2.1.

Excitation Light Source

The excitation light source is a critical part of the CRD, as it provides the tissue-exciting light and needs to be optimized for low OOB light in the UV and visible. It is based on a custom light source using a $75\text{-}\mathrm{W}$ Xenon short arc lamp [UX-75XE, Photon Technology International (PTI), Birmingham, New Jersey] in a lamp housing (A1010B and LPS-220B, PTI) with an f/2.5 elliptical reflector creating a spot size of $3.2\mathrm{mm}$ (FWHM). This results in a high power density at the focus, which is ideal for coupling light into a small fiber bundle. An integrated igniter (LPS-221B, PTI) provides the required high preignition voltage to start the lamp and interferes minimally with other electronic equipment. A shutter with a $6\text{-}\mathrm{mm}$ aperture and blades coated with $\mathrm{Al}\mathrm{Mg}{\mathrm{F}}_2$ (VS2581ZMO, Vincent Associates, Rochester, New York) controls light exposure that is synchronized with the camera electronics. A plano-convex quartz lens $(\mathrm{f}=50\mathrm{mm})$ is used for collimating the light beam through a 10-position filter wheel (Lambda 10, Sutter, Inc., Novato, California) carrying 10 filters (Semrock, Rochester, New York, Chroma Technologies, Rockingham, Vermont, and Omega Optical, Brattleboro, Vermont) with low autofluorescence (AF) and high OOB suppression. One filter is a long-pass filter with a cut-on at $320\mathrm{nm}$ for reflectance measurements, and the other nine filters were bandpass filters used to excite fluorescence. Their center wavelengths $\left({\lambda }_{\mathit{\text{center}}}\right)$ were selected to excite endogenous fluorophores mentioned earlier that have the potential to serve as biomarkers. A second lens similar to the first is used to focus the light beam into the fiber bundle. The lenses were chosen (infinite conjugates) to match the NA (0.22 in-air) of the fiber bundle. A pinhole aperture is placed in the beam path to block stray light. The light source assembly is cooled by an external low-noise fan (624N, ebm-papst, Inc., Mulfingen, Germany).

2.2.

Spectrometric Detection

Detection of fluorescence and reflectance signals is carried out using a diffraction-grating-based spectrometer (Inspectrum 300, Acton Research Corporation, Acton, Massachusetts). This spectrometer uses a Czerny-Turner optical configuration, which provides better stray light suppression than other compact spectrometers that are based on single mirror devices. It also has a built-in filter wheel, which is necessary to suppress excitation light reflected into the spectrometer. The mirrors are corrected for astigmatism, allowing multiple independent spectra to be collected independently. The spectrometer has an integrated back-illuminated CCD (INS-122B, Hamamatsu, $1024\times 122\text{pixels}$ , $24.6\mathrm{mm}\times 3\mathrm{mm}$ ) that is cooled to $-20°\mathrm{C}$ to lower the dark current when exposure reaches multiple seconds. Back-illumination provides higher quantum efficiency $(\eta \sim 85\mathrm{\%})$ than a comparable UV-coated front-illuminated CCD. The spectrometer has an internal shutter and mechanism to adjust the input slit size. The optical fibers themselves form a $200\text{-}\mu \mathrm{m}$ limiting aperture at the slit. A single $150\text{grooves}∕\mathrm{mm}$ grating (1-015-300, Newport Corporation, Irvine, California) was set to a center wavelength $\left({\lambda }_{\mathit{\text{center}}}\right)$ of $546\mathrm{nm}$ to match the spectral range of interest and to match the peak wavelength of the mercury spectrum $\left(546.1\mathrm{nm}\right)$ . The grating was optimized for the UV-blue ( $\eta$ at $300\mathrm{nm}=73\mathrm{\%}$ ) to minimize the exposure when the sample is illuminated with short wavelengths.

For fluorescence measurements, excitation light can be reflected into the detection channels and should therefore be well-suppressed. This is achieved by placing a rotating five-position filter wheel in the CCD’s field of view. Five long-pass filters were selected to provide the suppression needed for the excitation light. Due to the limited number of filter positions (total of 5 positions), one long-pass filter position had to suppress more than one excitation wavelength (total of 10 positions). These filters are dielectric coated and thus have minimal AF, and also they are placed away from the conjugate plane of the CCD.

For reflection measurement, the detection spectrum has second-order light, as the spectrometer used is a grating-based device. Thus, two reflection measurements are carried out, one with the shortest long-pass filter and the other with long-pass filter at around $400\mathrm{nm}$ .

2.3.

Fiber Optic Probe

The fiber optic probe is designed to transport the excitation light to the ovarian surface and carry back the emission and reflectance for spectrometric detection.²⁹ Through the duration of this project, three generations of fiber optic probes were developed. The differences in the three generations of the probe are summarized in Table 1 . Their development was based on feedback from the physician and also practical limitations observed while using them to perform clinical measurements. The excitation and emission collection elements are identical in all three versions, and they primarily differ in their ergonomics and support of biopsy collection. Carbon dioxide insufflated peritoneal pressure³⁰ is maintained in all three versions. We tested each of the three generations of fiber optic probes in a laparoscopy simulating training module (Karl Storz GmbH & Co., Tuttlingen, Germany). For this purpose, a phantom based on porcine collagen was used to simulate the surface toughness and texture of an ovary.

Table 1

Comparative summary of fiber optic probe features over its three generations.

For biopsy collection of the probed sites, we attempted in situ biopsy; however, standard retrieval with forceps is not feasible, and biopsy with surgical blades imposes additional risks to the patient. Thus, annotation of these sites is important for biopsy retrieval once the ovary is excised from the body. This is addressed by using a $20\text{-gauge}$ needle ( $0.032\text{-}\mathrm{in}$ OD, $0.020\text{-}\mathrm{in}$ ID, HTX-21R, Small Parts) custom-made (Vita Needle Company, Inc., Needham, Massachusetts) through a syringe channel in the probe. A few $\mu \mathrm{L}$ of sterile tissue-staining-dye (methylene blue) is used at its tip.³¹ A pump (NE-1000, New Era Pump Systems, Inc., Wantagh, New York) is used for dispensing the dye using a $12\text{-}\mathrm{ft}$ medical-grade tubing (US Plastics, Lima, Ohio). The needle is introduced into the fiber optic probe by the OR technician before the probe is inserted into the trocar. A spring (CSXX-0080, Small Parts) prevents the syringe from protruding out of the fiber optic probe. Often, the indentation of the syringe is sufficient to identify the measurement locations. Synchronized illumination of the field of view was incorporated, as the endoscope illumination must be turned off during measurements (as to not interfere with tissue emission collection). The probe tip uses one inch of transparent medical-grade polycarbonate (Zelux GS, Westlake Plastics, Lenni, Pennsylvania), and the surface is rough-polished with a grit size of 220. Illumination light is coupled into its proximal end, and when it reaches the rough surface, it is scattered, which illuminates a larger field compared to the previous probes. This allows the probe to be held steady when the endoscope illumination is turned off. In addition, at 120 lumens, the white-light LED (LXK2-PW14-V00, Philips Lumileds Lighting Co., San Jose, California) is three times brighter than that of the first-/second-generation probes.

The packaging of the third-generation probe and the layout of the optical fibers representing its different functioning legs are illustrated in Figs. 3 and 3 . This figure shows the configuration of the fibers used for excitation, illumination, emission, calibration, and quality control. Four fibers route the excitation light to the tissue, and the tissue emission is collected by three fibers that route it to the spectrometer. These seven fibers ( $200\mu \mathrm{m}$ , Polymicro) are interfaced to a short single central fiber (CF) $\left(800\mu \mathrm{m}\right)$ , allowing homogenous illumination over an $800\text{-}\mu \mathrm{m}$ area at the probe tip. Light from the light source is directly coupled with a fiber $\left(200\mu \mathrm{m}\right)$ for power measurements. A feedback fiber $\left(70\mu \mathrm{m}\right)$ couples light from the light source directly into the spectrometer to monitor bandpass filter failures. A fiber $\left(200\mu \mathrm{m}\right)$ carries light from an integrating sphere to the spectrometer and could be used for calibration purposes. NA (in air) of all the fibers employed is 0.22. There are 21 illumination fibers that provide auxiliary illumination for visual feedback, and these fibers terminate into the polycarbonate tip of the probe, providing illumination of the surgical field. The length of the probe was chosen to be comparable to an OR suction irrigator, which is $14\mathrm{in}$ in length and fits down a standard $5\text{-}\mathrm{mm}$ trocar. A stainless steel hypodermic tube (OD $4.77\mathrm{mm}$ , HTX 06-1/2R, Small Parts) defines the length and diameter of the rigid part. The optical sampling fibers are housed in a smaller tube (OD $1.27\mathrm{mm}$ , HTX-18T, Small Parts) and stick out about $0.5\mathrm{mm}$ at the probe tip. Between the rigid probe and the instrument, braided stainless steel monocoil with PVC (Armor Associates, Inc., Malvern, Pennsylvania) is used for protection of the fibers.

Fig. 3

(a) Packaging of the third-generation fiber optic probe. Section AA shows the cross section of the probe tip where the excitation light, diffused illumination light, and spring-loaded syringe emerge. (b) Illustration of the optical wiring diagram. EX and SPX are the excitation and emission fibers interfacing with CF, the central $800\text{-}\mu \mathrm{m}$ fiber. IC represents the 21 illumination channel fibers. Section BB presents the cross section of the probe where the IC fibers $(3\times 7)$ terminate and the syringe channel (SC) goes through. Section CC illustrates the seven EX-SPX fibers, the 21 IC fibers and the through syringe channel. PM, FB-1, 2, and IS are the power meter, feedback, and integration sphere fibers, respectively. The EX, PM, and FB-1 fibers are bundled together and terminated with a FC connector (section DD). They are simultaneously illuminated by the light source. IS is the integrating sphere fiber. The interface FB-1-FB-2 is for monitoring bandpass filter failures. The SPX, FB-2, and IS fibers are bundled together and terminated in a linear arrangement with a ferrule that fits the entrance slit of the spectrometer (section EE).

The fiber optic probe consists of several materials. The components that come in direct contact with the patient and physician are made of medical-grade materials such as stainless steel assembled using medical-grade epoxy. Chemically inert materials were used at all other assembly locations of the probe. Two assembly locations are of particular interest with respect to the type of glue employed: the junction connecting the excitation–emission fibers to the single fiber used for light mixing, and the probe tip where excitation light can be scattered by the sample and epoxy. These areas could fluoresce and add to the measurement signal. To reduce this, care was taken during fiber optic probe assembly to keep those areas glue-free. To minimize AF of the epoxy used, several industry-standard medical-grade epoxies were tested for their AF to determine the one that could be used for probe assembly. Measurements were performed on glue samples embedded in identically sized hypodermic needles, and the fiber optic probe was placed in direct contact with the glue samples. Measurements were performed on the Fluorolog using the fiber optic probe. The excitation and emission configurations covered the spectral region used for patient measurement. Furthermore, the excitation and emission fibers used have a high-OH fused silica core to allow ultraviolet C light transmission and have an aluminum jacket to keep their AF low in the presence of excitation light.

After each clinical use, the fiber optic probe and the dye delivery tubing are rinsed with Medical Enzyme detergent (Enzyme Solutions, Inc., Garrett, Indiana), cleaned with ethanol and cold sterilized using 100% ethylene oxide (EtO) at 90 to 135F using a dual cycle Steri-Vac 5XL gas sterilizer and aerator (3M, St. Paul, Minnesota).

2.4.

Calibration

Calibration of the data is important to factor out device-related parameters and was divided into power, spectral, and wavelength calibrations. Power calibration ensures that power incident on the tissue is always recorded. It is performed by taking a ratio of power at the probe tip and at the termination of a fiber directly coupled to the light source, each by using a power meter (1830-C, Newport) with a UV-enhanced silicon-based power meter sensor (883-UV, Newport).

Spectral calibration ensures that the emission is collected independent of the sensitivity fluctuations over the spectral range of interest. Deuterium and tungsten-halogen light (DH2000, Ocean Optics, Dunedin, Florida) sources cover the spectral range of interest, and their output is measured at each emission filter configuration and compared with the manufacturer’s data to generate a sensitivity calibration curve. The spectral calibration ranges are $200\text{to}400\mathrm{nm}$ and $380\text{to}800\mathrm{nm}$ , respectively, with a spectral overlap from $380\text{to}400\mathrm{nm}$ .

Peaks of a mercury-argon calibration source (HG-1, Ocean Optics) are used to convert the linear array of pixels needed to wavelength-calibrate the spectrometer. Because the spectrometer rotates the grating turret during each power-up, this calibration is performed as part of each calibration sequence.

2.5.

Data Processing

When analyzing the light path from source to sample to detector, one can identify the transfer functions involved. The light source is subjected to an excitation power transfer function prior to exciting the sample, which is determined by the throughputs of the filters and the fiber optic probe. Based on its composition, the sample converts the excitation light to fluorescence, and the amount collected also depends on the probe geometry. We did not further consider those, as they were kept constant by establishing probe–tissue physical contact. The fluorescence emitted is subjected to a detection transfer function based on spectral transmission of the fiber optic probe, emission filters, wavelength-dependent efficiency of the grating, and the detector’s spectral sensitivity. Integrating over the excitation bandwidth, one can show that the recorded fluorescence is proportional to the parameters described in the equation:

The recorded fluorescence is the intensity values on the CCD detector. The exposed area is kept constant in our study, and the exposure time is set by the control software. It is kept constant for all excitation wavelengths but is changed when measuring calibration standards. The excitation power is obtained by power calibration. The emission–collection efficiency transfer function is determined based on the spectral calibration.

In order to produce calibrated fluorescence spectra, the process described in Fig. 4 is employed. Spectrally dispersed light is recorded on the spectrometer output plane. For each measurement, an image is recorded with the light source shutter closed. The subtraction of those two images will eliminate room light and dark current of the sensor. The vertical direction of the image encodes for fiber location on the spectrometer input slit. Data along that direction are summed to create a single spectrum. The horizontal direction of the image encodes for wavelength, and based on the wavelength calibration, we convert pixel location to wavelength. Since the detector oversamples the spectrum, noise removal with a Savitsky-Golay³² filter was employed to match the systems optical resolution. By the multiplication of the computed power calibration factor of the excitation light and correcting by the spectral sensitivity of the detector, we can compute a calibrated fluorescence spectrum. Assembling all the emission spectra into a matrix yields an excitation emission matrix (EEM).

Fig. 4

Flowchart showing the CRD data processing sequence. Raw images from the spectrometer camera and background data (excitation shutter closed) were acquired for every site during patient measurements. Wavelength calibration transformed the background-subtracted data from pixel-space to wavelength-space. Noise removal was performed using Savitsky-Golay filtering ( $\text{order}=2$ , window $\text{size}=15$ ) (Ref. 32) to match the system’s resolution. The filtered spectrum was calibrated by the excitation power and the spectral sensitivity of the device. Several spectra were finally assembled to an excitation–emission matrix (EEM).

2.6.

Quality Control

In addition to calibration, it is important to develop a quality control protocol to ensure repeatable performance of the CRD throughout its lifetime. The spectral output and power of the light source at each excitation filter configuration are measured. The feedback fiber illustrated in Fig. 3 is used for monitoring any bandpass filter leaks. The passbands of those filters can be monitored by coupling light from the fiber optic probe to the input port of an integrating sphere (IS) and monitoring its output port with an IS fiber [Fig. 2 and Fig. 3]. Positive and negative fluorescence and reflectance standards that cover the spectral range of interest were included in the measurement procedure. These standards had peaks at excitation and emission wavelengths throughout the spectral range of the CRD. The positive standards include p-Terphenyl, Tetraphenylbutadiene, Rhodamine, and Fluorescein (Starna Cells, Atascadero, California). The first three standards are fluorescence standards dissolved in methylmethacrylate and polymerized to produce a polymethyl methacrylate (PMMA), while Fluorescein is a liquid fluorescence standard in a quartz cuvette. The negative fluorescence standard is deionized (DI) water in a dark bottle, which has a very low AF. For reflectance measurements, the positive standards are microspheres dispersed in DI-UV water in a $1\text{-}\mathrm{cm}$ path length quartz cuvette and an integrating sphere coated with broadband reflecting Spectralon³³ of $2\text{-}\mathrm{in}$ inner diameter.

In order to calculate the signal-to-noise ratio of the measured signals, we filtered the raw data with a Savitsky-Golay filtering ( $\text{order}=2$ , window $\text{size}=15\text{pixels}$ ). This processing resulted in a filtered spectrum that has the same number of elements as the raw spectrum but matches the spectrometer’s resolution. The noise spectrum was then obtained by subtracting the original spectrum from the filtered spectrum. A single number describing the noise was determined by calculating the standard deviation over a window of $150\text{pixels}$ $\left(75\mathrm{nm}\right)$ located at the maximum of the fluorescence spectrum. This number describes the noise of the raw spectrum. As the filter processes occurred over a window size of fifteen pixels, one can estimate the noise remaining in the filtered spectrum to be a square-root of 15 smaller. SNR is the peak intensity of the fluorescence signal divided by the standard deviation of the noise.

The calibration and quality control measurements are performed within the OR using a dark box to baffle room light. LabView (National Instruments, Austin, Texas) software is used to automate all the measurement procedures and for data storage. The drivers developed for the shutters, filter wheel, power meter, camera, and spectrometer are interfaced by custom-written LabView programs (RCubed, LLC, Lawrenceville, New Jersey). A script engine is used to select scripts that automate patient measurements, calibration, and quality control routines, and a separate batch file is generated to plot the data.

3.1.

Excitation Light Source

The design goal of the light source was to achieve an OOB suppression of four orders or better at wavelengths where fluorescence occurs. Several bandpass filters from three different manufacturers (Semrock, Chroma, and Omega) were investigated, and we found that it was more difficult to achieve the design goal with filters below $400\mathrm{nm}$ . The output of the configuration was analyzed by the Fluorolog. Figure 5 illustrates the best and worst cases of our configuration. Several filters did not operate at the desired specifications, and the suppression levels were improved by placing two filters in series, resulting in the desired OOB level. Table 2 illustrates the filter configuration that we implemented in our system and lists the combination of the excitation filter with corresponding emission long-pass filter in the spectrometer, which is needed to suppress the reflected excitation light. We chose long-pass filters with dielectric coatings with OD 5 or better OOB suppression and having cut-on wavelengths at 300, 372, 420 (Chroma), 500, and $590\mathrm{nm}$ (Omega). Excitation light used for reflectance measurement passes a $320\text{-}\mathrm{nm}$ longpass filter and is additionally attenuated by three orders of magnitude to avoid detector saturation. The total UV tissue exposure was calculated using a technique published by ACGIH wherein its threshold limiting value (TLV) was computed based on the effective irradiance relative to a monochromatic source at $270\mathrm{nm}$ . The exposure was determined to be less than 12% of the TLV set for UV exposure when light exposure is $0.2\mathrm{s}$ at each excitation wavelength. Of the 12%, 11 is from exposure at $280\mathrm{nm}$ (Table 2). Reflectance is collected by overlapping (300 to 600) and $\left(420\text{to}800\mathrm{nm}\right)$ with two measurements to achieve data from $300\text{to}800\mathrm{nm}$ .

Fig. 5

(a) Best and (b) worst cases of out-of-band suppression of the excitation light source measured for $400\text{-}\mathrm{nm}$ and $280\text{-}\mathrm{nm}$ excitation filters, respectively. A and B in both graphs represent the pass-band for the excitation configuration and its second order. The regions between them, in both cases, represent the out-of-band region where fluorescence is measured. C represents the light leaking through the filter in the infrared. The out-of-band light intensity in (a) is on average 5 orders lower than the pass-band, while in (b), it varies from 4.1 to 4.9 orders. The remaining excitation configurations have an out-of-band suppression of OD 4.8 or better.

Table 2

Filters employed in the CRD.

3.2.

Spectrometric Detection

The image plane of the detector contains three individual spectral tracks, with the largest track occupying the light collected from the sample and two additional tracks allowing feedback and quality control measurements of the instrument. Our fiber arrangement at the input slit of the spectrometer resulted in minimal cross talk between the tracks measuring sample spectra and that conducting instrument quality control.

The camera driver allowed binning of the data in spectral (horizontal) and measurement track (vertical) direction. Since binning occurs on the detector, it reduces readout noise and the amount of data that needs to be read out and transferred. Binning was optimized to not saturate the readout register and to not compromise system resolution while minimizing time needed to readout the data. In the vertical, a pixel-binning of eight was achieved and increased read-out speed by 55% when compared to unbinned readout. Binning in the horizontal direction did not improve read-out speed, indicating that the detector does not include electronic hardware for this operation. Binning is not appropriate for high signal levels such as some of our calibration standards, because it would saturate the readout register. Signal intensity can be decreased by reducing the exposure time; however, the camera drivers take significantly longer to reprogram the exposure time as compared to the binning. Therefore, the exposure time was optimized for the low light levels, and binning is disabled for high-light-level signals.

Correction factors due to wavelength-dependent sensitivity of the instrument are illustrated in Fig. 6 for measurement configurations. Fluctuations in the pass band and steep cut-on characteristics are typical for dielectric coated long-pass filters. An approximation of the system’s spectral sensitivity without long-pass filters is represented with the radiant sensitivity curve, which is directly proportional to the wavelength-dependent quantum efficiency of the detector and the grating efficiency. The spectral resolution of the detection system is determined by the size of the entrance slit image on the detector. In our case, the entrance aperture is a $200\text{-}\mu \mathrm{m}$ -diam fiber. Given a pixel size of $24\mu \mathrm{m}$ , a grating with $150\text{grooves}∕\mathrm{mm}$ and a focal length of $300\mathrm{mm}$ , the spectral resolution is $5\mathrm{nm}$ and 9 times oversampled.

Fig. 6

The CRD spectral-sensitivities for the five different long-pass filter configurations (300, 372, 420, 500, and 590). The first three filters (Chroma Technology) have smoother responses in comparison to the last two filters (Omega Filters). In addition, the curve composed of the device’s CCD quantum efficiency and the spectrometer’s grating efficiency is plotted and shows relative good agreement in the visible. The detection sensitivity is highest in the visible and is lower in the UV and near-infrared. The detector and spectrometer have been optimized for UV sensitivity.

3.3.

Fiber Optic Probe

The fiber optic probe development primarily improved ergonomics, as the sample illumination and signal collection were not altered compared to previously employed fiber optic probes.²⁶ We were able to incorporate both site annotation as well as visual illumination features. Annotation of the measurement site with the dye proved to be difficult, as delivery of precise dose depends on length of the tubing and hydrostatic configuration between annotation site and dye delivery pump. Also, excess dye blocked the tip of the probe. It turned out that a small needle puncture was sufficient to identify samples for later biopsy and processing, and inclusion of dye was abandoned. Using a $5\times$ optical eyepiece along with an anatomical sketch of the measurement location on the ovary improved the probability of finding the punctured site. This technique is ineffective for cystic ovaries, as they are fluid filled and should not be punctured. In our experience, about half of the punctures could be located. In the case of sites that did not have an identifiable puncture, we consulted the anatomical sketch to estimate the most probable location for the biopsy matching the measurement site. This approach is relatively straightforward, as the measurement is taken and the location is noted in the context of the blood supply, the tube, and the attachment to the uterus. Even if a puncture mark is not seen, the location can be approximated in a reasonably accurate manner. The approximate imaging depth of our probe is $50\text{to}120\mu \mathrm{m}$ , assuming that glandular breast tissue³⁴ and ovary have similar optical properties and that the optical diffusion constant³⁵ approximates the average imaging depth.

We required the glue used to assemble the fiber optic probe to have lower AF in comparison to tissue fluorescence. To confirm this AF, four different types of glues were compared in fluorescence to average normal ovary using the Fluorolog (Fig. 7 ). The results showed that among the tested excitation wavelengths, $280\mathrm{nm}$ gave the strongest AF for all the glues and was nearly comparable to that of tissue fluorescence. Except below $320\mathrm{nm}$ , the AF from Ångström Bond 9320 is lower than tissue fluorescence and is the lowest among the other glues tested and so was selected as the glue of choice. The other components of the fiber optic probe that do not come in contact with the sample or the excitation light were assembled with glues that have high metal–metal shear strength and relatively low curing temperatures. Furthermore, multiple ETO sterilizations of the fiber optic probe and the dye tubing did not degrade them with time.

Fig. 7

Choice of glue employed in the fiber optic probe. Except below $320\mathrm{nm}$ , the AF from AB9320 is lower than tissue fluorescence and is the lowest among the other glues tested. 353ND black works better below $320\mathrm{nm}$ and has the lowest AF and is lower than tissue fluorescence by a factor of $\sim 4.5$ , while AB9320 is $\sim 1.5$ times higher. At all other excitation wavelengths, AB9320 exhibits the lowest AF.

3.4.

CRD Performance

To ensure electrical safety, a hospital-grade isolation transformer with a $1000\text{-}\mathrm{W}$ capacity (ILC-1000MED4, Shoreview, Minnesota) was used to reduce the leakage current to less than $40\mu \mathrm{A}$ , a tenth of the standard hospital electrical safety limit, while the ground continuity was less than $210\mathrm{m}\Omega$ , two-thirds of the maximum allowed. Thus, there was no electrical continuity to the fiber optic probe. The total power consumption of the CRD was measured to be about $750\mathrm{W}$ .

Since our protocol requires that measurements are completed within $10\min$ , we evaluated the timing efficiency of our instrument. With the latest firmware upgrades and the automated LabView programming sequences, each measurement sequence takes approximately $20\mathrm{s}$ (Table 3 ). The total sample exposure time was $4.4\mathrm{s}$ . We were able to start changing the excitation filters when camera data were transferred to the computer; however, this was not possible for the emission filter wheel.

Table 3

Measurement timing efficiency for a single measurement.

The emission fibers carrying the fluorescence/reflectance signals operated at a working F/# of 2.27. Since the spectrometer used for detection of these signals operated at a system $\mathrm{F}∕\#=4$ and $300\text{-}\mathrm{mm}$ focal length, there is a loss of 60% of the detected emission signals.

Results from positive and negative standards with their respective comparison to the measurements on the Fluorolog are illustrated in Fig. 8 . Emission and excitation spectra were extracted from the respective EEMs. Good agreement was found for the systems, indicating that the calibration procedure for both excitation and emission–collection transfers were successful throughout the whole measurement wavelength range.

Fig. 8

The CRD measurement accuracy in comparison to measurements performed on the Fluorolog. The plots show the comparison of the excitation and emission spectra for the two devices using the same fiber optic probe. Measurements are performed on fluorescence standards. (a) through (d) show that the CRD measurement results (o, +) closely follow those from the Fluorolog. (a) Excitation range $\left(\mathrm{EX}\right)=270\text{to}290\mathrm{nm}$ , emission range $\left(\mathrm{EM}\right)=305\text{to}440\mathrm{nm}$ ; (b) $\mathrm{EX}=310\text{to}355\mathrm{nm}$ , $\mathrm{EM}=390\text{to}510\mathrm{nm}$ ; (c) $\mathrm{EX}=400\text{to}490\mathrm{nm}$ , $\mathrm{EM}=504\text{to}530$ ; and (d) $\mathrm{EX}=400\text{to}450\mathrm{nm}$ , $\mathrm{EM}=596\text{to}620\mathrm{nm}$ .

Because the power of our light source is decreasing with time, we monitor the power to determine optimal time of bulb replacement. There is a 3 to 8% power fluctuation per hour, while the fluctuations from one measurement day to the next one are 13 to 15%. This illustrates the importance to record power data for every CRD measurement. The power calibration factors (between power at probe tip and power at feed-back fiber) were at $3.7\pm 0.1$ at all wavelengths except at $280\mathrm{nm}$ , where it was 3.2. As expected, the ratio is not varying much over the excitation range as the fiber throughput and chromatic aberrations of the coupling system did not appreciably vary.

SNR of $⩾200$ is accomplished for all measurements involving calibration standards without saturating the camera. Our ongoing clinical study indicates that all patient measurements have an SNR of larger than 110, which occurs at $555\text{-}\mathrm{nm}$ excitation when using an exposure time of $0.2\mathrm{s}$ .

Results from a clinical measurement are presented in Fig. 9 from a normal ovary of a $62\text{-year}$ -old postmenopausal female. Emission spectra from the EEM are illustrated in the subplots and with the corresponding background measurement in water. In this example, measurements from the tissue are at least half an order larger than the background at all excitation wavelengths. Highest signal-to-background ratio was found at $320\mathrm{nm}$ and $340\mathrm{nm}$ , where AF form the tissue was high. AF of tissue at $280\mathrm{nm}$ is 2 times larger than in UV-A range, which is 12 times larger compared to the blue excitation range. A slight dip at $420\mathrm{nm}$ and ripples at $560\mathrm{nm}$ are due to hemoglobin reabsorption of the emission. In summary, the key CRD parameters and their values for the excitation and emission detection configurations, its clinical use, and the features of the fiber optic probe are listed in Table 4 .

Fig. 9

Single calibrated and processed clinical data compared with its background (water). (a) Complete excitation–emission matrix (EEM) representing fluorescence intensity data at the measured site. The contours represent areas of equal intensity. (b) The emissions collected when the site is excited at 280, 320, and 340 are at two or more orders higher in magnitude than the corresponding background. (c) The ones at 365, 400, and 415 are an order or more higher, while (d) the ones at 440, 480, and 555 result in higher than half an order when compared to background. Fluorescence spectra are background-subtracted. Fluorescence in (b) is an order of magnitude larger than (c), which is approximately 5 times larger than (d).

Table 4

Performance summary of the CRD.