5 nm. Spectra of fluorescence of suspensions were recorded by fluorimeter FluoroMax-4 (Horiba Jobin Yvon). The obtained fluorescence spectra were corrected for absorption. The measured values of the quantum yields of CD and ND are 3.07% (at a wavelength of 405 nm) and 0.47%, respectively. Previously, neural network algorithms were tested by us to separate the fluorescence signal of the CD from background autofluorescence.54 In this paper, it is shown that ANN provides much greater sensitivity of detection of the fluorescent signal: using neural network algorithms, a method of recognition of
Journal of Biomedical Optics

We carried out a comparative analysis of the temporal stability of the fluorescence of water suspensions of CDs, the typical detonation ND, and organic dye Alexa Fluor 488 with the same concentration, 0.3 gl. A cuvette with a suspension of nanoparticles was fixed in the cell holder and constantly exposed to laser radiation (argon laser, at a wavelength of 488 nm, power density in cuvette about 10 Wcm2 in a defocused beam) for 2 h. Spectra were measured every 15 min in the spectral range 500 to 800 nm using a photomultiplier tube (PMT) (Hamamatsu, H-8259-01) in photon counting mode. The geometry of the experiment and the experimental conditions were exactly the same for all samples. Spectra were corrected for laser power, data acquisition time, sensitivity of the PMT, and absorption of samples. Figure 1 shows the dependence of the integral intensity of the fluorescence of the studied suspensions as a function of irradiation time. The experimental results clearly demonstrate the advantages of photostability of carbon nanoparticles as compared with the organic dye.

3
3.1

Fluorescence Spectroscopy of Biological Objects with Carbon Nanoparticles Ion
Laser Spectrometer

Fluorescence and Raman signals from aqueous suspensions of nanoparticles and biological objects containing nanoparticles were excited by a diode laser (at a wavelength of 405 nm, power incident on the sample 50 mW). Spectra were recorded by PMT (Hamamatsu, H-8259-01) in the range 430 to 750 nm. The practical spectral resolution was 0.5 nm. The temperature of the samples was stabilized at 22.0 Ї 0.1ђC. Spectra were corrected for the laser power and data acquisition time. Further, mathematical data processing consisted of the subtraction of the background caused by light scattering in the cuvette with the sample and
November 2014
§

117007-3

Vol. 19(11)

Downloaded From: http://biomedicaloptics.spiedigitallibrary.org/ on 12/17/2014 Terms of Use: http://spiedl.org/terms


Dolenko et al.: Optical imaging of fluorescent carbon biomarkers using artificial neural networks

normalization of the obtained spectra by the area under the Raman valence band of water.

3.2

Analysis of the Fluorescence Spectra of Egg White with Introduced Carbon Nanoparticles

Figures 2 and 3 show Raman and fluorescence spectra of egg white, water suspensions of CD and ND, and egg white with introduced nanoparticles. The relatively narrow band near 470 nm is the Raman valence band of OH-groups either of water or of biological media (percentage of water in the egg white is about 85%) under excitation at a wavelength of 405 nm. This band is caused by the valence vibrations of molecular OHgroups (wavenumber of this band maximum is 3400 cm-1 and width is about 400 cm-1 ). The spectrum of autofluorescence of egg white in the studied range consists of several bands: an intense broad structureless band in the range 430 to 730 nm with a maximum at 480 nm, and weaker bands with maxima at 640 nm, 655 to 660 and 675 nm. It is known4 that the basic fluorescence band of egg white is formed by the fluorescence contributions of pyridoxine, NADF, flavins, and lipo pigments. Weak fluorescence bands

near 640 to 670 nm are caused by the fluorescence of porphyrins. As seen from Figs. 2 and 3, when excited at a wavelength of 405 nm, CDs exhibited fluorescence in the spectral region between 430 and 680 nm with a maximum near 500 to 505 nm, while NDs emit in the region 430 to 680 nm with a maximum near 520 to 525 nm. Egg white emits in the region 420 to 700 nm with a maximum which varies for different eggs within 480 to 520 nm (see below). Thus, the fluorescence spectra of nanoparticles and egg white are largely overlapping structureless bands. They differ in the position of the maximum and the position and extent of the fluorescence spectral bands. These characteristics are the basis for the extraction of the fluorescent signal of the nanocarbon particles. Obviously, if the concentration of nanoparticles in the egg white changes, the band of the integral fluorescence varies significantly for several reasons. The main reasons are as follows: (1) when the concentration of nanoparticles changes, the intensity of the fluorescence of individual particles changes and (2) due to the interactions of CD and ND with components of the egg white, both the fluorescence of the egg white and that of nanoparticles change. These interactions are very complex and still not well understood in part because the carbon nanoparticles have been relatively recently synthesized. Using conventional methods, it is impossible to construct a mathematical model of the change of the overall fluorescence of an egg white and nanoparticles as a function of their concentrations (for example, during movement of nanoparticles in biotissue). This means that traditional mathematical methods cannot solve either the direct problem of spectra modeling, or the inverse problem of extracting the fluorescent contribution of varying amount of nanoparticles against a background fluorescence of egg white. Therefore, the algorithms of ANNs have been used.

4
4.1

Artificial Neural Networks
Multilayer Perceptron

Fig. 2 Raman and fluorescence spectra of water, egg white, aqueous suspensions of CD and egg white containing CD.

In this study, we use the most widespread type of ANNs such as the multilayer perceptron (MLP). The elementary unit used to construct a MLP is a so-called formal neuron, a unit with several inputs and a single output. Each of the inputs is characterized by its own weight coefficient, wj . The value y at the output of a neuron is calculated as a weighted sum of the values xj at its inputs, put through a nonlinear function F called activation function:

y®F

N X j®0

! wj x
j

;

x0 1 :

(1)

Fig. 3 Raman and fluorescence spectra of water, egg white, aqueous suspensions of ND and egg white containing the ND.

Now, identical neurons, different only by their weights, are combined into blocks, which are used to create layers. Each neuron is connected with each of the neurons in the preceding and the next layers (fully connected scheme) (Fig. 4). The signal is fed, according to Eq. (1), from the input of the first layer of the MLP to the output of the last one. The first layer of an MLP is called the input layer, as it is used to feed data into the MLP, and serves as the input for the whole network. There are as many neurons in the input layer as there are features describing each of the processed data samples. The neurons of the input layer perform no calculations; they simply distribute the input signal to all the neurons of the next layer.
November 2014
§

Journal of Biomedical Optics

117007-4

Vol. 19(11)

Downloaded From: http://biomedicaloptics.spiedigitallibrary.org/ on 12/17/2014 Terms of Use: http://spiedl.org/terms


Dolenko et al.: Optical imaging of fluorescent carbon biomarkers using artificial neural networks

Fig. 4 Multilayer perceptron.

The last layer of an MLP is called the output layer; it serves as the output of the whole network. There are as many neurons in the output layer as there are values that are to be simultaneously determined for each data sample. One or several layers between the input and the output layers are called hidden layers. The numbers of neurons in the hidden layer(s) determine the complexity of the network; it can be estimated by some half-empirical equations, but often it is determined by trial and error. When the MLP is applied, its output depends on the values of the weight coefficients w (1). Hence, to provide correct answers for an MLP, it should be trained prior to its application. This is done by analyzing the answers of the MLP on the samples of the training set of data, for which the correct (desired) answers are known, with subsequent tuning of the weight coefficients. The goal of training is to minimize the average error of the MLP over the training set by changing the values of the weight coefficients of all the neurons. Most often this is performed by the so-called error backpropagation algorithm, which performs gradient descent in the space of weight coefficients.46

In the "quasimodel" approach, in order to obtain representative datasets, simulated spectra are used. Unlike the "model" approach in which the data for the learning sample are calculated according to the known analytical expression describing the spectra, in the "quasimodel" approach, first a parametric "quasimodel" describing the spectra on the basis of a small set of experimental data is constructed, and then it is used to compute the data for the learning sample. Obviously, in this way, a sufficient number of patterns can be obtained providing a good representativity of all sets for ANN training. However, the accuracy of solving the inverse problem with the "quasimodel" approach largely depends on two factors: (1) the error of the "quasimodel" used for the calculation (i.e., correspondence of the selected or calculated "quasimodel" with reality) and (2) differences of the noise in calculated patterns and the noise in real experimental data. In the "experiment-based" approach, experimental data are used to train the neural network. The disadvantage of this approach is the low representativity of sets because getting a very large amount of experimental material is very time consuming. The main advantages of this approach include: training neural network directly on the experimental data taking all molecular interactions into account, and the fact that the network is trained on real experimental noise, which increases the accuracy of the solution of inverse problems. As was mentioned above, a problem of recognition of the nanoparticles' fluorescence in the presence of the background fluorescence of egg white cannot be solved based on the "model" approach due to the lack of a correct analytical description of the fluorescence spectra of nanoparticles and the egg white. In addition, because the object of the study is a living biological material and its state can vary significantly with time, it is especially important to train the ANN on real signals of objects containing noise. Thus, in this paper, the inverse problem was solved by ANN in the framework of the "experimentbased" approach.

5
5.1

Results and Discussions
Experimental Learning Sample for Artificial Neural Networks

4.2

Use of Artificial Neural Networks to Solve Inverse Problems

Application of ANN for solving inverse problems of optical spectroscopy is possible using three approaches: "model," "quasi-model," and "experiment-based."48 In the "model" approach to obtain a learning sample for ANN training, one applies an existing analytical or computational model of the solution of the direct problem. In the presence of such a model, it is possible to provide the required representativity of all the datasets necessary for ANN training (description of the necessary datasets into which the total learning sample should be divided is provided in Sec. 5.3). However, the quality of solution in this case depends on the adequacy of the model used. In situations where the development of an adequate model is impossible because of the complexity of the object, this approach is not feasible.
Journal of Biomedical Optics

Two series of Raman and fluorescence spectra were obtained for two different egg whites containing CD. The CD concentration varied in the range from 0 to 20 gml with increments of 0.75 and 1.5 gml. Three series of three egg whites containing ND were obtained. The concentration of ND varied from 0 to 30 gml with increments of 1, 1.5, and 3 gml. To improve the stability of the solution relative to the biological object model, different series of spectra were recorded with different egg whites. As seen from Fig. 5, such an approach is justified since the fluorescence spectra of different egg whites are significantly different. At that position and shape of integral fluorescence band of pyridoxine, NADF, flavins, lipo pigments (430 to 730 nm), and band of porphyrins (640 to 670 nm) remain practically unchanged. Figure 6 shows examples of the experimental Raman and fluorescence spectra of egg whites with different concentrations of nanoparticles. As seen from Fig. 6, the water Raman valence band (in the region 470 nm) and fluorescence bands of porphyrins (bands near 640 to 670 nm) practically do not change while the fluorescence contribution of CD in the integral fluorescence
November 2014
§

117007-5

Vol. 19(11)

Downloaded From: http://biomedicaloptics.spiedigitallibrary.org/ on 12/17/2014 Terms of Use: http://spiedl.org/terms


Dolenko et al.: Optical imaging of fluorescent carbon biomarkers using artificial neural networks

standard deviation of the values in a channel, proportional to the amount of information contained in that channel. These studies are important not only from the viewpoint of increasing the sensitivity of the method, but also from a practical point of view. Thus, the aggregation of every four channels corresponds to a coarsening of the spectral resolution of the registration devices by four times (from 0.5 to 2 nm). The spectral range of registration is usually chosen in a way to include the entire bands of interesting objects. Discarding insignificant input features allows one to select the most informative region of the spectra and to use the device with registration in the more narrow spectral range.

5.3
Fig. 5 Fluorescence spectra for different egg whites: a, b, c--egg whites from different vendors; d--egg white after 3 weeks.

Results of Using Artificial Neural Networks

Fig. 6 One of the series of Raman and fluorescence spectra of egg whites with different concentrations of introduced carbon dots.

band in the region 430 to 600 nm changes. The obtained datasets for CD and ND were used for the training of ANN.

5.2

Preliminary Processing of Input Data

The considered problem in its original formulation is characterized by an extremely unfavorable ratio of the number of patterns in the training set (30 to 45) and the number of input features (651 pointsнwavelengths of spectra). Therefore, an important area of resear