2.1.

Time-Resolved Fluorescence Spectroscopy–Instrumentation (Point Spectroscopy)

The time-resolved laser-induced fluorescence spectroscopy (TR-LIFS) system developed in our lab and used in numerous studies^{43, 44, 46, 47} concerning the characterization of the biological tissues was based on a time-resolved time-domain pulse sampling and gated detection technique described in detail elsewhere.⁴⁸ It consists of a modular design including:

In the configuration described earlier, however, practical clinical application of TR-LIFS still faces challenges. These include the time-expensive data acquisition due to scanning of the monochromator. Recently, we reported⁴⁹ a novel concept that allows for near real-time acquisition of spectrally resolved fluorescence intensity transients through the combination of optical fibers and bandpass filters (Fig. 1 ). A single detector is used to simultaneously record multiple fluorescence intensity decay profiles (fluorescence response pulses) in response to a single pulse excitation event. Simultaneous time- and wavelength-resolved fluorescence spectroscopy (STWRFS) allows the recording of both fluorescence lifetime and spectral intensity information within hundreds of nanoseconds within multiple spectral bands. This system can be optimized for the characterization of the atherosclerotic tissue within suitable spectral bands, thus simplifying the data analysis and tissue classification algorithms. Since the objective of medical diagnosis is to identify conditions in tissue by rapidly recognizing signatures corresponding to specific states, this approach has the inherent potential for a direct recording of representative tissue fluorescence features. Once this apparatus is interfaced with classification models that enable online diagnosis, the STWRFS device can evolve into a clinical diagnosis tool. This research direction is currently being pursued in our laboratory and includes application of this apparatus for cardiovascular diagnostics.

Fig. 1

Schematic of the simultaneous time- and wavelength resolved fluorescence spectroscopy (STWRFS) system. By combining multiple bandpass and dichroic filters ( $405∕40$ , $460∕50$ , and $550∕50$ ) with different lengths of optical fiber (1, 10, and $19\mathrm{m}$ ) acting as optical delay, this system enables the near real-time acquisition and characterization of time-resolved fluorescence spectra using a single detector and excitation input. The recording of multiple fluorescence response pulses at selected wavelengths can be completed in hundreds of nanoseconds, which provides the capability of a real-time characterization of tissues. Configuration adapted for single-fiber excitation-collection geometry. The initial system was reported in Ref. 49.

2.2.

Fluorescence Lifetime Imaging Microscopy (FLIM)—Instrumentation

In contrast to point spectroscopy systems, FLIM allows the localization and mapping of fluorophore distributions so that structural, chemical, and environmental information may be contrasted across the field of interest. The FLIM apparatus (Fig. 2 ) constructed in our lab was recently reported⁵⁰ and consists of a gated intensified CCD camera (ICCD), a pulsed laser, a flexible fiber-optics-based endoscope, and a filter wheel. Tissue autofluorescence is induced by $337\text{-}\mathrm{nm}$ laser with $700\text{-}\mathrm{ps}$ pulse width (nitrogen laser, MNL 205, LTB Lasertechnik Berlin, Germany) and collected using a graded-refractive-index (GRIN) lens (GRINTECH GmbH, Jena, Germany, NA 0.5, $0.5\text{-}\mathrm{mm}$ diameter, $4\text{-}\mathrm{mm}$ field of view) cemented to a fiber bundle ( $0.6\mathrm{mm}$ diameter, $2\mathrm{m}$ long, 10,000 fibers). For intravascular compatibility, the probe was made to be side viewing by cementing a BK7 $90\text{-}\mathrm{deg}$ $0.5\text{-}\mathrm{mm}$ edge length polar prism on top of a GRIN objective lens. A $20\times$ microscope objective is used to magnify fluorescence images to the ICCD. An electronically changeable filter wheel enables the selection of up to six emission wavelengths. Advanced electro-optics and fiber optics enabled the compact and robust design of this FLIM apparatus. One of the key components is the compact ICCD 4 Picos (Stanford Computer Optics, Berkeley, California) with fully integrated electronic and optical devices in a small enclosure $(24\times 14\times 11{\mathrm{cm}}^3)$ . It includes a multichannel plate (MCP) image intensifier (gate time down to $0.2\mathrm{ns}$ , repetition rate $200\mathrm{kHz}$ ).

Fig. 2

Schematic of the FLIM experimental setup to transmit the $4\text{-}\mathrm{mm}$ field of view at $4\text{-}\mathrm{mm}$ working distance through the 10,000 coherent fiber bundle ( $0.5\text{-}\mathrm{mm}$ outer bundle diameter) attached to the graded index (GRIN) objective lens and onto the gated camera. Adapted from Ref. 50.

2.3.

Time-Resolved Fluorescence Spectroscopy/Imaging—Data Analysis Methods

In the context of time-domain time-resolved fluorescence measurements, the fluorescence impulse response function (FIRF) contains all the temporal information of a single fluorescence decay measurement. Mathematically, the measured fluorescence intensity decay is the convolution of the IRF with the instrument response. Thus, to estimate the FIRF of a sample, the instrument response must be deconvolved from the measured fluorescence intensity pulse.⁵¹ Traditionally, the most commonly applied deconvolution method is the least-square iterative reconvolution (LSIR).^{51, 52} LSIR applies nonlinear least-square optimization methods (i.e., Gauss-Newton, Lavenberg-Marquardt) to estimate the parameters of a multiexponential IRF that would best fit its convolution with the system response to the fluorescence decay data. Since the optimization process involves iterative convolutions, LSIR is computationally expensive. Moreover, since exponential functions do not form an orthonormal basis, different multiexponential expressions can be fitted to fluorescence decay data equally well, due to the correlation of the fitting parameters in the multiexponential model. As noted earlier, fluorescence emission in tissues originates from several endogenous fluorophores and is affected by light absorption and scattering. From such a complex medium, it is not entirely adequate to analyze time-resolved fluorescence decay in terms of multiexponential components, since they cannot be always interpreted in terms of fluorophore content or number of lifetime components.²³ Thus, for tissue fluorescence lifetime data analysis, there is an advantage in avoiding any a priori assumption about the functional form of the IRF decay.

Consequently, we applied an alternative model-free deconvolution method for TR-LIFS data analysis, based on the Laguerre expansion of the kernel technique (LET).^{53, 54} This method was found to provide important advantages over the more traditional methods when used in the context of tissue characterization. LET is based on the expansion of the kernels (IRF, for linear systems) on a set of ortho-normal discrete Laguerre functions (DLFs), allowing fast converging kernel estimation from short input-output data records using least-square calculation of the expansion coefficients. Taking advantage of the asymptotic exponential decline characteristics of the DLFs, the LET was recently adapted for the deconvolution of TR-LIFS and FLIM decay data and estimation of fluorescence IRFs.^{42, 43, 50, 53, 55, 56} The resulting Laguerre deconvolution method has been proven to be a fast, robust, and model-free alternative for the analysis of time-resolved fluorescence data, properties highly desirable for tissue diagnostic applications. In addition, we have showed that the LECs can be used for direct characterization and diagnosis of tissue. We applied this method in the characterization of the fluorescence decay from atherosclerotic plaques.⁴⁴

In brief, we used the orthonormal Laguerre functions $b_j^{\alpha }\left(n\right)$ to expand the FIRF and to estimate the Laguerre expansion coefficients (LEC) $c_j$ (Refs. 44, 57). Once the FIRFs were estimated for each emission wavelength, the steady-state spectrum $\left(I_{\lambda }\right)$ , was computed by integrating each intensity decay curve as a function of time. Further, to characterize the temporal dynamics of the fluorescence decay, two sets of parameters were used: (1) the average lifetime $\left({\tau }_f\right)$ computed as the interpolated time at which the FIRF decays to 1/e of its maximum value; and (2) the normalized value of the corresponding LECs. Thus, a complete description of each sample fluorescence as a function of emission wavelength, ${\lambda }_E$ , was given by the variation of a set of spectroscopic parameters ( $I_{\lambda }$ , ${\tau }_f$ , and LECs).

2.4.

Intraluminal Catheter for Combined TR-LIFS and IVUS

As noted earlier, a combination of optical spectroscopy parameters that provideds biochemical information and imaging technique that provides structural information could provide a more complete evaluation of features involved in plaque vulnerability. As a first step in validating such an approach, we have developed⁵⁸ a novel approach that enables integration of optical (TR-LIFS) and ultrasound (IVUS) modalities to operate in a beneficial and complementary way in a single multimodal catheter (Fig. 3 ). IVUS is a commonly used catheter-based ultrasonic imaging technique to diagnose and guide interventional therapeutic procedures with real-time arterial cross-section images. It provides specific lesion site identification and can provide direct guidance to an intraluminal location, and subsequent relocation, of specific lesion sites of interest. IVUS, as a catheter standard for ease of use,⁵⁹ is an obvious choice for the intraluminal guidance of the optical probe for use in obtaining TR-LIFS data from sites of suspected vulnerable plaque. Overall, a practical means to implement the TR-LIFS technique involves first an intravascular means of visual guidance for the optical fiber and second a steering method to co-register the optical fiber close to a particular arterial site in a moving blood environment. In brief, the TR-LIFS data is acquired via a side-viewing optical fiber (SVOF), which is shown in Fig. 3 as OF. It consists of a $400\text{-}\mu \mathrm{m}$ -diam fused silica fiber (multimode, $\mathrm{NA}=0.22$ ) with a 45-deg polished metalized mirror termination for beam deflection at 90 deg to the fiber axis. The fiber optic is coupled to a dichroic mirror, which is coupled to the TR-LIFS system in our laboratory described earlier. The IVUS imaging catheter is a 3-Fr $\left(1\text{-}\mathrm{mm}\right)$ mechanical rotating type (SR Pro BSC, Fremont, California) catheter that operates in the $30\text{-}\text{to}40\text{-}\mathrm{MHz}$ range with its ClearViewUltra (BSC) imaging system.

Fig. 3

(a) Picture, (b) schematic, and (c) cross section of the three components: $400\text{-}\mu \mathrm{m}$ side-viewing optic fiber (OF); 3 Fr IVUS catheter; and $356\text{-}\mathrm{mm}$ steering wire. These three components are integrated in a 5.4-Fr multimodal catheter (MMC). (b) and (c) show deployment of lateral (steering) movement with the steering wire (SW). The normally compact multimodal catheter components are separated in the photo for clearer viewing. The SW in the deployed state (b) is near the IVUS sheath and is mounted so that deployment by pushing on the wire will move the multimodal catheter close to the target. Adapted from Ref. 58.

3.1.

Material and Methods

In all these studies, TR-LIFS measurements were obtained with serial scanning of the monochromator across a spectral range from $360\text{to}550\mathrm{nm}$ in increments of $10\mathrm{nm}$ . The total acquisition time across the scanned emission spectrum was $\sim 30\mathrm{s}$ . After each measurement sequence, the laser pulse temporal profile was measured at a wavelength slightly below the excitation laser line. This profile was later used as input to the deconvolution algorithm for the estimation of fluorescence lifetimes. Laser excitation output measured at the tip of the probe was set at $2\mu \mathrm{J}$ /pulse (fluence $1.8\mu \mathrm{J}∕{\mathrm{mm}}^2$ per pulse, fluence rate $54\mu \mathrm{W}∕{\mathrm{mm}}^2$ at tissue level). This output was found as being a reasonable compromise between an adequate signal-to-noise ratio and the photobleaching of the sample.

Following TR-LIFS measurements, all spectroscopically investigated plaque areas underwent conventional histopathologic evaluation. The samples were fixed (10% buffered formalin), processed routinely, and embedded in paraffin. Four sequential $4\text{-}\mu \mathrm{m}$ -thick cross sections were cut from each segment and were stained with hematoxylin and eosin (H&E), a trichrome/elastin method, CD68 (for macrophages), and CD45 (for leukocytes), respectively. Histopathological analysis was performed by two pathologists specialized in cardiovascular diseases and blinded to the TR-LIFS data. The composition of artery wall was assessed within a region of interest (ROI) under the ink mark, defined by the fiber-optic excitation-collection geometry ( $\sim 1.1\text{-}\mathrm{mm}$ illuminated diameter area at tissue surface) and the light penetration depth ( $\sim 250\mu \mathrm{m}$ for $337\mathrm{nm}$ in arterial tissue).⁴² Each section was evaluated by light microscopy.

A stepwise linear discriminant analysis (SLDA) approach was adopted to generate a classification model (discriminant functions) and for sample classification. The discriminant function analysis provides an effective means for classifying spectroscopic data of unknown origin.⁶¹ The set of features (spectroscopic parameters) selected from the statistical analysis were used for the optimization of discriminant functions. The classification accuracy was determined using a leave-one-out cross-validation approach,⁶¹ and then values of sensitivity (SE), specificity (SP), and overall classification performance (% of samples correctly classified) were computed.

3.2.

Aorta

The first systematic study characterizing the time-resolved fluorescence emission of normal artery and various stages of atherosclerosis was conducted in human aortic specimens (94 samples, postmortem).⁶⁰ The Laguerre deconvolution method for TR-LIFS data analysis was applied to estimate the intrinsic fluorescence decays, from which the time-integrated fluorescence emission spectrum and fluorescence lifetime were computed. This study established that the time-resolved fluorescence emission spectra of aortic samples vary with the progression of atherosclerosis. For example, average lifetime at $390\mathrm{nm}$ gradually increased from $2.4\pm 0.1\mathrm{ns}$ (normal aorta) to $3.9\pm 0.1\mathrm{ns}$ (advanced lesions), while fluorescence intensity was markedly decreased above $430\mathrm{nm}$ in intermediate and advance lesions. Characteristic changes in parameters derived from the time-resolved spectra were related to the type of lesion assigned to the sample based on histological examination. Spectral and temporal features of the emission were interpreted for normal aortic wall and lesions ranging from early type I to advanced type V in terms of intimal content in fluorescent compounds and intimal thickness. Trends in the average lifetime at selected wavelengths were identified, which could serve as diagnostic markers for in situ optical analysis of the aortic wall. Also, it was determined⁶⁰ that a set of five predictors variables (fluorescence time-decay parameters at $390\mathrm{nm}$ and $460\mathrm{nm}$ and fluorescence intensity values at $490\mathrm{nm}$ ) allows for discrimination of the lipid-rich lesions (rupture-prone) from collagenous/fibrous lesions (stable) with a sensitivity and specificity higher than 95%.

3.3.

Coronary Artery

In a following study,^{43, 60} time-resolved fluorescence emission of normal and atherosclerotic human coronary arteries (58 coronary segments, postmortem) were analyzed using also the Laguerre deconvolution method. Similar to the study in aorta, these results showed that analysis of the time-resolved fluorescence spectra can be used to enhance the discrimination between different grades of atherosclerotic lesions as defined by the American Heart Association (AHA). Also, it demonstrated that the lipid-rich lesions (e.g., average lifetime at $390\mathrm{nm}$ : $2.6\pm 0.1\mathrm{ns}$ ) can be differentiated from the other lesion types—in particular, fibrous lesions $(3.2\pm 0.15\mathrm{ns})$ and normal arterial wall $(2.1\pm 0.1\mathrm{ns})$ . It was determined that spectroscopic features derived for lipid components are reflected in the emission of lipid-rich lesions, whereas characteristics of type I collagen are identified in the emission of fibrous lesions. In addition, the results suggested that a few spectroscopic parameters that combine spectral features at longer wavelengths and time-resolved characteristics from the peak emission region are the best selections for coronary artery lesion discrimination. For example, parameters derived from time-resolved spectra, such as the lifetime and the fast-time decay constants determined using a biexponential approximation of the intensity decay, are most likely to differentiate between lipid-rich (more unstable) and fibrous lesions (more stable) and be used for diagnosis. These findings further validated the potential of TR-LIFS for discriminating lipid-rich plaques from fibrotic lesions.

3.4.

Carotid Artery

More recently,⁴⁴ we conducted a TR-LIFS study in fresh excised carotid plaques from 65 endarterectomy patients (831 distinct areas). This study was designed (1) to demonstrate whether TR-LIFS signatures can distinguish fibrotic caps rich in macrophages and inflammatory cells and plaques with a necrotic/lipid core under a thin cap (rupture-prone) from plaques with caps rich in collagen (stable); and (2) to determine the TR-LIFS derived spectroscopic parameters that can be correlated to features of plaque vulnerability. Results of this study (Fig. 4 ) indicated that spectral features derived from the normalized time-integrated fluorescence emission spectra provide means for discriminating early lesions or intima thickening (IT) from more advanced stable lesions such as fibrotic (FP) and fibrocalcified (FC) and unstable lesions such as inflamed (INF) and necrotic (NEC) plaques. Also, the study indicated that time-resolved properties of the autofluorescence emission enable detection of superficial necrotic core, one additional feature of plaque vulnerability.

Fig. 4

Left panels: Spectroscopic parameters as a function of emission wavelength derived from the time-resolved spectra measured in ex vivo samples. Right panels: Statistical group comparisons $(\text{mean}\pm \mathrm{S.E.})$ for specific wavelengths. Spectroscopic parameters: normalized intensity (a), average lifetime (b), LEC-0 (c), and LEC-2 (d). Adapted from Ref. 44.

This study also demonstrated the important role that the LECs play in atherosclerotic plaques discrimination. While in previous studies, the Laguerre deconvolution method was primarily used to estimate the intrinsic fluorescence decay emission, in this study, in addition the resulting LECs were adopted for the first time as additional parameters for quantifying the time-resolved fluorescence emission (Fig. 4). Notably, one of these expansion coefficients (LEC-2) enabled discrimination of both inflamed and necrotic lesions from the stable plaques. Moreover, a combination of spectral and time-resolved fluorescence parameters, including LECs, were used as features in a linear discriminant analysis (LDA)–based classifier, which was able to discern inflamed and necrotic lesions from early and advanced fibrotic lesions with high sensitivity $(>80\mathrm{\%})$ and specificity $(>90\mathrm{\%})$ . Results of the classification aiming to discriminate IT, FP/FC, and INF/NEC lesions are shown in Table 1 . The overall cross-validation classification performance was 74.3%. IT was discriminated with sensitivity (SE) and specificity (SP) larger than 80%. However, there was an appreciable overlap between IT and FP/FC. Also, the INF and NEC lesions were detected with high SE (larger than 80%) and high SP (larger than 90%). These results provide strong indication that the LECs offer a new domain for representing the time-resolved information of tissue fluorescence emission in a very compact, accurate, complete, and computationally efficient way. This study also clearly demonstrated that the LECs derived from the TR-LIFS data analysis can characterize arterial tissue composition and, most important, detect features of vulnerable plaques, such as superficial necrosis and inflammation.

Table 1

Classification accuracy (sensitivity and specificity): discrimination of IT, FP/FC, and INF/NEC (from Ref. 44).

This study in carotid artery demonstrated also that a linear correlation between plaque biochemical content and spectroscopic parameters can be determined. Plaque elastin content, ranging from near 0% (in the advanced FP, FC, INF, and NEC) to 50% (in IT) showed a positive correlation with the intensity values at $440\text{-}\mathrm{nm}$ emission. Necrotic area, ranging from 0% (in stable plaques) to 70% (in NEC), presented a negative correlation with the LEC-0 at $550\text{-}\mathrm{nm}$ emission. A similar correlation was determined for the average lifetime values. The inflammatory cell (lymphocyte/macrophage) content, ranging from 0% (in stable plaques) to 60% (in INF and NEC), showed a positive correlation with the LEC-2 at $550\mathrm{nm}$ . In contrast, LEC-2 at $550\mathrm{nm}$ was negatively correlated to the collagen/SMC content.