Real time optical Biopsy: Time-resolved Fluorescence Spectroscopy instrumentation and validation

The Time-resolved fluorescence spectroscopy (TR-FS) has the potential to differentiate tumor and normal tissue in real time during surgical excision. In this manuscript, we describe the design of a novel TR-FS device, along with preliminary data on detection accuracy for fluorophores in a mixture. The instrument is capable of near real-time fluorescence lifetime acquisition in multiple spectral bands and analysis. It is also able to recover fluorescence lifetime with sub-20ps accuracy as validated with individual organic fluorescence dyes and dye mixtures yielding lifetime values for standard fluorescence dyes that closely match with published data. We also show that TR-FS is able to quantify the relative concentration of fluorescence dyes in a mixture by the unmixing of lifetime decays. We show that the TR-FS prototype is able to identify in near-real time the concentrations of dyes in a complex mixture based on previously trained data. As a result, we demonstrate that in complex mixtures of fluorophores, the relative concentration information is encoded in the fluorescence lifetime across multiple spectral bands. We show for the first time the temporal and spectral measurements of a mixture of fluorochromes and the ability to differentiate relative concentrations of each fluorochrome mixture in real time.


Spectral Demuxer
The spectral channel demuxer (Figure 2) takes the output from six 400 µm 0.22NA collection fibers as a single bundle and separates out the spectral channel through a series of dichroic beam splitters and band-pass filters.
The input is collimated by an 11mm focal length aspheric fiber port (PAF-SMA-11-A, Thorlabs) and long-pass filtered at 365nm to block the 355nm laser excitation. The first dichroic beam-splitter reflects the wavelengths shorter than 495nm and transmits longer wavelength of the fluorescence emission. This is the first split with two identical wings, one for the longer wavelengths and one for the shorter to filter the six total bands while minimizing transmission or reflection through more than three dichroic beam-splitters. This configuration also optimizes the dichroic beam-splitters performance since the 495nm split is clean but the 452nm is less efficient and passes through unwanted bands that require band-pass filters for cleanup.   There are two 60mm convex achromat lenses (Thorlabs AC254-060-A-ML) to narrow the divergence of the rays from the fiber ports due to the long distance required between entry and exit fiber ports. Figure 2 also shows relative spectral intensities measured for each of the six channels. Note that the channels with the weakest fluorescence (channels 5 and 6) are matched to the demuxer channels with the highest efficiency.

Optical Delay
A delay line was fabricated with 9 channels ranging in length from 5 ft. to 255 ft. at 50-foot intervals. The six channels on the demuxer are connected to the 5ft, 55ft, 105ft, 155ft, 205 ft., and 255 ft. delay lines.
The 50 ft. separation enables measuring longer lifetimes (45 ns total extinction) since for many tissues, the lifetime at 1/e is under 2 ns but the 10% extinction can be quite long. The delay fibers are 400 µm 0.22NA UV grade fused silica fibers and the fabrication of the delay unit was from Fiberguide Industries (Caldwell, ID, USA).

Fluorescent excitation laser source
The laser source is a Teem Photonics (Meylan Cedex, France) PNV-M02510-1x0 passively Q-switched MicroChip laser with 350ps pulse, 1 kHz repetition rate and 25µJ pulse energy output per pulse. There is a photodiode output for synchronization, but the rise time is too slow and the output voltage is too high to trigger the digitizer. Therefore a dichroic beam splitter (Semrock, 365nm long pass) that allows 94% refection and 6% transmission is used to power a separate photodiode (SV2-FC, 2GHz 320-1100nm, Thorlabs). The splitter is directly connected to the laser output with a fiber port to focus the 6% transmission of the beam splitter into a 200µm fiber which in turn connects to the photodiode. The output of the photodiode is terminated into the 50 Ohm trigger input on the digitizer. David S. Kittle, Fartash Vasefi, Chirag G. Patil,Adam Mamelak, Keith L. Black, and Pramod V Butte

Collection and Excitation Probe
The excitation laser pulse is transmitted to the sample by a 3.0 meter 0.11NA 600µm UV-grade silica non-solarizing multimode, step index fiber (probe manufactured by FiberGuide) and the collection is carried by twelve 0.22NA 200µm fibers surrounding the excitation fiber ( Figure 4). The collection fibers are beveled at a 10-degree angle in order to improve excitation collection overlap for small distances between the probe tip and sample. The collection fibers are located on a concentric circle 480 microns in radius around the central excitation fiber. These twelve fibers are then divided into two bundles of six each for two subminiature version A (SMA) terminated, bundled collection fibers. Every other fiber is routed to the opposing collection fiber. The fibers are built into a rigid, metal probe 7-cm long and at the end of the 3.0 meters of flexible cable. The entire probe is Sterrad (Steris) sterilizable up to fifty times without significant loss in transmission. Separate collection and excitation fibers are used to minimize the intrinsic fluorescence of the fiber due to the high-power of the excitation laser. The optimal distance for the probe tip is 3 mm above the tissue sample giving an illumination area defined by the NA of the excitation fiber of 2 mm 2 . David S. Kittle, Fartash Vasefi, Chirag G. Patil,Adam Mamelak, Keith L. Black, and Pramod V Butte

Laser Safety
The energy output of the laser at the probe tip is 4 µJ, giving the fluence per pulse at the tissue sample of 2µJ/mm 2 . The current accepted energy of 6.0 mJ/cm 2 or 60 µJ/mm 2 by the American Conference of Governmental Industrial Hygienists (ACGIH) is 30 times above this value. At a repetition rate of 1 millisecond, this gives an average power of 2 mW/mm 2 . The maximum permissible exposure (MPE) allowed at UV-A (315-400nm) is 1mW /cm 2 (0.01mW mm 2 ) for 8 hours. Over the short duration of tissue exposure, even though 200 times higher than the MPE, is well below damage threshold due to the short duration of pulses and low energy of each pulse.

Digitizer and photomultiplier tube
The output of the spectral demuxer is mixed and collected by a Photek photomultiplier tube (PMT), model PMT210 with an 80 ps rise time and gain of 1 million. The PMT is gated for 10 µs for each laser pulse due to a 2-3 µs jitter of the laser pulse and a 500ns settling time of the gate module (Photek GM200-3N). The PMT is supplied by a remotely controlled high voltage supply (Photek) that allows for automatic adjustment of the PMT voltage for various fluorescence signal strengths.
The PMT voltage ranges between -3 kV and -4.2 kV with a maximum permissible value of -5 kV.
Synthetic dyes that are highly fluorescent require a voltage of -3.0 kV to -3.4 kV. Tissue samples can require up to 4.2 kV, especially in the operating environment where the probe tip can get dirty. To accommodate this, after each pulse sequence, the signal intensity from the digitizer is assessed for saturation or low signal and a corresponding look-up table is accessed to set a new voltage for the PMT that optimizes the collection sensitivity. Due to the ramp-up of the PMT voltage, this usually takes a few cycles before stabilizing to the set voltage.

Calibration and standardization
The fluorescence decay measured at the digitizer is broadened by multiple factors: 1) finite pulse width of the laser (350ps), 2) finite bandwidth of the PMT and digitizer electronics, 3) intermodal dispersion in the multimode fibers used for the probe collection and excitation and the delay fibers. The actual impulse response of the system at each spectral band is then required to recover the fluorescence decay measured by the system since the measured signal is the convolution of the instrument response function and laser pulse width.

Data Processing and Deconvolution
The intrinsic fluorescence impulse response functions for each spectral channel can be recovered by numerical deconvolution of the measured system impulse response. Given a characterization of a linear time-invariant (LTI) system, the input x(t) is mapped to the output y(t) through the system H. In context of TRLIFS, the system obeys both superposition (the output is a linear combination of any input signals) and the system is time-invariant (adding delay to the input signal only adds delay to the output) and can be modeled by a Laguerre expansion of kernels for deconvolution to directly recover the system models.
The Laguere functions ( ), which are orthonormal, are used to expand the impulse response functions and estimate the Laguerre expansion coefficients . The fluorescence impulse response functions are then estimated for each of the six spectral channels and the steady-state spectrum can then be computed by integrating each intensity decay curve with respect to time. David S. Kittle,Fartash Vasefi,Chirag G. Patil,Adam Mamelak,Keith L. Black,and Pramod V Butte 7

Instrument Response function
In order to derive the actual fluorescence lifetime it is important to eliminate the effect of the optical and electronic components. This is achieved by deconvolving the instrument response. We have suggested employing fluorescence dyes with extremely fast decay response to mimic the IRF calculation. We chose 4-dimethylamino-4-cyanostilbene (DCS, ChemBridge Corporation, San Diego, CA) solution in Cyclohexane, 25°C, with emission range of 300 nm -500 nm to cover IRF for spectral channels 1-4. For channel 5 and 6, we used 1 mM concentration of the 2-(p-dimethylaminosotyryl)pyridylmethyl iodide (2-DASPI, Sigma-Aldrich Cat. 280135) in ethanol with maximum emission spectra at 550 nm. The DCS and 2-DASPI average lifetime is 66 ps and 30 ps, respectively.
In order to calculate pure fluorescence intensity decay from the TR-FS measurement, we need to independently characterize the instrument response function (IRF). The IRF can be determined by evaluation of the optical throughput including multimode fiber dispersion in probe and delay lines, laser optical properties, spectral demuxer, and detector

Lifetime Measurement Validation
Accurate lifetimes over all six spectral channels depend on an accurate system response function measurement and proper data fitting during deconvolution, which includes choosing the alpha decay constant and order of the Lagurre Polynomial. The decay constant alpha has a significant effect on lifetime calculation, shown in Figure 5. The order likewise effects lifetime, shown in Figure 6. The dye used for this comparison is K4-503-FLT (SETA BioMedicals) with a fluorescence peak at 415nm and lifetime of 1.25ns. Figure 6 also demonstrates the dependence of wavelength on impulse response and deconvolution. The impulse response for all three channels was measured at 355nm rather than the correct wavelengths, effectively shortening the lifetime calculation with longer wavelength channels. David S. Kittle, Fartash Vasefi, Chirag G. Patil,Adam Mamelak, Keith L. Black, and Pramod V Butte  Figure 7 MSE versus number of averaged signals comparing the signal variation between acquisitions. Note that after 30 averages, the MSE drops very little, showing little reduction in error, the signals are almost on top of each other and by 300 averages, there is almost no distinction between them. The error continues to drop beyond 300, but little value is gained in the additional measurements. For weaker signals, this trade-off can increase even more, therefore 1000 averages are always acquired for future flexibility.

Real time optical Biopsy: Time-resolved Fluorescence Spectroscopy instrumentation and validation
Supplementary video legend: The supplementary video shows the TR-FS prototype which calculates the fluorescence lifetime in near-real time and classifies the fluorescence mixtures based on fluorescence lifetime parameters. We demonstrate this by training the TR-FS to identify the samples of Rose Bengal and Rhodamine B and their four different mixtures. The scatterplot shows the classification marker which moves to the predefined regions (defined during the training procedure) while the optical probe moves between each fluorescent mixture.