The translation of optical devices based on fibre-optic light delivery/collection into the clinical setting has gained significant momentum driven by its diagnostic/therapeutic value and its ease of use. Due to the biophysical properties of light, optical measurements uniquely sample tissue at the molecular level, and thus can probe cells, proteins, and tissue. The biological information retrieved from optical measurements depends on the light-tissue interaction taking place, from assessing the changes in the concentration of biomolecules to probing changes in their shape and orientation that manifest during pathogenesis [1]. In the case of diffuse reflectance measurements, the detected signal is a result of multiple elastic scattering and absorption events. Extracting the wavelength-dependent absorption coefficient allows the determination of the concentration of macromolecules such as haemoglobin, lipids, and melanin [1], while the reduced scattering coefficient is modulated by the concentration and distribution of organelles with varying index of refraction such as at the interface of cellular membranes, collagen fibrils, lysosomes and mitochondria [1]. During breast tumor surgery, for example, changes in the reduced scattering coefficient provided more useful diagnostic information compared to monitoring blood oxygenation and total haemoglobin content [2]. In addition, when fluorescence is combined with diffuse reflectance measurements, the fluorophore concentration can be extracted [3].

Fibre-optical probes used in research hospitals are being evaluated for its efficacy to guide surgery [4–7], monitor tissue response to treatment [8, 9], and to obtain an optical biopsy [10–12]. The architecture of fibred probes differs according to application as well as the optical contrast that is being measured. A fibre-optic probe measuring both the steady-state diffuse reflectance and fluorescence can be configured so that both excitation light sources are guided by the same fibre, while several collection fibres are placed in the periphery or linearly at different source detector separations so that the tissue optical properties can be extracted from the diffusion Eq. 13. While previously lagging behind, probes measuring time-resolved fluorescence are slowly entering clinical studies with the advent of faster detectors and pulsed lasers [13]. A promising example is the bifurcated sterilizable probe designed for in vivo intensity and time-resolved fluorescence detection to distinguish between low-grade glioma and normal white matter [14]. Although point-based measurements suffer from mis-sampling errors, fibre-based optical spectroscopy still surpasses imaging in its sensitivity, acquisition speed, practicality and precision, providing point of care diagnosis. The unique advantage of using fibre-optic probes over wide-field imaging is that the sampled volume is well-defined by the source-detector geometry and that the functionality of multiple optical techniques can be incorporated into one probe so that more information about relevant physiological parameters could be extracted without extra bulkiness or cost, while also mitigating the limitations of each [15]. Since lifetime measurements are insensitive to variations in probe to tissue distances or tissue optical properties, they are well suited for a dual modality system such as the system developed by Dochow et al [16].

Translation of intraoperative multimodal fibred-based detection in to the neurosurgical suite is severely lagging due to additional constraints such as further miniaturization of the probe, a sterilizable distal end and the need for rapid acquisition and analysis. Here, we present a novel bimodal spectroscopy device, which combines spectrally- and temporally-resolved fluorescence measurements and is optimally designed for seamless surgical guidance with high temporal (35 ps), spatial (<0.3 mm), and spectral resolution (1 nm) obtained in under one second. Three main characteristics make our probe novel: 1) its ability to simultaneously detect five endogenous fluorophores under one excitation light (405 nm); 2) its hardware is modular in the sense that the probe head used by the surgeon can be sterilized, while the proximal end of the probe (5 m) and black box housing the remaining optical components can be placed in the non-sterile area of the operation room (OR) without affecting the quality of the measurements or intervention; 3) the custom software allows for immediate visualization of the autofluorescence spectrum and lifetime decay profile for each fluorophore with a single-click operation. Post measurement, a custom python-based program is implemented for spectral and lifetime fitting upon one click to retrieve quantitative information. Although the probe design and its development are optimized for epileptic and brain tumor surgery based on the feedback from the co-author at GHU Paris Psychiatrie et Neuroscience Hospital (B.D). It is compatible and useful for any open (or minimally invasive) surgery or in situ tissue monitoring during medical intervention. The probe specifically measures the fluorescence emitted by NADH, flavins (FAD), lipopigments, porphyrin I and porphyrin II, all of which are altered in the brain with the onset of disease [10, 17, 18]. The quantitative metrics of interest in brain cancer or epilepsy are primarily related to the energetic metabolism as mirrored in the FAD/NADH ratio [18], the oxidative stress, as portrayed by the Lipopigments to porphyrin ratio [19, 20], and the degree of vascularization as portrayed by the porphyrin to NADH ratio [21]. Additionally, changes in the short and long fluorescence lifetime of NADH (FAD) are directly related to the conversion of free NAD + (FAD) to bound NADH(FAD) signifying altered metabolic states as seen in hypoxic or inflamed regions [22].

This report starts with Monte Carlo (MC) simulations to characterize the collection efficiency and tissue volumes interrogated by the probe using a customized version of the Monte Carlo multi-layered code by Wang et al. [23]. The spatial resolution and optimal probe to tissue distance is theoretically calculated for a non-turbid medium and then experimentally determined for a turbid medium. Moreover, fluorescent dyes and tissue-simulating phantoms are both used to validate the linearity, the stability, the sensitivity as well as the temporal impulse response of the system. Ex vivo measurements of different organs taken from a rodent model and from human brain biopsy samples are performed. The spectral shapes are compared with those obtained from a 405 nm confocal fluorescence microscope at similar regions of interests, while the fluorescence lifetimes are compared to literature. Finally, the effect of pressure applied to tissue by the probe is experimentally tested. Ongoing ex vivo measurements of freshly excised brain tissue are being performed to delineate optical markers that can best characterize different brain tissue types, healthy and diseased, before moving the system into the OR. The end goal is to enhance needle biopsy and resection surgery by a high information content comprising relevant physiological and metabolic information in situ, unavailable by any other intraoperative technology.

You may typeset. The probe that is presented in this work is a multi-fibred optic probe specifically designed to be utilized for brain tissue classification during neurosurgery for a wide range of diagnosis including epilepsy, meningioma, and glioma. Two main biophysical properties are measured that are expected to vary with tissue type and pathology, the fluorescence spectra from five endogenous molecules as well as their fluorescence lifetime [17–19, 21, 22, 27–29, 31, 41, 42]. Since the geometry and physical properties of the collection and illumination fibres directly impacts the extent of tissue being optically sampled and the collection efficiency [43], it is imperative to fully characterize the fibred-optic probe, which is intendent for in vivo diagnosis. This was accomplished by MC simulations (Figures 2, 3) and by experiments with tissue simulating phantoms (Figure 4, 6; Tables 1–3). The modelled behavior of fluorophore absorption and fluorescence emission area at the tissue surface (Figure 3) is similar to the absorption and emission maps obtained for fibres of similar geometry, fibre diameter size, and illumination-collection fibre separation distance [43, 44]. The diameter of the collection fibre in our probe is modestly larger than those typically reported (360 µm vs. 200 µm) [43, 44], which results in higher collection efficiency and mean path lengths, but does not significantly increase the probing depth [43]. The asymmetric intensity profile as a function of probe to sample distance (Figure 4A) was also described in the bifurcated probes designed by Papaioannou et al and by Pfefer et al [43, 44]. The optimal probe to sample distance was found at larger distances for larger source-collection separation distances so that maximum overlap between the illumination and collection cones is achieved. This was found to be 1.5 mm for our probe. The deviations observed between the simulated intensity profile and the measured intensities (Figure 4A) depend on the probe to sample distance. At distances shorter than 1.5 mm, the simulated data are underestimated due to the low probability in fluorescence detection, where only laterally scattered light reaches the detector. At distances beyond 1.5 mm, the overlap between the illumination and collection cones decreases more rapidly than predicted by MC as the model does not account for the direction at which the fibres are beveled. Based on the above observation, one can deduce that the fibres are beveled away from each other so that the intersection between the illumination and detection areas decreases faster than if the fibres were beveled toward each other at the same angle (11°). At a fixed probe to sample distance (1.5 mm), tissue optical properties affect the depth at which fluorescent photons originate from, where the fluorescence signal is more surface weighted for higher μ_a and μ_s values as determined here. The higher scattering values used explains the shorter depths than those reported in similar studies [42, 43]. Nevertheless, the 600 µm separation distance between the source and collection fibre ensures relatively homogenous sampling with depth. The range of optical properties selected for MC and tissue phantom measurements aimed at modelling a wide range of optical properties as observed in vivo in brain tissue [4, 29]. The angular orientation of the detection fibre does not affect the sample volume as shown by rotating the emission matrices, therefore adding the emission maps from all five detectors does not affect the optically interrogated volume. This is because the illumination angle is maintained perpendicular to the sample and the separation distance and angle between the source and collection fibres is fixed and therefore the final emission is the summation of the photons collected by the five detectors.

The spatial and temporal impulse response of the probe in a non-turbid medium defines the spatial and temporal resolution. These were found to be 150 µm and 35 ps, respectively. It is important to note that the fluorescence lifetime of the five endogenous molecules of interest have their lifetimes in the nanosecond range and thus the temporal resolution is more than sufficient for in vivo measurements. Additionally, the nanosecond lifetimes ensure that the temporally-resolved fluorescence measurements interrogate the same tissue volumes as the steady-state intensity measurements [43], which is the main confounding factor in imaging devices with multimodal abilities. A recent study showed that reconstructing an optical coherence tomography (OCT) based image from a three dimensional image to a two dimensional “en face” OCT map, enables one to extract structural contrast within the same tissue area interrogated by a scanning fibre-optic probe acquiring the autofluorescence spectrum [45]. The steady-state MC simulations predicted the spatial extent of the incoming beam at the tissue surface to be ∼200 μm, close to the measured spatial impulse response of 150 µm. 200 µm corresponds to the size of the illumination fibre indicating minimal diffusion induced divergence of the excitation light, since the excitation photons are readily absorbed. The MC simulations were limited to modelling one fluorophore that was homogenously distributed throughout a semi-infinite turbid medium with homogenous optical properties at the excitation wavelength. For the fluorescence emission simulations, wavelength-dependent µ_am, µ_sm, and the extinction coefficient of FAD were included. A recent tetrahedral mesh-based MC program developed by Lilge et al. can incorporate different optical properties and tissue geometries to model white and grey matter as well as tumor inclusions [46]. Since the purpose of the MC simulations was to calculate the interrogated tissue volume, the optimal probe to sample distance, and the depths of maximum sensitivity, the assumptions in the modified mcml code were sufficient. Detailed validation of the modified code with multi-layered fluorescence inclusion measurements as well as comparison to the conventional MCML code has been studied previously [25, 26].

The minimum detection limit of 4 µM is only a magnitude lower than the [FAD] concentrations expected in vivo [31], however this limit was obtained for an acquisition time of 40 ms (Figure 6). Increasing the acquisition time to 400 ms is possible without causing practical delays or photobleaching. The detection limits at the other emission bands were not characterized here as their concentrations found in brain tissue are not reported in the literature, although the [NADH] is expected to be in the mM range [39]. For ex vivo measurements of biopsied brain tissue, an acquisition time of 500 ms is typically used to ensure high signal to noise ratios in both the spectral and temporal measurements. The accuracy of recovering the fluorescence lifetime of several fluorescent dyes was tested as a function of fluorophore concentration (Table 2) and pH (Table 3). The lifetime values for each dye is consistent with those found in literature for the given concentrations and dilution solutions [47–49]. The reproducibility can be estimated based on the standard deviations obtained after repeating the measurement three times. This varied between 2% and 6%. The dichotomic behaviour of RhdB as a function of concentration has also been observed elsewhere [48, 49]. At RhdB dye concentrations lower than 10⁻⁴ M, its fluorescence lifetime was longer (1.98–2.24 ns) that its value at higher dye concentrations (1.56–1.68 ns) due to static quenching [48]. The opposite trend is observed for FAD, where higher FAD concentrations (≥10⁻⁴) resulted in longer lifetime values due to the dynamic intramolecular electron transfer between the photoexcited isoalloxazine moiety and the adenine moiety of FAD, which in turn depends on the dielectricity and viscosity of the surrounding environment [40]. The longer lifetime at the high concentration is closer to the lifetime of flavin mononucleotide (467 ns) [50], while lifetime values for FAD for the lower concentrations are not found in literature even though they are closer to the concentrations observed in vivo. Moreover, the probe measurements also recapitulated the independence of the fluorescence lifetimes on pH for fluorescein and RhdB, as also observed in other studies, indicating that these dyes are not sensitive markers of pH [47–49].

In the ex vivo rodent and human study, steady-state and lifetime measurements were conducted on fresh and fixed tissue samples to understand the changes in the biophysical properties that occur due to tissue fixation. This is critical so that optical measurements performed on fixed tissue can at least be correctly extrapolated to what is expected in vivo when needed. Fixed tissue analysis is more practical and is more useful when correlating or comparing to findings from histopathology. An interesting finding here was that keeping the sample “fresh” is not necessary for conventional confocal fluorescence microscopy as the spectral specificity is inherently lost by the limited spectral resolution of the microscope itself. In fact, tissue fixation is preferred as the fluorescence is enhanced (Figures 5D, 6B, 7B), particularly that from NADH and FAD. The relative contribution of porphyrins I and II are significantly reduced. Another explanation of the difference in the spectral shape obtained by our probe and the confocal microscope is that in the latter case, the excited and emitted light is minimally distorted by tissue optical properties as scattered photons are not detected and hence do not modulate the fluorescence spectral shapes. After tissue fixation, the spectral shapes acquired from our probe and microscopy are similar. Since Formalin fixation reduces haemoglobin absorption [51], the fluorescence intensities under the oxy- and deoxy-haemoglobin peaks (440, 555, and 580 nm) are enhanced. These correspond to NADH, FAD and lipopigments. Another result due to haemoglobin changes is that the concentration and distribution of porphyrins are altered [51, 52], hence their relative contributions to the overall spectra are different from those found in fresh samples. Quenching of the fluorescence of porphyrins following Formalin fixation (Figures 6B, 7B) could have also resulted in such changes. Interestingly, although the spectral shapes for each fresh tissue type (brain, liver, and lung) are unique, the lifetime values in the rodent study for each fluorophore is relatively constant as they are independent of tissue optical properties and fluorophore concentration. In all three tissue types, although less for the lung sample, the fluorescence lifetime is systemically longer for each fluorophore upon tissue fixation (Table 4), which was also found recently with FLIM of brain tissue samples [53]. The interaction of each fluorophore with Formalin changes its protein-binding affinity, directly affecting the average lifetime. On the other hand, the amplitude weighted mean lifetimes of all five fluorophores, except porphyrin II, recovered from fresh human cortical tissue did not differ significantly after tissue fixation (p > 0.05) (Table 5). However, when comparing the free and bound lifetime values of FAD and NADH, it was found that the lifetime values of free NADH (τ1) and bound FAD (τ1) were higher upon tissue fixation. Similar findings were reported by Wang et al. [39], when cultured Hela cells were Formalin fixed. The authors also found that changes in the fluorescence lifetime of NADH and FAD depended on the time of measurement post-fixation [39]. Specifically, the lifetime of NADH increased sharply (33%) following fixation and remained constant for 14 days. FAD increased steadily right after tissue fixation, but then jumped after day 11. Lifetime measurements were not conducted after day 14. In the present study, the fixed human brain samples were measured one-month post-fixation, while the animal tissue samples were measured 6 days post-fixation. Therefore, the time of measurement after fixation may explain the differences observed. Nevertheless, the lifetime values obtained ex vivo do mirror the values reported in vivo for FAD and NADH, which varies from 2.3 to 2.9 ns for free FAD and 2–2.3 ns for bound NADH [42]. Additionally the lifetime values reported ex vivo on fresh cortical brain tissue for all five fluorophores are within the ranges reported previously by our group using a dual fibre-based fluorescence setup that was stationed adjacent to the OR [29]. With that setup as well, fixed tissue samples had generally longer lifetime values [53]. These comparisons further demonstrate the accuracy of the lifetime measurements acquired by our probe on the one hand, and the different fluorescence response as a function of tissue type and tissue status on the other.

Finally, the pressure study highlighted the importance of maintaining a relatively constant pressure applied to the probe when in contact to tissue. Although the autofluorescence spectra did not vary significantly between light pressure and medium pressure or between the two heavier ones (Figure 10A), as also seen in other studies [33–36], the pressure threshold to cause a difference in the intensity was not quantified. In any case, when comparing with other reported studies, it is evident that the effect of probe pressure is tissue site and probe specific. The diffuse reflectance acquired in vivo by a single fibre diffuse reflectance spectroscopy, showed that increased pressure reduces µ_a and increases µ'_s [33–35]. The reduced absorption is a result of reduced total haemoglobin content, vessel diameter and oxygenation level as pressure expels blood cells within the interrogated area and constrains the supply of incoming oxygen [33]. Another study involved testing the effect pressure applied on a clinical probe in contact with the cervix for cervical cancer screening [36]. The authors concluded that the autofluorescence spectral shapes remained unchanged with applied pressure. To correct for intensities differences, each spectrum was normalized to its peak intensity (∼380 nm). Pressure-induced signal changes did not affect the accuracy in identifying the different pathologies investigated [36]. A similar finding to Nath et al., was demonstrated here (Figures 10A, B). Light and medium contact did not significantly alter the fluorescence signals, however there is a significant difference between strong and light contact, which is systemic across the spectral range of interest. This observation motivated the choice of normalizing each spectrum by the 520 nm peak. This peak is one of haemoglobin’s isobestic wavelengths, where oxy- and deoxy-haemoglobin is equally affected and hence the normalization step accounts for total haemoglobin concentration changes. This decrease as well as an increase in the scattering parameters result in higher fluorescence signals. However, since applying pressure causes more oxygenated blood to become de-oxygenated [33–35], the total absorption coefficient is lower in magnitude at wavelengths beyond 600 nm. This explains the slightly higher porphyrin I fluorescence peak obtained from the heavy pressure measurement despite the normalization scheme (Figure 10B). The slight increase in the standard deviations pertaining to Lipopigments after normalization indicates that the normalization scheme overcorrects the fluorescence intensities since the fluorescence was least altered as lipopigments are not involved in cell respiration or blood flow. One additional important factor affecting probe pressure studies was found to be the duration of measurement, i.e. the duration the probe is in contact with tissue [33]. Minimal effects on the optically-derived physiological parameters, including [NADH], [haemoglobin], and oxygenation saturation levels, are seen for acquisition times shorter than 2 seconds, while significant changes were found for longer contact times [33]. The probe measurements do not take more than 1 s per site. These results show that the probe measurements can be conducted for in situ tissue diagnosis without being concerned about the interfering factor of pressure variation.

Overall, this paper aims to characterize a bimodal label-free fibred probe targeted for tissue characterization during brain tumor surgery. The complementary information extracted by spectrally- and temporally-resolved fluorescence has been utilized by other groups [7, 54], but the unique advantage of our custom-made optical probe is its practicality and ability to extract several physiologically relevant quantitative markers in near real-time. As a pilot validation study, our probe is now stationed adjacent to the OR so that the ex vivo measurements closely mimic the in vivo scenario.