1. INTRODUCTION

The photobiochemical effects observed at the cellular level result from the interaction of electromagnetic radiation with cells and tissues [1,2]. They include the phenomena of biological stimulation, inhibition, and cell death [3]. Currently, the most common therapy based on these effects is low-level laser therapy (LLLT). This phototherapy uses light in the wavelength range from red to near infrared (600–1100 nm) [4,5] with low energy or power densities to irradiate cells or tissue [2]. The current state of knowledge is that the specific effects of this therapy are due to light absorption by chromophores, including cytochrome c oxidase [6,7]. As indicated in the literature, LLLT has a number of important beneficial effects at different biological levels of organization, from molecular to tissue. It has been shown that this therapy can influence the regulation of cell proliferation [5,8,9], which currently constitutes one of the main photobiochemical effects applicable in clinical therapy [2], including wound healing [6,10,11]. Influence of LLLT on the proliferation of biological cells indicates a variety of effects of this therapy, both stimulating and inhibiting, depending on the type of cells studied, but also on the choice of irradiation parameters [6,9,12]. Indeed, it has been indicated that LLLT affects the proliferation of keratinocytes [13,14], fibroblasts [15,16], and other types of biological cells [5,17–19].

In terms of anti-cancer therapy, LLLT may have beneficial or negative effects depending on the wavelength of light and energy density used, as was shown in a systematic review by da Silva et al. [12]. On one hand, it can significantly influence the effectiveness of anticancer therapy by treating its side effects [20], which include oral mucositis, [21] and by inhibition of cancer cell proliferation. On the other hand, an increase in the proliferation of cancer cells [20,22] and also their aggressiveness [23] can be observed. The inconsistency of the results and of used irradiation parameters makes this area of application an especially controversial issue [20]. LLLT potentially offers great opportunities, but it must be stressed that the efficiency of this procedure is related to the choice of dosimetry parameters, including wavelength, exposure time, energy, and power density [5,24]. Their inadequate choice may be associated with a decrease in the effectiveness of the phototherapy and often is the main reason for the negative results of experimental and clinical studies [25]. For this reason, many medical groups are skeptical of the use of LLLT in medical practice [26]. In view of the above, further clinical as well as experimental studies are needed to determine the optimal irradiation parameters, which may result in better comprehension of this method.

 

Fig. 1. Main steps in the experiment.

Download Full Size | PDF

A key aspect in observing the specific photobiochemical effects underlying LLLT is the temporal analysis of its influence on cells. This requires capturing a series of measurements in time-lapse mode, which in turn means that investigated cells cannot be affected by any factors during the whole analysis. Since fluorescence microscopy requires utilization of phototoxic staining, it is not a suitable tool for this application. Another important requirement is to provide quantitative results of phase measurement. This allows for reliable assessment of the changes in cells and combining these results with setting proper LLLT parameters as well as a reliable comparison of experiments conducted in different laboratories. One technique that fulfills these requirements is digital holographic microscopy (DHM). It provides the possibility to perform label-free and quantitative phase imaging and allows observation of the integrated phase distribution in living cells, which can be converted to a dry mass density distribution without the use of cell markers affecting cell survival [27]. Determination of the dry mass density distribution allows the assessment of the current stage of cell cycle [28].

DHM has been widely used to study a variety of cell types (e.g.,  neurons, fibroblasts [29], erythrocytes [30], epithelial cells [31], stem cells, and cancer cells [28,29,32]). DHM is successfully applied to time-lapse measurements in a noninvasive manner, as indicated in the study of erythrocyte membrane fluctuations [33], morphology of living cells in different biological states [34], monitoring of cell biological processes such as cell division [31], cell death [34], and observation of cell motion [35]. Moreover, Bettenworth et al. in their work indicated the advantages of using DHM in observing the wound healing processes in a time-lapse mode [27].

In this paper, we propose an experimental procedure based on DHM phase measurements well fitted to the requirements of long-term monitoring of LLLT effects on investigated biological cells. We propose the full methodology of statistical analysis of temporal changes in biophysical features of cells under different irradiation doses. We have focused on solving crucial problems at each stage of the measurement methodology. The effectiveness of this approach is tested through investigation of SHSY-5Y neuroblastoma cells and HaCaT keratinocytes exposed to red light radiation.

2. METHODS

The proposed experimental–numerical procedure that employs DHM to study the effects of LLLT therapy on cells includes five basic steps, as shown in Fig. 1. The steps are:

Each of the steps is associated with different types of challenges which, through preliminary experiments, we have been able to answer. In this section, we will address these challenges.

A. Phase Images Specific Preprocessing and Stitching

Phase images require preprocessing before the stitching procedure. This is because phase images contain phase terms that are connected with: (1) aberrations of the system and (2) with the relative nature of phase measurements. Using the preprocessing steps we can maintain the quantitative character of the synthesized image [36,37].

The aberrations are removed in two steps. First, the systematic aberrations of the optical setup are estimated using averaging techniques, and they are subtracted from the individual phase images [36,37]. After this step, each image is corrected to remove phase trends that are caused by sample and system instabilities due to automated stage movement [37]. The systematic aberrations and phase trends are not symmetrically distributed within the FoV. Additionally, phase trends are different for each phase image; hence, they introduce different errors in the overlapping regions of two neighboring images. Therefore, the above corrections are necessary to perform the final correction that considers the differences between the images in the overlapping regions.

Because QPI techniques measure only the optical path difference and not the absolute optical path, we must establish a common baseline before the images are fused together. This is done for each image by finding an offset value based on differences in overlapping regions. The differences between the overlapping regions are globally minimized in an optimization procedure and final offset values are added to corresponding images, hence establishing a common baseline [37]. This step is crucial in the case of the quantitative analysis of near-full confluence samples because the baseline value must be unified based on the duplicate measurement in the overlapping regions due to lack of a cell-free region in some FoVs. This means that only one image with a visible baseline region is required to perform quantitative analysis of the full synthetic aperture image. The alignment between images does not need to be corrected before the baseline correction because the method is robust enough not to require a perfectly accurate size of the overlapping regions [37].

After the preprocessing of all single FoV phase images, a typical microscopy image stitching is performed that consists of mutual image alignment and image fusing. For this step, we use the BigStitcher [38], a FIJI plug-in. The physical coordinates provided by the motorized stage are not precise enough to directly fuse images [39]; hence, the phase correlation method [39] is used for image alignment. Image fusion is performed with smooth blending of the overlaps; that is, with a cosine-shaped fade-out toward the frame border.

B. Segmentation

Two types of cell culture are usually used in LLLT experiments: a typical cell culture with low confluence and one in which cells reach almost full confluence, such as in wound healing assays. The stitched image is preprocessed before segmentation. Preprocessing consists of the following steps: The constant component is subtracted from the image to set the background level to 0. Then the fragments with values below zero are removed. The segmentation procedure depends on the type of cell culture.

Images of a conventional cell culture were segmented with an adaptive image threshold using local first-order statistics [40,41]. In the case of the analyzed data, the thresholding sensitivity was set to a value of 0.4, and it was defined such that the foreground pixels (biological cells) were brighter than the surrounding background. In addition, the Gaussian weighted mean procedure was chosen to compute the local image background threshold. According to this procedure, for a given neighborhood area the pixels are first multiplied with Gaussian kernel coefficients and then summed. The resulting weighted mean is assigned to the pixel in the center of the specified neighborhood area. The image smoothing obtained with this approach allowed for partial noise reduction [42]. Other parameters of the preprocessing function are set to their default values [41]. In the next step, the image binarization is performed on the basis of the determined threshold, after which segments containing less than 300 pixels are removed and the holes are filled, finally creating cell masks. After preprocessing, the problem of residual aberrations and noise was not completely eliminated, which resulted in some background pixels being classified as cell pixels. Therefore, for regions where the background areas were erroneously segmented, the threshold value was manually adjusted.

Images of a cell culture with near-full confluence were segmented using a graph-based technique called a graph cut [43,44]. Semi-automatic segmentation was performed using the image segmenter tool available in Matlab R2018b software. After loading the images, foreground sections (cell segments) and background areas (gaps between cell segments) were manually selected.

C. Calculation of Cell Parameters and Their Statistics

The analysis of LLLT effects on cells was performed based on three parameters directly related to holographic phase measurements: the optical path difference (OPD), the projected area of the cells, and their dry mass density. For cell cultures with low cell confluence, the projected area ${S_{\rm{cell}}}$ is calculated according to

The second parameter, the dry mass density $\rho$, which quantifies the density of nonaqueous cell material [Eq. (2)], was calculated using the dry mass of cells $M$ [Eq. (3)] and the projected cell area ${S_{\rm{cell}}}$:

 

Fig. 2. Digital holographic microscope based on Mach–Zehnder interferometer architecture: (a) DHM configuration and (b) White light configuration. C1, C2, collimator lenses; M1, M2, M3, mirrors; MO1, imaging microscope objective; MO2, beam-expanding microscope objective; TL1, TL, tube lenses; S, sample on automated stage; CCD. camera: and WL, white light LED.

Download Full Size | PDF

Phase variance was calculated from Eq. (4) and phase kurtosis was calculated from Eq. (5):

Phase variance shows how the values of the cell OPD is spread out and phase kurtosis corresponds to the extremity of deviations (or outliers) [28].

D. Experimental Setup of Digital Holographic Microscope

The DHM measurement system was based on the Mach–Zehnder interferometer architecture, as shown in Fig. 2(a). Illumination was supplied by a laser diode with wavelength $\lambda = 0.632\;\unicode{x00B5}{\rm m}$. The specimen was imaged using the microscope objective tube lens (MO-TL) system with a magnification $M = 21.3$. The reason for the nonstandard value of $M$ was twofold: The focal length of the TL was 200 mm instead of the 180 mm proposed by the manufacturer of the MO (Olympus LUCPLFLN20X) and the MO-TL module was slightly nonconfocal. The numerical aperture of the MO was equal to ${\rm{NA}} = 0.45$. The reference beam has been expanded with the MO2-TL2 module to match the diameter and intensity of the object beam. The holograms were captured with a camera with a pixel size of 3.45 µm. The microscope was also equipped with the brightfield guiding mode, as shown in Fig. 2(b).

The holograms were captured in the image plane with a spatial carrier frequency introduced by the controlled tilt of the reference beam to separate ${+}1$ and ${-}1$ diffraction orders. For this reason, we could retrieve the phase directly from the hologram by means of the Fourier transform method [45,46]. The retrieval algorithm performs fast Fourier transform and locates the highest energy component (omitting the DC component) that indicates the center of the ${+}1$ or ${-}1$ diffraction order in the Fourier domain. The ambiguity between the ${+}1$ and ${-}1$ is irrelevant, considering the assumption that the cells introduce a higher phase shift compared to the background. If the phase map contradicts the assumption, we simply take the negative of the final image. Finally, the algorithm truncates components beyond the bandwidth of the diffraction order and inverts the Fourier transform of the remaining spectrum. The bandwidth is calculated from NA, $M$, and $\lambda$ based on the mathematical model of DHM provided by Sánchez-Ortiga et al. [47]. The phase image used for further analysis is the phase component of the inverted signal after processing with the unwrapping algorithm [48]. The retrieved phase image dimensions were $454 \times 378$ pixels, with a transverse resolution of 0.79 µm. Due to the nature of the DHM measurement (namely, the retrieval of integrated phase), the smallest substructures in the biological cells are not visible in the phase map; however, this is not due to low resolution but rather to low phase contrast of such overlapping structures observed in a single projection.

E. LLLT System

Cell irradiation was performed with a 10 W LED with a wavelength of 660 nm, 140° beam angle, and a luminous flux of 300 lm to 320 lm. The radiation power was measured in the sample plane using a photodiode power sensor (Thorlabs S121C). The measured optical power was $470\;{\rm{mW}} \pm 3\%$ and the optical power density was $500\;{\rm{mW}}\,{\rm{c}}{{\rm{m}}^{- 2}}$. An electronic system with a microcontroller has been designed to control the energy dose through changing the exposure time. Additional elements such as the device operation indicator LED and the start switch have also been added to the system. Their purpose is to facilitate and ensure the reproducibility of the irradiation process of cell samples.

 

Fig. 3. Analysis of SHSY-5Y dry mass density change for three irradiation values: $0\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ (blue), $5\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ (orange), and $20\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ (yellow). (a) Absolute dry mass density and (b) relative dry mass density change presented as a percentage change with respect to the first measurement.

Download Full Size | PDF

F. Cell Lines and Cell Culture

The proposed methodology was tested through investigations of two different cell cultures and measurement scenarios. The first experiment was devoted to investigations of SHSY-5Y neuroblastoma cells in a conventional cell culture. The second one was a wound healing experiment performed with HaCaT cells. For LLLT studies, the cells were grown in routine culture medium in the 35 mm ibi-Treat cell culture dish with polymer coverslip bottom and low wall (Ibidi, 80136). For the wound healing experiment, we used the Culture-Insert 2 Well in a µ-Dish 35 mm from Ibidi. During the measurements, the holographic microscope was equipped with a heating system (ibidi, 37°C) and the gas incubation system (ibidi, 5% ${{\rm{CO}}_2}$ at a flow rate of $10\;{\rm{L}}{{\rm{h}}^{- 1}}$ with 90% humidity).

1. SHSY-5Y

The SH-SY5Y cell line is the subline of the neuroblastoma SH-N-SH cell line. For routine culture, the SH-SY5Y cells were seeded in the amount of $2.5 \times {10^4}\;{\rm{cells/c}}{{\rm{m}}^2}$. They were cultured in a DMEM/F12 (1:1) medium (PAN Biotech, P04-41450) supplemented with 1% pen strep (Gibco, 15140-122), 2 mM L-glutamine (Gibco, 25030-024), and 10% heat-inactivated fetal bovine serum (FBS, Gibco, 10500-064). Cells were grown under standard conditions (37°C, 5% ${{\rm{CO}}_2}$, 90% humidity). The confluent cells were detached once a week using 1 ml of 0.25% trypsin-EDTA and seeded in a new Petri dish at the original density.

2. HaCaT

The HaCaT cell line is a noncancerous, immortalized human keratinocyte cell line. The cell cultures were incubated in a commercial incubator Galaxy 48R from Eppendorf at 37 °C, with a humid atmosphere with a 5% ${{\rm{CO}}_2}$. The medium chosen for breeding is Dulbecco’s Modified Eagle Medium (DMEM) with a high content of glucose (4.5 g/l), a stable form of L-glutamine, without sodium pyruvate (DMEM high glucose, with a stable glutamine w/o sodium pyruvate) with 10% FBS and 1% antibiotics (streptomycin at 0.1 mg/ml and penicillin at 100 U/ml). The dishes used for culturing were a Petri dish ($\mu$-Dish 35 mm high from Ibidi). Cells were cultured, passaged, and measured under sterile conditions: All dish openings were carried out in a laminar chamber.

3. RESULTS AND DISCUSSION

In this section, an experimental verification of the approach is shown. The holograms during the time-lapse measurement were recorded every 10 min for 18–20 h. For sufficient statistical analysis, the SFoV comprised of the arrays of $7 \times 6$ single FoVs for SHSY-5Y cells and $5 \times 5$ single FoVs for HaCaT cells phase maps, with a 7% overlap between neighboring frames. SHSY-5Y cell cultures were exposed to $5\;{\rm{J}}\,{\rm{c}}{{\rm{m}}^{- 2}}$ and $20\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ doses. HaCaT cell culture was exposed to a $5\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ dose.

Figure 3(a) shows the results from experiment with two irradiation doses on a SHSY-5Y neuroblastoma cell and its mean dry mass density changes during 18 h of measurements in the whole SFoV. It should be noted that the starting dry mass value differs for the three measurements. This is due to the variation between the cell number in the studied FoVs and a certain time variation that occurs during cell culture preparation and measurement; namely, the time between the moment when the cells are cultured and ready for the measurement and the measurement itself. To present the influence of LED irradiation on cells independently of the starting dry mass density, the relative values presented as a percentage change of the density with respect to the first measurement are presented in Fig. 3(b). The results show that the dry mass for all series (two irradiated series and the control group) has an increasing trend with the same rate until $t = 4h$, then the increase in the control group is faster compared to the irradiated cells. In each case, the shape of the plot is similar and shows a clear peak located approximately in the middle of the total time of the experiment. However, the higher the irradiation dose, the later the peak occurs. The values of the average dry mass are marked in Fig. 3(a) and the time points at the peak are marked in the Fig. 3(b). In addition, each plot shows an increasing trend toward the end of the measurement (for $0\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ starting at $12h$, for $5\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ at $11h$, and for $20\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ at $15h$). It should be noted that the cells after irradiation of $5\;{\rm{J}}\;{\rm{c}}{{\rm{m}}^{- 2}}$ have the least distinctive increasing trend, which can be also seen as the most fuzzy peak in the global plot (Fig. 3).

To check if the described trends are also confirmed in the behavior of selected cells, the local values of the mean dry mass density as a function of time are shown in Fig. 4 for the selected small regions of interest (ROIs). In the top row, the phase images of the whole synthetic FoV and the selected ROI measured at $t = 0h$ are presented. For each illumination dose and for each individual ROI, the same trend in the obtained plots can be observed as for global statistical results, as shown in Fig. 3(a).

 

Fig. 4. Analysis of local mean dry mass density. Top row: phase images in ${\rm{t}} = {{0}}\;{\rm{h}}$, with selected ROIs; middle row: mean dry mass density plot for selected ROIs; and bottom row: phase images in ${\rm{t}} = {{18}}\;{\rm{h}}$. The vertical red lines indicate the position of the middle peak in Fig. 3.

Download Full Size | PDF

Each individual ROI was also used to calculate phase statistical parameters: phase variance and kurtosis (Fig. 5). Figure 4 shows graphs with the changes in cell dry mass over time under irradiation with different doses in small, individual ROIs. These changes will also be visible in the phase statistical parameters. The increase in cell dry mass values following the trend noted and described in Fig. 3, which can also be seen in the graphs in Fig. 4, is reflected in the plots of the phase statistical parameters. Following the previously described trend, we can observe an increase in the phase variance (red lines in Figs. 4 and 5). The kurtosis value for normal distribution according to Eq. (5), which was used to calculate phase kurtosis, is 3. In the beginning of the measurement (up to $t = 5h$), the kurtosis values of all three measurements are close to the result for the normal distribution, but then it can be observed in the case of $5\;{\rm{Jc}}{{\rm{m}}^{- 2}}$ and $20\;{\rm{Jc}}{{\rm{m}}^{- 2}}$ that the kurtosis increases in the region of the peak observed in Fig. 3, which means that the intensity of the extreme values is higher. This may be interpreted as a condensation of OPD value distributions of the cells and linked to the process of cell division or death, when the cells change their geometry and become smaller, spherical, and therefore more dense.

 

Fig. 5. Statistical phase parameters over time in individual, small ROI for each three irradiation doses: (a) $0\;{\rm{Jc}}{{\rm{m}}^{- 2}}$, (b) $5\;{\rm{Jc}}{{\rm{m}}^{- 2}}$, and (c) $20\;{\rm{Jc}}{{\rm{m}}^{- 2}}$. The vertical red lines in phase variance graph correspond with the position of the red line in middle row in Fig. 4.

Download Full Size | PDF

 

Fig. 6. Projected area change analysis. (a) and (e) Projected cells area change in SHSY-5Y culture with and without light exposure over time. (c) Four phase maps from time points marked with red dots. (b) and (f) Projected wound area change in HaCaT wound healing assay with and without exposure over time. (d) Four phase maps from time points marked with red dots.

Download Full Size | PDF

Another important element in the analysis of the effect of LLLT treatment on cells is the observation of changes in the surface occupied by the cells over time. Figures 6(a) and 6(e) illustrates the projected area change in the SHSY-5Y cells with and without light exposure over time and Fig. 6(c) has four phase maps from time points marked with red dots. Both graphs show an initial decrease in the area, then a minimum and again an increase in the projected cell area. In both graphs, the minimum appears around $t = 8h$, and the maximum around $t = 12h$. Contrary to the analysis of cells in regular cell culture, a change in the wound surface was used to analyze the growth of HaCaT cells in wound healing assay, so that the speed of wound overgrowth and the effect of LLLT on this parameter was calculated. Figures 6(b) and 6(f) illustrate the wound area change in a HaCaT wound healing assay with and without light exposure over time and Fig. 6(d) presents four phase maps from time points marked with red dots. In the case of the unexposed sample, there was no full wound overgrowth, the overgrowth speed was $0.1\,\,{\rm{mm}}^2/{\rm{h}}$, and around $t = 8h$ the cells stopped growing, as cshown in Fig. 5(b). The sample irradiated with $5\;{\rm{Jc}}{{\rm{m}}^{- 2}}$ was completely overgrown and the rate of overgrowth was also $0.1\,\,{\rm{mm}}^2/{\rm{h}}$.

 

Fig. 7. Phase image of HaCaT cells from wound healing assay: (a) single field of view; (b) large field of view, stitched from ${{5}} \times {{5}}$ array of single FoV holograms.

Download Full Size | PDF

 

Fig. 8. Tracking cell motility and migration. Top row: Three time points from 18 h time-lapse measurement of SHSY-5Y cells with group of cells moved from one FoV to another marked by yellow rectangle. Bottom row: Three time points from 18 h wound healing assay of HaCaT cells after LLLT treatment. Red lines: Borders of single FoV from DHM measurement.

Download Full Size | PDF

While conducting biological experiments, we focused on the challenges in methodology for LLLT-influenced cellular research using DHM. The first issue is associated with long-term cell measurement. To achieve reliable results, it is important to limit the exposition of the sample to a DHM laser light. Otherwise, in addition to the planned irradiation of the specimen by the LLLT light source, the influence of a source with a different wavelength should be taken into account, which highly complicates the analysis of the obtained results. To address this issue, in the proposed approach the laser beam is shut with a mechanical shutter in between consecutive measurements and the laser power is limited. As a result, during a single measurement, the laser irradiates the cells with a dose of $0.01\;{\rm{Jc}}{{\rm{m}}^{- 2}}$. However small, this value may look like, it must be stressed out that during the whole experiment the cells are exposed to this energy multiple times. In the case of our experiments, where images were captured every 10 min for 18 h, the total energy density provided by the DHM laser equals $1.08\;{\rm{Jc}}{{\rm{m}}^{- 2}}$, which is still a value significantly smaller than the energy density provided by the LLLT system. It also is still possible to go down with this value through further limitation of the laser power and exposition time.

Another challenge is associated with the analysis of cells with near-full confluence, like in the wound healing experiment, where LLLT therapy is widely applied [49,50]. The cells used in such experiments are most frequently forming monolayers, as shown in Fig. 7(a). In such cases, it is impossible to directly measure the phase difference between the cells and the background, since in numerous FoVs there is no background. To mitigate this problem, a proper processing is needed to establish a common background for all phase images (Section 2.A).

One of the important steps in the proposed methodology is the generation of a SFoV for each time step. The reason to analyze the SFoV instead individual FoVs is that the overlapping regions are needed to establish the common baseline mentioned above. If the images are now analyzed independently without stitching, the presence of those duplicate regions would cause the problem of doubly counting the analyzed features of the culture, such as dry mass, area, or the number of individual cells. Once the baseline correction is performed, the actual generation of SFoV is done equivalently to the stitching of images captured by nonquantitative microscope modalities. Hence, this step relies on well-automated open source solutions such as BigStitcher [38]. It should be noted that it is possible to avoid this procedure, either by switching to low magnification MO or by using a high space-bandwidth product QPI modality like lensless holographic microscopy [51] or Fourier ptychography [52]. However, the drawback of the former is lower resolution, while the latter are not available currently as commercial products, contrary to conventional DHM microscopes [53] and other conventional QPI solutions like wavefront sensors [54].

Another important reason for stitching FoVs is the analysis of cell motility and migration. Cellular motility is the spontaneous movement of a cell from one location to another by the consumption of energy and is regarded as random cell movement occurring in almost every cell culture, while cell migration is a response to a cell attractant or repellent. Cell motility is required for many important physiological processes during cell development, such as for tissue regeneration or cell aggregation. Unregulated cell migration can be the cause for the progression of cancers (e.g, during metastasis). Experiments requiring cell tracking or observation of cell mobility need measurements over a large FoV. Observing only a single FoV over time carries the risk that the cell we are observing leaves the FoV (Fig. 8, top row). In addition, without stitching the images, we would not be able to reliably observe the wound healing process, as shown in the bottom row of Fig. 8.

4. CONCLUSIONS

Through our study, we show that DHM can be successfully used as the measurement tool to quantify the effects of light irradiation on biological cells. The proposed method allows us to analyze the effects of different light doses on the long-term behavior of different cell types. The statistical, quantitative analysis requires the careful preprocessing of the phase data. In particular, we proposed enhanced procedures for phase data stitching and their segmentation for cases of cell cultures with low and full confluence. We present the full methodology to create correct synthetic aperture phase maps, which could be further analyzed for highly accurate statistical analysis. The presented preliminary experiments and analysis of biological cells parameters showed changes in cell dry mass density and cells projected area under the influence of different irradiation doses. The results obtained during this study are very promising; however, it is crucial to carry out further systematic studies on the effects of LLLT on cell parameters using quantitative phase imaging with longer time lapse experiments to observe effects of LLLT on the full cell cycle. In addition, several of the challenges described in our work, such as the minimization of laser irradiation of cells during measurement, the automation of segmentation procedures, and the comparable and reproducible process of cell culture preparation for statistical comparisons of cell cultures, require attention and research for possible further solutions.

Minimization of additional irradiation of cells during measurement by lowering laser power, additional gray filters or shortening the measurement time will allow more precise analysis of LLLT therapy results, while the use of advanced segmentation procedures based on neural networks and machine learning will allow automation of the process, speed up the analysis and increase the accuracy of the segmentation [55,56]. With the use of these techniques, it would be also possible to perform segmentation of subcellular structures and analyze the effect of irradiation on individual internal organelles [57]. By gaining the ability to analyze the cell nuclei under LLLT therapy, it would be possible to more precisely analyze and describe the process of cell proliferation and differentiation under irradiation.

Further experiments should be directed toward extended systematic measurements of different irradiation doses and wavelengths; however, the proposed methodology provides a basis for many other biological experiments using quantitative phase imaging, such as drug testing or an analysis of the effect of other external factors such as fluorescent staining or sample fixation.