Highly-selective optical filter for NADH fluorescence detection in multiphoton microscopy

Colorectal cancer (CRC) is a pressing global health concern, emphasizing the need for early detection tools. In this study an optical filter for precise detection of nicotinamide adenine dinucleotide (NADH) fluorescence via two-photon excitation fluorescence (TPEF) was developed. Fabricated with silicon dioxide and titanium dioxide thin films in a Fabry-Perot structure, the filter achieved a peak transmittance of about 95% at 483 nm, with a 12 nm full-width at half maximum. TPEF measurements using a tailored setup and NADH liquid phantoms underscored the filter's significance in selectively capturing NADH fluorescence while mitigating interference from other fluorophores. This work marks a substantial stride towards integrating multiphoton microscopy into conventional colonoscopy, enabling non-invasive, objective optical biopsy for colorectal tissue analysis. Further refinements of the experimental setup are imperative to advance tissue differentiation and enhance CRC diagnosis.


Introduction
Colorectal cancer (CRC) is the third most commonly diagnosed cancer and the second most deadly cancer in the world, which poses an increasing global public health challenge [1].The stage of cancer development at the time of diagnosis significantly impacts treatment outcomes and patient survival.Therefore, early cancer detection is an important factor in preventing metastasis, reducing mortality rates, and improving prognosis and quality of life.However, it is concerning that estimations indicate more than one-third of CRC patients receive their diagnosis after lymph node metastasis, which poses a potential challenge to patient survival [1][2][3].
Traditional clinical examinations of patients with suspect CRC include history taking, fecal occult blood test, colonoscopy (with or without contrast), and parallel biopsy [4,5].The conventional white-light colonoscopy is limited to the diagnosis of gross morphologic changes and the excision of the tissue for later biopsy.However, it may be insufficient for diminutive lesions or microscopic abnormalities among large fields of diffusive disease, and in these cases, a biopsy can be difficult and time-consuming [6,7].Consequently, there is a constant need for non-invasive and objective diagnosis tools for early cancer detection.
The precise risk stratification of colorectal polyps through endoscopic evaluation serves to enhance the quality and safety of patient care.This is achieved by mitigating the potential risks associated with unnecessary tissue removal for biopsy and reducing the overall procedure duration [8].The concept of "optical biopsy," which enables real-time and in vivo analysis of colorectal tissues, has gained considerable attention in recent years, primarily due to advancements in imaging technologies [8][9][10].Among these optical imaging technologies, fluorescence and multiphoton microscopy (MPM) have emerged as promising approaches, offering high spatial and temporal resolution.These techniques can be integrated in conventional colonoscopy for optical biopsy of the tissues [11,12].
MPM is a specific type of microscopy technique that harnesses the principles of multiphoton excitation (MPE) and second harmonic generation signals.A notable example of MPE is two photon excitation fluorescence (TPEF), which has witnessed substantial advancements in cancer research.TPEF employs a near-infrared laser and relies on a phenomenon where the absorption of two photons simultaneously excites a fluorophore molecule.Due to the requirement of two excitation photons, each carrying half the energy compared to traditional fluorescence techniques, TPEF uses higher-wavelength light.Consequently, TPEF enables deeper tissue penetration while minimizing phototoxicity and photobleaching effects [13,14].Squirrell et al. conducted a meticulous study, employing two-photon microscopy to monitor mitochondrial dynamics in hamster embryos over a 24-hour period.Their findings highlight the superior safety of TPEF compared to confocal microscopy (one-photon fluorescence).The researchers successfully attained and sustained blastocyst and fetal development competence during TPEF imaging, a result not replicated with confocal microscopy [15].
The potential application of TPEF in CRC diagnosis during colonoscopy hinges upon the presence or absence of intrinsic fluorophores within the tissues.Among the intrinsic fluorophores found in human tissues, nicotinamide adenine dinucleotide (NADH) can be considered a promising biomarker for CRC diagnosis, using TPEF [16][17][18].
NADH plays a significant role in various metabolic processes in mammalian cells: apart from the regular oxidation-reduction activity, elevated concentration of NADH provides the required energy for cancer cells growth through aggravated mitochondrial respiration.The high levels of NADH in cancer cells accelerate the electron transport chain of the oxidative phosphorylation pathway, which ultimately accelerates the formation of reactive oxygen species in mitochondria.Therefore, NADH is a biomarker for different metabolic irregularities in cancer cells and other pathogenic conditions, including diabetes, neurodegenerative diseases, etc. [18,19].Previous studies have demonstrated that the NADH fluorescence signal could be used to distinguish normal and cancerous tissues [16,18,20,21].
Incorporating MPM into a commercial colonoscope requires addressing the challenges associated with the size of conventional equipment.Thus, efforts are necessary to achieve the miniaturization of the MPM setup.A key aspect of this endeavor involves the design, simulation, and fabrication of a highly-selective optical filter, specifically tailored to selectively transmit the wavelength emitted by the targeted fluorophore of interest, such as NADH in this specific case.This paper presents the details of this optical filter development, based on a thin-film Fabry-Perot structure, followed by the performance evaluation through TPEF measurements conducted in a custom-made setup, using liquid phantoms.This research represents a significant milestone towards the ultimate objective of miniaturizing and integrating MPM into conventional colonoscopy for tissue optical biopsy.The fabrication of thin-film optical filters enables seamless integration into compact devices, where cost-effectiveness and superior optical performance are imperative prerequisites.
The primary innovation of this paper resides in the optimization of the optical filter design.This involves careful consideration of elements such as the number of layers, the thickness of each layer, and the materials used, all geared towards selectively isolating the 485 nm wavelength.Moreover, it introduces an additional layer of novelty by conducting a full examination of the optimal wavelength for exciting NADH through two-photon absorption and determining the most effective wavelength for detecting the ensuing fluorescence, when the fluorophore is excited at 750 nm.This holds particular significance, given the current lack of definitive guidance in existing literature regarding these specific values.

NADH biomarker and fluorescence microscopies
Gastric tissues possess numerous intrinsic fluorophores, including NADH, which is significant for investigating cancer progression.NADH is an enzyme closely associated with cellular metabolism, and its fluorescence emission can be modified in cancerous cells due to alterations in glucose metabolism.Notably, human esophageal normal cells present a NADH concentration of approximately 90 µM [22] and in CRC tissues, the concentration is between 1.4 and 1.6 times higher [16].NADH exhibits a single-photon excitation fluorescence (SPEF) emission peak wavelength of approximately 460 nm.Regarding NADH excitation, it can be achieved through SPEF with a peak wavelength of around 350 nm, or through TPEF with a peak wavelength of around 750 nm [16,[22][23][24].
SPEF and TPEF are two types of fluorescence that result in the emission of photons during a relaxation process of an excited electron, after a different excitation process.In SPEF, only a single excitation photon needs to be absorbed, usually using ultraviolet light, where the energy of the incident photon is equal to the energy difference between the ground and excited electronic states (bandgap energy).In TPEF, two excitation photons need to be absorbed simultaneously, usually using near-infrared light, and the sum of the energies of the two incident photons is equal to the bandgap energy.Fluorescence is a selective process, since the bandgap energy is a molecular feature, making TPEF even more selective than SPEF since it is a non-linear process that requires the absorption of two photons simultaneously [25].
TPEF is a promising technique for depth-resolved and confocal measurements.This optical technique can be used to monitor colon cancer progression, with low photodamage and superior optical penetration due to the use of higher wavelengths.Furthermore, it is a label-free technique with inherent three-dimensional resolution [23,24].TPEF requires the use of an ultrafast mode-locked laser, usually a Ti:sapphire laser tunable between 700 and 1100 nm that reaches pulses of 100 fs, with repetition rates of 80 MHz, and peak powers of 1 to 2 W. In a TPEF setup, optical filters are incorporated in the emission path to facilitate the selective transmission of the emitted fluorescence light, while concurrently rejecting the excitation wavelength and other non-NADH fluorescence sources.This ensures that only the fluorescence generated by NADH is captured and detected, while effectively excluding fluorescence originating from other fluorophores present in the sample.Additionally, signal detection is commonly accomplished using a photomultiplier tube, complemented by other optical components such as lenses, and mirrors that optimize the optical path and enhance the efficiency of signal acquisition [26][27][28].Section 3.1 will detail the TPEF setup used in our work.

Design
The optical filter design relies on the implementation of a Fabry-Perot structure, which constitutes an interferometer comprising two parallel and highly reflective mirrors spaced by a resonant cavity (Fig. 1).This structure is founded on the principles of multiple beam interference, wherein the transmitted light's wavelength (λ) is governed by the equation: assuming a collimated beam at normal incidence, and n represents the refractive index of the resonant cavity.The thickness of the resonant cavity is denoted by d, while q is the interference order of the filter.Considering the case of employing dielectric mirrors (thin-films multilayer), the thickness of each thin-film (d m ) follows the condition: where n m is the refractive index of each thin-film.The use of dielectric mirrors enables low energy absorption rates and high transmittance within a specific range of wavelengths [29,30].
For the implementation of the optical filter, silicon dioxide (SiO 2 ) and titanium dioxide (TiO 2 ) thin-films were used as low and high refractive index materials, respectively.

Fabrication
SiO 2 and TiO 2 thin-films were deposited with radio frequency (RF) sputtering, which is a physical vapor deposition technique.This process occurs within a vacuum chamber with a low pressure of an inert gas (usually argon, Ar).A high voltage alternating current power source is used to generate energetic waves along the chamber, leading to the ionization of the Ar gas, creating Ar + ions.These high-energy ions then collide with the target (cathode), sputtering off its atoms that subsequently deposit onto the substrate (anode).RF sputtering is particularly advantageous for dielectric targets as the alternating mode prevents the accumulation of charges on the nonconductive target surface, ensuring the continuous ionization process [31,32].In this study, SiO 2 thin-films were deposited from a ceramic target (SiO 2 , 99.995% pure) with 150 W power supply.An Ar flow rate of 15 sccm and a pressure of 2 × 10 −3 mbar were used during deposition.A deposition rate of approximately 0.55 Å/s was achieved.The TiO 2 thin-films were deposited from a metallic target (Ti, 99.7% pure) with 200 W power supply.Reactive RF sputtering was used with Ar and O 2 flow rate atmosphere of 10 and 2 sccm, respectively, with a pressure of 4 × 10 −3 mbar during deposition.The deposition rate was approximately 0.15 Å/s.
The thin-films were deposited on a BK7 substrate, known for its high purity of raw materials, high transmittance in the range of visible and near-infrared wavelengths, and low dispersion.Prior to deposition, the substrate underwent a meticulous cleaning process to eliminate any impurities that could potentially impact the filter's performance.The cleaning procedure involved successive steps, starting with mechanical cleaning using isopropyl alcohol (IPA), followed by an IPA ultrasound cleaning, and finally rinsing with IPA and distilled water.Subsequently, the substrate was dried using a nitrogen gun to ensure a pristine surface for thin-film deposition.This rigorous cleaning process was imperative to ensure the integrity and optimal functioning of the optical filter.

Characterization
Ellipsometry (alpha-SE Ellipsometer, J.A. Woollam Co.) was used to characterize the deposited dielectric thin-films (SiO 2 and TiO 2 ).Before the fabrication of the optical filter, these thin-films underwent characterization in test samples.The purpose is to determine the experimental refractive indices for a range of wavelengths and their variation with film thickness.The measured refractive indices were used during the simulations of the optical filter.The ellipsometer used in this study covers a spectral range spanning from 380 nm to 890 nm.The incident angle of the light on the sample was set at three different angles, specifically 65°, 70°, and 75°.The acquired ellipsometry data were subsequently processed using a Cauchy model to determine the refractive index of the thin-films.
Following the fabrication, the optical filter was characterized in terms of transmittance and bandwidth, using the transmission mode of the same ellipsometer.The samples were properly oriented relative to both the emitter and receiver to ensure that the incidence of light was normal (perpendicular) to the filter.
Finally, the fabricated optical filter viability to measure NADH fluorescence emission was evaluated with liquid phantoms and a custom-made TPEF setup.Detailed analysis of this setup will be provided in a subsequent section.

Two-photon fluorescence setup
The TPEF measurements were carried out using the experimental setup shown in Fig. 2.This setup employed a mode-locked Ti:sapphire laser (Mira, Coherent Inc.) as the excitation source.The incident power was controlled by a zero order half-wave plate (464-4215, Eksma) followed by a calcite Glan Taylor polarizer.The beam was focused onto the samples using a × 10 objective (Nikon CFI Plan Fluor).The samples were positioned in the focal plane using a xyz piezo-controlled translation stage (Mad City Labs), with sub micrometer precision.The incident power in the sample ranges from 80 mW to 147 mW, approximately.Despite the theoretical Fourier limit of the pulse duration being approximately 85 fs, we estimated that the actual duration stretched to around 120 fs upon incidence on the sample.This elongation in pulse duration can be attributed to the combined effect of the calcite polarizer and the microscope objective in the experimental setup.In the detection arm along the transillumination direction, a 40 mm focal length best form lens (LBF254-040-A, Thorlabs) was employed to collimate the fluorescent light, followed by a calcite Glan-Taylor polarization controller.A long-pass dichroic mirror (DMLP650, Thorlabs) was used to filter out most of the incident light while reflecting 99% of the fluorescence light.Subsequently, a short focal length lens (C220TME-A f = 11.0 nm Aspheric Lens, Thorlabs) focuses the beam through a short-pass filter (FESH0600, Thorlabs) and the fabricated band-pass filter, and finally onto a fiber bundle coupled to an imaging spectrometer (Shamrock 300i, Andor Technology), equipped with a cooled CCD array (Newton DU920 UVB, Andor Technology).
The phantom was placed in a cuvette and inserted into the piezo-controlled translation stage.The cuvette position was adjusted to maximize the intensity of the fluorescence signal.The fluorescence signal was acquired using the following protocol: at each laser power the CCD signal was integrated for 60 s; the background signal is subtracted from the acquired fluorescence spectra, followed by an algorithmic filtering process to remove cosmic peaks; lastly, a Savitzky Golay filter was applied to further reduce the high frequency noise present in the signal.

NADH excitation and emission spectra
NADH fluorescence properties were studied using SPEF and TPEF.To characterize the excitation, the single-photon absorption spectrum of an NADH liquid phantom with a concentration of 128 µM was obtained, using commercial equipment (Shimadzu UV 3101PC).The NADH used in this work was acquired from Merck (N6005) and was dissolved in sodium hydroxide (Merck 160309).Figure 3 shows the measured absorption spectrum of NADH, exhibiting significant light absorption capabilities within the wavelength range from 300 to 400 nm, with a peak excitation wavelength of approximately 340 nm.The SPEF emission of NADH was characterized using the commercial equipment SPEX Fluorolog 2, and a liquid phantom containing NADH at a concentration of 128 µM. Figure 4(a) displays the experimental emission spectra of NADH when excited at 375 nm.As can be seen, NADH exhibits a broad emission band with a peak wavelength at approximately 470 nm.The observed significant SPEF with an excitation peak wavelength of 375 nm suggests that a corresponding excitation peak wavelength of 750 nm can be used for TPEF to generate NADH fluorescence.For the TPEF emission characterization, a custom-made setup, described in detail in section 3.1, and an NADH liquid phantom with a concentration of 90 µM were used.Figure 4(b) shows the experimental emission spectra of NADH with an excitation peak wavelength at 750 nm.Two different laser powers were employed (100 mW and 147 mW).It is important to note that at this stage the setup was used without the fabricated band-pass optical filter.As can be seen, NADH exhibits again a broad emission band with a peak wavelength of approximately 485 nm for both presented laser powers.A notable observation is the shift towards higher wavelength in the TPEF emission compared to SPEF.This phenomenon aligns with the finding of Bestvater et al., who conducted a comprehensive study comparing TPEF and SPEF from various fluorophores.Their study concluded that the emission spectra of the most fluorophores undergo a shift towards higher wavelengths, which corroborates the observations made in the current study [33].It is important to notice that although 375 nm is not the peak absorption wavelength of NADH, it is employed as the excitation wavelength to obtain the emission spectrum for SPEF.This choice is based on the fact that 375 nm falls within a range of high absorption wavelengths, and it is also half the value of 750 nm.The latter wavelength is the minimum achievable value in the developed two-photon fluorescence setup, considering the limitation imposed by the laser power.This constraint is directly associated with the characteristics of the laser used in the experiment.

Ellipsometry
The experimental refractive indices of the deposited SiO 2 and TiO 2 were obtained as a function of wavelength and for different thicknesses.For ellipsometry characterization the following considerations were taken: the SiO 2 thin-film was chosen as the material with a low refractive index for the optical filter mirror and resonant cavity, while the TiO 2 thin-film was selected as the material with a high refractive index for the optical filter mirror.Figure 5 (a) and (b) shows the obtained refractive indices for the SiO 2 and TiO 2 , respectively.These experimental refractive indices were then used in the optical filter simulation before fabrication.
The measured refractive indices exhibit a dependency on both wavelength and thin-film thickness.According to the literature, the refractive index at 500 nm for SiO 2 is approximately 1.46, while for TiO 2 is approximately 2.71 [34].The obtained experimental values differ slightly from the literature, which can be correlated to the deposition process parameters and the thicknesses of the thin-films, both of which influence the optical properties of the thin-films.

Optical filter simulations
The Fabry-Perot optical filter was simulated and optimized using OpenFilters software.The filter design consists of 11 thin films with alternating low (SiO 2 ) and high (TiO 2 ) refractive index materials.The experimental refractive indices were used in the simulations, considering the wavelength and thin-film thickness variations between the mirrors and the resonance cavity of the Febry-Perot structure (Fig. 5).
The objective is to achieve a selective optical filter with a transmittance peak near to 485 nm, which corresponds to a wavelength in the range of the TPEF peak emission (see Fig. 4(b)).The thickness of each thin film was determined using ( 1) and ( 2) and subsequently optimized using OpenFilters.The resulting thickness values are presented in Table 1.The simulation of the optical filter's transmittance is depicted in Fig. 6.These results reveal that at a wavelength of 485 nm the transmittance is approximately 96%, with a full width at half maximum (FWHM) of 12 nm.

Optical filter characterization
The fabricated optical filter was characterized using transmittance measurements conducted with a commercial ellipsometer.Figure 7 shows the transmittance curve of the fabricated filter, exhibiting a peak transmittance of approximately 95% at 483 nm, with a FWHM of 12 nm.The experimental results display slight deviations from the simulation results, which is probably related to the thin-films optical variations (e.g., refractive index dependence on thickness and wavelength) during the fabrication process.Despite the differences between the design and experimental results, it remains crucial to evaluate the suitability of the fabricated optical filter for the intended application, which involves accurately extracting the NADH fluorescence emission using TPEF.The validation of the filter through experimental tests conducted with the TFEF setup will be presented in section 4.2.

Two-photon fluorescence with phantoms
To assess the effectiveness of the fabricated optical filter in the specific application, the setup presented in section 3.1 was used to generate TPEF in NADH phantoms.Two different phantoms were developed for this purpose: one with a concentration of NADH of 90 µM (normal tissue), and another with a concentration of 126 µM (tumoral tissue).The acquired fluorescence results are presented in Fig. 8.
Figure 8(a) and 8(b) depict the acquired fluorescence curves without the use of the band-pass fabricated filter in the setup for normal and tumoral phantoms, respectively.Figure 8(c  phantoms, respectively.The use of the optical filter proficiently decreased the counts within wavelength ranges that do not correspond to the transmission zone of the filter.This allows the setup to selectively detects NADH fluorescence while disregarding the fluorescence of other intrinsic fluorophores that can be presented in tissues, such as FAD, which exhibits intense fluorescence between 500 nm and 575 nm [35].Without the filter, the FAD signal would interfere with the NADH signal, in the case of performing in vivo measurements.
When comparing the counts values between normal and tumoral phantoms, no significant difference is observed.This may be attributed, in part, to the limitations of the current setup, particularly the use of a spectrometer with an integration time of 60 s, which makes it challenging to maximize the fluorescence signal during the positioning of the cuvette.
By comparing the acquired fluorescence curves obtained at two different laser powers (80 and 100 mW), it is evident that the maximum number of counts increases approximately 1.6 times in both phantoms when the power is increased 1.25 times.This observation supports the conclusion that the fluorescence signal obtained is generated through the process of two-photon excitation.Since, unlike one-photon fluorescence, which exhibits a linear correlation between signal intensity and laser power, two-photon fluorescence is expected to be dependent on the square of the power.

Conclusion
In this study, we aimed to develop and evaluate an optical filter for the selective detection of NADH fluorescence emission using TPEF in the context of CRC diagnosis.The optical filter was designed based on a Fabry-Perot structure, using thin-film technology with SiO 2 and TiO 2 as the low and high refractive index materials, respectively.The filter was fabricated through RF sputtering deposition of thin-films, which were characterized by ellipsometry.
The transmittance measurements of the fabricated optical filter demonstrated a peak transmittance of approximately 95% at 483 nm, with a FWHM of 12 nm.Although slight deviations were observed between the experimental and simulation results, likely due to optical variations during the fabrication process, it remains crucial to evaluate the filter's suitability for the intended application.
To assess the effectiveness of the optical filter, TPEF measurements were performed using a custom-made setup and liquid phantoms containing NADH.Despite the custom-made setup detecting the NADH fluorescence without the use of an optical filter, the acquired fluorescence results demonstrated the importance of the band-pass filter in selectively capturing the NADH fluorescence emission, while reducing counts in other wavelengths where the contribution from other fluorophores is high.This enables the setup to distinguish NADH fluorescence from other intrinsic fluorophores, such as FAD, which could interfere with the analysis.However, no significant differences in counts were observed between normal and tumoral phantoms, potentially attributed to limitations in maximizing fluorescence signal during cuvette positioning.
Overall, this research represents a significant step towards the miniaturization and integration of MPM into conventional colonoscopy for optical biopsy of colorectal tissues.The developed optical filter provides promising capabilities for selective NADH fluorescence detection, contributing to the potential for early cancer detection and improved patient outcomes.Future work should focus on refining the experimental setup (photomultiplier tube for signal acquisition) to enhance the differentiation between normal and tumoral tissues, ultimately advancing the application of optical biopsy in CRC diagnosis.

Fig. 6 .
Fig. 6.Simulated transmittance curve of the optical filter for 485 nm transmittance peak.
) and8(d)  show the acquired fluorescence with the use of the fabricated filter, for normal and tumoral

Fig. 8 .
Fig. 8. TPEF measurements on NADH phantoms using the TPEF setup: (a) phantom with a NADH concentration of 90 µM without the fabricated filter; (b) phantom with a NADH concentration of 126 µM without the fabricated filter; (c) phantom with a NADH concentration of 90 µM with the fabricated filter; (d) phantom with a NADH concentration of 126 µM with the fabricated filter.