Optimized Detection of Hypoglycemic Glucose Ranges in

Human Serum by Raman Spectroscopy with

532 nm Laser Excitation

Ata Golparvar

1,* a

, Assim Boukhayma

1,2

, Christian Enz

and Sandro Carrara

Integrated Circuit Laboratory, École Polytechnique Fédérale de Lausanne (EPFL), CH-2002 Neuchâtel, Switzerland

Senbiosys SA, CH-2002 Neuchâtel, Switzerland

Keywords: Blood Plasma, Continuous Monitoring, Glucometer, Glucose Monitoring, Human Serum Analysis,

Non-invasive, Raman Effect, Vibrational Spectroscopy.

Abstract: Raman scattering-based biomedical detection has usually been proposed with near-infrared laser sources.

However, a low-cost CMOS imager’s quantum efficiency is optimum around green wavelength, and their

sensitivity substantially decreases in near-infrared wavelengths. Additionally, since Raman scattering

intensity is proportional to λ

-4

, where λ is the laser wavelength, the increase of wavelength directly results in

less sensitive measurement. These facts contribute to limiting the transfer of detection methodologies based

on Raman spectroscopy to portable and low-cost point-of-care medical devices. Therefore, here we propose

532 nm green laser-induced Raman spectroscopy for low human serum glucose level detection. However, in

532 nm Raman spectroscopy of carotenoid containing biological systems, such as human serum, resonance

Raman occurs, and total carotenoids resonance bands dominate the spectra. To demonstrate serum glucose

detection on concentration levels typical in severe hypoglycemic ranges, this study optimizes laser focal depth,

laser excitation duration, and laser power to extend the sensitivity by exploiting the glucose Raman shift peak

at 1125 ± 7.5 cm

–1

. By applying experimentally tuned parameters, our findings suggest sensitive detection of

serum glucose in the range of 0–10 mmol/l with 1.2 mmol/l theoretical limit of detection (LOD) by using

spontaneous (non-enhanced) Raman spectroscopy.

1 INTRODUCTION

The number of diabetes patients has increased

significantly (B. Zhou et al., 2016), and “diabetes

management” has been a severe public health burden

expected to double by 2030 compared to 2015

statistics (Bommer et al., 2018). Diabetes mellitus is

a chronic disorder that impairs glucose homeostasis

(Chege, Birech, Mwangi, & Bukachi, 2019). To

survive, diabetic patients must prevent its severe

secondary complications by frequent monitoring to

keep the glucose level under control through adequate

insulin injection (Zimmet, Alberti, & Shaw, 2001).

To monitor blood glucose levels reliably, patients

depend on frequent finger-prick tests to draw out

capillary blood, which is painful and inconvenient

with the potential cross-contamination risk when the

https://orcid.org/0000-0002-1107-6380

* Corresponding author

lancet is reused or not properly sterilized (Ju et al.,

2020). Additionally, finger pricking is closely related

to diabetes burnout—a state of detachment from

diabetic care (Zimmet et al., 2001), directly related to

diabetes-induced morbidity and mortality (Abdoli,

Hessler, Vora, Smither, & Stuckey, 2020). Therefore,

optical detection techniques have been proposed for

non-invasive glucose monitoring and reviewed

extensively (Smith, 2015). Among them, Raman

spectroscopy holds great potential thanks to its high

specificity due to the unique chemical “fingerprint”

signature of inelastic scattering of photons from each

specific analyte (Singh, Goh, Canzoneri, & Ram,

2015). Furthermore, it promises an excellent

alternative for rapid, label‐free, and non-invasive

detection in biomedical applications (Lawson, Barry,

Williams, & Edwards, 1997), such as continuous

glucose concentration monitoring in human tissue

158

Golparvar, A., Boukhayma, A., Enz, C. and Carrara, S.

Optimized Detection of Hypoglycemic Glucose Ranges in Human Serum by Raman Spectroscopy with 532 nm Laser Excitation.

DOI: 10.5220/0010981300003121

In Proceedings of the 10th International Conference on Photonics, Optics and Laser Technology (PHOTOPTICS 2022), pages 158-165

ISBN: 978-989-758-554-8; ISSN: 2184-4364

due to its water-insensitive probing (Gulyamov et al.,

2021; Kang et al., 2020). Raman effect occurs when

a molecule interacts with an incident photon and is

driven to a virtual state. With a slim chance the

molecule relaxes to a vibrational state other than the

original ground state, due to the quantum energy

exchange with the incident photon and the molecule’s

vibrational modes, dictated by their unique energy

difference (Krishnan & Shankar, 1981).

Although spontaneous Raman offers a small

scattering cross-section, it is still more robust for

glucose level estimation than other absorption-based

vibrational spectroscopy techniques due to the

water’s inadequate scattering response but high

absorbance signature (Li, Deen, Kumar, &

Selvaganapathy, 2014). However, since Raman

scattering is a weak process, its application to low

glucose concentration (i.e., below 5 mmol/l) detection

is problematic and to perform such measurements,

often long acquisition time or high laser powers are

needed. On the other hand, to enhance the Raman

effect, which may indeed lower the measurement’s

required acquisition time and laser power, techniques

such as plasmonic surface enhancement, resonance

Raman exploitation, and non-linear coherent process

have been suggested over the years (Kiefer, 2007; Li

et al., 2014). Even though surface enhancement is

most effective with metallic nanoparticle surfaces, the

subcutaneous injection of metallic materials can

produce toxicity (Asharani, Wu, Gong, &

Valiyaveettil, 2008) and require surgical implant

placement (Stuart et al., 2006), which removes the

technique from the non-invasive application. On the

other hand, resonant Raman spectroscopy usually

requires a low wavelength in ultra-violate ranges for

the excitation laser. However, such low wavelengths

hold risks of photochemical damage to the tissue

being investigated (Mhlanga, Tetyana, Nyembe, &

Sikhwivhilu, 2021), and thus this hinders the

technique from the non-invasive application as well.

Nonetheless, coherent Raman scattering might be

favored in the light of newly emerged technologies on

the side of cheap and low noise CMOS image sensors

(Boukhayma, 2018).

Even though spontaneous Raman has not been

primarily considered for low glucose level detection,

mainly due to the expressed reasons, we propose

spontaneous Raman spectroscopy to address the need

to improve the sensitivity toward low concentration

range. In fact, in the absence of any Raman

enhancement, the excitation laser wavelength

selection is perhaps the most critical design

parameter. In particular, Raman scattering intensity is

proportional to λ

-4

, where λ is the laser wavelength

(Šugar & Bouř, 2016). Due to that, 532 nm green laser

is theoretically 4.7 times more efficient than 785 nm

near-infrared laser and theoretically 16 times better

efficiency than a 1064 nm infrared laser source. Of

course, blue or violate lasers are even better than

green lasers, but they are still costly (Greer, Petrov, &

Yakovlev, 2013), and the quantum efficiency of low-

cost silicon-CMOS imagers decays rapidly below the

green wavelengths (Boukhayma, Peizerat, & Enz,

2016).

Recent investigations have shown that human

serum does not produce strong autofluorescence to

completely mask the Raman spectra in the visible

range with 532 nm green lasers and can be filtered out

using the available chemometric tools, different from

what is observed already with 660 nm red laser

(Medipally et al., 2017). Thus, acquiring 532 nm is

highly advantageous for improving the intensity of

the Raman scattering measurement. Furthermore, a

532 nm laser choice is optimum due to the highest

quantum efficiency of the CMOS imagers in the

green wavelengths (Wróbel, 2016), and therefore a

much better solution from the perspective of portable

Raman devices for glucose sensing. To succeed on

that, we need to consider the interference by

carotenoids, which resonance Raman shift is at 1153

-1

, extremely close to that of the glucose that shows

its characteristic band around 1125 cm

-1

. Therefore,

this study deeply investigated the possible

optimizations on laser focal depth, laser excitation

duration, and laser power to optimize especially for

serum glucose levels in severe hypoglycemic ranges

by exploiting the Raman shift peak of 1125 ± 7.5 cm

–

with spontaneous and non-enhanced Raman

spectroscopy.

2 METHODOLOGY

2.1 Sample Preparation

D-(+)-glucose powder (C

, purity ≥ 99.5%) and

human serum solution (male AB plasma, sterile-

filtered, stored in -20 °C) were purchased from

Sigma-Aldrich (MilliporeSigma, MO, USA).

Reagents were analytical grade and were used as

received. Glucose stock solutions with concentrations

of 200 mmol/l and 100 mmol/l (25 ml each) were

prepared to dilute and spike the serum’s glucose

concentration to various amounts to explore the

dynamic range and sensitivity of the measurements.

The powder was carefully measured with a highly

precise scale and wholly dissolved in nanopure water.

The concentrations were selected to cover a wide

Optimized Detection of Hypoglycemic Glucose Ranges in Human Serum by Raman Spectroscopy with 532 nm Laser Excitation

159

range of human blood serum glucose levels to

simulate normal and unstable conditions (i.e.,

hypoglycemia or hyperglycemia) as well as to more

extensive concentration ranges (up 100 mmol/l) to the

straightforward demonstration of the detection

principle. Hypoglycemia was defined as a blood

glucose level smaller than 3.9 mmol/l (70 mg/dl),

whereas hyperglycemia was identified when it is

above 10 mmol/l (180 mg/dl) (Brinati et al., 2021).

Therefore, glucose stock solutions were diluted to

prepare 1.5 mL spiked serum solutions (0.75 mL

glucose mixed with 0.75 mL serum) with overall

glucose concentrations of 1–10 mmol/l (18–180

mg/dl) and 20-100 mmol/l (360–1800 mg/dl) with

intervals of 1 mmol/l and 20 mmol/l, respectively,

and were refrigerated overnight. Each sample was

first stirred during the measurement session, and then

a 20-μl droplet of each liquid was placed into a

concave glass microscope slides with well depths of

~ 800 μm (Electron Microscopy Sciences PA, USA)

using a micropipette (Gilson International, France).

2.2 Data Acquisition and Optimization

In the backscattered configuration, the Raman spectra

of human serum solutions were obtained with

confocal micro-Raman microscopy (LabRAM HR,

Horiba, Japan), exploiting the spectral region of 200–

1900 cm

–1

using a liquid-nitrogen-cooled CCD

camera. The excitation source was a 532 nm single-

frequency green laser (Cobolt 05, Hubner Photonics,

Germany). Different laser powers varying from 0.4

mW to 400 mW and different acquisition durations

changing from 10 s to 180 s were tested to optimize

the best sensitivity within the fastest acquisition time

for low glucose level detection in human serum. To

further optimize the Raman scattering intensity, the

filtered beam was focused to the surface as well as to

the 200 μm, 400 μm, and 600 μm below the surface

of the droplet using a long working distance ×50

objective lens with NA of 0.50 (LMPLFLN, Olympus

Corporation, Japan). The beam quality was M2 < 1.1,

beam diameter (1/e²) at the objective input was 2 mm,

and objective lens focal length was 180 mm. The

spectrometer was adjusted to groove density of 600

g/mm, the slit size of 100 μm, and the confocal hole

size of 200 μm. Higher grating values increase the

spectral resolution while decreasing spectral

coverage, and larger slit and confocal hole size

increase the intensity at the cost of spectral resolution

(F Adar, Lee, Mamedov, & Whitley, 2010; Tuschel,

2020). The spectral resolution of this univariate

analysis study is critical since the targeted glucose

1125 cm

-1

band is essentially a single shoulder peak.

Therefore, the spectral resolution should be high

enough to identify the main Raman peak of interest

by the nearby prominent carotenoids resonance 1153

-1

peak. At the same time, Raman scattering

intensity is crucial to obtain high sensitivity in low

concentration levels. Therefore, we decided to select

a configuration to maintain a fair balance between

spectral resolution and Raman intensity. Calibration

of the spectrometer was carried with 520 cm

–1

characteristic peak of silicon. Three consecutive

spectra were obtained using different droplets to

compute the measurement error, and no photodamage

or photo-degradation was observed. Throughout the

experiments, the room was dark, and 24°C was

maintained.

All data processing was performed with Origin

(OriginLab Corporation, MA, USA). For each

spectrum, the autofluorescence induced baseline was

subtracted using the asymmetric least-square fit

(asymmetric factor 0.001, threshold 0.02, smoothing

factor 5, and iteration 10), then the Savitzky-Golay

filter (polynomial order 3, window length 13) was

applied to smooth the spectrum further (Zimmet et al.,

2001). The absolute area under the Raman shift peak

of 1125 ± 7.5 cm

–1

was integrated to perform the

univariate analysis and predict the serum glucose

level, and linear regression fit was used to draw the

calibration curve and calculate the measurement

sensitivity.

3 RESULTS AND DISCUSSION

Figure 1a illustrates the processed Raman scattering

spectra of human serum spiked with glucose, pure

aqueous glucose, and pure serum solutions. The

intensities of Raman scattering bands are directly

proportional to the concentration of solution analytes,

and for aqueous glucose, this is typically observed in

multiple peaks around 437, 518, 1060, 1125, and

1365 cm

–1

bands (Figure 1a, blue line) (Wang,

Mizaikoff, & Kranz, 2009). It has previously been

reported that selecting the spectral region of 1030 to

1400 cm

−1

improves the sensitivity for the glucose

prediction model (Parachalil et al., 2019). However,

our previous study shows that only the 1125 cm

–1

band is sensitive enough for pathophysiologically

relevant low glucose level detection and other peaks

are highly disrupted below 5 mmol/l. Thus we

concluded that univariate data analysis can preferred

over multivariate analysis (Golparvar et al., 2021).

The glucose Raman shift peak of 1125 cm

–1

has been

assigned to C-O-H bonds’ bending mode or C-O-C

bonds’ antisymmetric stretching mode (Dudek et al.,

PHOTOPTICS 2022 - 10th International Conference on Photonics, Optics and Laser Technology

160

Figure 1: (a) Processed (i.e., background subtracted and

smoothed) 532 nm green laser-induced Raman scattering

spectra of spiked human serum with 100 mmol/l glucose

(red line), human serum without glucose spike (green line),

and 100 mmol/l glucose aqueous solution (blue line). The

Raman spectrum of the mixture (serum + glucose) consists

of the superposition of the individual spectrum of each

solution (red line). Laser power and acquisition time were

adjusted to 200 mW and 1 minute, respectively.

Characteristic peaks of glucose (blue bands) and serum

(green bands) are highlighted. (b) The calibration curve for

glucose concentration prediction is computed by the area

under the curve at 1125 ± 7.5 cm

–1

; three consecutive

measurements return the standard deviation error bars.

2019; Fujihara, Nishimoto, Yasuda, & Takeshita,

2019). Although our previous study is based on

measurement acquired only on water-based solutions

(not human serum), it is imperative to address the

issue of measuring low concentrations in serum

successfully. On the other hand, when induced by a

785 nm near-infrared laser, the strongest serum

Raman bands in the fingerprint region appear around

820, 1044, 1335, 1383, 1442, and 1542 cm−1, and

they are widely associated with CH2 and CH3 groups

(i.e., lipids and proteins) (Huang et al., 2011).

However, when human serum is excited by a 532 nm

green laser, the Raman peaks of α- and β-carotenes,

which in total their concentration is only a few

hundred nmol/l in blood serum, resonate and

dominate the spectrum (Figure 1a, green line) (Bohn,

2018; Medipally et al., 2017).

Typically, if the laser excitation frequency is

close to the frequency of the electronic transition of a

molecule, resonance Raman occurs and enhances the

otherwise spontaneous Raman effect (Schmitt &

Popp, 2006). This well-known phenomenon is

traditionally used to study molecules in extremely

low concentrations (X. Zhou et al., 2019).

Carotenoids’ electronic absorption band is

anticipated to be between 400 nm and 550 nm, and

thus, 532 nm laser energy lies close enough to its

electronic transition to trigger the resonance Raman

effect (Fran Adar, 2017). However, in serum glucose

detection, this is an unwanted enhancement and the

very stable 1153 cm

−1

resonant Raman band of total

carotenoids interferes with the targeted glucose 1125

–1

band. Although for univariate analysis of high

glucose levels (10–100 mmol/l), this is not an issue,

and the calibration curve is linearly fit with an R

value of 0.99 and sensitivity of ~ 242 counts/mM, in

lower glucose levels (1–8 mmol/l), the detection is

limited. In high glucose concentrations, the 1125 cm

–

peak is strong and distinguishable as same as the

1153 cm

−1

peak (Figure 1a, red line), but when the

intensity of glucose 1125 cm

–1

peak decreases due to

the decrease in concentration, it becomes a shoulder

peak to the 1153 cm

−1

and

completely vanishes below

~ 8 mmol/l. As a result, the measurement should go

through specific optimizations to detect low serum

glucose levels with 532 nm green laser-induced

spontaneous Raman spectroscopy.

The optimal focus depth of the laser beam in the

solutions affects the number of received inelastically

scattered photons and can be tuned to increase

detection sensitivity (Dubessy, Lhomme, Boiron, &

Rull, 2002). This is validated using four focus depths

while concentration, laser power, and acquisition

time were kept constant (Figure 2a). Furthermore,

different droplets of the same solution were used each

time to ensure the intensity difference between

measurements is not induced by heat changes in the

region of focus due to the laser power. By fine-tuning

the z-axis of the objective stage, the focus depth

changed. Figure 2a illustrates that the laser beam

yields the lowest Raman intensity when focused on

the aqueous glucose solution’s surface (depth = 0

um). Instead, the intensity slightly decreases with the

focus depth below the solution’s surface (e.g., more

scattering at a depth of 200 um compared to 400 um).

Optimized Detection of Hypoglycemic Glucose Ranges in Human Serum by Raman Spectroscopy with 532 nm Laser Excitation

161

Figure 2: Optimizations for univariate detection of low

glucose concentration using merely 1125 cm

-1

band from

human serum with 532 nm green laser-induced Raman

spectroscopy in interference with carotenoids’ resonance

Raman band of 1153 cm

-1

. Intensity dependency of 1125

-1

band to (a) depth of focus variation, (b) acquisition

time, and (c) laser power variation.

On the other hand, increasing the laser power and

the acquisition time increases the sensitivity of

Raman spectra (Braun et al., 2016). Figure 2b

illustrates

the processed Raman spectra of 5 mmol/l

Figure 3: Unprocessed Raman scattering spectra of 532 nm

green laser-induced human serum with glucose

concentrations ranging from 10 mmol/l to severe

hypoglycemic ranges and nanopure water. The depth of

laser focus is 200 μm inside the solution droplet, the

excitation duration is 3 minutes, and the laser power is 400

mW. The highlighted characteristic Raman peak of glucose

1125 cm

-1

appears as a small shoulder to the strong

carotenoid resonance peak of 1153 cm

-1

but is still

distinguishable due to the applied optimizations. Offsets are

added for clarity.

glucose spiked serum with 400 mW of power with

different acquisition times varied from 10 s to 180 s

in the signature region. Data show that the signal by

glucose at 1125 cm

-1

stays stable, although the strong

total carotenoid resonance peaks saturate in the

acquisition time of 180 s. Figure 2c illustrates the

processed Raman spectra for the same samples but

with an acquisition time of 120 s and different laser

power varied from 0.4 mW to 200 mW. Figure 2c

shows that the 200 mW is not enough to detect low

glucose levels in serum because of the interference

with the total carotenoids resonance band.

Figure 3. illustrates the unprocessed serum

Raman spectra with glucose concentrations ranging

from 0 mmol/l to 10 mmol/l obtained by considering

all the discussed optimization (depth of laser focus

200 μm, excitation duration 3 minutes, and laser

power 400 mW). The highlighted glucose Raman

shift peak is stable and can be used in the univariate

analysis, calculating merely the area under the curve

at 1125 ± 7.5 cm

–1

, even though the 1153 cm

–1

and

1512 cm

–1

resonance carotenoid pecks saturated

already at 4 mmol/l of glucose in the serum.

Figure 4a. illustrates the filtered spectra from

Figure 3 in the region of interest, as highlighted, and

Figure 4b shows the obtained calibration curve by

recording an excellent linear fit (randomly scattered

PHOTOPTICS 2022 - 10th International Conference on Photonics, Optics and Laser Technology

162

data in the residual plot was recorded) with R

value of

0.98, the sensitivity of ~ 2606 counts/mM, and root

mean square of the error or residual standard deviation

of 0.6. The theoretical limit of detection (LOD) is

calculated with (1) in accordance with recommenda-

tions by the International Union of Pure and Applied

Chemistry (IUPAC) definitions (Stacey, Mader, &

Sammon, 2017; Vandenabeele & Moens, 2012).

LOD = Kẟ/S ≈ 1.2 mmol/l (1)

Where K is the confidence coefficient (usually K=3

with a confidence level of 99.86%), ẟ is the standard

deviation of the blank measurement (here ~ 1066.9

counts), and S is the slope of the calibration curve

(here ~ 2606 counts/mM).

Figure 4: (a) Processed Raman scattering spectra of human

serum with glucose concentrations below 10 mmol/l and

water; the characteristic peak under analysis is highlighted.

The visual evaluation of the Raman shift around 1125 cm

–1

indicates the glucose level increase as a function of its

concentration. (b) The calibration curve for glucose

concentration prediction is computed by the area under the

curve at 1125 ± 7.5 cm

–1

; three consecutive measurements

return the standard deviation error bars. Sensitivity ≈ 2606

counts/mM and LOD ≈ 1.2 mmol/l.

4 CONCLUSION

The 532 nm green laser-induced autofluorescence

appears as background noise in the human serum

Raman spectra but can be filtered out. Therefore, it is

feasible to acquire 532 nm excitation source for

glucose detection, while it appears highly

advantageous to improve the efficiency of the Raman

scattering (up to almost 16 times) with respect to

near-infrared wavelengths. Furthermore, 532 nm

laser choice is optimal for low-cost applications with

CMOS imagers. In fact, their highest quantum

efficiency (QE) at this wavelength is increased by a

further 37.5%: i.e., typically CMOS imagers QE at

532 nm is ~ 95% while is only 65% at 785 nm). Of

course, total carotenoids resonance appears when 532

nm laser is utilized in spontaneous Raman

spectroscopy on full human serum. Although this is

not an issue for high glucose concentrations (e.g., in

the case of hyperglycemia monitoring), this

resonance by carotenoids limits the detection of lower

glucose levels (below 10 mmol/l, or 180 mg/dl).

These levels are of top importance to monitor because

they typically correspond to those of hypoglycemic

patients, indeed.

Therefore, this study investigated possible

optimizations on laser focal depth, laser excitation

duration, and laser power to extend the sensitivity of

the measurement to low glucose ranges in order to

open the monitoring of hypoglycemic ranges below

2.2 mmol/l (40 mg/dl), usually neglected in

applications of Raman detection of glucose for the

limitations we have described in this paper. Indeed,

the calibration curve we have recorded on serum

samples by applying the mentioned optimized tuned

parameters was enough sensitive and linear in the

range of 0–10 mmol/l, with a limit of detection (LOD)

at 1.2 mmol/l only. Therefore, this study confirms that

Raman spectroscopy is useful to measure endogenous

compounds in lower concentrations also in the case

their peak are typically present close to the region of

Raman shift where carotenoid-interference is usually

present.

Future developments will be focused to directly

detect urea and lactate in lower concentration ranges

as well, in human serum, with 532 nm-induced

Raman spectroscopy, by merely analyzing their

single Raman signatures bands. Lactate has a

characteristic peak around 861 cm

-1

, and urea has a

strong characteristic presence around 1005 cm

-1

(Golparvar et al., 2021). Although lactate

characteristic peak hypothetically will not interfere

with any of the carotenoids resonance bands, urea’s

characteristic peak will interfere with the resonance

Optimized Detection of Hypoglycemic Glucose Ranges in Human Serum by Raman Spectroscopy with 532 nm Laser Excitation

163

band of total carotenoids at 1002 cm

-1

, and further

investigations should be carried out similarly to

optimize the selectivity for urea.

ACKNOWLEDGMENTS

This work was supported by the École Polytechnique

Fédérale de Lausanne (EPFL) research fund. The

authors gratefully thank Dr. Richard Gaal for highly

fruitful discussions about Raman spectroscopy.

REFERENCES

Abdoli, S., Hessler, D., Vora, A., Smither, B., & Stuckey,

H. (2020). Descriptions of diabetes burnout from

individuals with Type 1 diabetes: an analysis of

YouTube videos. Diabetic Medicine, 37(8), 1344-1351.

Adar, F. (2017). Carotenoids-their resonance Raman

spectra and how they can be helpful in characterizing a

number of biological systems. Spectroscopy, 32(6), 12-

20.

Adar, F., Lee, E., Mamedov, S., & Whitley, A. (2010).

Experimental evaluation of the depth resolution of a

Raman microscope. Microscopy and Microanalysis,

16(S2), 360-361.

Asharani, P., Wu, Y. L., Gong, Z., & Valiyaveettil, S.

(2008). Toxicity of silver nanoparticles in zebrafish

models. Nanotechnology, 19(25), 255102.

Bohn, T. (2018). Metabolic fate of bioaccessible and non-

bioaccessible carotenoids. Non-Extractable

Polyphenols and Carotenoids: Importance in Human

Nutrition and Health, 165-200.

Bommer, C., Sagalova, V., Heesemann, E., Manne-

Goehler, J., Atun, R., Bärnighausen, T., . . . Vollmer, S.

(2018). Global economic burden of diabetes in adults:

projections from 2015 to 2030. Diabetes care, 41(5),

963-970.

Boukhayma, A. (2018). Low-noise CMOS image sensors.

In Ultra Low Noise CMOS Image Sensors (pp. 13-34):

Springer.

Boukhayma, A., Peizerat, A., & Enz, C. (2016). A sub-0.5

electron read noise VGA image sensor in a standard

CMOS process. IEEE Journal of solid-state circuits,

51(9), 2180-2191.

Braun, F., Schwolow, S., Seltenreich, J., Kockmann, N.,

Röder, T., Gretz, N., & Rädle, M. (2016). Highly

sensitive Raman spectroscopy with low laser power for

fast in-line reaction and multiphase flow monitoring.

Analytical chemistry, 88(19), 9368-9374.

Brinati, L. M., de Fátima Januário, C., Balbino, P. C.,

Gonçalves Rezende Macieira, T., Cardoso, S. A.,

Moreira, T. R., & de Oliveira Salgado, P. (2021).

Incidence and Prediction of Unstable Blood Glucose

Level among Critically Ill Patients: A Cohort Study.

International Journal of Nursing Knowledge, 32(2), 96-

102.

Chege, B. M., Birech, Z., Mwangi, P. W., & Bukachi, F. O.

(2019). Utility of Raman spectroscopy in diabetes

detection based on biomarker Raman bands and in

antidiabetic efficacy studies of herbal extract Rotheca

myricoides Hochst. Journal of Raman Spectroscopy,

50(10), 1358-1366.

Dubessy, J., Lhomme, T., Boiron, M.-C., & Rull, F. (2002).

Determination of chlorinity in aqueous fluids using

Raman spectroscopy of the stretching band of water at

room temperature: application to fluid inclusions.

Applied spectroscopy, 56(1), 99-106.

Dudek, M., Zajac, G., Szafraniec, E., Wiercigroch, E., Tott,

S., Malek, K., . . . Baranska, M. (2019). Raman Optical

Activity and Raman spectroscopy of carbohydrates in

solution. Spectrochimica Acta Part A: Molecular and

Biomolecular Spectroscopy, 206, 597-612.

Fujihara, J., Nishimoto, N., Yasuda, T., & Takeshita, H.

(2019). Discrimination between infant and adult

bloodstains using micro‐Raman spectroscopy: A

preliminary study. Journal of forensic sciences, 64(3),

698-701.

Golparvar, A., Boukhayma, A., Loayza, T., Caizzone, A.,

Enz, C., & Carrara, S. (2021). Very Selective Detection

of Low Physiopathological Glucose Levels by

Spontaneous Raman Spectroscopy with Univariate

Data Analysis. BioNanoScience, 1-7.

Greer, J. S., Petrov, G. I., & Yakovlev, V. V. (2013). Raman

spectroscopy with LED excitation source. Journal of

Raman Spectroscopy, 44(7), 1058-1059.

Gulyamov, S., Shamshiddinova, M., Bae, W. H., Park, Y.

C., Kim, H. J., Cho, W. B., & Lee, Y. M. (2021).

Identification of biomarkers on kidney failure by

Raman spectroscopy. Journal of Raman Spectroscopy.

Huang, N., Short, M., Zhao, J., Wang, H., Lui, H., Korbelik,

M., & Zeng, H. (2011). Full range characterization of

the Raman spectra of organs in a murine model. Optics

express, 19(23), 22892-22909.

Ju, J., Hsieh, C.-M., Tian, Y., Kang, J., Chia, R., Chang, H.,

. . . Liu, Q. (2020). Surface enhanced Raman

spectroscopy based biosensor with a microneedle array

for minimally invasive in vivo glucose measurements.

ACS sensors, 5(6), 1777-1785.

Kang, J. W., Park, Y. S., Chang, H., Lee, W., Singh, S. P.,

Choi, W., . . . Park, J. (2020). Direct observation of

glucose fingerprint using in vivo Raman spectroscopy.

Science Advances, 6(4), eaay5206.

Kiefer, W. (2007). Recent advances in linear and non-linear

Raman spectroscopy I. Journal of Raman

Spectroscopy: An International Journal for Original

Work in all Aspects of Raman Spectroscopy, Including

Higher Order Processes, and also Brillouin and

Rayleigh Scattering, 38(12), 1538-1553.

Krishnan, R., & Shankar, R. (1981). Raman effect: History

of the discovery. Journal of Raman Spectroscopy,

10(1), 1-8.

Lawson, E., Barry, B., Williams, A., & Edwards, H. (1997).

Biomedical applications of Raman spectroscopy.

Journal of Raman Spectroscopy, 28(2‐3), 111-117.

Li, Z., Deen, M. J., Kumar, S., & Selvaganapathy, P. R.

(2014). Raman spectroscopy for in-line water quality

PHOTOPTICS 2022 - 10th International Conference on Photonics, Optics and Laser Technology

164

monitoring—Instrumentation and potential. Sensors,

14(9), 17275-17303.

Medipally, D. K., Maguire, A., Bryant, J., Armstrong, J.,

Dunne, M., Finn, M., . . . Meade, A. D. (2017).

Development of a high throughput (HT) Raman

spectroscopy method for rapid screening of liquid blood

plasma from prostate cancer patients. Analyst, 142(8),

1216-1226.

Mhlanga, N., Tetyana, P., Nyembe, S., & Sikhwivhilu, L.

(2021). Application of Raman Spectroscopy in

Biomedical Diagnostics. In Recent Developments in

Atomic Force Microscopy and Raman Spectroscopy for

Materials Characterization: IntechOpen.

Parachalil, D. R., Bruno, C., Bonnier, F., Blasco, H.,

Chourpa, I., McIntyre, J., & Byrne, H. J. (2019). Raman

spectroscopic screening of high and low molecular

weight fractions of human serum. Analyst, 144(14),

4295-4311.

Schmitt, M., & Popp, J. (2006). Raman spectroscopy at the

beginning of the twenty‐first century. Journal of Raman

Spectroscopy: An International Journal for Original

Work in all Aspects of Raman Spectroscopy, Including

Higher Order Processes, and also Brillouin and

Rayleigh Scattering, 37(1‐3), 20-28.

Singh, G. P., Goh, S., Canzoneri, M., & Ram, R. J. (2015).

Raman spectroscopy of complex defined media:

biopharmaceutical applications. Journal of Raman

Spectroscopy, 46(6), 545-550.

Smith, J. L. (2015). The pursuit of non-invasive glucose:

hunting the deceitful turkey. Revised and Expanded,

Stacey, P., Mader, K. T., & Sammon, C. (2017). Feasibility

of the quantification of respirable crystalline silica by

mass on aerosol sampling filters using Raman

microscopy. Journal of Raman Spectroscopy, 48(5),

720-725.

Stuart, D. A., Yuen, J. M., Shah, N., Lyandres, O., Yonzon,

C. R., Glucksberg, M. R., . . . Van Duyne, R. P. (2006).

In vivo glucose measurement by surface-enhanced

Raman spectroscopy. Analytical chemistry, 78(20),

7211-7215.

Šugar, J., & Bouř, P. (2016). Quantitative analysis of sugar

composition in honey using 532‐nm excitation Raman

and Raman optical activity spectra. Journal of Raman

Spectroscopy, 47(11), 1298-1303.

Tuschel, D. (2020). Spectral Resolution and Dispersion in

Raman Spectroscopy. Spectroscopy, 35, 9.

Vandenabeele, P., & Moens, L. (2012). Some ideas on the

definition of Raman spectroscopic detection limits for

the analysis of art and archaeological objects. Journal

of Raman Spectroscopy, 43(11), 1545-1550.

Wang, L., Mizaikoff, B., & Kranz, C. (2009).

Quantification of sugar mixtures with near-infrared

Raman spectroscopy and multivariate data analysis. A

quantitative analysis laboratory experiment. Journal of

chemical education, 86(11), 1322.

Wróbel, M. (2016). Non-invasive blood glucose monitoring

with Raman spectroscopy: prospects for device

miniaturization. Paper presented at the IOP Conference

Series: Materials Science and Engineering.

Zhou, B., Lu, Y., Hajifathalian, K., Bentham, J., Di Cesare,

M., Danaei, G., . . . Taddei, C. (2016). Worldwide

trends in diabetes since 1980: a pooled analysis of 751

population-based studies with 4ꞏ 4 million participants.

The lancet, 387(10027), 1513-1530.

Zhou, X., Qin, M., Zhu, J., Wang, C., Zhu, G., Wang, H.,

& Yang, L. (2019). Rapid and sensitive surface‐

enhanced resonance Raman spectroscopy detection for

norepinephrine in biofluids. Journal of Raman

Spectroscopy, 50(3), 314-321.

Zimmet, P., Alberti, K., & Shaw, J. (2001). Global and

societal implications of the diabetes epidemic. Nature,

414(6865), 782-787.

Optimized Detection of Hypoglycemic Glucose Ranges in Human Serum by Raman Spectroscopy with 532 nm Laser Excitation

165